No OneTemporary
Actions

Size

6 MB

Referenced Files

None

Subscribers

None

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.

	diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.c b/sys/contrib/openzfs/cmd/zdb/zdb.c
	index 92df3dd167bf..96db9c4b9a76 100644
	--- a/sys/contrib/openzfs/cmd/zdb/zdb.c
	+++ b/sys/contrib/openzfs/cmd/zdb/zdb.c
	@@ -1,9013 +1,9025 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2017, 2018 Lawrence Livermore National Security, LLC.
	* Copyright (c) 2015, 2017, Intel Corporation.
	* Copyright (c) 2020 Datto Inc.
	* Copyright (c) 2020, The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	* Copyright (c) 2021 Allan Jude
	* Copyright (c) 2021 Toomas Soome <tsoome@me.com>
	*/

	#include <stdio.h>
	#include <unistd.h>
	#include <stdlib.h>
	#include <ctype.h>
	#include <getopt.h>
	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/zap.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_sa.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/dbuf.h>
	#include <sys/zil.h>
	#include <sys/zil_impl.h>
	#include <sys/stat.h>
	#include <sys/resource.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_traverse.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/zfs_fuid.h>
	#include <sys/arc.h>
	#include <sys/arc_impl.h>
	#include <sys/ddt.h>
	#include <sys/zfeature.h>
	#include <sys/abd.h>
	#include <sys/blkptr.h>
	#include <sys/dsl_crypt.h>
	#include <sys/dsl_scan.h>
	#include <sys/btree.h>
	#include <zfs_comutil.h>
	#include <sys/zstd/zstd.h>

	#include <libnvpair.h>
	#include <libzutil.h>

	#include "zdb.h"

	#define ZDB_COMPRESS_NAME(idx) ((idx) < ZIO_COMPRESS_FUNCTIONS ? \
	zio_compress_table[(idx)].ci_name : "UNKNOWN")
	#define ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ? \
	zio_checksum_table[(idx)].ci_name : "UNKNOWN")
	#define ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : \
	(idx) == DMU_OTN_ZAP_DATA \|\| (idx) == DMU_OTN_ZAP_METADATA ? \
	DMU_OT_ZAP_OTHER : \
	(idx) == DMU_OTN_UINT64_DATA \|\| (idx) == DMU_OTN_UINT64_METADATA ? \
	DMU_OT_UINT64_OTHER : DMU_OT_NUMTYPES)

	/* Some platforms require part of inode IDs to be remapped */
	#ifdef __APPLE__
	#define ZDB_MAP_OBJECT_ID(obj) INO_XNUTOZFS(obj, 2)
	#else
	#define ZDB_MAP_OBJECT_ID(obj) (obj)
	#endif

	static const char *
	zdb_ot_name(dmu_object_type_t type)
	{
	if (type < DMU_OT_NUMTYPES)
	return (dmu_ot[type].ot_name);
	else if ((type & DMU_OT_NEWTYPE) &&
	((type & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS))
	return (dmu_ot_byteswap[type & DMU_OT_BYTESWAP_MASK].ob_name);
	else
	return ("UNKNOWN");
	}

	extern int reference_tracking_enable;
	extern int zfs_recover;
	extern unsigned long zfs_arc_meta_min, zfs_arc_meta_limit;
	-extern int zfs_vdev_async_read_max_active;
	+extern uint_t zfs_vdev_async_read_max_active;
	extern boolean_t spa_load_verify_dryrun;
	extern boolean_t spa_mode_readable_spacemaps;
	-extern int zfs_reconstruct_indirect_combinations_max;
	+extern uint_t zfs_reconstruct_indirect_combinations_max;
	extern uint_t zfs_btree_verify_intensity;

	static const char cmdname[] = "zdb";
	uint8_t dump_opt[256];

	typedef void object_viewer_t(objset_t , uint64_t, void data, size_t size);

	uint64_t *zopt_metaslab = NULL;
	static unsigned zopt_metaslab_args = 0;

	typedef struct zopt_object_range {
	uint64_t zor_obj_start;
	uint64_t zor_obj_end;
	uint64_t zor_flags;
	} zopt_object_range_t;
	zopt_object_range_t *zopt_object_ranges = NULL;
	static unsigned zopt_object_args = 0;

	static int flagbits[256];

	#define ZOR_FLAG_PLAIN_FILE 0x0001
	#define ZOR_FLAG_DIRECTORY 0x0002
	#define ZOR_FLAG_SPACE_MAP 0x0004
	#define ZOR_FLAG_ZAP 0x0008
	#define ZOR_FLAG_ALL_TYPES -1
	#define ZOR_SUPPORTED_FLAGS (ZOR_FLAG_PLAIN_FILE \| \
	ZOR_FLAG_DIRECTORY \| \
	ZOR_FLAG_SPACE_MAP \| \
	ZOR_FLAG_ZAP)

	#define ZDB_FLAG_CHECKSUM 0x0001
	#define ZDB_FLAG_DECOMPRESS 0x0002
	#define ZDB_FLAG_BSWAP 0x0004
	#define ZDB_FLAG_GBH 0x0008
	#define ZDB_FLAG_INDIRECT 0x0010
	#define ZDB_FLAG_RAW 0x0020
	#define ZDB_FLAG_PRINT_BLKPTR 0x0040
	#define ZDB_FLAG_VERBOSE 0x0080

	uint64_t max_inflight_bytes = 256 * 1024 * 1024; /* 256MB */
	static int leaked_objects = 0;
	static range_tree_t *mos_refd_objs;

	static void snprintf_blkptr_compact(char , size_t, const blkptr_t ,
	boolean_t);
	static void mos_obj_refd(uint64_t);
	static void mos_obj_refd_multiple(uint64_t);
	static int dump_bpobj_cb(void arg, const blkptr_t bp, boolean_t free,
	dmu_tx_t *tx);

	typedef struct sublivelist_verify {
	/* FREE's that haven't yet matched to an ALLOC, in one sub-livelist */
	zfs_btree_t sv_pair;

	/* ALLOC's without a matching FREE, accumulates across sub-livelists */
	zfs_btree_t sv_leftover;
	} sublivelist_verify_t;

	static int
	livelist_compare(const void larg, const void rarg)
	{
	const blkptr_t *l = larg;
	const blkptr_t *r = rarg;

	/* Sort them according to dva[0] */
	uint64_t l_dva0_vdev, r_dva0_vdev;
	l_dva0_vdev = DVA_GET_VDEV(&l->blk_dva[0]);
	r_dva0_vdev = DVA_GET_VDEV(&r->blk_dva[0]);
	if (l_dva0_vdev < r_dva0_vdev)
	return (-1);
	else if (l_dva0_vdev > r_dva0_vdev)
	return (+1);

	/* if vdevs are equal, sort by offsets. */
	uint64_t l_dva0_offset;
	uint64_t r_dva0_offset;
	l_dva0_offset = DVA_GET_OFFSET(&l->blk_dva[0]);
	r_dva0_offset = DVA_GET_OFFSET(&r->blk_dva[0]);
	if (l_dva0_offset < r_dva0_offset) {
	return (-1);
	} else if (l_dva0_offset > r_dva0_offset) {
	return (+1);
	}

	/*
	* Since we're storing blkptrs without cancelling FREE/ALLOC pairs,
	* it's possible the offsets are equal. In that case, sort by txg
	*/
	if (l->blk_birth < r->blk_birth) {
	return (-1);
	} else if (l->blk_birth > r->blk_birth) {
	return (+1);
	}
	return (0);
	}

	typedef struct sublivelist_verify_block {
	dva_t svb_dva;

	/*
	* We need this to check if the block marked as allocated
	* in the livelist was freed (and potentially reallocated)
	* in the metaslab spacemaps at a later TXG.
	*/
	uint64_t svb_allocated_txg;
	} sublivelist_verify_block_t;

	static void zdb_print_blkptr(const blkptr_t *bp, int flags);

	typedef struct sublivelist_verify_block_refcnt {
	/* block pointer entry in livelist being verified */
	blkptr_t svbr_blk;

	/*
	* Refcount gets incremented to 1 when we encounter the first
	* FREE entry for the svfbr block pointer and a node for it
	* is created in our ZDB verification/tracking metadata.
	*
	* As we encounter more FREE entries we increment this counter
	* and similarly decrement it whenever we find the respective
	* ALLOC entries for this block.
	*
	* When the refcount gets to 0 it means that all the FREE and
	* ALLOC entries of this block have paired up and we no longer
	* need to track it in our verification logic (e.g. the node
	* containing this struct in our verification data structure
	* should be freed).
	*
	* [refer to sublivelist_verify_blkptr() for the actual code]
	*/
	uint32_t svbr_refcnt;
	} sublivelist_verify_block_refcnt_t;

	static int
	sublivelist_block_refcnt_compare(const void larg, const void rarg)
	{
	const sublivelist_verify_block_refcnt_t *l = larg;
	const sublivelist_verify_block_refcnt_t *r = rarg;
	return (livelist_compare(&l->svbr_blk, &r->svbr_blk));
	}

	static int
	sublivelist_verify_blkptr(void arg, const blkptr_t bp, boolean_t free,
	dmu_tx_t *tx)
	{
	ASSERT3P(tx, ==, NULL);
	struct sublivelist_verify *sv = arg;
	sublivelist_verify_block_refcnt_t current = {
	.svbr_blk = *bp,

	/*
	* Start with 1 in case this is the first free entry.
	* This field is not used for our B-Tree comparisons
	* anyway.
	*/
	.svbr_refcnt = 1,
	};

	zfs_btree_index_t where;
	sublivelist_verify_block_refcnt_t *pair =
	zfs_btree_find(&sv->sv_pair, &current, &where);
	if (free) {
	if (pair == NULL) {
	/* first free entry for this block pointer */
	zfs_btree_add(&sv->sv_pair, &current);
	} else {
	pair->svbr_refcnt++;
	}
	} else {
	if (pair == NULL) {
	/* block that is currently marked as allocated */
	for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
	if (DVA_IS_EMPTY(&bp->blk_dva[i]))
	break;
	sublivelist_verify_block_t svb = {
	.svb_dva = bp->blk_dva[i],
	.svb_allocated_txg = bp->blk_birth
	};

	if (zfs_btree_find(&sv->sv_leftover, &svb,
	&where) == NULL) {
	zfs_btree_add_idx(&sv->sv_leftover,
	&svb, &where);
	}
	}
	} else {
	/* alloc matches a free entry */
	pair->svbr_refcnt--;
	if (pair->svbr_refcnt == 0) {
	/* all allocs and frees have been matched */
	zfs_btree_remove_idx(&sv->sv_pair, &where);
	}
	}
	}

	return (0);
	}

	static int
	sublivelist_verify_func(void args, dsl_deadlist_entry_t dle)
	{
	int err;
	struct sublivelist_verify *sv = args;

	zfs_btree_create(&sv->sv_pair, sublivelist_block_refcnt_compare,
	sizeof (sublivelist_verify_block_refcnt_t));

	err = bpobj_iterate_nofree(&dle->dle_bpobj, sublivelist_verify_blkptr,
	sv, NULL);

	sublivelist_verify_block_refcnt_t *e;
	zfs_btree_index_t *cookie = NULL;
	while ((e = zfs_btree_destroy_nodes(&sv->sv_pair, &cookie)) != NULL) {
	char blkbuf[BP_SPRINTF_LEN];
	snprintf_blkptr_compact(blkbuf, sizeof (blkbuf),
	&e->svbr_blk, B_TRUE);
	(void) printf("\tERROR: %d unmatched FREE(s): %s\n",
	e->svbr_refcnt, blkbuf);
	}
	zfs_btree_destroy(&sv->sv_pair);

	return (err);
	}

	static int
	livelist_block_compare(const void larg, const void rarg)
	{
	const sublivelist_verify_block_t *l = larg;
	const sublivelist_verify_block_t *r = rarg;

	if (DVA_GET_VDEV(&l->svb_dva) < DVA_GET_VDEV(&r->svb_dva))
	return (-1);
	else if (DVA_GET_VDEV(&l->svb_dva) > DVA_GET_VDEV(&r->svb_dva))
	return (+1);

	if (DVA_GET_OFFSET(&l->svb_dva) < DVA_GET_OFFSET(&r->svb_dva))
	return (-1);
	else if (DVA_GET_OFFSET(&l->svb_dva) > DVA_GET_OFFSET(&r->svb_dva))
	return (+1);

	if (DVA_GET_ASIZE(&l->svb_dva) < DVA_GET_ASIZE(&r->svb_dva))
	return (-1);
	else if (DVA_GET_ASIZE(&l->svb_dva) > DVA_GET_ASIZE(&r->svb_dva))
	return (+1);

	return (0);
	}

	/*
	* Check for errors in a livelist while tracking all unfreed ALLOCs in the
	* sublivelist_verify_t: sv->sv_leftover
	*/
	static void
	livelist_verify(dsl_deadlist_t dl, void arg)
	{
	sublivelist_verify_t *sv = arg;
	dsl_deadlist_iterate(dl, sublivelist_verify_func, sv);
	}

	/*
	* Check for errors in the livelist entry and discard the intermediary
	* data structures
	*/
	static int
	sublivelist_verify_lightweight(void args, dsl_deadlist_entry_t dle)
	{
	(void) args;
	sublivelist_verify_t sv;
	zfs_btree_create(&sv.sv_leftover, livelist_block_compare,
	sizeof (sublivelist_verify_block_t));
	int err = sublivelist_verify_func(&sv, dle);
	zfs_btree_clear(&sv.sv_leftover);
	zfs_btree_destroy(&sv.sv_leftover);
	return (err);
	}

	typedef struct metaslab_verify {
	/*
	* Tree containing all the leftover ALLOCs from the livelists
	* that are part of this metaslab.
	*/
	zfs_btree_t mv_livelist_allocs;

	/*
	* Metaslab information.
	*/
	uint64_t mv_vdid;
	uint64_t mv_msid;
	uint64_t mv_start;
	uint64_t mv_end;

	/*
	* What's currently allocated for this metaslab.
	*/
	range_tree_t *mv_allocated;
	} metaslab_verify_t;

	typedef void ll_iter_t(dsl_deadlist_t ll, void arg);

	typedef int (zdb_log_sm_cb_t)(spa_t spa, space_map_entry_t *sme, uint64_t txg,
	void *arg);

	typedef struct unflushed_iter_cb_arg {
	spa_t *uic_spa;
	uint64_t uic_txg;
	void *uic_arg;
	zdb_log_sm_cb_t uic_cb;
	} unflushed_iter_cb_arg_t;

	static int
	iterate_through_spacemap_logs_cb(space_map_entry_t sme, void arg)
	{
	unflushed_iter_cb_arg_t *uic = arg;
	return (uic->uic_cb(uic->uic_spa, sme, uic->uic_txg, uic->uic_arg));
	}

	static void
	iterate_through_spacemap_logs(spa_t spa, zdb_log_sm_cb_t cb, void arg)
	{
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
	return;

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
	sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls)) {
	space_map_t *sm = NULL;
	VERIFY0(space_map_open(&sm, spa_meta_objset(spa),
	sls->sls_sm_obj, 0, UINT64_MAX, SPA_MINBLOCKSHIFT));

	unflushed_iter_cb_arg_t uic = {
	.uic_spa = spa,
	.uic_txg = sls->sls_txg,
	.uic_arg = arg,
	.uic_cb = cb
	};
	VERIFY0(space_map_iterate(sm, space_map_length(sm),
	iterate_through_spacemap_logs_cb, &uic));
	space_map_close(sm);
	}
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	static void
	verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg,
	uint64_t offset, uint64_t size)
	{
	sublivelist_verify_block_t svb;
	DVA_SET_VDEV(&svb.svb_dva, mv->mv_vdid);
	DVA_SET_OFFSET(&svb.svb_dva, offset);
	DVA_SET_ASIZE(&svb.svb_dva, size);
	zfs_btree_index_t where;
	uint64_t end_offset = offset + size;

	/*
	* Look for an exact match for spacemap entry in the livelist entries.
	* Then, look for other livelist entries that fall within the range
	* of the spacemap entry as it may have been condensed
	*/
	sublivelist_verify_block_t *found =
	zfs_btree_find(&mv->mv_livelist_allocs, &svb, &where);
	if (found == NULL) {
	found = zfs_btree_next(&mv->mv_livelist_allocs, &where, &where);
	}
	for (; found != NULL && DVA_GET_VDEV(&found->svb_dva) == mv->mv_vdid &&
	DVA_GET_OFFSET(&found->svb_dva) < end_offset;
	found = zfs_btree_next(&mv->mv_livelist_allocs, &where, &where)) {
	if (found->svb_allocated_txg <= txg) {
	(void) printf("ERROR: Livelist ALLOC [%llx:%llx] "
	"from TXG %llx FREED at TXG %llx\n",
	(u_longlong_t)DVA_GET_OFFSET(&found->svb_dva),
	(u_longlong_t)DVA_GET_ASIZE(&found->svb_dva),
	(u_longlong_t)found->svb_allocated_txg,
	(u_longlong_t)txg);
	}
	}
	}

	static int
	metaslab_spacemap_validation_cb(space_map_entry_t sme, void arg)
	{
	metaslab_verify_t *mv = arg;
	uint64_t offset = sme->sme_offset;
	uint64_t size = sme->sme_run;
	uint64_t txg = sme->sme_txg;

	if (sme->sme_type == SM_ALLOC) {
	if (range_tree_contains(mv->mv_allocated,
	offset, size)) {
	(void) printf("ERROR: DOUBLE ALLOC: "
	"%llu [%llx:%llx] "
	"%llu:%llu LOG_SM\n",
	(u_longlong_t)txg, (u_longlong_t)offset,
	(u_longlong_t)size, (u_longlong_t)mv->mv_vdid,
	(u_longlong_t)mv->mv_msid);
	} else {
	range_tree_add(mv->mv_allocated,
	offset, size);
	}
	} else {
	if (!range_tree_contains(mv->mv_allocated,
	offset, size)) {
	(void) printf("ERROR: DOUBLE FREE: "
	"%llu [%llx:%llx] "
	"%llu:%llu LOG_SM\n",
	(u_longlong_t)txg, (u_longlong_t)offset,
	(u_longlong_t)size, (u_longlong_t)mv->mv_vdid,
	(u_longlong_t)mv->mv_msid);
	} else {
	range_tree_remove(mv->mv_allocated,
	offset, size);
	}
	}

	if (sme->sme_type != SM_ALLOC) {
	/*
	* If something is freed in the spacemap, verify that
	* it is not listed as allocated in the livelist.
	*/
	verify_livelist_allocs(mv, txg, offset, size);
	}
	return (0);
	}

	static int
	spacemap_check_sm_log_cb(spa_t spa, space_map_entry_t sme,
	uint64_t txg, void *arg)
	{
	metaslab_verify_t *mv = arg;
	uint64_t offset = sme->sme_offset;
	uint64_t vdev_id = sme->sme_vdev;

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);

	/* skip indirect vdevs */
	if (!vdev_is_concrete(vd))
	return (0);

	if (vdev_id != mv->mv_vdid)
	return (0);

	metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];
	if (ms->ms_id != mv->mv_msid)
	return (0);

	if (txg < metaslab_unflushed_txg(ms))
	return (0);


	ASSERT3U(txg, ==, sme->sme_txg);
	return (metaslab_spacemap_validation_cb(sme, mv));
	}

	static void
	spacemap_check_sm_log(spa_t spa, metaslab_verify_t mv)
	{
	iterate_through_spacemap_logs(spa, spacemap_check_sm_log_cb, mv);
	}

	static void
	spacemap_check_ms_sm(space_map_t sm, metaslab_verify_t mv)
	{
	if (sm == NULL)
	return;

	VERIFY0(space_map_iterate(sm, space_map_length(sm),
	metaslab_spacemap_validation_cb, mv));
	}

	static void iterate_deleted_livelists(spa_t spa, ll_iter_t func, void arg);

	/*
	* Transfer blocks from sv_leftover tree to the mv_livelist_allocs if
	* they are part of that metaslab (mv_msid).
	*/
	static void
	mv_populate_livelist_allocs(metaslab_verify_t mv, sublivelist_verify_t sv)
	{
	zfs_btree_index_t where;
	sublivelist_verify_block_t *svb;
	ASSERT3U(zfs_btree_numnodes(&mv->mv_livelist_allocs), ==, 0);
	for (svb = zfs_btree_first(&sv->sv_leftover, &where);
	svb != NULL;
	svb = zfs_btree_next(&sv->sv_leftover, &where, &where)) {
	if (DVA_GET_VDEV(&svb->svb_dva) != mv->mv_vdid)
	continue;

	if (DVA_GET_OFFSET(&svb->svb_dva) < mv->mv_start &&
	(DVA_GET_OFFSET(&svb->svb_dva) +
	DVA_GET_ASIZE(&svb->svb_dva)) > mv->mv_start) {
	(void) printf("ERROR: Found block that crosses "
	"metaslab boundary: <%llu:%llx:%llx>\n",
	(u_longlong_t)DVA_GET_VDEV(&svb->svb_dva),
	(u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva),
	(u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva));
	continue;
	}

	if (DVA_GET_OFFSET(&svb->svb_dva) < mv->mv_start)
	continue;

	if (DVA_GET_OFFSET(&svb->svb_dva) >= mv->mv_end)
	continue;

	if ((DVA_GET_OFFSET(&svb->svb_dva) +
	DVA_GET_ASIZE(&svb->svb_dva)) > mv->mv_end) {
	(void) printf("ERROR: Found block that crosses "
	"metaslab boundary: <%llu:%llx:%llx>\n",
	(u_longlong_t)DVA_GET_VDEV(&svb->svb_dva),
	(u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva),
	(u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva));
	continue;
	}

	zfs_btree_add(&mv->mv_livelist_allocs, svb);
	}

	for (svb = zfs_btree_first(&mv->mv_livelist_allocs, &where);
	svb != NULL;
	svb = zfs_btree_next(&mv->mv_livelist_allocs, &where, &where)) {
	zfs_btree_remove(&sv->sv_leftover, svb);
	}
	}

	/*
	* [Livelist Check]
	* Iterate through all the sublivelists and:
	* - report leftover frees (**)
	* - record leftover ALLOCs together with their TXG [see Cross Check]
	*
	* (**) Note: Double ALLOCs are valid in datasets that have dedup
	* enabled. Similarly double FREEs are allowed as well but
	* only if they pair up with a corresponding ALLOC entry once
	* we our done with our sublivelist iteration.
	*
	* [Spacemap Check]
	* for each metaslab:
	* - iterate over spacemap and then the metaslab's entries in the
	* spacemap log, then report any double FREEs and ALLOCs (do not
	* blow up).
	*
	* [Cross Check]
	* After finishing the Livelist Check phase and while being in the
	* Spacemap Check phase, we find all the recorded leftover ALLOCs
	* of the livelist check that are part of the metaslab that we are
	* currently looking at in the Spacemap Check. We report any entries
	* that are marked as ALLOCs in the livelists but have been actually
	* freed (and potentially allocated again) after their TXG stamp in
	* the spacemaps. Also report any ALLOCs from the livelists that
	* belong to indirect vdevs (e.g. their vdev completed removal).
	*
	* Note that this will miss Log Spacemap entries that cancelled each other
	* out before being flushed to the metaslab, so we are not guaranteed
	* to match all erroneous ALLOCs.
	*/
	static void
	livelist_metaslab_validate(spa_t *spa)
	{
	(void) printf("Verifying deleted livelist entries\n");

	sublivelist_verify_t sv;
	zfs_btree_create(&sv.sv_leftover, livelist_block_compare,
	sizeof (sublivelist_verify_block_t));
	iterate_deleted_livelists(spa, livelist_verify, &sv);

	(void) printf("Verifying metaslab entries\n");
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	if (!vdev_is_concrete(vd))
	continue;

	for (uint64_t mid = 0; mid < vd->vdev_ms_count; mid++) {
	metaslab_t *m = vd->vdev_ms[mid];

	(void) fprintf(stderr,
	"\rverifying concrete vdev %llu, "
	"metaslab %llu of %llu ...",
	(longlong_t)vd->vdev_id,
	(longlong_t)mid,
	(longlong_t)vd->vdev_ms_count);

	uint64_t shift, start;
	range_seg_type_t type =
	metaslab_calculate_range_tree_type(vd, m,
	&start, &shift);
	metaslab_verify_t mv;
	mv.mv_allocated = range_tree_create(NULL,
	type, NULL, start, shift);
	mv.mv_vdid = vd->vdev_id;
	mv.mv_msid = m->ms_id;
	mv.mv_start = m->ms_start;
	mv.mv_end = m->ms_start + m->ms_size;
	zfs_btree_create(&mv.mv_livelist_allocs,
	livelist_block_compare,
	sizeof (sublivelist_verify_block_t));

	mv_populate_livelist_allocs(&mv, &sv);

	spacemap_check_ms_sm(m->ms_sm, &mv);
	spacemap_check_sm_log(spa, &mv);

	range_tree_vacate(mv.mv_allocated, NULL, NULL);
	range_tree_destroy(mv.mv_allocated);
	zfs_btree_clear(&mv.mv_livelist_allocs);
	zfs_btree_destroy(&mv.mv_livelist_allocs);
	}
	}
	(void) fprintf(stderr, "\n");

	/*
	* If there are any segments in the leftover tree after we walked
	* through all the metaslabs in the concrete vdevs then this means
	* that we have segments in the livelists that belong to indirect
	* vdevs and are marked as allocated.
	*/
	if (zfs_btree_numnodes(&sv.sv_leftover) == 0) {
	zfs_btree_destroy(&sv.sv_leftover);
	return;
	}
	(void) printf("ERROR: Found livelist blocks marked as allocated "
	"for indirect vdevs:\n");

	zfs_btree_index_t *where = NULL;
	sublivelist_verify_block_t *svb;
	while ((svb = zfs_btree_destroy_nodes(&sv.sv_leftover, &where)) !=
	NULL) {
	int vdev_id = DVA_GET_VDEV(&svb->svb_dva);
	ASSERT3U(vdev_id, <, rvd->vdev_children);
	vdev_t *vd = rvd->vdev_child[vdev_id];
	ASSERT(!vdev_is_concrete(vd));
	(void) printf("<%d:%llx:%llx> TXG %llx\n",
	vdev_id, (u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva),
	(u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva),
	(u_longlong_t)svb->svb_allocated_txg);
	}
	(void) printf("\n");
	zfs_btree_destroy(&sv.sv_leftover);
	}

	/*
	* These libumem hooks provide a reasonable set of defaults for the allocator's
	* debugging facilities.
	*/
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

	const char *
	_umem_logging_init(void)
	{
	return ("fail,contents"); /* $UMEM_LOGGING setting */
	}

	static void
	usage(void)
	{
	(void) fprintf(stderr,
	"Usage:\t%s [-AbcdDFGhikLMPsvXy] [-e [-V] [-p <path> ...]] "
	"[-I <inflight I/Os>]\n"
	"\t\t[-o <var>=<value>]... [-t <txg>] [-U <cache>] [-x <dumpdir>]\n"
	"\t\t[<poolname>[/<dataset \| objset id>] [<object \| range> ...]]\n"
	"\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
	"\t\t[<poolname>[/<dataset \| objset id>] [<object \| range> ...]\n"
	"\t%s [-v] <bookmark>\n"
	"\t%s -C [-A] [-U <cache>]\n"
	"\t%s -l [-Aqu] <device>\n"
	"\t%s -m [-AFLPX] [-e [-V] [-p <path> ...]] [-t <txg>] "
	"[-U <cache>]\n\t\t<poolname> [<vdev> [<metaslab> ...]]\n"
	"\t%s -O <dataset> <path>\n"
	"\t%s -r <dataset> <path> <destination>\n"
	"\t%s -R [-A] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
	"\t\t<poolname> <vdev>:<offset>:<size>[:<flags>]\n"
	"\t%s -E [-A] word0:word1:...:word15\n"
	"\t%s -S [-AP] [-e [-V] [-p <path> ...]] [-U <cache>] "
	"<poolname>\n\n",
	cmdname, cmdname, cmdname, cmdname, cmdname, cmdname, cmdname,
	cmdname, cmdname, cmdname, cmdname);

	(void) fprintf(stderr, " Dataset name must include at least one "
	"separator character '/' or '@'\n");
	(void) fprintf(stderr, " If dataset name is specified, only that "
	"dataset is dumped\n");
	(void) fprintf(stderr, " If object numbers or object number "
	"ranges are specified, only those\n"
	" objects or ranges are dumped.\n\n");
	(void) fprintf(stderr,
	" Object ranges take the form <start>:<end>[:<flags>]\n"
	" start Starting object number\n"
	" end Ending object number, or -1 for no upper bound\n"
	" flags Optional flags to select object types:\n"
	" A All objects (this is the default)\n"
	" d ZFS directories\n"
	" f ZFS files \n"
	" m SPA space maps\n"
	" z ZAPs\n"
	" - Negate effect of next flag\n\n");
	(void) fprintf(stderr, " Options to control amount of output:\n");
	(void) fprintf(stderr, " -b --block-stats "
	"block statistics\n");
	(void) fprintf(stderr, " -c --checksum "
	"checksum all metadata (twice for all data) blocks\n");
	(void) fprintf(stderr, " -C --config "
	"config (or cachefile if alone)\n");
	(void) fprintf(stderr, " -d --datasets "
	"dataset(s)\n");
	(void) fprintf(stderr, " -D --dedup-stats "
	"dedup statistics\n");
	(void) fprintf(stderr, " -E --embedded-block-pointer=INTEGER\n"
	" decode and display block "
	"from an embedded block pointer\n");
	(void) fprintf(stderr, " -h --history "
	"pool history\n");
	(void) fprintf(stderr, " -i --intent-logs "
	"intent logs\n");
	(void) fprintf(stderr, " -l --label "
	"read label contents\n");
	(void) fprintf(stderr, " -k --checkpointed-state "
	"examine the checkpointed state of the pool\n");
	(void) fprintf(stderr, " -L --disable-leak-tracking "
	"disable leak tracking (do not load spacemaps)\n");
	(void) fprintf(stderr, " -m --metaslabs "
	"metaslabs\n");
	(void) fprintf(stderr, " -M --metaslab-groups "
	"metaslab groups\n");
	(void) fprintf(stderr, " -O --object-lookups "
	"perform object lookups by path\n");
	(void) fprintf(stderr, " -r --copy-object "
	"copy an object by path to file\n");
	(void) fprintf(stderr, " -R --read-block "
	"read and display block from a device\n");
	(void) fprintf(stderr, " -s --io-stats "
	"report stats on zdb's I/O\n");
	(void) fprintf(stderr, " -S --simulate-dedup "
	"simulate dedup to measure effect\n");
	(void) fprintf(stderr, " -v --verbose "
	"verbose (applies to all others)\n");
	(void) fprintf(stderr, " -y --livelist "
	"perform livelist and metaslab validation on any livelists being "
	"deleted\n\n");
	(void) fprintf(stderr, " Below options are intended for use "
	"with other options:\n");
	(void) fprintf(stderr, " -A --ignore-assertions "
	"ignore assertions (-A), enable panic recovery (-AA) or both "
	"(-AAA)\n");
	(void) fprintf(stderr, " -e --exported "
	"pool is exported/destroyed/has altroot/not in a cachefile\n");
	(void) fprintf(stderr, " -F --automatic-rewind "
	"attempt automatic rewind within safe range of transaction "
	"groups\n");
	(void) fprintf(stderr, " -G --dump-debug-msg "
	"dump zfs_dbgmsg buffer before exiting\n");
	(void) fprintf(stderr, " -I --inflight=INTEGER "
	"specify the maximum number of checksumming I/Os "
	"[default is 200]\n");
	(void) fprintf(stderr, " -o --option=\"OPTION=INTEGER\" "
	"set global variable to an unsigned 32-bit integer\n");
	(void) fprintf(stderr, " -p --path==PATH "
	"use one or more with -e to specify path to vdev dir\n");
	(void) fprintf(stderr, " -P --parseable "
	"print numbers in parseable form\n");
	(void) fprintf(stderr, " -q --skip-label "
	"don't print label contents\n");
	(void) fprintf(stderr, " -t --txg=INTEGER "
	"highest txg to use when searching for uberblocks\n");
	(void) fprintf(stderr, " -u --uberblock "
	"uberblock\n");
	(void) fprintf(stderr, " -U --cachefile=PATH "
	"use alternate cachefile\n");
	(void) fprintf(stderr, " -V --verbatim "
	"do verbatim import\n");
	(void) fprintf(stderr, " -x --dump-blocks=PATH "
	"dump all read blocks into specified directory\n");
	(void) fprintf(stderr, " -X --extreme-rewind "
	"attempt extreme rewind (does not work with dataset)\n");
	(void) fprintf(stderr, " -Y --all-reconstruction "
	"attempt all reconstruction combinations for split blocks\n");
	(void) fprintf(stderr, " -Z --zstd-headers "
	"show ZSTD headers \n");
	(void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
	"to make only that option verbose\n");
	(void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
	exit(1);
	}

	static void
	dump_debug_buffer(void)
	{
	if (dump_opt['G']) {
	(void) printf("\n");
	(void) fflush(stdout);
	zfs_dbgmsg_print("zdb");
	}
	}

	/*
	* Called for usage errors that are discovered after a call to spa_open(),
	* dmu_bonus_hold(), or pool_match(). abort() is called for other errors.
	*/

	static void
	fatal(const char *fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);
	(void) fprintf(stderr, "%s: ", cmdname);
	(void) vfprintf(stderr, fmt, ap);
	va_end(ap);
	(void) fprintf(stderr, "\n");

	dump_debug_buffer();

	exit(1);
	}

	static void
	dump_packed_nvlist(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) size;
	nvlist_t *nv;
	size_t nvsize = (uint64_t )data;
	char *packed = umem_alloc(nvsize, UMEM_NOFAIL);

	VERIFY(0 == dmu_read(os, object, 0, nvsize, packed, DMU_READ_PREFETCH));

	VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0);

	umem_free(packed, nvsize);

	dump_nvlist(nv, 8);

	nvlist_free(nv);
	}

	static void
	dump_history_offsets(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) size;
	spa_history_phys_t *shp = data;

	if (shp == NULL)
	return;

	(void) printf("\t\tpool_create_len = %llu\n",
	(u_longlong_t)shp->sh_pool_create_len);
	(void) printf("\t\tphys_max_off = %llu\n",
	(u_longlong_t)shp->sh_phys_max_off);
	(void) printf("\t\tbof = %llu\n",
	(u_longlong_t)shp->sh_bof);
	(void) printf("\t\teof = %llu\n",
	(u_longlong_t)shp->sh_eof);
	(void) printf("\t\trecords_lost = %llu\n",
	(u_longlong_t)shp->sh_records_lost);
	}

	static void
	zdb_nicenum(uint64_t num, char *buf, size_t buflen)
	{
	if (dump_opt['P'])
	(void) snprintf(buf, buflen, "%llu", (longlong_t)num);
	else
	- nicenum(num, buf, sizeof (buf));
	+ nicenum(num, buf, buflen);
	}

	static const char histo_stars[] = "****************************************";
	static const uint64_t histo_width = sizeof (histo_stars) - 1;

	static void
	dump_histogram(const uint64_t *histo, int size, int offset)
	{
	int i;
	int minidx = size - 1;
	int maxidx = 0;
	uint64_t max = 0;

	for (i = 0; i < size; i++) {
	if (histo[i] > max)
	max = histo[i];
	if (histo[i] > 0 && i > maxidx)
	maxidx = i;
	if (histo[i] > 0 && i < minidx)
	minidx = i;
	}

	if (max < histo_width)
	max = histo_width;

	for (i = minidx; i <= maxidx; i++) {
	(void) printf("\t\t\t%3u: %6llu %s\n",
	i + offset, (u_longlong_t)histo[i],
	&histo_stars[(max - histo[i]) * histo_width / max]);
	}
	}

	static void
	dump_zap_stats(objset_t *os, uint64_t object)
	{
	int error;
	zap_stats_t zs;

	error = zap_get_stats(os, object, &zs);
	if (error)
	return;

	if (zs.zs_ptrtbl_len == 0) {
	ASSERT(zs.zs_num_blocks == 1);
	(void) printf("\tmicrozap: %llu bytes, %llu entries\n",
	(u_longlong_t)zs.zs_blocksize,
	(u_longlong_t)zs.zs_num_entries);
	return;
	}

	(void) printf("\tFat ZAP stats:\n");

	(void) printf("\t\tPointer table:\n");
	(void) printf("\t\t\t%llu elements\n",
	(u_longlong_t)zs.zs_ptrtbl_len);
	(void) printf("\t\t\tzt_blk: %llu\n",
	(u_longlong_t)zs.zs_ptrtbl_zt_blk);
	(void) printf("\t\t\tzt_numblks: %llu\n",
	(u_longlong_t)zs.zs_ptrtbl_zt_numblks);
	(void) printf("\t\t\tzt_shift: %llu\n",
	(u_longlong_t)zs.zs_ptrtbl_zt_shift);
	(void) printf("\t\t\tzt_blks_copied: %llu\n",
	(u_longlong_t)zs.zs_ptrtbl_blks_copied);
	(void) printf("\t\t\tzt_nextblk: %llu\n",
	(u_longlong_t)zs.zs_ptrtbl_nextblk);

	(void) printf("\t\tZAP entries: %llu\n",
	(u_longlong_t)zs.zs_num_entries);
	(void) printf("\t\tLeaf blocks: %llu\n",
	(u_longlong_t)zs.zs_num_leafs);
	(void) printf("\t\tTotal blocks: %llu\n",
	(u_longlong_t)zs.zs_num_blocks);
	(void) printf("\t\tzap_block_type: 0x%llx\n",
	(u_longlong_t)zs.zs_block_type);
	(void) printf("\t\tzap_magic: 0x%llx\n",
	(u_longlong_t)zs.zs_magic);
	(void) printf("\t\tzap_salt: 0x%llx\n",
	(u_longlong_t)zs.zs_salt);

	(void) printf("\t\tLeafs with 2^n pointers:\n");
	dump_histogram(zs.zs_leafs_with_2n_pointers, ZAP_HISTOGRAM_SIZE, 0);

	(void) printf("\t\tBlocks with n*5 entries:\n");
	dump_histogram(zs.zs_blocks_with_n5_entries, ZAP_HISTOGRAM_SIZE, 0);

	(void) printf("\t\tBlocks n/10 full:\n");
	dump_histogram(zs.zs_blocks_n_tenths_full, ZAP_HISTOGRAM_SIZE, 0);

	(void) printf("\t\tEntries with n chunks:\n");
	dump_histogram(zs.zs_entries_using_n_chunks, ZAP_HISTOGRAM_SIZE, 0);

	(void) printf("\t\tBuckets with n entries:\n");
	dump_histogram(zs.zs_buckets_with_n_entries, ZAP_HISTOGRAM_SIZE, 0);
	}

	static void
	dump_none(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	}

	static void
	dump_unknown(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	(void) printf("\tUNKNOWN OBJECT TYPE\n");
	}

	static void
	dump_uint8(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	}

	static void
	dump_uint64(objset_t os, uint64_t object, void data, size_t size)
	{
	uint64_t *arr;
	uint64_t oursize;
	if (dump_opt['d'] < 6)
	return;

	if (data == NULL) {
	dmu_object_info_t doi;

	VERIFY0(dmu_object_info(os, object, &doi));
	size = doi.doi_max_offset;
	/*
	* We cap the size at 1 mebibyte here to prevent
	* allocation failures and nigh-infinite printing if the
	* object is extremely large.
	*/
	oursize = MIN(size, 1 << 20);
	arr = kmem_alloc(oursize, KM_SLEEP);

	int err = dmu_read(os, object, 0, oursize, arr, 0);
	if (err != 0) {
	(void) printf("got error %u from dmu_read\n", err);
	kmem_free(arr, oursize);
	return;
	}
	} else {
	/*
	* Even though the allocation is already done in this code path,
	* we still cap the size to prevent excessive printing.
	*/
	oursize = MIN(size, 1 << 20);
	arr = data;
	}

	if (size == 0) {
	+ if (data == NULL)
	+ kmem_free(arr, oursize);
	(void) printf("\t\t[]\n");
	return;
	}

	(void) printf("\t\t[%0llx", (u_longlong_t)arr[0]);
	for (size_t i = 1; i * sizeof (uint64_t) < oursize; i++) {
	if (i % 4 != 0)
	(void) printf(", %0llx", (u_longlong_t)arr[i]);
	else
	(void) printf(",\n\t\t%0llx", (u_longlong_t)arr[i]);
	}
	if (oursize != size)
	(void) printf(", ... ");
	(void) printf("]\n");

	if (data == NULL)
	kmem_free(arr, oursize);
	}

	static void
	dump_zap(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	zap_cursor_t zc;
	zap_attribute_t attr;
	void *prop;
	unsigned i;

	dump_zap_stats(os, object);
	(void) printf("\n");

	for (zap_cursor_init(&zc, os, object);
	zap_cursor_retrieve(&zc, &attr) == 0;
	zap_cursor_advance(&zc)) {
	(void) printf("\t\t%s = ", attr.za_name);
	if (attr.za_num_integers == 0) {
	(void) printf("\n");
	continue;
	}
	prop = umem_zalloc(attr.za_num_integers *
	attr.za_integer_length, UMEM_NOFAIL);
	(void) zap_lookup(os, object, attr.za_name,
	attr.za_integer_length, attr.za_num_integers, prop);
	if (attr.za_integer_length == 1) {
	if (strcmp(attr.za_name,
	DSL_CRYPTO_KEY_MASTER_KEY) == 0 \|\|
	strcmp(attr.za_name,
	DSL_CRYPTO_KEY_HMAC_KEY) == 0 \|\|
	strcmp(attr.za_name, DSL_CRYPTO_KEY_IV) == 0 \|\|
	strcmp(attr.za_name, DSL_CRYPTO_KEY_MAC) == 0 \|\|
	strcmp(attr.za_name, DMU_POOL_CHECKSUM_SALT) == 0) {
	uint8_t *u8 = prop;

	for (i = 0; i < attr.za_num_integers; i++) {
	(void) printf("%02x", u8[i]);
	}
	} else {
	(void) printf("%s", (char *)prop);
	}
	} else {
	for (i = 0; i < attr.za_num_integers; i++) {
	switch (attr.za_integer_length) {
	case 2:
	(void) printf("%u ",
	((uint16_t *)prop)[i]);
	break;
	case 4:
	(void) printf("%u ",
	((uint32_t *)prop)[i]);
	break;
	case 8:
	(void) printf("%lld ",
	(u_longlong_t)((int64_t *)prop)[i]);
	break;
	}
	}
	}
	(void) printf("\n");
	umem_free(prop, attr.za_num_integers * attr.za_integer_length);
	}
	zap_cursor_fini(&zc);
	}

	static void
	dump_bpobj(objset_t os, uint64_t object, void data, size_t size)
	{
	bpobj_phys_t *bpop = data;
	uint64_t i;
	char bytes[32], comp[32], uncomp[32];

	/* make sure the output won't get truncated */
	_Static_assert(sizeof (bytes) >= NN_NUMBUF_SZ, "bytes truncated");
	_Static_assert(sizeof (comp) >= NN_NUMBUF_SZ, "comp truncated");
	_Static_assert(sizeof (uncomp) >= NN_NUMBUF_SZ, "uncomp truncated");

	if (bpop == NULL)
	return;

	zdb_nicenum(bpop->bpo_bytes, bytes, sizeof (bytes));
	zdb_nicenum(bpop->bpo_comp, comp, sizeof (comp));
	zdb_nicenum(bpop->bpo_uncomp, uncomp, sizeof (uncomp));

	(void) printf("\t\tnum_blkptrs = %llu\n",
	(u_longlong_t)bpop->bpo_num_blkptrs);
	(void) printf("\t\tbytes = %s\n", bytes);
	if (size >= BPOBJ_SIZE_V1) {
	(void) printf("\t\tcomp = %s\n", comp);
	(void) printf("\t\tuncomp = %s\n", uncomp);
	}
	if (size >= BPOBJ_SIZE_V2) {
	(void) printf("\t\tsubobjs = %llu\n",
	(u_longlong_t)bpop->bpo_subobjs);
	(void) printf("\t\tnum_subobjs = %llu\n",
	(u_longlong_t)bpop->bpo_num_subobjs);
	}
	if (size >= sizeof (*bpop)) {
	(void) printf("\t\tnum_freed = %llu\n",
	(u_longlong_t)bpop->bpo_num_freed);
	}

	if (dump_opt['d'] < 5)
	return;

	for (i = 0; i < bpop->bpo_num_blkptrs; i++) {
	char blkbuf[BP_SPRINTF_LEN];
	blkptr_t bp;

	int err = dmu_read(os, object,
	i * sizeof (bp), sizeof (bp), &bp, 0);
	if (err != 0) {
	(void) printf("got error %u from dmu_read\n", err);
	break;
	}
	snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), &bp,
	BP_GET_FREE(&bp));
	(void) printf("\t%s\n", blkbuf);
	}
	}

	static void
	dump_bpobj_subobjs(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	dmu_object_info_t doi;
	int64_t i;

	VERIFY0(dmu_object_info(os, object, &doi));
	uint64_t *subobjs = kmem_alloc(doi.doi_max_offset, KM_SLEEP);

	int err = dmu_read(os, object, 0, doi.doi_max_offset, subobjs, 0);
	if (err != 0) {
	(void) printf("got error %u from dmu_read\n", err);
	kmem_free(subobjs, doi.doi_max_offset);
	return;
	}

	int64_t last_nonzero = -1;
	for (i = 0; i < doi.doi_max_offset / 8; i++) {
	if (subobjs[i] != 0)
	last_nonzero = i;
	}

	for (i = 0; i <= last_nonzero; i++) {
	(void) printf("\t%llu\n", (u_longlong_t)subobjs[i]);
	}
	kmem_free(subobjs, doi.doi_max_offset);
	}

	static void
	dump_ddt_zap(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	dump_zap_stats(os, object);
	/* contents are printed elsewhere, properly decoded */
	}

	static void
	dump_sa_attrs(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	zap_cursor_t zc;
	zap_attribute_t attr;

	dump_zap_stats(os, object);
	(void) printf("\n");

	for (zap_cursor_init(&zc, os, object);
	zap_cursor_retrieve(&zc, &attr) == 0;
	zap_cursor_advance(&zc)) {
	(void) printf("\t\t%s = ", attr.za_name);
	if (attr.za_num_integers == 0) {
	(void) printf("\n");
	continue;
	}
	(void) printf(" %llx : [%d:%d:%d]\n",
	(u_longlong_t)attr.za_first_integer,
	(int)ATTR_LENGTH(attr.za_first_integer),
	(int)ATTR_BSWAP(attr.za_first_integer),
	(int)ATTR_NUM(attr.za_first_integer));
	}
	zap_cursor_fini(&zc);
	}

	static void
	dump_sa_layouts(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	zap_cursor_t zc;
	zap_attribute_t attr;
	uint16_t *layout_attrs;
	unsigned i;

	dump_zap_stats(os, object);
	(void) printf("\n");

	for (zap_cursor_init(&zc, os, object);
	zap_cursor_retrieve(&zc, &attr) == 0;
	zap_cursor_advance(&zc)) {
	(void) printf("\t\t%s = [", attr.za_name);
	if (attr.za_num_integers == 0) {
	(void) printf("\n");
	continue;
	}

	VERIFY(attr.za_integer_length == 2);
	layout_attrs = umem_zalloc(attr.za_num_integers *
	attr.za_integer_length, UMEM_NOFAIL);

	VERIFY(zap_lookup(os, object, attr.za_name,
	attr.za_integer_length,
	attr.za_num_integers, layout_attrs) == 0);

	for (i = 0; i != attr.za_num_integers; i++)
	(void) printf(" %d ", (int)layout_attrs[i]);
	(void) printf("]\n");
	umem_free(layout_attrs,
	attr.za_num_integers * attr.za_integer_length);
	}
	zap_cursor_fini(&zc);
	}

	static void
	dump_zpldir(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	zap_cursor_t zc;
	zap_attribute_t attr;
	const char *typenames[] = {
	/* 0 */ "not specified",
	/* 1 */ "FIFO",
	/* 2 */ "Character Device",
	/* 3 */ "3 (invalid)",
	/* 4 */ "Directory",
	/* 5 */ "5 (invalid)",
	/* 6 */ "Block Device",
	/* 7 */ "7 (invalid)",
	/* 8 */ "Regular File",
	/* 9 */ "9 (invalid)",
	/* 10 */ "Symbolic Link",
	/* 11 */ "11 (invalid)",
	/* 12 */ "Socket",
	/* 13 */ "Door",
	/* 14 */ "Event Port",
	/* 15 */ "15 (invalid)",
	};

	dump_zap_stats(os, object);
	(void) printf("\n");

	for (zap_cursor_init(&zc, os, object);
	zap_cursor_retrieve(&zc, &attr) == 0;
	zap_cursor_advance(&zc)) {
	(void) printf("\t\t%s = %lld (type: %s)\n",
	attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer),
	typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]);
	}
	zap_cursor_fini(&zc);
	}

	static int
	get_dtl_refcount(vdev_t *vd)
	{
	int refcount = 0;

	if (vd->vdev_ops->vdev_op_leaf) {
	space_map_t *sm = vd->vdev_dtl_sm;

	if (sm != NULL &&
	sm->sm_dbuf->db_size == sizeof (space_map_phys_t))
	return (1);
	return (0);
	}

	for (unsigned c = 0; c < vd->vdev_children; c++)
	refcount += get_dtl_refcount(vd->vdev_child[c]);
	return (refcount);
	}

	static int
	get_metaslab_refcount(vdev_t *vd)
	{
	int refcount = 0;

	if (vd->vdev_top == vd) {
	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
	space_map_t *sm = vd->vdev_ms[m]->ms_sm;

	if (sm != NULL &&
	sm->sm_dbuf->db_size == sizeof (space_map_phys_t))
	refcount++;
	}
	}
	for (unsigned c = 0; c < vd->vdev_children; c++)
	refcount += get_metaslab_refcount(vd->vdev_child[c]);

	return (refcount);
	}

	static int
	get_obsolete_refcount(vdev_t *vd)
	{
	uint64_t obsolete_sm_object;
	int refcount = 0;

	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (vd->vdev_top == vd && obsolete_sm_object != 0) {
	dmu_object_info_t doi;
	VERIFY0(dmu_object_info(vd->vdev_spa->spa_meta_objset,
	obsolete_sm_object, &doi));
	if (doi.doi_bonus_size == sizeof (space_map_phys_t)) {
	refcount++;
	}
	} else {
	ASSERT3P(vd->vdev_obsolete_sm, ==, NULL);
	ASSERT3U(obsolete_sm_object, ==, 0);
	}
	for (unsigned c = 0; c < vd->vdev_children; c++) {
	refcount += get_obsolete_refcount(vd->vdev_child[c]);
	}

	return (refcount);
	}

	static int
	get_prev_obsolete_spacemap_refcount(spa_t *spa)
	{
	uint64_t prev_obj =
	spa->spa_condensing_indirect_phys.scip_prev_obsolete_sm_object;
	if (prev_obj != 0) {
	dmu_object_info_t doi;
	VERIFY0(dmu_object_info(spa->spa_meta_objset, prev_obj, &doi));
	if (doi.doi_bonus_size == sizeof (space_map_phys_t)) {
	return (1);
	}
	}
	return (0);
	}

	static int
	get_checkpoint_refcount(vdev_t *vd)
	{
	int refcount = 0;

	if (vd->vdev_top == vd && vd->vdev_top_zap != 0 &&
	zap_contains(spa_meta_objset(vd->vdev_spa),
	vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) == 0)
	refcount++;

	for (uint64_t c = 0; c < vd->vdev_children; c++)
	refcount += get_checkpoint_refcount(vd->vdev_child[c]);

	return (refcount);
	}

	static int
	get_log_spacemap_refcount(spa_t *spa)
	{
	return (avl_numnodes(&spa->spa_sm_logs_by_txg));
	}

	static int
	verify_spacemap_refcounts(spa_t *spa)
	{
	uint64_t expected_refcount = 0;
	uint64_t actual_refcount;

	(void) feature_get_refcount(spa,
	&spa_feature_table[SPA_FEATURE_SPACEMAP_HISTOGRAM],
	&expected_refcount);
	actual_refcount = get_dtl_refcount(spa->spa_root_vdev);
	actual_refcount += get_metaslab_refcount(spa->spa_root_vdev);
	actual_refcount += get_obsolete_refcount(spa->spa_root_vdev);
	actual_refcount += get_prev_obsolete_spacemap_refcount(spa);
	actual_refcount += get_checkpoint_refcount(spa->spa_root_vdev);
	actual_refcount += get_log_spacemap_refcount(spa);

	if (expected_refcount != actual_refcount) {
	(void) printf("space map refcount mismatch: expected %lld != "
	"actual %lld\n",
	(longlong_t)expected_refcount,
	(longlong_t)actual_refcount);
	return (2);
	}
	return (0);
	}

	static void
	dump_spacemap(objset_t os, space_map_t sm)
	{
	const char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
	"INVALID", "INVALID", "INVALID", "INVALID" };

	if (sm == NULL)
	return;

	(void) printf("space map object %llu:\n",
	(longlong_t)sm->sm_object);
	(void) printf(" smp_length = 0x%llx\n",
	(longlong_t)sm->sm_phys->smp_length);
	(void) printf(" smp_alloc = 0x%llx\n",
	(longlong_t)sm->sm_phys->smp_alloc);

	if (dump_opt['d'] < 6 && dump_opt['m'] < 4)
	return;

	/*
	* Print out the freelist entries in both encoded and decoded form.
	*/
	uint8_t mapshift = sm->sm_shift;
	int64_t alloc = 0;
	uint64_t word, entry_id = 0;
	for (uint64_t offset = 0; offset < space_map_length(sm);
	offset += sizeof (word)) {

	VERIFY0(dmu_read(os, space_map_object(sm), offset,
	sizeof (word), &word, DMU_READ_PREFETCH));

	if (sm_entry_is_debug(word)) {
	uint64_t de_txg = SM_DEBUG_TXG_DECODE(word);
	uint64_t de_sync_pass = SM_DEBUG_SYNCPASS_DECODE(word);
	if (de_txg == 0) {
	(void) printf(
	"\t [%6llu] PADDING\n",
	(u_longlong_t)entry_id);
	} else {
	(void) printf(
	"\t [%6llu] %s: txg %llu pass %llu\n",
	(u_longlong_t)entry_id,
	ddata[SM_DEBUG_ACTION_DECODE(word)],
	(u_longlong_t)de_txg,
	(u_longlong_t)de_sync_pass);
	}
	entry_id++;
	continue;
	}

	uint8_t words;
	char entry_type;
	uint64_t entry_off, entry_run, entry_vdev = SM_NO_VDEVID;

	if (sm_entry_is_single_word(word)) {
	entry_type = (SM_TYPE_DECODE(word) == SM_ALLOC) ?
	'A' : 'F';
	entry_off = (SM_OFFSET_DECODE(word) << mapshift) +
	sm->sm_start;
	entry_run = SM_RUN_DECODE(word) << mapshift;
	words = 1;
	} else {
	/* it is a two-word entry so we read another word */
	ASSERT(sm_entry_is_double_word(word));

	uint64_t extra_word;
	offset += sizeof (extra_word);
	VERIFY0(dmu_read(os, space_map_object(sm), offset,
	sizeof (extra_word), &extra_word,
	DMU_READ_PREFETCH));

	ASSERT3U(offset, <=, space_map_length(sm));

	entry_run = SM2_RUN_DECODE(word) << mapshift;
	entry_vdev = SM2_VDEV_DECODE(word);
	entry_type = (SM2_TYPE_DECODE(extra_word) == SM_ALLOC) ?
	'A' : 'F';
	entry_off = (SM2_OFFSET_DECODE(extra_word) <<
	mapshift) + sm->sm_start;
	words = 2;
	}

	(void) printf("\t [%6llu] %c range:"
	" %010llx-%010llx size: %06llx vdev: %06llu words: %u\n",
	(u_longlong_t)entry_id,
	entry_type, (u_longlong_t)entry_off,
	(u_longlong_t)(entry_off + entry_run),
	(u_longlong_t)entry_run,
	(u_longlong_t)entry_vdev, words);

	if (entry_type == 'A')
	alloc += entry_run;
	else
	alloc -= entry_run;
	entry_id++;
	}
	if (alloc != space_map_allocated(sm)) {
	(void) printf("space_map_object alloc (%lld) INCONSISTENT "
	"with space map summary (%lld)\n",
	(longlong_t)space_map_allocated(sm), (longlong_t)alloc);
	}
	}

	static void
	dump_metaslab_stats(metaslab_t *msp)
	{
	char maxbuf[32];
	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_t *t = &msp->ms_allocatable_by_size;
	int free_pct = range_tree_space(rt) * 100 / msp->ms_size;

	/* max sure nicenum has enough space */
	_Static_assert(sizeof (maxbuf) >= NN_NUMBUF_SZ, "maxbuf truncated");

	zdb_nicenum(metaslab_largest_allocatable(msp), maxbuf, sizeof (maxbuf));

	(void) printf("\t %25s %10lu %7s %6s %4s %4d%%\n",
	"segments", zfs_btree_numnodes(t), "maxsize", maxbuf,
	"freepct", free_pct);
	(void) printf("\tIn-memory histogram:\n");
	dump_histogram(rt->rt_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
	}

	static void
	dump_metaslab(metaslab_t *msp)
	{
	vdev_t *vd = msp->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	space_map_t *sm = msp->ms_sm;
	char freebuf[32];

	zdb_nicenum(msp->ms_size - space_map_allocated(sm), freebuf,
	sizeof (freebuf));

	(void) printf(
	"\tmetaslab %6llu offset %12llx spacemap %6llu free %5s\n",
	(u_longlong_t)msp->ms_id, (u_longlong_t)msp->ms_start,
	(u_longlong_t)space_map_object(sm), freebuf);

	if (dump_opt['m'] > 2 && !dump_opt['L']) {
	mutex_enter(&msp->ms_lock);
	VERIFY0(metaslab_load(msp));
	range_tree_stat_verify(msp->ms_allocatable);
	dump_metaslab_stats(msp);
	metaslab_unload(msp);
	mutex_exit(&msp->ms_lock);
	}

	if (dump_opt['m'] > 1 && sm != NULL &&
	spa_feature_is_active(spa, SPA_FEATURE_SPACEMAP_HISTOGRAM)) {
	/*
	* The space map histogram represents free space in chunks
	* of sm_shift (i.e. bucket 0 refers to 2^sm_shift).
	*/
	(void) printf("\tOn-disk histogram:\t\tfragmentation %llu\n",
	(u_longlong_t)msp->ms_fragmentation);
	dump_histogram(sm->sm_phys->smp_histogram,
	SPACE_MAP_HISTOGRAM_SIZE, sm->sm_shift);
	}

	if (vd->vdev_ops == &vdev_draid_ops)
	ASSERT3U(msp->ms_size, <=, 1ULL << vd->vdev_ms_shift);
	else
	ASSERT3U(msp->ms_size, ==, 1ULL << vd->vdev_ms_shift);

	dump_spacemap(spa->spa_meta_objset, msp->ms_sm);

	if (spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
	(void) printf("\tFlush data:\n\tunflushed txg=%llu\n\n",
	(u_longlong_t)metaslab_unflushed_txg(msp));
	}
	}

	static void
	print_vdev_metaslab_header(vdev_t *vd)
	{
	vdev_alloc_bias_t alloc_bias = vd->vdev_alloc_bias;
	const char *bias_str = "";
	if (alloc_bias == VDEV_BIAS_LOG \|\| vd->vdev_islog) {
	bias_str = VDEV_ALLOC_BIAS_LOG;
	} else if (alloc_bias == VDEV_BIAS_SPECIAL) {
	bias_str = VDEV_ALLOC_BIAS_SPECIAL;
	} else if (alloc_bias == VDEV_BIAS_DEDUP) {
	bias_str = VDEV_ALLOC_BIAS_DEDUP;
	}

	uint64_t ms_flush_data_obj = 0;
	if (vd->vdev_top_zap != 0) {
	int error = zap_lookup(spa_meta_objset(vd->vdev_spa),
	vd->vdev_top_zap, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS,
	sizeof (uint64_t), 1, &ms_flush_data_obj);
	if (error != ENOENT) {
	ASSERT0(error);
	}
	}

	(void) printf("\tvdev %10llu %s",
	(u_longlong_t)vd->vdev_id, bias_str);

	if (ms_flush_data_obj != 0) {
	(void) printf(" ms_unflushed_phys object %llu",
	(u_longlong_t)ms_flush_data_obj);
	}

	(void) printf("\n\t%-10s%5llu %-19s %-15s %-12s\n",
	"metaslabs", (u_longlong_t)vd->vdev_ms_count,
	"offset", "spacemap", "free");
	(void) printf("\t%15s %19s %15s %12s\n",
	"---------------", "-------------------",
	"---------------", "------------");
	}

	static void
	dump_metaslab_groups(spa_t *spa, boolean_t show_special)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	metaslab_class_t *mc = spa_normal_class(spa);
	metaslab_class_t *smc = spa_special_class(spa);
	uint64_t fragmentation;

	metaslab_class_histogram_verify(mc);

	for (unsigned c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = tvd->vdev_mg;

	if (mg == NULL \|\| (mg->mg_class != mc &&
	(!show_special \|\| mg->mg_class != smc)))
	continue;

	metaslab_group_histogram_verify(mg);
	mg->mg_fragmentation = metaslab_group_fragmentation(mg);

	(void) printf("\tvdev %10llu\t\tmetaslabs%5llu\t\t"
	"fragmentation",
	(u_longlong_t)tvd->vdev_id,
	(u_longlong_t)tvd->vdev_ms_count);
	if (mg->mg_fragmentation == ZFS_FRAG_INVALID) {
	(void) printf("%3s\n", "-");
	} else {
	(void) printf("%3llu%%\n",
	(u_longlong_t)mg->mg_fragmentation);
	}
	dump_histogram(mg->mg_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
	}

	(void) printf("\tpool %s\tfragmentation", spa_name(spa));
	fragmentation = metaslab_class_fragmentation(mc);
	if (fragmentation == ZFS_FRAG_INVALID)
	(void) printf("\t%3s\n", "-");
	else
	(void) printf("\t%3llu%%\n", (u_longlong_t)fragmentation);
	dump_histogram(mc->mc_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0);
	}

	static void
	print_vdev_indirect(vdev_t *vd)
	{
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	vdev_indirect_births_t *vib = vd->vdev_indirect_births;

	if (vim == NULL) {
	ASSERT3P(vib, ==, NULL);
	return;
	}

	ASSERT3U(vdev_indirect_mapping_object(vim), ==,
	vic->vic_mapping_object);
	ASSERT3U(vdev_indirect_births_object(vib), ==,
	vic->vic_births_object);

	(void) printf("indirect births obj %llu:\n",
	(longlong_t)vic->vic_births_object);
	(void) printf(" vib_count = %llu\n",
	(longlong_t)vdev_indirect_births_count(vib));
	for (uint64_t i = 0; i < vdev_indirect_births_count(vib); i++) {
	vdev_indirect_birth_entry_phys_t *cur_vibe =
	&vib->vib_entries[i];
	(void) printf("\toffset %llx -> txg %llu\n",
	(longlong_t)cur_vibe->vibe_offset,
	(longlong_t)cur_vibe->vibe_phys_birth_txg);
	}
	(void) printf("\n");

	(void) printf("indirect mapping obj %llu:\n",
	(longlong_t)vic->vic_mapping_object);
	(void) printf(" vim_max_offset = 0x%llx\n",
	(longlong_t)vdev_indirect_mapping_max_offset(vim));
	(void) printf(" vim_bytes_mapped = 0x%llx\n",
	(longlong_t)vdev_indirect_mapping_bytes_mapped(vim));
	(void) printf(" vim_count = %llu\n",
	(longlong_t)vdev_indirect_mapping_num_entries(vim));

	if (dump_opt['d'] <= 5 && dump_opt['m'] <= 3)
	return;

	uint32_t *counts = vdev_indirect_mapping_load_obsolete_counts(vim);

	for (uint64_t i = 0; i < vdev_indirect_mapping_num_entries(vim); i++) {
	vdev_indirect_mapping_entry_phys_t *vimep =
	&vim->vim_entries[i];
	(void) printf("\t<%llx:%llx:%llx> -> "
	"<%llx:%llx:%llx> (%x obsolete)\n",
	(longlong_t)vd->vdev_id,
	(longlong_t)DVA_MAPPING_GET_SRC_OFFSET(vimep),
	(longlong_t)DVA_GET_ASIZE(&vimep->vimep_dst),
	(longlong_t)DVA_GET_VDEV(&vimep->vimep_dst),
	(longlong_t)DVA_GET_OFFSET(&vimep->vimep_dst),
	(longlong_t)DVA_GET_ASIZE(&vimep->vimep_dst),
	counts[i]);
	}
	(void) printf("\n");

	uint64_t obsolete_sm_object;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	(void) printf("obsolete space map object %llu:\n",
	(u_longlong_t)obsolete_sm_object);
	ASSERT(vd->vdev_obsolete_sm != NULL);
	ASSERT3U(space_map_object(vd->vdev_obsolete_sm), ==,
	obsolete_sm_object);
	dump_spacemap(mos, vd->vdev_obsolete_sm);
	(void) printf("\n");
	}
	}

	static void
	dump_metaslabs(spa_t *spa)
	{
	vdev_t vd, rvd = spa->spa_root_vdev;
	uint64_t m, c = 0, children = rvd->vdev_children;

	(void) printf("\nMetaslabs:\n");

	if (!dump_opt['d'] && zopt_metaslab_args > 0) {
	c = zopt_metaslab[0];

	if (c >= children)
	(void) fatal("bad vdev id: %llu", (u_longlong_t)c);

	if (zopt_metaslab_args > 1) {
	vd = rvd->vdev_child[c];
	print_vdev_metaslab_header(vd);

	for (m = 1; m < zopt_metaslab_args; m++) {
	if (zopt_metaslab[m] < vd->vdev_ms_count)
	dump_metaslab(
	vd->vdev_ms[zopt_metaslab[m]]);
	else
	(void) fprintf(stderr, "bad metaslab "
	"number %llu\n",
	(u_longlong_t)zopt_metaslab[m]);
	}
	(void) printf("\n");
	return;
	}
	children = c + 1;
	}
	for (; c < children; c++) {
	vd = rvd->vdev_child[c];
	print_vdev_metaslab_header(vd);

	print_vdev_indirect(vd);

	for (m = 0; m < vd->vdev_ms_count; m++)
	dump_metaslab(vd->vdev_ms[m]);
	(void) printf("\n");
	}
	}

	static void
	dump_log_spacemaps(spa_t *spa)
	{
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
	return;

	(void) printf("\nLog Space Maps in Pool:\n");
	for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
	sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls)) {
	space_map_t *sm = NULL;
	VERIFY0(space_map_open(&sm, spa_meta_objset(spa),
	sls->sls_sm_obj, 0, UINT64_MAX, SPA_MINBLOCKSHIFT));

	(void) printf("Log Spacemap object %llu txg %llu\n",
	(u_longlong_t)sls->sls_sm_obj, (u_longlong_t)sls->sls_txg);
	dump_spacemap(spa->spa_meta_objset, sm);
	space_map_close(sm);
	}
	(void) printf("\n");
	}

	static void
	dump_dde(const ddt_t ddt, const ddt_entry_t dde, uint64_t index)
	{
	const ddt_phys_t *ddp = dde->dde_phys;
	const ddt_key_t *ddk = &dde->dde_key;
	const char *types[4] = { "ditto", "single", "double", "triple" };
	char blkbuf[BP_SPRINTF_LEN];
	blkptr_t blk;
	int p;

	for (p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0)
	continue;
	ddt_bp_create(ddt->ddt_checksum, ddk, ddp, &blk);
	snprintf_blkptr(blkbuf, sizeof (blkbuf), &blk);
	(void) printf("index %llx refcnt %llu %s %s\n",
	(u_longlong_t)index, (u_longlong_t)ddp->ddp_refcnt,
	types[p], blkbuf);
	}
	}

	static void
	dump_dedup_ratio(const ddt_stat_t *dds)
	{
	double rL, rP, rD, D, dedup, compress, copies;

	if (dds->dds_blocks == 0)
	return;

	rL = (double)dds->dds_ref_lsize;
	rP = (double)dds->dds_ref_psize;
	rD = (double)dds->dds_ref_dsize;
	D = (double)dds->dds_dsize;

	dedup = rD / D;
	compress = rL / rP;
	copies = rD / rP;

	(void) printf("dedup = %.2f, compress = %.2f, copies = %.2f, "
	"dedup * compress / copies = %.2f\n\n",
	dedup, compress, copies, dedup * compress / copies);
	}

	static void
	dump_ddt(ddt_t *ddt, enum ddt_type type, enum ddt_class class)
	{
	char name[DDT_NAMELEN];
	ddt_entry_t dde;
	uint64_t walk = 0;
	dmu_object_info_t doi;
	uint64_t count, dspace, mspace;
	int error;

	error = ddt_object_info(ddt, type, class, &doi);

	if (error == ENOENT)
	return;
	ASSERT(error == 0);

	error = ddt_object_count(ddt, type, class, &count);
	ASSERT(error == 0);
	if (count == 0)
	return;

	dspace = doi.doi_physical_blocks_512 << 9;
	mspace = doi.doi_fill_count * doi.doi_data_block_size;

	ddt_object_name(ddt, type, class, name);

	(void) printf("%s: %llu entries, size %llu on disk, %llu in core\n",
	name,
	(u_longlong_t)count,
	(u_longlong_t)(dspace / count),
	(u_longlong_t)(mspace / count));

	if (dump_opt['D'] < 3)
	return;

	zpool_dump_ddt(NULL, &ddt->ddt_histogram[type][class]);

	if (dump_opt['D'] < 4)
	return;

	if (dump_opt['D'] < 5 && class == DDT_CLASS_UNIQUE)
	return;

	(void) printf("%s contents:\n\n", name);

	while ((error = ddt_object_walk(ddt, type, class, &walk, &dde)) == 0)
	dump_dde(ddt, &dde, walk);

	ASSERT3U(error, ==, ENOENT);

	(void) printf("\n");
	}

	static void
	dump_all_ddts(spa_t *spa)
	{
	ddt_histogram_t ddh_total = {{{0}}};
	ddt_stat_t dds_total = {0};

	for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) {
	ddt_t *ddt = spa->spa_ddt[c];
	for (enum ddt_type type = 0; type < DDT_TYPES; type++) {
	for (enum ddt_class class = 0; class < DDT_CLASSES;
	class++) {
	dump_ddt(ddt, type, class);
	}
	}
	}

	ddt_get_dedup_stats(spa, &dds_total);

	if (dds_total.dds_blocks == 0) {
	(void) printf("All DDTs are empty\n");
	return;
	}

	(void) printf("\n");

	if (dump_opt['D'] > 1) {
	(void) printf("DDT histogram (aggregated over all DDTs):\n");
	ddt_get_dedup_histogram(spa, &ddh_total);
	zpool_dump_ddt(&dds_total, &ddh_total);
	}

	dump_dedup_ratio(&dds_total);
	}

	static void
	dump_dtl_seg(void *arg, uint64_t start, uint64_t size)
	{
	char *prefix = arg;

	(void) printf("%s [%llu,%llu) length %llu\n",
	prefix,
	(u_longlong_t)start,
	(u_longlong_t)(start + size),
	(u_longlong_t)(size));
	}

	static void
	dump_dtl(vdev_t *vd, int indent)
	{
	spa_t *spa = vd->vdev_spa;
	boolean_t required;
	const char *name[DTL_TYPES] = { "missing", "partial", "scrub",
	"outage" };
	char prefix[256];

	spa_vdev_state_enter(spa, SCL_NONE);
	required = vdev_dtl_required(vd);
	(void) spa_vdev_state_exit(spa, NULL, 0);

	if (indent == 0)
	(void) printf("\nDirty time logs:\n\n");

	(void) printf("\t%*s%s [%s]\n", indent, "",
	vd->vdev_path ? vd->vdev_path :
	vd->vdev_parent ? vd->vdev_ops->vdev_op_type : spa_name(spa),
	required ? "DTL-required" : "DTL-expendable");

	for (int t = 0; t < DTL_TYPES; t++) {
	range_tree_t *rt = vd->vdev_dtl[t];
	if (range_tree_space(rt) == 0)
	continue;
	(void) snprintf(prefix, sizeof (prefix), "\t%*s%s",
	indent + 2, "", name[t]);
	range_tree_walk(rt, dump_dtl_seg, prefix);
	if (dump_opt['d'] > 5 && vd->vdev_children == 0)
	dump_spacemap(spa->spa_meta_objset,
	vd->vdev_dtl_sm);
	}

	for (unsigned c = 0; c < vd->vdev_children; c++)
	dump_dtl(vd->vdev_child[c], indent + 4);
	}

	static void
	dump_history(spa_t *spa)
	{
	nvlist_t **events = NULL;
	char *buf;
	uint64_t resid, len, off = 0;
	uint_t num = 0;
	int error;
	char tbuf[30];

	if ((buf = malloc(SPA_OLD_MAXBLOCKSIZE)) == NULL) {
	(void) fprintf(stderr, "%s: unable to allocate I/O buffer\n",
	__func__);
	return;
	}

	do {
	len = SPA_OLD_MAXBLOCKSIZE;

	if ((error = spa_history_get(spa, &off, &len, buf)) != 0) {
	(void) fprintf(stderr, "Unable to read history: "
	"error %d\n", error);
	free(buf);
	return;
	}

	if (zpool_history_unpack(buf, len, &resid, &events, &num) != 0)
	break;

	off -= resid;
	} while (len != 0);

	(void) printf("\nHistory:\n");
	for (unsigned i = 0; i < num; i++) {
	boolean_t printed = B_FALSE;

	if (nvlist_exists(events[i], ZPOOL_HIST_TIME)) {
	time_t tsec;
	struct tm t;

	tsec = fnvlist_lookup_uint64(events[i],
	ZPOOL_HIST_TIME);
	(void) localtime_r(&tsec, &t);
	(void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t);
	} else {
	tbuf[0] = '\0';
	}

	if (nvlist_exists(events[i], ZPOOL_HIST_CMD)) {
	(void) printf("%s %s\n", tbuf,
	fnvlist_lookup_string(events[i], ZPOOL_HIST_CMD));
	} else if (nvlist_exists(events[i], ZPOOL_HIST_INT_EVENT)) {
	uint64_t ievent;

	ievent = fnvlist_lookup_uint64(events[i],
	ZPOOL_HIST_INT_EVENT);
	if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS)
	goto next;

	(void) printf(" %s [internal %s txg:%ju] %s\n",
	tbuf,
	zfs_history_event_names[ievent],
	fnvlist_lookup_uint64(events[i],
	ZPOOL_HIST_TXG),
	fnvlist_lookup_string(events[i],
	ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(events[i], ZPOOL_HIST_INT_NAME)) {
	(void) printf("%s [txg:%ju] %s", tbuf,
	fnvlist_lookup_uint64(events[i],
	ZPOOL_HIST_TXG),
	fnvlist_lookup_string(events[i],
	ZPOOL_HIST_INT_NAME));

	if (nvlist_exists(events[i], ZPOOL_HIST_DSNAME)) {
	(void) printf(" %s (%llu)",
	fnvlist_lookup_string(events[i],
	ZPOOL_HIST_DSNAME),
	(u_longlong_t)fnvlist_lookup_uint64(
	events[i],
	ZPOOL_HIST_DSID));
	}

	(void) printf(" %s\n", fnvlist_lookup_string(events[i],
	ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(events[i], ZPOOL_HIST_IOCTL)) {
	(void) printf("%s ioctl %s\n", tbuf,
	fnvlist_lookup_string(events[i],
	ZPOOL_HIST_IOCTL));

	if (nvlist_exists(events[i], ZPOOL_HIST_INPUT_NVL)) {
	(void) printf(" input:\n");
	dump_nvlist(fnvlist_lookup_nvlist(events[i],
	ZPOOL_HIST_INPUT_NVL), 8);
	}
	if (nvlist_exists(events[i], ZPOOL_HIST_OUTPUT_NVL)) {
	(void) printf(" output:\n");
	dump_nvlist(fnvlist_lookup_nvlist(events[i],
	ZPOOL_HIST_OUTPUT_NVL), 8);
	}
	if (nvlist_exists(events[i], ZPOOL_HIST_ERRNO)) {
	(void) printf(" errno: %lld\n",
	(longlong_t)fnvlist_lookup_int64(events[i],
	ZPOOL_HIST_ERRNO));
	}
	} else {
	goto next;
	}

	printed = B_TRUE;
	next:
	if (dump_opt['h'] > 1) {
	if (!printed)
	(void) printf("unrecognized record:\n");
	dump_nvlist(events[i], 2);
	}
	}
	free(buf);
	}

	static void
	dump_dnode(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	}

	static uint64_t
	blkid2offset(const dnode_phys_t dnp, const blkptr_t bp,
	const zbookmark_phys_t *zb)
	{
	if (dnp == NULL) {
	ASSERT(zb->zb_level < 0);
	if (zb->zb_object == 0)
	return (zb->zb_blkid);
	return (zb->zb_blkid * BP_GET_LSIZE(bp));
	}

	ASSERT(zb->zb_level >= 0);

	return ((zb->zb_blkid <<
	(zb->zb_level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) *
	dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
	}

	static void
	snprintf_zstd_header(spa_t spa, char blkbuf, size_t buflen,
	const blkptr_t *bp)
	{
	abd_t *pabd;
	void *buf;
	zio_t *zio;
	zfs_zstdhdr_t zstd_hdr;
	int error;

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_ZSTD)
	return;

	if (BP_IS_HOLE(bp))
	return;

	if (BP_IS_EMBEDDED(bp)) {
	buf = malloc(SPA_MAXBLOCKSIZE);
	if (buf == NULL) {
	(void) fprintf(stderr, "out of memory\n");
	exit(1);
	}
	decode_embedded_bp_compressed(bp, buf);
	memcpy(&zstd_hdr, buf, sizeof (zstd_hdr));
	free(buf);
	zstd_hdr.c_len = BE_32(zstd_hdr.c_len);
	zstd_hdr.raw_version_level = BE_32(zstd_hdr.raw_version_level);
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	" ZSTD:size=%u:version=%u:level=%u:EMBEDDED",
	zstd_hdr.c_len, zfs_get_hdrversion(&zstd_hdr),
	zfs_get_hdrlevel(&zstd_hdr));
	return;
	}

	pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
	zio = zio_root(spa, NULL, NULL, 0);

	/* Decrypt but don't decompress so we can read the compression header */
	zio_nowait(zio_read(zio, spa, bp, pabd, BP_GET_PSIZE(bp), NULL, NULL,
	ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW_COMPRESS,
	NULL));
	error = zio_wait(zio);
	if (error) {
	(void) fprintf(stderr, "read failed: %d\n", error);
	return;
	}
	buf = abd_borrow_buf_copy(pabd, BP_GET_LSIZE(bp));
	memcpy(&zstd_hdr, buf, sizeof (zstd_hdr));
	zstd_hdr.c_len = BE_32(zstd_hdr.c_len);
	zstd_hdr.raw_version_level = BE_32(zstd_hdr.raw_version_level);

	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	" ZSTD:size=%u:version=%u:level=%u:NORMAL",
	zstd_hdr.c_len, zfs_get_hdrversion(&zstd_hdr),
	zfs_get_hdrlevel(&zstd_hdr));

	abd_return_buf_copy(pabd, buf, BP_GET_LSIZE(bp));
	}

	static void
	snprintf_blkptr_compact(char blkbuf, size_t buflen, const blkptr_t bp,
	boolean_t bp_freed)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = dump_opt['d'] > 5 ? BP_GET_NDVAS(bp) : 1;
	int i;

	if (dump_opt['b'] >= 6) {
	snprintf_blkptr(blkbuf, buflen, bp);
	if (bp_freed) {
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf), " %s", "FREE");
	}
	return;
	}

	if (BP_IS_EMBEDDED(bp)) {
	(void) sprintf(blkbuf,
	"EMBEDDED et=%u %llxL/%llxP B=%llu",
	(int)BPE_GET_ETYPE(bp),
	(u_longlong_t)BPE_GET_LSIZE(bp),
	(u_longlong_t)BPE_GET_PSIZE(bp),
	(u_longlong_t)bp->blk_birth);
	return;
	}

	blkbuf[0] = '\0';

	for (i = 0; i < ndvas; i++)
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf), "%llu:%llx:%llx ",
	(u_longlong_t)DVA_GET_VDEV(&dva[i]),
	(u_longlong_t)DVA_GET_OFFSET(&dva[i]),
	(u_longlong_t)DVA_GET_ASIZE(&dva[i]));

	if (BP_IS_HOLE(bp)) {
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	"%llxL B=%llu",
	(u_longlong_t)BP_GET_LSIZE(bp),
	(u_longlong_t)bp->blk_birth);
	} else {
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	"%llxL/%llxP F=%llu B=%llu/%llu",
	(u_longlong_t)BP_GET_LSIZE(bp),
	(u_longlong_t)BP_GET_PSIZE(bp),
	(u_longlong_t)BP_GET_FILL(bp),
	(u_longlong_t)bp->blk_birth,
	(u_longlong_t)BP_PHYSICAL_BIRTH(bp));
	if (bp_freed)
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf), " %s", "FREE");
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf), " cksum=%llx:%llx:%llx:%llx",
	(u_longlong_t)bp->blk_cksum.zc_word[0],
	(u_longlong_t)bp->blk_cksum.zc_word[1],
	(u_longlong_t)bp->blk_cksum.zc_word[2],
	(u_longlong_t)bp->blk_cksum.zc_word[3]);
	}
	}

	static void
	print_indirect(spa_t spa, blkptr_t bp, const zbookmark_phys_t *zb,
	const dnode_phys_t *dnp)
	{
	char blkbuf[BP_SPRINTF_LEN];
	int l;

	if (!BP_IS_EMBEDDED(bp)) {
	ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
	ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
	}

	(void) printf("%16llx ", (u_longlong_t)blkid2offset(dnp, bp, zb));

	ASSERT(zb->zb_level >= 0);

	for (l = dnp->dn_nlevels - 1; l >= -1; l--) {
	if (l == zb->zb_level) {
	(void) printf("L%llx", (u_longlong_t)zb->zb_level);
	} else {
	(void) printf(" ");
	}
	}

	snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp, B_FALSE);
	if (dump_opt['Z'] && BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD)
	snprintf_zstd_header(spa, blkbuf, sizeof (blkbuf), bp);
	(void) printf("%s\n", blkbuf);
	}

	static int
	visit_indirect(spa_t spa, const dnode_phys_t dnp,
	blkptr_t bp, const zbookmark_phys_t zb)
	{
	int err = 0;

	if (bp->blk_birth == 0)
	return (0);

	print_indirect(spa, bp, zb, dnp);

	if (BP_GET_LEVEL(bp) > 0 && !BP_IS_HOLE(bp)) {
	arc_flags_t flags = ARC_FLAG_WAIT;
	int i;
	blkptr_t *cbp;
	int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
	arc_buf_t *buf;
	uint64_t fill = 0;
	ASSERT(!BP_IS_REDACTED(bp));

	err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
	if (err)
	return (err);
	ASSERT(buf->b_data);

	/* recursively visit blocks below this */
	cbp = buf->b_data;
	for (i = 0; i < epb; i++, cbp++) {
	zbookmark_phys_t czb;

	SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
	zb->zb_level - 1,
	zb->zb_blkid * epb + i);
	err = visit_indirect(spa, dnp, cbp, &czb);
	if (err)
	break;
	fill += BP_GET_FILL(cbp);
	}
	if (!err)
	ASSERT3U(fill, ==, BP_GET_FILL(bp));
	arc_buf_destroy(buf, &buf);
	}

	return (err);
	}

	static void
	dump_indirect(dnode_t *dn)
	{
	dnode_phys_t *dnp = dn->dn_phys;
	zbookmark_phys_t czb;

	(void) printf("Indirect blocks:\n");

	SET_BOOKMARK(&czb, dmu_objset_id(dn->dn_objset),
	dn->dn_object, dnp->dn_nlevels - 1, 0);
	for (int j = 0; j < dnp->dn_nblkptr; j++) {
	czb.zb_blkid = j;
	(void) visit_indirect(dmu_objset_spa(dn->dn_objset), dnp,
	&dnp->dn_blkptr[j], &czb);
	}

	(void) printf("\n");
	}

	static void
	dump_dsl_dir(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object;
	dsl_dir_phys_t *dd = data;
	time_t crtime;
	char nice[32];

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (nice) >= NN_NUMBUF_SZ, "nice truncated");

	if (dd == NULL)
	return;

	ASSERT3U(size, >=, sizeof (dsl_dir_phys_t));

	crtime = dd->dd_creation_time;
	(void) printf("\t\tcreation_time = %s", ctime(&crtime));
	(void) printf("\t\thead_dataset_obj = %llu\n",
	(u_longlong_t)dd->dd_head_dataset_obj);
	(void) printf("\t\tparent_dir_obj = %llu\n",
	(u_longlong_t)dd->dd_parent_obj);
	(void) printf("\t\torigin_obj = %llu\n",
	(u_longlong_t)dd->dd_origin_obj);
	(void) printf("\t\tchild_dir_zapobj = %llu\n",
	(u_longlong_t)dd->dd_child_dir_zapobj);
	zdb_nicenum(dd->dd_used_bytes, nice, sizeof (nice));
	(void) printf("\t\tused_bytes = %s\n", nice);
	zdb_nicenum(dd->dd_compressed_bytes, nice, sizeof (nice));
	(void) printf("\t\tcompressed_bytes = %s\n", nice);
	zdb_nicenum(dd->dd_uncompressed_bytes, nice, sizeof (nice));
	(void) printf("\t\tuncompressed_bytes = %s\n", nice);
	zdb_nicenum(dd->dd_quota, nice, sizeof (nice));
	(void) printf("\t\tquota = %s\n", nice);
	zdb_nicenum(dd->dd_reserved, nice, sizeof (nice));
	(void) printf("\t\treserved = %s\n", nice);
	(void) printf("\t\tprops_zapobj = %llu\n",
	(u_longlong_t)dd->dd_props_zapobj);
	(void) printf("\t\tdeleg_zapobj = %llu\n",
	(u_longlong_t)dd->dd_deleg_zapobj);
	(void) printf("\t\tflags = %llx\n",
	(u_longlong_t)dd->dd_flags);

	#define DO(which) \
	zdb_nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice, \
	sizeof (nice)); \
	(void) printf("\t\tused_breakdown[" #which "] = %s\n", nice)
	DO(HEAD);
	DO(SNAP);
	DO(CHILD);
	DO(CHILD_RSRV);
	DO(REFRSRV);
	#undef DO
	(void) printf("\t\tclones = %llu\n",
	(u_longlong_t)dd->dd_clones);
	}

	static void
	dump_dsl_dataset(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object;
	dsl_dataset_phys_t *ds = data;
	time_t crtime;
	char used[32], compressed[32], uncompressed[32], unique[32];
	char blkbuf[BP_SPRINTF_LEN];

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (used) >= NN_NUMBUF_SZ, "used truncated");
	_Static_assert(sizeof (compressed) >= NN_NUMBUF_SZ,
	"compressed truncated");
	_Static_assert(sizeof (uncompressed) >= NN_NUMBUF_SZ,
	"uncompressed truncated");
	_Static_assert(sizeof (unique) >= NN_NUMBUF_SZ, "unique truncated");

	if (ds == NULL)
	return;

	ASSERT(size == sizeof (*ds));
	crtime = ds->ds_creation_time;
	zdb_nicenum(ds->ds_referenced_bytes, used, sizeof (used));
	zdb_nicenum(ds->ds_compressed_bytes, compressed, sizeof (compressed));
	zdb_nicenum(ds->ds_uncompressed_bytes, uncompressed,
	sizeof (uncompressed));
	zdb_nicenum(ds->ds_unique_bytes, unique, sizeof (unique));
	snprintf_blkptr(blkbuf, sizeof (blkbuf), &ds->ds_bp);

	(void) printf("\t\tdir_obj = %llu\n",
	(u_longlong_t)ds->ds_dir_obj);
	(void) printf("\t\tprev_snap_obj = %llu\n",
	(u_longlong_t)ds->ds_prev_snap_obj);
	(void) printf("\t\tprev_snap_txg = %llu\n",
	(u_longlong_t)ds->ds_prev_snap_txg);
	(void) printf("\t\tnext_snap_obj = %llu\n",
	(u_longlong_t)ds->ds_next_snap_obj);
	(void) printf("\t\tsnapnames_zapobj = %llu\n",
	(u_longlong_t)ds->ds_snapnames_zapobj);
	(void) printf("\t\tnum_children = %llu\n",
	(u_longlong_t)ds->ds_num_children);
	(void) printf("\t\tuserrefs_obj = %llu\n",
	(u_longlong_t)ds->ds_userrefs_obj);
	(void) printf("\t\tcreation_time = %s", ctime(&crtime));
	(void) printf("\t\tcreation_txg = %llu\n",
	(u_longlong_t)ds->ds_creation_txg);
	(void) printf("\t\tdeadlist_obj = %llu\n",
	(u_longlong_t)ds->ds_deadlist_obj);
	(void) printf("\t\tused_bytes = %s\n", used);
	(void) printf("\t\tcompressed_bytes = %s\n", compressed);
	(void) printf("\t\tuncompressed_bytes = %s\n", uncompressed);
	(void) printf("\t\tunique = %s\n", unique);
	(void) printf("\t\tfsid_guid = %llu\n",
	(u_longlong_t)ds->ds_fsid_guid);
	(void) printf("\t\tguid = %llu\n",
	(u_longlong_t)ds->ds_guid);
	(void) printf("\t\tflags = %llx\n",
	(u_longlong_t)ds->ds_flags);
	(void) printf("\t\tnext_clones_obj = %llu\n",
	(u_longlong_t)ds->ds_next_clones_obj);
	(void) printf("\t\tprops_obj = %llu\n",
	(u_longlong_t)ds->ds_props_obj);
	(void) printf("\t\tbp = %s\n", blkbuf);
	}

	static int
	dump_bptree_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	(void) arg, (void) tx;
	char blkbuf[BP_SPRINTF_LEN];

	if (bp->blk_birth != 0) {
	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	(void) printf("\t%s\n", blkbuf);
	}
	return (0);
	}

	static void
	dump_bptree(objset_t os, uint64_t obj, const char name)
	{
	char bytes[32];
	bptree_phys_t *bt;
	dmu_buf_t *db;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (bytes) >= NN_NUMBUF_SZ, "bytes truncated");

	if (dump_opt['d'] < 3)
	return;

	VERIFY3U(0, ==, dmu_bonus_hold(os, obj, FTAG, &db));
	bt = db->db_data;
	zdb_nicenum(bt->bt_bytes, bytes, sizeof (bytes));
	(void) printf("\n %s: %llu datasets, %s\n",
	name, (unsigned long long)(bt->bt_end - bt->bt_begin), bytes);
	dmu_buf_rele(db, FTAG);

	if (dump_opt['d'] < 5)
	return;

	(void) printf("\n");

	(void) bptree_iterate(os, obj, B_FALSE, dump_bptree_cb, NULL, NULL);
	}

	static int
	dump_bpobj_cb(void arg, const blkptr_t bp, boolean_t bp_freed, dmu_tx_t *tx)
	{
	(void) arg, (void) tx;
	char blkbuf[BP_SPRINTF_LEN];

	ASSERT(bp->blk_birth != 0);
	snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp, bp_freed);
	(void) printf("\t%s\n", blkbuf);
	return (0);
	}

	static void
	dump_full_bpobj(bpobj_t bpo, const char name, int indent)
	{
	char bytes[32];
	char comp[32];
	char uncomp[32];
	uint64_t i;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (bytes) >= NN_NUMBUF_SZ, "bytes truncated");
	_Static_assert(sizeof (comp) >= NN_NUMBUF_SZ, "comp truncated");
	_Static_assert(sizeof (uncomp) >= NN_NUMBUF_SZ, "uncomp truncated");

	if (dump_opt['d'] < 3)
	return;

	zdb_nicenum(bpo->bpo_phys->bpo_bytes, bytes, sizeof (bytes));
	if (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_subobjs != 0) {
	zdb_nicenum(bpo->bpo_phys->bpo_comp, comp, sizeof (comp));
	zdb_nicenum(bpo->bpo_phys->bpo_uncomp, uncomp, sizeof (uncomp));
	if (bpo->bpo_havefreed) {
	(void) printf(" %*s: object %llu, %llu local "
	"blkptrs, %llu freed, %llu subobjs in object %llu, "
	"%s (%s/%s comp)\n",
	indent * 8, name,
	(u_longlong_t)bpo->bpo_object,
	(u_longlong_t)bpo->bpo_phys->bpo_num_blkptrs,
	(u_longlong_t)bpo->bpo_phys->bpo_num_freed,
	(u_longlong_t)bpo->bpo_phys->bpo_num_subobjs,
	(u_longlong_t)bpo->bpo_phys->bpo_subobjs,
	bytes, comp, uncomp);
	} else {
	(void) printf(" %*s: object %llu, %llu local "
	"blkptrs, %llu subobjs in object %llu, "
	"%s (%s/%s comp)\n",
	indent * 8, name,
	(u_longlong_t)bpo->bpo_object,
	(u_longlong_t)bpo->bpo_phys->bpo_num_blkptrs,
	(u_longlong_t)bpo->bpo_phys->bpo_num_subobjs,
	(u_longlong_t)bpo->bpo_phys->bpo_subobjs,
	bytes, comp, uncomp);
	}

	for (i = 0; i < bpo->bpo_phys->bpo_num_subobjs; i++) {
	uint64_t subobj;
	bpobj_t subbpo;
	int error;
	VERIFY0(dmu_read(bpo->bpo_os,
	bpo->bpo_phys->bpo_subobjs,
	i * sizeof (subobj), sizeof (subobj), &subobj, 0));
	error = bpobj_open(&subbpo, bpo->bpo_os, subobj);
	if (error != 0) {
	(void) printf("ERROR %u while trying to open "
	"subobj id %llu\n",
	error, (u_longlong_t)subobj);
	continue;
	}
	dump_full_bpobj(&subbpo, "subobj", indent + 1);
	bpobj_close(&subbpo);
	}
	} else {
	if (bpo->bpo_havefreed) {
	(void) printf(" %*s: object %llu, %llu blkptrs, "
	"%llu freed, %s\n",
	indent * 8, name,
	(u_longlong_t)bpo->bpo_object,
	(u_longlong_t)bpo->bpo_phys->bpo_num_blkptrs,
	(u_longlong_t)bpo->bpo_phys->bpo_num_freed,
	bytes);
	} else {
	(void) printf(" %*s: object %llu, %llu blkptrs, "
	"%s\n",
	indent * 8, name,
	(u_longlong_t)bpo->bpo_object,
	(u_longlong_t)bpo->bpo_phys->bpo_num_blkptrs,
	bytes);
	}
	}

	if (dump_opt['d'] < 5)
	return;


	if (indent == 0) {
	(void) bpobj_iterate_nofree(bpo, dump_bpobj_cb, NULL, NULL);
	(void) printf("\n");
	}
	}

	static int
	dump_bookmark(dsl_pool_t dp, char name, boolean_t print_redact,
	boolean_t print_list)
	{
	int err = 0;
	zfs_bookmark_phys_t prop;
	objset_t *mos = dp->dp_spa->spa_meta_objset;
	err = dsl_bookmark_lookup(dp, name, NULL, &prop);

	if (err != 0) {
	return (err);
	}

	(void) printf("\t#%s: ", strchr(name, '#') + 1);
	(void) printf("{guid: %llx creation_txg: %llu creation_time: "
	"%llu redaction_obj: %llu}\n", (u_longlong_t)prop.zbm_guid,
	(u_longlong_t)prop.zbm_creation_txg,
	(u_longlong_t)prop.zbm_creation_time,
	(u_longlong_t)prop.zbm_redaction_obj);

	IMPLY(print_list, print_redact);
	if (!print_redact \|\| prop.zbm_redaction_obj == 0)
	return (0);

	redaction_list_t *rl;
	VERIFY0(dsl_redaction_list_hold_obj(dp,
	prop.zbm_redaction_obj, FTAG, &rl));

	redaction_list_phys_t *rlp = rl->rl_phys;
	(void) printf("\tRedacted:\n\t\tProgress: ");
	if (rlp->rlp_last_object != UINT64_MAX \|\|
	rlp->rlp_last_blkid != UINT64_MAX) {
	(void) printf("%llu %llu (incomplete)\n",
	(u_longlong_t)rlp->rlp_last_object,
	(u_longlong_t)rlp->rlp_last_blkid);
	} else {
	(void) printf("complete\n");
	}
	(void) printf("\t\tSnapshots: [");
	for (unsigned int i = 0; i < rlp->rlp_num_snaps; i++) {
	if (i > 0)
	(void) printf(", ");
	(void) printf("%0llu",
	(u_longlong_t)rlp->rlp_snaps[i]);
	}
	(void) printf("]\n\t\tLength: %llu\n",
	(u_longlong_t)rlp->rlp_num_entries);

	if (!print_list) {
	dsl_redaction_list_rele(rl, FTAG);
	return (0);
	}

	if (rlp->rlp_num_entries == 0) {
	dsl_redaction_list_rele(rl, FTAG);
	(void) printf("\t\tRedaction List: []\n\n");
	return (0);
	}

	redact_block_phys_t *rbp_buf;
	uint64_t size;
	dmu_object_info_t doi;

	VERIFY0(dmu_object_info(mos, prop.zbm_redaction_obj, &doi));
	size = doi.doi_max_offset;
	rbp_buf = kmem_alloc(size, KM_SLEEP);

	err = dmu_read(mos, prop.zbm_redaction_obj, 0, size,
	rbp_buf, 0);
	if (err != 0) {
	dsl_redaction_list_rele(rl, FTAG);
	kmem_free(rbp_buf, size);
	return (err);
	}

	(void) printf("\t\tRedaction List: [{object: %llx, offset: "
	"%llx, blksz: %x, count: %llx}",
	(u_longlong_t)rbp_buf[0].rbp_object,
	(u_longlong_t)rbp_buf[0].rbp_blkid,
	(uint_t)(redact_block_get_size(&rbp_buf[0])),
	(u_longlong_t)redact_block_get_count(&rbp_buf[0]));

	for (size_t i = 1; i < rlp->rlp_num_entries; i++) {
	(void) printf(",\n\t\t{object: %llx, offset: %llx, "
	"blksz: %x, count: %llx}",
	(u_longlong_t)rbp_buf[i].rbp_object,
	(u_longlong_t)rbp_buf[i].rbp_blkid,
	(uint_t)(redact_block_get_size(&rbp_buf[i])),
	(u_longlong_t)redact_block_get_count(&rbp_buf[i]));
	}
	dsl_redaction_list_rele(rl, FTAG);
	kmem_free(rbp_buf, size);
	(void) printf("]\n\n");
	return (0);
	}

	static void
	dump_bookmarks(objset_t *os, int verbosity)
	{
	zap_cursor_t zc;
	zap_attribute_t attr;
	dsl_dataset_t *ds = dmu_objset_ds(os);
	dsl_pool_t *dp = spa_get_dsl(os->os_spa);
	objset_t *mos = os->os_spa->spa_meta_objset;
	if (verbosity < 4)
	return;
	dsl_pool_config_enter(dp, FTAG);

	for (zap_cursor_init(&zc, mos, ds->ds_bookmarks_obj);
	zap_cursor_retrieve(&zc, &attr) == 0;
	zap_cursor_advance(&zc)) {
	char osname[ZFS_MAX_DATASET_NAME_LEN];
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	dmu_objset_name(os, osname);
	VERIFY3S(0, <=, snprintf(buf, sizeof (buf), "%s#%s", osname,
	attr.za_name));
	(void) dump_bookmark(dp, buf, verbosity >= 5, verbosity >= 6);
	}
	zap_cursor_fini(&zc);
	dsl_pool_config_exit(dp, FTAG);
	}

	static void
	bpobj_count_refd(bpobj_t *bpo)
	{
	mos_obj_refd(bpo->bpo_object);

	if (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_subobjs != 0) {
	mos_obj_refd(bpo->bpo_phys->bpo_subobjs);
	for (uint64_t i = 0; i < bpo->bpo_phys->bpo_num_subobjs; i++) {
	uint64_t subobj;
	bpobj_t subbpo;
	int error;
	VERIFY0(dmu_read(bpo->bpo_os,
	bpo->bpo_phys->bpo_subobjs,
	i * sizeof (subobj), sizeof (subobj), &subobj, 0));
	error = bpobj_open(&subbpo, bpo->bpo_os, subobj);
	if (error != 0) {
	(void) printf("ERROR %u while trying to open "
	"subobj id %llu\n",
	error, (u_longlong_t)subobj);
	continue;
	}
	bpobj_count_refd(&subbpo);
	bpobj_close(&subbpo);
	}
	}
	}

	static int
	dsl_deadlist_entry_count_refd(void arg, dsl_deadlist_entry_t dle)
	{
	spa_t *spa = arg;
	uint64_t empty_bpobj = spa->spa_dsl_pool->dp_empty_bpobj;
	if (dle->dle_bpobj.bpo_object != empty_bpobj)
	bpobj_count_refd(&dle->dle_bpobj);
	return (0);
	}

	static int
	dsl_deadlist_entry_dump(void arg, dsl_deadlist_entry_t dle)
	{
	ASSERT(arg == NULL);
	if (dump_opt['d'] >= 5) {
	char buf[128];
	(void) snprintf(buf, sizeof (buf),
	"mintxg %llu -> obj %llu",
	(longlong_t)dle->dle_mintxg,
	(longlong_t)dle->dle_bpobj.bpo_object);

	dump_full_bpobj(&dle->dle_bpobj, buf, 0);
	} else {
	(void) printf("mintxg %llu -> obj %llu\n",
	(longlong_t)dle->dle_mintxg,
	(longlong_t)dle->dle_bpobj.bpo_object);
	}
	return (0);
	}

	static void
	dump_blkptr_list(dsl_deadlist_t dl, const char name)
	{
	char bytes[32];
	char comp[32];
	char uncomp[32];
	char entries[32];
	spa_t *spa = dmu_objset_spa(dl->dl_os);
	uint64_t empty_bpobj = spa->spa_dsl_pool->dp_empty_bpobj;

	if (dl->dl_oldfmt) {
	if (dl->dl_bpobj.bpo_object != empty_bpobj)
	bpobj_count_refd(&dl->dl_bpobj);
	} else {
	mos_obj_refd(dl->dl_object);
	dsl_deadlist_iterate(dl, dsl_deadlist_entry_count_refd, spa);
	}

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (bytes) >= NN_NUMBUF_SZ, "bytes truncated");
	_Static_assert(sizeof (comp) >= NN_NUMBUF_SZ, "comp truncated");
	_Static_assert(sizeof (uncomp) >= NN_NUMBUF_SZ, "uncomp truncated");
	_Static_assert(sizeof (entries) >= NN_NUMBUF_SZ, "entries truncated");

	if (dump_opt['d'] < 3)
	return;

	if (dl->dl_oldfmt) {
	dump_full_bpobj(&dl->dl_bpobj, "old-format deadlist", 0);
	return;
	}

	zdb_nicenum(dl->dl_phys->dl_used, bytes, sizeof (bytes));
	zdb_nicenum(dl->dl_phys->dl_comp, comp, sizeof (comp));
	zdb_nicenum(dl->dl_phys->dl_uncomp, uncomp, sizeof (uncomp));
	zdb_nicenum(avl_numnodes(&dl->dl_tree), entries, sizeof (entries));
	(void) printf("\n %s: %s (%s/%s comp), %s entries\n",
	name, bytes, comp, uncomp, entries);

	if (dump_opt['d'] < 4)
	return;

	(void) putchar('\n');

	dsl_deadlist_iterate(dl, dsl_deadlist_entry_dump, NULL);
	}

	static int
	verify_dd_livelist(objset_t *os)
	{
	uint64_t ll_used, used, ll_comp, comp, ll_uncomp, uncomp;
	dsl_pool_t *dp = spa_get_dsl(os->os_spa);
	dsl_dir_t *dd = os->os_dsl_dataset->ds_dir;

	ASSERT(!dmu_objset_is_snapshot(os));
	if (!dsl_deadlist_is_open(&dd->dd_livelist))
	return (0);

	/* Iterate through the livelist to check for duplicates */
	dsl_deadlist_iterate(&dd->dd_livelist, sublivelist_verify_lightweight,
	NULL);

	dsl_pool_config_enter(dp, FTAG);
	dsl_deadlist_space(&dd->dd_livelist, &ll_used,
	&ll_comp, &ll_uncomp);

	dsl_dataset_t *origin_ds;
	ASSERT(dsl_pool_config_held(dp));
	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin_ds));
	VERIFY0(dsl_dataset_space_written(origin_ds, os->os_dsl_dataset,
	&used, &comp, &uncomp));
	dsl_dataset_rele(origin_ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	/*
	* It's possible that the dataset's uncomp space is larger than the
	* livelist's because livelists do not track embedded block pointers
	*/
	if (used != ll_used \|\| comp != ll_comp \|\| uncomp < ll_uncomp) {
	char nice_used[32], nice_comp[32], nice_uncomp[32];
	(void) printf("Discrepancy in space accounting:\n");
	zdb_nicenum(used, nice_used, sizeof (nice_used));
	zdb_nicenum(comp, nice_comp, sizeof (nice_comp));
	zdb_nicenum(uncomp, nice_uncomp, sizeof (nice_uncomp));
	(void) printf("dir: used %s, comp %s, uncomp %s\n",
	nice_used, nice_comp, nice_uncomp);
	zdb_nicenum(ll_used, nice_used, sizeof (nice_used));
	zdb_nicenum(ll_comp, nice_comp, sizeof (nice_comp));
	zdb_nicenum(ll_uncomp, nice_uncomp, sizeof (nice_uncomp));
	(void) printf("livelist: used %s, comp %s, uncomp %s\n",
	nice_used, nice_comp, nice_uncomp);
	return (1);
	}
	return (0);
	}

	static avl_tree_t idx_tree;
	static avl_tree_t domain_tree;
	static boolean_t fuid_table_loaded;
	static objset_t *sa_os = NULL;
	static sa_attr_type_t *sa_attr_table = NULL;

	static int
	open_objset(const char path, const void tag, objset_t **osp)
	{
	int err;
	uint64_t sa_attrs = 0;
	uint64_t version = 0;

	VERIFY3P(sa_os, ==, NULL);
	/*
	* We can't own an objset if it's redacted. Therefore, we do this
	* dance: hold the objset, then acquire a long hold on its dataset, then
	* release the pool (which is held as part of holding the objset).
	*/
	err = dmu_objset_hold(path, tag, osp);
	if (err != 0) {
	(void) fprintf(stderr, "failed to hold dataset '%s': %s\n",
	path, strerror(err));
	return (err);
	}
	dsl_dataset_long_hold(dmu_objset_ds(*osp), tag);
	dsl_pool_rele(dmu_objset_pool(*osp), tag);

	if (dmu_objset_type(osp) == DMU_OST_ZFS && !(osp)->os_encrypted) {
	(void) zap_lookup(*osp, MASTER_NODE_OBJ, ZPL_VERSION_STR,
	8, 1, &version);
	if (version >= ZPL_VERSION_SA) {
	(void) zap_lookup(*osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS,
	8, 1, &sa_attrs);
	}
	err = sa_setup(*osp, sa_attrs, zfs_attr_table, ZPL_END,
	&sa_attr_table);
	if (err != 0) {
	(void) fprintf(stderr, "sa_setup failed: %s\n",
	strerror(err));
	dsl_dataset_long_rele(dmu_objset_ds(*osp), tag);
	dsl_dataset_rele(dmu_objset_ds(*osp), tag);
	*osp = NULL;
	}
	}
	sa_os = *osp;

	return (0);
	}

	static void
	close_objset(objset_t os, const void tag)
	{
	VERIFY3P(os, ==, sa_os);
	if (os->os_sa != NULL)
	sa_tear_down(os);
	dsl_dataset_long_rele(dmu_objset_ds(os), tag);
	dsl_dataset_rele(dmu_objset_ds(os), tag);
	sa_attr_table = NULL;
	sa_os = NULL;
	}

	static void
	fuid_table_destroy(void)
	{
	if (fuid_table_loaded) {
	zfs_fuid_table_destroy(&idx_tree, &domain_tree);
	fuid_table_loaded = B_FALSE;
	}
	}

	/*
	* print uid or gid information.
	* For normal POSIX id just the id is printed in decimal format.
	* For CIFS files with FUID the fuid is printed in hex followed by
	* the domain-rid string.
	*/
	static void
	print_idstr(uint64_t id, const char *id_type)
	{
	if (FUID_INDEX(id)) {
	const char *domain =
	zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id));
	(void) printf("\t%s %llx [%s-%d]\n", id_type,
	(u_longlong_t)id, domain, (int)FUID_RID(id));
	} else {
	(void) printf("\t%s %llu\n", id_type, (u_longlong_t)id);
	}

	}

	static void
	dump_uidgid(objset_t *os, uint64_t uid, uint64_t gid)
	{
	uint32_t uid_idx, gid_idx;

	uid_idx = FUID_INDEX(uid);
	gid_idx = FUID_INDEX(gid);

	/* Load domain table, if not already loaded */
	if (!fuid_table_loaded && (uid_idx \|\| gid_idx)) {
	uint64_t fuid_obj;

	/* first find the fuid object. It lives in the master node */
	VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES,
	8, 1, &fuid_obj) == 0);
	zfs_fuid_avl_tree_create(&idx_tree, &domain_tree);
	(void) zfs_fuid_table_load(os, fuid_obj,
	&idx_tree, &domain_tree);
	fuid_table_loaded = B_TRUE;
	}

	print_idstr(uid, "uid");
	print_idstr(gid, "gid");
	}

	static void
	dump_znode_sa_xattr(sa_handle_t *hdl)
	{
	nvlist_t *sa_xattr;
	nvpair_t *elem = NULL;
	int sa_xattr_size = 0;
	int sa_xattr_entries = 0;
	int error;
	char *sa_xattr_packed;

	error = sa_size(hdl, sa_attr_table[ZPL_DXATTR], &sa_xattr_size);
	if (error \|\| sa_xattr_size == 0)
	return;

	sa_xattr_packed = malloc(sa_xattr_size);
	if (sa_xattr_packed == NULL)
	return;

	error = sa_lookup(hdl, sa_attr_table[ZPL_DXATTR],
	sa_xattr_packed, sa_xattr_size);
	if (error) {
	free(sa_xattr_packed);
	return;
	}

	error = nvlist_unpack(sa_xattr_packed, sa_xattr_size, &sa_xattr, 0);
	if (error) {
	free(sa_xattr_packed);
	return;
	}

	while ((elem = nvlist_next_nvpair(sa_xattr, elem)) != NULL)
	sa_xattr_entries++;

	(void) printf("\tSA xattrs: %d bytes, %d entries\n\n",
	sa_xattr_size, sa_xattr_entries);
	while ((elem = nvlist_next_nvpair(sa_xattr, elem)) != NULL) {
	uchar_t *value;
	uint_t cnt, idx;

	(void) printf("\t\t%s = ", nvpair_name(elem));
	nvpair_value_byte_array(elem, &value, &cnt);
	for (idx = 0; idx < cnt; ++idx) {
	if (isprint(value[idx]))
	(void) putchar(value[idx]);
	else
	(void) printf("\\%3.3o", value[idx]);
	}
	(void) putchar('\n');
	}

	nvlist_free(sa_xattr);
	free(sa_xattr_packed);
	}

	static void
	dump_znode_symlink(sa_handle_t *hdl)
	{
	int sa_symlink_size = 0;
	char linktarget[MAXPATHLEN];
	int error;

	error = sa_size(hdl, sa_attr_table[ZPL_SYMLINK], &sa_symlink_size);
	if (error \|\| sa_symlink_size == 0) {
	return;
	}
	if (sa_symlink_size >= sizeof (linktarget)) {
	(void) printf("symlink size %d is too large\n",
	sa_symlink_size);
	return;
	}
	linktarget[sa_symlink_size] = '\0';
	if (sa_lookup(hdl, sa_attr_table[ZPL_SYMLINK],
	&linktarget, sa_symlink_size) == 0)
	(void) printf("\ttarget %s\n", linktarget);
	}

	static void
	dump_znode(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) data, (void) size;
	char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */
	sa_handle_t *hdl;
	uint64_t xattr, rdev, gen;
	uint64_t uid, gid, mode, fsize, parent, links;
	uint64_t pflags;
	uint64_t acctm[2], modtm[2], chgtm[2], crtm[2];
	time_t z_crtime, z_atime, z_mtime, z_ctime;
	sa_bulk_attr_t bulk[12];
	int idx = 0;
	int error;

	VERIFY3P(os, ==, sa_os);
	if (sa_handle_get(os, object, NULL, SA_HDL_PRIVATE, &hdl)) {
	(void) printf("Failed to get handle for SA znode\n");
	return;
	}

	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_UID], NULL, &uid, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_GID], NULL, &gid, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_LINKS], NULL,
	&links, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_GEN], NULL, &gen, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_MODE], NULL,
	&mode, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_PARENT],
	NULL, &parent, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_SIZE], NULL,
	&fsize, 8);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_ATIME], NULL,
	acctm, 16);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_MTIME], NULL,
	modtm, 16);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_CRTIME], NULL,
	crtm, 16);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_CTIME], NULL,
	chgtm, 16);
	SA_ADD_BULK_ATTR(bulk, idx, sa_attr_table[ZPL_FLAGS], NULL,
	&pflags, 8);

	if (sa_bulk_lookup(hdl, bulk, idx)) {
	(void) sa_handle_destroy(hdl);
	return;
	}

	z_crtime = (time_t)crtm[0];
	z_atime = (time_t)acctm[0];
	z_mtime = (time_t)modtm[0];
	z_ctime = (time_t)chgtm[0];

	if (dump_opt['d'] > 4) {
	error = zfs_obj_to_path(os, object, path, sizeof (path));
	if (error == ESTALE) {
	(void) snprintf(path, sizeof (path), "on delete queue");
	} else if (error != 0) {
	leaked_objects++;
	(void) snprintf(path, sizeof (path),
	"path not found, possibly leaked");
	}
	(void) printf("\tpath %s\n", path);
	}

	if (S_ISLNK(mode))
	dump_znode_symlink(hdl);
	dump_uidgid(os, uid, gid);
	(void) printf("\tatime %s", ctime(&z_atime));
	(void) printf("\tmtime %s", ctime(&z_mtime));
	(void) printf("\tctime %s", ctime(&z_ctime));
	(void) printf("\tcrtime %s", ctime(&z_crtime));
	(void) printf("\tgen %llu\n", (u_longlong_t)gen);
	(void) printf("\tmode %llo\n", (u_longlong_t)mode);
	(void) printf("\tsize %llu\n", (u_longlong_t)fsize);
	(void) printf("\tparent %llu\n", (u_longlong_t)parent);
	(void) printf("\tlinks %llu\n", (u_longlong_t)links);
	(void) printf("\tpflags %llx\n", (u_longlong_t)pflags);
	if (dmu_objset_projectquota_enabled(os) && (pflags & ZFS_PROJID)) {
	uint64_t projid;

	if (sa_lookup(hdl, sa_attr_table[ZPL_PROJID], &projid,
	sizeof (uint64_t)) == 0)
	(void) printf("\tprojid %llu\n", (u_longlong_t)projid);
	}
	if (sa_lookup(hdl, sa_attr_table[ZPL_XATTR], &xattr,
	sizeof (uint64_t)) == 0)
	(void) printf("\txattr %llu\n", (u_longlong_t)xattr);
	if (sa_lookup(hdl, sa_attr_table[ZPL_RDEV], &rdev,
	sizeof (uint64_t)) == 0)
	(void) printf("\trdev 0x%016llx\n", (u_longlong_t)rdev);
	dump_znode_sa_xattr(hdl);
	sa_handle_destroy(hdl);
	}

	static void
	dump_acl(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	}

	static void
	dump_dmu_objset(objset_t os, uint64_t object, void data, size_t size)
	{
	(void) os, (void) object, (void) data, (void) size;
	}

	static object_viewer_t *object_viewer[DMU_OT_NUMTYPES + 1] = {
	dump_none, /* unallocated */
	dump_zap, /* object directory */
	dump_uint64, /* object array */
	dump_none, /* packed nvlist */
	dump_packed_nvlist, /* packed nvlist size */
	dump_none, /* bpobj */
	dump_bpobj, /* bpobj header */
	dump_none, /* SPA space map header */
	dump_none, /* SPA space map */
	dump_none, /* ZIL intent log */
	dump_dnode, /* DMU dnode */
	dump_dmu_objset, /* DMU objset */
	dump_dsl_dir, /* DSL directory */
	dump_zap, /* DSL directory child map */
	dump_zap, /* DSL dataset snap map */
	dump_zap, /* DSL props */
	dump_dsl_dataset, /* DSL dataset */
	dump_znode, /* ZFS znode */
	dump_acl, /* ZFS V0 ACL */
	dump_uint8, /* ZFS plain file */
	dump_zpldir, /* ZFS directory */
	dump_zap, /* ZFS master node */
	dump_zap, /* ZFS delete queue */
	dump_uint8, /* zvol object */
	dump_zap, /* zvol prop */
	dump_uint8, /* other uint8[] */
	dump_uint64, /* other uint64[] */
	dump_zap, /* other ZAP */
	dump_zap, /* persistent error log */
	dump_uint8, /* SPA history */
	dump_history_offsets, /* SPA history offsets */
	dump_zap, /* Pool properties */
	dump_zap, /* DSL permissions */
	dump_acl, /* ZFS ACL */
	dump_uint8, /* ZFS SYSACL */
	dump_none, /* FUID nvlist */
	dump_packed_nvlist, /* FUID nvlist size */
	dump_zap, /* DSL dataset next clones */
	dump_zap, /* DSL scrub queue */
	dump_zap, /* ZFS user/group/project used */
	dump_zap, /* ZFS user/group/project quota */
	dump_zap, /* snapshot refcount tags */
	dump_ddt_zap, /* DDT ZAP object */
	dump_zap, /* DDT statistics */
	dump_znode, /* SA object */
	dump_zap, /* SA Master Node */
	dump_sa_attrs, /* SA attribute registration */
	dump_sa_layouts, /* SA attribute layouts */
	dump_zap, /* DSL scrub translations */
	dump_none, /* fake dedup BP */
	dump_zap, /* deadlist */
	dump_none, /* deadlist hdr */
	dump_zap, /* dsl clones */
	dump_bpobj_subobjs, /* bpobj subobjs */
	dump_unknown, /* Unknown type, must be last */
	};

	static boolean_t
	match_object_type(dmu_object_type_t obj_type, uint64_t flags)
	{
	boolean_t match = B_TRUE;

	switch (obj_type) {
	case DMU_OT_DIRECTORY_CONTENTS:
	if (!(flags & ZOR_FLAG_DIRECTORY))
	match = B_FALSE;
	break;
	case DMU_OT_PLAIN_FILE_CONTENTS:
	if (!(flags & ZOR_FLAG_PLAIN_FILE))
	match = B_FALSE;
	break;
	case DMU_OT_SPACE_MAP:
	if (!(flags & ZOR_FLAG_SPACE_MAP))
	match = B_FALSE;
	break;
	default:
	if (strcmp(zdb_ot_name(obj_type), "zap") == 0) {
	if (!(flags & ZOR_FLAG_ZAP))
	match = B_FALSE;
	break;
	}

	/*
	* If all bits except some of the supported flags are
	* set, the user combined the all-types flag (A) with
	* a negated flag to exclude some types (e.g. A-f to
	* show all object types except plain files).
	*/
	if ((flags \| ZOR_SUPPORTED_FLAGS) != ZOR_FLAG_ALL_TYPES)
	match = B_FALSE;

	break;
	}

	return (match);
	}

	static void
	dump_object(objset_t *os, uint64_t object, int verbosity,
	boolean_t print_header, uint64_t dnode_slots_used, uint64_t flags)
	{
	dmu_buf_t *db = NULL;
	dmu_object_info_t doi;
	dnode_t *dn;
	boolean_t dnode_held = B_FALSE;
	void *bonus = NULL;
	size_t bsize = 0;
	char iblk[32], dblk[32], lsize[32], asize[32], fill[32], dnsize[32];
	char bonus_size[32];
	char aux[50];
	int error;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (iblk) >= NN_NUMBUF_SZ, "iblk truncated");
	_Static_assert(sizeof (dblk) >= NN_NUMBUF_SZ, "dblk truncated");
	_Static_assert(sizeof (lsize) >= NN_NUMBUF_SZ, "lsize truncated");
	_Static_assert(sizeof (asize) >= NN_NUMBUF_SZ, "asize truncated");
	_Static_assert(sizeof (bonus_size) >= NN_NUMBUF_SZ,
	"bonus_size truncated");

	if (*print_header) {
	(void) printf("\n%10s %3s %5s %5s %5s %6s %5s %6s %s\n",
	"Object", "lvl", "iblk", "dblk", "dsize", "dnsize",
	"lsize", "%full", "type");
	*print_header = 0;
	}

	if (object == 0) {
	dn = DMU_META_DNODE(os);
	dmu_object_info_from_dnode(dn, &doi);
	} else {
	/*
	* Encrypted datasets will have sensitive bonus buffers
	* encrypted. Therefore we cannot hold the bonus buffer and
	* must hold the dnode itself instead.
	*/
	error = dmu_object_info(os, object, &doi);
	if (error)
	fatal("dmu_object_info() failed, errno %u", error);

	if (os->os_encrypted &&
	DMU_OT_IS_ENCRYPTED(doi.doi_bonus_type)) {
	error = dnode_hold(os, object, FTAG, &dn);
	if (error)
	fatal("dnode_hold() failed, errno %u", error);
	dnode_held = B_TRUE;
	} else {
	error = dmu_bonus_hold(os, object, FTAG, &db);
	if (error)
	fatal("dmu_bonus_hold(%llu) failed, errno %u",
	object, error);
	bonus = db->db_data;
	bsize = db->db_size;
	dn = DB_DNODE((dmu_buf_impl_t *)db);
	}
	}

	/*
	* Default to showing all object types if no flags were specified.
	*/
	if (flags != 0 && flags != ZOR_FLAG_ALL_TYPES &&
	!match_object_type(doi.doi_type, flags))
	goto out;

	if (dnode_slots_used)
	*dnode_slots_used = doi.doi_dnodesize / DNODE_MIN_SIZE;

	zdb_nicenum(doi.doi_metadata_block_size, iblk, sizeof (iblk));
	zdb_nicenum(doi.doi_data_block_size, dblk, sizeof (dblk));
	zdb_nicenum(doi.doi_max_offset, lsize, sizeof (lsize));
	zdb_nicenum(doi.doi_physical_blocks_512 << 9, asize, sizeof (asize));
	zdb_nicenum(doi.doi_bonus_size, bonus_size, sizeof (bonus_size));
	zdb_nicenum(doi.doi_dnodesize, dnsize, sizeof (dnsize));
	(void) sprintf(fill, "%6.2f", 100.0 * doi.doi_fill_count *
	doi.doi_data_block_size / (object == 0 ? DNODES_PER_BLOCK : 1) /
	doi.doi_max_offset);

	aux[0] = '\0';

	if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT \|\| verbosity >= 6) {
	(void) snprintf(aux + strlen(aux), sizeof (aux) - strlen(aux),
	" (K=%s)", ZDB_CHECKSUM_NAME(doi.doi_checksum));
	}

	if (doi.doi_compress == ZIO_COMPRESS_INHERIT &&
	ZIO_COMPRESS_HASLEVEL(os->os_compress) && verbosity >= 6) {
	const char *compname = NULL;
	if (zfs_prop_index_to_string(ZFS_PROP_COMPRESSION,
	ZIO_COMPRESS_RAW(os->os_compress, os->os_complevel),
	&compname) == 0) {
	(void) snprintf(aux + strlen(aux),
	sizeof (aux) - strlen(aux), " (Z=inherit=%s)",
	compname);
	} else {
	(void) snprintf(aux + strlen(aux),
	sizeof (aux) - strlen(aux),
	" (Z=inherit=%s-unknown)",
	ZDB_COMPRESS_NAME(os->os_compress));
	}
	} else if (doi.doi_compress == ZIO_COMPRESS_INHERIT && verbosity >= 6) {
	(void) snprintf(aux + strlen(aux), sizeof (aux) - strlen(aux),
	" (Z=inherit=%s)", ZDB_COMPRESS_NAME(os->os_compress));
	} else if (doi.doi_compress != ZIO_COMPRESS_INHERIT \|\| verbosity >= 6) {
	(void) snprintf(aux + strlen(aux), sizeof (aux) - strlen(aux),
	" (Z=%s)", ZDB_COMPRESS_NAME(doi.doi_compress));
	}

	(void) printf("%10lld %3u %5s %5s %5s %6s %5s %6s %s%s\n",
	(u_longlong_t)object, doi.doi_indirection, iblk, dblk,
	asize, dnsize, lsize, fill, zdb_ot_name(doi.doi_type), aux);

	if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) {
	(void) printf("%10s %3s %5s %5s %5s %5s %5s %6s %s\n",
	"", "", "", "", "", "", bonus_size, "bonus",
	zdb_ot_name(doi.doi_bonus_type));
	}

	if (verbosity >= 4) {
	(void) printf("\tdnode flags: %s%s%s%s\n",
	(dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) ?
	"USED_BYTES " : "",
	(dn->dn_phys->dn_flags & DNODE_FLAG_USERUSED_ACCOUNTED) ?
	"USERUSED_ACCOUNTED " : "",
	(dn->dn_phys->dn_flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) ?
	"USEROBJUSED_ACCOUNTED " : "",
	(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) ?
	"SPILL_BLKPTR" : "");
	(void) printf("\tdnode maxblkid: %llu\n",
	(longlong_t)dn->dn_phys->dn_maxblkid);

	if (!dnode_held) {
	object_viewer[ZDB_OT_TYPE(doi.doi_bonus_type)](os,
	object, bonus, bsize);
	} else {
	(void) printf("\t\t(bonus encrypted)\n");
	}

	if (!os->os_encrypted \|\| !DMU_OT_IS_ENCRYPTED(doi.doi_type)) {
	object_viewer[ZDB_OT_TYPE(doi.doi_type)](os, object,
	NULL, 0);
	} else {
	(void) printf("\t\t(object encrypted)\n");
	}

	*print_header = B_TRUE;
	}

	if (verbosity >= 5) {
	if (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	char blkbuf[BP_SPRINTF_LEN];
	snprintf_blkptr_compact(blkbuf, sizeof (blkbuf),
	DN_SPILL_BLKPTR(dn->dn_phys), B_FALSE);
	(void) printf("\nSpill block: %s\n", blkbuf);
	}
	dump_indirect(dn);
	}

	if (verbosity >= 5) {
	/*
	* Report the list of segments that comprise the object.
	*/
	uint64_t start = 0;
	uint64_t end;
	uint64_t blkfill = 1;
	int minlvl = 1;

	if (dn->dn_type == DMU_OT_DNODE) {
	minlvl = 0;
	blkfill = DNODES_PER_BLOCK;
	}

	for (;;) {
	char segsize[32];
	/* make sure nicenum has enough space */
	_Static_assert(sizeof (segsize) >= NN_NUMBUF_SZ,
	"segsize truncated");
	error = dnode_next_offset(dn,
	0, &start, minlvl, blkfill, 0);
	if (error)
	break;
	end = start;
	error = dnode_next_offset(dn,
	DNODE_FIND_HOLE, &end, minlvl, blkfill, 0);
	zdb_nicenum(end - start, segsize, sizeof (segsize));
	(void) printf("\t\tsegment [%016llx, %016llx)"
	" size %5s\n", (u_longlong_t)start,
	(u_longlong_t)end, segsize);
	if (error)
	break;
	start = end;
	}
	}

	out:
	if (db != NULL)
	dmu_buf_rele(db, FTAG);
	if (dnode_held)
	dnode_rele(dn, FTAG);
	}

	static void
	count_dir_mos_objects(dsl_dir_t *dd)
	{
	mos_obj_refd(dd->dd_object);
	mos_obj_refd(dsl_dir_phys(dd)->dd_child_dir_zapobj);
	mos_obj_refd(dsl_dir_phys(dd)->dd_deleg_zapobj);
	mos_obj_refd(dsl_dir_phys(dd)->dd_props_zapobj);
	mos_obj_refd(dsl_dir_phys(dd)->dd_clones);

	/*
	* The dd_crypto_obj can be referenced by multiple dsl_dir's.
	* Ignore the references after the first one.
	*/
	mos_obj_refd_multiple(dd->dd_crypto_obj);
	}

	static void
	count_ds_mos_objects(dsl_dataset_t *ds)
	{
	mos_obj_refd(ds->ds_object);
	mos_obj_refd(dsl_dataset_phys(ds)->ds_next_clones_obj);
	mos_obj_refd(dsl_dataset_phys(ds)->ds_props_obj);
	mos_obj_refd(dsl_dataset_phys(ds)->ds_userrefs_obj);
	mos_obj_refd(dsl_dataset_phys(ds)->ds_snapnames_zapobj);
	mos_obj_refd(ds->ds_bookmarks_obj);

	if (!dsl_dataset_is_snapshot(ds)) {
	count_dir_mos_objects(ds->ds_dir);
	}
	}

	static const char *const objset_types[DMU_OST_NUMTYPES] = {
	"NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };

	/*
	* Parse a string denoting a range of object IDs of the form
	* <start>[:<end>[:flags]], and store the results in zor.
	* Return 0 on success. On error, return 1 and update the msg
	* pointer to point to a descriptive error message.
	*/
	static int
	parse_object_range(char range, zopt_object_range_t zor, const char **msg)
	{
	uint64_t flags = 0;
	char p, s, dup, flagstr, *tmp = NULL;
	size_t len;
	int i;
	int rc = 0;

	if (strchr(range, ':') == NULL) {
	zor->zor_obj_start = strtoull(range, &p, 0);
	if (*p != '\0') {
	*msg = "Invalid characters in object ID";
	rc = 1;
	}
	zor->zor_obj_start = ZDB_MAP_OBJECT_ID(zor->zor_obj_start);
	zor->zor_obj_end = zor->zor_obj_start;
	return (rc);
	}

	if (strchr(range, ':') == range) {
	*msg = "Invalid leading colon";
	rc = 1;
	return (rc);
	}

	len = strlen(range);
	if (range[len - 1] == ':') {
	*msg = "Invalid trailing colon";
	rc = 1;
	return (rc);
	}

	dup = strdup(range);
	s = strtok_r(dup, ":", &tmp);
	zor->zor_obj_start = strtoull(s, &p, 0);

	if (*p != '\0') {
	*msg = "Invalid characters in start object ID";
	rc = 1;
	goto out;
	}

	s = strtok_r(NULL, ":", &tmp);
	zor->zor_obj_end = strtoull(s, &p, 0);

	if (*p != '\0') {
	*msg = "Invalid characters in end object ID";
	rc = 1;
	goto out;
	}

	if (zor->zor_obj_start > zor->zor_obj_end) {
	*msg = "Start object ID may not exceed end object ID";
	rc = 1;
	goto out;
	}

	s = strtok_r(NULL, ":", &tmp);
	if (s == NULL) {
	zor->zor_flags = ZOR_FLAG_ALL_TYPES;
	goto out;
	} else if (strtok_r(NULL, ":", &tmp) != NULL) {
	*msg = "Invalid colon-delimited field after flags";
	rc = 1;
	goto out;
	}

	flagstr = s;
	for (i = 0; flagstr[i]; i++) {
	int bit;
	boolean_t negation = (flagstr[i] == '-');

	if (negation) {
	i++;
	if (flagstr[i] == '\0') {
	*msg = "Invalid trailing negation operator";
	rc = 1;
	goto out;
	}
	}
	bit = flagbits[(uchar_t)flagstr[i]];
	if (bit == 0) {
	*msg = "Invalid flag";
	rc = 1;
	goto out;
	}
	if (negation)
	flags &= ~bit;
	else
	flags \|= bit;
	}
	zor->zor_flags = flags;

	zor->zor_obj_start = ZDB_MAP_OBJECT_ID(zor->zor_obj_start);
	zor->zor_obj_end = ZDB_MAP_OBJECT_ID(zor->zor_obj_end);

	out:
	free(dup);
	return (rc);
	}

	static void
	dump_objset(objset_t *os)
	{
	dmu_objset_stats_t dds = { 0 };
	uint64_t object, object_count;
	uint64_t refdbytes, usedobjs, scratch;
	char numbuf[32];
	char blkbuf[BP_SPRINTF_LEN + 20];
	char osname[ZFS_MAX_DATASET_NAME_LEN];
	const char *type = "UNKNOWN";
	int verbosity = dump_opt['d'];
	boolean_t print_header;
	unsigned i;
	int error;
	uint64_t total_slots_used = 0;
	uint64_t max_slot_used = 0;
	uint64_t dnode_slots;
	uint64_t obj_start;
	uint64_t obj_end;
	uint64_t flags;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (numbuf) >= NN_NUMBUF_SZ, "numbuf truncated");

	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
	dmu_objset_fast_stat(os, &dds);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);

	print_header = B_TRUE;

	if (dds.dds_type < DMU_OST_NUMTYPES)
	type = objset_types[dds.dds_type];

	if (dds.dds_type == DMU_OST_META) {
	dds.dds_creation_txg = TXG_INITIAL;
	usedobjs = BP_GET_FILL(os->os_rootbp);
	refdbytes = dsl_dir_phys(os->os_spa->spa_dsl_pool->dp_mos_dir)->
	dd_used_bytes;
	} else {
	dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch);
	}

	ASSERT3U(usedobjs, ==, BP_GET_FILL(os->os_rootbp));

	zdb_nicenum(refdbytes, numbuf, sizeof (numbuf));

	if (verbosity >= 4) {
	(void) snprintf(blkbuf, sizeof (blkbuf), ", rootbp ");
	(void) snprintf_blkptr(blkbuf + strlen(blkbuf),
	sizeof (blkbuf) - strlen(blkbuf), os->os_rootbp);
	} else {
	blkbuf[0] = '\0';
	}

	dmu_objset_name(os, osname);

	(void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
	"%s, %llu objects%s%s\n",
	osname, type, (u_longlong_t)dmu_objset_id(os),
	(u_longlong_t)dds.dds_creation_txg,
	numbuf, (u_longlong_t)usedobjs, blkbuf,
	(dds.dds_inconsistent) ? " (inconsistent)" : "");

	for (i = 0; i < zopt_object_args; i++) {
	obj_start = zopt_object_ranges[i].zor_obj_start;
	obj_end = zopt_object_ranges[i].zor_obj_end;
	flags = zopt_object_ranges[i].zor_flags;

	object = obj_start;
	if (object == 0 \|\| obj_start == obj_end)
	dump_object(os, object, verbosity, &print_header, NULL,
	flags);
	else
	object--;

	while ((dmu_object_next(os, &object, B_FALSE, 0) == 0) &&
	object <= obj_end) {
	dump_object(os, object, verbosity, &print_header, NULL,
	flags);
	}
	}

	if (zopt_object_args > 0) {
	(void) printf("\n");
	return;
	}

	if (dump_opt['i'] != 0 \|\| verbosity >= 2)
	dump_intent_log(dmu_objset_zil(os));

	if (dmu_objset_ds(os) != NULL) {
	dsl_dataset_t *ds = dmu_objset_ds(os);
	dump_blkptr_list(&ds->ds_deadlist, "Deadlist");
	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
	!dmu_objset_is_snapshot(os)) {
	dump_blkptr_list(&ds->ds_dir->dd_livelist, "Livelist");
	if (verify_dd_livelist(os) != 0)
	fatal("livelist is incorrect");
	}

	if (dsl_dataset_remap_deadlist_exists(ds)) {
	(void) printf("ds_remap_deadlist:\n");
	dump_blkptr_list(&ds->ds_remap_deadlist, "Deadlist");
	}
	count_ds_mos_objects(ds);
	}

	if (dmu_objset_ds(os) != NULL)
	dump_bookmarks(os, verbosity);

	if (verbosity < 2)
	return;

	if (BP_IS_HOLE(os->os_rootbp))
	return;

	dump_object(os, 0, verbosity, &print_header, NULL, 0);
	object_count = 0;
	if (DMU_USERUSED_DNODE(os) != NULL &&
	DMU_USERUSED_DNODE(os)->dn_type != 0) {
	dump_object(os, DMU_USERUSED_OBJECT, verbosity, &print_header,
	NULL, 0);
	dump_object(os, DMU_GROUPUSED_OBJECT, verbosity, &print_header,
	NULL, 0);
	}

	if (DMU_PROJECTUSED_DNODE(os) != NULL &&
	DMU_PROJECTUSED_DNODE(os)->dn_type != 0)
	dump_object(os, DMU_PROJECTUSED_OBJECT, verbosity,
	&print_header, NULL, 0);

	object = 0;
	while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) {
	dump_object(os, object, verbosity, &print_header, &dnode_slots,
	0);
	object_count++;
	total_slots_used += dnode_slots;
	max_slot_used = object + dnode_slots - 1;
	}

	(void) printf("\n");

	(void) printf(" Dnode slots:\n");
	(void) printf("\tTotal used: %10llu\n",
	(u_longlong_t)total_slots_used);
	(void) printf("\tMax used: %10llu\n",
	(u_longlong_t)max_slot_used);
	(void) printf("\tPercent empty: %10lf\n",
	(double)(max_slot_used - total_slots_used)*100 /
	(double)max_slot_used);
	(void) printf("\n");

	if (error != ESRCH) {
	(void) fprintf(stderr, "dmu_object_next() = %d\n", error);
	abort();
	}

	ASSERT3U(object_count, ==, usedobjs);

	if (leaked_objects != 0) {
	(void) printf("%d potentially leaked objects detected\n",
	leaked_objects);
	leaked_objects = 0;
	}
	}

	static void
	dump_uberblock(uberblock_t ub, const char header, const char *footer)
	{
	time_t timestamp = ub->ub_timestamp;

	(void) printf("%s", header ? header : "");
	(void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic);
	(void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version);
	(void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg);
	(void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum);
	(void) printf("\ttimestamp = %llu UTC = %s",
	(u_longlong_t)ub->ub_timestamp, ctime(&timestamp));

	(void) printf("\tmmp_magic = %016llx\n",
	(u_longlong_t)ub->ub_mmp_magic);
	if (MMP_VALID(ub)) {
	(void) printf("\tmmp_delay = %0llu\n",
	(u_longlong_t)ub->ub_mmp_delay);
	if (MMP_SEQ_VALID(ub))
	(void) printf("\tmmp_seq = %u\n",
	(unsigned int) MMP_SEQ(ub));
	if (MMP_FAIL_INT_VALID(ub))
	(void) printf("\tmmp_fail = %u\n",
	(unsigned int) MMP_FAIL_INT(ub));
	if (MMP_INTERVAL_VALID(ub))
	(void) printf("\tmmp_write = %u\n",
	(unsigned int) MMP_INTERVAL(ub));
	/* After MMP_* to make summarize_uberblock_mmp cleaner */
	(void) printf("\tmmp_valid = %x\n",
	(unsigned int) ub->ub_mmp_config & 0xFF);
	}

	if (dump_opt['u'] >= 4) {
	char blkbuf[BP_SPRINTF_LEN];
	snprintf_blkptr(blkbuf, sizeof (blkbuf), &ub->ub_rootbp);
	(void) printf("\trootbp = %s\n", blkbuf);
	}
	(void) printf("\tcheckpoint_txg = %llu\n",
	(u_longlong_t)ub->ub_checkpoint_txg);
	(void) printf("%s", footer ? footer : "");
	}

	static void
	dump_config(spa_t *spa)
	{
	dmu_buf_t *db;
	size_t nvsize = 0;
	int error = 0;


	error = dmu_bonus_hold(spa->spa_meta_objset,
	spa->spa_config_object, FTAG, &db);

	if (error == 0) {
	nvsize = (uint64_t )db->db_data;
	dmu_buf_rele(db, FTAG);

	(void) printf("\nMOS Configuration:\n");
	dump_packed_nvlist(spa->spa_meta_objset,
	spa->spa_config_object, (void *)&nvsize, 1);
	} else {
	(void) fprintf(stderr, "dmu_bonus_hold(%llu) failed, errno %d",
	(u_longlong_t)spa->spa_config_object, error);
	}
	}

	static void
	dump_cachefile(const char *cachefile)
	{
	int fd;
	struct stat64 statbuf;
	char *buf;
	nvlist_t *config;

	if ((fd = open64(cachefile, O_RDONLY)) < 0) {
	(void) printf("cannot open '%s': %s\n", cachefile,
	strerror(errno));
	exit(1);
	}

	if (fstat64(fd, &statbuf) != 0) {
	(void) printf("failed to stat '%s': %s\n", cachefile,
	strerror(errno));
	exit(1);
	}

	if ((buf = malloc(statbuf.st_size)) == NULL) {
	(void) fprintf(stderr, "failed to allocate %llu bytes\n",
	(u_longlong_t)statbuf.st_size);
	exit(1);
	}

	if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
	(void) fprintf(stderr, "failed to read %llu bytes\n",
	(u_longlong_t)statbuf.st_size);
	exit(1);
	}

	(void) close(fd);

	if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) {
	(void) fprintf(stderr, "failed to unpack nvlist\n");
	exit(1);
	}

	free(buf);

	dump_nvlist(config, 0);

	nvlist_free(config);
	}

	/*
	* ZFS label nvlist stats
	*/
	typedef struct zdb_nvl_stats {
	int zns_list_count;
	int zns_leaf_count;
	size_t zns_leaf_largest;
	size_t zns_leaf_total;
	nvlist_t *zns_string;
	nvlist_t *zns_uint64;
	nvlist_t *zns_boolean;
	} zdb_nvl_stats_t;

	static void
	collect_nvlist_stats(nvlist_t nvl, zdb_nvl_stats_t stats)
	{
	nvlist_t list, *array;
	nvpair_t *nvp = NULL;
	char *name;
	uint_t i, items;

	stats->zns_list_count++;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	name = nvpair_name(nvp);

	switch (nvpair_type(nvp)) {
	case DATA_TYPE_STRING:
	fnvlist_add_string(stats->zns_string, name,
	fnvpair_value_string(nvp));
	break;
	case DATA_TYPE_UINT64:
	fnvlist_add_uint64(stats->zns_uint64, name,
	fnvpair_value_uint64(nvp));
	break;
	case DATA_TYPE_BOOLEAN:
	fnvlist_add_boolean(stats->zns_boolean, name);
	break;
	case DATA_TYPE_NVLIST:
	if (nvpair_value_nvlist(nvp, &list) == 0)
	collect_nvlist_stats(list, stats);
	break;
	case DATA_TYPE_NVLIST_ARRAY:
	if (nvpair_value_nvlist_array(nvp, &array, &items) != 0)
	break;

	for (i = 0; i < items; i++) {
	collect_nvlist_stats(array[i], stats);

	/* collect stats on leaf vdev */
	if (strcmp(name, "children") == 0) {
	size_t size;

	(void) nvlist_size(array[i], &size,
	NV_ENCODE_XDR);
	stats->zns_leaf_total += size;
	if (size > stats->zns_leaf_largest)
	stats->zns_leaf_largest = size;
	stats->zns_leaf_count++;
	}
	}
	break;
	default:
	(void) printf("skip type %d!\n", (int)nvpair_type(nvp));
	}
	}
	}

	static void
	dump_nvlist_stats(nvlist_t *nvl, size_t cap)
	{
	zdb_nvl_stats_t stats = { 0 };
	size_t size, sum = 0, total;
	size_t noise;

	/* requires nvlist with non-unique names for stat collection */
	VERIFY0(nvlist_alloc(&stats.zns_string, 0, 0));
	VERIFY0(nvlist_alloc(&stats.zns_uint64, 0, 0));
	VERIFY0(nvlist_alloc(&stats.zns_boolean, 0, 0));
	VERIFY0(nvlist_size(stats.zns_boolean, &noise, NV_ENCODE_XDR));

	(void) printf("\n\nZFS Label NVList Config Stats:\n");

	VERIFY0(nvlist_size(nvl, &total, NV_ENCODE_XDR));
	(void) printf(" %d bytes used, %d bytes free (using %4.1f%%)\n\n",
	(int)total, (int)(cap - total), 100.0 * total / cap);

	collect_nvlist_stats(nvl, &stats);

	VERIFY0(nvlist_size(stats.zns_uint64, &size, NV_ENCODE_XDR));
	size -= noise;
	sum += size;
	(void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "integers:",
	(int)fnvlist_num_pairs(stats.zns_uint64),
	(int)size, 100.0 * size / total);

	VERIFY0(nvlist_size(stats.zns_string, &size, NV_ENCODE_XDR));
	size -= noise;
	sum += size;
	(void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "strings:",
	(int)fnvlist_num_pairs(stats.zns_string),
	(int)size, 100.0 * size / total);

	VERIFY0(nvlist_size(stats.zns_boolean, &size, NV_ENCODE_XDR));
	size -= noise;
	sum += size;
	(void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "booleans:",
	(int)fnvlist_num_pairs(stats.zns_boolean),
	(int)size, 100.0 * size / total);

	size = total - sum; /* treat remainder as nvlist overhead */
	(void) printf("%12s %4d %6d bytes (%5.2f%%)\n\n", "nvlists:",
	stats.zns_list_count, (int)size, 100.0 * size / total);

	if (stats.zns_leaf_count > 0) {
	size_t average = stats.zns_leaf_total / stats.zns_leaf_count;

	(void) printf("%12s %4d %6d bytes average\n", "leaf vdevs:",
	stats.zns_leaf_count, (int)average);
	(void) printf("%24d bytes largest\n",
	(int)stats.zns_leaf_largest);

	if (dump_opt['l'] >= 3 && average > 0)
	(void) printf(" space for %d additional leaf vdevs\n",
	(int)((cap - total) / average));
	}
	(void) printf("\n");

	nvlist_free(stats.zns_string);
	nvlist_free(stats.zns_uint64);
	nvlist_free(stats.zns_boolean);
	}

	typedef struct cksum_record {
	zio_cksum_t cksum;
	boolean_t labels[VDEV_LABELS];
	avl_node_t link;
	} cksum_record_t;

	static int
	cksum_record_compare(const void x1, const void x2)
	{
	const cksum_record_t l = (cksum_record_t )x1;
	const cksum_record_t r = (cksum_record_t )x2;
	int arraysize = ARRAY_SIZE(l->cksum.zc_word);
	int difference = 0;

	for (int i = 0; i < arraysize; i++) {
	difference = TREE_CMP(l->cksum.zc_word[i], r->cksum.zc_word[i]);
	if (difference)
	break;
	}

	return (difference);
	}

	static cksum_record_t *
	cksum_record_alloc(zio_cksum_t *cksum, int l)
	{
	cksum_record_t *rec;

	rec = umem_zalloc(sizeof (*rec), UMEM_NOFAIL);
	rec->cksum = *cksum;
	rec->labels[l] = B_TRUE;

	return (rec);
	}

	static cksum_record_t *
	cksum_record_lookup(avl_tree_t tree, zio_cksum_t cksum)
	{
	cksum_record_t lookup = { .cksum = *cksum };
	avl_index_t where;

	return (avl_find(tree, &lookup, &where));
	}

	static cksum_record_t *
	cksum_record_insert(avl_tree_t tree, zio_cksum_t cksum, int l)
	{
	cksum_record_t *rec;

	rec = cksum_record_lookup(tree, cksum);
	if (rec) {
	rec->labels[l] = B_TRUE;
	} else {
	rec = cksum_record_alloc(cksum, l);
	avl_add(tree, rec);
	}

	return (rec);
	}

	static int
	first_label(cksum_record_t *rec)
	{
	for (int i = 0; i < VDEV_LABELS; i++)
	if (rec->labels[i])
	return (i);

	return (-1);
	}

	static void
	print_label_numbers(const char prefix, const cksum_record_t rec)
	{
	fputs(prefix, stdout);
	for (int i = 0; i < VDEV_LABELS; i++)
	if (rec->labels[i] == B_TRUE)
	printf("%d ", i);
	putchar('\n');
	}

	#define MAX_UBERBLOCK_COUNT (VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT)

	typedef struct zdb_label {
	vdev_label_t label;
	uint64_t label_offset;
	nvlist_t *config_nv;
	cksum_record_t *config;
	cksum_record_t *uberblocks[MAX_UBERBLOCK_COUNT];
	boolean_t header_printed;
	boolean_t read_failed;
	boolean_t cksum_valid;
	} zdb_label_t;

	static void
	print_label_header(zdb_label_t *label, int l)
	{

	if (dump_opt['q'])
	return;

	if (label->header_printed == B_TRUE)
	return;

	(void) printf("------------------------------------\n");
	(void) printf("LABEL %d %s\n", l,
	label->cksum_valid ? "" : "(Bad label cksum)");
	(void) printf("------------------------------------\n");

	label->header_printed = B_TRUE;
	}

	static void
	print_l2arc_header(void)
	{
	(void) printf("------------------------------------\n");
	(void) printf("L2ARC device header\n");
	(void) printf("------------------------------------\n");
	}

	static void
	print_l2arc_log_blocks(void)
	{
	(void) printf("------------------------------------\n");
	(void) printf("L2ARC device log blocks\n");
	(void) printf("------------------------------------\n");
	}

	static void
	dump_l2arc_log_entries(uint64_t log_entries,
	l2arc_log_ent_phys_t *le, uint64_t i)
	{
	for (int j = 0; j < log_entries; j++) {
	dva_t dva = le[j].le_dva;
	(void) printf("lb[%4llu]\tle[%4d]\tDVA asize: %llu, "
	"vdev: %llu, offset: %llu\n",
	(u_longlong_t)i, j + 1,
	(u_longlong_t)DVA_GET_ASIZE(&dva),
	(u_longlong_t)DVA_GET_VDEV(&dva),
	(u_longlong_t)DVA_GET_OFFSET(&dva));
	(void) printf("\|\t\t\t\tbirth: %llu\n",
	(u_longlong_t)le[j].le_birth);
	(void) printf("\|\t\t\t\tlsize: %llu\n",
	(u_longlong_t)L2BLK_GET_LSIZE((&le[j])->le_prop));
	(void) printf("\|\t\t\t\tpsize: %llu\n",
	(u_longlong_t)L2BLK_GET_PSIZE((&le[j])->le_prop));
	(void) printf("\|\t\t\t\tcompr: %llu\n",
	(u_longlong_t)L2BLK_GET_COMPRESS((&le[j])->le_prop));
	(void) printf("\|\t\t\t\tcomplevel: %llu\n",
	(u_longlong_t)(&le[j])->le_complevel);
	(void) printf("\|\t\t\t\ttype: %llu\n",
	(u_longlong_t)L2BLK_GET_TYPE((&le[j])->le_prop));
	(void) printf("\|\t\t\t\tprotected: %llu\n",
	(u_longlong_t)L2BLK_GET_PROTECTED((&le[j])->le_prop));
	(void) printf("\|\t\t\t\tprefetch: %llu\n",
	(u_longlong_t)L2BLK_GET_PREFETCH((&le[j])->le_prop));
	(void) printf("\|\t\t\t\taddress: %llu\n",
	(u_longlong_t)le[j].le_daddr);
	(void) printf("\|\t\t\t\tARC state: %llu\n",
	(u_longlong_t)L2BLK_GET_STATE((&le[j])->le_prop));
	(void) printf("\|\n");
	}
	(void) printf("\n");
	}

	static void
	dump_l2arc_log_blkptr(l2arc_log_blkptr_t lbps)
	{
	(void) printf("\|\t\tdaddr: %llu\n", (u_longlong_t)lbps.lbp_daddr);
	(void) printf("\|\t\tpayload_asize: %llu\n",
	(u_longlong_t)lbps.lbp_payload_asize);
	(void) printf("\|\t\tpayload_start: %llu\n",
	(u_longlong_t)lbps.lbp_payload_start);
	(void) printf("\|\t\tlsize: %llu\n",
	(u_longlong_t)L2BLK_GET_LSIZE((&lbps)->lbp_prop));
	(void) printf("\|\t\tasize: %llu\n",
	(u_longlong_t)L2BLK_GET_PSIZE((&lbps)->lbp_prop));
	(void) printf("\|\t\tcompralgo: %llu\n",
	(u_longlong_t)L2BLK_GET_COMPRESS((&lbps)->lbp_prop));
	(void) printf("\|\t\tcksumalgo: %llu\n",
	(u_longlong_t)L2BLK_GET_CHECKSUM((&lbps)->lbp_prop));
	(void) printf("\|\n\n");
	}

	static void
	dump_l2arc_log_blocks(int fd, l2arc_dev_hdr_phys_t l2dhdr,
	l2arc_dev_hdr_phys_t *rebuild)
	{
	l2arc_log_blk_phys_t this_lb;
	uint64_t asize;
	l2arc_log_blkptr_t lbps[2];
	abd_t *abd;
	zio_cksum_t cksum;
	int failed = 0;
	l2arc_dev_t dev;

	if (!dump_opt['q'])
	print_l2arc_log_blocks();
	memcpy(lbps, l2dhdr.dh_start_lbps, sizeof (lbps));

	dev.l2ad_evict = l2dhdr.dh_evict;
	dev.l2ad_start = l2dhdr.dh_start;
	dev.l2ad_end = l2dhdr.dh_end;

	if (l2dhdr.dh_start_lbps[0].lbp_daddr == 0) {
	/* no log blocks to read */
	if (!dump_opt['q']) {
	(void) printf("No log blocks to read\n");
	(void) printf("\n");
	}
	return;
	} else {
	dev.l2ad_hand = lbps[0].lbp_daddr +
	L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
	}

	dev.l2ad_first = !!(l2dhdr.dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);

	for (;;) {
	if (!l2arc_log_blkptr_valid(&dev, &lbps[0]))
	break;

	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	asize = L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
	if (pread64(fd, &this_lb, asize, lbps[0].lbp_daddr) != asize) {
	if (!dump_opt['q']) {
	(void) printf("Error while reading next log "
	"block\n\n");
	}
	break;
	}

	fletcher_4_native_varsize(&this_lb, asize, &cksum);
	if (!ZIO_CHECKSUM_EQUAL(cksum, lbps[0].lbp_cksum)) {
	failed++;
	if (!dump_opt['q']) {
	(void) printf("Invalid cksum\n");
	dump_l2arc_log_blkptr(lbps[0]);
	}
	break;
	}

	switch (L2BLK_GET_COMPRESS((&lbps[0])->lbp_prop)) {
	case ZIO_COMPRESS_OFF:
	break;
	default:
	abd = abd_alloc_for_io(asize, B_TRUE);
	abd_copy_from_buf_off(abd, &this_lb, 0, asize);
	- zio_decompress_data(L2BLK_GET_COMPRESS(
	+ if (zio_decompress_data(L2BLK_GET_COMPRESS(
	(&lbps[0])->lbp_prop), abd, &this_lb,
	- asize, sizeof (this_lb), NULL);
	+ asize, sizeof (this_lb), NULL) != 0) {
	+ (void) printf("L2ARC block decompression "
	+ "failed\n");
	+ abd_free(abd);
	+ goto out;
	+ }
	abd_free(abd);
	break;
	}

	if (this_lb.lb_magic == BSWAP_64(L2ARC_LOG_BLK_MAGIC))
	byteswap_uint64_array(&this_lb, sizeof (this_lb));
	if (this_lb.lb_magic != L2ARC_LOG_BLK_MAGIC) {
	if (!dump_opt['q'])
	(void) printf("Invalid log block magic\n\n");
	break;
	}

	rebuild->dh_lb_count++;
	rebuild->dh_lb_asize += asize;
	if (dump_opt['l'] > 1 && !dump_opt['q']) {
	(void) printf("lb[%4llu]\tmagic: %llu\n",
	(u_longlong_t)rebuild->dh_lb_count,
	(u_longlong_t)this_lb.lb_magic);
	dump_l2arc_log_blkptr(lbps[0]);
	}

	if (dump_opt['l'] > 2 && !dump_opt['q'])
	dump_l2arc_log_entries(l2dhdr.dh_log_entries,
	this_lb.lb_entries,
	rebuild->dh_lb_count);

	if (l2arc_range_check_overlap(lbps[1].lbp_payload_start,
	lbps[0].lbp_payload_start, dev.l2ad_evict) &&
	!dev.l2ad_first)
	break;

	lbps[0] = lbps[1];
	lbps[1] = this_lb.lb_prev_lbp;
	}
	-
	+out:
	if (!dump_opt['q']) {
	(void) printf("log_blk_count:\t %llu with valid cksum\n",
	(u_longlong_t)rebuild->dh_lb_count);
	(void) printf("\t\t %d with invalid cksum\n", failed);
	(void) printf("log_blk_asize:\t %llu\n\n",
	(u_longlong_t)rebuild->dh_lb_asize);
	}
	}

	static int
	dump_l2arc_header(int fd)
	{
	l2arc_dev_hdr_phys_t l2dhdr = {0}, rebuild = {0};
	int error = B_FALSE;

	if (pread64(fd, &l2dhdr, sizeof (l2dhdr),
	VDEV_LABEL_START_SIZE) != sizeof (l2dhdr)) {
	error = B_TRUE;
	} else {
	if (l2dhdr.dh_magic == BSWAP_64(L2ARC_DEV_HDR_MAGIC))
	byteswap_uint64_array(&l2dhdr, sizeof (l2dhdr));

	if (l2dhdr.dh_magic != L2ARC_DEV_HDR_MAGIC)
	error = B_TRUE;
	}

	if (error) {
	(void) printf("L2ARC device header not found\n\n");
	/* Do not return an error here for backward compatibility */
	return (0);
	} else if (!dump_opt['q']) {
	print_l2arc_header();

	(void) printf(" magic: %llu\n",
	(u_longlong_t)l2dhdr.dh_magic);
	(void) printf(" version: %llu\n",
	(u_longlong_t)l2dhdr.dh_version);
	(void) printf(" pool_guid: %llu\n",
	(u_longlong_t)l2dhdr.dh_spa_guid);
	(void) printf(" flags: %llu\n",
	(u_longlong_t)l2dhdr.dh_flags);
	(void) printf(" start_lbps[0]: %llu\n",
	(u_longlong_t)
	l2dhdr.dh_start_lbps[0].lbp_daddr);
	(void) printf(" start_lbps[1]: %llu\n",
	(u_longlong_t)
	l2dhdr.dh_start_lbps[1].lbp_daddr);
	(void) printf(" log_blk_ent: %llu\n",
	(u_longlong_t)l2dhdr.dh_log_entries);
	(void) printf(" start: %llu\n",
	(u_longlong_t)l2dhdr.dh_start);
	(void) printf(" end: %llu\n",
	(u_longlong_t)l2dhdr.dh_end);
	(void) printf(" evict: %llu\n",
	(u_longlong_t)l2dhdr.dh_evict);
	(void) printf(" lb_asize_refcount: %llu\n",
	(u_longlong_t)l2dhdr.dh_lb_asize);
	(void) printf(" lb_count_refcount: %llu\n",
	(u_longlong_t)l2dhdr.dh_lb_count);
	(void) printf(" trim_action_time: %llu\n",
	(u_longlong_t)l2dhdr.dh_trim_action_time);
	(void) printf(" trim_state: %llu\n\n",
	(u_longlong_t)l2dhdr.dh_trim_state);
	}

	dump_l2arc_log_blocks(fd, l2dhdr, &rebuild);
	/*
	* The total aligned size of log blocks and the number of log blocks
	* reported in the header of the device may be less than what zdb
	* reports by dump_l2arc_log_blocks() which emulates l2arc_rebuild().
	* This happens because dump_l2arc_log_blocks() lacks the memory
	* pressure valve that l2arc_rebuild() has. Thus, if we are on a system
	* with low memory, l2arc_rebuild will exit prematurely and dh_lb_asize
	* and dh_lb_count will be lower to begin with than what exists on the
	* device. This is normal and zdb should not exit with an error. The
	* opposite case should never happen though, the values reported in the
	* header should never be higher than what dump_l2arc_log_blocks() and
	* l2arc_rebuild() report. If this happens there is a leak in the
	* accounting of log blocks.
	*/
	if (l2dhdr.dh_lb_asize > rebuild.dh_lb_asize \|\|
	l2dhdr.dh_lb_count > rebuild.dh_lb_count)
	return (1);

	return (0);
	}

	static void
	dump_config_from_label(zdb_label_t *label, size_t buflen, int l)
	{
	if (dump_opt['q'])
	return;

	if ((dump_opt['l'] < 3) && (first_label(label->config) != l))
	return;

	print_label_header(label, l);
	dump_nvlist(label->config_nv, 4);
	print_label_numbers(" labels = ", label->config);

	if (dump_opt['l'] >= 2)
	dump_nvlist_stats(label->config_nv, buflen);
	}

	#define ZDB_MAX_UB_HEADER_SIZE 32

	static void
	dump_label_uberblocks(zdb_label_t *label, uint64_t ashift, int label_num)
	{

	vdev_t vd;
	char header[ZDB_MAX_UB_HEADER_SIZE];

	vd.vdev_ashift = ashift;
	vd.vdev_top = &vd;

	for (int i = 0; i < VDEV_UBERBLOCK_COUNT(&vd); i++) {
	uint64_t uoff = VDEV_UBERBLOCK_OFFSET(&vd, i);
	uberblock_t ub = (void )((char *)&label->label + uoff);
	cksum_record_t *rec = label->uberblocks[i];

	if (rec == NULL) {
	if (dump_opt['u'] >= 2) {
	print_label_header(label, label_num);
	(void) printf(" Uberblock[%d] invalid\n", i);
	}
	continue;
	}

	if ((dump_opt['u'] < 3) && (first_label(rec) != label_num))
	continue;

	if ((dump_opt['u'] < 4) &&
	(ub->ub_mmp_magic == MMP_MAGIC) && ub->ub_mmp_delay &&
	(i >= VDEV_UBERBLOCK_COUNT(&vd) - MMP_BLOCKS_PER_LABEL))
	continue;

	print_label_header(label, label_num);
	(void) snprintf(header, ZDB_MAX_UB_HEADER_SIZE,
	" Uberblock[%d]\n", i);
	dump_uberblock(ub, header, "");
	print_label_numbers(" labels = ", rec);
	}
	}

	static char curpath[PATH_MAX];

	/*
	* Iterate through the path components, recursively passing
	* current one's obj and remaining path until we find the obj
	* for the last one.
	*/
	static int
	dump_path_impl(objset_t os, uint64_t obj, char name, uint64_t *retobj)
	{
	int err;
	boolean_t header = B_TRUE;
	uint64_t child_obj;
	char *s;
	dmu_buf_t *db;
	dmu_object_info_t doi;

	if ((s = strchr(name, '/')) != NULL)
	*s = '\0';
	err = zap_lookup(os, obj, name, 8, 1, &child_obj);

	(void) strlcat(curpath, name, sizeof (curpath));

	if (err != 0) {
	(void) fprintf(stderr, "failed to lookup %s: %s\n",
	curpath, strerror(err));
	return (err);
	}

	child_obj = ZFS_DIRENT_OBJ(child_obj);
	err = sa_buf_hold(os, child_obj, FTAG, &db);
	if (err != 0) {
	(void) fprintf(stderr,
	"failed to get SA dbuf for obj %llu: %s\n",
	(u_longlong_t)child_obj, strerror(err));
	return (EINVAL);
	}
	dmu_object_info_from_db(db, &doi);
	sa_buf_rele(db, FTAG);

	if (doi.doi_bonus_type != DMU_OT_SA &&
	doi.doi_bonus_type != DMU_OT_ZNODE) {
	(void) fprintf(stderr, "invalid bonus type %d for obj %llu\n",
	doi.doi_bonus_type, (u_longlong_t)child_obj);
	return (EINVAL);
	}

	if (dump_opt['v'] > 6) {
	(void) printf("obj=%llu %s type=%d bonustype=%d\n",
	(u_longlong_t)child_obj, curpath, doi.doi_type,
	doi.doi_bonus_type);
	}

	(void) strlcat(curpath, "/", sizeof (curpath));

	switch (doi.doi_type) {
	case DMU_OT_DIRECTORY_CONTENTS:
	if (s != NULL && *(s + 1) != '\0')
	return (dump_path_impl(os, child_obj, s + 1, retobj));
	zfs_fallthrough;
	case DMU_OT_PLAIN_FILE_CONTENTS:
	if (retobj != NULL) {
	*retobj = child_obj;
	} else {
	dump_object(os, child_obj, dump_opt['v'], &header,
	NULL, 0);
	}
	return (0);
	default:
	(void) fprintf(stderr, "object %llu has non-file/directory "
	"type %d\n", (u_longlong_t)obj, doi.doi_type);
	break;
	}

	return (EINVAL);
	}

	/*
	* Dump the blocks for the object specified by path inside the dataset.
	*/
	static int
	dump_path(char ds, char path, uint64_t *retobj)
	{
	int err;
	objset_t *os;
	uint64_t root_obj;

	err = open_objset(ds, FTAG, &os);
	if (err != 0)
	return (err);

	err = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, &root_obj);
	if (err != 0) {
	(void) fprintf(stderr, "can't lookup root znode: %s\n",
	strerror(err));
	close_objset(os, FTAG);
	return (EINVAL);
	}

	(void) snprintf(curpath, sizeof (curpath), "dataset=%s path=/", ds);

	err = dump_path_impl(os, root_obj, path, retobj);

	close_objset(os, FTAG);
	return (err);
	}

	static int
	zdb_copy_object(objset_t os, uint64_t srcobj, char destfile)
	{
	int err = 0;
	uint64_t size, readsize, oursize, offset;
	ssize_t writesize;
	sa_handle_t *hdl;

	(void) printf("Copying object %" PRIu64 " to file %s\n", srcobj,
	destfile);

	VERIFY3P(os, ==, sa_os);
	if ((err = sa_handle_get(os, srcobj, NULL, SA_HDL_PRIVATE, &hdl))) {
	(void) printf("Failed to get handle for SA znode\n");
	return (err);
	}
	if ((err = sa_lookup(hdl, sa_attr_table[ZPL_SIZE], &size, 8))) {
	(void) sa_handle_destroy(hdl);
	return (err);
	}
	(void) sa_handle_destroy(hdl);

	(void) printf("Object %" PRIu64 " is %" PRIu64 " bytes\n", srcobj,
	size);
	if (size == 0) {
	return (EINVAL);
	}

	int fd = open(destfile, O_WRONLY \| O_CREAT \| O_TRUNC, 0644);
	if (fd == -1)
	return (errno);
	/*
	* We cap the size at 1 mebibyte here to prevent
	* allocation failures and nigh-infinite printing if the
	* object is extremely large.
	*/
	oursize = MIN(size, 1 << 20);
	offset = 0;
	char *buf = kmem_alloc(oursize, KM_NOSLEEP);
	if (buf == NULL) {
	(void) close(fd);
	return (ENOMEM);
	}

	while (offset < size) {
	readsize = MIN(size - offset, 1 << 20);
	err = dmu_read(os, srcobj, offset, readsize, buf, 0);
	if (err != 0) {
	(void) printf("got error %u from dmu_read\n", err);
	kmem_free(buf, oursize);
	(void) close(fd);
	return (err);
	}
	if (dump_opt['v'] > 3) {
	(void) printf("Read offset=%" PRIu64 " size=%" PRIu64
	" error=%d\n", offset, readsize, err);
	}

	writesize = write(fd, buf, readsize);
	if (writesize < 0) {
	err = errno;
	break;
	} else if (writesize != readsize) {
	/* Incomplete write */
	(void) fprintf(stderr, "Short write, only wrote %llu of"
	" %" PRIu64 " bytes, exiting...\n",
	(u_longlong_t)writesize, readsize);
	break;
	}

	offset += readsize;
	}

	(void) close(fd);

	if (buf != NULL)
	kmem_free(buf, oursize);

	return (err);
	}

	static boolean_t
	label_cksum_valid(vdev_label_t *label, uint64_t offset)
	{
	zio_checksum_info_t *ci = &zio_checksum_table[ZIO_CHECKSUM_LABEL];
	zio_cksum_t expected_cksum;
	zio_cksum_t actual_cksum;
	zio_cksum_t verifier;
	zio_eck_t *eck;
	int byteswap;

	void data = (char )label + offsetof(vdev_label_t, vl_vdev_phys);
	eck = (zio_eck_t )((char )(data) + VDEV_PHYS_SIZE) - 1;

	offset += offsetof(vdev_label_t, vl_vdev_phys);
	ZIO_SET_CHECKSUM(&verifier, offset, 0, 0, 0);

	byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC));
	if (byteswap)
	byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));

	expected_cksum = eck->zec_cksum;
	eck->zec_cksum = verifier;

	abd_t *abd = abd_get_from_buf(data, VDEV_PHYS_SIZE);
	ci->ci_func[byteswap](abd, VDEV_PHYS_SIZE, NULL, &actual_cksum);
	abd_free(abd);

	if (byteswap)
	byteswap_uint64_array(&expected_cksum, sizeof (zio_cksum_t));

	if (ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
	return (B_TRUE);

	return (B_FALSE);
	}

	static int
	dump_label(const char *dev)
	{
	char path[MAXPATHLEN];
	zdb_label_t labels[VDEV_LABELS] = {{{{0}}}};
	uint64_t psize, ashift, l2cache;
	struct stat64 statbuf;
	boolean_t config_found = B_FALSE;
	boolean_t error = B_FALSE;
	boolean_t read_l2arc_header = B_FALSE;
	avl_tree_t config_tree;
	avl_tree_t uberblock_tree;
	void node, cookie;
	int fd;

	/*
	* Check if we were given absolute path and use it as is.
	* Otherwise if the provided vdev name doesn't point to a file,
	* try prepending expected disk paths and partition numbers.
	*/
	(void) strlcpy(path, dev, sizeof (path));
	if (dev[0] != '/' && stat64(path, &statbuf) != 0) {
	int error;

	error = zfs_resolve_shortname(dev, path, MAXPATHLEN);
	if (error == 0 && zfs_dev_is_whole_disk(path)) {
	if (zfs_append_partition(path, MAXPATHLEN) == -1)
	error = ENOENT;
	}

	if (error \|\| (stat64(path, &statbuf) != 0)) {
	(void) printf("failed to find device %s, try "
	"specifying absolute path instead\n", dev);
	return (1);
	}
	}

	if ((fd = open64(path, O_RDONLY)) < 0) {
	(void) printf("cannot open '%s': %s\n", path, strerror(errno));
	exit(1);
	}

	if (fstat64_blk(fd, &statbuf) != 0) {
	(void) printf("failed to stat '%s': %s\n", path,
	strerror(errno));
	(void) close(fd);
	exit(1);
	}

	if (S_ISBLK(statbuf.st_mode) && zfs_dev_flush(fd) != 0)
	(void) printf("failed to invalidate cache '%s' : %s\n", path,
	strerror(errno));

	avl_create(&config_tree, cksum_record_compare,
	sizeof (cksum_record_t), offsetof(cksum_record_t, link));
	avl_create(&uberblock_tree, cksum_record_compare,
	sizeof (cksum_record_t), offsetof(cksum_record_t, link));

	psize = statbuf.st_size;
	psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
	ashift = SPA_MINBLOCKSHIFT;

	/*
	* 1. Read the label from disk
	* 2. Verify label cksum
	* 3. Unpack the configuration and insert in config tree.
	* 4. Traverse all uberblocks and insert in uberblock tree.
	*/
	for (int l = 0; l < VDEV_LABELS; l++) {
	zdb_label_t *label = &labels[l];
	char *buf = label->label.vl_vdev_phys.vp_nvlist;
	size_t buflen = sizeof (label->label.vl_vdev_phys.vp_nvlist);
	nvlist_t *config;
	cksum_record_t *rec;
	zio_cksum_t cksum;
	vdev_t vd;

	label->label_offset = vdev_label_offset(psize, l, 0);

	if (pread64(fd, &label->label, sizeof (label->label),
	label->label_offset) != sizeof (label->label)) {
	if (!dump_opt['q'])
	(void) printf("failed to read label %d\n", l);
	label->read_failed = B_TRUE;
	error = B_TRUE;
	continue;
	}

	label->read_failed = B_FALSE;
	label->cksum_valid = label_cksum_valid(&label->label,
	label->label_offset);

	if (nvlist_unpack(buf, buflen, &config, 0) == 0) {
	nvlist_t *vdev_tree = NULL;
	size_t size;

	if ((nvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE, &vdev_tree) != 0) \|\|
	(nvlist_lookup_uint64(vdev_tree,
	ZPOOL_CONFIG_ASHIFT, &ashift) != 0))
	ashift = SPA_MINBLOCKSHIFT;

	if (nvlist_size(config, &size, NV_ENCODE_XDR) != 0)
	size = buflen;

	/* If the device is a cache device clear the header. */
	if (!read_l2arc_header) {
	if (nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_POOL_STATE, &l2cache) == 0 &&
	l2cache == POOL_STATE_L2CACHE) {
	read_l2arc_header = B_TRUE;
	}
	}

	fletcher_4_native_varsize(buf, size, &cksum);
	rec = cksum_record_insert(&config_tree, &cksum, l);

	label->config = rec;
	label->config_nv = config;
	config_found = B_TRUE;
	} else {
	error = B_TRUE;
	}

	vd.vdev_ashift = ashift;
	vd.vdev_top = &vd;

	for (int i = 0; i < VDEV_UBERBLOCK_COUNT(&vd); i++) {
	uint64_t uoff = VDEV_UBERBLOCK_OFFSET(&vd, i);
	uberblock_t ub = (void )((char *)label + uoff);

	if (uberblock_verify(ub))
	continue;

	fletcher_4_native_varsize(ub, sizeof (*ub), &cksum);
	rec = cksum_record_insert(&uberblock_tree, &cksum, l);

	label->uberblocks[i] = rec;
	}
	}

	/*
	* Dump the label and uberblocks.
	*/
	for (int l = 0; l < VDEV_LABELS; l++) {
	zdb_label_t *label = &labels[l];
	size_t buflen = sizeof (label->label.vl_vdev_phys.vp_nvlist);

	if (label->read_failed == B_TRUE)
	continue;

	if (label->config_nv) {
	dump_config_from_label(label, buflen, l);
	} else {
	if (!dump_opt['q'])
	(void) printf("failed to unpack label %d\n", l);
	}

	if (dump_opt['u'])
	dump_label_uberblocks(label, ashift, l);

	nvlist_free(label->config_nv);
	}

	/*
	* Dump the L2ARC header, if existent.
	*/
	if (read_l2arc_header)
	error \|= dump_l2arc_header(fd);

	cookie = NULL;
	while ((node = avl_destroy_nodes(&config_tree, &cookie)) != NULL)
	umem_free(node, sizeof (cksum_record_t));

	cookie = NULL;
	while ((node = avl_destroy_nodes(&uberblock_tree, &cookie)) != NULL)
	umem_free(node, sizeof (cksum_record_t));

	avl_destroy(&config_tree);
	avl_destroy(&uberblock_tree);

	(void) close(fd);

	return (config_found == B_FALSE ? 2 :
	(error == B_TRUE ? 1 : 0));
	}

	static uint64_t dataset_feature_count[SPA_FEATURES];
	static uint64_t global_feature_count[SPA_FEATURES];
	static uint64_t remap_deadlist_count = 0;

	static int
	dump_one_objset(const char dsname, void arg)
	{
	(void) arg;
	int error;
	objset_t *os;
	spa_feature_t f;

	error = open_objset(dsname, FTAG, &os);
	if (error != 0)
	return (0);

	for (f = 0; f < SPA_FEATURES; f++) {
	if (!dsl_dataset_feature_is_active(dmu_objset_ds(os), f))
	continue;
	ASSERT(spa_feature_table[f].fi_flags &
	ZFEATURE_FLAG_PER_DATASET);
	dataset_feature_count[f]++;
	}

	if (dsl_dataset_remap_deadlist_exists(dmu_objset_ds(os))) {
	remap_deadlist_count++;
	}

	for (dsl_bookmark_node_t *dbn =
	avl_first(&dmu_objset_ds(os)->ds_bookmarks); dbn != NULL;
	dbn = AVL_NEXT(&dmu_objset_ds(os)->ds_bookmarks, dbn)) {
	mos_obj_refd(dbn->dbn_phys.zbm_redaction_obj);
	if (dbn->dbn_phys.zbm_redaction_obj != 0)
	global_feature_count[SPA_FEATURE_REDACTION_BOOKMARKS]++;
	if (dbn->dbn_phys.zbm_flags & ZBM_FLAG_HAS_FBN)
	global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN]++;
	}

	if (dsl_deadlist_is_open(&dmu_objset_ds(os)->ds_dir->dd_livelist) &&
	!dmu_objset_is_snapshot(os)) {
	global_feature_count[SPA_FEATURE_LIVELIST]++;
	}

	dump_objset(os);
	close_objset(os, FTAG);
	fuid_table_destroy();
	return (0);
	}

	/*
	* Block statistics.
	*/
	#define PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
	typedef struct zdb_blkstats {
	uint64_t zb_asize;
	uint64_t zb_lsize;
	uint64_t zb_psize;
	uint64_t zb_count;
	uint64_t zb_gangs;
	uint64_t zb_ditto_samevdev;
	uint64_t zb_ditto_same_ms;
	uint64_t zb_psize_histogram[PSIZE_HISTO_SIZE];
	} zdb_blkstats_t;

	/*
	* Extended object types to report deferred frees and dedup auto-ditto blocks.
	*/
	#define ZDB_OT_DEFERRED (DMU_OT_NUMTYPES + 0)
	#define ZDB_OT_DITTO (DMU_OT_NUMTYPES + 1)
	#define ZDB_OT_OTHER (DMU_OT_NUMTYPES + 2)
	#define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 3)

	static const char *zdb_ot_extname[] = {
	"deferred free",
	"dedup ditto",
	"other",
	"Total",
	};

	#define ZB_TOTAL DN_MAX_LEVELS
	#define SPA_MAX_FOR_16M (SPA_MAXBLOCKSHIFT+1)

	typedef struct zdb_cb {
	zdb_blkstats_t zcb_type[ZB_TOTAL + 1][ZDB_OT_TOTAL + 1];
	uint64_t zcb_removing_size;
	uint64_t zcb_checkpoint_size;
	uint64_t zcb_dedup_asize;
	uint64_t zcb_dedup_blocks;
	uint64_t zcb_psize_count[SPA_MAX_FOR_16M];
	uint64_t zcb_lsize_count[SPA_MAX_FOR_16M];
	uint64_t zcb_asize_count[SPA_MAX_FOR_16M];
	uint64_t zcb_psize_len[SPA_MAX_FOR_16M];
	uint64_t zcb_lsize_len[SPA_MAX_FOR_16M];
	uint64_t zcb_asize_len[SPA_MAX_FOR_16M];
	uint64_t zcb_psize_total;
	uint64_t zcb_lsize_total;
	uint64_t zcb_asize_total;
	uint64_t zcb_embedded_blocks[NUM_BP_EMBEDDED_TYPES];
	uint64_t zcb_embedded_histogram[NUM_BP_EMBEDDED_TYPES]
	[BPE_PAYLOAD_SIZE + 1];
	uint64_t zcb_start;
	hrtime_t zcb_lastprint;
	uint64_t zcb_totalasize;
	uint64_t zcb_errors[256];
	int zcb_readfails;
	int zcb_haderrors;
	spa_t *zcb_spa;
	uint32_t **zcb_vd_obsolete_counts;
	} zdb_cb_t;

	/* test if two DVA offsets from same vdev are within the same metaslab */
	static boolean_t
	same_metaslab(spa_t *spa, uint64_t vdev, uint64_t off1, uint64_t off2)
	{
	vdev_t *vd = vdev_lookup_top(spa, vdev);
	uint64_t ms_shift = vd->vdev_ms_shift;

	return ((off1 >> ms_shift) == (off2 >> ms_shift));
	}

	/*
	* Used to simplify reporting of the histogram data.
	*/
	typedef struct one_histo {
	const char *name;
	uint64_t *count;
	uint64_t *len;
	uint64_t cumulative;
	} one_histo_t;

	/*
	* The number of separate histograms processed for psize, lsize and asize.
	*/
	#define NUM_HISTO 3

	/*
	* This routine will create a fixed column size output of three different
	* histograms showing by blocksize of 512 - 2^ SPA_MAX_FOR_16M
	* the count, length and cumulative length of the psize, lsize and
	* asize blocks.
	*
	* All three types of blocks are listed on a single line
	*
	* By default the table is printed in nicenumber format (e.g. 123K) but
	* if the '-P' parameter is specified then the full raw number (parseable)
	* is printed out.
	*/
	static void
	dump_size_histograms(zdb_cb_t *zcb)
	{
	/*
	* A temporary buffer that allows us to convert a number into
	* a string using zdb_nicenumber to allow either raw or human
	* readable numbers to be output.
	*/
	char numbuf[32];

	/*
	* Define titles which are used in the headers of the tables
	* printed by this routine.
	*/
	const char blocksize_title1[] = "block";
	const char blocksize_title2[] = "size";
	const char count_title[] = "Count";
	const char length_title[] = "Size";
	const char cumulative_title[] = "Cum.";

	/*
	* Setup the histogram arrays (psize, lsize, and asize).
	*/
	one_histo_t parm_histo[NUM_HISTO];

	parm_histo[0].name = "psize";
	parm_histo[0].count = zcb->zcb_psize_count;
	parm_histo[0].len = zcb->zcb_psize_len;
	parm_histo[0].cumulative = 0;

	parm_histo[1].name = "lsize";
	parm_histo[1].count = zcb->zcb_lsize_count;
	parm_histo[1].len = zcb->zcb_lsize_len;
	parm_histo[1].cumulative = 0;

	parm_histo[2].name = "asize";
	parm_histo[2].count = zcb->zcb_asize_count;
	parm_histo[2].len = zcb->zcb_asize_len;
	parm_histo[2].cumulative = 0;


	(void) printf("\nBlock Size Histogram\n");
	/*
	* Print the first line titles
	*/
	if (dump_opt['P'])
	(void) printf("\n%s\t", blocksize_title1);
	else
	(void) printf("\n%7s ", blocksize_title1);

	for (int j = 0; j < NUM_HISTO; j++) {
	if (dump_opt['P']) {
	if (j < NUM_HISTO - 1) {
	(void) printf("%s\t\t\t", parm_histo[j].name);
	} else {
	/* Don't print trailing spaces */
	(void) printf(" %s", parm_histo[j].name);
	}
	} else {
	if (j < NUM_HISTO - 1) {
	/* Left aligned strings in the output */
	(void) printf("%-7s ",
	parm_histo[j].name);
	} else {
	/* Don't print trailing spaces */
	(void) printf("%s", parm_histo[j].name);
	}
	}
	}
	(void) printf("\n");

	/*
	* Print the second line titles
	*/
	if (dump_opt['P']) {
	(void) printf("%s\t", blocksize_title2);
	} else {
	(void) printf("%7s ", blocksize_title2);
	}

	for (int i = 0; i < NUM_HISTO; i++) {
	if (dump_opt['P']) {
	(void) printf("%s\t%s\t%s\t",
	count_title, length_title, cumulative_title);
	} else {
	(void) printf("%7s%7s%7s",
	count_title, length_title, cumulative_title);
	}
	}
	(void) printf("\n");

	/*
	* Print the rows
	*/
	for (int i = SPA_MINBLOCKSHIFT; i < SPA_MAX_FOR_16M; i++) {

	/*
	* Print the first column showing the blocksize
	*/
	zdb_nicenum((1ULL << i), numbuf, sizeof (numbuf));

	if (dump_opt['P']) {
	printf("%s", numbuf);
	} else {
	printf("%7s:", numbuf);
	}

	/*
	* Print the remaining set of 3 columns per size:
	* for psize, lsize and asize
	*/
	for (int j = 0; j < NUM_HISTO; j++) {
	parm_histo[j].cumulative += parm_histo[j].len[i];

	zdb_nicenum(parm_histo[j].count[i],
	numbuf, sizeof (numbuf));
	if (dump_opt['P'])
	(void) printf("\t%s", numbuf);
	else
	(void) printf("%7s", numbuf);

	zdb_nicenum(parm_histo[j].len[i],
	numbuf, sizeof (numbuf));
	if (dump_opt['P'])
	(void) printf("\t%s", numbuf);
	else
	(void) printf("%7s", numbuf);

	zdb_nicenum(parm_histo[j].cumulative,
	numbuf, sizeof (numbuf));
	if (dump_opt['P'])
	(void) printf("\t%s", numbuf);
	else
	(void) printf("%7s", numbuf);
	}
	(void) printf("\n");
	}
	}

	static void
	zdb_count_block(zdb_cb_t zcb, zilog_t zilog, const blkptr_t *bp,
	dmu_object_type_t type)
	{
	uint64_t refcnt = 0;
	int i;

	ASSERT(type < ZDB_OT_TOTAL);

	if (zilog && zil_bp_tree_add(zilog, bp) != 0)
	return;

	spa_config_enter(zcb->zcb_spa, SCL_CONFIG, FTAG, RW_READER);

	for (i = 0; i < 4; i++) {
	int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL;
	int t = (i & 1) ? type : ZDB_OT_TOTAL;
	int equal;
	zdb_blkstats_t *zb = &zcb->zcb_type[l][t];

	zb->zb_asize += BP_GET_ASIZE(bp);
	zb->zb_lsize += BP_GET_LSIZE(bp);
	zb->zb_psize += BP_GET_PSIZE(bp);
	zb->zb_count++;

	/*
	* The histogram is only big enough to record blocks up to
	* SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
	* "other", bucket.
	*/
	unsigned idx = BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT;
	idx = MIN(idx, SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 1);
	zb->zb_psize_histogram[idx]++;

	zb->zb_gangs += BP_COUNT_GANG(bp);

	switch (BP_GET_NDVAS(bp)) {
	case 2:
	if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[1])) {
	zb->zb_ditto_samevdev++;

	if (same_metaslab(zcb->zcb_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[1])))
	zb->zb_ditto_same_ms++;
	}
	break;
	case 3:
	equal = (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[1])) +
	(DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[2])) +
	(DVA_GET_VDEV(&bp->blk_dva[1]) ==
	DVA_GET_VDEV(&bp->blk_dva[2]));
	if (equal != 0) {
	zb->zb_ditto_samevdev++;

	if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[1]) &&
	same_metaslab(zcb->zcb_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[1])))
	zb->zb_ditto_same_ms++;
	else if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[2]) &&
	same_metaslab(zcb->zcb_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[0]),
	DVA_GET_OFFSET(&bp->blk_dva[2])))
	zb->zb_ditto_same_ms++;
	else if (DVA_GET_VDEV(&bp->blk_dva[1]) ==
	DVA_GET_VDEV(&bp->blk_dva[2]) &&
	same_metaslab(zcb->zcb_spa,
	DVA_GET_VDEV(&bp->blk_dva[1]),
	DVA_GET_OFFSET(&bp->blk_dva[1]),
	DVA_GET_OFFSET(&bp->blk_dva[2])))
	zb->zb_ditto_same_ms++;
	}
	break;
	}
	}

	spa_config_exit(zcb->zcb_spa, SCL_CONFIG, FTAG);

	if (BP_IS_EMBEDDED(bp)) {
	zcb->zcb_embedded_blocks[BPE_GET_ETYPE(bp)]++;
	zcb->zcb_embedded_histogram[BPE_GET_ETYPE(bp)]
	[BPE_GET_PSIZE(bp)]++;
	return;
	}
	/*
	* The binning histogram bins by powers of two up to
	* SPA_MAXBLOCKSIZE rather than creating bins for
	* every possible blocksize found in the pool.
	*/
	int bin = highbit64(BP_GET_PSIZE(bp)) - 1;

	zcb->zcb_psize_count[bin]++;
	zcb->zcb_psize_len[bin] += BP_GET_PSIZE(bp);
	zcb->zcb_psize_total += BP_GET_PSIZE(bp);

	bin = highbit64(BP_GET_LSIZE(bp)) - 1;

	zcb->zcb_lsize_count[bin]++;
	zcb->zcb_lsize_len[bin] += BP_GET_LSIZE(bp);
	zcb->zcb_lsize_total += BP_GET_LSIZE(bp);

	bin = highbit64(BP_GET_ASIZE(bp)) - 1;

	zcb->zcb_asize_count[bin]++;
	zcb->zcb_asize_len[bin] += BP_GET_ASIZE(bp);
	zcb->zcb_asize_total += BP_GET_ASIZE(bp);

	if (dump_opt['L'])
	return;

	if (BP_GET_DEDUP(bp)) {
	ddt_t *ddt;
	ddt_entry_t *dde;

	ddt = ddt_select(zcb->zcb_spa, bp);
	ddt_enter(ddt);
	dde = ddt_lookup(ddt, bp, B_FALSE);

	if (dde == NULL) {
	refcnt = 0;
	} else {
	ddt_phys_t *ddp = ddt_phys_select(dde, bp);
	ddt_phys_decref(ddp);
	refcnt = ddp->ddp_refcnt;
	if (ddt_phys_total_refcnt(dde) == 0)
	ddt_remove(ddt, dde);
	}
	ddt_exit(ddt);
	}

	VERIFY3U(zio_wait(zio_claim(NULL, zcb->zcb_spa,
	refcnt ? 0 : spa_min_claim_txg(zcb->zcb_spa),
	bp, NULL, NULL, ZIO_FLAG_CANFAIL)), ==, 0);
	}

	static void
	zdb_blkptr_done(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	int ioerr = zio->io_error;
	zdb_cb_t *zcb = zio->io_private;
	zbookmark_phys_t *zb = &zio->io_bookmark;

	mutex_enter(&spa->spa_scrub_lock);
	spa->spa_load_verify_bytes -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);

	if (ioerr && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
	char blkbuf[BP_SPRINTF_LEN];

	zcb->zcb_haderrors = 1;
	zcb->zcb_errors[ioerr]++;

	if (dump_opt['b'] >= 2)
	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	else
	blkbuf[0] = '\0';

	(void) printf("zdb_blkptr_cb: "
	"Got error %d reading "
	"<%llu, %llu, %lld, %llx> %s -- skipping\n",
	ioerr,
	(u_longlong_t)zb->zb_objset,
	(u_longlong_t)zb->zb_object,
	(u_longlong_t)zb->zb_level,
	(u_longlong_t)zb->zb_blkid,
	blkbuf);
	}
	mutex_exit(&spa->spa_scrub_lock);

	abd_free(zio->io_abd);
	}

	static int
	zdb_blkptr_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	zdb_cb_t *zcb = arg;
	dmu_object_type_t type;
	boolean_t is_metadata;

	if (zb->zb_level == ZB_DNODE_LEVEL)
	return (0);

	if (dump_opt['b'] >= 5 && bp->blk_birth > 0) {
	char blkbuf[BP_SPRINTF_LEN];
	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	(void) printf("objset %llu object %llu "
	"level %lld offset 0x%llx %s\n",
	(u_longlong_t)zb->zb_objset,
	(u_longlong_t)zb->zb_object,
	(longlong_t)zb->zb_level,
	(u_longlong_t)blkid2offset(dnp, bp, zb),
	blkbuf);
	}

	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp))
	return (0);

	type = BP_GET_TYPE(bp);

	zdb_count_block(zcb, zilog, bp,
	(type & DMU_OT_NEWTYPE) ? ZDB_OT_OTHER : type);

	is_metadata = (BP_GET_LEVEL(bp) != 0 \|\| DMU_OT_IS_METADATA(type));

	if (!BP_IS_EMBEDDED(bp) &&
	(dump_opt['c'] > 1 \|\| (dump_opt['c'] && is_metadata))) {
	size_t size = BP_GET_PSIZE(bp);
	abd_t *abd = abd_alloc(size, B_FALSE);
	int flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SCRUB \| ZIO_FLAG_RAW;

	/* If it's an intent log block, failure is expected. */
	if (zb->zb_level == ZB_ZIL_LEVEL)
	flags \|= ZIO_FLAG_SPECULATIVE;

	mutex_enter(&spa->spa_scrub_lock);
	while (spa->spa_load_verify_bytes > max_inflight_bytes)
	cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
	spa->spa_load_verify_bytes += size;
	mutex_exit(&spa->spa_scrub_lock);

	zio_nowait(zio_read(NULL, spa, bp, abd, size,
	zdb_blkptr_done, zcb, ZIO_PRIORITY_ASYNC_READ, flags, zb));
	}

	zcb->zcb_readfails = 0;

	/* only call gethrtime() every 100 blocks */
	static int iters;
	if (++iters > 100)
	iters = 0;
	else
	return (0);

	if (dump_opt['b'] < 5 && gethrtime() > zcb->zcb_lastprint + NANOSEC) {
	uint64_t now = gethrtime();
	char buf[10];
	uint64_t bytes = zcb->zcb_type[ZB_TOTAL][ZDB_OT_TOTAL].zb_asize;
	uint64_t kb_per_sec =
	1 + bytes / (1 + ((now - zcb->zcb_start) / 1000 / 1000));
	uint64_t sec_remaining =
	(zcb->zcb_totalasize - bytes) / 1024 / kb_per_sec;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (buf) >= NN_NUMBUF_SZ, "buf truncated");

	zfs_nicebytes(bytes, buf, sizeof (buf));
	(void) fprintf(stderr,
	"\r%5s completed (%4"PRIu64"MB/s) "
	"estimated time remaining: "
	"%"PRIu64"hr %02"PRIu64"min %02"PRIu64"sec ",
	buf, kb_per_sec / 1024,
	sec_remaining / 60 / 60,
	sec_remaining / 60 % 60,
	sec_remaining % 60);

	zcb->zcb_lastprint = now;
	}

	return (0);
	}

	static void
	zdb_leak(void *arg, uint64_t start, uint64_t size)
	{
	vdev_t *vd = arg;

	(void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size);
	}

	static metaslab_ops_t zdb_metaslab_ops = {
	NULL /* alloc */
	};

	static int
	load_unflushed_svr_segs_cb(spa_t spa, space_map_entry_t sme,
	uint64_t txg, void *arg)
	{
	spa_vdev_removal_t *svr = arg;

	uint64_t offset = sme->sme_offset;
	uint64_t size = sme->sme_run;

	/* skip vdevs we don't care about */
	if (sme->sme_vdev != svr->svr_vdev_id)
	return (0);

	vdev_t *vd = vdev_lookup_top(spa, sme->sme_vdev);
	metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];
	ASSERT(sme->sme_type == SM_ALLOC \|\| sme->sme_type == SM_FREE);

	if (txg < metaslab_unflushed_txg(ms))
	return (0);

	if (sme->sme_type == SM_ALLOC)
	range_tree_add(svr->svr_allocd_segs, offset, size);
	else
	range_tree_remove(svr->svr_allocd_segs, offset, size);

	return (0);
	}

	static void
	claim_segment_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner_offset, (void) arg;

	/*
	* This callback was called through a remap from
	* a device being removed. Therefore, the vdev that
	* this callback is applied to is a concrete
	* vdev.
	*/
	ASSERT(vdev_is_concrete(vd));

	VERIFY0(metaslab_claim_impl(vd, offset, size,
	spa_min_claim_txg(vd->vdev_spa)));
	}

	static void
	claim_segment_cb(void *arg, uint64_t offset, uint64_t size)
	{
	vdev_t *vd = arg;

	vdev_indirect_ops.vdev_op_remap(vd, offset, size,
	claim_segment_impl_cb, NULL);
	}

	/*
	* After accounting for all allocated blocks that are directly referenced,
	* we might have missed a reference to a block from a partially complete
	* (and thus unused) indirect mapping object. We perform a secondary pass
	* through the metaslabs we have already mapped and claim the destination
	* blocks.
	*/
	static void
	zdb_claim_removing(spa_t spa, zdb_cb_t zcb)
	{
	if (dump_opt['L'])
	return;

	if (spa->spa_vdev_removal == NULL)
	return;

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	ASSERT0(range_tree_space(svr->svr_allocd_segs));

	range_tree_t *allocs = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	for (uint64_t msi = 0; msi < vd->vdev_ms_count; msi++) {
	metaslab_t *msp = vd->vdev_ms[msi];

	ASSERT0(range_tree_space(allocs));
	if (msp->ms_sm != NULL)
	VERIFY0(space_map_load(msp->ms_sm, allocs, SM_ALLOC));
	range_tree_vacate(allocs, range_tree_add, svr->svr_allocd_segs);
	}
	range_tree_destroy(allocs);

	iterate_through_spacemap_logs(spa, load_unflushed_svr_segs_cb, svr);

	/*
	* Clear everything past what has been synced,
	* because we have not allocated mappings for
	* it yet.
	*/
	range_tree_clear(svr->svr_allocd_segs,
	vdev_indirect_mapping_max_offset(vim),
	vd->vdev_asize - vdev_indirect_mapping_max_offset(vim));

	zcb->zcb_removing_size += range_tree_space(svr->svr_allocd_segs);
	range_tree_vacate(svr->svr_allocd_segs, claim_segment_cb, vd);

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	static int
	increment_indirect_mapping_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	(void) tx;
	zdb_cb_t *zcb = arg;
	spa_t *spa = zcb->zcb_spa;
	vdev_t *vd;
	const dva_t *dva = &bp->blk_dva[0];

	ASSERT(!bp_freed);
	ASSERT(!dump_opt['L']);
	ASSERT3U(BP_GET_NDVAS(bp), ==, 1);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	vd = vdev_lookup_top(zcb->zcb_spa, DVA_GET_VDEV(dva));
	ASSERT3P(vd, !=, NULL);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	ASSERT(vd->vdev_indirect_config.vic_mapping_object != 0);
	ASSERT3P(zcb->zcb_vd_obsolete_counts[vd->vdev_id], !=, NULL);

	vdev_indirect_mapping_increment_obsolete_count(
	vd->vdev_indirect_mapping,
	DVA_GET_OFFSET(dva), DVA_GET_ASIZE(dva),
	zcb->zcb_vd_obsolete_counts[vd->vdev_id]);

	return (0);
	}

	static uint32_t *
	zdb_load_obsolete_counts(vdev_t *vd)
	{
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	spa_t *spa = vd->vdev_spa;
	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	uint64_t obsolete_sm_object;
	uint32_t *counts;

	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	EQUIV(obsolete_sm_object != 0, vd->vdev_obsolete_sm != NULL);
	counts = vdev_indirect_mapping_load_obsolete_counts(vim);
	if (vd->vdev_obsolete_sm != NULL) {
	vdev_indirect_mapping_load_obsolete_spacemap(vim, counts,
	vd->vdev_obsolete_sm);
	}
	if (scip->scip_vdev == vd->vdev_id &&
	scip->scip_prev_obsolete_sm_object != 0) {
	space_map_t *prev_obsolete_sm = NULL;
	VERIFY0(space_map_open(&prev_obsolete_sm, spa->spa_meta_objset,
	scip->scip_prev_obsolete_sm_object, 0, vd->vdev_asize, 0));
	vdev_indirect_mapping_load_obsolete_spacemap(vim, counts,
	prev_obsolete_sm);
	space_map_close(prev_obsolete_sm);
	}
	return (counts);
	}

	static void
	zdb_ddt_leak_init(spa_t spa, zdb_cb_t zcb)
	{
	ddt_bookmark_t ddb = {0};
	ddt_entry_t dde;
	int error;
	int p;

	ASSERT(!dump_opt['L']);

	while ((error = ddt_walk(spa, &ddb, &dde)) == 0) {
	blkptr_t blk;
	ddt_phys_t *ddp = dde.dde_phys;

	if (ddb.ddb_class == DDT_CLASS_UNIQUE)
	return;

	ASSERT(ddt_phys_total_refcnt(&dde) > 1);

	for (p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0)
	continue;
	ddt_bp_create(ddb.ddb_checksum,
	&dde.dde_key, ddp, &blk);
	if (p == DDT_PHYS_DITTO) {
	zdb_count_block(zcb, NULL, &blk, ZDB_OT_DITTO);
	} else {
	zcb->zcb_dedup_asize +=
	BP_GET_ASIZE(&blk) * (ddp->ddp_refcnt - 1);
	zcb->zcb_dedup_blocks++;
	}
	}
	ddt_t *ddt = spa->spa_ddt[ddb.ddb_checksum];
	ddt_enter(ddt);
	VERIFY(ddt_lookup(ddt, &blk, B_TRUE) != NULL);
	ddt_exit(ddt);
	}

	ASSERT(error == ENOENT);
	}

	typedef struct checkpoint_sm_exclude_entry_arg {
	vdev_t *cseea_vd;
	uint64_t cseea_checkpoint_size;
	} checkpoint_sm_exclude_entry_arg_t;

	static int
	checkpoint_sm_exclude_entry_cb(space_map_entry_t sme, void arg)
	{
	checkpoint_sm_exclude_entry_arg_t *cseea = arg;
	vdev_t *vd = cseea->cseea_vd;
	metaslab_t *ms = vd->vdev_ms[sme->sme_offset >> vd->vdev_ms_shift];
	uint64_t end = sme->sme_offset + sme->sme_run;

	ASSERT(sme->sme_type == SM_FREE);

	/*
	* Since the vdev_checkpoint_sm exists in the vdev level
	* and the ms_sm space maps exist in the metaslab level,
	* an entry in the checkpoint space map could theoretically
	* cross the boundaries of the metaslab that it belongs.
	*
	* In reality, because of the way that we populate and
	* manipulate the checkpoint's space maps currently,
	* there shouldn't be any entries that cross metaslabs.
	* Hence the assertion below.
	*
	* That said, there is no fundamental requirement that
	* the checkpoint's space map entries should not cross
	* metaslab boundaries. So if needed we could add code
	* that handles metaslab-crossing segments in the future.
	*/
	VERIFY3U(sme->sme_offset, >=, ms->ms_start);
	VERIFY3U(end, <=, ms->ms_start + ms->ms_size);

	/*
	* By removing the entry from the allocated segments we
	* also verify that the entry is there to begin with.
	*/
	mutex_enter(&ms->ms_lock);
	range_tree_remove(ms->ms_allocatable, sme->sme_offset, sme->sme_run);
	mutex_exit(&ms->ms_lock);

	cseea->cseea_checkpoint_size += sme->sme_run;
	return (0);
	}

	static void
	zdb_leak_init_vdev_exclude_checkpoint(vdev_t vd, zdb_cb_t zcb)
	{
	spa_t *spa = vd->vdev_spa;
	space_map_t *checkpoint_sm = NULL;
	uint64_t checkpoint_sm_obj;

	/*
	* If there is no vdev_top_zap, we are in a pool whose
	* version predates the pool checkpoint feature.
	*/
	if (vd->vdev_top_zap == 0)
	return;

	/*
	* If there is no reference of the vdev_checkpoint_sm in
	* the vdev_top_zap, then one of the following scenarios
	* is true:
	*
	* 1] There is no checkpoint
	* 2] There is a checkpoint, but no checkpointed blocks
	* have been freed yet
	* 3] The current vdev is indirect
	*
	* In these cases we return immediately.
	*/
	if (zap_contains(spa_meta_objset(spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) != 0)
	return;

	VERIFY0(zap_lookup(spa_meta_objset(spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM, sizeof (uint64_t), 1,
	&checkpoint_sm_obj));

	checkpoint_sm_exclude_entry_arg_t cseea;
	cseea.cseea_vd = vd;
	cseea.cseea_checkpoint_size = 0;

	VERIFY0(space_map_open(&checkpoint_sm, spa_meta_objset(spa),
	checkpoint_sm_obj, 0, vd->vdev_asize, vd->vdev_ashift));

	VERIFY0(space_map_iterate(checkpoint_sm,
	space_map_length(checkpoint_sm),
	checkpoint_sm_exclude_entry_cb, &cseea));
	space_map_close(checkpoint_sm);

	zcb->zcb_checkpoint_size += cseea.cseea_checkpoint_size;
	}

	static void
	zdb_leak_init_exclude_checkpoint(spa_t spa, zdb_cb_t zcb)
	{
	ASSERT(!dump_opt['L']);

	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	ASSERT3U(c, ==, rvd->vdev_child[c]->vdev_id);
	zdb_leak_init_vdev_exclude_checkpoint(rvd->vdev_child[c], zcb);
	}
	}

	static int
	count_unflushed_space_cb(spa_t spa, space_map_entry_t sme,
	uint64_t txg, void *arg)
	{
	int64_t *ualloc_space = arg;

	uint64_t offset = sme->sme_offset;
	uint64_t vdev_id = sme->sme_vdev;

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	if (!vdev_is_concrete(vd))
	return (0);

	metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];
	ASSERT(sme->sme_type == SM_ALLOC \|\| sme->sme_type == SM_FREE);

	if (txg < metaslab_unflushed_txg(ms))
	return (0);

	if (sme->sme_type == SM_ALLOC)
	*ualloc_space += sme->sme_run;
	else
	*ualloc_space -= sme->sme_run;

	return (0);
	}

	static int64_t
	get_unflushed_alloc_space(spa_t *spa)
	{
	if (dump_opt['L'])
	return (0);

	int64_t ualloc_space = 0;
	iterate_through_spacemap_logs(spa, count_unflushed_space_cb,
	&ualloc_space);
	return (ualloc_space);
	}

	static int
	load_unflushed_cb(spa_t spa, space_map_entry_t sme, uint64_t txg, void *arg)
	{
	maptype_t *uic_maptype = arg;

	uint64_t offset = sme->sme_offset;
	uint64_t size = sme->sme_run;
	uint64_t vdev_id = sme->sme_vdev;

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);

	/* skip indirect vdevs */
	if (!vdev_is_concrete(vd))
	return (0);

	metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	ASSERT(sme->sme_type == SM_ALLOC \|\| sme->sme_type == SM_FREE);
	ASSERT(uic_maptype == SM_ALLOC \|\| uic_maptype == SM_FREE);

	if (txg < metaslab_unflushed_txg(ms))
	return (0);

	if (*uic_maptype == sme->sme_type)
	range_tree_add(ms->ms_allocatable, offset, size);
	else
	range_tree_remove(ms->ms_allocatable, offset, size);

	return (0);
	}

	static void
	load_unflushed_to_ms_allocatables(spa_t *spa, maptype_t maptype)
	{
	iterate_through_spacemap_logs(spa, load_unflushed_cb, &maptype);
	}

	static void
	load_concrete_ms_allocatable_trees(spa_t *spa, maptype_t maptype)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];

	ASSERT3U(i, ==, vd->vdev_id);

	if (vd->vdev_ops == &vdev_indirect_ops)
	continue;

	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
	metaslab_t *msp = vd->vdev_ms[m];

	(void) fprintf(stderr,
	"\rloading concrete vdev %llu, "
	"metaslab %llu of %llu ...",
	(longlong_t)vd->vdev_id,
	(longlong_t)msp->ms_id,
	(longlong_t)vd->vdev_ms_count);

	mutex_enter(&msp->ms_lock);
	range_tree_vacate(msp->ms_allocatable, NULL, NULL);

	/*
	* We don't want to spend the CPU manipulating the
	* size-ordered tree, so clear the range_tree ops.
	*/
	msp->ms_allocatable->rt_ops = NULL;

	if (msp->ms_sm != NULL) {
	VERIFY0(space_map_load(msp->ms_sm,
	msp->ms_allocatable, maptype));
	}
	if (!msp->ms_loaded)
	msp->ms_loaded = B_TRUE;
	mutex_exit(&msp->ms_lock);
	}
	}

	load_unflushed_to_ms_allocatables(spa, maptype);
	}

	/*
	* vm_idxp is an in-out parameter which (for indirect vdevs) is the
	* index in vim_entries that has the first entry in this metaslab.
	* On return, it will be set to the first entry after this metaslab.
	*/
	static void
	load_indirect_ms_allocatable_tree(vdev_t vd, metaslab_t msp,
	uint64_t *vim_idxp)
	{
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	mutex_enter(&msp->ms_lock);
	range_tree_vacate(msp->ms_allocatable, NULL, NULL);

	/*
	* We don't want to spend the CPU manipulating the
	* size-ordered tree, so clear the range_tree ops.
	*/
	msp->ms_allocatable->rt_ops = NULL;

	for (; *vim_idxp < vdev_indirect_mapping_num_entries(vim);
	(*vim_idxp)++) {
	vdev_indirect_mapping_entry_phys_t *vimep =
	&vim->vim_entries[*vim_idxp];
	uint64_t ent_offset = DVA_MAPPING_GET_SRC_OFFSET(vimep);
	uint64_t ent_len = DVA_GET_ASIZE(&vimep->vimep_dst);
	ASSERT3U(ent_offset, >=, msp->ms_start);
	if (ent_offset >= msp->ms_start + msp->ms_size)
	break;

	/*
	* Mappings do not cross metaslab boundaries,
	* because we create them by walking the metaslabs.
	*/
	ASSERT3U(ent_offset + ent_len, <=,
	msp->ms_start + msp->ms_size);
	range_tree_add(msp->ms_allocatable, ent_offset, ent_len);
	}

	if (!msp->ms_loaded)
	msp->ms_loaded = B_TRUE;
	mutex_exit(&msp->ms_lock);
	}

	static void
	zdb_leak_init_prepare_indirect_vdevs(spa_t spa, zdb_cb_t zcb)
	{
	ASSERT(!dump_opt['L']);

	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	ASSERT3U(c, ==, vd->vdev_id);

	if (vd->vdev_ops != &vdev_indirect_ops)
	continue;

	/*
	* Note: we don't check for mapping leaks on
	* removing vdevs because their ms_allocatable's
	* are used to look for leaks in allocated space.
	*/
	zcb->zcb_vd_obsolete_counts[c] = zdb_load_obsolete_counts(vd);

	/*
	* Normally, indirect vdevs don't have any
	* metaslabs. We want to set them up for
	* zio_claim().
	*/
	vdev_metaslab_group_create(vd);
	VERIFY0(vdev_metaslab_init(vd, 0));

	vdev_indirect_mapping_t *vim __maybe_unused =
	vd->vdev_indirect_mapping;
	uint64_t vim_idx = 0;
	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {

	(void) fprintf(stderr,
	"\rloading indirect vdev %llu, "
	"metaslab %llu of %llu ...",
	(longlong_t)vd->vdev_id,
	(longlong_t)vd->vdev_ms[m]->ms_id,
	(longlong_t)vd->vdev_ms_count);

	load_indirect_ms_allocatable_tree(vd, vd->vdev_ms[m],
	&vim_idx);
	}
	ASSERT3U(vim_idx, ==, vdev_indirect_mapping_num_entries(vim));
	}
	}

	static void
	zdb_leak_init(spa_t spa, zdb_cb_t zcb)
	{
	zcb->zcb_spa = spa;

	if (dump_opt['L'])
	return;

	dsl_pool_t *dp = spa->spa_dsl_pool;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* We are going to be changing the meaning of the metaslab's
	* ms_allocatable. Ensure that the allocator doesn't try to
	* use the tree.
	*/
	spa->spa_normal_class->mc_ops = &zdb_metaslab_ops;
	spa->spa_log_class->mc_ops = &zdb_metaslab_ops;
	spa->spa_embedded_log_class->mc_ops = &zdb_metaslab_ops;

	zcb->zcb_vd_obsolete_counts =
	umem_zalloc(rvd->vdev_children * sizeof (uint32_t *),
	UMEM_NOFAIL);

	/*
	* For leak detection, we overload the ms_allocatable trees
	* to contain allocated segments instead of free segments.
	* As a result, we can't use the normal metaslab_load/unload
	* interfaces.
	*/
	zdb_leak_init_prepare_indirect_vdevs(spa, zcb);
	load_concrete_ms_allocatable_trees(spa, SM_ALLOC);

	/*
	* On load_concrete_ms_allocatable_trees() we loaded all the
	* allocated entries from the ms_sm to the ms_allocatable for
	* each metaslab. If the pool has a checkpoint or is in the
	* middle of discarding a checkpoint, some of these blocks
	* may have been freed but their ms_sm may not have been
	* updated because they are referenced by the checkpoint. In
	* order to avoid false-positives during leak-detection, we
	* go through the vdev's checkpoint space map and exclude all
	* its entries from their relevant ms_allocatable.
	*
	* We also aggregate the space held by the checkpoint and add
	* it to zcb_checkpoint_size.
	*
	* Note that at this point we are also verifying that all the
	* entries on the checkpoint_sm are marked as allocated in
	* the ms_sm of their relevant metaslab.
	* [see comment in checkpoint_sm_exclude_entry_cb()]
	*/
	zdb_leak_init_exclude_checkpoint(spa, zcb);
	ASSERT3U(zcb->zcb_checkpoint_size, ==, spa_get_checkpoint_space(spa));

	/* for cleaner progress output */
	(void) fprintf(stderr, "\n");

	if (bpobj_is_open(&dp->dp_obsolete_bpobj)) {
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_DEVICE_REMOVAL));
	(void) bpobj_iterate_nofree(&dp->dp_obsolete_bpobj,
	increment_indirect_mapping_cb, zcb, NULL);
	}

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	zdb_ddt_leak_init(spa, zcb);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	static boolean_t
	zdb_check_for_obsolete_leaks(vdev_t vd, zdb_cb_t zcb)
	{
	boolean_t leaks = B_FALSE;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	uint64_t total_leaked = 0;
	boolean_t are_precise = B_FALSE;

	ASSERT(vim != NULL);

	for (uint64_t i = 0; i < vdev_indirect_mapping_num_entries(vim); i++) {
	vdev_indirect_mapping_entry_phys_t *vimep =
	&vim->vim_entries[i];
	uint64_t obsolete_bytes = 0;
	uint64_t offset = DVA_MAPPING_GET_SRC_OFFSET(vimep);
	metaslab_t *msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	/*
	* This is not very efficient but it's easy to
	* verify correctness.
	*/
	for (uint64_t inner_offset = 0;
	inner_offset < DVA_GET_ASIZE(&vimep->vimep_dst);
	- inner_offset += 1 << vd->vdev_ashift) {
	+ inner_offset += 1ULL << vd->vdev_ashift) {
	if (range_tree_contains(msp->ms_allocatable,
	- offset + inner_offset, 1 << vd->vdev_ashift)) {
	- obsolete_bytes += 1 << vd->vdev_ashift;
	+ offset + inner_offset, 1ULL << vd->vdev_ashift)) {
	+ obsolete_bytes += 1ULL << vd->vdev_ashift;
	}
	}

	int64_t bytes_leaked = obsolete_bytes -
	zcb->zcb_vd_obsolete_counts[vd->vdev_id][i];
	ASSERT3U(DVA_GET_ASIZE(&vimep->vimep_dst), >=,
	zcb->zcb_vd_obsolete_counts[vd->vdev_id][i]);

	VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	if (bytes_leaked != 0 && (are_precise \|\| dump_opt['d'] >= 5)) {
	(void) printf("obsolete indirect mapping count "
	"mismatch on %llu:%llx:%llx : %llx bytes leaked\n",
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)DVA_MAPPING_GET_SRC_OFFSET(vimep),
	(u_longlong_t)DVA_GET_ASIZE(&vimep->vimep_dst),
	(u_longlong_t)bytes_leaked);
	}
	total_leaked += ABS(bytes_leaked);
	}

	VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	if (!are_precise && total_leaked > 0) {
	int pct_leaked = total_leaked * 100 /
	vdev_indirect_mapping_bytes_mapped(vim);
	(void) printf("cannot verify obsolete indirect mapping "
	"counts of vdev %llu because precise feature was not "
	"enabled when it was removed: %d%% (%llx bytes) of mapping"
	"unreferenced\n",
	(u_longlong_t)vd->vdev_id, pct_leaked,
	(u_longlong_t)total_leaked);
	} else if (total_leaked > 0) {
	(void) printf("obsolete indirect mapping count mismatch "
	"for vdev %llu -- %llx total bytes mismatched\n",
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)total_leaked);
	leaks \|= B_TRUE;
	}

	vdev_indirect_mapping_free_obsolete_counts(vim,
	zcb->zcb_vd_obsolete_counts[vd->vdev_id]);
	zcb->zcb_vd_obsolete_counts[vd->vdev_id] = NULL;

	return (leaks);
	}

	static boolean_t
	zdb_leak_fini(spa_t spa, zdb_cb_t zcb)
	{
	if (dump_opt['L'])
	return (B_FALSE);

	boolean_t leaks = B_FALSE;
	vdev_t *rvd = spa->spa_root_vdev;
	for (unsigned c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	if (zcb->zcb_vd_obsolete_counts[c] != NULL) {
	leaks \|= zdb_check_for_obsolete_leaks(vd, zcb);
	}

	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
	metaslab_t *msp = vd->vdev_ms[m];
	ASSERT3P(msp->ms_group, ==, (msp->ms_group->mg_class ==
	spa_embedded_log_class(spa)) ?
	vd->vdev_log_mg : vd->vdev_mg);

	/*
	* ms_allocatable has been overloaded
	* to contain allocated segments. Now that
	* we finished traversing all blocks, any
	* block that remains in the ms_allocatable
	* represents an allocated block that we
	* did not claim during the traversal.
	* Claimed blocks would have been removed
	* from the ms_allocatable. For indirect
	* vdevs, space remaining in the tree
	* represents parts of the mapping that are
	* not referenced, which is not a bug.
	*/
	if (vd->vdev_ops == &vdev_indirect_ops) {
	range_tree_vacate(msp->ms_allocatable,
	NULL, NULL);
	} else {
	range_tree_vacate(msp->ms_allocatable,
	zdb_leak, vd);
	}
	if (msp->ms_loaded) {
	msp->ms_loaded = B_FALSE;
	}
	}
	}

	umem_free(zcb->zcb_vd_obsolete_counts,
	rvd->vdev_children * sizeof (uint32_t *));
	zcb->zcb_vd_obsolete_counts = NULL;

	return (leaks);
	}

	static int
	count_block_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	(void) tx;
	zdb_cb_t *zcb = arg;

	if (dump_opt['b'] >= 5) {
	char blkbuf[BP_SPRINTF_LEN];
	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	(void) printf("[%s] %s\n",
	"deferred free", blkbuf);
	}
	zdb_count_block(zcb, NULL, bp, ZDB_OT_DEFERRED);
	return (0);
	}

	/*
	* Iterate over livelists which have been destroyed by the user but
	* are still present in the MOS, waiting to be freed
	*/
	static void
	iterate_deleted_livelists(spa_t spa, ll_iter_t func, void arg)
	{
	objset_t *mos = spa->spa_meta_objset;
	uint64_t zap_obj;
	int err = zap_lookup(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_DELETED_CLONES, sizeof (uint64_t), 1, &zap_obj);
	if (err == ENOENT)
	return;
	ASSERT0(err);

	zap_cursor_t zc;
	zap_attribute_t attr;
	dsl_deadlist_t ll;
	/* NULL out os prior to dsl_deadlist_open in case it's garbage */
	ll.dl_os = NULL;
	for (zap_cursor_init(&zc, mos, zap_obj);
	zap_cursor_retrieve(&zc, &attr) == 0;
	(void) zap_cursor_advance(&zc)) {
	dsl_deadlist_open(&ll, mos, attr.za_first_integer);
	func(&ll, arg);
	dsl_deadlist_close(&ll);
	}
	zap_cursor_fini(&zc);
	}

	static int
	bpobj_count_block_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(!bp_freed);
	return (count_block_cb(arg, bp, tx));
	}

	static int
	livelist_entry_count_blocks_cb(void args, dsl_deadlist_entry_t dle)
	{
	zdb_cb_t *zbc = args;
	bplist_t blks;
	bplist_create(&blks);
	/* determine which blocks have been alloc'd but not freed */
	VERIFY0(dsl_process_sub_livelist(&dle->dle_bpobj, &blks, NULL, NULL));
	/* count those blocks */
	(void) bplist_iterate(&blks, count_block_cb, zbc, NULL);
	bplist_destroy(&blks);
	return (0);
	}

	static void
	livelist_count_blocks(dsl_deadlist_t ll, void arg)
	{
	dsl_deadlist_iterate(ll, livelist_entry_count_blocks_cb, arg);
	}

	/*
	* Count the blocks in the livelists that have been destroyed by the user
	* but haven't yet been freed.
	*/
	static void
	deleted_livelists_count_blocks(spa_t spa, zdb_cb_t zbc)
	{
	iterate_deleted_livelists(spa, livelist_count_blocks, zbc);
	}

	static void
	dump_livelist_cb(dsl_deadlist_t ll, void arg)
	{
	ASSERT3P(arg, ==, NULL);
	global_feature_count[SPA_FEATURE_LIVELIST]++;
	dump_blkptr_list(ll, "Deleted Livelist");
	dsl_deadlist_iterate(ll, sublivelist_verify_lightweight, NULL);
	}

	/*
	* Print out, register object references to, and increment feature counts for
	* livelists that have been destroyed by the user but haven't yet been freed.
	*/
	static void
	deleted_livelists_dump_mos(spa_t *spa)
	{
	uint64_t zap_obj;
	objset_t *mos = spa->spa_meta_objset;
	int err = zap_lookup(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_DELETED_CLONES, sizeof (uint64_t), 1, &zap_obj);
	if (err == ENOENT)
	return;
	mos_obj_refd(zap_obj);
	iterate_deleted_livelists(spa, dump_livelist_cb, NULL);
	}

	static int
	dump_block_stats(spa_t *spa)
	{
	zdb_cb_t *zcb;
	zdb_blkstats_t zb, tzb;
	uint64_t norm_alloc, norm_space, total_alloc, total_found;
	int flags = TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \|
	TRAVERSE_NO_DECRYPT \| TRAVERSE_HARD;
	boolean_t leaks = B_FALSE;
	int e, c, err;
	bp_embedded_type_t i;

	zcb = umem_zalloc(sizeof (zdb_cb_t), UMEM_NOFAIL);

	(void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
	(dump_opt['c'] \|\| !dump_opt['L']) ? "to verify " : "",
	(dump_opt['c'] == 1) ? "metadata " : "",
	dump_opt['c'] ? "checksums " : "",
	(dump_opt['c'] && !dump_opt['L']) ? "and verify " : "",
	!dump_opt['L'] ? "nothing leaked " : "");

	/*
	* When leak detection is enabled we load all space maps as SM_ALLOC
	* maps, then traverse the pool claiming each block we discover. If
	* the pool is perfectly consistent, the segment trees will be empty
	* when we're done. Anything left over is a leak; any block we can't
	* claim (because it's not part of any space map) is a double
	* allocation, reference to a freed block, or an unclaimed log block.
	*
	* When leak detection is disabled (-L option) we still traverse the
	* pool claiming each block we discover, but we skip opening any space
	* maps.
	*/
	zdb_leak_init(spa, zcb);

	/*
	* If there's a deferred-free bplist, process that first.
	*/
	(void) bpobj_iterate_nofree(&spa->spa_deferred_bpobj,
	bpobj_count_block_cb, zcb, NULL);

	if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
	(void) bpobj_iterate_nofree(&spa->spa_dsl_pool->dp_free_bpobj,
	bpobj_count_block_cb, zcb, NULL);
	}

	zdb_claim_removing(spa, zcb);

	if (spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY)) {
	VERIFY3U(0, ==, bptree_iterate(spa->spa_meta_objset,
	spa->spa_dsl_pool->dp_bptree_obj, B_FALSE, count_block_cb,
	zcb, NULL));
	}

	deleted_livelists_count_blocks(spa, zcb);

	if (dump_opt['c'] > 1)
	flags \|= TRAVERSE_PREFETCH_DATA;

	zcb->zcb_totalasize = metaslab_class_get_alloc(spa_normal_class(spa));
	zcb->zcb_totalasize += metaslab_class_get_alloc(spa_special_class(spa));
	zcb->zcb_totalasize += metaslab_class_get_alloc(spa_dedup_class(spa));
	zcb->zcb_totalasize +=
	metaslab_class_get_alloc(spa_embedded_log_class(spa));
	zcb->zcb_start = zcb->zcb_lastprint = gethrtime();
	err = traverse_pool(spa, 0, flags, zdb_blkptr_cb, zcb);

	/*
	* If we've traversed the data blocks then we need to wait for those
	* I/Os to complete. We leverage "The Godfather" zio to wait on
	* all async I/Os to complete.
	*/
	if (dump_opt['c']) {
	for (c = 0; c < max_ncpus; c++) {
	(void) zio_wait(spa->spa_async_zio_root[c]);
	spa->spa_async_zio_root[c] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);
	}
	}
	ASSERT0(spa->spa_load_verify_bytes);

	/*
	* Done after zio_wait() since zcb_haderrors is modified in
	* zdb_blkptr_done()
	*/
	zcb->zcb_haderrors \|= err;

	if (zcb->zcb_haderrors) {
	(void) printf("\nError counts:\n\n");
	(void) printf("\t%5s %s\n", "errno", "count");
	for (e = 0; e < 256; e++) {
	if (zcb->zcb_errors[e] != 0) {
	(void) printf("\t%5d %llu\n",
	e, (u_longlong_t)zcb->zcb_errors[e]);
	}
	}
	}

	/*
	* Report any leaked segments.
	*/
	leaks \|= zdb_leak_fini(spa, zcb);

	tzb = &zcb->zcb_type[ZB_TOTAL][ZDB_OT_TOTAL];

	norm_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	norm_space = metaslab_class_get_space(spa_normal_class(spa));

	total_alloc = norm_alloc +
	metaslab_class_get_alloc(spa_log_class(spa)) +
	metaslab_class_get_alloc(spa_embedded_log_class(spa)) +
	metaslab_class_get_alloc(spa_special_class(spa)) +
	metaslab_class_get_alloc(spa_dedup_class(spa)) +
	get_unflushed_alloc_space(spa);
	total_found = tzb->zb_asize - zcb->zcb_dedup_asize +
	zcb->zcb_removing_size + zcb->zcb_checkpoint_size;

	if (total_found == total_alloc && !dump_opt['L']) {
	(void) printf("\n\tNo leaks (block sum matches space"
	" maps exactly)\n");
	} else if (!dump_opt['L']) {
	(void) printf("block traversal size %llu != alloc %llu "
	"(%s %lld)\n",
	(u_longlong_t)total_found,
	(u_longlong_t)total_alloc,
	(dump_opt['L']) ? "unreachable" : "leaked",
	(longlong_t)(total_alloc - total_found));
	leaks = B_TRUE;
	}

	if (tzb->zb_count == 0) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (2);
	}

	(void) printf("\n");
	(void) printf("\t%-16s %14llu\n", "bp count:",
	(u_longlong_t)tzb->zb_count);
	(void) printf("\t%-16s %14llu\n", "ganged count:",
	(longlong_t)tzb->zb_gangs);
	(void) printf("\t%-16s %14llu avg: %6llu\n", "bp logical:",
	(u_longlong_t)tzb->zb_lsize,
	(u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
	(void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
	"bp physical:", (u_longlong_t)tzb->zb_psize,
	(u_longlong_t)(tzb->zb_psize / tzb->zb_count),
	(double)tzb->zb_lsize / tzb->zb_psize);
	(void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
	"bp allocated:", (u_longlong_t)tzb->zb_asize,
	(u_longlong_t)(tzb->zb_asize / tzb->zb_count),
	(double)tzb->zb_lsize / tzb->zb_asize);
	(void) printf("\t%-16s %14llu ref>1: %6llu deduplication: %6.2f\n",
	"bp deduped:", (u_longlong_t)zcb->zcb_dedup_asize,
	(u_longlong_t)zcb->zcb_dedup_blocks,
	(double)zcb->zcb_dedup_asize / tzb->zb_asize + 1.0);
	(void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
	(u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);

	if (spa_special_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_special_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_special_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Special class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	if (spa_dedup_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_dedup_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_dedup_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Dedup class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	if (spa_embedded_log_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_embedded_log_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_embedded_log_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Embedded log class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	for (i = 0; i < NUM_BP_EMBEDDED_TYPES; i++) {
	if (zcb->zcb_embedded_blocks[i] == 0)
	continue;
	(void) printf("\n");
	(void) printf("\tadditional, non-pointer bps of type %u: "
	"%10llu\n",
	i, (u_longlong_t)zcb->zcb_embedded_blocks[i]);

	if (dump_opt['b'] >= 3) {
	(void) printf("\t number of (compressed) bytes: "
	"number of bps\n");
	dump_histogram(zcb->zcb_embedded_histogram[i],
	sizeof (zcb->zcb_embedded_histogram[i]) /
	sizeof (zcb->zcb_embedded_histogram[i][0]), 0);
	}
	}

	if (tzb->zb_ditto_samevdev != 0) {
	(void) printf("\tDittoed blocks on same vdev: %llu\n",
	(longlong_t)tzb->zb_ditto_samevdev);
	}
	if (tzb->zb_ditto_same_ms != 0) {
	(void) printf("\tDittoed blocks in same metaslab: %llu\n",
	(longlong_t)tzb->zb_ditto_same_ms);
	}

	for (uint64_t v = 0; v < spa->spa_root_vdev->vdev_children; v++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[v];
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	if (vim == NULL) {
	continue;
	}

	char mem[32];
	zdb_nicenum(vdev_indirect_mapping_num_entries(vim),
	mem, vdev_indirect_mapping_size(vim));

	(void) printf("\tindirect vdev id %llu has %llu segments "
	"(%s in memory)\n",
	(longlong_t)vd->vdev_id,
	(longlong_t)vdev_indirect_mapping_num_entries(vim), mem);
	}

	if (dump_opt['b'] >= 2) {
	int l, t, level;
	(void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
	"\t avg\t comp\t%%Total\tType\n");

	for (t = 0; t <= ZDB_OT_TOTAL; t++) {
	char csize[32], lsize[32], psize[32], asize[32];
	char avg[32], gang[32];
	const char *typename;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (csize) >= NN_NUMBUF_SZ,
	"csize truncated");
	_Static_assert(sizeof (lsize) >= NN_NUMBUF_SZ,
	"lsize truncated");
	_Static_assert(sizeof (psize) >= NN_NUMBUF_SZ,
	"psize truncated");
	_Static_assert(sizeof (asize) >= NN_NUMBUF_SZ,
	"asize truncated");
	_Static_assert(sizeof (avg) >= NN_NUMBUF_SZ,
	"avg truncated");
	_Static_assert(sizeof (gang) >= NN_NUMBUF_SZ,
	"gang truncated");

	if (t < DMU_OT_NUMTYPES)
	typename = dmu_ot[t].ot_name;
	else
	typename = zdb_ot_extname[t - DMU_OT_NUMTYPES];

	if (zcb->zcb_type[ZB_TOTAL][t].zb_asize == 0) {
	(void) printf("%6s\t%5s\t%5s\t%5s"
	"\t%5s\t%5s\t%6s\t%s\n",
	"-",
	"-",
	"-",
	"-",
	"-",
	"-",
	"-",
	typename);
	continue;
	}

	for (l = ZB_TOTAL - 1; l >= -1; l--) {
	level = (l == -1 ? ZB_TOTAL : l);
	zb = &zcb->zcb_type[level][t];

	if (zb->zb_asize == 0)
	continue;

	if (dump_opt['b'] < 3 && level != ZB_TOTAL)
	continue;

	if (level == 0 && zb->zb_asize ==
	zcb->zcb_type[ZB_TOTAL][t].zb_asize)
	continue;

	zdb_nicenum(zb->zb_count, csize,
	sizeof (csize));
	zdb_nicenum(zb->zb_lsize, lsize,
	sizeof (lsize));
	zdb_nicenum(zb->zb_psize, psize,
	sizeof (psize));
	zdb_nicenum(zb->zb_asize, asize,
	sizeof (asize));
	zdb_nicenum(zb->zb_asize / zb->zb_count, avg,
	sizeof (avg));
	zdb_nicenum(zb->zb_gangs, gang, sizeof (gang));

	(void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
	"\t%5.2f\t%6.2f\t",
	csize, lsize, psize, asize, avg,
	(double)zb->zb_lsize / zb->zb_psize,
	100.0 * zb->zb_asize / tzb->zb_asize);

	if (level == ZB_TOTAL)
	(void) printf("%s\n", typename);
	else
	(void) printf(" L%d %s\n",
	level, typename);

	if (dump_opt['b'] >= 3 && zb->zb_gangs > 0) {
	(void) printf("\t number of ganged "
	"blocks: %s\n", gang);
	}

	if (dump_opt['b'] >= 4) {
	(void) printf("psize "
	"(in 512-byte sectors): "
	"number of blocks\n");
	dump_histogram(zb->zb_psize_histogram,
	PSIZE_HISTO_SIZE, 0);
	}
	}
	}

	/* Output a table summarizing block sizes in the pool */
	if (dump_opt['b'] >= 2) {
	dump_size_histograms(zcb);
	}
	}

	(void) printf("\n");

	if (leaks) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (2);
	}

	if (zcb->zcb_haderrors) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (3);
	}

	umem_free(zcb, sizeof (zdb_cb_t));
	return (0);
	}

	typedef struct zdb_ddt_entry {
	ddt_key_t zdde_key;
	uint64_t zdde_ref_blocks;
	uint64_t zdde_ref_lsize;
	uint64_t zdde_ref_psize;
	uint64_t zdde_ref_dsize;
	avl_node_t zdde_node;
	} zdb_ddt_entry_t;

	static int
	zdb_ddt_add_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	(void) zilog, (void) dnp;
	avl_tree_t *t = arg;
	avl_index_t where;
	zdb_ddt_entry_t *zdde, zdde_search;

	if (zb->zb_level == ZB_DNODE_LEVEL \|\| BP_IS_HOLE(bp) \|\|
	BP_IS_EMBEDDED(bp))
	return (0);

	if (dump_opt['S'] > 1 && zb->zb_level == ZB_ROOT_LEVEL) {
	(void) printf("traversing objset %llu, %llu objects, "
	"%lu blocks so far\n",
	(u_longlong_t)zb->zb_objset,
	(u_longlong_t)BP_GET_FILL(bp),
	avl_numnodes(t));
	}

	if (BP_IS_HOLE(bp) \|\| BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_OFF \|\|
	BP_GET_LEVEL(bp) > 0 \|\| DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
	return (0);

	ddt_key_fill(&zdde_search.zdde_key, bp);

	zdde = avl_find(t, &zdde_search, &where);

	if (zdde == NULL) {
	zdde = umem_zalloc(sizeof (*zdde), UMEM_NOFAIL);
	zdde->zdde_key = zdde_search.zdde_key;
	avl_insert(t, zdde, where);
	}

	zdde->zdde_ref_blocks += 1;
	zdde->zdde_ref_lsize += BP_GET_LSIZE(bp);
	zdde->zdde_ref_psize += BP_GET_PSIZE(bp);
	zdde->zdde_ref_dsize += bp_get_dsize_sync(spa, bp);

	return (0);
	}

	static void
	dump_simulated_ddt(spa_t *spa)
	{
	avl_tree_t t;
	void *cookie = NULL;
	zdb_ddt_entry_t *zdde;
	ddt_histogram_t ddh_total = {{{0}}};
	ddt_stat_t dds_total = {0};

	avl_create(&t, ddt_entry_compare,
	sizeof (zdb_ddt_entry_t), offsetof(zdb_ddt_entry_t, zdde_node));

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	(void) traverse_pool(spa, 0, TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \|
	TRAVERSE_NO_DECRYPT, zdb_ddt_add_cb, &t);

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	while ((zdde = avl_destroy_nodes(&t, &cookie)) != NULL) {
	ddt_stat_t dds;
	uint64_t refcnt = zdde->zdde_ref_blocks;
	ASSERT(refcnt != 0);

	dds.dds_blocks = zdde->zdde_ref_blocks / refcnt;
	dds.dds_lsize = zdde->zdde_ref_lsize / refcnt;
	dds.dds_psize = zdde->zdde_ref_psize / refcnt;
	dds.dds_dsize = zdde->zdde_ref_dsize / refcnt;

	dds.dds_ref_blocks = zdde->zdde_ref_blocks;
	dds.dds_ref_lsize = zdde->zdde_ref_lsize;
	dds.dds_ref_psize = zdde->zdde_ref_psize;
	dds.dds_ref_dsize = zdde->zdde_ref_dsize;

	ddt_stat_add(&ddh_total.ddh_stat[highbit64(refcnt) - 1],
	&dds, 0);

	umem_free(zdde, sizeof (*zdde));
	}

	avl_destroy(&t);

	ddt_histogram_stat(&dds_total, &ddh_total);

	(void) printf("Simulated DDT histogram:\n");

	zpool_dump_ddt(&dds_total, &ddh_total);

	dump_dedup_ratio(&dds_total);
	}

	static int
	verify_device_removal_feature_counts(spa_t *spa)
	{
	uint64_t dr_feature_refcount = 0;
	uint64_t oc_feature_refcount = 0;
	uint64_t indirect_vdev_count = 0;
	uint64_t precise_vdev_count = 0;
	uint64_t obsolete_counts_object_count = 0;
	uint64_t obsolete_sm_count = 0;
	uint64_t obsolete_counts_count = 0;
	uint64_t scip_count = 0;
	uint64_t obsolete_bpobj_count = 0;
	int ret = 0;

	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	if (scip->scip_next_mapping_object != 0) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[scip->scip_vdev];
	ASSERT(scip->scip_prev_obsolete_sm_object != 0);
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	(void) printf("Condensing indirect vdev %llu: new mapping "
	"object %llu, prev obsolete sm %llu\n",
	(u_longlong_t)scip->scip_vdev,
	(u_longlong_t)scip->scip_next_mapping_object,
	(u_longlong_t)scip->scip_prev_obsolete_sm_object);
	if (scip->scip_prev_obsolete_sm_object != 0) {
	space_map_t *prev_obsolete_sm = NULL;
	VERIFY0(space_map_open(&prev_obsolete_sm,
	spa->spa_meta_objset,
	scip->scip_prev_obsolete_sm_object,
	0, vd->vdev_asize, 0));
	dump_spacemap(spa->spa_meta_objset, prev_obsolete_sm);
	(void) printf("\n");
	space_map_close(prev_obsolete_sm);
	}

	scip_count += 2;
	}

	for (uint64_t i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;

	if (vic->vic_mapping_object != 0) {
	ASSERT(vd->vdev_ops == &vdev_indirect_ops \|\|
	vd->vdev_removing);
	indirect_vdev_count++;

	if (vd->vdev_indirect_mapping->vim_havecounts) {
	obsolete_counts_count++;
	}
	}

	boolean_t are_precise;
	VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	if (are_precise) {
	ASSERT(vic->vic_mapping_object != 0);
	precise_vdev_count++;
	}

	uint64_t obsolete_sm_object;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	ASSERT(vic->vic_mapping_object != 0);
	obsolete_sm_count++;
	}
	}

	(void) feature_get_refcount(spa,
	&spa_feature_table[SPA_FEATURE_DEVICE_REMOVAL],
	&dr_feature_refcount);
	(void) feature_get_refcount(spa,
	&spa_feature_table[SPA_FEATURE_OBSOLETE_COUNTS],
	&oc_feature_refcount);

	if (dr_feature_refcount != indirect_vdev_count) {
	ret = 1;
	(void) printf("Number of indirect vdevs (%llu) " \
	"does not match feature count (%llu)\n",
	(u_longlong_t)indirect_vdev_count,
	(u_longlong_t)dr_feature_refcount);
	} else {
	(void) printf("Verified device_removal feature refcount " \
	"of %llu is correct\n",
	(u_longlong_t)dr_feature_refcount);
	}

	if (zap_contains(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ) == 0) {
	obsolete_bpobj_count++;
	}


	obsolete_counts_object_count = precise_vdev_count;
	obsolete_counts_object_count += obsolete_sm_count;
	obsolete_counts_object_count += obsolete_counts_count;
	obsolete_counts_object_count += scip_count;
	obsolete_counts_object_count += obsolete_bpobj_count;
	obsolete_counts_object_count += remap_deadlist_count;

	if (oc_feature_refcount != obsolete_counts_object_count) {
	ret = 1;
	(void) printf("Number of obsolete counts objects (%llu) " \
	"does not match feature count (%llu)\n",
	(u_longlong_t)obsolete_counts_object_count,
	(u_longlong_t)oc_feature_refcount);
	(void) printf("pv:%llu os:%llu oc:%llu sc:%llu "
	"ob:%llu rd:%llu\n",
	(u_longlong_t)precise_vdev_count,
	(u_longlong_t)obsolete_sm_count,
	(u_longlong_t)obsolete_counts_count,
	(u_longlong_t)scip_count,
	(u_longlong_t)obsolete_bpobj_count,
	(u_longlong_t)remap_deadlist_count);
	} else {
	(void) printf("Verified indirect_refcount feature refcount " \
	"of %llu is correct\n",
	(u_longlong_t)oc_feature_refcount);
	}
	return (ret);
	}

	static void
	zdb_set_skip_mmp(char *target)
	{
	spa_t *spa;

	/*
	* Disable the activity check to allow examination of
	* active pools.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(target)) != NULL) {
	spa->spa_import_flags \|= ZFS_IMPORT_SKIP_MMP;
	}
	mutex_exit(&spa_namespace_lock);
	}

	#define BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
	/*
	* Import the checkpointed state of the pool specified by the target
	* parameter as readonly. The function also accepts a pool config
	* as an optional parameter, else it attempts to infer the config by
	* the name of the target pool.
	*
	* Note that the checkpointed state's pool name will be the name of
	* the original pool with the above suffix appended to it. In addition,
	* if the target is not a pool name (e.g. a path to a dataset) then
	* the new_path parameter is populated with the updated path to
	* reflect the fact that we are looking into the checkpointed state.
	*
	* The function returns a newly-allocated copy of the name of the
	* pool containing the checkpointed state. When this copy is no
	* longer needed it should be freed with free(3C). Same thing
	* applies to the new_path parameter if allocated.
	*/
	static char *
	import_checkpointed_state(char target, nvlist_t cfg, char **new_path)
	{
	int error = 0;
	char poolname, bogus_name = NULL;
	boolean_t freecfg = B_FALSE;

	/* If the target is not a pool, the extract the pool name */
	char *path_start = strchr(target, '/');
	if (path_start != NULL) {
	size_t poolname_len = path_start - target;
	poolname = strndup(target, poolname_len);
	} else {
	poolname = target;
	}

	if (cfg == NULL) {
	zdb_set_skip_mmp(poolname);
	error = spa_get_stats(poolname, &cfg, NULL, 0);
	if (error != 0) {
	fatal("Tried to read config of pool \"%s\" but "
	"spa_get_stats() failed with error %d\n",
	poolname, error);
	}
	freecfg = B_TRUE;
	}

	- if (asprintf(&bogus_name, "%s%s", poolname, BOGUS_SUFFIX) == -1)
	+ if (asprintf(&bogus_name, "%s%s", poolname, BOGUS_SUFFIX) == -1) {
	+ if (target != poolname)
	+ free(poolname);
	return (NULL);
	+ }
	fnvlist_add_string(cfg, ZPOOL_CONFIG_POOL_NAME, bogus_name);

	error = spa_import(bogus_name, cfg, NULL,
	ZFS_IMPORT_MISSING_LOG \| ZFS_IMPORT_CHECKPOINT \|
	ZFS_IMPORT_SKIP_MMP);
	if (freecfg)
	nvlist_free(cfg);
	if (error != 0) {
	fatal("Tried to import pool \"%s\" but spa_import() failed "
	"with error %d\n", bogus_name, error);
	}

	if (new_path != NULL && path_start != NULL) {
	if (asprintf(new_path, "%s%s", bogus_name, path_start) == -1) {
	+ free(bogus_name);
	if (path_start != NULL)
	free(poolname);
	return (NULL);
	}
	}

	if (target != poolname)
	free(poolname);

	return (bogus_name);
	}

	typedef struct verify_checkpoint_sm_entry_cb_arg {
	vdev_t *vcsec_vd;

	/* the following fields are only used for printing progress */
	uint64_t vcsec_entryid;
	uint64_t vcsec_num_entries;
	} verify_checkpoint_sm_entry_cb_arg_t;

	#define ENTRIES_PER_PROGRESS_UPDATE 10000

	static int
	verify_checkpoint_sm_entry_cb(space_map_entry_t sme, void arg)
	{
	verify_checkpoint_sm_entry_cb_arg_t *vcsec = arg;
	vdev_t *vd = vcsec->vcsec_vd;
	metaslab_t *ms = vd->vdev_ms[sme->sme_offset >> vd->vdev_ms_shift];
	uint64_t end = sme->sme_offset + sme->sme_run;

	ASSERT(sme->sme_type == SM_FREE);

	if ((vcsec->vcsec_entryid % ENTRIES_PER_PROGRESS_UPDATE) == 0) {
	(void) fprintf(stderr,
	"\rverifying vdev %llu, space map entry %llu of %llu ...",
	(longlong_t)vd->vdev_id,
	(longlong_t)vcsec->vcsec_entryid,
	(longlong_t)vcsec->vcsec_num_entries);
	}
	vcsec->vcsec_entryid++;

	/*
	* See comment in checkpoint_sm_exclude_entry_cb()
	*/
	VERIFY3U(sme->sme_offset, >=, ms->ms_start);
	VERIFY3U(end, <=, ms->ms_start + ms->ms_size);

	/*
	* The entries in the vdev_checkpoint_sm should be marked as
	* allocated in the checkpointed state of the pool, therefore
	* their respective ms_allocateable trees should not contain them.
	*/
	mutex_enter(&ms->ms_lock);
	range_tree_verify_not_present(ms->ms_allocatable,
	sme->sme_offset, sme->sme_run);
	mutex_exit(&ms->ms_lock);

	return (0);
	}

	/*
	* Verify that all segments in the vdev_checkpoint_sm are allocated
	* according to the checkpoint's ms_sm (i.e. are not in the checkpoint's
	* ms_allocatable).
	*
	* Do so by comparing the checkpoint space maps (vdev_checkpoint_sm) of
	* each vdev in the current state of the pool to the metaslab space maps
	* (ms_sm) of the checkpointed state of the pool.
	*
	* Note that the function changes the state of the ms_allocatable
	* trees of the current spa_t. The entries of these ms_allocatable
	* trees are cleared out and then repopulated from with the free
	* entries of their respective ms_sm space maps.
	*/
	static void
	verify_checkpoint_vdev_spacemaps(spa_t checkpoint, spa_t current)
	{
	vdev_t *ckpoint_rvd = checkpoint->spa_root_vdev;
	vdev_t *current_rvd = current->spa_root_vdev;

	load_concrete_ms_allocatable_trees(checkpoint, SM_FREE);

	for (uint64_t c = 0; c < ckpoint_rvd->vdev_children; c++) {
	vdev_t *ckpoint_vd = ckpoint_rvd->vdev_child[c];
	vdev_t *current_vd = current_rvd->vdev_child[c];

	space_map_t *checkpoint_sm = NULL;
	uint64_t checkpoint_sm_obj;

	if (ckpoint_vd->vdev_ops == &vdev_indirect_ops) {
	/*
	* Since we don't allow device removal in a pool
	* that has a checkpoint, we expect that all removed
	* vdevs were removed from the pool before the
	* checkpoint.
	*/
	ASSERT3P(current_vd->vdev_ops, ==, &vdev_indirect_ops);
	continue;
	}

	/*
	* If the checkpoint space map doesn't exist, then nothing
	* here is checkpointed so there's nothing to verify.
	*/
	if (current_vd->vdev_top_zap == 0 \|\|
	zap_contains(spa_meta_objset(current),
	current_vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) != 0)
	continue;

	VERIFY0(zap_lookup(spa_meta_objset(current),
	current_vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM,
	sizeof (uint64_t), 1, &checkpoint_sm_obj));

	VERIFY0(space_map_open(&checkpoint_sm, spa_meta_objset(current),
	checkpoint_sm_obj, 0, current_vd->vdev_asize,
	current_vd->vdev_ashift));

	verify_checkpoint_sm_entry_cb_arg_t vcsec;
	vcsec.vcsec_vd = ckpoint_vd;
	vcsec.vcsec_entryid = 0;
	vcsec.vcsec_num_entries =
	space_map_length(checkpoint_sm) / sizeof (uint64_t);
	VERIFY0(space_map_iterate(checkpoint_sm,
	space_map_length(checkpoint_sm),
	verify_checkpoint_sm_entry_cb, &vcsec));
	if (dump_opt['m'] > 3)
	dump_spacemap(current->spa_meta_objset, checkpoint_sm);
	space_map_close(checkpoint_sm);
	}

	/*
	* If we've added vdevs since we took the checkpoint, ensure
	* that their checkpoint space maps are empty.
	*/
	if (ckpoint_rvd->vdev_children < current_rvd->vdev_children) {
	for (uint64_t c = ckpoint_rvd->vdev_children;
	c < current_rvd->vdev_children; c++) {
	vdev_t *current_vd = current_rvd->vdev_child[c];
	VERIFY3P(current_vd->vdev_checkpoint_sm, ==, NULL);
	}
	}

	/* for cleaner progress output */
	(void) fprintf(stderr, "\n");
	}

	/*
	* Verifies that all space that's allocated in the checkpoint is
	* still allocated in the current version, by checking that everything
	* in checkpoint's ms_allocatable (which is actually allocated, not
	* allocatable/free) is not present in current's ms_allocatable.
	*
	* Note that the function changes the state of the ms_allocatable
	* trees of both spas when called. The entries of all ms_allocatable
	* trees are cleared out and then repopulated from their respective
	* ms_sm space maps. In the checkpointed state we load the allocated
	* entries, and in the current state we load the free entries.
	*/
	static void
	verify_checkpoint_ms_spacemaps(spa_t checkpoint, spa_t current)
	{
	vdev_t *ckpoint_rvd = checkpoint->spa_root_vdev;
	vdev_t *current_rvd = current->spa_root_vdev;

	load_concrete_ms_allocatable_trees(checkpoint, SM_ALLOC);
	load_concrete_ms_allocatable_trees(current, SM_FREE);

	for (uint64_t i = 0; i < ckpoint_rvd->vdev_children; i++) {
	vdev_t *ckpoint_vd = ckpoint_rvd->vdev_child[i];
	vdev_t *current_vd = current_rvd->vdev_child[i];

	if (ckpoint_vd->vdev_ops == &vdev_indirect_ops) {
	/*
	* See comment in verify_checkpoint_vdev_spacemaps()
	*/
	ASSERT3P(current_vd->vdev_ops, ==, &vdev_indirect_ops);
	continue;
	}

	for (uint64_t m = 0; m < ckpoint_vd->vdev_ms_count; m++) {
	metaslab_t *ckpoint_msp = ckpoint_vd->vdev_ms[m];
	metaslab_t *current_msp = current_vd->vdev_ms[m];

	(void) fprintf(stderr,
	"\rverifying vdev %llu of %llu, "
	"metaslab %llu of %llu ...",
	(longlong_t)current_vd->vdev_id,
	(longlong_t)current_rvd->vdev_children,
	(longlong_t)current_vd->vdev_ms[m]->ms_id,
	(longlong_t)current_vd->vdev_ms_count);

	/*
	* We walk through the ms_allocatable trees that
	* are loaded with the allocated blocks from the
	* ms_sm spacemaps of the checkpoint. For each
	* one of these ranges we ensure that none of them
	* exists in the ms_allocatable trees of the
	* current state which are loaded with the ranges
	* that are currently free.
	*
	* This way we ensure that none of the blocks that
	* are part of the checkpoint were freed by mistake.
	*/
	range_tree_walk(ckpoint_msp->ms_allocatable,
	(range_tree_func_t *)range_tree_verify_not_present,
	current_msp->ms_allocatable);
	}
	}

	/* for cleaner progress output */
	(void) fprintf(stderr, "\n");
	}

	static void
	verify_checkpoint_blocks(spa_t *spa)
	{
	ASSERT(!dump_opt['L']);

	spa_t *checkpoint_spa;
	char *checkpoint_pool;
	int error = 0;

	/*
	* We import the checkpointed state of the pool (under a different
	* name) so we can do verification on it against the current state
	* of the pool.
	*/
	checkpoint_pool = import_checkpointed_state(spa->spa_name, NULL,
	NULL);
	ASSERT(strcmp(spa->spa_name, checkpoint_pool) != 0);

	error = spa_open(checkpoint_pool, &checkpoint_spa, FTAG);
	if (error != 0) {
	fatal("Tried to open pool \"%s\" but spa_open() failed with "
	"error %d\n", checkpoint_pool, error);
	}

	/*
	* Ensure that ranges in the checkpoint space maps of each vdev
	* are allocated according to the checkpointed state's metaslab
	* space maps.
	*/
	verify_checkpoint_vdev_spacemaps(checkpoint_spa, spa);

	/*
	* Ensure that allocated ranges in the checkpoint's metaslab
	* space maps remain allocated in the metaslab space maps of
	* the current state.
	*/
	verify_checkpoint_ms_spacemaps(checkpoint_spa, spa);

	/*
	* Once we are done, we get rid of the checkpointed state.
	*/
	spa_close(checkpoint_spa, FTAG);
	free(checkpoint_pool);
	}

	static void
	dump_leftover_checkpoint_blocks(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];

	space_map_t *checkpoint_sm = NULL;
	uint64_t checkpoint_sm_obj;

	if (vd->vdev_top_zap == 0)
	continue;

	if (zap_contains(spa_meta_objset(spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) != 0)
	continue;

	VERIFY0(zap_lookup(spa_meta_objset(spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM,
	sizeof (uint64_t), 1, &checkpoint_sm_obj));

	VERIFY0(space_map_open(&checkpoint_sm, spa_meta_objset(spa),
	checkpoint_sm_obj, 0, vd->vdev_asize, vd->vdev_ashift));
	dump_spacemap(spa->spa_meta_objset, checkpoint_sm);
	space_map_close(checkpoint_sm);
	}
	}

	static int
	verify_checkpoint(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error;

	if (!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
	return (0);

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error == ENOENT && !dump_opt['L']) {
	/*
	* If the feature is active but the uberblock is missing
	* then we must be in the middle of discarding the
	* checkpoint.
	*/
	(void) printf("\nPartially discarded checkpoint "
	"state found:\n");
	if (dump_opt['m'] > 3)
	dump_leftover_checkpoint_blocks(spa);
	return (0);
	} else if (error != 0) {
	(void) printf("lookup error %d when looking for "
	"checkpointed uberblock in MOS\n", error);
	return (error);
	}
	dump_uberblock(&checkpoint, "\nCheckpointed uberblock found:\n", "\n");

	if (checkpoint.ub_checkpoint_txg == 0) {
	(void) printf("\nub_checkpoint_txg not set in checkpointed "
	"uberblock\n");
	error = 3;
	}

	if (error == 0 && !dump_opt['L'])
	verify_checkpoint_blocks(spa);

	return (error);
	}

	static void
	mos_leaks_cb(void *arg, uint64_t start, uint64_t size)
	{
	(void) arg;
	for (uint64_t i = start; i < size; i++) {
	(void) printf("MOS object %llu referenced but not allocated\n",
	(u_longlong_t)i);
	}
	}

	static void
	mos_obj_refd(uint64_t obj)
	{
	if (obj != 0 && mos_refd_objs != NULL)
	range_tree_add(mos_refd_objs, obj, 1);
	}

	/*
	* Call on a MOS object that may already have been referenced.
	*/
	static void
	mos_obj_refd_multiple(uint64_t obj)
	{
	if (obj != 0 && mos_refd_objs != NULL &&
	!range_tree_contains(mos_refd_objs, obj, 1))
	range_tree_add(mos_refd_objs, obj, 1);
	}

	static void
	mos_leak_vdev_top_zap(vdev_t *vd)
	{
	uint64_t ms_flush_data_obj;
	int error = zap_lookup(spa_meta_objset(vd->vdev_spa),
	vd->vdev_top_zap, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS,
	sizeof (ms_flush_data_obj), 1, &ms_flush_data_obj);
	if (error == ENOENT)
	return;
	ASSERT0(error);

	mos_obj_refd(ms_flush_data_obj);
	}

	static void
	mos_leak_vdev(vdev_t *vd)
	{
	mos_obj_refd(vd->vdev_dtl_object);
	mos_obj_refd(vd->vdev_ms_array);
	mos_obj_refd(vd->vdev_indirect_config.vic_births_object);
	mos_obj_refd(vd->vdev_indirect_config.vic_mapping_object);
	mos_obj_refd(vd->vdev_leaf_zap);
	if (vd->vdev_checkpoint_sm != NULL)
	mos_obj_refd(vd->vdev_checkpoint_sm->sm_object);
	if (vd->vdev_indirect_mapping != NULL) {
	mos_obj_refd(vd->vdev_indirect_mapping->
	vim_phys->vimp_counts_object);
	}
	if (vd->vdev_obsolete_sm != NULL)
	mos_obj_refd(vd->vdev_obsolete_sm->sm_object);

	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
	metaslab_t *ms = vd->vdev_ms[m];
	mos_obj_refd(space_map_object(ms->ms_sm));
	}

	if (vd->vdev_top_zap != 0) {
	mos_obj_refd(vd->vdev_top_zap);
	mos_leak_vdev_top_zap(vd);
	}

	for (uint64_t c = 0; c < vd->vdev_children; c++) {
	mos_leak_vdev(vd->vdev_child[c]);
	}
	}

	static void
	mos_leak_log_spacemaps(spa_t *spa)
	{
	uint64_t spacemap_zap;
	int error = zap_lookup(spa_meta_objset(spa),
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_LOG_SPACEMAP_ZAP,
	sizeof (spacemap_zap), 1, &spacemap_zap);
	if (error == ENOENT)
	return;
	ASSERT0(error);

	mos_obj_refd(spacemap_zap);
	for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
	sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls))
	mos_obj_refd(sls->sls_sm_obj);
	}

	static int
	dump_mos_leaks(spa_t *spa)
	{
	int rv = 0;
	objset_t *mos = spa->spa_meta_objset;
	dsl_pool_t *dp = spa->spa_dsl_pool;

	/* Visit and mark all referenced objects in the MOS */

	mos_obj_refd(DMU_POOL_DIRECTORY_OBJECT);
	mos_obj_refd(spa->spa_pool_props_object);
	mos_obj_refd(spa->spa_config_object);
	mos_obj_refd(spa->spa_ddt_stat_object);
	mos_obj_refd(spa->spa_feat_desc_obj);
	mos_obj_refd(spa->spa_feat_enabled_txg_obj);
	mos_obj_refd(spa->spa_feat_for_read_obj);
	mos_obj_refd(spa->spa_feat_for_write_obj);
	mos_obj_refd(spa->spa_history);
	mos_obj_refd(spa->spa_errlog_last);
	mos_obj_refd(spa->spa_errlog_scrub);
	mos_obj_refd(spa->spa_all_vdev_zaps);
	mos_obj_refd(spa->spa_dsl_pool->dp_bptree_obj);
	mos_obj_refd(spa->spa_dsl_pool->dp_tmp_userrefs_obj);
	mos_obj_refd(spa->spa_dsl_pool->dp_scan->scn_phys.scn_queue_obj);
	bpobj_count_refd(&spa->spa_deferred_bpobj);
	mos_obj_refd(dp->dp_empty_bpobj);
	bpobj_count_refd(&dp->dp_obsolete_bpobj);
	bpobj_count_refd(&dp->dp_free_bpobj);
	mos_obj_refd(spa->spa_l2cache.sav_object);
	mos_obj_refd(spa->spa_spares.sav_object);

	if (spa->spa_syncing_log_sm != NULL)
	mos_obj_refd(spa->spa_syncing_log_sm->sm_object);
	mos_leak_log_spacemaps(spa);

	mos_obj_refd(spa->spa_condensing_indirect_phys.
	scip_next_mapping_object);
	mos_obj_refd(spa->spa_condensing_indirect_phys.
	scip_prev_obsolete_sm_object);
	if (spa->spa_condensing_indirect_phys.scip_next_mapping_object != 0) {
	vdev_indirect_mapping_t *vim =
	vdev_indirect_mapping_open(mos,
	spa->spa_condensing_indirect_phys.scip_next_mapping_object);
	mos_obj_refd(vim->vim_phys->vimp_counts_object);
	vdev_indirect_mapping_close(vim);
	}
	deleted_livelists_dump_mos(spa);

	if (dp->dp_origin_snap != NULL) {
	dsl_dataset_t *ds;

	dsl_pool_config_enter(dp, FTAG);
	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(dp->dp_origin_snap)->ds_next_snap_obj,
	FTAG, &ds));
	count_ds_mos_objects(ds);
	dump_blkptr_list(&ds->ds_deadlist, "Deadlist");
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);

	count_ds_mos_objects(dp->dp_origin_snap);
	dump_blkptr_list(&dp->dp_origin_snap->ds_deadlist, "Deadlist");
	}
	count_dir_mos_objects(dp->dp_mos_dir);
	if (dp->dp_free_dir != NULL)
	count_dir_mos_objects(dp->dp_free_dir);
	if (dp->dp_leak_dir != NULL)
	count_dir_mos_objects(dp->dp_leak_dir);

	mos_leak_vdev(spa->spa_root_vdev);

	for (uint64_t class = 0; class < DDT_CLASSES; class++) {
	for (uint64_t type = 0; type < DDT_TYPES; type++) {
	for (uint64_t cksum = 0;
	cksum < ZIO_CHECKSUM_FUNCTIONS; cksum++) {
	ddt_t *ddt = spa->spa_ddt[cksum];
	mos_obj_refd(ddt->ddt_object[type][class]);
	}
	}
	}

	/*
	* Visit all allocated objects and make sure they are referenced.
	*/
	uint64_t object = 0;
	while (dmu_object_next(mos, &object, B_FALSE, 0) == 0) {
	if (range_tree_contains(mos_refd_objs, object, 1)) {
	range_tree_remove(mos_refd_objs, object, 1);
	} else {
	dmu_object_info_t doi;
	const char *name;
	- dmu_object_info(mos, object, &doi);
	+ VERIFY0(dmu_object_info(mos, object, &doi));
	if (doi.doi_type & DMU_OT_NEWTYPE) {
	dmu_object_byteswap_t bswap =
	DMU_OT_BYTESWAP(doi.doi_type);
	name = dmu_ot_byteswap[bswap].ob_name;
	} else {
	name = dmu_ot[doi.doi_type].ot_name;
	}

	(void) printf("MOS object %llu (%s) leaked\n",
	(u_longlong_t)object, name);
	rv = 2;
	}
	}
	(void) range_tree_walk(mos_refd_objs, mos_leaks_cb, NULL);
	if (!range_tree_is_empty(mos_refd_objs))
	rv = 2;
	range_tree_vacate(mos_refd_objs, NULL, NULL);
	range_tree_destroy(mos_refd_objs);
	return (rv);
	}

	typedef struct log_sm_obsolete_stats_arg {
	uint64_t lsos_current_txg;

	uint64_t lsos_total_entries;
	uint64_t lsos_valid_entries;

	uint64_t lsos_sm_entries;
	uint64_t lsos_valid_sm_entries;
	} log_sm_obsolete_stats_arg_t;

	static int
	log_spacemap_obsolete_stats_cb(spa_t spa, space_map_entry_t sme,
	uint64_t txg, void *arg)
	{
	log_sm_obsolete_stats_arg_t *lsos = arg;

	uint64_t offset = sme->sme_offset;
	uint64_t vdev_id = sme->sme_vdev;

	if (lsos->lsos_current_txg == 0) {
	/* this is the first log */
	lsos->lsos_current_txg = txg;
	} else if (lsos->lsos_current_txg < txg) {
	/* we just changed log - print stats and reset */
	(void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
	(u_longlong_t)lsos->lsos_valid_sm_entries,
	(u_longlong_t)lsos->lsos_sm_entries,
	(u_longlong_t)lsos->lsos_current_txg);
	lsos->lsos_valid_sm_entries = 0;
	lsos->lsos_sm_entries = 0;
	lsos->lsos_current_txg = txg;
	}
	ASSERT3U(lsos->lsos_current_txg, ==, txg);

	lsos->lsos_sm_entries++;
	lsos->lsos_total_entries++;

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	if (!vdev_is_concrete(vd))
	return (0);

	metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift];
	ASSERT(sme->sme_type == SM_ALLOC \|\| sme->sme_type == SM_FREE);

	if (txg < metaslab_unflushed_txg(ms))
	return (0);
	lsos->lsos_valid_sm_entries++;
	lsos->lsos_valid_entries++;
	return (0);
	}

	static void
	dump_log_spacemap_obsolete_stats(spa_t *spa)
	{
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
	return;

	log_sm_obsolete_stats_arg_t lsos = {0};

	(void) printf("Log Space Map Obsolete Entry Statistics:\n");

	iterate_through_spacemap_logs(spa,
	log_spacemap_obsolete_stats_cb, &lsos);

	/* print stats for latest log */
	(void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
	(u_longlong_t)lsos.lsos_valid_sm_entries,
	(u_longlong_t)lsos.lsos_sm_entries,
	(u_longlong_t)lsos.lsos_current_txg);

	(void) printf("%-8llu valid entries out of %-8llu - total\n\n",
	(u_longlong_t)lsos.lsos_valid_entries,
	(u_longlong_t)lsos.lsos_total_entries);
	}

	static void
	dump_zpool(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);
	int rc = 0;

	if (dump_opt['y']) {
	livelist_metaslab_validate(spa);
	}

	if (dump_opt['S']) {
	dump_simulated_ddt(spa);
	return;
	}

	if (!dump_opt['e'] && dump_opt['C'] > 1) {
	(void) printf("\nCached configuration:\n");
	dump_nvlist(spa->spa_config, 8);
	}

	if (dump_opt['C'])
	dump_config(spa);

	if (dump_opt['u'])
	dump_uberblock(&spa->spa_uberblock, "\nUberblock:\n", "\n");

	if (dump_opt['D'])
	dump_all_ddts(spa);

	if (dump_opt['d'] > 2 \|\| dump_opt['m'])
	dump_metaslabs(spa);
	if (dump_opt['M'])
	dump_metaslab_groups(spa, dump_opt['M'] > 1);
	if (dump_opt['d'] > 2 \|\| dump_opt['m']) {
	dump_log_spacemaps(spa);
	dump_log_spacemap_obsolete_stats(spa);
	}

	if (dump_opt['d'] \|\| dump_opt['i']) {
	spa_feature_t f;
	mos_refd_objs = range_tree_create(NULL, RANGE_SEG64, NULL, 0,
	0);
	dump_objset(dp->dp_meta_objset);

	if (dump_opt['d'] >= 3) {
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dump_full_bpobj(&spa->spa_deferred_bpobj,
	"Deferred frees", 0);
	if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
	dump_full_bpobj(&dp->dp_free_bpobj,
	"Pool snapshot frees", 0);
	}
	if (bpobj_is_open(&dp->dp_obsolete_bpobj)) {
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_DEVICE_REMOVAL));
	dump_full_bpobj(&dp->dp_obsolete_bpobj,
	"Pool obsolete blocks", 0);
	}

	if (spa_feature_is_active(spa,
	SPA_FEATURE_ASYNC_DESTROY)) {
	dump_bptree(spa->spa_meta_objset,
	dp->dp_bptree_obj,
	"Pool dataset frees");
	}
	dump_dtl(spa->spa_root_vdev, 0);
	}

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++)
	global_feature_count[f] = UINT64_MAX;
	global_feature_count[SPA_FEATURE_REDACTION_BOOKMARKS] = 0;
	global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN] = 0;
	global_feature_count[SPA_FEATURE_LIVELIST] = 0;

	(void) dmu_objset_find(spa_name(spa), dump_one_objset,
	NULL, DS_FIND_SNAPSHOTS \| DS_FIND_CHILDREN);

	if (rc == 0 && !dump_opt['L'])
	rc = dump_mos_leaks(spa);

	for (f = 0; f < SPA_FEATURES; f++) {
	uint64_t refcount;

	uint64_t *arr;
	if (!(spa_feature_table[f].fi_flags &
	ZFEATURE_FLAG_PER_DATASET)) {
	if (global_feature_count[f] == UINT64_MAX)
	continue;
	if (!spa_feature_is_enabled(spa, f)) {
	ASSERT0(global_feature_count[f]);
	continue;
	}
	arr = global_feature_count;
	} else {
	if (!spa_feature_is_enabled(spa, f)) {
	ASSERT0(dataset_feature_count[f]);
	continue;
	}
	arr = dataset_feature_count;
	}
	if (feature_get_refcount(spa, &spa_feature_table[f],
	&refcount) == ENOTSUP)
	continue;
	if (arr[f] != refcount) {
	(void) printf("%s feature refcount mismatch: "
	"%lld consumers != %lld refcount\n",
	spa_feature_table[f].fi_uname,
	(longlong_t)arr[f], (longlong_t)refcount);
	rc = 2;
	} else {
	(void) printf("Verified %s feature refcount "
	"of %llu is correct\n",
	spa_feature_table[f].fi_uname,
	(longlong_t)refcount);
	}
	}

	if (rc == 0)
	rc = verify_device_removal_feature_counts(spa);
	}

	if (rc == 0 && (dump_opt['b'] \|\| dump_opt['c']))
	rc = dump_block_stats(spa);

	if (rc == 0)
	rc = verify_spacemap_refcounts(spa);

	if (dump_opt['s'])
	show_pool_stats(spa);

	if (dump_opt['h'])
	dump_history(spa);

	if (rc == 0)
	rc = verify_checkpoint(spa);

	if (rc != 0) {
	dump_debug_buffer();
	exit(rc);
	}
	}

	#define ZDB_FLAG_CHECKSUM 0x0001
	#define ZDB_FLAG_DECOMPRESS 0x0002
	#define ZDB_FLAG_BSWAP 0x0004
	#define ZDB_FLAG_GBH 0x0008
	#define ZDB_FLAG_INDIRECT 0x0010
	#define ZDB_FLAG_RAW 0x0020
	#define ZDB_FLAG_PRINT_BLKPTR 0x0040
	#define ZDB_FLAG_VERBOSE 0x0080

	static int flagbits[256];
	static char flagbitstr[16];

	static void
	zdb_print_blkptr(const blkptr_t *bp, int flags)
	{
	char blkbuf[BP_SPRINTF_LEN];

	if (flags & ZDB_FLAG_BSWAP)
	byteswap_uint64_array((void *)bp, sizeof (blkptr_t));

	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	(void) printf("%s\n", blkbuf);
	}

	static void
	zdb_dump_indirect(blkptr_t *bp, int nbps, int flags)
	{
	int i;

	for (i = 0; i < nbps; i++)
	zdb_print_blkptr(&bp[i], flags);
	}

	static void
	zdb_dump_gbh(void *buf, int flags)
	{
	zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags);
	}

	static void
	zdb_dump_block_raw(void *buf, uint64_t size, int flags)
	{
	if (flags & ZDB_FLAG_BSWAP)
	byteswap_uint64_array(buf, size);
	VERIFY(write(fileno(stdout), buf, size) == size);
	}

	static void
	zdb_dump_block(char label, void buf, uint64_t size, int flags)
	{
	uint64_t d = (uint64_t )buf;
	unsigned nwords = size / sizeof (uint64_t);
	int do_bswap = !!(flags & ZDB_FLAG_BSWAP);
	unsigned i, j;
	const char *hdr;
	char *c;


	if (do_bswap)
	hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
	else
	hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f";

	(void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr);

	#ifdef _LITTLE_ENDIAN
	/* correct the endianness */
	do_bswap = !do_bswap;
	#endif
	for (i = 0; i < nwords; i += 2) {
	(void) printf("%06llx: %016llx %016llx ",
	(u_longlong_t)(i * sizeof (uint64_t)),
	(u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]),
	(u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1]));

	c = (char *)&d[i];
	for (j = 0; j < 2 * sizeof (uint64_t); j++)
	(void) printf("%c", isprint(c[j]) ? c[j] : '.');
	(void) printf("\n");
	}
	}

	/*
	* There are two acceptable formats:
	* leaf_name - For example: c1t0d0 or /tmp/ztest.0a
	* child[.child]* - For example: 0.1.1
	*
	* The second form can be used to specify arbitrary vdevs anywhere
	* in the hierarchy. For example, in a pool with a mirror of
	* RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
	*/
	static vdev_t *
	zdb_vdev_lookup(vdev_t vdev, const char path)
	{
	char s, p, *q;
	unsigned i;

	if (vdev == NULL)
	return (NULL);

	/* First, assume the x.x.x.x format */
	i = strtoul(path, &s, 10);
	if (s == path \|\| (s && s != '.' && s != '\0'))
	goto name;
	if (i >= vdev->vdev_children)
	return (NULL);

	vdev = vdev->vdev_child[i];
	if (s && *s == '\0')
	return (vdev);
	return (zdb_vdev_lookup(vdev, s+1));

	name:
	for (i = 0; i < vdev->vdev_children; i++) {
	vdev_t *vc = vdev->vdev_child[i];

	if (vc->vdev_path == NULL) {
	vc = zdb_vdev_lookup(vc, path);
	if (vc == NULL)
	continue;
	else
	return (vc);
	}

	p = strrchr(vc->vdev_path, '/');
	p = p ? p + 1 : vc->vdev_path;
	q = &vc->vdev_path[strlen(vc->vdev_path) - 2];

	if (strcmp(vc->vdev_path, path) == 0)
	return (vc);
	if (strcmp(p, path) == 0)
	return (vc);
	if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0)
	return (vc);
	}

	return (NULL);
	}

	static int
	name_from_objset_id(spa_t spa, uint64_t objset_id, char outstr)
	{
	dsl_dataset_t *ds;

	dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
	int error = dsl_dataset_hold_obj(spa->spa_dsl_pool, objset_id,
	NULL, &ds);
	if (error != 0) {
	(void) fprintf(stderr, "failed to hold objset %llu: %s\n",
	(u_longlong_t)objset_id, strerror(error));
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	return (error);
	}
	dsl_dataset_name(ds, outstr);
	dsl_dataset_rele(ds, NULL);
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	return (0);
	}

	static boolean_t
	zdb_parse_block_sizes(char sizes, uint64_t lsize, uint64_t *psize)
	{
	char s0, s1, *tmp = NULL;

	if (sizes == NULL)
	return (B_FALSE);

	s0 = strtok_r(sizes, "/", &tmp);
	if (s0 == NULL)
	return (B_FALSE);
	s1 = strtok_r(NULL, "/", &tmp);
	*lsize = strtoull(s0, NULL, 16);
	psize = s1 ? strtoull(s1, NULL, 16) : lsize;
	return (lsize >= psize && *psize > 0);
	}

	#define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))

	static boolean_t
	zdb_decompress_block(abd_t pabd, void buf, void *lbuf, uint64_t lsize,
	uint64_t psize, int flags)
	{
	(void) buf;
	boolean_t exceeded = B_FALSE;
	/*
	* We don't know how the data was compressed, so just try
	* every decompress function at every inflated blocksize.
	*/
	void *lbuf2 = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);
	int cfuncs[ZIO_COMPRESS_FUNCTIONS] = { 0 };
	int *cfuncp = cfuncs;
	uint64_t maxlsize = SPA_MAXBLOCKSIZE;
	uint64_t mask = ZIO_COMPRESS_MASK(ON) \| ZIO_COMPRESS_MASK(OFF) \|
	ZIO_COMPRESS_MASK(INHERIT) \| ZIO_COMPRESS_MASK(EMPTY) \|
	(getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE) : 0);
	*cfuncp++ = ZIO_COMPRESS_LZ4;
	*cfuncp++ = ZIO_COMPRESS_LZJB;
	mask \|= ZIO_COMPRESS_MASK(LZ4) \| ZIO_COMPRESS_MASK(LZJB);
	for (int c = 0; c < ZIO_COMPRESS_FUNCTIONS; c++)
	if (((1ULL << c) & mask) == 0)
	*cfuncp++ = c;

	/*
	* On the one hand, with SPA_MAXBLOCKSIZE at 16MB, this
	* could take a while and we should let the user know
	* we are not stuck. On the other hand, printing progress
	* info gets old after a while. User can specify 'v' flag
	* to see the progression.
	*/
	if (lsize == psize)
	lsize += SPA_MINBLOCKSIZE;
	else
	maxlsize = lsize;
	for (; lsize <= maxlsize; lsize += SPA_MINBLOCKSIZE) {
	for (cfuncp = cfuncs; *cfuncp; cfuncp++) {
	if (flags & ZDB_FLAG_VERBOSE) {
	(void) fprintf(stderr,
	"Trying %05llx -> %05llx (%s)\n",
	(u_longlong_t)psize,
	(u_longlong_t)lsize,
	zio_compress_table[*cfuncp].\
	ci_name);
	}

	/*
	* We randomize lbuf2, and decompress to both
	* lbuf and lbuf2. This way, we will know if
	* decompression fill exactly to lsize.
	*/
	VERIFY0(random_get_pseudo_bytes(lbuf2, lsize));

	if (zio_decompress_data(*cfuncp, pabd,
	lbuf, psize, lsize, NULL) == 0 &&
	zio_decompress_data(*cfuncp, pabd,
	lbuf2, psize, lsize, NULL) == 0 &&
	memcmp(lbuf, lbuf2, lsize) == 0)
	break;
	}
	if (*cfuncp != 0)
	break;
	}
	umem_free(lbuf2, SPA_MAXBLOCKSIZE);

	if (lsize > maxlsize) {
	exceeded = B_TRUE;
	}
	if (*cfuncp == ZIO_COMPRESS_ZLE) {
	printf("\nZLE decompression was selected. If you "
	"suspect the results are wrong,\ntry avoiding ZLE "
	"by setting and exporting ZDB_NO_ZLE=\"true\"\n");
	}

	return (exceeded);
	}

	/*
	* Read a block from a pool and print it out. The syntax of the
	* block descriptor is:
	*
	* pool:vdev_specifier:offset:[lsize/]psize[:flags]
	*
	* pool - The name of the pool you wish to read from
	* vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
	* offset - offset, in hex, in bytes
	* size - Amount of data to read, in hex, in bytes
	* flags - A string of characters specifying options
	* b: Decode a blkptr at given offset within block
	* c: Calculate and display checksums
	* d: Decompress data before dumping
	* e: Byteswap data before dumping
	* g: Display data as a gang block header
	* i: Display as an indirect block
	* r: Dump raw data to stdout
	* v: Verbose
	*
	*/
	static void
	zdb_read_block(char thing, spa_t spa)
	{
	blkptr_t blk, *bp = &blk;
	dva_t *dva = bp->blk_dva;
	int flags = 0;
	uint64_t offset = 0, psize = 0, lsize = 0, blkptr_offset = 0;
	zio_t *zio;
	vdev_t *vd;
	abd_t *pabd;
	void lbuf, buf;
	char s, p, dup, flagstr, sizes, tmp = NULL;
	const char vdev, errmsg = NULL;
	int i, error;
	boolean_t borrowed = B_FALSE, found = B_FALSE;

	dup = strdup(thing);
	s = strtok_r(dup, ":", &tmp);
	vdev = s ?: "";
	s = strtok_r(NULL, ":", &tmp);
	offset = strtoull(s ? s : "", NULL, 16);
	sizes = strtok_r(NULL, ":", &tmp);
	s = strtok_r(NULL, ":", &tmp);
	flagstr = strdup(s ?: "");

	if (!zdb_parse_block_sizes(sizes, &lsize, &psize))
	errmsg = "invalid size(s)";
	if (!IS_P2ALIGNED(psize, DEV_BSIZE) \|\| !IS_P2ALIGNED(lsize, DEV_BSIZE))
	errmsg = "size must be a multiple of sector size";
	if (!IS_P2ALIGNED(offset, DEV_BSIZE))
	errmsg = "offset must be a multiple of sector size";
	if (errmsg) {
	(void) printf("Invalid block specifier: %s - %s\n",
	thing, errmsg);
	goto done;
	}

	tmp = NULL;
	for (s = strtok_r(flagstr, ":", &tmp);
	s != NULL;
	s = strtok_r(NULL, ":", &tmp)) {
	for (i = 0; i < strlen(flagstr); i++) {
	int bit = flagbits[(uchar_t)flagstr[i]];

	if (bit == 0) {
	(void) printf("***Ignoring flag: %c\n",
	(uchar_t)flagstr[i]);
	continue;
	}
	found = B_TRUE;
	flags \|= bit;

	p = &flagstr[i + 1];
	if (p != ':' && p != '\0') {
	int j = 0, nextbit = flagbits[(uchar_t)*p];
	char *end, offstr[8] = { 0 };
	if ((bit == ZDB_FLAG_PRINT_BLKPTR) &&
	(nextbit == 0)) {
	/* look ahead to isolate the offset */
	while (nextbit == 0 &&
	strchr(flagbitstr, *p) == NULL) {
	offstr[j] = *p;
	j++;
	if (i + j > strlen(flagstr))
	break;
	p++;
	nextbit = flagbits[(uchar_t)*p];
	}
	blkptr_offset = strtoull(offstr, &end,
	16);
	i += j;
	} else if (nextbit == 0) {
	(void) printf("***Ignoring flag arg:"
	" '%c'\n", (uchar_t)*p);
	}
	}
	}
	}
	if (blkptr_offset % sizeof (blkptr_t)) {
	printf("Block pointer offset 0x%llx "
	"must be divisible by 0x%x\n",
	(longlong_t)blkptr_offset, (int)sizeof (blkptr_t));
	goto done;
	}
	if (found == B_FALSE && strlen(flagstr) > 0) {
	printf("Invalid flag arg: '%s'\n", flagstr);
	goto done;
	}

	vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev);
	if (vd == NULL) {
	(void) printf("***Invalid vdev: %s\n", vdev);
	- free(dup);
	- return;
	+ goto done;
	} else {
	if (vd->vdev_path)
	(void) fprintf(stderr, "Found vdev: %s\n",
	vd->vdev_path);
	else
	(void) fprintf(stderr, "Found vdev type: %s\n",
	vd->vdev_ops->vdev_op_type);
	}

	pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
	lbuf = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);

	BP_ZERO(bp);

	DVA_SET_VDEV(&dva[0], vd->vdev_id);
	DVA_SET_OFFSET(&dva[0], offset);
	DVA_SET_GANG(&dva[0], !!(flags & ZDB_FLAG_GBH));
	DVA_SET_ASIZE(&dva[0], vdev_psize_to_asize(vd, psize));

	BP_SET_BIRTH(bp, TXG_INITIAL, TXG_INITIAL);

	BP_SET_LSIZE(bp, lsize);
	BP_SET_PSIZE(bp, psize);
	BP_SET_COMPRESS(bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(bp, ZIO_CHECKSUM_OFF);
	BP_SET_TYPE(bp, DMU_OT_NONE);
	BP_SET_LEVEL(bp, 0);
	BP_SET_DEDUP(bp, 0);
	BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	zio = zio_root(spa, NULL, NULL, 0);

	if (vd == vd->vdev_top) {
	/*
	* Treat this as a normal block read.
	*/
	zio_nowait(zio_read(zio, spa, bp, pabd, psize, NULL, NULL,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW, NULL));
	} else {
	/*
	* Treat this as a vdev child I/O.
	*/
	zio_nowait(zio_vdev_child_io(zio, bp, vd, offset, pabd,
	psize, ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \|
	ZIO_FLAG_OPTIONAL, NULL, NULL));
	}

	error = zio_wait(zio);
	spa_config_exit(spa, SCL_STATE, FTAG);

	if (error) {
	(void) printf("Read of %s failed, error: %d\n", thing, error);
	goto out;
	}

	uint64_t orig_lsize = lsize;
	buf = lbuf;
	if (flags & ZDB_FLAG_DECOMPRESS) {
	boolean_t failed = zdb_decompress_block(pabd, buf, lbuf,
	lsize, psize, flags);
	if (failed) {
	(void) printf("Decompress of %s failed\n", thing);
	goto out;
	}
	} else {
	buf = abd_borrow_buf_copy(pabd, lsize);
	borrowed = B_TRUE;
	}
	/*
	* Try to detect invalid block pointer. If invalid, try
	* decompressing.
	*/
	if ((flags & ZDB_FLAG_PRINT_BLKPTR \|\| flags & ZDB_FLAG_INDIRECT) &&
	!(flags & ZDB_FLAG_DECOMPRESS)) {
	const blkptr_t b = (const blkptr_t )(void *)
	((uintptr_t)buf + (uintptr_t)blkptr_offset);
	if (zfs_blkptr_verify(spa, b, B_FALSE, BLK_VERIFY_ONLY) ==
	B_FALSE) {
	abd_return_buf_copy(pabd, buf, lsize);
	borrowed = B_FALSE;
	buf = lbuf;
	boolean_t failed = zdb_decompress_block(pabd, buf,
	lbuf, lsize, psize, flags);
	b = (const blkptr_t )(void )
	((uintptr_t)buf + (uintptr_t)blkptr_offset);
	if (failed \|\| zfs_blkptr_verify(spa, b, B_FALSE,
	BLK_VERIFY_LOG) == B_FALSE) {
	printf("invalid block pointer at this DVA\n");
	goto out;
	}
	}
	}

	if (flags & ZDB_FLAG_PRINT_BLKPTR)
	zdb_print_blkptr((blkptr_t )(void )
	((uintptr_t)buf + (uintptr_t)blkptr_offset), flags);
	else if (flags & ZDB_FLAG_RAW)
	zdb_dump_block_raw(buf, lsize, flags);
	else if (flags & ZDB_FLAG_INDIRECT)
	zdb_dump_indirect((blkptr_t *)buf,
	orig_lsize / sizeof (blkptr_t), flags);
	else if (flags & ZDB_FLAG_GBH)
	zdb_dump_gbh(buf, flags);
	else
	zdb_dump_block(thing, buf, lsize, flags);

	/*
	* If :c was specified, iterate through the checksum table to
	* calculate and display each checksum for our specified
	* DVA and length.
	*/
	if ((flags & ZDB_FLAG_CHECKSUM) && !(flags & ZDB_FLAG_RAW) &&
	!(flags & ZDB_FLAG_GBH)) {
	zio_t *czio;
	(void) printf("\n");
	for (enum zio_checksum ck = ZIO_CHECKSUM_LABEL;
	ck < ZIO_CHECKSUM_FUNCTIONS; ck++) {

	if ((zio_checksum_table[ck].ci_flags &
	ZCHECKSUM_FLAG_EMBEDDED) \|\|
	ck == ZIO_CHECKSUM_NOPARITY) {
	continue;
	}
	BP_SET_CHECKSUM(bp, ck);
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	czio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	czio->io_bp = bp;

	if (vd == vd->vdev_top) {
	zio_nowait(zio_read(czio, spa, bp, pabd, psize,
	NULL, NULL,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \|
	ZIO_FLAG_DONT_RETRY, NULL));
	} else {
	zio_nowait(zio_vdev_child_io(czio, bp, vd,
	offset, pabd, psize, ZIO_TYPE_READ,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_DONT_CACHE \|
	ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \|
	ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_OPTIONAL, NULL, NULL));
	}
	error = zio_wait(czio);
	if (error == 0 \|\| error == ECKSUM) {
	zio_t *ck_zio = zio_root(spa, NULL, NULL, 0);
	ck_zio->io_offset =
	DVA_GET_OFFSET(&bp->blk_dva[0]);
	ck_zio->io_bp = bp;
	zio_checksum_compute(ck_zio, ck, pabd, lsize);
	printf("%12s\tcksum=%llx:%llx:%llx:%llx\n",
	zio_checksum_table[ck].ci_name,
	(u_longlong_t)bp->blk_cksum.zc_word[0],
	(u_longlong_t)bp->blk_cksum.zc_word[1],
	(u_longlong_t)bp->blk_cksum.zc_word[2],
	(u_longlong_t)bp->blk_cksum.zc_word[3]);
	zio_wait(ck_zio);
	} else {
	printf("error %d reading block\n", error);
	}
	spa_config_exit(spa, SCL_STATE, FTAG);
	}
	}

	if (borrowed)
	abd_return_buf_copy(pabd, buf, lsize);

	out:
	abd_free(pabd);
	umem_free(lbuf, SPA_MAXBLOCKSIZE);
	done:
	free(flagstr);
	free(dup);
	}

	static void
	zdb_embedded_block(char *thing)
	{
	blkptr_t bp = {{{{0}}}};
	unsigned long long words = (void )&bp;
	char *buf;
	int err;

	err = sscanf(thing, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
	"%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
	words + 0, words + 1, words + 2, words + 3,
	words + 4, words + 5, words + 6, words + 7,
	words + 8, words + 9, words + 10, words + 11,
	words + 12, words + 13, words + 14, words + 15);
	if (err != 16) {
	(void) fprintf(stderr, "invalid input format\n");
	exit(1);
	}
	ASSERT3U(BPE_GET_LSIZE(&bp), <=, SPA_MAXBLOCKSIZE);
	buf = malloc(SPA_MAXBLOCKSIZE);
	if (buf == NULL) {
	(void) fprintf(stderr, "out of memory\n");
	exit(1);
	}
	err = decode_embedded_bp(&bp, buf, BPE_GET_LSIZE(&bp));
	if (err != 0) {
	(void) fprintf(stderr, "decode failed: %u\n", err);
	exit(1);
	}
	zdb_dump_block_raw(buf, BPE_GET_LSIZE(&bp), 0);
	free(buf);
	}

	/* check for valid hex or decimal numeric string */
	static boolean_t
	zdb_numeric(char *str)
	{
	int i = 0;

	if (strlen(str) == 0)
	return (B_FALSE);
	if (strncmp(str, "0x", 2) == 0 \|\| strncmp(str, "0X", 2) == 0)
	i = 2;
	for (; i < strlen(str); i++) {
	if (!isxdigit(str[i]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	int
	main(int argc, char **argv)
	{
	int c;
	spa_t *spa = NULL;
	objset_t *os = NULL;
	int dump_all = 1;
	int verbose = 0;
	int error = 0;
	char **searchdirs = NULL;
	int nsearch = 0;
	char target, target_pool, dsname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *policy = NULL;
	uint64_t max_txg = UINT64_MAX;
	int64_t objset_id = -1;
	uint64_t object;
	int flags = ZFS_IMPORT_MISSING_LOG;
	int rewind = ZPOOL_NEVER_REWIND;
	char spa_config_path_env, objset_str;
	boolean_t target_is_spa = B_TRUE, dataset_lookup = B_FALSE;
	nvlist_t *cfg = NULL;

	dprintf_setup(&argc, argv);

	/*
	* If there is an environment variable SPA_CONFIG_PATH it overrides
	* default spa_config_path setting. If -U flag is specified it will
	* override this environment variable settings once again.
	*/
	spa_config_path_env = getenv("SPA_CONFIG_PATH");
	if (spa_config_path_env != NULL)
	spa_config_path = spa_config_path_env;

	/*
	* For performance reasons, we set this tunable down. We do so before
	* the arg parsing section so that the user can override this value if
	* they choose.
	*/
	zfs_btree_verify_intensity = 3;

	struct option long_options[] = {
	{"ignore-assertions", no_argument, NULL, 'A'},
	{"block-stats", no_argument, NULL, 'b'},
	{"checksum", no_argument, NULL, 'c'},
	{"config", no_argument, NULL, 'C'},
	{"datasets", no_argument, NULL, 'd'},
	{"dedup-stats", no_argument, NULL, 'D'},
	{"exported", no_argument, NULL, 'e'},
	{"embedded-block-pointer", no_argument, NULL, 'E'},
	{"automatic-rewind", no_argument, NULL, 'F'},
	{"dump-debug-msg", no_argument, NULL, 'G'},
	{"history", no_argument, NULL, 'h'},
	{"intent-logs", no_argument, NULL, 'i'},
	{"inflight", required_argument, NULL, 'I'},
	{"checkpointed-state", no_argument, NULL, 'k'},
	{"label", no_argument, NULL, 'l'},
	{"disable-leak-tracking", no_argument, NULL, 'L'},
	{"metaslabs", no_argument, NULL, 'm'},
	{"metaslab-groups", no_argument, NULL, 'M'},
	{"numeric", no_argument, NULL, 'N'},
	{"option", required_argument, NULL, 'o'},
	{"object-lookups", no_argument, NULL, 'O'},
	{"path", required_argument, NULL, 'p'},
	{"parseable", no_argument, NULL, 'P'},
	{"skip-label", no_argument, NULL, 'q'},
	{"copy-object", no_argument, NULL, 'r'},
	{"read-block", no_argument, NULL, 'R'},
	{"io-stats", no_argument, NULL, 's'},
	{"simulate-dedup", no_argument, NULL, 'S'},
	{"txg", required_argument, NULL, 't'},
	{"uberblock", no_argument, NULL, 'u'},
	{"cachefile", required_argument, NULL, 'U'},
	{"verbose", no_argument, NULL, 'v'},
	{"verbatim", no_argument, NULL, 'V'},
	{"dump-blocks", required_argument, NULL, 'x'},
	{"extreme-rewind", no_argument, NULL, 'X'},
	{"all-reconstruction", no_argument, NULL, 'Y'},
	{"livelist", no_argument, NULL, 'y'},
	{"zstd-headers", no_argument, NULL, 'Z'},
	{0, 0, 0, 0}
	};

	while ((c = getopt_long(argc, argv,
	"AbcCdDeEFGhiI:klLmMNo:Op:PqrRsSt:uU:vVx:XYyZ",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'b':
	case 'c':
	case 'C':
	case 'd':
	case 'D':
	case 'E':
	case 'G':
	case 'h':
	case 'i':
	case 'l':
	case 'm':
	case 'M':
	case 'N':
	case 'O':
	case 'r':
	case 'R':
	case 's':
	case 'S':
	case 'u':
	case 'y':
	case 'Z':
	dump_opt[c]++;
	dump_all = 0;
	break;
	case 'A':
	case 'e':
	case 'F':
	case 'k':
	case 'L':
	case 'P':
	case 'q':
	case 'X':
	dump_opt[c]++;
	break;
	case 'Y':
	zfs_reconstruct_indirect_combinations_max = INT_MAX;
	zfs_deadman_enabled = 0;
	break;
	/* NB: Sort single match options below. */
	case 'I':
	max_inflight_bytes = strtoull(optarg, NULL, 0);
	if (max_inflight_bytes == 0) {
	(void) fprintf(stderr, "maximum number "
	"of inflight bytes must be greater "
	"than 0\n");
	usage();
	}
	break;
	case 'o':
	error = set_global_var(optarg);
	if (error != 0)
	usage();
	break;
	case 'p':
	if (searchdirs == NULL) {
	searchdirs = umem_alloc(sizeof (char *),
	UMEM_NOFAIL);
	} else {
	char *tmp = umem_alloc((nsearch + 1)
	sizeof (char *), UMEM_NOFAIL);
	memcpy(tmp, searchdirs, nsearch *
	sizeof (char *));
	umem_free(searchdirs,
	nsearch * sizeof (char *));
	searchdirs = tmp;
	}
	searchdirs[nsearch++] = optarg;
	break;
	case 't':
	max_txg = strtoull(optarg, NULL, 0);
	if (max_txg < TXG_INITIAL) {
	(void) fprintf(stderr, "incorrect txg "
	"specified: %s\n", optarg);
	usage();
	}
	break;
	case 'U':
	spa_config_path = optarg;
	if (spa_config_path[0] != '/') {
	(void) fprintf(stderr,
	"cachefile must be an absolute path "
	"(i.e. start with a slash)\n");
	usage();
	}
	break;
	case 'v':
	verbose++;
	break;
	case 'V':
	flags = ZFS_IMPORT_VERBATIM;
	break;
	case 'x':
	vn_dumpdir = optarg;
	break;
	default:
	usage();
	break;
	}
	}

	if (!dump_opt['e'] && searchdirs != NULL) {
	(void) fprintf(stderr, "-p option requires use of -e\n");
	usage();
	}
	#if defined(_LP64)
	/*
	* ZDB does not typically re-read blocks; therefore limit the ARC
	* to 256 MB, which can be used entirely for metadata.
	*/
	zfs_arc_min = zfs_arc_meta_min = 2ULL << SPA_MAXBLOCKSHIFT;
	zfs_arc_max = zfs_arc_meta_limit = 256 * 1024 * 1024;
	#endif

	/*
	* "zdb -c" uses checksum-verifying scrub i/os which are async reads.
	* "zdb -b" uses traversal prefetch which uses async reads.
	* For good performance, let several of them be active at once.
	*/
	zfs_vdev_async_read_max_active = 10;

	/*
	* Disable reference tracking for better performance.
	*/
	reference_tracking_enable = B_FALSE;

	/*
	* Do not fail spa_load when spa_load_verify fails. This is needed
	* to load non-idle pools.
	*/
	spa_load_verify_dryrun = B_TRUE;

	/*
	* ZDB should have ability to read spacemaps.
	*/
	spa_mode_readable_spacemaps = B_TRUE;

	kernel_init(SPA_MODE_READ);

	if (dump_all)
	verbose = MAX(verbose, 1);

	for (c = 0; c < 256; c++) {
	if (dump_all && strchr("AeEFklLNOPrRSXy", c) == NULL)
	dump_opt[c] = 1;
	if (dump_opt[c])
	dump_opt[c] += verbose;
	}

	libspl_set_assert_ok((dump_opt['A'] == 1) \|\| (dump_opt['A'] > 2));
	zfs_recover = (dump_opt['A'] > 1);

	argc -= optind;
	argv += optind;
	if (argc < 2 && dump_opt['R'])
	usage();

	if (dump_opt['E']) {
	if (argc != 1)
	usage();
	zdb_embedded_block(argv[0]);
	return (0);
	}

	if (argc < 1) {
	if (!dump_opt['e'] && dump_opt['C']) {
	dump_cachefile(spa_config_path);
	return (0);
	}
	usage();
	}

	if (dump_opt['l'])
	return (dump_label(argv[0]));

	if (dump_opt['O']) {
	if (argc != 2)
	usage();
	dump_opt['v'] = verbose + 3;
	return (dump_path(argv[0], argv[1], NULL));
	}
	if (dump_opt['r']) {
	target_is_spa = B_FALSE;
	if (argc != 3)
	usage();
	dump_opt['v'] = verbose;
	error = dump_path(argv[0], argv[1], &object);
	+ if (error != 0)
	+ fatal("internal error: %s", strerror(error));
	}

	if (dump_opt['X'] \|\| dump_opt['F'])
	rewind = ZPOOL_DO_REWIND \|
	(dump_opt['X'] ? ZPOOL_EXTREME_REWIND : 0);

	/* -N implies -d */
	if (dump_opt['N'] && dump_opt['d'] == 0)
	dump_opt['d'] = dump_opt['N'];

	if (nvlist_alloc(&policy, NV_UNIQUE_NAME_TYPE, 0) != 0 \|\|
	nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, max_txg) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY, rewind) != 0)
	fatal("internal error: %s", strerror(ENOMEM));

	error = 0;
	target = argv[0];

	if (strpbrk(target, "/@") != NULL) {
	size_t targetlen;

	target_pool = strdup(target);
	*strpbrk(target_pool, "/@") = '\0';

	target_is_spa = B_FALSE;
	targetlen = strlen(target);
	if (targetlen && target[targetlen - 1] == '/')
	target[targetlen - 1] = '\0';

	/*
	* See if an objset ID was supplied (-d <pool>/<objset ID>).
	* To disambiguate tank/100, consider the 100 as objsetID
	* if -N was given, otherwise 100 is an objsetID iff
	* tank/100 as a named dataset fails on lookup.
	*/
	objset_str = strchr(target, '/');
	if (objset_str && strlen(objset_str) > 1 &&
	zdb_numeric(objset_str + 1)) {
	char *endptr;
	errno = 0;
	objset_str++;
	objset_id = strtoull(objset_str, &endptr, 0);
	/* dataset 0 is the same as opening the pool */
	if (errno == 0 && endptr != objset_str &&
	objset_id != 0) {
	if (dump_opt['N'])
	dataset_lookup = B_TRUE;
	}
	/* normal dataset name not an objset ID */
	if (endptr == objset_str) {
	objset_id = -1;
	}
	} else if (objset_str && !zdb_numeric(objset_str + 1) &&
	dump_opt['N']) {
	printf("Supply a numeric objset ID with -N\n");
	exit(1);
	}
	} else {
	target_pool = target;
	}

	if (dump_opt['e']) {
	importargs_t args = { 0 };

	args.paths = nsearch;
	args.path = searchdirs;
	args.can_be_active = B_TRUE;

	error = zpool_find_config(NULL, target_pool, &cfg, &args,
	&libzpool_config_ops);

	if (error == 0) {

	if (nvlist_add_nvlist(cfg,
	ZPOOL_LOAD_POLICY, policy) != 0) {
	fatal("can't open '%s': %s",
	target, strerror(ENOMEM));
	}

	if (dump_opt['C'] > 1) {
	(void) printf("\nConfiguration for import:\n");
	dump_nvlist(cfg, 8);
	}

	/*
	* Disable the activity check to allow examination of
	* active pools.
	*/
	error = spa_import(target_pool, cfg, NULL,
	flags \| ZFS_IMPORT_SKIP_MMP);
	}
	}

	if (searchdirs != NULL) {
	umem_free(searchdirs, nsearch * sizeof (char *));
	searchdirs = NULL;
	}

	/*
	* import_checkpointed_state makes the assumption that the
	* target pool that we pass it is already part of the spa
	* namespace. Because of that we need to make sure to call
	* it always after the -e option has been processed, which
	* imports the pool to the namespace if it's not in the
	* cachefile.
	*/
	char *checkpoint_pool = NULL;
	char *checkpoint_target = NULL;
	if (dump_opt['k']) {
	checkpoint_pool = import_checkpointed_state(target, cfg,
	&checkpoint_target);

	if (checkpoint_target != NULL)
	target = checkpoint_target;
	}

	if (cfg != NULL) {
	nvlist_free(cfg);
	cfg = NULL;
	}

	if (target_pool != target)
	free(target_pool);

	if (error == 0) {
	if (dump_opt['k'] && (target_is_spa \|\| dump_opt['R'])) {
	ASSERT(checkpoint_pool != NULL);
	ASSERT(checkpoint_target == NULL);

	error = spa_open(checkpoint_pool, &spa, FTAG);
	if (error != 0) {
	fatal("Tried to open pool \"%s\" but "
	"spa_open() failed with error %d\n",
	checkpoint_pool, error);
	}

	} else if (target_is_spa \|\| dump_opt['R'] \|\| objset_id == 0) {
	zdb_set_skip_mmp(target);
	error = spa_open_rewind(target, &spa, FTAG, policy,
	NULL);
	if (error) {
	/*
	* If we're missing the log device then
	* try opening the pool after clearing the
	* log state.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(target)) != NULL &&
	spa->spa_log_state == SPA_LOG_MISSING) {
	spa->spa_log_state = SPA_LOG_CLEAR;
	error = 0;
	}
	mutex_exit(&spa_namespace_lock);

	if (!error) {
	error = spa_open_rewind(target, &spa,
	FTAG, policy, NULL);
	}
	}
	} else if (strpbrk(target, "#") != NULL) {
	dsl_pool_t *dp;
	error = dsl_pool_hold(target, FTAG, &dp);
	if (error != 0) {
	fatal("can't dump '%s': %s", target,
	strerror(error));
	}
	error = dump_bookmark(dp, target, B_TRUE, verbose > 1);
	dsl_pool_rele(dp, FTAG);
	if (error != 0) {
	fatal("can't dump '%s': %s", target,
	strerror(error));
	}
	return (error);
	} else {
	target_pool = strdup(target);
	if (strpbrk(target, "/@") != NULL)
	*strpbrk(target_pool, "/@") = '\0';

	zdb_set_skip_mmp(target);
	/*
	* If -N was supplied, the user has indicated that
	* zdb -d <pool>/<objsetID> is in effect. Otherwise
	* we first assume that the dataset string is the
	* dataset name. If dmu_objset_hold fails with the
	* dataset string, and we have an objset_id, retry the
	* lookup with the objsetID.
	*/
	boolean_t retry = B_TRUE;
	retry_lookup:
	if (dataset_lookup == B_TRUE) {
	/*
	* Use the supplied id to get the name
	* for open_objset.
	*/
	error = spa_open(target_pool, &spa, FTAG);
	if (error == 0) {
	error = name_from_objset_id(spa,
	objset_id, dsname);
	spa_close(spa, FTAG);
	if (error == 0)
	target = dsname;
	}
	}
	if (error == 0) {
	if (objset_id > 0 && retry) {
	int err = dmu_objset_hold(target, FTAG,
	&os);
	if (err) {
	dataset_lookup = B_TRUE;
	retry = B_FALSE;
	goto retry_lookup;
	} else {
	dmu_objset_rele(os, FTAG);
	}
	}
	error = open_objset(target, FTAG, &os);
	}
	if (error == 0)
	spa = dmu_objset_spa(os);
	free(target_pool);
	}
	}
	nvlist_free(policy);

	if (error)
	fatal("can't open '%s': %s", target, strerror(error));

	/*
	* Set the pool failure mode to panic in order to prevent the pool
	* from suspending. A suspended I/O will have no way to resume and
	* can prevent the zdb(8) command from terminating as expected.
	*/
	if (spa != NULL)
	spa->spa_failmode = ZIO_FAILURE_MODE_PANIC;

	argv++;
	argc--;
	if (dump_opt['r']) {
	error = zdb_copy_object(os, object, argv[1]);
	} else if (!dump_opt['R']) {
	flagbits['d'] = ZOR_FLAG_DIRECTORY;
	flagbits['f'] = ZOR_FLAG_PLAIN_FILE;
	flagbits['m'] = ZOR_FLAG_SPACE_MAP;
	flagbits['z'] = ZOR_FLAG_ZAP;
	flagbits['A'] = ZOR_FLAG_ALL_TYPES;

	if (argc > 0 && dump_opt['d']) {
	zopt_object_args = argc;
	zopt_object_ranges = calloc(zopt_object_args,
	sizeof (zopt_object_range_t));
	for (unsigned i = 0; i < zopt_object_args; i++) {
	int err;
	const char *msg = NULL;

	err = parse_object_range(argv[i],
	&zopt_object_ranges[i], &msg);
	if (err != 0)
	fatal("Bad object or range: '%s': %s\n",
	argv[i], msg ?: "");
	}
	} else if (argc > 0 && dump_opt['m']) {
	zopt_metaslab_args = argc;
	zopt_metaslab = calloc(zopt_metaslab_args,
	sizeof (uint64_t));
	for (unsigned i = 0; i < zopt_metaslab_args; i++) {
	errno = 0;
	zopt_metaslab[i] = strtoull(argv[i], NULL, 0);
	if (zopt_metaslab[i] == 0 && errno != 0)
	fatal("bad number %s: %s", argv[i],
	strerror(errno));
	}
	}
	if (os != NULL) {
	dump_objset(os);
	} else if (zopt_object_args > 0 && !dump_opt['m']) {
	dump_objset(spa->spa_meta_objset);
	} else {
	dump_zpool(spa);
	}
	} else {
	flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR;
	flagbits['c'] = ZDB_FLAG_CHECKSUM;
	flagbits['d'] = ZDB_FLAG_DECOMPRESS;
	flagbits['e'] = ZDB_FLAG_BSWAP;
	flagbits['g'] = ZDB_FLAG_GBH;
	flagbits['i'] = ZDB_FLAG_INDIRECT;
	flagbits['r'] = ZDB_FLAG_RAW;
	flagbits['v'] = ZDB_FLAG_VERBOSE;

	for (int i = 0; i < argc; i++)
	zdb_read_block(argv[i], spa);
	}

	if (dump_opt['k']) {
	free(checkpoint_pool);
	if (!target_is_spa)
	free(checkpoint_target);
	}

	if (os != NULL) {
	close_objset(os, FTAG);
	} else {
	spa_close(spa, FTAG);
	}

	fuid_table_destroy();

	dump_debug_buffer();

	kernel_fini();

	return (error);
	}
	diff --git a/sys/contrib/openzfs/cmd/zed/agents/zfs_agents.c b/sys/contrib/openzfs/cmd/zed/agents/zfs_agents.c
	index 2bc84a4f57d1..fb07266dae21 100644
	--- a/sys/contrib/openzfs/cmd/zed/agents/zfs_agents.c
	+++ b/sys/contrib/openzfs/cmd/zed/agents/zfs_agents.c
	@@ -1,434 +1,458 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License Version 1.0 (CDDL-1.0).
	* You can obtain a copy of the license from the top-level file
	* "OPENSOLARIS.LICENSE" or at <http://opensource.org/licenses/CDDL-1.0>.
	* You may not use this file except in compliance with the license.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2016, Intel Corporation.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
	* Copyright (c) 2021 Hewlett Packard Enterprise Development LP
	*/

	#include <libnvpair.h>
	#include <libzfs.h>
	#include <stddef.h>
	#include <stdlib.h>
	#include <string.h>
	#include <sys/list.h>
	#include <sys/time.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/sysevent/dev.h>
	#include <sys/fm/protocol.h>
	#include <sys/fm/fs/zfs.h>
	#include <pthread.h>
	#include <unistd.h>

	#include "zfs_agents.h"
	#include "fmd_api.h"
	#include "../zed_log.h"

	/*
	* agent dispatch code
	*/

	static pthread_mutex_t agent_lock = PTHREAD_MUTEX_INITIALIZER;
	static pthread_cond_t agent_cond = PTHREAD_COND_INITIALIZER;
	static list_t agent_events; /* list of pending events */
	static int agent_exiting;

	typedef struct agent_event {
	char ae_class[64];
	char ae_subclass[32];
	nvlist_t *ae_nvl;
	list_node_t ae_node;
	} agent_event_t;

	pthread_t g_agents_tid;

	libzfs_handle_t *g_zfs_hdl;

	/* guid search data */
	typedef enum device_type {
	DEVICE_TYPE_L2ARC, /* l2arc device */
	DEVICE_TYPE_SPARE, /* spare device */
	DEVICE_TYPE_PRIMARY /* any primary pool storage device */
	} device_type_t;

	typedef struct guid_search {
	uint64_t gs_pool_guid;
	uint64_t gs_vdev_guid;
	char *gs_devid;
	device_type_t gs_vdev_type;
	uint64_t gs_vdev_expandtime; /* vdev expansion time */
	} guid_search_t;

	/*
	* Walks the vdev tree recursively looking for a matching devid.
	* Returns B_TRUE as soon as a matching device is found, B_FALSE otherwise.
	*/
	static boolean_t
	zfs_agent_iter_vdev(zpool_handle_t zhp, nvlist_t nvl, void *arg)
	{
	guid_search_t *gsp = arg;
	char *path = NULL;
	uint_t c, children;
	nvlist_t **child;
	+ uint64_t vdev_guid;

	/*
	* First iterate over any children.
	*/
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
	gsp->gs_vdev_type = DEVICE_TYPE_PRIMARY;
	return (B_TRUE);
	}
	}
	}
	/*
	* Iterate over any spares and cache devices
	*/
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
	- gsp->gs_vdev_type = DEVICE_TYPE_L2ARC;
	+ gsp->gs_vdev_type = DEVICE_TYPE_SPARE;
	return (B_TRUE);
	}
	}
	}
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
	- gsp->gs_vdev_type = DEVICE_TYPE_SPARE;
	+ gsp->gs_vdev_type = DEVICE_TYPE_L2ARC;
	return (B_TRUE);
	}
	}
	}
	/*
	* On a devid match, grab the vdev guid and expansion time, if any.
	*/
	if (gsp->gs_devid != NULL &&
	(nvlist_lookup_string(nvl, ZPOOL_CONFIG_DEVID, &path) == 0) &&
	(strcmp(gsp->gs_devid, path) == 0)) {
	(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID,
	&gsp->gs_vdev_guid);
	(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_EXPANSION_TIME,
	&gsp->gs_vdev_expandtime);
	return (B_TRUE);
	}
	+ /*
	+ * Otherwise, on a vdev guid match, grab the devid and expansion
	+ * time. The devid might be missing on removal since its not part
	+ * of blkid cache and L2ARC VDEV does not contain pool guid in its
	+ * blkid, so this is a special case for L2ARC VDEV.
	+ */
	+ else if (gsp->gs_vdev_guid != 0 && gsp->gs_devid == NULL &&
	+ nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID, &vdev_guid) == 0 &&
	+ gsp->gs_vdev_guid == vdev_guid) {
	+ (void) nvlist_lookup_string(nvl, ZPOOL_CONFIG_DEVID,
	+ &gsp->gs_devid);
	+ (void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_EXPANSION_TIME,
	+ &gsp->gs_vdev_expandtime);
	+ return (B_TRUE);
	+ }

	return (B_FALSE);
	}

	static int
	zfs_agent_iter_pool(zpool_handle_t zhp, void arg)
	{
	guid_search_t *gsp = arg;
	nvlist_t config, nvl;

	/*
	* For each vdev in this pool, look for a match by devid
	*/
	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvl) == 0) {
	(void) zfs_agent_iter_vdev(zhp, nvl, gsp);
	}
	}
	/*
	* if a match was found then grab the pool guid
	*/
	- if (gsp->gs_vdev_guid) {
	+ if (gsp->gs_vdev_guid && gsp->gs_devid) {
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&gsp->gs_pool_guid);
	}

	zpool_close(zhp);
	return (gsp->gs_vdev_guid != 0);
	}

	void
	zfs_agent_post_event(const char class, const char subclass, nvlist_t *nvl)
	{
	agent_event_t *event;

	if (subclass == NULL)
	subclass = "";

	event = malloc(sizeof (agent_event_t));
	if (event == NULL \|\| nvlist_dup(nvl, &event->ae_nvl, 0) != 0) {
	if (event)
	free(event);
	return;
	}

	if (strcmp(class, "sysevent.fs.zfs.vdev_check") == 0) {
	class = EC_ZFS;
	subclass = ESC_ZFS_VDEV_CHECK;
	}

	/*
	* On Linux, we don't get the expected FM_RESOURCE_REMOVED ereport
	* from the vdev_disk layer after a hot unplug. Fortunately we do
	* get an EC_DEV_REMOVE from our disk monitor and it is a suitable
	* proxy so we remap it here for the benefit of the diagnosis engine.
	* Starting in OpenZFS 2.0, we do get FM_RESOURCE_REMOVED from the spa
	* layer. Processing multiple FM_RESOURCE_REMOVED events is not harmful.
	*/
	if ((strcmp(class, EC_DEV_REMOVE) == 0) &&
	(strcmp(subclass, ESC_DISK) == 0) &&
	(nvlist_exists(nvl, ZFS_EV_VDEV_GUID) \|\|
	nvlist_exists(nvl, DEV_IDENTIFIER))) {
	nvlist_t *payload = event->ae_nvl;
	struct timeval tv;
	int64_t tod[2];
	uint64_t pool_guid = 0, vdev_guid = 0;
	guid_search_t search = { 0 };
	device_type_t devtype = DEVICE_TYPE_PRIMARY;
	+ char *devid = NULL;

	class = "resource.fs.zfs.removed";
	subclass = "";

	(void) nvlist_add_string(payload, FM_CLASS, class);
	+ (void) nvlist_lookup_string(nvl, DEV_IDENTIFIER, &devid);
	(void) nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID, &pool_guid);
	(void) nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &vdev_guid);

	(void) gettimeofday(&tv, NULL);
	tod[0] = tv.tv_sec;
	tod[1] = tv.tv_usec;
	(void) nvlist_add_int64_array(payload, FM_EREPORT_TIME, tod, 2);

	/*
	+ * If devid is missing but vdev_guid is available, find devid
	+ * and pool_guid from vdev_guid.
	* For multipath, spare and l2arc devices ZFS_EV_VDEV_GUID or
	* ZFS_EV_POOL_GUID may be missing so find them.
	*/
	- if (pool_guid == 0 \|\| vdev_guid == 0) {
	- if ((nvlist_lookup_string(nvl, DEV_IDENTIFIER,
	- &search.gs_devid) == 0) &&
	- (zpool_iter(g_zfs_hdl, zfs_agent_iter_pool, &search)
	- == 1)) {
	- if (pool_guid == 0)
	- pool_guid = search.gs_pool_guid;
	- if (vdev_guid == 0)
	- vdev_guid = search.gs_vdev_guid;
	- devtype = search.gs_vdev_type;
	- }
	+ if (devid == NULL \|\| pool_guid == 0 \|\| vdev_guid == 0) {
	+ if (devid == NULL)
	+ search.gs_vdev_guid = vdev_guid;
	+ else
	+ search.gs_devid = devid;
	+ zpool_iter(g_zfs_hdl, zfs_agent_iter_pool, &search);
	+ if (devid == NULL)
	+ devid = search.gs_devid;
	+ if (pool_guid == 0)
	+ pool_guid = search.gs_pool_guid;
	+ if (vdev_guid == 0)
	+ vdev_guid = search.gs_vdev_guid;
	+ devtype = search.gs_vdev_type;
	}

	/*
	* We want to avoid reporting "remove" events coming from
	* libudev for VDEVs which were expanded recently (10s) and
	* avoid activating spares in response to partitions being
	* deleted and created in rapid succession.
	*/
	if (search.gs_vdev_expandtime != 0 &&
	search.gs_vdev_expandtime + 10 > tv.tv_sec) {
	zed_log_msg(LOG_INFO, "agent post event: ignoring '%s' "
	"for recently expanded device '%s'", EC_DEV_REMOVE,
	- search.gs_devid);
	+ devid);
	+ fnvlist_free(payload);
	+ free(event);
	goto out;
	}

	(void) nvlist_add_uint64(payload,
	FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, pool_guid);
	(void) nvlist_add_uint64(payload,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vdev_guid);
	switch (devtype) {
	case DEVICE_TYPE_L2ARC:
	(void) nvlist_add_string(payload,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
	VDEV_TYPE_L2CACHE);
	break;
	case DEVICE_TYPE_SPARE:
	(void) nvlist_add_string(payload,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, VDEV_TYPE_SPARE);
	break;
	case DEVICE_TYPE_PRIMARY:
	(void) nvlist_add_string(payload,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, VDEV_TYPE_DISK);
	break;
	}

	zed_log_msg(LOG_INFO, "agent post event: mapping '%s' to '%s'",
	EC_DEV_REMOVE, class);
	}

	(void) strlcpy(event->ae_class, class, sizeof (event->ae_class));
	(void) strlcpy(event->ae_subclass, subclass,
	sizeof (event->ae_subclass));

	(void) pthread_mutex_lock(&agent_lock);
	list_insert_tail(&agent_events, event);
	(void) pthread_mutex_unlock(&agent_lock);

	out:
	(void) pthread_cond_signal(&agent_cond);
	}

	static void
	zfs_agent_dispatch(const char class, const char subclass, nvlist_t *nvl)
	{
	/*
	* The diagnosis engine subscribes to the following events.
	* On illumos these subscriptions reside in:
	* /usr/lib/fm/fmd/plugins/zfs-diagnosis.conf
	*/
	if (strstr(class, "ereport.fs.zfs.") != NULL \|\|
	strstr(class, "resource.fs.zfs.") != NULL \|\|
	strcmp(class, "sysevent.fs.zfs.vdev_remove") == 0 \|\|
	strcmp(class, "sysevent.fs.zfs.vdev_remove_dev") == 0 \|\|
	strcmp(class, "sysevent.fs.zfs.pool_destroy") == 0) {
	fmd_module_recv(fmd_module_hdl("zfs-diagnosis"), nvl, class);
	}

	/*
	* The retire agent subscribes to the following events.
	* On illumos these subscriptions reside in:
	* /usr/lib/fm/fmd/plugins/zfs-retire.conf
	*
	* NOTE: faults events come directly from our diagnosis engine
	* and will not pass through the zfs kernel module.
	*/
	if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0 \|\|
	strcmp(class, "resource.fs.zfs.removed") == 0 \|\|
	strcmp(class, "resource.fs.zfs.statechange") == 0 \|\|
	strcmp(class, "sysevent.fs.zfs.vdev_remove") == 0) {
	fmd_module_recv(fmd_module_hdl("zfs-retire"), nvl, class);
	}

	/*
	* The SLM module only consumes disk events and vdev check events
	*
	* NOTE: disk events come directly from disk monitor and will
	* not pass through the zfs kernel module.
	*/
	if (strstr(class, "EC_dev_") != NULL \|\|
	strcmp(class, EC_ZFS) == 0) {
	(void) zfs_slm_event(class, subclass, nvl);
	}
	}

	/*
	* Events are consumed and dispatched from this thread
	* An agent can also post an event so event list lock
	* is not held when calling an agent.
	* One event is consumed at a time.
	*/
	static void *
	zfs_agent_consumer_thread(void *arg)
	{
	(void) arg;

	for (;;) {
	agent_event_t *event;

	(void) pthread_mutex_lock(&agent_lock);

	/* wait for an event to show up */
	while (!agent_exiting && list_is_empty(&agent_events))
	(void) pthread_cond_wait(&agent_cond, &agent_lock);

	if (agent_exiting) {
	(void) pthread_mutex_unlock(&agent_lock);
	zed_log_msg(LOG_INFO, "zfs_agent_consumer_thread: "
	"exiting");
	return (NULL);
	}

	if ((event = (list_head(&agent_events))) != NULL) {
	list_remove(&agent_events, event);

	(void) pthread_mutex_unlock(&agent_lock);

	/* dispatch to all event subscribers */
	zfs_agent_dispatch(event->ae_class, event->ae_subclass,
	event->ae_nvl);

	nvlist_free(event->ae_nvl);
	free(event);
	continue;
	}

	(void) pthread_mutex_unlock(&agent_lock);
	}

	return (NULL);
	}

	void
	zfs_agent_init(libzfs_handle_t *zfs_hdl)
	{
	fmd_hdl_t *hdl;

	g_zfs_hdl = zfs_hdl;

	if (zfs_slm_init() != 0)
	zed_log_die("Failed to initialize zfs slm");
	zed_log_msg(LOG_INFO, "Add Agent: init");

	hdl = fmd_module_hdl("zfs-diagnosis");
	_zfs_diagnosis_init(hdl);
	if (!fmd_module_initialized(hdl))
	zed_log_die("Failed to initialize zfs diagnosis");

	hdl = fmd_module_hdl("zfs-retire");
	_zfs_retire_init(hdl);
	if (!fmd_module_initialized(hdl))
	zed_log_die("Failed to initialize zfs retire");

	list_create(&agent_events, sizeof (agent_event_t),
	offsetof(struct agent_event, ae_node));

	if (pthread_create(&g_agents_tid, NULL, zfs_agent_consumer_thread,
	NULL) != 0) {
	list_destroy(&agent_events);
	zed_log_die("Failed to initialize agents");
	}
	pthread_setname_np(g_agents_tid, "agents");
	}

	void
	zfs_agent_fini(void)
	{
	fmd_hdl_t *hdl;
	agent_event_t *event;

	agent_exiting = 1;
	(void) pthread_cond_signal(&agent_cond);

	/* wait for zfs_enum_pools thread to complete */
	(void) pthread_join(g_agents_tid, NULL);

	/* drain any pending events */
	while ((event = (list_head(&agent_events))) != NULL) {
	list_remove(&agent_events, event);
	nvlist_free(event->ae_nvl);
	free(event);
	}

	list_destroy(&agent_events);

	if ((hdl = fmd_module_hdl("zfs-retire")) != NULL) {
	_zfs_retire_fini(hdl);
	fmd_hdl_unregister(hdl);
	}
	if ((hdl = fmd_module_hdl("zfs-diagnosis")) != NULL) {
	_zfs_diagnosis_fini(hdl);
	fmd_hdl_unregister(hdl);
	}

	zed_log_msg(LOG_INFO, "Add Agent: fini");
	zfs_slm_fini();

	g_zfs_hdl = NULL;
	}
	diff --git a/sys/contrib/openzfs/cmd/zed/agents/zfs_retire.c b/sys/contrib/openzfs/cmd/zed/agents/zfs_retire.c
	index a9e8baaa2c54..f4b6dff48176 100644
	--- a/sys/contrib/openzfs/cmd/zed/agents/zfs_retire.c
	+++ b/sys/contrib/openzfs/cmd/zed/agents/zfs_retire.c
	@@ -1,565 +1,588 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
	*
	* Copyright (c) 2016, Intel Corporation.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
	*/

	/*
	* The ZFS retire agent is responsible for managing hot spares across all pools.
	* When we see a device fault or a device removal, we try to open the associated
	* pool and look for any hot spares. We iterate over any available hot spares
	* and attempt a 'zpool replace' for each one.
	*
	* For vdevs diagnosed as faulty, the agent is also responsible for proactively
	* marking the vdev FAULTY (for I/O errors) or DEGRADED (for checksum errors).
	*/

	#include <sys/fs/zfs.h>
	#include <sys/fm/protocol.h>
	#include <sys/fm/fs/zfs.h>
	#include <libzutil.h>
	#include <libzfs.h>
	#include <string.h>
	#include <libgen.h>

	#include "zfs_agents.h"
	#include "fmd_api.h"


	typedef struct zfs_retire_repaired {
	struct zfs_retire_repaired *zrr_next;
	uint64_t zrr_pool;
	uint64_t zrr_vdev;
	} zfs_retire_repaired_t;

	typedef struct zfs_retire_data {
	libzfs_handle_t *zrd_hdl;
	zfs_retire_repaired_t *zrd_repaired;
	} zfs_retire_data_t;

	static void
	zfs_retire_clear_data(fmd_hdl_t hdl, zfs_retire_data_t zdp)
	{
	zfs_retire_repaired_t *zrp;

	while ((zrp = zdp->zrd_repaired) != NULL) {
	zdp->zrd_repaired = zrp->zrr_next;
	fmd_hdl_free(hdl, zrp, sizeof (zfs_retire_repaired_t));
	}
	}

	/*
	* Find a pool with a matching GUID.
	*/
	typedef struct find_cbdata {
	uint64_t cb_guid;
	zpool_handle_t *cb_zhp;
	nvlist_t *cb_vdev;
	} find_cbdata_t;

	static int
	find_pool(zpool_handle_t zhp, void data)
	{
	find_cbdata_t *cbp = data;

	if (cbp->cb_guid ==
	zpool_get_prop_int(zhp, ZPOOL_PROP_GUID, NULL)) {
	cbp->cb_zhp = zhp;
	return (1);
	}

	zpool_close(zhp);
	return (0);
	}

	/*
	* Find a vdev within a tree with a matching GUID.
	*/
	static nvlist_t *
	find_vdev(libzfs_handle_t zhdl, nvlist_t nv, uint64_t search_guid)
	{
	uint64_t guid;
	nvlist_t **child;
	uint_t c, children;
	nvlist_t *ret;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
	guid == search_guid) {
	fmd_hdl_debug(fmd_module_hdl("zfs-retire"),
	"matched vdev %llu", guid);
	return (nv);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return (NULL);

	for (c = 0; c < children; c++) {
	if ((ret = find_vdev(zhdl, child[c], search_guid)) != NULL)
	return (ret);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) != 0)
	return (NULL);

	for (c = 0; c < children; c++) {
	if ((ret = find_vdev(zhdl, child[c], search_guid)) != NULL)
	return (ret);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) != 0)
	return (NULL);

	for (c = 0; c < children; c++) {
	if ((ret = find_vdev(zhdl, child[c], search_guid)) != NULL)
	return (ret);
	}

	return (NULL);
	}

	/*
	* Given a (pool, vdev) GUID pair, find the matching pool and vdev.
	*/
	static zpool_handle_t *
	find_by_guid(libzfs_handle_t *zhdl, uint64_t pool_guid, uint64_t vdev_guid,
	nvlist_t **vdevp)
	{
	find_cbdata_t cb;
	zpool_handle_t *zhp;
	nvlist_t config, nvroot;

	/*
	* Find the corresponding pool and make sure the vdev still exists.
	*/
	cb.cb_guid = pool_guid;
	if (zpool_iter(zhdl, find_pool, &cb) != 1)
	return (NULL);

	zhp = cb.cb_zhp;
	config = zpool_get_config(zhp, NULL);
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) != 0) {
	zpool_close(zhp);
	return (NULL);
	}

	if (vdev_guid != 0) {
	if ((*vdevp = find_vdev(zhdl, nvroot, vdev_guid)) == NULL) {
	zpool_close(zhp);
	return (NULL);
	}
	}

	return (zhp);
	}

	/*
	* Given a vdev, attempt to replace it with every known spare until one
	* succeeds or we run out of devices to try.
	* Return whether we were successful or not in replacing the device.
	*/
	static boolean_t
	replace_with_spare(fmd_hdl_t hdl, zpool_handle_t zhp, nvlist_t *vdev)
	{
	nvlist_t config, nvroot, *replacement;
	nvlist_t **spares;
	uint_t s, nspares;
	char *dev_name;
	zprop_source_t source;
	int ashift;

	config = zpool_get_config(zhp, NULL);
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) != 0)
	return (B_FALSE);

	/*
	* Find out if there are any hot spares available in the pool.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) != 0)
	return (B_FALSE);

	/*
	* lookup "ashift" pool property, we may need it for the replacement
	*/
	ashift = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &source);

	replacement = fmd_nvl_alloc(hdl, FMD_SLEEP);

	(void) nvlist_add_string(replacement, ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_ROOT);

	dev_name = zpool_vdev_name(NULL, zhp, vdev, B_FALSE);

	/*
	* Try to replace each spare, ending when we successfully
	* replace it.
	*/
	for (s = 0; s < nspares; s++) {
	boolean_t rebuild = B_FALSE;
	char spare_name, type;

	if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
	&spare_name) != 0)
	continue;

	/* prefer sequential resilvering for distributed spares */
	if ((nvlist_lookup_string(spares[s], ZPOOL_CONFIG_TYPE,
	&type) == 0) && strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0)
	rebuild = B_TRUE;

	/* if set, add the "ashift" pool property to the spare nvlist */
	if (source != ZPROP_SRC_DEFAULT)
	(void) nvlist_add_uint64(spares[s],
	ZPOOL_CONFIG_ASHIFT, ashift);

	(void) nvlist_add_nvlist_array(replacement,
	ZPOOL_CONFIG_CHILDREN, (const nvlist_t **)&spares[s], 1);

	fmd_hdl_debug(hdl, "zpool_vdev_replace '%s' with spare '%s'",
	dev_name, zfs_basename(spare_name));

	if (zpool_vdev_attach(zhp, dev_name, spare_name,
	replacement, B_TRUE, rebuild) == 0) {
	free(dev_name);
	nvlist_free(replacement);
	return (B_TRUE);
	}
	}

	free(dev_name);
	nvlist_free(replacement);

	return (B_FALSE);
	}

	/*
	* Repair this vdev if we had diagnosed a 'fault.fs.zfs.device' and
	* ASRU is now usable. ZFS has found the device to be present and
	* functioning.
	*/
	static void
	zfs_vdev_repair(fmd_hdl_t hdl, nvlist_t nvl)
	{
	zfs_retire_data_t *zdp = fmd_hdl_getspecific(hdl);
	zfs_retire_repaired_t *zrp;
	uint64_t pool_guid, vdev_guid;
	if (nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
	&pool_guid) != 0 \|\| nvlist_lookup_uint64(nvl,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, &vdev_guid) != 0)
	return;

	/*
	* Before checking the state of the ASRU, go through and see if we've
	* already made an attempt to repair this ASRU. This list is cleared
	* whenever we receive any kind of list event, and is designed to
	* prevent us from generating a feedback loop when we attempt repairs
	* against a faulted pool. The problem is that checking the unusable
	* state of the ASRU can involve opening the pool, which can post
	* statechange events but otherwise leave the pool in the faulted
	* state. This list allows us to detect when a statechange event is
	* due to our own request.
	*/
	for (zrp = zdp->zrd_repaired; zrp != NULL; zrp = zrp->zrr_next) {
	if (zrp->zrr_pool == pool_guid &&
	zrp->zrr_vdev == vdev_guid)
	return;
	}

	zrp = fmd_hdl_alloc(hdl, sizeof (zfs_retire_repaired_t), FMD_SLEEP);
	zrp->zrr_next = zdp->zrd_repaired;
	zrp->zrr_pool = pool_guid;
	zrp->zrr_vdev = vdev_guid;
	zdp->zrd_repaired = zrp;

	fmd_hdl_debug(hdl, "marking repaired vdev %llu on pool %llu",
	vdev_guid, pool_guid);
	}

	static void
	zfs_retire_recv(fmd_hdl_t hdl, fmd_event_t ep, nvlist_t *nvl,
	const char *class)
	{
	(void) ep;
	uint64_t pool_guid, vdev_guid;
	zpool_handle_t *zhp;
	nvlist_t resource, fault;
	nvlist_t **faults;
	uint_t f, nfaults;
	zfs_retire_data_t *zdp = fmd_hdl_getspecific(hdl);
	libzfs_handle_t *zhdl = zdp->zrd_hdl;
	boolean_t fault_device, degrade_device;
	boolean_t is_repair;
	char *scheme;
	nvlist_t *vdev = NULL;
	char *uuid;
	int repair_done = 0;
	boolean_t retire;
	boolean_t is_disk;
	vdev_aux_t aux;
	uint64_t state = 0;
	+ int l2arc;
	+ vdev_stat_t *vs;
	+ unsigned int c;

	fmd_hdl_debug(hdl, "zfs_retire_recv: '%s'", class);

	- nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, &state);
	+ (void) nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE,
	+ &state);

	/*
	* If this is a resource notifying us of device removal then simply
	* check for an available spare and continue unless the device is a
	* l2arc vdev, in which case we just offline it.
	*/
	if (strcmp(class, "resource.fs.zfs.removed") == 0 \|\|
	(strcmp(class, "resource.fs.zfs.statechange") == 0 &&
	(state == VDEV_STATE_REMOVED \|\| state == VDEV_STATE_FAULTED))) {
	char *devtype;
	char *devname;

	if (nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
	&pool_guid) != 0 \|\|
	nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
	&vdev_guid) != 0)
	return;

	if ((zhp = find_by_guid(zhdl, pool_guid, vdev_guid,
	&vdev)) == NULL)
	return;

	devname = zpool_vdev_name(NULL, zhp, vdev, B_FALSE);

	- /* Can't replace l2arc with a spare: offline the device */
	- if (nvlist_lookup_string(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
	- &devtype) == 0 && strcmp(devtype, VDEV_TYPE_L2CACHE) == 0) {
	- fmd_hdl_debug(hdl, "zpool_vdev_offline '%s'", devname);
	- zpool_vdev_offline(zhp, devname, B_TRUE);
	- } else if (!fmd_prop_get_int32(hdl, "spare_on_remove") \|\|
	- replace_with_spare(hdl, zhp, vdev) == B_FALSE) {
	+ nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
	+ (uint64_t **)&vs, &c);
	+
	+ /*
	+ * If state removed is requested for already removed vdev,
	+ * its a loopback event from spa_async_remove(). Just
	+ * ignore it.
	+ */
	+ if (vs->vs_state == VDEV_STATE_REMOVED &&
	+ state == VDEV_STATE_REMOVED)
	+ return;
	+
	+ l2arc = (nvlist_lookup_string(nvl,
	+ FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, &devtype) == 0 &&
	+ strcmp(devtype, VDEV_TYPE_L2CACHE) == 0);
	+
	+ /* Remove the vdev since device is unplugged */
	+ if (l2arc \|\| (strcmp(class, "resource.fs.zfs.removed") == 0)) {
	+ int status = zpool_vdev_remove_wanted(zhp, devname);
	+ fmd_hdl_debug(hdl, "zpool_vdev_remove_wanted '%s'"
	+ ", ret:%d", devname, status);
	+ }
	+
	+ /* Replace the vdev with a spare if its not a l2arc */
	+ if (!l2arc && (!fmd_prop_get_int32(hdl, "spare_on_remove") \|\|
	+ replace_with_spare(hdl, zhp, vdev) == B_FALSE)) {
	/* Could not handle with spare */
	fmd_hdl_debug(hdl, "no spare for '%s'", devname);
	}

	free(devname);
	zpool_close(zhp);
	return;
	}

	if (strcmp(class, FM_LIST_RESOLVED_CLASS) == 0)
	return;

	/*
	* Note: on Linux statechange events are more than just
	* healthy ones so we need to confirm the actual state value.
	*/
	if (strcmp(class, "resource.fs.zfs.statechange") == 0 &&
	state == VDEV_STATE_HEALTHY) {
	zfs_vdev_repair(hdl, nvl);
	return;
	}
	if (strcmp(class, "sysevent.fs.zfs.vdev_remove") == 0) {
	zfs_vdev_repair(hdl, nvl);
	return;
	}

	zfs_retire_clear_data(hdl, zdp);

	if (strcmp(class, FM_LIST_REPAIRED_CLASS) == 0)
	is_repair = B_TRUE;
	else
	is_repair = B_FALSE;

	/*
	* We subscribe to zfs faults as well as all repair events.
	*/
	if (nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST,
	&faults, &nfaults) != 0)
	return;

	for (f = 0; f < nfaults; f++) {
	fault = faults[f];

	fault_device = B_FALSE;
	degrade_device = B_FALSE;
	is_disk = B_FALSE;

	if (nvlist_lookup_boolean_value(fault, FM_SUSPECT_RETIRE,
	&retire) == 0 && retire == 0)
	continue;

	/*
	* While we subscribe to fault.fs.zfs.*, we only take action
	* for faults targeting a specific vdev (open failure or SERD
	* failure). We also subscribe to fault.io.* events, so that
	* faulty disks will be faulted in the ZFS configuration.
	*/
	if (fmd_nvl_class_match(hdl, fault, "fault.fs.zfs.vdev.io")) {
	fault_device = B_TRUE;
	} else if (fmd_nvl_class_match(hdl, fault,
	"fault.fs.zfs.vdev.checksum")) {
	degrade_device = B_TRUE;
	} else if (fmd_nvl_class_match(hdl, fault,
	"fault.fs.zfs.device")) {
	fault_device = B_FALSE;
	} else if (fmd_nvl_class_match(hdl, fault, "fault.io.*")) {
	is_disk = B_TRUE;
	fault_device = B_TRUE;
	} else {
	continue;
	}

	if (is_disk) {
	continue;
	} else {
	/*
	* This is a ZFS fault. Lookup the resource, and
	* attempt to find the matching vdev.
	*/
	if (nvlist_lookup_nvlist(fault, FM_FAULT_RESOURCE,
	&resource) != 0 \|\|
	nvlist_lookup_string(resource, FM_FMRI_SCHEME,
	&scheme) != 0)
	continue;

	if (strcmp(scheme, FM_FMRI_SCHEME_ZFS) != 0)
	continue;

	if (nvlist_lookup_uint64(resource, FM_FMRI_ZFS_POOL,
	&pool_guid) != 0)
	continue;

	if (nvlist_lookup_uint64(resource, FM_FMRI_ZFS_VDEV,
	&vdev_guid) != 0) {
	if (is_repair)
	vdev_guid = 0;
	else
	continue;
	}

	if ((zhp = find_by_guid(zhdl, pool_guid, vdev_guid,
	&vdev)) == NULL)
	continue;

	aux = VDEV_AUX_ERR_EXCEEDED;
	}

	if (vdev_guid == 0) {
	/*
	* For pool-level repair events, clear the entire pool.
	*/
	fmd_hdl_debug(hdl, "zpool_clear of pool '%s'",
	zpool_get_name(zhp));
	(void) zpool_clear(zhp, NULL, NULL);
	zpool_close(zhp);
	continue;
	}

	/*
	* If this is a repair event, then mark the vdev as repaired and
	* continue.
	*/
	if (is_repair) {
	repair_done = 1;
	fmd_hdl_debug(hdl, "zpool_clear of pool '%s' vdev %llu",
	zpool_get_name(zhp), vdev_guid);
	(void) zpool_vdev_clear(zhp, vdev_guid);
	zpool_close(zhp);
	continue;
	}

	/*
	* Actively fault the device if needed.
	*/
	if (fault_device)
	(void) zpool_vdev_fault(zhp, vdev_guid, aux);
	if (degrade_device)
	(void) zpool_vdev_degrade(zhp, vdev_guid, aux);

	if (fault_device \|\| degrade_device)
	fmd_hdl_debug(hdl, "zpool_vdev_%s: vdev %llu on '%s'",
	fault_device ? "fault" : "degrade", vdev_guid,
	zpool_get_name(zhp));

	/*
	* Attempt to substitute a hot spare.
	*/
	(void) replace_with_spare(hdl, zhp, vdev);

	zpool_close(zhp);
	}

	if (strcmp(class, FM_LIST_REPAIRED_CLASS) == 0 && repair_done &&
	nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) == 0)
	fmd_case_uuresolved(hdl, uuid);
	}

	static const fmd_hdl_ops_t fmd_ops = {
	zfs_retire_recv, /* fmdo_recv */
	NULL, /* fmdo_timeout */
	NULL, /* fmdo_close */
	NULL, /* fmdo_stats */
	NULL, /* fmdo_gc */
	};

	static const fmd_prop_t fmd_props[] = {
	{ "spare_on_remove", FMD_TYPE_BOOL, "true" },
	{ NULL, 0, NULL }
	};

	static const fmd_hdl_info_t fmd_info = {
	"ZFS Retire Agent", "1.0", &fmd_ops, fmd_props
	};

	void
	_zfs_retire_init(fmd_hdl_t *hdl)
	{
	zfs_retire_data_t *zdp;
	libzfs_handle_t *zhdl;

	if ((zhdl = libzfs_init()) == NULL)
	return;

	if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) {
	libzfs_fini(zhdl);
	return;
	}

	zdp = fmd_hdl_zalloc(hdl, sizeof (zfs_retire_data_t), FMD_SLEEP);
	zdp->zrd_hdl = zhdl;

	fmd_hdl_setspecific(hdl, zdp);
	}

	void
	_zfs_retire_fini(fmd_hdl_t *hdl)
	{
	zfs_retire_data_t *zdp = fmd_hdl_getspecific(hdl);

	if (zdp != NULL) {
	zfs_retire_clear_data(hdl, zdp);
	libzfs_fini(zdp->zrd_hdl);
	fmd_hdl_free(hdl, zdp, sizeof (zfs_retire_data_t));
	}
	}
	diff --git a/sys/contrib/openzfs/cmd/zed/zed_log.h b/sys/contrib/openzfs/cmd/zed/zed_log.h
	index ed88ad41d7e2..18ab0553ea40 100644
	--- a/sys/contrib/openzfs/cmd/zed/zed_log.h
	+++ b/sys/contrib/openzfs/cmd/zed/zed_log.h
	@@ -1,44 +1,45 @@
	/*
	* This file is part of the ZFS Event Daemon (ZED).
	*
	* Developed at Lawrence Livermore National Laboratory (LLNL-CODE-403049).
	* Copyright (C) 2013-2014 Lawrence Livermore National Security, LLC.
	* Refer to the OpenZFS git commit log for authoritative copyright attribution.
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License Version 1.0 (CDDL-1.0).
	* You can obtain a copy of the license from the top-level file
	* "OPENSOLARIS.LICENSE" or at <http://opensource.org/licenses/CDDL-1.0>.
	* You may not use this file except in compliance with the license.
	*/

	#ifndef ZED_LOG_H
	#define ZED_LOG_H

	#include <syslog.h>

	void zed_log_init(const char *identity);

	void zed_log_fini(void);

	void zed_log_pipe_open(void);

	void zed_log_pipe_close_reads(void);

	void zed_log_pipe_close_writes(void);

	void zed_log_pipe_wait(void);

	void zed_log_stderr_open(int priority);

	void zed_log_stderr_close(void);

	void zed_log_syslog_open(int facility);

	void zed_log_syslog_close(void);

	void zed_log_msg(int priority, const char *fmt, ...);

	+__attribute__((format(printf, 1, 2), __noreturn__))
	void zed_log_die(const char *fmt, ...);

	#endif /* !ZED_LOG_H */
	diff --git a/sys/contrib/openzfs/cmd/zfs/zfs_main.c b/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	index 008f1bea0ec9..7bda3f5292b3 100644
	--- a/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	+++ b/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	@@ -1,8820 +1,8824 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2012 Milan Jurik. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
	* Copyright 2019 Joyent, Inc.
	* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <sys/debug.h>
	#include <errno.h>
	#include <getopt.h>
	#include <libgen.h>
	#include <libintl.h>
	#include <libuutil.h>
	#include <libnvpair.h>
	#include <locale.h>
	#include <stddef.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <zone.h>
	#include <grp.h>
	#include <pwd.h>
	#include <umem.h>
	#include <pthread.h>
	#include <signal.h>
	#include <sys/list.h>
	#include <sys/mkdev.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/mount.h>
	#include <sys/stat.h>
	#include <sys/fs/zfs.h>
	#include <sys/systeminfo.h>
	#include <sys/types.h>
	#include <time.h>
	#include <sys/zfs_project.h>

	#include <libzfs.h>
	#include <libzfs_core.h>
	#include <zfs_prop.h>
	#include <zfs_deleg.h>
	#include <libzutil.h>
	#ifdef HAVE_IDMAP
	#include <aclutils.h>
	#include <directory.h>
	#endif /* HAVE_IDMAP */

	#include "zfs_iter.h"
	#include "zfs_util.h"
	#include "zfs_comutil.h"
	#include "zfs_projectutil.h"

	libzfs_handle_t *g_zfs;

	static char history_str[HIS_MAX_RECORD_LEN];
	static boolean_t log_history = B_TRUE;

	static int zfs_do_clone(int argc, char **argv);
	static int zfs_do_create(int argc, char **argv);
	static int zfs_do_destroy(int argc, char **argv);
	static int zfs_do_get(int argc, char **argv);
	static int zfs_do_inherit(int argc, char **argv);
	static int zfs_do_list(int argc, char **argv);
	static int zfs_do_mount(int argc, char **argv);
	static int zfs_do_rename(int argc, char **argv);
	static int zfs_do_rollback(int argc, char **argv);
	static int zfs_do_set(int argc, char **argv);
	static int zfs_do_upgrade(int argc, char **argv);
	static int zfs_do_snapshot(int argc, char **argv);
	static int zfs_do_unmount(int argc, char **argv);
	static int zfs_do_share(int argc, char **argv);
	static int zfs_do_unshare(int argc, char **argv);
	static int zfs_do_send(int argc, char **argv);
	static int zfs_do_receive(int argc, char **argv);
	static int zfs_do_promote(int argc, char **argv);
	static int zfs_do_userspace(int argc, char **argv);
	static int zfs_do_allow(int argc, char **argv);
	static int zfs_do_unallow(int argc, char **argv);
	static int zfs_do_hold(int argc, char **argv);
	static int zfs_do_holds(int argc, char **argv);
	static int zfs_do_release(int argc, char **argv);
	static int zfs_do_diff(int argc, char **argv);
	static int zfs_do_bookmark(int argc, char **argv);
	static int zfs_do_channel_program(int argc, char **argv);
	static int zfs_do_load_key(int argc, char **argv);
	static int zfs_do_unload_key(int argc, char **argv);
	static int zfs_do_change_key(int argc, char **argv);
	static int zfs_do_project(int argc, char **argv);
	static int zfs_do_version(int argc, char **argv);
	static int zfs_do_redact(int argc, char **argv);
	static int zfs_do_wait(int argc, char **argv);

	#ifdef __FreeBSD__
	static int zfs_do_jail(int argc, char **argv);
	static int zfs_do_unjail(int argc, char **argv);
	#endif

	#ifdef __linux__
	static int zfs_do_zone(int argc, char **argv);
	static int zfs_do_unzone(int argc, char **argv);
	#endif

	/*
	* Enable a reasonable set of defaults for libumem debugging on DEBUG builds.
	*/

	#ifdef DEBUG
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

	const char *
	_umem_logging_init(void)
	{
	return ("fail,contents"); /* $UMEM_LOGGING setting */
	}
	#endif

	typedef enum {
	HELP_CLONE,
	HELP_CREATE,
	HELP_DESTROY,
	HELP_GET,
	HELP_INHERIT,
	HELP_UPGRADE,
	HELP_LIST,
	HELP_MOUNT,
	HELP_PROMOTE,
	HELP_RECEIVE,
	HELP_RENAME,
	HELP_ROLLBACK,
	HELP_SEND,
	HELP_SET,
	HELP_SHARE,
	HELP_SNAPSHOT,
	HELP_UNMOUNT,
	HELP_UNSHARE,
	HELP_ALLOW,
	HELP_UNALLOW,
	HELP_USERSPACE,
	HELP_GROUPSPACE,
	HELP_PROJECTSPACE,
	HELP_PROJECT,
	HELP_HOLD,
	HELP_HOLDS,
	HELP_RELEASE,
	HELP_DIFF,
	HELP_BOOKMARK,
	HELP_CHANNEL_PROGRAM,
	HELP_LOAD_KEY,
	HELP_UNLOAD_KEY,
	HELP_CHANGE_KEY,
	HELP_VERSION,
	HELP_REDACT,
	HELP_JAIL,
	HELP_UNJAIL,
	HELP_WAIT,
	HELP_ZONE,
	HELP_UNZONE,
	} zfs_help_t;

	typedef struct zfs_command {
	const char *name;
	int (func)(int argc, char *argv);
	zfs_help_t usage;
	} zfs_command_t;

	/*
	* Master command table. Each ZFS command has a name, associated function, and
	* usage message. The usage messages need to be internationalized, so we have
	* to have a function to return the usage message based on a command index.
	*
	* These commands are organized according to how they are displayed in the usage
	* message. An empty command (one with a NULL name) indicates an empty line in
	* the generic usage message.
	*/
	static zfs_command_t command_table[] = {
	{ "version", zfs_do_version, HELP_VERSION },
	{ NULL },
	{ "create", zfs_do_create, HELP_CREATE },
	{ "destroy", zfs_do_destroy, HELP_DESTROY },
	{ NULL },
	{ "snapshot", zfs_do_snapshot, HELP_SNAPSHOT },
	{ "rollback", zfs_do_rollback, HELP_ROLLBACK },
	{ "clone", zfs_do_clone, HELP_CLONE },
	{ "promote", zfs_do_promote, HELP_PROMOTE },
	{ "rename", zfs_do_rename, HELP_RENAME },
	{ "bookmark", zfs_do_bookmark, HELP_BOOKMARK },
	{ "program", zfs_do_channel_program, HELP_CHANNEL_PROGRAM },
	{ NULL },
	{ "list", zfs_do_list, HELP_LIST },
	{ NULL },
	{ "set", zfs_do_set, HELP_SET },
	{ "get", zfs_do_get, HELP_GET },
	{ "inherit", zfs_do_inherit, HELP_INHERIT },
	{ "upgrade", zfs_do_upgrade, HELP_UPGRADE },
	{ NULL },
	{ "userspace", zfs_do_userspace, HELP_USERSPACE },
	{ "groupspace", zfs_do_userspace, HELP_GROUPSPACE },
	{ "projectspace", zfs_do_userspace, HELP_PROJECTSPACE },
	{ NULL },
	{ "project", zfs_do_project, HELP_PROJECT },
	{ NULL },
	{ "mount", zfs_do_mount, HELP_MOUNT },
	{ "unmount", zfs_do_unmount, HELP_UNMOUNT },
	{ "share", zfs_do_share, HELP_SHARE },
	{ "unshare", zfs_do_unshare, HELP_UNSHARE },
	{ NULL },
	{ "send", zfs_do_send, HELP_SEND },
	{ "receive", zfs_do_receive, HELP_RECEIVE },
	{ NULL },
	{ "allow", zfs_do_allow, HELP_ALLOW },
	{ NULL },
	{ "unallow", zfs_do_unallow, HELP_UNALLOW },
	{ NULL },
	{ "hold", zfs_do_hold, HELP_HOLD },
	{ "holds", zfs_do_holds, HELP_HOLDS },
	{ "release", zfs_do_release, HELP_RELEASE },
	{ "diff", zfs_do_diff, HELP_DIFF },
	{ "load-key", zfs_do_load_key, HELP_LOAD_KEY },
	{ "unload-key", zfs_do_unload_key, HELP_UNLOAD_KEY },
	{ "change-key", zfs_do_change_key, HELP_CHANGE_KEY },
	{ "redact", zfs_do_redact, HELP_REDACT },
	{ "wait", zfs_do_wait, HELP_WAIT },

	#ifdef __FreeBSD__
	{ "jail", zfs_do_jail, HELP_JAIL },
	{ "unjail", zfs_do_unjail, HELP_UNJAIL },
	#endif

	#ifdef __linux__
	{ "zone", zfs_do_zone, HELP_ZONE },
	{ "unzone", zfs_do_unzone, HELP_UNZONE },
	#endif
	};

	#define NCOMMAND (sizeof (command_table) / sizeof (command_table[0]))

	zfs_command_t *current_command;

	static const char *
	get_usage(zfs_help_t idx)
	{
	switch (idx) {
	case HELP_CLONE:
	return (gettext("\tclone [-p] [-o property=value] ... "
	"<snapshot> <filesystem\|volume>\n"));
	case HELP_CREATE:
	return (gettext("\tcreate [-Pnpuv] [-o property=value] ... "
	"<filesystem>\n"
	"\tcreate [-Pnpsv] [-b blocksize] [-o property=value] ... "
	"-V <size> <volume>\n"));
	case HELP_DESTROY:
	return (gettext("\tdestroy [-fnpRrv] <filesystem\|volume>\n"
	"\tdestroy [-dnpRrv] "
	"<filesystem\|volume>@<snap>[%<snap>][,...]\n"
	"\tdestroy <filesystem\|volume>#<bookmark>\n"));
	case HELP_GET:
	return (gettext("\tget [-rHp] [-d max] "
	"[-o \"all\" \| field[,...]]\n"
	"\t [-t type[,...]] [-s source[,...]]\n"
	"\t <\"all\" \| property[,...]> "
	"[filesystem\|volume\|snapshot\|bookmark] ...\n"));
	case HELP_INHERIT:
	return (gettext("\tinherit [-rS] <property> "
	"<filesystem\|volume\|snapshot> ...\n"));
	case HELP_UPGRADE:
	return (gettext("\tupgrade [-v]\n"
	"\tupgrade [-r] [-V version] <-a \| filesystem ...>\n"));
	case HELP_LIST:
	return (gettext("\tlist [-Hp] [-r\|-d max] [-o property[,...]] "
	"[-s property]...\n\t [-S property]... [-t type[,...]] "
	"[filesystem\|volume\|snapshot] ...\n"));
	case HELP_MOUNT:
	return (gettext("\tmount\n"
	"\tmount [-flvO] [-o opts] <-a \| filesystem>\n"));
	case HELP_PROMOTE:
	return (gettext("\tpromote <clone-filesystem>\n"));
	case HELP_RECEIVE:
	return (gettext("\treceive [-vMnsFhu] "
	"[-o <property>=<value>] ... [-x <property>] ...\n"
	"\t <filesystem\|volume\|snapshot>\n"
	"\treceive [-vMnsFhu] [-o <property>=<value>] ... "
	"[-x <property>] ... \n"
	"\t [-d \| -e] <filesystem>\n"
	"\treceive -A <filesystem\|volume>\n"));
	case HELP_RENAME:
	return (gettext("\trename [-f] <filesystem\|volume\|snapshot> "
	"<filesystem\|volume\|snapshot>\n"
	"\trename -p [-f] <filesystem\|volume> <filesystem\|volume>\n"
	"\trename -u [-f] <filesystem> <filesystem>\n"
	"\trename -r <snapshot> <snapshot>\n"));
	case HELP_ROLLBACK:
	return (gettext("\trollback [-rRf] <snapshot>\n"));
	case HELP_SEND:
	return (gettext("\tsend [-DLPbcehnpsvw] "
	"[-i\|-I snapshot]\n"
	"\t [-R [-X dataset[,dataset]...]] <snapshot>\n"
	"\tsend [-DnvPLecw] [-i snapshot\|bookmark] "
	"<filesystem\|volume\|snapshot>\n"
	"\tsend [-DnPpvLec] [-i bookmark\|snapshot] "
	"--redact <bookmark> <snapshot>\n"
	"\tsend [-nvPe] -t <receive_resume_token>\n"
	"\tsend [-Pnv] --saved filesystem\n"));
	case HELP_SET:
	return (gettext("\tset <property=value> ... "
	"<filesystem\|volume\|snapshot> ...\n"));
	case HELP_SHARE:
	return (gettext("\tshare [-l] <-a [nfs\|smb] \| filesystem>\n"));
	case HELP_SNAPSHOT:
	return (gettext("\tsnapshot [-r] [-o property=value] ... "
	"<filesystem\|volume>@<snap> ...\n"));
	case HELP_UNMOUNT:
	return (gettext("\tunmount [-fu] "
	"<-a \| filesystem\|mountpoint>\n"));
	case HELP_UNSHARE:
	return (gettext("\tunshare "
	"<-a [nfs\|smb] \| filesystem\|mountpoint>\n"));
	case HELP_ALLOW:
	return (gettext("\tallow <filesystem\|volume>\n"
	"\tallow [-ldug] "
	"<\"everyone\"\|user\|group>[,...] <perm\|@setname>[,...]\n"
	"\t <filesystem\|volume>\n"
	"\tallow [-ld] -e <perm\|@setname>[,...] "
	"<filesystem\|volume>\n"
	"\tallow -c <perm\|@setname>[,...] <filesystem\|volume>\n"
	"\tallow -s @setname <perm\|@setname>[,...] "
	"<filesystem\|volume>\n"));
	case HELP_UNALLOW:
	return (gettext("\tunallow [-rldug] "
	"<\"everyone\"\|user\|group>[,...]\n"
	"\t [<perm\|@setname>[,...]] <filesystem\|volume>\n"
	"\tunallow [-rld] -e [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"
	"\tunallow [-r] -c [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"
	"\tunallow [-r] -s @setname [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"));
	case HELP_USERSPACE:
	return (gettext("\tuserspace [-Hinp] [-o field[,...]] "
	"[-s field] ...\n"
	"\t [-S field] ... [-t type[,...]] "
	"<filesystem\|snapshot\|path>\n"));
	case HELP_GROUPSPACE:
	return (gettext("\tgroupspace [-Hinp] [-o field[,...]] "
	"[-s field] ...\n"
	"\t [-S field] ... [-t type[,...]] "
	"<filesystem\|snapshot\|path>\n"));
	case HELP_PROJECTSPACE:
	return (gettext("\tprojectspace [-Hp] [-o field[,...]] "
	"[-s field] ... \n"
	"\t [-S field] ... <filesystem\|snapshot\|path>\n"));
	case HELP_PROJECT:
	return (gettext("\tproject [-d\|-r] <directory\|file ...>\n"
	"\tproject -c [-0] [-d\|-r] [-p id] <directory\|file ...>\n"
	"\tproject -C [-k] [-r] <directory ...>\n"
	"\tproject [-p id] [-r] [-s] <directory ...>\n"));
	case HELP_HOLD:
	return (gettext("\thold [-r] <tag> <snapshot> ...\n"));
	case HELP_HOLDS:
	return (gettext("\tholds [-rH] <snapshot> ...\n"));
	case HELP_RELEASE:
	return (gettext("\trelease [-r] <tag> <snapshot> ...\n"));
	case HELP_DIFF:
	return (gettext("\tdiff [-FHt] <snapshot> "
	"[snapshot\|filesystem]\n"));
	case HELP_BOOKMARK:
	return (gettext("\tbookmark <snapshot\|bookmark> "
	"<newbookmark>\n"));
	case HELP_CHANNEL_PROGRAM:
	return (gettext("\tprogram [-jn] [-t <instruction limit>] "
	"[-m <memory limit (b)>]\n"
	"\t <pool> <program file> [lua args...]\n"));
	case HELP_LOAD_KEY:
	return (gettext("\tload-key [-rn] [-L <keylocation>] "
	"<-a \| filesystem\|volume>\n"));
	case HELP_UNLOAD_KEY:
	return (gettext("\tunload-key [-r] "
	"<-a \| filesystem\|volume>\n"));
	case HELP_CHANGE_KEY:
	return (gettext("\tchange-key [-l] [-o keyformat=<value>]\n"
	"\t [-o keylocation=<value>] [-o pbkdf2iters=<value>]\n"
	"\t <filesystem\|volume>\n"
	"\tchange-key -i [-l] <filesystem\|volume>\n"));
	case HELP_VERSION:
	return (gettext("\tversion\n"));
	case HELP_REDACT:
	return (gettext("\tredact <snapshot> <bookmark> "
	"<redaction_snapshot> ...\n"));
	case HELP_JAIL:
	return (gettext("\tjail <jailid\|jailname> <filesystem>\n"));
	case HELP_UNJAIL:
	return (gettext("\tunjail <jailid\|jailname> <filesystem>\n"));
	case HELP_WAIT:
	return (gettext("\twait [-t <activity>] <filesystem>\n"));
	case HELP_ZONE:
	return (gettext("\tzone <nsfile> <filesystem>\n"));
	case HELP_UNZONE:
	return (gettext("\tunzone <nsfile> <filesystem>\n"));
	default:
	__builtin_unreachable();
	}
	}

	void
	nomem(void)
	{
	(void) fprintf(stderr, gettext("internal error: out of memory\n"));
	exit(1);
	}

	/*
	* Utility function to guarantee malloc() success.
	*/

	void *
	safe_malloc(size_t size)
	{
	void *data;

	if ((data = calloc(1, size)) == NULL)
	nomem();

	return (data);
	}

	static void *
	safe_realloc(void *data, size_t size)
	{
	void *newp;
	if ((newp = realloc(data, size)) == NULL) {
	free(data);
	nomem();
	}

	return (newp);
	}

	static char *
	safe_strdup(const char *str)
	{
	char *dupstr = strdup(str);

	if (dupstr == NULL)
	nomem();

	return (dupstr);
	}

	/*
	* Callback routine that will print out information for each of
	* the properties.
	*/
	static int
	usage_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-15s ", zfs_prop_to_name(prop));

	if (zfs_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, "YES ");

	if (zfs_prop_inheritable(prop))
	(void) fprintf(fp, " YES ");
	else
	(void) fprintf(fp, " NO ");

	(void) fprintf(fp, "%s\n", zfs_prop_values(prop) ?: "-");

	return (ZPROP_CONT);
	}

	/*
	* Display usage message. If we're inside a command, display only the usage for
	* that command. Otherwise, iterate over the entire command table and display
	* a complete usage message.
	*/
	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	int i;
	boolean_t show_properties = B_FALSE;
	FILE *fp = requested ? stdout : stderr;

	if (current_command == NULL) {

	(void) fprintf(fp, gettext("usage: zfs command args ...\n"));
	(void) fprintf(fp,
	gettext("where 'command' is one of the following:\n\n"));

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	(void) fprintf(fp, "\n");
	else
	(void) fprintf(fp, "%s",
	get_usage(command_table[i].usage));
	}

	(void) fprintf(fp, gettext("\nEach dataset is of the form: "
	"pool/[dataset/]*dataset[@name]\n"));
	} else {
	(void) fprintf(fp, gettext("usage:\n"));
	(void) fprintf(fp, "%s", get_usage(current_command->usage));
	}

	if (current_command != NULL &&
	(strcmp(current_command->name, "set") == 0 \|\|
	strcmp(current_command->name, "get") == 0 \|\|
	strcmp(current_command->name, "inherit") == 0 \|\|
	strcmp(current_command->name, "list") == 0))
	show_properties = B_TRUE;

	if (show_properties) {
	(void) fprintf(fp,
	gettext("\nThe following properties are supported:\n"));

	(void) fprintf(fp, "\n\t%-14s %s %s %s\n\n",
	"PROPERTY", "EDIT", "INHERIT", "VALUES");

	/* Iterate over all properties */
	(void) zprop_iter(usage_prop_cb, fp, B_FALSE, B_TRUE,
	ZFS_TYPE_DATASET);

	(void) fprintf(fp, "\t%-15s ", "userused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "groupused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "projectused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "userobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "groupobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "projectobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "userquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "groupquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "projectquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "userobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "groupobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "projectobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "written@<snap>");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "written#<bookmark>");
	(void) fprintf(fp, " NO NO <size>\n");

	(void) fprintf(fp, gettext("\nSizes are specified in bytes "
	"with standard units such as K, M, G, etc.\n"));
	(void) fprintf(fp, "%s", gettext("\nUser-defined properties "
	"can be specified by using a name containing a colon "
	"(:).\n"));
	(void) fprintf(fp, gettext("\nThe {user\|group\|project}"
	"[obj]{used\|quota}@ properties must be appended with\n"
	"a user\|group\|project specifier of one of these forms:\n"
	" POSIX name (eg: \"matt\")\n"
	" POSIX id (eg: \"126829\")\n"
	" SMB name@domain (eg: \"matt@sun\")\n"
	" SMB SID (eg: \"S-1-234-567-89\")\n"));
	} else {
	(void) fprintf(fp,
	gettext("\nFor the property list, run: %s\n"),
	"zfs set\|get");
	(void) fprintf(fp,
	gettext("\nFor the delegated permission list, run: %s\n"),
	"zfs allow\|unallow");
	}

	/*
	* See comments at end of main().
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	exit(requested ? 0 : 2);
	}

	/*
	* Take a property=value argument string and add it to the given nvlist.
	* Modifies the argument inplace.
	*/
	static boolean_t
	parseprop(nvlist_t props, char propname)
	{
	char *propval;

	if ((propval = strchr(propname, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for property=value argument\n"));
	return (B_FALSE);
	}
	*propval = '\0';
	propval++;
	if (nvlist_exists(props, propname)) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (B_FALSE);
	}
	if (nvlist_add_string(props, propname, propval) != 0)
	nomem();
	return (B_TRUE);
	}

	/*
	* Take a property name argument and add it to the given nvlist.
	* Modifies the argument inplace.
	*/
	static boolean_t
	parsepropname(nvlist_t props, char propname)
	{
	if (strchr(propname, '=') != NULL) {
	(void) fprintf(stderr, gettext("invalid character "
	"'=' in property argument\n"));
	return (B_FALSE);
	}
	if (nvlist_exists(props, propname)) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (B_FALSE);
	}
	if (nvlist_add_boolean(props, propname) != 0)
	nomem();
	return (B_TRUE);
	}

	static int
	parse_depth(char opt, int flags)
	{
	char *tmp;
	int depth;

	depth = (int)strtol(opt, &tmp, 0);
	if (*tmp) {
	(void) fprintf(stderr,
	gettext("%s is not an integer\n"), optarg);
	usage(B_FALSE);
	}
	if (depth < 0) {
	(void) fprintf(stderr,
	gettext("Depth can not be negative.\n"));
	usage(B_FALSE);
	}
	*flags \|= (ZFS_ITER_DEPTH_LIMIT\|ZFS_ITER_RECURSE);
	return (depth);
	}

	#define PROGRESS_DELAY 2 /* seconds */

	static const char *pt_reverse =
	"\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b";
	static time_t pt_begin;
	static char *pt_header = NULL;
	static boolean_t pt_shown;

	static void
	start_progress_timer(void)
	{
	pt_begin = time(NULL) + PROGRESS_DELAY;
	pt_shown = B_FALSE;
	}

	static void
	set_progress_header(const char *header)
	{
	assert(pt_header == NULL);
	pt_header = safe_strdup(header);
	if (pt_shown) {
	(void) printf("%s: ", header);
	(void) fflush(stdout);
	}
	}

	static void
	update_progress(const char *update)
	{
	if (!pt_shown && time(NULL) > pt_begin) {
	int len = strlen(update);

	(void) printf("%s: %s%.s", pt_header, update, len, len,
	pt_reverse);
	(void) fflush(stdout);
	pt_shown = B_TRUE;
	} else if (pt_shown) {
	int len = strlen(update);

	(void) printf("%s%.s", update, len, len, pt_reverse);
	(void) fflush(stdout);
	}
	}

	static void
	finish_progress(const char *done)
	{
	if (pt_shown) {
	(void) puts(done);
	(void) fflush(stdout);
	}
	free(pt_header);
	pt_header = NULL;
	}

	static int
	zfs_mount_and_share(libzfs_handle_t hdl, const char dataset, zfs_type_t type)
	{
	zfs_handle_t *zhp = NULL;
	int ret = 0;

	zhp = zfs_open(hdl, dataset, type);
	if (zhp == NULL)
	return (1);

	/*
	* Volumes may neither be mounted or shared. Potentially in the
	* future filesystems detected on these volumes could be mounted.
	*/
	if (zfs_get_type(zhp) == ZFS_TYPE_VOLUME) {
	zfs_close(zhp);
	return (0);
	}

	/*
	* Mount and/or share the new filesystem as appropriate. We provide a
	* verbose error message to let the user know that their filesystem was
	* in fact created, even if we failed to mount or share it.
	*
	* If the user doesn't want the dataset automatically mounted, then
	* skip the mount/share step
	*/
	if (zfs_prop_valid_for_type(ZFS_PROP_CANMOUNT, type, B_FALSE) &&
	zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_ON) {
	if (zfs_mount_delegation_check()) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but it may only be "
	"mounted by root\n"));
	ret = 1;
	} else if (zfs_mount(zhp, NULL, 0) != 0) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but not mounted\n"));
	ret = 1;
	} else if (zfs_share(zhp, NULL) != 0) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but not shared\n"));
	ret = 1;
	}
	zfs_commit_shares(NULL);
	}

	zfs_close(zhp);

	return (ret);
	}

	/*
	* zfs clone [-p] [-o prop=value] ... <snap> <fs \| vol>
	*
	* Given an existing dataset, create a writable copy whose initial contents
	* are the same as the source. The newly created dataset maintains a
	* dependency on the original; the original cannot be destroyed so long as
	* the clone exists.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	*/
	static int
	zfs_do_clone(int argc, char **argv)
	{
	zfs_handle_t *zhp = NULL;
	boolean_t parents = B_FALSE;
	nvlist_t *props;
	int ret = 0;
	int c;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, "o:p")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	return (1);
	}
	break;
	case 'p':
	parents = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source dataset "
	"argument\n"));
	goto usage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing target dataset "
	"argument\n"));
	goto usage;
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	goto usage;
	}

	/* open the source dataset */
	if ((zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	if (parents && zfs_name_valid(argv[1], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) {
	/*
	* Now create the ancestors of the target dataset. If the
	* target already exists and '-p' option was used we should not
	* complain.
	*/
	if (zfs_dataset_exists(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) {
	zfs_close(zhp);
	nvlist_free(props);
	return (0);
	}
	if (zfs_create_ancestors(g_zfs, argv[1]) != 0) {
	zfs_close(zhp);
	nvlist_free(props);
	return (1);
	}
	}

	/* pass to libzfs */
	ret = zfs_clone(zhp, argv[1], props);

	/* create the mountpoint if necessary */
	if (ret == 0) {
	if (log_history) {
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	ret = zfs_mount_and_share(g_zfs, argv[1], ZFS_TYPE_DATASET);
	}

	zfs_close(zhp);
	nvlist_free(props);

	return (!!ret);

	usage:
	ASSERT3P(zhp, ==, NULL);
	nvlist_free(props);
	usage(B_FALSE);
	return (-1);
	}

	/*
	* Return a default volblocksize for the pool which always uses more than
	* half of the data sectors. This primarily applies to dRAID which always
	* writes full stripe widths.
	*/
	static uint64_t
	default_volblocksize(zpool_handle_t zhp, nvlist_t props)
	{
	uint64_t volblocksize, asize = SPA_MINBLOCKSIZE;
	nvlist_t tree, *vdevs;
	uint_t nvdevs;

	nvlist_t *config = zpool_get_config(zhp, NULL);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 \|\|
	nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
	&vdevs, &nvdevs) != 0) {
	return (ZVOL_DEFAULT_BLOCKSIZE);
	}

	for (int i = 0; i < nvdevs; i++) {
	nvlist_t *nv = vdevs[i];
	uint64_t ashift, ndata, nparity;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &ashift) != 0)
	continue;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DRAID_NDATA,
	&ndata) == 0) {
	/* dRAID minimum allocation width */
	asize = MAX(asize, ndata * (1ULL << ashift));
	} else if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
	&nparity) == 0) {
	/* raidz minimum allocation width */
	if (nparity == 1)
	asize = MAX(asize, 2 * (1ULL << ashift));
	else
	asize = MAX(asize, 4 * (1ULL << ashift));
	} else {
	/* mirror or (non-redundant) leaf vdev */
	asize = MAX(asize, 1ULL << ashift);
	}
	}

	/*
	* Calculate the target volblocksize such that more than half
	* of the asize is used. The following table is for 4k sectors.
	*
	* n asize blksz used \| n asize blksz used
	* -------------------------+---------------------------------
	* 1 4,096 8,192 100% \| 9 36,864 32,768 88%
	* 2 8,192 8,192 100% \| 10 40,960 32,768 80%
	* 3 12,288 8,192 66% \| 11 45,056 32,768 72%
	* 4 16,384 16,384 100% \| 12 49,152 32,768 66%
	* 5 20,480 16,384 80% \| 13 53,248 32,768 61%
	* 6 24,576 16,384 66% \| 14 57,344 32,768 57%
	* 7 28,672 16,384 57% \| 15 61,440 32,768 53%
	* 8 32,768 32,768 100% \| 16 65,536 65,636 100%
	*
	* This is primarily a concern for dRAID which always allocates
	* a full stripe width. For dRAID the default stripe width is
	* n=8 in which case the volblocksize is set to 32k. Ignoring
	* compression there are no unused sectors. This same reasoning
	* applies to raidz[2,3] so target 4 sectors to minimize waste.
	*/
	uint64_t tgt_volblocksize = ZVOL_DEFAULT_BLOCKSIZE;
	while (tgt_volblocksize * 2 <= asize)
	tgt_volblocksize *= 2;

	const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE);
	if (nvlist_lookup_uint64(props, prop, &volblocksize) == 0) {

	/* Issue a warning when a non-optimal size is requested. */
	if (volblocksize < ZVOL_DEFAULT_BLOCKSIZE) {
	(void) fprintf(stderr, gettext("Warning: "
	"volblocksize (%llu) is less than the default "
	"minimum block size (%llu).\nTo reduce wasted "
	"space a volblocksize of %llu is recommended.\n"),
	(u_longlong_t)volblocksize,
	(u_longlong_t)ZVOL_DEFAULT_BLOCKSIZE,
	(u_longlong_t)tgt_volblocksize);
	} else if (volblocksize < tgt_volblocksize) {
	(void) fprintf(stderr, gettext("Warning: "
	"volblocksize (%llu) is much less than the "
	"minimum allocation\nunit (%llu), which wastes "
	"at least %llu%% of space. To reduce wasted "
	"space,\nuse a larger volblocksize (%llu is "
	"recommended), fewer dRAID data disks\n"
	"per group, or smaller sector size (ashift).\n"),
	(u_longlong_t)volblocksize, (u_longlong_t)asize,
	(u_longlong_t)((100 * (asize - volblocksize)) /
	asize), (u_longlong_t)tgt_volblocksize);
	}
	} else {
	volblocksize = tgt_volblocksize;
	fnvlist_add_uint64(props, prop, volblocksize);
	}

	return (volblocksize);
	}

	/*
	* zfs create [-Pnpv] [-o prop=value] ... fs
	* zfs create [-Pnpsv] [-b blocksize] [-o prop=value] ... -V vol size
	*
	* Create a new dataset. This command can be used to create filesystems
	* and volumes. Snapshot creation is handled by 'zfs snapshot'.
	* For volumes, the user must specify a size to be used.
	*
	* The '-s' flag applies only to volumes, and indicates that we should not try
	* to set the reservation for this volume. By default we set a reservation
	* equal to the size for any volume. For pools with SPA_VERSION >=
	* SPA_VERSION_REFRESERVATION, we set a refreservation instead.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	*
	* The '-n' flag is no-op (dry run) mode. This will perform a user-space sanity
	* check of arguments and properties, but does not check for permissions,
	* available space, etc.
	*
	* The '-u' flag prevents the newly created file system from being mounted.
	*
	* The '-v' flag is for verbose output.
	*
	* The '-P' flag is used for parseable output. It implies '-v'.
	*/
	static int
	zfs_do_create(int argc, char **argv)
	{
	zfs_type_t type = ZFS_TYPE_FILESYSTEM;
	zpool_handle_t *zpool_handle = NULL;
	nvlist_t *real_props = NULL;
	uint64_t volsize = 0;
	int c;
	boolean_t noreserve = B_FALSE;
	boolean_t bflag = B_FALSE;
	boolean_t parents = B_FALSE;
	boolean_t dryrun = B_FALSE;
	boolean_t nomount = B_FALSE;
	boolean_t verbose = B_FALSE;
	boolean_t parseable = B_FALSE;
	int ret = 1;
	nvlist_t *props;
	uint64_t intval;
	char *strval;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, ":PV:b:nso:puv")) != -1) {
	switch (c) {
	case 'V':
	type = ZFS_TYPE_VOLUME;
	if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) {
	(void) fprintf(stderr, gettext("bad volume "
	"size '%s': %s\n"), optarg,
	libzfs_error_description(g_zfs));
	goto error;
	}

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), intval) != 0)
	nomem();
	volsize = intval;
	break;
	case 'P':
	verbose = B_TRUE;
	parseable = B_TRUE;
	break;
	case 'p':
	parents = B_TRUE;
	break;
	case 'b':
	bflag = B_TRUE;
	if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) {
	(void) fprintf(stderr, gettext("bad volume "
	"block size '%s': %s\n"), optarg,
	libzfs_error_description(g_zfs));
	goto error;
	}

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	intval) != 0)
	nomem();
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'o':
	if (!parseprop(props, optarg))
	goto error;
	break;
	case 's':
	noreserve = B_TRUE;
	break;
	case 'u':
	nomount = B_TRUE;
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing size "
	"argument\n"));
	goto badusage;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto badusage;
	}
	}

	if ((bflag \|\| noreserve) && type != ZFS_TYPE_VOLUME) {
	(void) fprintf(stderr, gettext("'-s' and '-b' can only be "
	"used when creating a volume\n"));
	goto badusage;
	}
	if (nomount && type != ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("'-u' can only be "
	"used when creating a filesystem\n"));
	goto badusage;
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc == 0) {
	(void) fprintf(stderr, gettext("missing %s argument\n"),
	zfs_type_to_name(type));
	goto badusage;
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	goto badusage;
	}

	if (dryrun \|\| type == ZFS_TYPE_VOLUME) {
	char msg[ZFS_MAX_DATASET_NAME_LEN * 2];
	char *p;

	if ((p = strchr(argv[0], '/')) != NULL)
	*p = '\0';
	zpool_handle = zpool_open(g_zfs, argv[0]);
	if (p != NULL)
	*p = '/';
	if (zpool_handle == NULL)
	goto error;

	(void) snprintf(msg, sizeof (msg),
	dryrun ? gettext("cannot verify '%s'") :
	gettext("cannot create '%s'"), argv[0]);
	if (props && (real_props = zfs_valid_proplist(g_zfs, type,
	props, 0, NULL, zpool_handle, B_TRUE, msg)) == NULL) {
	zpool_close(zpool_handle);
	goto error;
	}
	}

	if (type == ZFS_TYPE_VOLUME) {
	const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE);
	uint64_t volblocksize = default_volblocksize(zpool_handle,
	real_props);

	if (volblocksize != ZVOL_DEFAULT_BLOCKSIZE &&
	nvlist_lookup_string(props, prop, &strval) != 0) {
	if (asprintf(&strval, "%llu",
	(u_longlong_t)volblocksize) == -1)
	nomem();
	nvlist_add_string(props, prop, strval);
	free(strval);
	}

	/*
	* If volsize is not a multiple of volblocksize, round it
	* up to the nearest multiple of the volblocksize.
	*/
	if (volsize % volblocksize) {
	volsize = P2ROUNDUP_TYPED(volsize, volblocksize,
	uint64_t);

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), volsize) != 0) {
	nvlist_free(props);
	nomem();
	}
	}
	}

	if (type == ZFS_TYPE_VOLUME && !noreserve) {
	uint64_t spa_version;
	zfs_prop_t resv_prop;

	spa_version = zpool_get_prop_int(zpool_handle,
	ZPOOL_PROP_VERSION, NULL);
	if (spa_version >= SPA_VERSION_REFRESERVATION)
	resv_prop = ZFS_PROP_REFRESERVATION;
	else
	resv_prop = ZFS_PROP_RESERVATION;

	volsize = zvol_volsize_to_reservation(zpool_handle, volsize,
	real_props);

	if (nvlist_lookup_string(props, zfs_prop_to_name(resv_prop),
	&strval) != 0) {
	if (nvlist_add_uint64(props,
	zfs_prop_to_name(resv_prop), volsize) != 0) {
	nvlist_free(props);
	nomem();
	}
	}
	}
	if (zpool_handle != NULL) {
	zpool_close(zpool_handle);
	nvlist_free(real_props);
	}

	if (parents && zfs_name_valid(argv[0], type)) {
	/*
	* Now create the ancestors of target dataset. If the target
	* already exists and '-p' option was used we should not
	* complain.
	*/
	if (zfs_dataset_exists(g_zfs, argv[0], type)) {
	ret = 0;
	goto error;
	}
	if (verbose) {
	(void) printf(parseable ? "create_ancestors\t%s\n" :
	dryrun ? "would create ancestors of %s\n" :
	"create ancestors of %s\n", argv[0]);
	}
	if (!dryrun) {
	if (zfs_create_ancestors(g_zfs, argv[0]) != 0) {
	goto error;
	}
	}
	}

	if (verbose) {
	nvpair_t *nvp = NULL;
	(void) printf(parseable ? "create\t%s\n" :
	dryrun ? "would create %s\n" : "create %s\n", argv[0]);
	while ((nvp = nvlist_next_nvpair(props, nvp)) != NULL) {
	uint64_t uval;
	char *sval;

	switch (nvpair_type(nvp)) {
	case DATA_TYPE_UINT64:
	VERIFY0(nvpair_value_uint64(nvp, &uval));
	(void) printf(parseable ?
	"property\t%s\t%llu\n" : "\t%s=%llu\n",
	nvpair_name(nvp), (u_longlong_t)uval);
	break;
	case DATA_TYPE_STRING:
	VERIFY0(nvpair_value_string(nvp, &sval));
	(void) printf(parseable ?
	"property\t%s\t%s\n" : "\t%s=%s\n",
	nvpair_name(nvp), sval);
	break;
	default:
	(void) fprintf(stderr, "property '%s' "
	"has illegal type %d\n",
	nvpair_name(nvp), nvpair_type(nvp));
	abort();
	}
	}
	}
	if (dryrun) {
	ret = 0;
	goto error;
	}

	/* pass to libzfs */
	if (zfs_create(g_zfs, argv[0], type, props) != 0)
	goto error;

	if (log_history) {
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	if (nomount) {
	ret = 0;
	goto error;
	}

	ret = zfs_mount_and_share(g_zfs, argv[0], ZFS_TYPE_DATASET);
	error:
	nvlist_free(props);
	return (ret);
	badusage:
	nvlist_free(props);
	usage(B_FALSE);
	return (2);
	}

	/*
	* zfs destroy [-rRf] <fs, vol>
	* zfs destroy [-rRd] <snap>
	*
	* -r Recursively destroy all children
	* -R Recursively destroy all dependents, including clones
	* -f Force unmounting of any dependents
	* -d If we can't destroy now, mark for deferred destruction
	*
	* Destroys the given dataset. By default, it will unmount any filesystems,
	* and refuse to destroy a dataset that has any dependents. A dependent can
	* either be a child, or a clone of a child.
	*/
	typedef struct destroy_cbdata {
	boolean_t cb_first;
	boolean_t cb_force;
	boolean_t cb_recurse;
	boolean_t cb_error;
	boolean_t cb_doclones;
	zfs_handle_t *cb_target;
	boolean_t cb_defer_destroy;
	boolean_t cb_verbose;
	boolean_t cb_parsable;
	boolean_t cb_dryrun;
	nvlist_t *cb_nvl;
	nvlist_t *cb_batchedsnaps;

	/* first snap in contiguous run */
	char *cb_firstsnap;
	/* previous snap in contiguous run */
	char *cb_prevsnap;
	int64_t cb_snapused;
	char *cb_snapspec;
	char *cb_bookmark;
	uint64_t cb_snap_count;
	} destroy_cbdata_t;

	/*
	* Check for any dependents based on the '-r' or '-R' flags.
	*/
	static int
	destroy_check_dependent(zfs_handle_t zhp, void data)
	{
	destroy_cbdata_t *cbp = data;
	const char *tname = zfs_get_name(cbp->cb_target);
	const char *name = zfs_get_name(zhp);

	if (strncmp(tname, name, strlen(tname)) == 0 &&
	(name[strlen(tname)] == '/' \|\| name[strlen(tname)] == '@')) {
	/*
	* This is a direct descendant, not a clone somewhere else in
	* the hierarchy.
	*/
	if (cbp->cb_recurse)
	goto out;

	if (cbp->cb_first) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"%s has children\n"),
	zfs_get_name(cbp->cb_target),
	zfs_type_to_name(zfs_get_type(cbp->cb_target)));
	(void) fprintf(stderr, gettext("use '-r' to destroy "
	"the following datasets:\n"));
	cbp->cb_first = B_FALSE;
	cbp->cb_error = B_TRUE;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));
	} else {
	/*
	* This is a clone. We only want to report this if the '-r'
	* wasn't specified, or the target is a snapshot.
	*/
	if (!cbp->cb_recurse &&
	zfs_get_type(cbp->cb_target) != ZFS_TYPE_SNAPSHOT)
	goto out;

	if (cbp->cb_first) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"%s has dependent clones\n"),
	zfs_get_name(cbp->cb_target),
	zfs_type_to_name(zfs_get_type(cbp->cb_target)));
	(void) fprintf(stderr, gettext("use '-R' to destroy "
	"the following datasets:\n"));
	cbp->cb_first = B_FALSE;
	cbp->cb_error = B_TRUE;
	cbp->cb_dryrun = B_TRUE;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));
	}

	out:
	zfs_close(zhp);
	return (0);
	}

	static int
	destroy_batched(destroy_cbdata_t *cb)
	{
	int error = zfs_destroy_snaps_nvl(g_zfs,
	cb->cb_batchedsnaps, B_FALSE);
	fnvlist_free(cb->cb_batchedsnaps);
	cb->cb_batchedsnaps = fnvlist_alloc();
	return (error);
	}

	static int
	destroy_callback(zfs_handle_t zhp, void data)
	{
	destroy_cbdata_t *cb = data;
	const char *name = zfs_get_name(zhp);
	int error;

	if (cb->cb_verbose) {
	if (cb->cb_parsable) {
	(void) printf("destroy\t%s\n", name);
	} else if (cb->cb_dryrun) {
	(void) printf(gettext("would destroy %s\n"),
	name);
	} else {
	(void) printf(gettext("will destroy %s\n"),
	name);
	}
	}

	/*
	* Ignore pools (which we've already flagged as an error before getting
	* here).
	*/
	if (strchr(zfs_get_name(zhp), '/') == NULL &&
	zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) {
	zfs_close(zhp);
	return (0);
	}
	if (cb->cb_dryrun) {
	zfs_close(zhp);
	return (0);
	}

	/*
	* We batch up all contiguous snapshots (even of different
	* filesystems) and destroy them with one ioctl. We can't
	* simply do all snap deletions and then all fs deletions,
	* because we must delete a clone before its origin.
	*/
	if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT) {
	cb->cb_snap_count++;
	fnvlist_add_boolean(cb->cb_batchedsnaps, name);
	if (cb->cb_snap_count % 10 == 0 && cb->cb_defer_destroy)
	error = destroy_batched(cb);
	} else {
	error = destroy_batched(cb);
	if (error != 0 \|\|
	zfs_unmount(zhp, NULL, cb->cb_force ? MS_FORCE : 0) != 0 \|\|
	zfs_destroy(zhp, cb->cb_defer_destroy) != 0) {
	zfs_close(zhp);
	/*
	* When performing a recursive destroy we ignore errors
	* so that the recursive destroy could continue
	* destroying past problem datasets
	*/
	if (cb->cb_recurse) {
	cb->cb_error = B_TRUE;
	return (0);
	}
	return (-1);
	}
	}

	zfs_close(zhp);
	return (0);
	}

	static int
	destroy_print_cb(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	const char *name = zfs_get_name(zhp);
	int err = 0;

	if (nvlist_exists(cb->cb_nvl, name)) {
	if (cb->cb_firstsnap == NULL)
	cb->cb_firstsnap = strdup(name);
	if (cb->cb_prevsnap != NULL)
	free(cb->cb_prevsnap);
	/* this snap continues the current range */
	cb->cb_prevsnap = strdup(name);
	if (cb->cb_firstsnap == NULL \|\| cb->cb_prevsnap == NULL)
	nomem();
	if (cb->cb_verbose) {
	if (cb->cb_parsable) {
	(void) printf("destroy\t%s\n", name);
	} else if (cb->cb_dryrun) {
	(void) printf(gettext("would destroy %s\n"),
	name);
	} else {
	(void) printf(gettext("will destroy %s\n"),
	name);
	}
	}
	} else if (cb->cb_firstsnap != NULL) {
	/* end of this range */
	uint64_t used = 0;
	err = lzc_snaprange_space(cb->cb_firstsnap,
	cb->cb_prevsnap, &used);
	cb->cb_snapused += used;
	free(cb->cb_firstsnap);
	cb->cb_firstsnap = NULL;
	free(cb->cb_prevsnap);
	cb->cb_prevsnap = NULL;
	}
	zfs_close(zhp);
	return (err);
	}

	static int
	destroy_print_snapshots(zfs_handle_t fs_zhp, destroy_cbdata_t cb)
	{
	int err;
	assert(cb->cb_firstsnap == NULL);
	assert(cb->cb_prevsnap == NULL);
	err = zfs_iter_snapshots_sorted(fs_zhp, destroy_print_cb, cb, 0, 0);
	if (cb->cb_firstsnap != NULL) {
	uint64_t used = 0;
	if (err == 0) {
	err = lzc_snaprange_space(cb->cb_firstsnap,
	cb->cb_prevsnap, &used);
	}
	cb->cb_snapused += used;
	free(cb->cb_firstsnap);
	cb->cb_firstsnap = NULL;
	free(cb->cb_prevsnap);
	cb->cb_prevsnap = NULL;
	}
	return (err);
	}

	static int
	snapshot_to_nvl_cb(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	int err = 0;

	/* Check for clones. */
	if (!cb->cb_doclones && !cb->cb_defer_destroy) {
	cb->cb_target = zhp;
	cb->cb_first = B_TRUE;
	err = zfs_iter_dependents(zhp, B_TRUE,
	destroy_check_dependent, cb);
	}

	if (err == 0) {
	if (nvlist_add_boolean(cb->cb_nvl, zfs_get_name(zhp)))
	nomem();
	}
	zfs_close(zhp);
	return (err);
	}

	static int
	gather_snapshots(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	int err = 0;

	err = zfs_iter_snapspec(zhp, cb->cb_snapspec, snapshot_to_nvl_cb, cb);
	if (err == ENOENT)
	err = 0;
	if (err != 0)
	goto out;

	if (cb->cb_verbose) {
	err = destroy_print_snapshots(zhp, cb);
	if (err != 0)
	goto out;
	}

	if (cb->cb_recurse)
	err = zfs_iter_filesystems(zhp, gather_snapshots, cb);

	out:
	zfs_close(zhp);
	return (err);
	}

	static int
	destroy_clones(destroy_cbdata_t *cb)
	{
	nvpair_t *pair;
	for (pair = nvlist_next_nvpair(cb->cb_nvl, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(cb->cb_nvl, pair)) {
	zfs_handle_t *zhp = zfs_open(g_zfs, nvpair_name(pair),
	ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	boolean_t defer = cb->cb_defer_destroy;
	int err;

	/*
	* We can't defer destroy non-snapshots, so set it to
	* false while destroying the clones.
	*/
	cb->cb_defer_destroy = B_FALSE;
	err = zfs_iter_dependents(zhp, B_FALSE,
	destroy_callback, cb);
	cb->cb_defer_destroy = defer;
	zfs_close(zhp);
	if (err != 0)
	return (err);
	}
	}
	return (0);
	}

	static int
	zfs_do_destroy(int argc, char **argv)
	{
	destroy_cbdata_t cb = { 0 };
	int rv = 0;
	int err = 0;
	int c;
	zfs_handle_t *zhp = NULL;
	char at, pound;
	zfs_type_t type = ZFS_TYPE_DATASET;

	/* check options */
	while ((c = getopt(argc, argv, "vpndfrR")) != -1) {
	switch (c) {
	case 'v':
	cb.cb_verbose = B_TRUE;
	break;
	case 'p':
	cb.cb_verbose = B_TRUE;
	cb.cb_parsable = B_TRUE;
	break;
	case 'n':
	cb.cb_dryrun = B_TRUE;
	break;
	case 'd':
	cb.cb_defer_destroy = B_TRUE;
	type = ZFS_TYPE_SNAPSHOT;
	break;
	case 'f':
	cb.cb_force = B_TRUE;
	break;
	case 'r':
	cb.cb_recurse = B_TRUE;
	break;
	case 'R':
	cb.cb_recurse = B_TRUE;
	cb.cb_doclones = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc == 0) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	at = strchr(argv[0], '@');
	pound = strchr(argv[0], '#');
	if (at != NULL) {

	/* Build the list of snaps to destroy in cb_nvl. */
	cb.cb_nvl = fnvlist_alloc();

	*at = '\0';
	zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	nvlist_free(cb.cb_nvl);
	return (1);
	}

	cb.cb_snapspec = at + 1;
	if (gather_snapshots(zfs_handle_dup(zhp), &cb) != 0 \|\|
	cb.cb_error) {
	rv = 1;
	goto out;
	}

	if (nvlist_empty(cb.cb_nvl)) {
	(void) fprintf(stderr, gettext("could not find any "
	"snapshots to destroy; check snapshot names.\n"));
	rv = 1;
	goto out;
	}

	if (cb.cb_verbose) {
	char buf[16];
	zfs_nicebytes(cb.cb_snapused, buf, sizeof (buf));
	if (cb.cb_parsable) {
	(void) printf("reclaim\t%llu\n",
	(u_longlong_t)cb.cb_snapused);
	} else if (cb.cb_dryrun) {
	(void) printf(gettext("would reclaim %s\n"),
	buf);
	} else {
	(void) printf(gettext("will reclaim %s\n"),
	buf);
	}
	}

	if (!cb.cb_dryrun) {
	if (cb.cb_doclones) {
	cb.cb_batchedsnaps = fnvlist_alloc();
	err = destroy_clones(&cb);
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs,
	cb.cb_batchedsnaps, B_FALSE);
	}
	if (err != 0) {
	rv = 1;
	goto out;
	}
	}
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_nvl,
	cb.cb_defer_destroy);
	}
	}

	if (err != 0)
	rv = 1;
	} else if (pound != NULL) {
	int err;
	nvlist_t *nvl;

	if (cb.cb_dryrun) {
	(void) fprintf(stderr,
	"dryrun is not supported with bookmark\n");
	return (-1);
	}

	if (cb.cb_defer_destroy) {
	(void) fprintf(stderr,
	"defer destroy is not supported with bookmark\n");
	return (-1);
	}

	if (cb.cb_recurse) {
	(void) fprintf(stderr,
	"recursive is not supported with bookmark\n");
	return (-1);
	}

	/*
	* Unfortunately, zfs_bookmark() doesn't honor the
	* casesensitivity setting. However, we can't simply
	* remove this check, because lzc_destroy_bookmarks()
	* ignores non-existent bookmarks, so this is necessary
	* to get a proper error message.
	*/
	if (!zfs_bookmark_exists(argv[0])) {
	(void) fprintf(stderr, gettext("bookmark '%s' "
	"does not exist.\n"), argv[0]);
	return (1);
	}

	nvl = fnvlist_alloc();
	fnvlist_add_boolean(nvl, argv[0]);

	err = lzc_destroy_bookmarks(nvl, NULL);
	if (err != 0) {
	(void) zfs_standard_error(g_zfs, err,
	"cannot destroy bookmark");
	}

	nvlist_free(nvl);

	return (err);
	} else {
	/* Open the given dataset */
	if ((zhp = zfs_open(g_zfs, argv[0], type)) == NULL)
	return (1);

	cb.cb_target = zhp;

	/*
	* Perform an explicit check for pools before going any further.
	*/
	if (!cb.cb_recurse && strchr(zfs_get_name(zhp), '/') == NULL &&
	zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"operation does not apply to pools\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use 'zfs destroy -r "
	"%s' to destroy all datasets in the pool\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use 'zpool destroy %s' "
	"to destroy the pool itself\n"), zfs_get_name(zhp));
	rv = 1;
	goto out;
	}

	/*
	* Check for any dependents and/or clones.
	*/
	cb.cb_first = B_TRUE;
	if (!cb.cb_doclones &&
	zfs_iter_dependents(zhp, B_TRUE, destroy_check_dependent,
	&cb) != 0) {
	rv = 1;
	goto out;
	}

	if (cb.cb_error) {
	rv = 1;
	goto out;
	}
	cb.cb_batchedsnaps = fnvlist_alloc();
	if (zfs_iter_dependents(zhp, B_FALSE, destroy_callback,
	&cb) != 0) {
	rv = 1;
	goto out;
	}

	/*
	* Do the real thing. The callback will close the
	* handle regardless of whether it succeeds or not.
	*/
	err = destroy_callback(zhp, &cb);
	zhp = NULL;
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs,
	cb.cb_batchedsnaps, cb.cb_defer_destroy);
	}
	if (err != 0 \|\| cb.cb_error == B_TRUE)
	rv = 1;
	}

	out:
	fnvlist_free(cb.cb_batchedsnaps);
	fnvlist_free(cb.cb_nvl);
	if (zhp != NULL)
	zfs_close(zhp);
	return (rv);
	}

	static boolean_t
	is_recvd_column(zprop_get_cbdata_t *cbp)
	{
	int i;
	zfs_get_column_t col;

	for (i = 0; i < ZFS_GET_NCOLS &&
	(col = cbp->cb_columns[i]) != GET_COL_NONE; i++)
	if (col == GET_COL_RECVD)
	return (B_TRUE);
	return (B_FALSE);
	}

	/*
	* zfs get [-rHp] [-o all \| field[,field]...] [-s source[,source]...]
	* < all \| property[,property]... > < fs \| snap \| vol > ...
	*
	* -r recurse over any child datasets
	* -H scripted mode. Headers are stripped, and fields are separated
	* by tabs instead of spaces.
	* -o Set of fields to display. One of "name,property,value,
	* received,source". Default is "name,property,value,source".
	* "all" is an alias for all five.
	* -s Set of sources to allow. One of
	* "local,default,inherited,received,temporary,none". Default is
	* all six.
	* -p Display values in parsable (literal) format.
	*
	* Prints properties for the given datasets. The user can control which
	* columns to display as well as which property types to allow.
	*/

	/*
	* Invoked to display the properties for a single dataset.
	*/
	static int
	get_callback(zfs_handle_t zhp, void data)
	{
	char buf[ZFS_MAXPROPLEN];
	char rbuf[ZFS_MAXPROPLEN];
	zprop_source_t sourcetype;
	char source[ZFS_MAX_DATASET_NAME_LEN];
	zprop_get_cbdata_t *cbp = data;
	nvlist_t *user_props = zfs_get_user_props(zhp);
	zprop_list_t *pl = cbp->cb_proplist;
	nvlist_t *propval;
	const char *strval;
	const char *sourceval;
	boolean_t received = is_recvd_column(cbp);

	for (; pl != NULL; pl = pl->pl_next) {
	char *recvdval = NULL;
	/*
	* Skip the special fake placeholder. This will also skip over
	* the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZFS_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop != ZPROP_USERPROP) {
	if (zfs_prop_get(zhp, pl->pl_prop, buf,
	sizeof (buf), &sourcetype, source,
	sizeof (source),
	cbp->cb_literal) != 0) {
	if (pl->pl_all)
	continue;
	if (!zfs_prop_valid_for_type(pl->pl_prop,
	ZFS_TYPE_DATASET, B_FALSE)) {
	(void) fprintf(stderr,
	gettext("No such property '%s'\n"),
	zfs_prop_to_name(pl->pl_prop));
	continue;
	}
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	if (received && (zfs_prop_get_recvd(zhp,
	zfs_prop_to_name(pl->pl_prop), rbuf, sizeof (rbuf),
	cbp->cb_literal) == 0))
	recvdval = rbuf;

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	zfs_prop_to_name(pl->pl_prop),
	buf, sourcetype, source, recvdval);
	} else if (zfs_prop_userquota(pl->pl_user_prop)) {
	sourcetype = ZPROP_SRC_LOCAL;

	if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
	buf, sizeof (buf), cbp->cb_literal) != 0) {
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, buf, sourcetype, source, NULL);
	} else if (zfs_prop_written(pl->pl_user_prop)) {
	sourcetype = ZPROP_SRC_LOCAL;

	if (zfs_prop_get_written(zhp, pl->pl_user_prop,
	buf, sizeof (buf), cbp->cb_literal) != 0) {
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, buf, sourcetype, source, NULL);
	} else {
	if (nvlist_lookup_nvlist(user_props,
	pl->pl_user_prop, &propval) != 0) {
	if (pl->pl_all)
	continue;
	sourcetype = ZPROP_SRC_NONE;
	strval = "-";
	} else {
	strval = fnvlist_lookup_string(propval,
	ZPROP_VALUE);
	sourceval = fnvlist_lookup_string(propval,
	ZPROP_SOURCE);

	if (strcmp(sourceval,
	zfs_get_name(zhp)) == 0) {
	sourcetype = ZPROP_SRC_LOCAL;
	} else if (strcmp(sourceval,
	ZPROP_SOURCE_VAL_RECVD) == 0) {
	sourcetype = ZPROP_SRC_RECEIVED;
	} else {
	sourcetype = ZPROP_SRC_INHERITED;
	(void) strlcpy(source,
	sourceval, sizeof (source));
	}
	}

	if (received && (zfs_prop_get_recvd(zhp,
	pl->pl_user_prop, rbuf, sizeof (rbuf),
	cbp->cb_literal) == 0))
	recvdval = rbuf;

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, strval, sourcetype,
	source, recvdval);
	}
	}

	return (0);
	}

	static int
	zfs_do_get(int argc, char **argv)
	{
	zprop_get_cbdata_t cb = { 0 };
	int i, c, flags = ZFS_ITER_ARGS_CAN_BE_PATHS;
	int types = ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK;
	char *fields;
	int ret = 0;
	int limit = 0;
	zprop_list_t fake_name = { 0 };

	/*
	* Set up default columns and sources.
	*/
	cb.cb_sources = ZPROP_SRC_ALL;
	cb.cb_columns[0] = GET_COL_NAME;
	cb.cb_columns[1] = GET_COL_PROPERTY;
	cb.cb_columns[2] = GET_COL_VALUE;
	cb.cb_columns[3] = GET_COL_SOURCE;
	cb.cb_type = ZFS_TYPE_DATASET;

	/* check options */
	while ((c = getopt(argc, argv, ":d:o:s:rt:Hp")) != -1) {
	switch (c) {
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'd':
	limit = parse_depth(optarg, &flags);
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case 'o':
	/*
	* Process the set of columns to display. We zero out
	* the structure to give us a blank slate.
	*/
	memset(&cb.cb_columns, 0, sizeof (cb.cb_columns));

	i = 0;
	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_subopts[] =
	{ "name", "property", "value",
	"received", "source", "all" };
	static const zfs_get_column_t col_subopt_col[] =
	{ GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE,
	GET_COL_RECVD, GET_COL_SOURCE };
	static const int col_subopt_flags[] =
	{ 0, 0, 0, ZFS_ITER_RECVD_PROPS, 0 };

	if (i == ZFS_GET_NCOLS) {
	(void) fprintf(stderr, gettext("too "
	"many fields given to -o "
	"option\n"));
	usage(B_FALSE);
	}

	for (c = 0; c < ARRAY_SIZE(col_subopts); ++c)
	if (strcmp(tok, col_subopts[c]) == 0)
	goto found;

	(void) fprintf(stderr,
	gettext("invalid column name '%s'\n"), tok);
	usage(B_FALSE);

	found:
	if (c >= 5) {
	if (i > 0) {
	(void) fprintf(stderr,
	gettext("\"all\" conflicts "
	"with specific fields "
	"given to -o option\n"));
	usage(B_FALSE);
	}

	memcpy(cb.cb_columns, col_subopt_col,
	sizeof (col_subopt_col));
	flags \|= ZFS_ITER_RECVD_PROPS;
	i = ZFS_GET_NCOLS;
	} else {
	cb.cb_columns[i++] = col_subopt_col[c];
	flags \|= col_subopt_flags[c];
	}
	}
	break;

	case 's':
	cb.cb_sources = 0;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const source_opt[] = {
	"local", "default",
	"inherited", "received",
	"temporary", "none" };
	static const int source_flg[] = {
	ZPROP_SRC_LOCAL, ZPROP_SRC_DEFAULT,
	ZPROP_SRC_INHERITED, ZPROP_SRC_RECEIVED,
	ZPROP_SRC_TEMPORARY, ZPROP_SRC_NONE };

	for (i = 0; i < ARRAY_SIZE(source_opt); ++i)
	if (strcmp(tok, source_opt[i]) == 0) {
	cb.cb_sources \|= source_flg[i];
	goto found2;
	}

	(void) fprintf(stderr,
	gettext("invalid source '%s'\n"), tok);
	usage(B_FALSE);
	found2:;
	}
	break;

	case 't':
	types = 0;
	flags &= ~ZFS_ITER_PROP_LISTSNAPS;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const type_opts[] = {
	"filesystem", "volume",
	"snapshot", "snap",
	"bookmark",
	"all" };
	static const int type_types[] = {
	ZFS_TYPE_FILESYSTEM, ZFS_TYPE_VOLUME,
	ZFS_TYPE_SNAPSHOT, ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_BOOKMARK,
	ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK };

	for (i = 0; i < ARRAY_SIZE(type_opts); ++i)
	if (strcmp(tok, type_opts[i]) == 0) {
	types \|= type_types[i];
	goto found3;
	}

	(void) fprintf(stderr,
	gettext("invalid type '%s'\n"), tok);
	usage(B_FALSE);
	found3:;
	}
	break;

	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property "
	"argument\n"));
	usage(B_FALSE);
	}

	fields = argv[0];

	/*
	* Handle users who want to get all snapshots or bookmarks
	* of a dataset (ex. 'zfs get -t snapshot refer <dataset>').
	*/
	if ((types == ZFS_TYPE_SNAPSHOT \|\| types == ZFS_TYPE_BOOKMARK) &&
	argc > 1 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) {
	flags \|= (ZFS_ITER_DEPTH_LIMIT \| ZFS_ITER_RECURSE);
	limit = 1;
	}

	if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET)
	!= 0)
	usage(B_FALSE);

	argc--;
	argv++;

	/*
	* As part of zfs_expand_proplist(), we keep track of the maximum column
	* width for each property. For the 'NAME' (and 'SOURCE') columns, we
	* need to know the maximum name length. However, the user likely did
	* not specify 'name' as one of the properties to fetch, so we need to
	* make sure we always include at least this property for
	* print_get_headers() to work properly.
	*/
	if (cb.cb_proplist != NULL) {
	fake_name.pl_prop = ZFS_PROP_NAME;
	fake_name.pl_width = strlen(gettext("NAME"));
	fake_name.pl_next = cb.cb_proplist;
	cb.cb_proplist = &fake_name;
	}

	cb.cb_first = B_TRUE;

	/* run for each object */
	ret = zfs_for_each(argc, argv, flags, types, NULL,
	&cb.cb_proplist, limit, get_callback, &cb);

	if (cb.cb_proplist == &fake_name)
	zprop_free_list(fake_name.pl_next);
	else
	zprop_free_list(cb.cb_proplist);

	return (ret);
	}

	/*
	* inherit [-rS] <property> <fs\|vol> ...
	*
	* -r Recurse over all children
	* -S Revert to received value, if any
	*
	* For each dataset specified on the command line, inherit the given property
	* from its parent. Inheriting a property at the pool level will cause it to
	* use the default value. The '-r' flag will recurse over all children, and is
	* useful for setting a property on a hierarchy-wide basis, regardless of any
	* local modifications for each dataset.
	*/

	typedef struct inherit_cbdata {
	const char *cb_propname;
	boolean_t cb_received;
	} inherit_cbdata_t;

	static int
	inherit_recurse_cb(zfs_handle_t zhp, void data)
	{
	inherit_cbdata_t *cb = data;
	zfs_prop_t prop = zfs_name_to_prop(cb->cb_propname);

	/*
	* If we're doing it recursively, then ignore properties that
	* are not valid for this type of dataset.
	*/
	if (prop != ZPROP_INVAL &&
	!zfs_prop_valid_for_type(prop, zfs_get_type(zhp), B_FALSE))
	return (0);

	return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0);
	}

	static int
	inherit_cb(zfs_handle_t zhp, void data)
	{
	inherit_cbdata_t *cb = data;

	return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0);
	}

	static int
	zfs_do_inherit(int argc, char **argv)
	{
	int c;
	zfs_prop_t prop;
	inherit_cbdata_t cb = { 0 };
	char *propname;
	int ret = 0;
	int flags = 0;
	boolean_t received = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "rS")) != -1) {
	switch (c) {
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'S':
	received = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}

	propname = argv[0];
	argc--;
	argv++;

	if ((prop = zfs_name_to_prop(propname)) != ZPROP_USERPROP) {
	if (zfs_prop_readonly(prop)) {
	(void) fprintf(stderr, gettext(
	"%s property is read-only\n"),
	propname);
	return (1);
	}
	if (!zfs_prop_inheritable(prop) && !received) {
	(void) fprintf(stderr, gettext("'%s' property cannot "
	"be inherited\n"), propname);
	if (prop == ZFS_PROP_QUOTA \|\|
	prop == ZFS_PROP_RESERVATION \|\|
	prop == ZFS_PROP_REFQUOTA \|\|
	prop == ZFS_PROP_REFRESERVATION) {
	(void) fprintf(stderr, gettext("use 'zfs set "
	"%s=none' to clear\n"), propname);
	(void) fprintf(stderr, gettext("use 'zfs "
	"inherit -S %s' to revert to received "
	"value\n"), propname);
	}
	return (1);
	}
	if (received && (prop == ZFS_PROP_VOLSIZE \|\|
	prop == ZFS_PROP_VERSION)) {
	(void) fprintf(stderr, gettext("'%s' property cannot "
	"be reverted to a received value\n"), propname);
	return (1);
	}
	} else if (!zfs_prop_user(propname)) {
	(void) fprintf(stderr, gettext("invalid property '%s'\n"),
	propname);
	usage(B_FALSE);
	}

	cb.cb_propname = propname;
	cb.cb_received = received;

	if (flags & ZFS_ITER_RECURSE) {
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
	NULL, NULL, 0, inherit_recurse_cb, &cb);
	} else {
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
	NULL, NULL, 0, inherit_cb, &cb);
	}

	return (ret);
	}

	typedef struct upgrade_cbdata {
	uint64_t cb_numupgraded;
	uint64_t cb_numsamegraded;
	uint64_t cb_numfailed;
	uint64_t cb_version;
	boolean_t cb_newer;
	boolean_t cb_foundone;
	char cb_lastfs[ZFS_MAX_DATASET_NAME_LEN];
	} upgrade_cbdata_t;

	static int
	same_pool(zfs_handle_t zhp, const char name)
	{
	int len1 = strcspn(name, "/@");
	const char *zhname = zfs_get_name(zhp);
	int len2 = strcspn(zhname, "/@");

	if (len1 != len2)
	return (B_FALSE);
	return (strncmp(name, zhname, len1) == 0);
	}

	static int
	upgrade_list_callback(zfs_handle_t zhp, void data)
	{
	upgrade_cbdata_t *cb = data;
	int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);

	/* list if it's old/new */
	if ((!cb->cb_newer && version < ZPL_VERSION) \|\|
	(cb->cb_newer && version > ZPL_VERSION)) {
	char *str;
	if (cb->cb_newer) {
	str = gettext("The following filesystems are "
	"formatted using a newer software version and\n"
	"cannot be accessed on the current system.\n\n");
	} else {
	str = gettext("The following filesystems are "
	"out of date, and can be upgraded. After being\n"
	"upgraded, these filesystems (and any 'zfs send' "
	"streams generated from\n"
	"subsequent snapshots) will no longer be "
	"accessible by older software versions.\n\n");
	}

	if (!cb->cb_foundone) {
	(void) puts(str);
	(void) printf(gettext("VER FILESYSTEM\n"));
	(void) printf(gettext("--- ------------\n"));
	cb->cb_foundone = B_TRUE;
	}

	(void) printf("%2u %s\n", version, zfs_get_name(zhp));
	}

	return (0);
	}

	static int
	upgrade_set_callback(zfs_handle_t zhp, void data)
	{
	upgrade_cbdata_t *cb = data;
	int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
	int needed_spa_version;
	int spa_version;

	if (zfs_spa_version(zhp, &spa_version) < 0)
	return (-1);

	needed_spa_version = zfs_spa_version_map(cb->cb_version);

	if (needed_spa_version < 0)
	return (-1);

	if (spa_version < needed_spa_version) {
	/* can't upgrade */
	(void) printf(gettext("%s: can not be "
	"upgraded; the pool version needs to first "
	"be upgraded\nto version %d\n\n"),
	zfs_get_name(zhp), needed_spa_version);
	cb->cb_numfailed++;
	return (0);
	}

	/* upgrade */
	if (version < cb->cb_version) {
	char verstr[24];
	(void) snprintf(verstr, sizeof (verstr),
	"%llu", (u_longlong_t)cb->cb_version);
	if (cb->cb_lastfs[0] && !same_pool(zhp, cb->cb_lastfs)) {
	/*
	* If they did "zfs upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}
	if (zfs_prop_set(zhp, "version", verstr) == 0)
	cb->cb_numupgraded++;
	else
	cb->cb_numfailed++;
	- (void) strcpy(cb->cb_lastfs, zfs_get_name(zhp));
	+ (void) strlcpy(cb->cb_lastfs, zfs_get_name(zhp),
	+ sizeof (cb->cb_lastfs));
	} else if (version > cb->cb_version) {
	/* can't downgrade */
	(void) printf(gettext("%s: can not be downgraded; "
	"it is already at version %u\n"),
	zfs_get_name(zhp), version);
	cb->cb_numfailed++;
	} else {
	cb->cb_numsamegraded++;
	}
	return (0);
	}

	/*
	* zfs upgrade
	* zfs upgrade -v
	* zfs upgrade [-r] [-V <version>] <-a \| filesystem>
	*/
	static int
	zfs_do_upgrade(int argc, char **argv)
	{
	boolean_t all = B_FALSE;
	boolean_t showversions = B_FALSE;
	int ret = 0;
	upgrade_cbdata_t cb = { 0 };
	int c;
	int flags = ZFS_ITER_ARGS_CAN_BE_PATHS;

	/* check options */
	while ((c = getopt(argc, argv, "rvV:a")) != -1) {
	switch (c) {
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'v':
	showversions = B_TRUE;
	break;
	case 'V':
	if (zfs_prop_string_to_index(ZFS_PROP_VERSION,
	optarg, &cb.cb_version) != 0) {
	(void) fprintf(stderr,
	gettext("invalid version %s\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 'a':
	all = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if ((!all && !argc) && ((flags & ZFS_ITER_RECURSE) \| cb.cb_version))
	usage(B_FALSE);
	if (showversions && (flags & ZFS_ITER_RECURSE \|\| all \|\|
	cb.cb_version \|\| argc))
	usage(B_FALSE);
	if ((all \|\| argc) && (showversions))
	usage(B_FALSE);
	if (all && argc)
	usage(B_FALSE);

	if (showversions) {
	/* Show info on available versions. */
	(void) printf(gettext("The following filesystem versions are "
	"supported:\n\n"));
	(void) printf(gettext("VER DESCRIPTION\n"));
	(void) printf("--- -----------------------------------------"
	"---------------\n");
	(void) printf(gettext(" 1 Initial ZFS filesystem version\n"));
	(void) printf(gettext(" 2 Enhanced directory entries\n"));
	(void) printf(gettext(" 3 Case insensitive and filesystem "
	"user identifier (FUID)\n"));
	(void) printf(gettext(" 4 userquota, groupquota "
	"properties\n"));
	(void) printf(gettext(" 5 System attributes\n"));
	(void) printf(gettext("\nFor more information on a particular "
	"version, including supported releases,\n"));
	(void) printf("see the ZFS Administration Guide.\n\n");
	ret = 0;
	} else if (argc \|\| all) {
	/* Upgrade filesystems */
	if (cb.cb_version == 0)
	cb.cb_version = ZPL_VERSION;
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_set_callback, &cb);
	(void) printf(gettext("%llu filesystems upgraded\n"),
	(u_longlong_t)cb.cb_numupgraded);
	if (cb.cb_numsamegraded) {
	(void) printf(gettext("%llu filesystems already at "
	"this version\n"),
	(u_longlong_t)cb.cb_numsamegraded);
	}
	if (cb.cb_numfailed != 0)
	ret = 1;
	} else {
	/* List old-version filesystems */
	boolean_t found;
	(void) printf(gettext("This system is currently running "
	"ZFS filesystem version %llu.\n\n"), ZPL_VERSION);

	flags \|= ZFS_ITER_RECURSE;
	ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_list_callback, &cb);

	found = cb.cb_foundone;
	cb.cb_foundone = B_FALSE;
	cb.cb_newer = B_TRUE;

	ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_list_callback, &cb);

	if (!cb.cb_foundone && !found) {
	(void) printf(gettext("All filesystems are "
	"formatted with the current version.\n"));
	}
	}

	return (ret);
	}

	/*
	* zfs userspace [-Hinp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] [-t type[,...]]
	* filesystem \| snapshot \| path
	* zfs groupspace [-Hinp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] [-t type[,...]]
	* filesystem \| snapshot \| path
	* zfs projectspace [-Hp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] filesystem \| snapshot \| path
	*
	* -H Scripted mode; elide headers and separate columns by tabs.
	* -i Translate SID to POSIX ID.
	* -n Print numeric ID instead of user/group name.
	* -o Control which fields to display.
	* -p Use exact (parsable) numeric output.
	* -s Specify sort columns, descending order.
	* -S Specify sort columns, ascending order.
	* -t Control which object types to display.
	*
	* Displays space consumed by, and quotas on, each user in the specified
	* filesystem or snapshot.
	*/

	/* us_field_types, us_field_hdr and us_field_names should be kept in sync */
	enum us_field_types {
	USFIELD_TYPE,
	USFIELD_NAME,
	USFIELD_USED,
	USFIELD_QUOTA,
	USFIELD_OBJUSED,
	USFIELD_OBJQUOTA
	};
	static const char *const us_field_hdr[] = { "TYPE", "NAME", "USED", "QUOTA",
	"OBJUSED", "OBJQUOTA" };
	static const char *const us_field_names[] = { "type", "name", "used", "quota",
	"objused", "objquota" };
	#define USFIELD_LAST (sizeof (us_field_names) / sizeof (char *))

	#define USTYPE_PSX_GRP (1 << 0)
	#define USTYPE_PSX_USR (1 << 1)
	#define USTYPE_SMB_GRP (1 << 2)
	#define USTYPE_SMB_USR (1 << 3)
	#define USTYPE_PROJ (1 << 4)
	#define USTYPE_ALL \
	(USTYPE_PSX_GRP \| USTYPE_PSX_USR \| USTYPE_SMB_GRP \| USTYPE_SMB_USR \| \
	USTYPE_PROJ)

	static int us_type_bits[] = {
	USTYPE_PSX_GRP,
	USTYPE_PSX_USR,
	USTYPE_SMB_GRP,
	USTYPE_SMB_USR,
	USTYPE_ALL
	};
	static const char *const us_type_names[] = { "posixgroup", "posixuser",
	"smbgroup", "smbuser", "all" };

	typedef struct us_node {
	nvlist_t *usn_nvl;
	uu_avl_node_t usn_avlnode;
	uu_list_node_t usn_listnode;
	} us_node_t;

	typedef struct us_cbdata {
	nvlist_t **cb_nvlp;
	uu_avl_pool_t *cb_avl_pool;
	uu_avl_t *cb_avl;
	boolean_t cb_numname;
	boolean_t cb_nicenum;
	boolean_t cb_sid2posix;
	zfs_userquota_prop_t cb_prop;
	zfs_sort_column_t *cb_sortcol;
	size_t cb_width[USFIELD_LAST];
	} us_cbdata_t;

	static boolean_t us_populated = B_FALSE;

	typedef struct {
	zfs_sort_column_t *si_sortcol;
	boolean_t si_numname;
	} us_sort_info_t;

	static int
	us_field_index(const char *field)
	{
	for (int i = 0; i < USFIELD_LAST; i++) {
	if (strcmp(field, us_field_names[i]) == 0)
	return (i);
	}

	return (-1);
	}

	static int
	us_compare(const void larg, const void rarg, void *unused)
	{
	const us_node_t *l = larg;
	const us_node_t *r = rarg;
	us_sort_info_t si = (us_sort_info_t )unused;
	zfs_sort_column_t *sortcol = si->si_sortcol;
	boolean_t numname = si->si_numname;
	nvlist_t *lnvl = l->usn_nvl;
	nvlist_t *rnvl = r->usn_nvl;
	int rc = 0;
	boolean_t lvb, rvb;

	for (; sortcol != NULL; sortcol = sortcol->sc_next) {
	char lvstr = (char )"";
	char rvstr = (char )"";
	uint32_t lv32 = 0;
	uint32_t rv32 = 0;
	uint64_t lv64 = 0;
	uint64_t rv64 = 0;
	zfs_prop_t prop = sortcol->sc_prop;
	const char *propname = NULL;
	boolean_t reverse = sortcol->sc_reverse;

	switch (prop) {
	case ZFS_PROP_TYPE:
	propname = "type";
	(void) nvlist_lookup_uint32(lnvl, propname, &lv32);
	(void) nvlist_lookup_uint32(rnvl, propname, &rv32);
	if (rv32 != lv32)
	rc = (rv32 < lv32) ? 1 : -1;
	break;
	case ZFS_PROP_NAME:
	propname = "name";
	if (numname) {
	compare_nums:
	(void) nvlist_lookup_uint64(lnvl, propname,
	&lv64);
	(void) nvlist_lookup_uint64(rnvl, propname,
	&rv64);
	if (rv64 != lv64)
	rc = (rv64 < lv64) ? 1 : -1;
	} else {
	if ((nvlist_lookup_string(lnvl, propname,
	&lvstr) == ENOENT) \|\|
	(nvlist_lookup_string(rnvl, propname,
	&rvstr) == ENOENT)) {
	goto compare_nums;
	}
	rc = strcmp(lvstr, rvstr);
	}
	break;
	case ZFS_PROP_USED:
	case ZFS_PROP_QUOTA:
	if (!us_populated)
	break;
	if (prop == ZFS_PROP_USED)
	propname = "used";
	else
	propname = "quota";
	(void) nvlist_lookup_uint64(lnvl, propname, &lv64);
	(void) nvlist_lookup_uint64(rnvl, propname, &rv64);
	if (rv64 != lv64)
	rc = (rv64 < lv64) ? 1 : -1;
	break;

	default:
	break;
	}

	if (rc != 0) {
	if (rc < 0)
	return (reverse ? 1 : -1);
	else
	return (reverse ? -1 : 1);
	}
	}

	/*
	* If entries still seem to be the same, check if they are of the same
	* type (smbentity is added only if we are doing SID to POSIX ID
	* translation where we can have duplicate type/name combinations).
	*/
	if (nvlist_lookup_boolean_value(lnvl, "smbentity", &lvb) == 0 &&
	nvlist_lookup_boolean_value(rnvl, "smbentity", &rvb) == 0 &&
	lvb != rvb)
	return (lvb < rvb ? -1 : 1);

	return (0);
	}

	static boolean_t
	zfs_prop_is_user(unsigned p)
	{
	return (p == ZFS_PROP_USERUSED \|\| p == ZFS_PROP_USERQUOTA \|\|
	p == ZFS_PROP_USEROBJUSED \|\| p == ZFS_PROP_USEROBJQUOTA);
	}

	static boolean_t
	zfs_prop_is_group(unsigned p)
	{
	return (p == ZFS_PROP_GROUPUSED \|\| p == ZFS_PROP_GROUPQUOTA \|\|
	p == ZFS_PROP_GROUPOBJUSED \|\| p == ZFS_PROP_GROUPOBJQUOTA);
	}

	static boolean_t
	zfs_prop_is_project(unsigned p)
	{
	return (p == ZFS_PROP_PROJECTUSED \|\| p == ZFS_PROP_PROJECTQUOTA \|\|
	p == ZFS_PROP_PROJECTOBJUSED \|\| p == ZFS_PROP_PROJECTOBJQUOTA);
	}

	static inline const char *
	us_type2str(unsigned field_type)
	{
	switch (field_type) {
	case USTYPE_PSX_USR:
	return ("POSIX User");
	case USTYPE_PSX_GRP:
	return ("POSIX Group");
	case USTYPE_SMB_USR:
	return ("SMB User");
	case USTYPE_SMB_GRP:
	return ("SMB Group");
	case USTYPE_PROJ:
	return ("Project");
	default:
	return ("Undefined");
	}
	}

	static int
	userspace_cb(void arg, const char domain, uid_t rid, uint64_t space)
	{
	us_cbdata_t cb = (us_cbdata_t )arg;
	zfs_userquota_prop_t prop = cb->cb_prop;
	char *name = NULL;
	const char *propname;
	char sizebuf[32];
	us_node_t *node;
	uu_avl_pool_t *avl_pool = cb->cb_avl_pool;
	uu_avl_t *avl = cb->cb_avl;
	uu_avl_index_t idx;
	nvlist_t *props;
	us_node_t *n;
	zfs_sort_column_t *sortcol = cb->cb_sortcol;
	unsigned type = 0;
	const char *typestr;
	size_t namelen;
	size_t typelen;
	size_t sizelen;
	int typeidx, nameidx, sizeidx;
	us_sort_info_t sortinfo = { sortcol, cb->cb_numname };
	boolean_t smbentity = B_FALSE;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	node = safe_malloc(sizeof (us_node_t));
	uu_avl_node_init(node, &node->usn_avlnode, avl_pool);
	node->usn_nvl = props;

	if (domain != NULL && domain[0] != '\0') {
	#ifdef HAVE_IDMAP
	/* SMB */
	char sid[MAXNAMELEN + 32];
	uid_t id;
	uint64_t classes;
	int err;
	directory_error_t e;

	smbentity = B_TRUE;

	(void) snprintf(sid, sizeof (sid), "%s-%u", domain, rid);

	if (prop == ZFS_PROP_GROUPUSED \|\| prop == ZFS_PROP_GROUPQUOTA) {
	type = USTYPE_SMB_GRP;
	err = sid_to_id(sid, B_FALSE, &id);
	} else {
	type = USTYPE_SMB_USR;
	err = sid_to_id(sid, B_TRUE, &id);
	}

	if (err == 0) {
	rid = id;
	if (!cb->cb_sid2posix) {
	e = directory_name_from_sid(NULL, sid, &name,
	&classes);
	if (e != NULL)
	directory_error_free(e);
	if (name == NULL)
	name = sid;
	}
	}
	#else
	nvlist_free(props);
	free(node);

	return (-1);
	#endif /* HAVE_IDMAP */
	}

	if (cb->cb_sid2posix \|\| domain == NULL \|\| domain[0] == '\0') {
	/* POSIX or -i */
	if (zfs_prop_is_group(prop)) {
	type = USTYPE_PSX_GRP;
	if (!cb->cb_numname) {
	struct group *g;

	if ((g = getgrgid(rid)) != NULL)
	name = g->gr_name;
	}
	} else if (zfs_prop_is_user(prop)) {
	type = USTYPE_PSX_USR;
	if (!cb->cb_numname) {
	struct passwd *p;

	if ((p = getpwuid(rid)) != NULL)
	name = p->pw_name;
	}
	} else {
	type = USTYPE_PROJ;
	}
	}

	/*
	* Make sure that the type/name combination is unique when doing
	* SID to POSIX ID translation (hence changing the type from SMB to
	* POSIX).
	*/
	if (cb->cb_sid2posix &&
	nvlist_add_boolean_value(props, "smbentity", smbentity) != 0)
	nomem();

	/* Calculate/update width of TYPE field */
	typestr = us_type2str(type);
	typelen = strlen(gettext(typestr));
	typeidx = us_field_index("type");
	if (typelen > cb->cb_width[typeidx])
	cb->cb_width[typeidx] = typelen;
	if (nvlist_add_uint32(props, "type", type) != 0)
	nomem();

	/* Calculate/update width of NAME field */
	if ((cb->cb_numname && cb->cb_sid2posix) \|\| name == NULL) {
	if (nvlist_add_uint64(props, "name", rid) != 0)
	nomem();
	namelen = snprintf(NULL, 0, "%u", rid);
	} else {
	if (nvlist_add_string(props, "name", name) != 0)
	nomem();
	namelen = strlen(name);
	}
	nameidx = us_field_index("name");
	if (nameidx >= 0 && namelen > cb->cb_width[nameidx])
	cb->cb_width[nameidx] = namelen;

	/*
	* Check if this type/name combination is in the list and update it;
	* otherwise add new node to the list.
	*/
	if ((n = uu_avl_find(avl, node, &sortinfo, &idx)) == NULL) {
	uu_avl_insert(avl, node, idx);
	} else {
	nvlist_free(props);
	free(node);
	node = n;
	props = node->usn_nvl;
	}

	/* Calculate/update width of USED/QUOTA fields */
	if (cb->cb_nicenum) {
	if (prop == ZFS_PROP_USERUSED \|\| prop == ZFS_PROP_GROUPUSED \|\|
	prop == ZFS_PROP_USERQUOTA \|\| prop == ZFS_PROP_GROUPQUOTA \|\|
	prop == ZFS_PROP_PROJECTUSED \|\|
	prop == ZFS_PROP_PROJECTQUOTA) {
	zfs_nicebytes(space, sizebuf, sizeof (sizebuf));
	} else {
	zfs_nicenum(space, sizebuf, sizeof (sizebuf));
	}
	} else {
	(void) snprintf(sizebuf, sizeof (sizebuf), "%llu",
	(u_longlong_t)space);
	}
	sizelen = strlen(sizebuf);
	if (prop == ZFS_PROP_USERUSED \|\| prop == ZFS_PROP_GROUPUSED \|\|
	prop == ZFS_PROP_PROJECTUSED) {
	propname = "used";
	if (!nvlist_exists(props, "quota"))
	(void) nvlist_add_uint64(props, "quota", 0);
	} else if (prop == ZFS_PROP_USERQUOTA \|\| prop == ZFS_PROP_GROUPQUOTA \|\|
	prop == ZFS_PROP_PROJECTQUOTA) {
	propname = "quota";
	if (!nvlist_exists(props, "used"))
	(void) nvlist_add_uint64(props, "used", 0);
	} else if (prop == ZFS_PROP_USEROBJUSED \|\|
	prop == ZFS_PROP_GROUPOBJUSED \|\| prop == ZFS_PROP_PROJECTOBJUSED) {
	propname = "objused";
	if (!nvlist_exists(props, "objquota"))
	(void) nvlist_add_uint64(props, "objquota", 0);
	} else if (prop == ZFS_PROP_USEROBJQUOTA \|\|
	prop == ZFS_PROP_GROUPOBJQUOTA \|\|
	prop == ZFS_PROP_PROJECTOBJQUOTA) {
	propname = "objquota";
	if (!nvlist_exists(props, "objused"))
	(void) nvlist_add_uint64(props, "objused", 0);
	} else {
	return (-1);
	}
	sizeidx = us_field_index(propname);
	if (sizeidx >= 0 && sizelen > cb->cb_width[sizeidx])
	cb->cb_width[sizeidx] = sizelen;

	if (nvlist_add_uint64(props, propname, space) != 0)
	nomem();

	return (0);
	}

	static void
	print_us_node(boolean_t scripted, boolean_t parsable, int *fields, int types,
	size_t width, us_node_t node)
	{
	nvlist_t *nvl = node->usn_nvl;
	char valstr[MAXNAMELEN];
	boolean_t first = B_TRUE;
	int cfield = 0;
	int field;
	uint32_t ustype;

	/* Check type */
	(void) nvlist_lookup_uint32(nvl, "type", &ustype);
	if (!(ustype & types))
	return;

	while ((field = fields[cfield]) != USFIELD_LAST) {
	nvpair_t *nvp = NULL;
	data_type_t type;
	uint32_t val32 = -1;
	uint64_t val64 = -1;
	const char *strval = "-";

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL)
	if (strcmp(nvpair_name(nvp),
	us_field_names[field]) == 0)
	break;

	type = nvp == NULL ? DATA_TYPE_UNKNOWN : nvpair_type(nvp);
	switch (type) {
	case DATA_TYPE_UINT32:
	val32 = fnvpair_value_uint32(nvp);
	break;
	case DATA_TYPE_UINT64:
	val64 = fnvpair_value_uint64(nvp);
	break;
	case DATA_TYPE_STRING:
	strval = fnvpair_value_string(nvp);
	break;
	case DATA_TYPE_UNKNOWN:
	break;
	default:
	(void) fprintf(stderr, "invalid data type\n");
	}

	switch (field) {
	case USFIELD_TYPE:
	if (type == DATA_TYPE_UINT32)
	strval = us_type2str(val32);
	break;
	case USFIELD_NAME:
	if (type == DATA_TYPE_UINT64) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	}
	break;
	case USFIELD_USED:
	case USFIELD_QUOTA:
	if (type == DATA_TYPE_UINT64) {
	if (parsable) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	} else if (field == USFIELD_QUOTA &&
	val64 == 0) {
	strval = "none";
	} else {
	zfs_nicebytes(val64, valstr,
	sizeof (valstr));
	strval = valstr;
	}
	}
	break;
	case USFIELD_OBJUSED:
	case USFIELD_OBJQUOTA:
	if (type == DATA_TYPE_UINT64) {
	if (parsable) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	} else if (field == USFIELD_OBJQUOTA &&
	val64 == 0) {
	strval = "none";
	} else {
	zfs_nicenum(val64, valstr,
	sizeof (valstr));
	strval = valstr;
	}
	}
	break;
	}

	if (!first) {
	if (scripted)
	(void) putchar('\t');
	else
	(void) fputs(" ", stdout);
	}
	if (scripted)
	(void) fputs(strval, stdout);
	else if (field == USFIELD_TYPE \|\| field == USFIELD_NAME)
	(void) printf("%-*s", (int)width[field], strval);
	else
	(void) printf("%*s", (int)width[field], strval);

	first = B_FALSE;
	cfield++;
	}

	(void) putchar('\n');
	}

	static void
	print_us(boolean_t scripted, boolean_t parsable, int *fields, int types,
	size_t width, boolean_t rmnode, uu_avl_t avl)
	{
	us_node_t *node;
	const char *col;
	int cfield = 0;
	int field;

	if (!scripted) {
	boolean_t first = B_TRUE;

	while ((field = fields[cfield]) != USFIELD_LAST) {
	col = gettext(us_field_hdr[field]);
	if (field == USFIELD_TYPE \|\| field == USFIELD_NAME) {
	(void) printf(first ? "%-s" : " %-s",
	(int)width[field], col);
	} else {
	(void) printf(first ? "%s" : " %s",
	(int)width[field], col);
	}
	first = B_FALSE;
	cfield++;
	}
	(void) printf("\n");
	}

	for (node = uu_avl_first(avl); node; node = uu_avl_next(avl, node)) {
	print_us_node(scripted, parsable, fields, types, width, node);
	if (rmnode)
	nvlist_free(node->usn_nvl);
	}
	}

	static int
	zfs_do_userspace(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	zfs_userquota_prop_t p;
	uu_avl_pool_t *avl_pool;
	uu_avl_t *avl_tree;
	uu_avl_walk_t *walk;
	char *delim;
	char deffields[] = "type,name,used,quota,objused,objquota";
	char *ofield = NULL;
	char *tfield = NULL;
	int cfield = 0;
	int fields[256];
	int i;
	boolean_t scripted = B_FALSE;
	boolean_t prtnum = B_FALSE;
	boolean_t parsable = B_FALSE;
	boolean_t sid2posix = B_FALSE;
	int ret = 0;
	int c;
	zfs_sort_column_t *sortcol = NULL;
	int types = USTYPE_PSX_USR \| USTYPE_SMB_USR;
	us_cbdata_t cb;
	us_node_t *node;
	us_node_t *rmnode;
	uu_list_pool_t *listpool;
	uu_list_t *list;
	uu_avl_index_t idx = 0;
	uu_list_index_t idx2 = 0;

	if (argc < 2)
	usage(B_FALSE);

	if (strcmp(argv[0], "groupspace") == 0) {
	/* Toggle default group types */
	types = USTYPE_PSX_GRP \| USTYPE_SMB_GRP;
	} else if (strcmp(argv[0], "projectspace") == 0) {
	types = USTYPE_PROJ;
	prtnum = B_TRUE;
	}

	while ((c = getopt(argc, argv, "nHpo:s:S:t:i")) != -1) {
	switch (c) {
	case 'n':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 'n'\n"));
	usage(B_FALSE);
	}
	prtnum = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 'o':
	ofield = optarg;
	break;
	case 's':
	case 'S':
	if (zfs_add_sort_column(&sortcol, optarg,
	c == 's' ? B_FALSE : B_TRUE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid field '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 't':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 't'\n"));
	usage(B_FALSE);
	}
	tfield = optarg;
	break;
	case 'i':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 'i'\n"));
	usage(B_FALSE);
	}
	sid2posix = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing dataset name\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/* Use default output fields if not specified using -o */
	if (ofield == NULL)
	ofield = deffields;
	do {
	if ((delim = strchr(ofield, ',')) != NULL)
	*delim = '\0';
	if ((fields[cfield++] = us_field_index(ofield)) == -1) {
	(void) fprintf(stderr, gettext("invalid type '%s' "
	"for -o option\n"), ofield);
	return (-1);
	}
	if (delim != NULL)
	ofield = delim + 1;
	} while (delim != NULL);
	fields[cfield] = USFIELD_LAST;

	/* Override output types (-t option) */
	if (tfield != NULL) {
	types = 0;

	do {
	boolean_t found = B_FALSE;

	if ((delim = strchr(tfield, ',')) != NULL)
	*delim = '\0';
	for (i = 0; i < sizeof (us_type_bits) / sizeof (int);
	i++) {
	if (strcmp(tfield, us_type_names[i]) == 0) {
	found = B_TRUE;
	types \|= us_type_bits[i];
	break;
	}
	}
	if (!found) {
	(void) fprintf(stderr, gettext("invalid type "
	"'%s' for -t option\n"), tfield);
	return (-1);
	}
	if (delim != NULL)
	tfield = delim + 1;
	} while (delim != NULL);
	}

	if ((zhp = zfs_path_to_zhandle(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_SNAPSHOT)) == NULL)
	return (1);
	if (zfs_get_underlying_type(zhp) != ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("operation is only applicable "
	"to filesystems and their snapshots\n"));
	zfs_close(zhp);
	return (1);
	}

	if ((avl_pool = uu_avl_pool_create("us_avl_pool", sizeof (us_node_t),
	offsetof(us_node_t, usn_avlnode), us_compare, UU_DEFAULT)) == NULL)
	nomem();
	if ((avl_tree = uu_avl_create(avl_pool, NULL, UU_DEFAULT)) == NULL)
	nomem();

	/* Always add default sorting columns */
	(void) zfs_add_sort_column(&sortcol, "type", B_FALSE);
	(void) zfs_add_sort_column(&sortcol, "name", B_FALSE);

	cb.cb_sortcol = sortcol;
	cb.cb_numname = prtnum;
	cb.cb_nicenum = !parsable;
	cb.cb_avl_pool = avl_pool;
	cb.cb_avl = avl_tree;
	cb.cb_sid2posix = sid2posix;

	for (i = 0; i < USFIELD_LAST; i++)
	cb.cb_width[i] = strlen(gettext(us_field_hdr[i]));

	for (p = 0; p < ZFS_NUM_USERQUOTA_PROPS; p++) {
	if ((zfs_prop_is_user(p) &&
	!(types & (USTYPE_PSX_USR \| USTYPE_SMB_USR))) \|\|
	(zfs_prop_is_group(p) &&
	!(types & (USTYPE_PSX_GRP \| USTYPE_SMB_GRP))) \|\|
	(zfs_prop_is_project(p) && types != USTYPE_PROJ))
	continue;

	cb.cb_prop = p;
	if ((ret = zfs_userspace(zhp, p, userspace_cb, &cb)) != 0) {
	zfs_close(zhp);
	return (ret);
	}
	}
	zfs_close(zhp);

	/* Sort the list */
	if ((node = uu_avl_first(avl_tree)) == NULL)
	return (0);

	us_populated = B_TRUE;

	listpool = uu_list_pool_create("tmplist", sizeof (us_node_t),
	offsetof(us_node_t, usn_listnode), NULL, UU_DEFAULT);
	list = uu_list_create(listpool, NULL, UU_DEFAULT);
	uu_list_node_init(node, &node->usn_listnode, listpool);

	while (node != NULL) {
	rmnode = node;
	node = uu_avl_next(avl_tree, node);
	uu_avl_remove(avl_tree, rmnode);
	if (uu_list_find(list, rmnode, NULL, &idx2) == NULL)
	uu_list_insert(list, rmnode, idx2);
	}

	for (node = uu_list_first(list); node != NULL;
	node = uu_list_next(list, node)) {
	us_sort_info_t sortinfo = { sortcol, cb.cb_numname };

	if (uu_avl_find(avl_tree, node, &sortinfo, &idx) == NULL)
	uu_avl_insert(avl_tree, node, idx);
	}

	uu_list_destroy(list);
	uu_list_pool_destroy(listpool);

	/* Print and free node nvlist memory */
	print_us(scripted, parsable, fields, types, cb.cb_width, B_TRUE,
	cb.cb_avl);

	zfs_free_sort_columns(sortcol);

	/* Clean up the AVL tree */
	if ((walk = uu_avl_walk_start(cb.cb_avl, UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((node = uu_avl_walk_next(walk)) != NULL) {
	uu_avl_remove(cb.cb_avl, node);
	free(node);
	}

	uu_avl_walk_end(walk);
	uu_avl_destroy(avl_tree);
	uu_avl_pool_destroy(avl_pool);

	return (ret);
	}

	/*
	* list [-Hp][-r\|-d max] [-o property[,...]] [-s property] ... [-S property]
	* [-t type[,...]] [filesystem\|volume\|snapshot] ...
	*
	* -H Scripted mode; elide headers and separate columns by tabs
	* -p Display values in parsable (literal) format.
	* -r Recurse over all children
	* -d Limit recursion by depth.
	* -o Control which fields to display.
	* -s Specify sort columns, descending order.
	* -S Specify sort columns, ascending order.
	* -t Control which object types to display.
	*
	* When given no arguments, list all filesystems in the system.
	* Otherwise, list the specified datasets, optionally recursing down them if
	* '-r' is specified.
	*/
	typedef struct list_cbdata {
	boolean_t cb_first;
	boolean_t cb_literal;
	boolean_t cb_scripted;
	zprop_list_t *cb_proplist;
	} list_cbdata_t;

	/*
	* Given a list of columns to display, output appropriate headers for each one.
	*/
	static void
	print_header(list_cbdata_t *cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	char headerbuf[ZFS_MAXPROPLEN];
	const char *header;
	int i;
	boolean_t first = B_TRUE;
	boolean_t right_justify;

	for (; pl != NULL; pl = pl->pl_next) {
	if (!first) {
	(void) printf(" ");
	} else {
	first = B_FALSE;
	}

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	header = zfs_prop_column_name(pl->pl_prop);
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else {
	for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
	headerbuf[i] = toupper(pl->pl_user_prop[i]);
	headerbuf[i] = '\0';
	header = headerbuf;
	}

	if (pl->pl_next == NULL && !right_justify)
	(void) printf("%s", header);
	else if (right_justify)
	(void) printf("%*s", (int)pl->pl_width, header);
	else
	(void) printf("%-*s", (int)pl->pl_width, header);
	}

	(void) printf("\n");
	}

	/*
	* Given a dataset and a list of fields, print out all the properties according
	* to the described layout.
	*/
	static void
	print_dataset(zfs_handle_t zhp, list_cbdata_t cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	boolean_t first = B_TRUE;
	char property[ZFS_MAXPROPLEN];
	nvlist_t *userprops = zfs_get_user_props(zhp);
	nvlist_t *propval;
	const char *propstr;
	boolean_t right_justify;

	for (; pl != NULL; pl = pl->pl_next) {
	if (!first) {
	if (cb->cb_scripted)
	(void) putchar('\t');
	else
	(void) fputs(" ", stdout);
	} else {
	first = B_FALSE;
	}

	if (pl->pl_prop == ZFS_PROP_NAME) {
	(void) strlcpy(property, zfs_get_name(zhp),
	sizeof (property));
	propstr = property;
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else if (pl->pl_prop != ZPROP_USERPROP) {
	if (zfs_prop_get(zhp, pl->pl_prop, property,
	sizeof (property), NULL, NULL, 0,
	cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else if (zfs_prop_userquota(pl->pl_user_prop)) {
	if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
	property, sizeof (property), cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = B_TRUE;
	} else if (zfs_prop_written(pl->pl_user_prop)) {
	if (zfs_prop_get_written(zhp, pl->pl_user_prop,
	property, sizeof (property), cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = B_TRUE;
	} else {
	if (nvlist_lookup_nvlist(userprops,
	pl->pl_user_prop, &propval) != 0)
	propstr = "-";
	else
	propstr = fnvlist_lookup_string(propval,
	ZPROP_VALUE);
	right_justify = B_FALSE;
	}

	/*
	* If this is being called in scripted mode, or if this is the
	* last column and it is left-justified, don't include a width
	* format specifier.
	*/
	if (cb->cb_scripted \|\| (pl->pl_next == NULL && !right_justify))
	(void) fputs(propstr, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)pl->pl_width, propstr);
	else
	(void) printf("%-*s", (int)pl->pl_width, propstr);
	}

	(void) putchar('\n');
	}

	/*
	* Generic callback function to list a dataset or snapshot.
	*/
	static int
	list_callback(zfs_handle_t zhp, void data)
	{
	list_cbdata_t *cbp = data;

	if (cbp->cb_first) {
	if (!cbp->cb_scripted)
	print_header(cbp);
	cbp->cb_first = B_FALSE;
	}

	print_dataset(zhp, cbp);

	return (0);
	}

	static int
	zfs_do_list(int argc, char **argv)
	{
	int c;
	char default_fields[] =
	"name,used,available,referenced,mountpoint";
	int types = ZFS_TYPE_DATASET;
	boolean_t types_specified = B_FALSE;
	char *fields = default_fields;
	list_cbdata_t cb = { 0 };
	int limit = 0;
	int ret = 0;
	zfs_sort_column_t *sortcol = NULL;
	int flags = ZFS_ITER_PROP_LISTSNAPS \| ZFS_ITER_ARGS_CAN_BE_PATHS;

	/* check options */
	while ((c = getopt(argc, argv, "HS:d:o:prs:t:")) != -1) {
	switch (c) {
	case 'o':
	fields = optarg;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	flags \|= ZFS_ITER_LITERAL_PROPS;
	break;
	case 'd':
	limit = parse_depth(optarg, &flags);
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 's':
	if (zfs_add_sort_column(&sortcol, optarg,
	B_FALSE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid property '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 'S':
	if (zfs_add_sort_column(&sortcol, optarg,
	B_TRUE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid property '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 't':
	types = 0;
	types_specified = B_TRUE;
	flags &= ~ZFS_ITER_PROP_LISTSNAPS;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const type_subopts[] = {
	"filesystem", "volume",
	"snapshot", "snap",
	"bookmark",
	"all" };
	static const int type_types[] = {
	ZFS_TYPE_FILESYSTEM, ZFS_TYPE_VOLUME,
	ZFS_TYPE_SNAPSHOT, ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_BOOKMARK,
	ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK };

	for (c = 0; c < ARRAY_SIZE(type_subopts); ++c)
	if (strcmp(tok, type_subopts[c]) == 0) {
	types \|= type_types[c];
	goto found3;
	}

	(void) fprintf(stderr,
	gettext("invalid type '%s'\n"), tok);
	usage(B_FALSE);
	found3:;
	}
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/*
	* If we are only going to list snapshot names and sort by name or
	* by createtxg, then we can use faster version.
	*/
	if (strcmp(fields, "name") == 0 &&
	(zfs_sort_only_by_name(sortcol) \|\|
	zfs_sort_only_by_createtxg(sortcol))) {
	flags \|= ZFS_ITER_SIMPLE;
	}

	/*
	* If "-o space" and no types were specified, don't display snapshots.
	*/
	if (strcmp(fields, "space") == 0 && types_specified == B_FALSE)
	types &= ~ZFS_TYPE_SNAPSHOT;

	/*
	* Handle users who want to list all snapshots or bookmarks
	* of the current dataset (ex. 'zfs list -t snapshot <dataset>').
	*/
	if ((types == ZFS_TYPE_SNAPSHOT \|\| types == ZFS_TYPE_BOOKMARK) &&
	argc > 0 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) {
	flags \|= (ZFS_ITER_DEPTH_LIMIT \| ZFS_ITER_RECURSE);
	limit = 1;
	}

	/*
	* If the user specifies '-o all', the zprop_get_list() doesn't
	* normally include the name of the dataset. For 'zfs list', we always
	* want this property to be first.
	*/
	if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET)
	!= 0)
	usage(B_FALSE);

	cb.cb_first = B_TRUE;

	ret = zfs_for_each(argc, argv, flags, types, sortcol, &cb.cb_proplist,
	limit, list_callback, &cb);

	zprop_free_list(cb.cb_proplist);
	zfs_free_sort_columns(sortcol);

	if (ret == 0 && cb.cb_first && !cb.cb_scripted)
	(void) fprintf(stderr, gettext("no datasets available\n"));

	return (ret);
	}

	/*
	* zfs rename [-fu] <fs \| snap \| vol> <fs \| snap \| vol>
	* zfs rename [-f] -p <fs \| vol> <fs \| vol>
	* zfs rename [-u] -r <snap> <snap>
	*
	* Renames the given dataset to another of the same type.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	* The '-u' flag prevents file systems from being remounted during rename.
	*/
	static int
	zfs_do_rename(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	renameflags_t flags = { 0 };
	int c;
	int ret = 0;
	int types;
	boolean_t parents = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "pruf")) != -1) {
	switch (c) {
	case 'p':
	parents = B_TRUE;
	break;
	case 'r':
	flags.recursive = B_TRUE;
	break;
	case 'u':
	flags.nounmount = B_TRUE;
	break;
	case 'f':
	flags.forceunmount = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source dataset "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing target dataset "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (flags.recursive && parents) {
	(void) fprintf(stderr, gettext("-p and -r options are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}

	if (flags.nounmount && parents) {
	(void) fprintf(stderr, gettext("-u and -p options are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}

	if (flags.recursive && strchr(argv[0], '@') == 0) {
	(void) fprintf(stderr, gettext("source dataset for recursive "
	"rename must be a snapshot\n"));
	usage(B_FALSE);
	}

	if (flags.nounmount)
	types = ZFS_TYPE_FILESYSTEM;
	else if (parents)
	types = ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME;
	else
	types = ZFS_TYPE_DATASET;

	if ((zhp = zfs_open(g_zfs, argv[0], types)) == NULL)
	return (1);

	/* If we were asked and the name looks good, try to create ancestors. */
	if (parents && zfs_name_valid(argv[1], zfs_get_type(zhp)) &&
	zfs_create_ancestors(g_zfs, argv[1]) != 0) {
	zfs_close(zhp);
	return (1);
	}

	ret = (zfs_rename(zhp, argv[1], flags) != 0);

	zfs_close(zhp);
	return (ret);
	}

	/*
	* zfs promote <fs>
	*
	* Promotes the given clone fs to be the parent
	*/
	static int
	zfs_do_promote(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	int ret = 0;

	/* check options */
	if (argc > 1 && argv[1][0] == '-') {
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	argv[1][1]);
	usage(B_FALSE);
	}

	/* check number of arguments */
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing clone filesystem"
	" argument\n"));
	usage(B_FALSE);
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return (1);

	ret = (zfs_promote(zhp) != 0);


	zfs_close(zhp);
	return (ret);
	}

	static int
	zfs_do_redact(int argc, char **argv)
	{
	char *snap = NULL;
	char *bookname = NULL;
	char **rsnaps = NULL;
	int numrsnaps = 0;
	argv++;
	argc--;
	if (argc < 3) {
	(void) fprintf(stderr, gettext("too few arguments\n"));
	usage(B_FALSE);
	}

	snap = argv[0];
	bookname = argv[1];
	rsnaps = argv + 2;
	numrsnaps = argc - 2;

	nvlist_t *rsnapnv = fnvlist_alloc();

	for (int i = 0; i < numrsnaps; i++) {
	fnvlist_add_boolean(rsnapnv, rsnaps[i]);
	}

	int err = lzc_redact(snap, bookname, rsnapnv);
	fnvlist_free(rsnapnv);

	switch (err) {
	case 0:
	break;
	case ENOENT:
	(void) fprintf(stderr,
	gettext("provided snapshot %s does not exist\n"), snap);
	break;
	case EEXIST:
	(void) fprintf(stderr, gettext("specified redaction bookmark "
	"(%s) provided already exists\n"), bookname);
	break;
	case ENAMETOOLONG:
	(void) fprintf(stderr, gettext("provided bookmark name cannot "
	"be used, final name would be too long\n"));
	break;
	case E2BIG:
	(void) fprintf(stderr, gettext("too many redaction snapshots "
	"specified\n"));
	break;
	case EINVAL:
	if (strchr(bookname, '#') != NULL)
	(void) fprintf(stderr, gettext(
	"redaction bookmark name must not contain '#'\n"));
	else
	(void) fprintf(stderr, gettext(
	"redaction snapshot must be descendent of "
	"snapshot being redacted\n"));
	break;
	case EALREADY:
	(void) fprintf(stderr, gettext("attempted to redact redacted "
	"dataset or with respect to redacted dataset\n"));
	break;
	case ENOTSUP:
	(void) fprintf(stderr, gettext("redaction bookmarks feature "
	"not enabled\n"));
	break;
	case EXDEV:
	(void) fprintf(stderr, gettext("potentially invalid redaction "
	"snapshot; full dataset names required\n"));
	break;
	default:
	(void) fprintf(stderr, gettext("internal error: %s\n"),
	strerror(errno));
	}

	return (err);
	}

	/*
	* zfs rollback [-rRf] <snapshot>
	*
	* -r Delete any intervening snapshots before doing rollback
	* -R Delete any snapshots and their clones
	* -f ignored for backwards compatibility
	*
	* Given a filesystem, rollback to a specific snapshot, discarding any changes
	* since then and making it the active dataset. If more recent snapshots exist,
	* the command will complain unless the '-r' flag is given.
	*/
	typedef struct rollback_cbdata {
	uint64_t cb_create;
	uint8_t cb_younger_ds_printed;
	boolean_t cb_first;
	int cb_doclones;
	char *cb_target;
	int cb_error;
	boolean_t cb_recurse;
	} rollback_cbdata_t;

	static int
	rollback_check_dependent(zfs_handle_t zhp, void data)
	{
	rollback_cbdata_t *cbp = data;

	if (cbp->cb_first && cbp->cb_recurse) {
	(void) fprintf(stderr, gettext("cannot rollback to "
	"'%s': clones of previous snapshots exist\n"),
	cbp->cb_target);
	(void) fprintf(stderr, gettext("use '-R' to "
	"force deletion of the following clones and "
	"dependents:\n"));
	cbp->cb_first = 0;
	cbp->cb_error = 1;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));

	zfs_close(zhp);
	return (0);
	}


	/*
	* Report some snapshots/bookmarks more recent than the one specified.
	* Used when '-r' is not specified. We reuse this same callback for the
	* snapshot dependents - if 'cb_dependent' is set, then this is a
	* dependent and we should report it without checking the transaction group.
	*/
	static int
	rollback_check(zfs_handle_t zhp, void data)
	{
	rollback_cbdata_t *cbp = data;
	/*
	* Max number of younger snapshots and/or bookmarks to display before
	* we stop the iteration.
	*/
	const uint8_t max_younger = 32;

	if (cbp->cb_doclones) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) {
	if (cbp->cb_first && !cbp->cb_recurse) {
	(void) fprintf(stderr, gettext("cannot "
	"rollback to '%s': more recent snapshots "
	"or bookmarks exist\n"),
	cbp->cb_target);
	(void) fprintf(stderr, gettext("use '-r' to "
	"force deletion of the following "
	"snapshots and bookmarks:\n"));
	cbp->cb_first = 0;
	cbp->cb_error = 1;
	}

	if (cbp->cb_recurse) {
	if (zfs_iter_dependents(zhp, B_TRUE,
	rollback_check_dependent, cbp) != 0) {
	zfs_close(zhp);
	return (-1);
	}
	} else {
	(void) fprintf(stderr, "%s\n",
	zfs_get_name(zhp));
	cbp->cb_younger_ds_printed++;
	}
	}
	zfs_close(zhp);

	if (cbp->cb_younger_ds_printed == max_younger) {
	/*
	* This non-recursive rollback is going to fail due to the
	* presence of snapshots and/or bookmarks that are younger than
	* the rollback target.
	* We printed some of the offending objects, now we stop
	* zfs_iter_snapshot/bookmark iteration so we can fail fast and
	* avoid iterating over the rest of the younger objects
	*/
	(void) fprintf(stderr, gettext("Output limited to %d "
	"snapshots/bookmarks\n"), max_younger);
	return (-1);
	}
	return (0);
	}

	static int
	zfs_do_rollback(int argc, char **argv)
	{
	int ret = 0;
	int c;
	boolean_t force = B_FALSE;
	rollback_cbdata_t cb = { 0 };
	zfs_handle_t zhp, snap;
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	char *delim;
	uint64_t min_txg = 0;

	/* check options */
	while ((c = getopt(argc, argv, "rRf")) != -1) {
	switch (c) {
	case 'r':
	cb.cb_recurse = 1;
	break;
	case 'R':
	cb.cb_recurse = 1;
	cb.cb_doclones = 1;
	break;
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/* open the snapshot */
	if ((snap = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL)
	return (1);

	/* open the parent dataset */
	(void) strlcpy(parentname, argv[0], sizeof (parentname));
	verify((delim = strrchr(parentname, '@')) != NULL);
	*delim = '\0';
	if ((zhp = zfs_open(g_zfs, parentname, ZFS_TYPE_DATASET)) == NULL) {
	zfs_close(snap);
	return (1);
	}

	/*
	* Check for more recent snapshots and/or clones based on the presence
	* of '-r' and '-R'.
	*/
	cb.cb_target = argv[0];
	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
	cb.cb_first = B_TRUE;
	cb.cb_error = 0;

	if (cb.cb_create > 0)
	min_txg = cb.cb_create;

	if ((ret = zfs_iter_snapshots(zhp, B_FALSE, rollback_check, &cb,
	min_txg, 0)) != 0)
	goto out;
	if ((ret = zfs_iter_bookmarks(zhp, rollback_check, &cb)) != 0)
	goto out;

	if ((ret = cb.cb_error) != 0)
	goto out;

	/*
	* Rollback parent to the given snapshot.
	*/
	ret = zfs_rollback(zhp, snap, force);

	out:
	zfs_close(snap);
	zfs_close(zhp);

	if (ret == 0)
	return (0);
	else
	return (1);
	}

	/*
	* zfs set property=value ... { fs \| snap \| vol } ...
	*
	* Sets the given properties for all datasets specified on the command line.
	*/

	static int
	set_callback(zfs_handle_t zhp, void data)
	{
	nvlist_t *props = data;

	if (zfs_prop_set_list(zhp, props) != 0) {
	switch (libzfs_errno(g_zfs)) {
	case EZFS_MOUNTFAILED:
	(void) fprintf(stderr, gettext("property may be set "
	"but unable to remount filesystem\n"));
	break;
	case EZFS_SHARENFSFAILED:
	(void) fprintf(stderr, gettext("property may be set "
	"but unable to reshare filesystem\n"));
	break;
	}
	return (1);
	}
	return (0);
	}

	static int
	zfs_do_set(int argc, char **argv)
	{
	nvlist_t *props = NULL;
	int ds_start = -1; /* argv idx of first dataset arg */
	int ret = 0;
	int i;

	/* check for options */
	if (argc > 1 && argv[1][0] == '-') {
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	argv[1][1]);
	usage(B_FALSE);
	}

	/* check number of arguments */
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing arguments\n"));
	usage(B_FALSE);
	}
	if (argc < 3) {
	if (strchr(argv[1], '=') == NULL) {
	(void) fprintf(stderr, gettext("missing property=value "
	"argument(s)\n"));
	} else {
	(void) fprintf(stderr, gettext("missing dataset "
	"name(s)\n"));
	}
	usage(B_FALSE);
	}

	/* validate argument order: prop=val args followed by dataset args */
	for (i = 1; i < argc; i++) {
	if (strchr(argv[i], '=') != NULL) {
	if (ds_start > 0) {
	/* out-of-order prop=val argument */
	(void) fprintf(stderr, gettext("invalid "
	"argument order\n"));
	usage(B_FALSE);
	}
	} else if (ds_start < 0) {
	ds_start = i;
	}
	}
	if (ds_start < 0) {
	(void) fprintf(stderr, gettext("missing dataset name(s)\n"));
	usage(B_FALSE);
	}

	/* Populate a list of property settings */
	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	for (i = 1; i < ds_start; i++) {
	if (!parseprop(props, argv[i])) {
	ret = -1;
	goto error;
	}
	}

	ret = zfs_for_each(argc - ds_start, argv + ds_start, 0,
	ZFS_TYPE_DATASET, NULL, NULL, 0, set_callback, props);

	error:
	nvlist_free(props);
	return (ret);
	}

	typedef struct snap_cbdata {
	nvlist_t *sd_nvl;
	boolean_t sd_recursive;
	const char *sd_snapname;
	} snap_cbdata_t;

	static int
	zfs_snapshot_cb(zfs_handle_t zhp, void arg)
	{
	snap_cbdata_t *sd = arg;
	char *name;
	int rv = 0;
	int error;

	if (sd->sd_recursive &&
	zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) != 0) {
	zfs_close(zhp);
	return (0);
	}

	error = asprintf(&name, "%s@%s", zfs_get_name(zhp), sd->sd_snapname);
	if (error == -1)
	nomem();
	fnvlist_add_boolean(sd->sd_nvl, name);
	free(name);

	if (sd->sd_recursive)
	rv = zfs_iter_filesystems(zhp, zfs_snapshot_cb, sd);
	zfs_close(zhp);
	return (rv);
	}

	/*
	* zfs snapshot [-r] [-o prop=value] ... <fs@snap>
	*
	* Creates a snapshot with the given name. While functionally equivalent to
	* 'zfs create', it is a separate command to differentiate intent.
	*/
	static int
	zfs_do_snapshot(int argc, char **argv)
	{
	int ret = 0;
	int c;
	nvlist_t *props;
	snap_cbdata_t sd = { 0 };
	boolean_t multiple_snaps = B_FALSE;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	if (nvlist_alloc(&sd.sd_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, "ro:")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	return (1);
	}
	break;
	case 'r':
	sd.sd_recursive = B_TRUE;
	multiple_snaps = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing snapshot argument\n"));
	goto usage;
	}

	if (argc > 1)
	multiple_snaps = B_TRUE;
	for (; argc > 0; argc--, argv++) {
	char *atp;
	zfs_handle_t *zhp;

	atp = strchr(argv[0], '@');
	if (atp == NULL)
	goto usage;
	*atp = '\0';
	sd.sd_snapname = atp + 1;
	zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	goto usage;
	if (zfs_snapshot_cb(zhp, &sd) != 0)
	goto usage;
	}

	ret = zfs_snapshot_nvl(g_zfs, sd.sd_nvl, props);
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	if (ret != 0 && multiple_snaps)
	(void) fprintf(stderr, gettext("no snapshots were created\n"));
	return (ret != 0);

	usage:
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	usage(B_FALSE);
	return (-1);
	}

	/*
	* Array of prefixes to exclude –
	* a linear search, even if executed for each dataset,
	* is plenty good enough.
	*/
	typedef struct zfs_send_exclude_arg {
	size_t count;
	const char **list;
	} zfs_send_exclude_arg_t;

	static boolean_t
	zfs_do_send_exclude(zfs_handle_t zhp, void context)
	{
	zfs_send_exclude_arg_t *excludes = context;
	const char *name = zfs_get_name(zhp);

	for (size_t i = 0; i < excludes->count; ++i) {
	size_t len = strlen(excludes->list[i]);
	if (strncmp(name, excludes->list[i], len) == 0 &&
	memchr("/@", name[len], sizeof ("/@")))
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	/*
	* Send a backup stream to stdout.
	*/
	static int
	zfs_do_send(int argc, char **argv)
	{
	char *fromname = NULL;
	char *toname = NULL;
	char *resume_token = NULL;
	char *cp;
	zfs_handle_t *zhp;
	sendflags_t flags = { 0 };
	int c, err;
	nvlist_t *dbgnv = NULL;
	char *redactbook = NULL;
	zfs_send_exclude_arg_t excludes = { 0 };

	struct option long_options[] = {
	{"replicate", no_argument, NULL, 'R'},
	{"skip-missing", no_argument, NULL, 's'},
	{"redact", required_argument, NULL, 'd'},
	{"props", no_argument, NULL, 'p'},
	{"parsable", no_argument, NULL, 'P'},
	{"dedup", no_argument, NULL, 'D'},
	{"verbose", no_argument, NULL, 'v'},
	{"dryrun", no_argument, NULL, 'n'},
	{"large-block", no_argument, NULL, 'L'},
	{"embed", no_argument, NULL, 'e'},
	{"resume", required_argument, NULL, 't'},
	{"compressed", no_argument, NULL, 'c'},
	{"raw", no_argument, NULL, 'w'},
	{"backup", no_argument, NULL, 'b'},
	{"holds", no_argument, NULL, 'h'},
	{"saved", no_argument, NULL, 'S'},
	{"exclude", required_argument, NULL, 'X'},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, ":i:I:RsDpvnPLeht:cwbd:SX:",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'X':
	for (char *ds; (ds = strsep(&optarg, ",")) != NULL; ) {
	if (!zfs_name_valid(ds, ZFS_TYPE_DATASET) \|\|
	strchr(ds, '/') == NULL) {
	(void) fprintf(stderr, gettext("-X %s: "
	"not a valid non-root dataset name"
	".\n"), ds);
	usage(B_FALSE);
	}
	excludes.list = safe_realloc(excludes.list,
	sizeof (char ) (excludes.count + 1));
	excludes.list[excludes.count++] = ds;
	}
	break;
	case 'i':
	if (fromname)
	usage(B_FALSE);
	fromname = optarg;
	break;
	case 'I':
	if (fromname)
	usage(B_FALSE);
	fromname = optarg;
	flags.doall = B_TRUE;
	break;
	case 'R':
	flags.replicate = B_TRUE;
	break;
	case 's':
	flags.skipmissing = B_TRUE;
	break;
	case 'd':
	redactbook = optarg;
	break;
	case 'p':
	flags.props = B_TRUE;
	break;
	case 'b':
	flags.backup = B_TRUE;
	break;
	case 'h':
	flags.holds = B_TRUE;
	break;
	case 'P':
	flags.parsable = B_TRUE;
	break;
	case 'v':
	flags.verbosity++;
	flags.progress = B_TRUE;
	break;
	case 'D':
	(void) fprintf(stderr,
	gettext("WARNING: deduplicated send is no "
	"longer supported. A regular,\n"
	"non-deduplicated stream will be generated.\n\n"));
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'L':
	flags.largeblock = B_TRUE;
	break;
	case 'e':
	flags.embed_data = B_TRUE;
	break;
	case 't':
	resume_token = optarg;
	break;
	case 'c':
	flags.compress = B_TRUE;
	break;
	case 'w':
	flags.raw = B_TRUE;
	flags.compress = B_TRUE;
	flags.embed_data = B_TRUE;
	flags.largeblock = B_TRUE;
	break;
	case 'S':
	flags.saved = B_TRUE;
	break;
	case ':':
	/*
	* If a parameter was not passed, optopt contains the
	* value that would normally lead us into the
	* appropriate case statement. If it's > 256, then this
	* must be a longopt and we should look at argv to get
	* the string. Otherwise it's just the character, so we
	* should use it directly.
	*/
	if (optopt <= UINT8_MAX) {
	(void) fprintf(stderr,
	gettext("missing argument for '%c' "
	"option\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("missing argument for '%s' "
	"option\n"), argv[optind - 1]);
	}
	usage(B_FALSE);
	break;
	case '?':
	default:
	/*
	* If an invalid flag was passed, optopt contains the
	* character if it was a short flag, or 0 if it was a
	* longopt.
	*/
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);

	}
	usage(B_FALSE);
	}
	}

	if (flags.parsable && flags.verbosity == 0)
	flags.verbosity = 1;

	if (excludes.count > 0 && !flags.replicate) {
	(void) fprintf(stderr, gettext("Cannot specify "
	"dataset exclusion (-X) on a non-recursive "
	"send.\n"));
	return (1);
	}

	argc -= optind;
	argv += optind;

	if (resume_token != NULL) {
	if (fromname != NULL \|\| flags.replicate \|\| flags.props \|\|
	flags.backup \|\| flags.holds \|\|
	flags.saved \|\| redactbook != NULL) {
	(void) fprintf(stderr,
	gettext("invalid flags combined with -t\n"));
	usage(B_FALSE);
	}
	if (argc > 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc < 1) {
	(void) fprintf(stderr,
	gettext("missing snapshot argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	}

	if (flags.saved) {
	if (fromname != NULL \|\| flags.replicate \|\| flags.props \|\|
	flags.doall \|\| flags.backup \|\|
	flags.holds \|\| flags.largeblock \|\| flags.embed_data \|\|
	flags.compress \|\| flags.raw \|\| redactbook != NULL) {
	(void) fprintf(stderr, gettext("incompatible flags "
	"combined with saved send flag\n"));
	usage(B_FALSE);
	}
	if (strchr(argv[0], '@') != NULL) {
	(void) fprintf(stderr, gettext("saved send must "
	"specify the dataset with partially-received "
	"state\n"));
	usage(B_FALSE);
	}
	}

	if (flags.raw && redactbook != NULL) {
	(void) fprintf(stderr,
	gettext("Error: raw sends may not be redacted.\n"));
	return (1);
	}

	if (!flags.dryrun && isatty(STDOUT_FILENO)) {
	(void) fprintf(stderr,
	gettext("Error: Stream can not be written to a terminal.\n"
	"You must redirect standard output.\n"));
	return (1);
	}

	if (flags.saved) {
	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (1);

	err = zfs_send_saved(zhp, &flags, STDOUT_FILENO,
	resume_token);
	zfs_close(zhp);
	return (err != 0);
	} else if (resume_token != NULL) {
	return (zfs_send_resume(g_zfs, &flags, STDOUT_FILENO,
	resume_token));
	}

	if (flags.skipmissing && !flags.replicate) {
	(void) fprintf(stderr,
	gettext("skip-missing flag can only be used in "
	"conjunction with replicate\n"));
	usage(B_FALSE);
	}

	/*
	* For everything except -R and -I, use the new, cleaner code path.
	*/
	if (!(flags.replicate \|\| flags.doall)) {
	char frombuf[ZFS_MAX_DATASET_NAME_LEN];

	if (fromname != NULL && (strchr(fromname, '#') == NULL &&
	strchr(fromname, '@') == NULL)) {
	/*
	* Neither bookmark or snapshot was specified. Print a
	* warning, and assume snapshot.
	*/
	(void) fprintf(stderr, "Warning: incremental source "
	"didn't specify type, assuming snapshot. Use '@' "
	"or '#' prefix to avoid ambiguity.\n");
	(void) snprintf(frombuf, sizeof (frombuf), "@%s",
	fromname);
	fromname = frombuf;
	}
	if (fromname != NULL &&
	(fromname[0] == '#' \|\| fromname[0] == '@')) {
	/*
	* Incremental source name begins with # or @.
	* Default to same fs as target.
	*/
	char tmpbuf[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(tmpbuf, fromname, sizeof (tmpbuf));
	(void) strlcpy(frombuf, argv[0], sizeof (frombuf));
	cp = strchr(frombuf, '@');
	if (cp != NULL)
	*cp = '\0';
	(void) strlcat(frombuf, tmpbuf, sizeof (frombuf));
	fromname = frombuf;
	}

	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (1);
	err = zfs_send_one(zhp, fromname, STDOUT_FILENO, &flags,
	redactbook);
	zfs_close(zhp);
	return (err != 0);
	}

	if (fromname != NULL && strchr(fromname, '#')) {
	(void) fprintf(stderr,
	gettext("Error: multiple snapshots cannot be "
	"sent from a bookmark.\n"));
	return (1);
	}

	if (redactbook != NULL) {
	(void) fprintf(stderr, gettext("Error: multiple snapshots "
	"cannot be sent redacted.\n"));
	return (1);
	}

	if ((cp = strchr(argv[0], '@')) == NULL) {
	(void) fprintf(stderr, gettext("Error: "
	"Unsupported flag with filesystem or bookmark.\n"));
	return (1);
	}
	*cp = '\0';
	toname = cp + 1;
	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return (1);

	/*
	* If they specified the full path to the snapshot, chop off
	* everything except the short name of the snapshot, but special
	* case if they specify the origin.
	*/
	if (fromname && (cp = strchr(fromname, '@')) != NULL) {
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	zprop_source_t src;

	(void) zfs_prop_get(zhp, ZFS_PROP_ORIGIN,
	origin, sizeof (origin), &src, NULL, 0, B_FALSE);

	if (strcmp(origin, fromname) == 0) {
	fromname = NULL;
	flags.fromorigin = B_TRUE;
	} else {
	*cp = '\0';
	if (cp != fromname && strcmp(argv[0], fromname)) {
	(void) fprintf(stderr,
	gettext("incremental source must be "
	"in same filesystem\n"));
	usage(B_FALSE);
	}
	fromname = cp + 1;
	if (strchr(fromname, '@') \|\| strchr(fromname, '/')) {
	(void) fprintf(stderr,
	gettext("invalid incremental source\n"));
	usage(B_FALSE);
	}
	}
	}

	if (flags.replicate && fromname == NULL)
	flags.doall = B_TRUE;

	err = zfs_send(zhp, fromname, toname, &flags, STDOUT_FILENO,
	excludes.count > 0 ? zfs_do_send_exclude : NULL,
	&excludes, flags.verbosity >= 3 ? &dbgnv : NULL);

	if (flags.verbosity >= 3 && dbgnv != NULL) {
	/*
	* dump_nvlist prints to stdout, but that's been
	* redirected to a file. Make it print to stderr
	* instead.
	*/
	(void) dup2(STDERR_FILENO, STDOUT_FILENO);
	dump_nvlist(dbgnv, 0);
	nvlist_free(dbgnv);
	}

	zfs_close(zhp);
	free(excludes.list);
	return (err != 0);
	}

	/*
	* Restore a backup stream from stdin.
	*/
	static int
	zfs_do_receive(int argc, char **argv)
	{
	int c, err = 0;
	recvflags_t flags = { 0 };
	boolean_t abort_resumable = B_FALSE;
	nvlist_t *props;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, ":o:x:dehMnuvFsAc")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'x':
	if (!parsepropname(props, optarg)) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'd':
	if (flags.istail) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -d and -e are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}
	flags.isprefix = B_TRUE;
	break;
	case 'e':
	if (flags.isprefix) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -d and -e are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}
	flags.istail = B_TRUE;
	break;
	case 'h':
	flags.skipholds = B_TRUE;
	break;
	case 'M':
	flags.forceunmount = B_TRUE;
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'u':
	flags.nomount = B_TRUE;
	break;
	case 'v':
	flags.verbose = B_TRUE;
	break;
	case 's':
	flags.resumable = B_TRUE;
	break;
	case 'F':
	flags.force = B_TRUE;
	break;
	case 'A':
	abort_resumable = B_TRUE;
	break;
	case 'c':
	flags.heal = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* zfs recv -e (use "tail" name) implies -d (remove dataset "head") */
	if (flags.istail)
	flags.isprefix = B_TRUE;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing snapshot argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (abort_resumable) {
	if (flags.isprefix \|\| flags.istail \|\| flags.dryrun \|\|
	flags.resumable \|\| flags.nomount) {
	(void) fprintf(stderr, gettext("invalid option\n"));
	usage(B_FALSE);
	}

	char namebuf[ZFS_MAX_DATASET_NAME_LEN];
	(void) snprintf(namebuf, sizeof (namebuf),
	"%s/%%recv", argv[0]);

	if (zfs_dataset_exists(g_zfs, namebuf,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME)) {
	zfs_handle_t *zhp = zfs_open(g_zfs,
	namebuf, ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	nvlist_free(props);
	return (1);
	}
	err = zfs_destroy(zhp, B_FALSE);
	zfs_close(zhp);
	} else {
	zfs_handle_t *zhp = zfs_open(g_zfs,
	argv[0], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	usage(B_FALSE);
	if (!zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) \|\|
	zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	NULL, 0, NULL, NULL, 0, B_TRUE) == -1) {
	(void) fprintf(stderr,
	gettext("'%s' does not have any "
	"resumable receive state to abort\n"),
	argv[0]);
	nvlist_free(props);
	zfs_close(zhp);
	return (1);
	}
	err = zfs_destroy(zhp, B_FALSE);
	zfs_close(zhp);
	}
	nvlist_free(props);
	return (err != 0);
	}

	if (isatty(STDIN_FILENO)) {
	(void) fprintf(stderr,
	gettext("Error: Backup stream can not be read "
	"from a terminal.\n"
	"You must redirect standard input.\n"));
	nvlist_free(props);
	return (1);
	}
	err = zfs_receive(g_zfs, argv[0], props, &flags, STDIN_FILENO, NULL);
	nvlist_free(props);

	return (err != 0);
	}

	/*
	* allow/unallow stuff
	*/
	/* copied from zfs/sys/dsl_deleg.h */
	#define ZFS_DELEG_PERM_CREATE "create"
	#define ZFS_DELEG_PERM_DESTROY "destroy"
	#define ZFS_DELEG_PERM_SNAPSHOT "snapshot"
	#define ZFS_DELEG_PERM_ROLLBACK "rollback"
	#define ZFS_DELEG_PERM_CLONE "clone"
	#define ZFS_DELEG_PERM_PROMOTE "promote"
	#define ZFS_DELEG_PERM_RENAME "rename"
	#define ZFS_DELEG_PERM_MOUNT "mount"
	#define ZFS_DELEG_PERM_SHARE "share"
	#define ZFS_DELEG_PERM_SEND "send"
	#define ZFS_DELEG_PERM_RECEIVE "receive"
	#define ZFS_DELEG_PERM_ALLOW "allow"
	#define ZFS_DELEG_PERM_USERPROP "userprop"
	#define ZFS_DELEG_PERM_VSCAN "vscan" /* ??? */
	#define ZFS_DELEG_PERM_USERQUOTA "userquota"
	#define ZFS_DELEG_PERM_GROUPQUOTA "groupquota"
	#define ZFS_DELEG_PERM_USERUSED "userused"
	#define ZFS_DELEG_PERM_GROUPUSED "groupused"
	#define ZFS_DELEG_PERM_USEROBJQUOTA "userobjquota"
	#define ZFS_DELEG_PERM_GROUPOBJQUOTA "groupobjquota"
	#define ZFS_DELEG_PERM_USEROBJUSED "userobjused"
	#define ZFS_DELEG_PERM_GROUPOBJUSED "groupobjused"

	#define ZFS_DELEG_PERM_HOLD "hold"
	#define ZFS_DELEG_PERM_RELEASE "release"
	#define ZFS_DELEG_PERM_DIFF "diff"
	#define ZFS_DELEG_PERM_BOOKMARK "bookmark"
	#define ZFS_DELEG_PERM_LOAD_KEY "load-key"
	#define ZFS_DELEG_PERM_CHANGE_KEY "change-key"

	#define ZFS_DELEG_PERM_PROJECTUSED "projectused"
	#define ZFS_DELEG_PERM_PROJECTQUOTA "projectquota"
	#define ZFS_DELEG_PERM_PROJECTOBJUSED "projectobjused"
	#define ZFS_DELEG_PERM_PROJECTOBJQUOTA "projectobjquota"

	#define ZFS_NUM_DELEG_NOTES ZFS_DELEG_NOTE_NONE

	static zfs_deleg_perm_tab_t zfs_deleg_perm_tbl[] = {
	{ ZFS_DELEG_PERM_ALLOW, ZFS_DELEG_NOTE_ALLOW },
	{ ZFS_DELEG_PERM_CLONE, ZFS_DELEG_NOTE_CLONE },
	{ ZFS_DELEG_PERM_CREATE, ZFS_DELEG_NOTE_CREATE },
	{ ZFS_DELEG_PERM_DESTROY, ZFS_DELEG_NOTE_DESTROY },
	{ ZFS_DELEG_PERM_DIFF, ZFS_DELEG_NOTE_DIFF},
	{ ZFS_DELEG_PERM_HOLD, ZFS_DELEG_NOTE_HOLD },
	{ ZFS_DELEG_PERM_MOUNT, ZFS_DELEG_NOTE_MOUNT },
	{ ZFS_DELEG_PERM_PROMOTE, ZFS_DELEG_NOTE_PROMOTE },
	{ ZFS_DELEG_PERM_RECEIVE, ZFS_DELEG_NOTE_RECEIVE },
	{ ZFS_DELEG_PERM_RELEASE, ZFS_DELEG_NOTE_RELEASE },
	{ ZFS_DELEG_PERM_RENAME, ZFS_DELEG_NOTE_RENAME },
	{ ZFS_DELEG_PERM_ROLLBACK, ZFS_DELEG_NOTE_ROLLBACK },
	{ ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_SEND },
	{ ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE },
	{ ZFS_DELEG_PERM_SNAPSHOT, ZFS_DELEG_NOTE_SNAPSHOT },
	{ ZFS_DELEG_PERM_BOOKMARK, ZFS_DELEG_NOTE_BOOKMARK },
	{ ZFS_DELEG_PERM_LOAD_KEY, ZFS_DELEG_NOTE_LOAD_KEY },
	{ ZFS_DELEG_PERM_CHANGE_KEY, ZFS_DELEG_NOTE_CHANGE_KEY },

	{ ZFS_DELEG_PERM_GROUPQUOTA, ZFS_DELEG_NOTE_GROUPQUOTA },
	{ ZFS_DELEG_PERM_GROUPUSED, ZFS_DELEG_NOTE_GROUPUSED },
	{ ZFS_DELEG_PERM_USERPROP, ZFS_DELEG_NOTE_USERPROP },
	{ ZFS_DELEG_PERM_USERQUOTA, ZFS_DELEG_NOTE_USERQUOTA },
	{ ZFS_DELEG_PERM_USERUSED, ZFS_DELEG_NOTE_USERUSED },
	{ ZFS_DELEG_PERM_USEROBJQUOTA, ZFS_DELEG_NOTE_USEROBJQUOTA },
	{ ZFS_DELEG_PERM_USEROBJUSED, ZFS_DELEG_NOTE_USEROBJUSED },
	{ ZFS_DELEG_PERM_GROUPOBJQUOTA, ZFS_DELEG_NOTE_GROUPOBJQUOTA },
	{ ZFS_DELEG_PERM_GROUPOBJUSED, ZFS_DELEG_NOTE_GROUPOBJUSED },
	{ ZFS_DELEG_PERM_PROJECTUSED, ZFS_DELEG_NOTE_PROJECTUSED },
	{ ZFS_DELEG_PERM_PROJECTQUOTA, ZFS_DELEG_NOTE_PROJECTQUOTA },
	{ ZFS_DELEG_PERM_PROJECTOBJUSED, ZFS_DELEG_NOTE_PROJECTOBJUSED },
	{ ZFS_DELEG_PERM_PROJECTOBJQUOTA, ZFS_DELEG_NOTE_PROJECTOBJQUOTA },
	{ NULL, ZFS_DELEG_NOTE_NONE }
	};

	/* permission structure */
	typedef struct deleg_perm {
	zfs_deleg_who_type_t dp_who_type;
	const char *dp_name;
	boolean_t dp_local;
	boolean_t dp_descend;
	} deleg_perm_t;

	/* */
	typedef struct deleg_perm_node {
	deleg_perm_t dpn_perm;

	uu_avl_node_t dpn_avl_node;
	} deleg_perm_node_t;

	typedef struct fs_perm fs_perm_t;

	/* permissions set */
	typedef struct who_perm {
	zfs_deleg_who_type_t who_type;
	const char who_name; / id */
	char who_ug_name[256]; /* user/group name */
	fs_perm_t who_fsperm; / uplink */

	uu_avl_t who_deleg_perm_avl; / permissions */
	} who_perm_t;

	/* */
	typedef struct who_perm_node {
	who_perm_t who_perm;
	uu_avl_node_t who_avl_node;
	} who_perm_node_t;

	typedef struct fs_perm_set fs_perm_set_t;
	/* fs permissions */
	struct fs_perm {
	const char *fsp_name;

	uu_avl_t fsp_sc_avl; / sets,create */
	uu_avl_t fsp_uge_avl; / user,group,everyone */

	fs_perm_set_t fsp_set; / uplink */
	};

	/* */
	typedef struct fs_perm_node {
	fs_perm_t fspn_fsperm;
	uu_avl_t *fspn_avl;

	uu_list_node_t fspn_list_node;
	} fs_perm_node_t;

	/* top level structure */
	struct fs_perm_set {
	uu_list_pool_t *fsps_list_pool;
	uu_list_t fsps_list; / list of fs_perms */

	uu_avl_pool_t *fsps_named_set_avl_pool;
	uu_avl_pool_t *fsps_who_perm_avl_pool;
	uu_avl_pool_t *fsps_deleg_perm_avl_pool;
	};

	static inline const char *
	deleg_perm_type(zfs_deleg_note_t note)
	{
	/* subcommands */
	switch (note) {
	/* SUBCOMMANDS */
	/* OTHER */
	case ZFS_DELEG_NOTE_GROUPQUOTA:
	case ZFS_DELEG_NOTE_GROUPUSED:
	case ZFS_DELEG_NOTE_USERPROP:
	case ZFS_DELEG_NOTE_USERQUOTA:
	case ZFS_DELEG_NOTE_USERUSED:
	case ZFS_DELEG_NOTE_USEROBJQUOTA:
	case ZFS_DELEG_NOTE_USEROBJUSED:
	case ZFS_DELEG_NOTE_GROUPOBJQUOTA:
	case ZFS_DELEG_NOTE_GROUPOBJUSED:
	case ZFS_DELEG_NOTE_PROJECTUSED:
	case ZFS_DELEG_NOTE_PROJECTQUOTA:
	case ZFS_DELEG_NOTE_PROJECTOBJUSED:
	case ZFS_DELEG_NOTE_PROJECTOBJQUOTA:
	/* other */
	return (gettext("other"));
	default:
	return (gettext("subcommand"));
	}
	}

	static int
	who_type2weight(zfs_deleg_who_type_t who_type)
	{
	int res;
	switch (who_type) {
	case ZFS_DELEG_NAMED_SET_SETS:
	case ZFS_DELEG_NAMED_SET:
	res = 0;
	break;
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_CREATE:
	res = 1;
	break;
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	res = 2;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	res = 3;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	res = 4;
	break;
	default:
	res = -1;
	}

	return (res);
	}

	static int
	who_perm_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const who_perm_node_t *l = larg;
	const who_perm_node_t *r = rarg;
	zfs_deleg_who_type_t ltype = l->who_perm.who_type;
	zfs_deleg_who_type_t rtype = r->who_perm.who_type;
	int lweight = who_type2weight(ltype);
	int rweight = who_type2weight(rtype);
	int res = lweight - rweight;
	if (res == 0)
	res = strncmp(l->who_perm.who_name, r->who_perm.who_name,
	ZFS_MAX_DELEG_NAME-1);

	if (res == 0)
	return (0);
	if (res > 0)
	return (1);
	else
	return (-1);
	}

	static int
	deleg_perm_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const deleg_perm_node_t *l = larg;
	const deleg_perm_node_t *r = rarg;
	int res = strncmp(l->dpn_perm.dp_name, r->dpn_perm.dp_name,
	ZFS_MAX_DELEG_NAME-1);

	if (res == 0)
	return (0);

	if (res > 0)
	return (1);
	else
	return (-1);
	}

	static inline void
	fs_perm_set_init(fs_perm_set_t *fspset)
	{
	memset(fspset, 0, sizeof (fs_perm_set_t));

	if ((fspset->fsps_list_pool = uu_list_pool_create("fsps_list_pool",
	sizeof (fs_perm_node_t), offsetof(fs_perm_node_t, fspn_list_node),
	NULL, UU_DEFAULT)) == NULL)
	nomem();
	if ((fspset->fsps_list = uu_list_create(fspset->fsps_list_pool, NULL,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_named_set_avl_pool = uu_avl_pool_create(
	"named_set_avl_pool", sizeof (who_perm_node_t), offsetof(
	who_perm_node_t, who_avl_node), who_perm_compare,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_who_perm_avl_pool = uu_avl_pool_create(
	"who_perm_avl_pool", sizeof (who_perm_node_t), offsetof(
	who_perm_node_t, who_avl_node), who_perm_compare,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_deleg_perm_avl_pool = uu_avl_pool_create(
	"deleg_perm_avl_pool", sizeof (deleg_perm_node_t), offsetof(
	deleg_perm_node_t, dpn_avl_node), deleg_perm_compare, UU_DEFAULT))
	== NULL)
	nomem();
	}

	static inline void fs_perm_fini(fs_perm_t *);
	static inline void who_perm_fini(who_perm_t *);

	static inline void
	fs_perm_set_fini(fs_perm_set_t *fspset)
	{
	fs_perm_node_t *node = uu_list_first(fspset->fsps_list);

	while (node != NULL) {
	fs_perm_node_t *next_node =
	uu_list_next(fspset->fsps_list, node);
	fs_perm_t *fsperm = &node->fspn_fsperm;
	fs_perm_fini(fsperm);
	uu_list_remove(fspset->fsps_list, node);
	free(node);
	node = next_node;
	}

	uu_avl_pool_destroy(fspset->fsps_named_set_avl_pool);
	uu_avl_pool_destroy(fspset->fsps_who_perm_avl_pool);
	uu_avl_pool_destroy(fspset->fsps_deleg_perm_avl_pool);
	}

	static inline void
	deleg_perm_init(deleg_perm_t *deleg_perm, zfs_deleg_who_type_t type,
	const char *name)
	{
	deleg_perm->dp_who_type = type;
	deleg_perm->dp_name = name;
	}

	static inline void
	who_perm_init(who_perm_t who_perm, fs_perm_t fsperm,
	zfs_deleg_who_type_t type, const char *name)
	{
	uu_avl_pool_t *pool;
	pool = fsperm->fsp_set->fsps_deleg_perm_avl_pool;

	memset(who_perm, 0, sizeof (who_perm_t));

	if ((who_perm->who_deleg_perm_avl = uu_avl_create(pool, NULL,
	UU_DEFAULT)) == NULL)
	nomem();

	who_perm->who_type = type;
	who_perm->who_name = name;
	who_perm->who_fsperm = fsperm;
	}

	static inline void
	who_perm_fini(who_perm_t *who_perm)
	{
	deleg_perm_node_t *node = uu_avl_first(who_perm->who_deleg_perm_avl);

	while (node != NULL) {
	deleg_perm_node_t *next_node =
	uu_avl_next(who_perm->who_deleg_perm_avl, node);

	uu_avl_remove(who_perm->who_deleg_perm_avl, node);
	free(node);
	node = next_node;
	}

	uu_avl_destroy(who_perm->who_deleg_perm_avl);
	}

	static inline void
	fs_perm_init(fs_perm_t fsperm, fs_perm_set_t fspset, const char *fsname)
	{
	uu_avl_pool_t *nset_pool = fspset->fsps_named_set_avl_pool;
	uu_avl_pool_t *who_pool = fspset->fsps_who_perm_avl_pool;

	memset(fsperm, 0, sizeof (fs_perm_t));

	if ((fsperm->fsp_sc_avl = uu_avl_create(nset_pool, NULL, UU_DEFAULT))
	== NULL)
	nomem();

	if ((fsperm->fsp_uge_avl = uu_avl_create(who_pool, NULL, UU_DEFAULT))
	== NULL)
	nomem();

	fsperm->fsp_set = fspset;
	fsperm->fsp_name = fsname;
	}

	static inline void
	fs_perm_fini(fs_perm_t *fsperm)
	{
	who_perm_node_t *node = uu_avl_first(fsperm->fsp_sc_avl);
	while (node != NULL) {
	who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_sc_avl,
	node);
	who_perm_t *who_perm = &node->who_perm;
	who_perm_fini(who_perm);
	uu_avl_remove(fsperm->fsp_sc_avl, node);
	free(node);
	node = next_node;
	}

	node = uu_avl_first(fsperm->fsp_uge_avl);
	while (node != NULL) {
	who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_uge_avl,
	node);
	who_perm_t *who_perm = &node->who_perm;
	who_perm_fini(who_perm);
	uu_avl_remove(fsperm->fsp_uge_avl, node);
	free(node);
	node = next_node;
	}

	uu_avl_destroy(fsperm->fsp_sc_avl);
	uu_avl_destroy(fsperm->fsp_uge_avl);
	}

	static void
	set_deleg_perm_node(uu_avl_t avl, deleg_perm_node_t node,
	zfs_deleg_who_type_t who_type, const char *name, char locality)
	{
	uu_avl_index_t idx = 0;

	deleg_perm_node_t *found_node = NULL;
	deleg_perm_t *deleg_perm = &node->dpn_perm;

	deleg_perm_init(deleg_perm, who_type, name);

	if ((found_node = uu_avl_find(avl, node, NULL, &idx))
	== NULL)
	uu_avl_insert(avl, node, idx);
	else {
	node = found_node;
	deleg_perm = &node->dpn_perm;
	}


	switch (locality) {
	case ZFS_DELEG_LOCAL:
	deleg_perm->dp_local = B_TRUE;
	break;
	case ZFS_DELEG_DESCENDENT:
	deleg_perm->dp_descend = B_TRUE;
	break;
	case ZFS_DELEG_NA:
	break;
	default:
	assert(B_FALSE); /* invalid locality */
	}
	}

	static inline int
	parse_who_perm(who_perm_t who_perm, nvlist_t nvl, char locality)
	{
	nvpair_t *nvp = NULL;
	fs_perm_set_t *fspset = who_perm->who_fsperm->fsp_set;
	uu_avl_t *avl = who_perm->who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_perm->who_type;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	data_type_t type = nvpair_type(nvp);
	uu_avl_pool_t *avl_pool = fspset->fsps_deleg_perm_avl_pool;
	deleg_perm_node_t *node =
	safe_malloc(sizeof (deleg_perm_node_t));

	VERIFY(type == DATA_TYPE_BOOLEAN);

	uu_avl_node_init(node, &node->dpn_avl_node, avl_pool);
	set_deleg_perm_node(avl, node, who_type, name, locality);
	}

	return (0);
	}

	static inline int
	parse_fs_perm(fs_perm_t fsperm, nvlist_t nvl)
	{
	nvpair_t *nvp = NULL;
	fs_perm_set_t *fspset = fsperm->fsp_set;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	nvlist_t *nvl2 = NULL;
	const char *name = nvpair_name(nvp);
	uu_avl_t *avl = NULL;
	uu_avl_pool_t *avl_pool = NULL;
	zfs_deleg_who_type_t perm_type = name[0];
	char perm_locality = name[1];
	const char *perm_name = name + 3;
	who_perm_t *who_perm = NULL;

	assert('$' == name[2]);

	if (nvpair_value_nvlist(nvp, &nvl2) != 0)
	return (-1);

	switch (perm_type) {
	case ZFS_DELEG_CREATE:
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_NAMED_SET:
	case ZFS_DELEG_NAMED_SET_SETS:
	avl_pool = fspset->fsps_named_set_avl_pool;
	avl = fsperm->fsp_sc_avl;
	break;
	case ZFS_DELEG_USER:
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_GROUP:
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_EVERYONE:
	case ZFS_DELEG_EVERYONE_SETS:
	avl_pool = fspset->fsps_who_perm_avl_pool;
	avl = fsperm->fsp_uge_avl;
	break;

	default:
	assert(!"unhandled zfs_deleg_who_type_t");
	}

	who_perm_node_t *found_node = NULL;
	who_perm_node_t *node = safe_malloc(
	sizeof (who_perm_node_t));
	who_perm = &node->who_perm;
	uu_avl_index_t idx = 0;

	uu_avl_node_init(node, &node->who_avl_node, avl_pool);
	who_perm_init(who_perm, fsperm, perm_type, perm_name);

	if ((found_node = uu_avl_find(avl, node, NULL, &idx))
	== NULL) {
	if (avl == fsperm->fsp_uge_avl) {
	uid_t rid = 0;
	struct passwd *p = NULL;
	struct group *g = NULL;
	const char *nice_name = NULL;

	switch (perm_type) {
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	rid = atoi(perm_name);
	p = getpwuid(rid);
	if (p)
	nice_name = p->pw_name;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	rid = atoi(perm_name);
	g = getgrgid(rid);
	if (g)
	nice_name = g->gr_name;
	break;

	default:
	break;
	}

	if (nice_name != NULL) {
	(void) strlcpy(
	node->who_perm.who_ug_name,
	nice_name, 256);
	} else {
	/* User or group unknown */
	(void) snprintf(
	node->who_perm.who_ug_name,
	sizeof (node->who_perm.who_ug_name),
	"(unknown: %d)", rid);
	}
	}

	uu_avl_insert(avl, node, idx);
	} else {
	node = found_node;
	who_perm = &node->who_perm;
	}

	assert(who_perm != NULL);
	(void) parse_who_perm(who_perm, nvl2, perm_locality);
	}

	return (0);
	}

	static inline int
	parse_fs_perm_set(fs_perm_set_t fspset, nvlist_t nvl)
	{
	nvpair_t *nvp = NULL;
	uu_avl_index_t idx = 0;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	nvlist_t *nvl2 = NULL;
	const char *fsname = nvpair_name(nvp);
	data_type_t type = nvpair_type(nvp);
	fs_perm_t *fsperm = NULL;
	fs_perm_node_t *node = safe_malloc(sizeof (fs_perm_node_t));

	fsperm = &node->fspn_fsperm;

	VERIFY(DATA_TYPE_NVLIST == type);

	uu_list_node_init(node, &node->fspn_list_node,
	fspset->fsps_list_pool);

	idx = uu_list_numnodes(fspset->fsps_list);
	fs_perm_init(fsperm, fspset, fsname);

	if (nvpair_value_nvlist(nvp, &nvl2) != 0)
	return (-1);

	(void) parse_fs_perm(fsperm, nvl2);

	uu_list_insert(fspset->fsps_list, node, idx);
	}

	return (0);
	}

	static inline const char *
	deleg_perm_comment(zfs_deleg_note_t note)
	{
	const char *str = "";

	/* subcommands */
	switch (note) {
	/* SUBCOMMANDS */
	case ZFS_DELEG_NOTE_ALLOW:
	str = gettext("Must also have the permission that is being"
	"\n\t\t\t\tallowed");
	break;
	case ZFS_DELEG_NOTE_CLONE:
	str = gettext("Must also have the 'create' ability and 'mount'"
	"\n\t\t\t\tability in the origin file system");
	break;
	case ZFS_DELEG_NOTE_CREATE:
	str = gettext("Must also have the 'mount' ability");
	break;
	case ZFS_DELEG_NOTE_DESTROY:
	str = gettext("Must also have the 'mount' ability");
	break;
	case ZFS_DELEG_NOTE_DIFF:
	str = gettext("Allows lookup of paths within a dataset;"
	"\n\t\t\t\tgiven an object number. Ordinary users need this"
	"\n\t\t\t\tin order to use zfs diff");
	break;
	case ZFS_DELEG_NOTE_HOLD:
	str = gettext("Allows adding a user hold to a snapshot");
	break;
	case ZFS_DELEG_NOTE_MOUNT:
	str = gettext("Allows mount/umount of ZFS datasets");
	break;
	case ZFS_DELEG_NOTE_PROMOTE:
	str = gettext("Must also have the 'mount'\n\t\t\t\tand"
	" 'promote' ability in the origin file system");
	break;
	case ZFS_DELEG_NOTE_RECEIVE:
	str = gettext("Must also have the 'mount' and 'create'"
	" ability");
	break;
	case ZFS_DELEG_NOTE_RELEASE:
	str = gettext("Allows releasing a user hold which\n\t\t\t\t"
	"might destroy the snapshot");
	break;
	case ZFS_DELEG_NOTE_RENAME:
	str = gettext("Must also have the 'mount' and 'create'"
	"\n\t\t\t\tability in the new parent");
	break;
	case ZFS_DELEG_NOTE_ROLLBACK:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_SEND:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_SHARE:
	str = gettext("Allows sharing file systems over NFS or SMB"
	"\n\t\t\t\tprotocols");
	break;
	case ZFS_DELEG_NOTE_SNAPSHOT:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_LOAD_KEY:
	str = gettext("Allows loading or unloading an encryption key");
	break;
	case ZFS_DELEG_NOTE_CHANGE_KEY:
	str = gettext("Allows changing or adding an encryption key");
	break;
	/*
	* case ZFS_DELEG_NOTE_VSCAN:
	* str = gettext("");
	* break;
	*/
	/* OTHER */
	case ZFS_DELEG_NOTE_GROUPQUOTA:
	str = gettext("Allows accessing any groupquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPUSED:
	str = gettext("Allows reading any groupused@... property");
	break;
	case ZFS_DELEG_NOTE_USERPROP:
	str = gettext("Allows changing any user property");
	break;
	case ZFS_DELEG_NOTE_USERQUOTA:
	str = gettext("Allows accessing any userquota@... property");
	break;
	case ZFS_DELEG_NOTE_USERUSED:
	str = gettext("Allows reading any userused@... property");
	break;
	case ZFS_DELEG_NOTE_USEROBJQUOTA:
	str = gettext("Allows accessing any userobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPOBJQUOTA:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"groupobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPOBJUSED:
	str = gettext("Allows reading any groupobjused@... property");
	break;
	case ZFS_DELEG_NOTE_USEROBJUSED:
	str = gettext("Allows reading any userobjused@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTQUOTA:
	str = gettext("Allows accessing any projectquota@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTOBJQUOTA:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"projectobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTUSED:
	str = gettext("Allows reading any projectused@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTOBJUSED:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"projectobjused@... property");
	break;
	/* other */
	default:
	str = "";
	}

	return (str);
	}

	struct allow_opts {
	boolean_t local;
	boolean_t descend;
	boolean_t user;
	boolean_t group;
	boolean_t everyone;
	boolean_t create;
	boolean_t set;
	boolean_t recursive; /* unallow only */
	boolean_t prt_usage;

	boolean_t prt_perms;
	char *who;
	char *perms;
	const char *dataset;
	};

	static inline int
	prop_cmp(const void a, const void b)
	{
	const char str1 = (const char **)a;
	const char str2 = (const char **)b;
	return (strcmp(str1, str2));
	}

	static void
	allow_usage(boolean_t un, boolean_t requested, const char *msg)
	{
	const char *opt_desc[] = {
	"-h", gettext("show this help message and exit"),
	"-l", gettext("set permission locally"),
	"-d", gettext("set permission for descents"),
	"-u", gettext("set permission for user"),
	"-g", gettext("set permission for group"),
	"-e", gettext("set permission for everyone"),
	"-c", gettext("set create time permission"),
	"-s", gettext("define permission set"),
	/* unallow only */
	"-r", gettext("remove permissions recursively"),
	};
	size_t unallow_size = sizeof (opt_desc) / sizeof (char *);
	size_t allow_size = unallow_size - 2;
	const char *props[ZFS_NUM_PROPS];
	int i;
	size_t count = 0;
	FILE *fp = requested ? stdout : stderr;
	zprop_desc_t *pdtbl = zfs_prop_get_table();
	const char *fmt = gettext("%-16s %-14s\t%s\n");

	(void) fprintf(fp, gettext("Usage: %s\n"), get_usage(un ? HELP_UNALLOW :
	HELP_ALLOW));
	(void) fprintf(fp, gettext("Options:\n"));
	for (i = 0; i < (un ? unallow_size : allow_size); i += 2) {
	const char *opt = opt_desc[i];
	const char *optdsc = opt_desc[i + 1];
	(void) fprintf(fp, gettext(" %-10s %s\n"), opt, optdsc);
	}

	(void) fprintf(fp, gettext("\nThe following permissions are "
	"supported:\n\n"));
	(void) fprintf(fp, fmt, gettext("NAME"), gettext("TYPE"),
	gettext("NOTES"));
	for (i = 0; i < ZFS_NUM_DELEG_NOTES; i++) {
	const char *perm_name = zfs_deleg_perm_tbl[i].z_perm;
	zfs_deleg_note_t perm_note = zfs_deleg_perm_tbl[i].z_note;
	const char *perm_type = deleg_perm_type(perm_note);
	const char *perm_comment = deleg_perm_comment(perm_note);
	(void) fprintf(fp, fmt, perm_name, perm_type, perm_comment);
	}

	for (i = 0; i < ZFS_NUM_PROPS; i++) {
	zprop_desc_t *pd = &pdtbl[i];
	if (pd->pd_visible != B_TRUE)
	continue;

	if (pd->pd_attr == PROP_READONLY)
	continue;

	props[count++] = pd->pd_name;
	}
	props[count] = NULL;

	qsort(props, count, sizeof (char *), prop_cmp);

	for (i = 0; i < count; i++)
	(void) fprintf(fp, fmt, props[i], gettext("property"), "");

	if (msg != NULL)
	(void) fprintf(fp, gettext("\nzfs: error: %s"), msg);

	exit(requested ? 0 : 2);
	}

	static inline const char *
	munge_args(int argc, char **argv, boolean_t un, size_t expected_argc,
	char **permsp)
	{
	if (un && argc == expected_argc - 1)
	*permsp = NULL;
	else if (argc == expected_argc)
	*permsp = argv[argc - 2];
	else
	allow_usage(un, B_FALSE,
	gettext("wrong number of parameters\n"));

	return (argv[argc - 1]);
	}

	static void
	parse_allow_args(int argc, char *argv, boolean_t un, struct allow_opts opts)
	{
	int uge_sum = opts->user + opts->group + opts->everyone;
	int csuge_sum = opts->create + opts->set + uge_sum;
	int ldcsuge_sum = csuge_sum + opts->local + opts->descend;
	int all_sum = un ? ldcsuge_sum + opts->recursive : ldcsuge_sum;

	if (uge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("-u, -g, and -e are mutually exclusive\n"));

	if (opts->prt_usage) {
	if (argc == 0 && all_sum == 0)
	allow_usage(un, B_TRUE, NULL);
	else
	usage(B_FALSE);
	}

	if (opts->set) {
	if (csuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -s\n"));

	opts->dataset = munge_args(argc, argv, un, 3, &opts->perms);
	if (argv[0][0] != '@')
	allow_usage(un, B_FALSE,
	gettext("invalid set name: missing '@' prefix\n"));
	opts->who = argv[0];
	} else if (opts->create) {
	if (ldcsuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -c\n"));
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (opts->everyone) {
	if (csuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -e\n"));
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (uge_sum == 0 && argc > 0 && strcmp(argv[0], "everyone")
	== 0) {
	opts->everyone = B_TRUE;
	argc--;
	argv++;
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (argc == 1 && !un) {
	opts->prt_perms = B_TRUE;
	opts->dataset = argv[argc-1];
	} else {
	opts->dataset = munge_args(argc, argv, un, 3, &opts->perms);
	opts->who = argv[0];
	}

	if (!opts->local && !opts->descend) {
	opts->local = B_TRUE;
	opts->descend = B_TRUE;
	}
	}

	static void
	store_allow_perm(zfs_deleg_who_type_t type, boolean_t local, boolean_t descend,
	const char who, char perms, nvlist_t *top_nvl)
	{
	int i;
	char ld[2] = { '\0', '\0' };
	char who_buf[MAXNAMELEN + 32];
	char base_type = '\0';
	char set_type = '\0';
	nvlist_t *base_nvl = NULL;
	nvlist_t *set_nvl = NULL;
	nvlist_t *nvl;

	if (nvlist_alloc(&base_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	if (nvlist_alloc(&set_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	switch (type) {
	case ZFS_DELEG_NAMED_SET_SETS:
	case ZFS_DELEG_NAMED_SET:
	set_type = ZFS_DELEG_NAMED_SET_SETS;
	base_type = ZFS_DELEG_NAMED_SET;
	ld[0] = ZFS_DELEG_NA;
	break;
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_CREATE:
	set_type = ZFS_DELEG_CREATE_SETS;
	base_type = ZFS_DELEG_CREATE;
	ld[0] = ZFS_DELEG_NA;
	break;
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	set_type = ZFS_DELEG_USER_SETS;
	base_type = ZFS_DELEG_USER;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	set_type = ZFS_DELEG_GROUP_SETS;
	base_type = ZFS_DELEG_GROUP;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	set_type = ZFS_DELEG_EVERYONE_SETS;
	base_type = ZFS_DELEG_EVERYONE;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;

	default:
	assert(set_type != '\0' && base_type != '\0');
	}

	if (perms != NULL) {
	char *curr = perms;
	char *end = curr + strlen(perms);

	while (curr < end) {
	char *delim = strchr(curr, ',');
	if (delim == NULL)
	delim = end;
	else
	*delim = '\0';

	if (curr[0] == '@')
	nvl = set_nvl;
	else
	nvl = base_nvl;

	(void) nvlist_add_boolean(nvl, curr);
	if (delim != end)
	*delim = ',';
	curr = delim + 1;
	}

	for (i = 0; i < 2; i++) {
	char locality = ld[i];
	if (locality == 0)
	continue;

	if (!nvlist_empty(base_nvl)) {
	if (who != NULL)
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$%s",
	base_type, locality, who);
	else
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$",
	base_type, locality);

	(void) nvlist_add_nvlist(top_nvl, who_buf,
	base_nvl);
	}


	if (!nvlist_empty(set_nvl)) {
	if (who != NULL)
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$%s",
	set_type, locality, who);
	else
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$",
	set_type, locality);

	(void) nvlist_add_nvlist(top_nvl, who_buf,
	set_nvl);
	}
	}
	} else {
	for (i = 0; i < 2; i++) {
	char locality = ld[i];
	if (locality == 0)
	continue;

	if (who != NULL)
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$%s", base_type, locality, who);
	else
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$", base_type, locality);
	(void) nvlist_add_boolean(top_nvl, who_buf);

	if (who != NULL)
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$%s", set_type, locality, who);
	else
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$", set_type, locality);
	(void) nvlist_add_boolean(top_nvl, who_buf);
	}
	}
	}

	static int
	construct_fsacl_list(boolean_t un, struct allow_opts opts, nvlist_t *nvlp)
	{
	if (nvlist_alloc(nvlp, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	if (opts->set) {
	store_allow_perm(ZFS_DELEG_NAMED_SET, opts->local,
	opts->descend, opts->who, opts->perms, *nvlp);
	} else if (opts->create) {
	store_allow_perm(ZFS_DELEG_CREATE, opts->local,
	opts->descend, NULL, opts->perms, *nvlp);
	} else if (opts->everyone) {
	store_allow_perm(ZFS_DELEG_EVERYONE, opts->local,
	opts->descend, NULL, opts->perms, *nvlp);
	} else {
	char *curr = opts->who;
	char *end = curr + strlen(curr);

	while (curr < end) {
	const char *who;
	zfs_deleg_who_type_t who_type = ZFS_DELEG_WHO_UNKNOWN;
	char *endch;
	char *delim = strchr(curr, ',');
	char errbuf[256];
	char id[64];
	struct passwd *p = NULL;
	struct group *g = NULL;

	uid_t rid;
	if (delim == NULL)
	delim = end;
	else
	*delim = '\0';

	rid = (uid_t)strtol(curr, &endch, 0);
	if (opts->user) {
	who_type = ZFS_DELEG_USER;
	if (*endch != '\0')
	p = getpwnam(curr);
	else
	p = getpwuid(rid);

	if (p != NULL)
	rid = p->pw_uid;
	else if (*endch != '\0') {
	(void) snprintf(errbuf, 256, gettext(
	"invalid user %s\n"), curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	} else if (opts->group) {
	who_type = ZFS_DELEG_GROUP;
	if (*endch != '\0')
	g = getgrnam(curr);
	else
	g = getgrgid(rid);

	if (g != NULL)
	rid = g->gr_gid;
	else if (*endch != '\0') {
	(void) snprintf(errbuf, 256, gettext(
	"invalid group %s\n"), curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	} else {
	if (*endch != '\0') {
	p = getpwnam(curr);
	} else {
	p = getpwuid(rid);
	}

	if (p == NULL) {
	if (*endch != '\0') {
	g = getgrnam(curr);
	} else {
	g = getgrgid(rid);
	}
	}

	if (p != NULL) {
	who_type = ZFS_DELEG_USER;
	rid = p->pw_uid;
	} else if (g != NULL) {
	who_type = ZFS_DELEG_GROUP;
	rid = g->gr_gid;
	} else {
	(void) snprintf(errbuf, 256, gettext(
	"invalid user/group %s\n"), curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	}

	(void) sprintf(id, "%u", rid);
	who = id;

	store_allow_perm(who_type, opts->local,
	opts->descend, who, opts->perms, *nvlp);
	curr = delim + 1;
	}
	}

	return (0);
	}

	static void
	print_set_creat_perms(uu_avl_t *who_avl)
	{
	const char *sc_title[] = {
	gettext("Permission sets:\n"),
	gettext("Create time permissions:\n"),
	NULL
	};
	who_perm_node_t *who_node = NULL;
	int prev_weight = -1;

	for (who_node = uu_avl_first(who_avl); who_node != NULL;
	who_node = uu_avl_next(who_avl, who_node)) {
	uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_node->who_perm.who_type;
	const char *who_name = who_node->who_perm.who_name;
	int weight = who_type2weight(who_type);
	boolean_t first = B_TRUE;
	deleg_perm_node_t *deleg_node;

	if (prev_weight != weight) {
	(void) printf("%s", sc_title[weight]);
	prev_weight = weight;
	}

	if (who_name == NULL \|\| strnlen(who_name, 1) == 0)
	(void) printf("\t");
	else
	(void) printf("\t%s ", who_name);

	for (deleg_node = uu_avl_first(avl); deleg_node != NULL;
	deleg_node = uu_avl_next(avl, deleg_node)) {
	if (first) {
	(void) printf("%s",
	deleg_node->dpn_perm.dp_name);
	first = B_FALSE;
	} else
	(void) printf(",%s",
	deleg_node->dpn_perm.dp_name);
	}

	(void) printf("\n");
	}
	}

	static void
	print_uge_deleg_perms(uu_avl_t *who_avl, boolean_t local, boolean_t descend,
	const char *title)
	{
	who_perm_node_t *who_node = NULL;
	boolean_t prt_title = B_TRUE;
	uu_avl_walk_t *walk;

	if ((walk = uu_avl_walk_start(who_avl, UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((who_node = uu_avl_walk_next(walk)) != NULL) {
	const char *who_name = who_node->who_perm.who_name;
	const char *nice_who_name = who_node->who_perm.who_ug_name;
	uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_node->who_perm.who_type;
	char delim = ' ';
	deleg_perm_node_t *deleg_node;
	boolean_t prt_who = B_TRUE;

	for (deleg_node = uu_avl_first(avl);
	deleg_node != NULL;
	deleg_node = uu_avl_next(avl, deleg_node)) {
	if (local != deleg_node->dpn_perm.dp_local \|\|
	descend != deleg_node->dpn_perm.dp_descend)
	continue;

	if (prt_who) {
	const char *who = NULL;
	if (prt_title) {
	prt_title = B_FALSE;
	(void) printf("%s", title);
	}

	switch (who_type) {
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	who = gettext("user");
	if (nice_who_name)
	who_name = nice_who_name;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	who = gettext("group");
	if (nice_who_name)
	who_name = nice_who_name;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	who = gettext("everyone");
	who_name = NULL;
	break;

	default:
	assert(who != NULL);
	}

	prt_who = B_FALSE;
	if (who_name == NULL)
	(void) printf("\t%s", who);
	else
	(void) printf("\t%s %s", who, who_name);
	}

	(void) printf("%c%s", delim,
	deleg_node->dpn_perm.dp_name);
	delim = ',';
	}

	if (!prt_who)
	(void) printf("\n");
	}

	uu_avl_walk_end(walk);
	}

	static void
	print_fs_perms(fs_perm_set_t *fspset)
	{
	fs_perm_node_t *node = NULL;
	char buf[MAXNAMELEN + 32];
	const char *dsname = buf;

	for (node = uu_list_first(fspset->fsps_list); node != NULL;
	node = uu_list_next(fspset->fsps_list, node)) {
	uu_avl_t *sc_avl = node->fspn_fsperm.fsp_sc_avl;
	uu_avl_t *uge_avl = node->fspn_fsperm.fsp_uge_avl;
	int left = 0;

	(void) snprintf(buf, sizeof (buf),
	gettext("---- Permissions on %s "),
	node->fspn_fsperm.fsp_name);
	(void) printf("%s", dsname);
	left = 70 - strlen(buf);
	while (left-- > 0)
	(void) printf("-");
	(void) printf("\n");

	print_set_creat_perms(sc_avl);
	print_uge_deleg_perms(uge_avl, B_TRUE, B_FALSE,
	gettext("Local permissions:\n"));
	print_uge_deleg_perms(uge_avl, B_FALSE, B_TRUE,
	gettext("Descendent permissions:\n"));
	print_uge_deleg_perms(uge_avl, B_TRUE, B_TRUE,
	gettext("Local+Descendent permissions:\n"));
	}
	}

	static fs_perm_set_t fs_perm_set = { NULL, NULL, NULL, NULL };

	struct deleg_perms {
	boolean_t un;
	nvlist_t *nvl;
	};

	static int
	set_deleg_perms(zfs_handle_t zhp, void data)
	{
	struct deleg_perms perms = (struct deleg_perms )data;
	zfs_type_t zfs_type = zfs_get_type(zhp);

	if (zfs_type != ZFS_TYPE_FILESYSTEM && zfs_type != ZFS_TYPE_VOLUME)
	return (0);

	return (zfs_set_fsacl(zhp, perms->un, perms->nvl));
	}

	static int
	zfs_do_allow_unallow_impl(int argc, char **argv, boolean_t un)
	{
	zfs_handle_t *zhp;
	nvlist_t *perm_nvl = NULL;
	nvlist_t *update_perm_nvl = NULL;
	int error = 1;
	int c;
	struct allow_opts opts = { 0 };

	const char *optstr = un ? "ldugecsrh" : "ldugecsh";

	/* check opts */
	while ((c = getopt(argc, argv, optstr)) != -1) {
	switch (c) {
	case 'l':
	opts.local = B_TRUE;
	break;
	case 'd':
	opts.descend = B_TRUE;
	break;
	case 'u':
	opts.user = B_TRUE;
	break;
	case 'g':
	opts.group = B_TRUE;
	break;
	case 'e':
	opts.everyone = B_TRUE;
	break;
	case 's':
	opts.set = B_TRUE;
	break;
	case 'c':
	opts.create = B_TRUE;
	break;
	case 'r':
	opts.recursive = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case 'h':
	opts.prt_usage = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check arguments */
	parse_allow_args(argc, argv, un, &opts);

	/* try to open the dataset */
	if ((zhp = zfs_open(g_zfs, opts.dataset, ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) == NULL) {
	(void) fprintf(stderr, "Failed to open dataset: %s\n",
	opts.dataset);
	return (-1);
	}

	if (zfs_get_fsacl(zhp, &perm_nvl) != 0)
	goto cleanup2;

	fs_perm_set_init(&fs_perm_set);
	if (parse_fs_perm_set(&fs_perm_set, perm_nvl) != 0) {
	(void) fprintf(stderr, "Failed to parse fsacl permissions\n");
	goto cleanup1;
	}

	if (opts.prt_perms)
	print_fs_perms(&fs_perm_set);
	else {
	(void) construct_fsacl_list(un, &opts, &update_perm_nvl);
	if (zfs_set_fsacl(zhp, un, update_perm_nvl) != 0)
	goto cleanup0;

	if (un && opts.recursive) {
	struct deleg_perms data = { un, update_perm_nvl };
	if (zfs_iter_filesystems(zhp, set_deleg_perms,
	&data) != 0)
	goto cleanup0;
	}
	}

	error = 0;

	cleanup0:
	nvlist_free(perm_nvl);
	nvlist_free(update_perm_nvl);
	cleanup1:
	fs_perm_set_fini(&fs_perm_set);
	cleanup2:
	zfs_close(zhp);

	return (error);
	}

	static int
	zfs_do_allow(int argc, char **argv)
	{
	return (zfs_do_allow_unallow_impl(argc, argv, B_FALSE));
	}

	static int
	zfs_do_unallow(int argc, char **argv)
	{
	return (zfs_do_allow_unallow_impl(argc, argv, B_TRUE));
	}

	static int
	zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
	{
	int errors = 0;
	int i;
	const char *tag;
	boolean_t recursive = B_FALSE;
	const char *opts = holding ? "rt" : "r";
	int c;

	/* check options */
	while ((c = getopt(argc, argv, opts)) != -1) {
	switch (c) {
	case 'r':
	recursive = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 2)
	usage(B_FALSE);

	tag = argv[0];
	--argc;
	++argv;

	if (holding && tag[0] == '.') {
	/* tags starting with '.' are reserved for libzfs */
	(void) fprintf(stderr, gettext("tag may not start with '.'\n"));
	usage(B_FALSE);
	}

	for (i = 0; i < argc; ++i) {
	zfs_handle_t *zhp;
	char parent[ZFS_MAX_DATASET_NAME_LEN];
	const char *delim;
	char *path = argv[i];

	delim = strchr(path, '@');
	if (delim == NULL) {
	(void) fprintf(stderr,
	gettext("'%s' is not a snapshot\n"), path);
	++errors;
	continue;
	}
	- (void) strncpy(parent, path, delim - path);
	- parent[delim - path] = '\0';
	+ (void) strlcpy(parent, path, MIN(sizeof (parent),
	+ delim - path + 1));

	zhp = zfs_open(g_zfs, parent,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	++errors;
	continue;
	}
	if (holding) {
	if (zfs_hold(zhp, delim+1, tag, recursive, -1) != 0)
	++errors;
	} else {
	if (zfs_release(zhp, delim+1, tag, recursive) != 0)
	++errors;
	}
	zfs_close(zhp);
	}

	return (errors != 0);
	}

	/*
	* zfs hold [-r] [-t] <tag> <snap> ...
	*
	* -r Recursively hold
	*
	* Apply a user-hold with the given tag to the list of snapshots.
	*/
	static int
	zfs_do_hold(int argc, char **argv)
	{
	return (zfs_do_hold_rele_impl(argc, argv, B_TRUE));
	}

	/*
	* zfs release [-r] <tag> <snap> ...
	*
	* -r Recursively release
	*
	* Release a user-hold with the given tag from the list of snapshots.
	*/
	static int
	zfs_do_release(int argc, char **argv)
	{
	return (zfs_do_hold_rele_impl(argc, argv, B_FALSE));
	}

	typedef struct holds_cbdata {
	boolean_t cb_recursive;
	const char *cb_snapname;
	nvlist_t **cb_nvlp;
	size_t cb_max_namelen;
	size_t cb_max_taglen;
	} holds_cbdata_t;

	#define STRFTIME_FMT_STR "%a %b %e %H:%M %Y"
	#define DATETIME_BUF_LEN (32)
	/*
	*
	*/
	static void
	print_holds(boolean_t scripted, int nwidth, int tagwidth, nvlist_t *nvl)
	{
	int i;
	nvpair_t *nvp = NULL;
	const char *const hdr_cols[] = { "NAME", "TAG", "TIMESTAMP" };
	const char *col;

	if (!scripted) {
	for (i = 0; i < 3; i++) {
	col = gettext(hdr_cols[i]);
	if (i < 2)
	(void) printf("%-*s ", i ? tagwidth : nwidth,
	col);
	else
	(void) printf("%s\n", col);
	}
	}

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	char *zname = nvpair_name(nvp);
	nvlist_t *nvl2;
	nvpair_t *nvp2 = NULL;
	(void) nvpair_value_nvlist(nvp, &nvl2);
	while ((nvp2 = nvlist_next_nvpair(nvl2, nvp2)) != NULL) {
	char tsbuf[DATETIME_BUF_LEN];
	const char *tagname = nvpair_name(nvp2);
	uint64_t val = 0;
	time_t time;
	struct tm t;

	(void) nvpair_value_uint64(nvp2, &val);
	time = (time_t)val;
	(void) localtime_r(&time, &t);
	(void) strftime(tsbuf, DATETIME_BUF_LEN,
	gettext(STRFTIME_FMT_STR), &t);

	if (scripted) {
	(void) printf("%s\t%s\t%s\n", zname,
	tagname, tsbuf);
	} else {
	(void) printf("%-s %-s %s\n", nwidth,
	zname, tagwidth, tagname, tsbuf);
	}
	}
	}
	}

	/*
	* Generic callback function to list a dataset or snapshot.
	*/
	static int
	holds_callback(zfs_handle_t zhp, void data)
	{
	holds_cbdata_t *cbp = data;
	nvlist_t top_nvl = cbp->cb_nvlp;
	nvlist_t *nvl = NULL;
	nvpair_t *nvp = NULL;
	const char *zname = zfs_get_name(zhp);
	size_t znamelen = strlen(zname);

	if (cbp->cb_recursive) {
	const char *snapname;
	char *delim = strchr(zname, '@');
	if (delim == NULL)
	return (0);

	snapname = delim + 1;
	if (strcmp(cbp->cb_snapname, snapname))
	return (0);
	}

	if (zfs_get_holds(zhp, &nvl) != 0)
	return (-1);

	if (znamelen > cbp->cb_max_namelen)
	cbp->cb_max_namelen = znamelen;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	const char *tag = nvpair_name(nvp);
	size_t taglen = strlen(tag);
	if (taglen > cbp->cb_max_taglen)
	cbp->cb_max_taglen = taglen;
	}

	return (nvlist_add_nvlist(top_nvl, zname, nvl));
	}

	/*
	* zfs holds [-rH] <snap> ...
	*
	* -r Lists holds that are set on the named snapshots recursively.
	* -H Scripted mode; elide headers and separate columns by tabs.
	*/
	static int
	zfs_do_holds(int argc, char **argv)
	{
	int c;
	boolean_t errors = B_FALSE;
	boolean_t scripted = B_FALSE;
	boolean_t recursive = B_FALSE;

	int types = ZFS_TYPE_SNAPSHOT;
	holds_cbdata_t cb = { 0 };

	int limit = 0;
	int ret = 0;
	int flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, "rH")) != -1) {
	switch (c) {
	case 'r':
	recursive = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (recursive) {
	types \|= ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME;
	flags \|= ZFS_ITER_RECURSE;
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1)
	usage(B_FALSE);

	nvlist_t *nvl = fnvlist_alloc();

	for (int i = 0; i < argc; ++i) {
	char *snapshot = argv[i];
	const char *delim;
	const char *snapname;

	delim = strchr(snapshot, '@');
	if (delim == NULL) {
	(void) fprintf(stderr,
	gettext("'%s' is not a snapshot\n"), snapshot);
	errors = B_TRUE;
	continue;
	}
	snapname = delim + 1;
	if (recursive)
	snapshot[delim - snapshot] = '\0';

	cb.cb_recursive = recursive;
	cb.cb_snapname = snapname;
	cb.cb_nvlp = &nvl;

	/*
	* 1. collect holds data, set format options
	*/
	ret = zfs_for_each(1, argv + i, flags, types, NULL, NULL, limit,
	holds_callback, &cb);
	if (ret != 0)
	errors = B_TRUE;
	}

	/*
	* 2. print holds data
	*/
	print_holds(scripted, cb.cb_max_namelen, cb.cb_max_taglen, nvl);

	if (nvlist_empty(nvl))
	(void) fprintf(stderr, gettext("no datasets available\n"));

	nvlist_free(nvl);

	return (errors);
	}

	#define CHECK_SPINNER 30
	#define SPINNER_TIME 3 /* seconds */
	#define MOUNT_TIME 1 /* seconds */

	typedef struct get_all_state {
	boolean_t ga_verbose;
	get_all_cb_t *ga_cbp;
	} get_all_state_t;

	static int
	get_one_dataset(zfs_handle_t zhp, void data)
	{
	static const char *const spin[] = { "-", "\\", "\|", "/" };
	static int spinval = 0;
	static int spincheck = 0;
	static time_t last_spin_time = (time_t)0;
	get_all_state_t *state = data;
	zfs_type_t type = zfs_get_type(zhp);

	if (state->ga_verbose) {
	if (--spincheck < 0) {
	time_t now = time(NULL);
	if (last_spin_time + SPINNER_TIME < now) {
	update_progress(spin[spinval++ % 4]);
	last_spin_time = now;
	}
	spincheck = CHECK_SPINNER;
	}
	}

	/*
	* Iterate over any nested datasets.
	*/
	if (zfs_iter_filesystems(zhp, get_one_dataset, data) != 0) {
	zfs_close(zhp);
	return (1);
	}

	/*
	* Skip any datasets whose type does not match.
	*/
	if ((type & ZFS_TYPE_FILESYSTEM) == 0) {
	zfs_close(zhp);
	return (0);
	}
	libzfs_add_handle(state->ga_cbp, zhp);
	assert(state->ga_cbp->cb_used <= state->ga_cbp->cb_alloc);

	return (0);
	}

	static void
	get_all_datasets(get_all_cb_t *cbp, boolean_t verbose)
	{
	get_all_state_t state = {
	.ga_verbose = verbose,
	.ga_cbp = cbp
	};

	if (verbose)
	set_progress_header(gettext("Reading ZFS config"));
	(void) zfs_iter_root(g_zfs, get_one_dataset, &state);

	if (verbose)
	finish_progress(gettext("done."));
	}

	/*
	* Generic callback for sharing or mounting filesystems. Because the code is so
	* similar, we have a common function with an extra parameter to determine which
	* mode we are using.
	*/
	typedef enum { OP_SHARE, OP_MOUNT } share_mount_op_t;

	typedef struct share_mount_state {
	share_mount_op_t sm_op;
	boolean_t sm_verbose;
	int sm_flags;
	char *sm_options;
	enum sa_protocol sm_proto; /* only valid for OP_SHARE */
	pthread_mutex_t sm_lock; /* protects the remaining fields */
	uint_t sm_total; /* number of filesystems to process */
	uint_t sm_done; /* number of filesystems processed */
	int sm_status; /* -1 if any of the share/mount operations failed */
	} share_mount_state_t;

	/*
	* Share or mount a dataset.
	*/
	static int
	share_mount_one(zfs_handle_t *zhp, int op, int flags, enum sa_protocol protocol,
	boolean_t explicit, const char *options)
	{
	char mountpoint[ZFS_MAXPROPLEN];
	char shareopts[ZFS_MAXPROPLEN];
	char smbshareopts[ZFS_MAXPROPLEN];
	const char *cmdname = op == OP_SHARE ? "share" : "mount";
	struct mnttab mnt;
	uint64_t zoned, canmount;
	boolean_t shared_nfs, shared_smb;

	assert(zfs_get_type(zhp) & ZFS_TYPE_FILESYSTEM);

	/*
	* Check to make sure we can mount/share this dataset. If we
	* are in the global zone and the filesystem is exported to a
	* local zone, or if we are in a local zone and the
	* filesystem is not exported, then it is an error.
	*/
	zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);

	if (zoned && getzoneid() == GLOBAL_ZONEID) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"dataset is exported to a local zone\n"), cmdname,
	zfs_get_name(zhp));
	return (1);

	} else if (!zoned && getzoneid() != GLOBAL_ZONEID) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"permission denied\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	}

	/*
	* Ignore any filesystems which don't apply to us. This
	* includes those with a legacy mountpoint, or those with
	* legacy share options.
	*/
	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mountpoint,
	sizeof (mountpoint), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, shareopts,
	sizeof (shareopts), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshareopts,
	sizeof (smbshareopts), NULL, NULL, 0, B_FALSE) == 0);

	if (op == OP_SHARE && strcmp(shareopts, "off") == 0 &&
	strcmp(smbshareopts, "off") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot share '%s': "
	"legacy share\n"), zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use exports(5) or "
	"smb.conf(5) to share this filesystem, or set "
	"the sharenfs or sharesmb property\n"));
	return (1);
	}

	/*
	* We cannot share or mount legacy filesystems. If the
	* shareopts is non-legacy but the mountpoint is legacy, we
	* treat it as a legacy share.
	*/
	if (strcmp(mountpoint, "legacy") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"legacy mountpoint\n"), cmdname, zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use %s(8) to "
	"%s this filesystem\n"), cmdname, cmdname);
	return (1);
	}

	if (strcmp(mountpoint, "none") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': no "
	"mountpoint set\n"), cmdname, zfs_get_name(zhp));
	return (1);
	}

	/*
	* canmount explicit outcome
	* on no pass through
	* on yes pass through
	* off no return 0
	* off yes display error, return 1
	* noauto no return 0
	* noauto yes pass through
	*/
	canmount = zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT);
	if (canmount == ZFS_CANMOUNT_OFF) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"'canmount' property is set to 'off'\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	} else if (canmount == ZFS_CANMOUNT_NOAUTO && !explicit) {
	/*
	* When performing a 'zfs mount -a', we skip any mounts for
	* datasets that have 'noauto' set. Sharing a dataset with
	* 'noauto' set is only allowed if it's mounted.
	*/
	if (op == OP_MOUNT)
	return (0);
	if (op == OP_SHARE && !zfs_is_mounted(zhp, NULL)) {
	/* also purge it from existing exports */
	zfs_unshare(zhp, mountpoint, NULL);
	return (0);
	}
	}

	/*
	* If this filesystem is encrypted and does not have
	* a loaded key, we can not mount it.
	*/
	if ((flags & MS_CRYPT) == 0 &&
	zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF &&
	zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) ==
	ZFS_KEYSTATUS_UNAVAILABLE) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"encryption key not loaded\n"), cmdname, zfs_get_name(zhp));
	return (1);
	}

	/*
	* If this filesystem is inconsistent and has a receive resume
	* token, we can not mount it.
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) &&
	zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	NULL, 0, NULL, NULL, 0, B_TRUE) == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"Contains partially-completed state from "
	"\"zfs receive -s\", which can be resumed with "
	"\"zfs send -t\"\n"),
	cmdname, zfs_get_name(zhp));
	return (1);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_REDACTED) && !(flags & MS_FORCE)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"Dataset is not complete, was created by receiving "
	"a redacted zfs send stream.\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	}

	/*
	* At this point, we have verified that the mountpoint and/or
	* shareopts are appropriate for auto management. If the
	* filesystem is already mounted or shared, return (failing
	* for explicit requests); otherwise mount or share the
	* filesystem.
	*/
	switch (op) {
	case OP_SHARE: {
	enum sa_protocol prot[] = {SA_PROTOCOL_NFS, SA_NO_PROTOCOL};
	shared_nfs = zfs_is_shared(zhp, NULL, prot);
	*prot = SA_PROTOCOL_SMB;
	shared_smb = zfs_is_shared(zhp, NULL, prot);

	if ((shared_nfs && shared_smb) \|\|
	(shared_nfs && strcmp(shareopts, "on") == 0 &&
	strcmp(smbshareopts, "off") == 0) \|\|
	(shared_smb && strcmp(smbshareopts, "on") == 0 &&
	strcmp(shareopts, "off") == 0)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot share "
	"'%s': filesystem already shared\n"),
	zfs_get_name(zhp));
	return (1);
	}

	if (!zfs_is_mounted(zhp, NULL) &&
	zfs_mount(zhp, NULL, flags) != 0)
	return (1);

	*prot = protocol;
	if (zfs_share(zhp, protocol == SA_NO_PROTOCOL ? NULL : prot))
	return (1);

	}
	break;

	case OP_MOUNT:
	mnt.mnt_mntopts = (char *)(options ?: "");

	if (!hasmntopt(&mnt, MNTOPT_REMOUNT) &&
	zfs_is_mounted(zhp, NULL)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot mount "
	"'%s': filesystem already mounted\n"),
	zfs_get_name(zhp));
	return (1);
	}

	if (zfs_mount(zhp, options, flags) != 0)
	return (1);
	break;
	}

	return (0);
	}

	/*
	* Reports progress in the form "(current/total)". Not thread-safe.
	*/
	static void
	report_mount_progress(int current, int total)
	{
	static time_t last_progress_time = 0;
	time_t now = time(NULL);
	char info[32];

	/* display header if we're here for the first time */
	if (current == 1) {
	set_progress_header(gettext("Mounting ZFS filesystems"));
	} else if (current != total && last_progress_time + MOUNT_TIME >= now) {
	/* too soon to report again */
	return;
	}

	last_progress_time = now;

	(void) sprintf(info, "(%d/%d)", current, total);

	if (current == total)
	finish_progress(info);
	else
	update_progress(info);
	}

	/*
	* zfs_foreach_mountpoint() callback that mounts or shares one filesystem and
	* updates the progress meter.
	*/
	static int
	share_mount_one_cb(zfs_handle_t zhp, void arg)
	{
	share_mount_state_t *sms = arg;
	int ret;

	ret = share_mount_one(zhp, sms->sm_op, sms->sm_flags, sms->sm_proto,
	B_FALSE, sms->sm_options);

	pthread_mutex_lock(&sms->sm_lock);
	if (ret != 0)
	sms->sm_status = ret;
	sms->sm_done++;
	if (sms->sm_verbose)
	report_mount_progress(sms->sm_done, sms->sm_total);
	pthread_mutex_unlock(&sms->sm_lock);
	return (ret);
	}

	static void
	append_options(char mntopts, char newopts)
	{
	int len = strlen(mntopts);

	/* original length plus new string to append plus 1 for the comma */
	if (len + 1 + strlen(newopts) >= MNT_LINE_MAX) {
	(void) fprintf(stderr, gettext("the opts argument for "
	"'%s' option is too long (more than %d chars)\n"),
	"-o", MNT_LINE_MAX);
	usage(B_FALSE);
	}

	if (*mntopts)
	mntopts[len++] = ',';

	(void) strcpy(&mntopts[len], newopts);
	}

	static enum sa_protocol
	sa_protocol_decode(const char *protocol)
	{
	for (enum sa_protocol i = 0; i < ARRAY_SIZE(sa_protocol_names); ++i)
	if (strcmp(protocol, sa_protocol_names[i]) == 0)
	return (i);

	(void) fputs(gettext("share type must be one of: "), stderr);
	for (enum sa_protocol i = 0;
	i < ARRAY_SIZE(sa_protocol_names); ++i)
	(void) fprintf(stderr, "%s%s",
	i != 0 ? ", " : "", sa_protocol_names[i]);
	(void) fputc('\n', stderr);
	usage(B_FALSE);
	}

	static int
	share_mount(int op, int argc, char **argv)
	{
	int do_all = 0;
	boolean_t verbose = B_FALSE;
	int c, ret = 0;
	char *options = NULL;
	int flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, op == OP_MOUNT ? ":alvo:Of" : "al"))
	!= -1) {
	switch (c) {
	case 'a':
	do_all = 1;
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case 'l':
	flags \|= MS_CRYPT;
	break;
	case 'o':
	if (*optarg == '\0') {
	(void) fprintf(stderr, gettext("empty mount "
	"options (-o) specified\n"));
	usage(B_FALSE);
	}

	if (options == NULL)
	options = safe_malloc(MNT_LINE_MAX + 1);

	/* option validation is done later */
	append_options(options, optarg);
	break;
	case 'O':
	flags \|= MS_OVERLAY;
	break;
	case 'f':
	flags \|= MS_FORCE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (do_all) {
	enum sa_protocol protocol = SA_NO_PROTOCOL;

	if (op == OP_SHARE && argc > 0) {
	protocol = sa_protocol_decode(argv[0]);
	argc--;
	argv++;
	}

	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	start_progress_timer();
	get_all_cb_t cb = { 0 };
	get_all_datasets(&cb, verbose);

	if (cb.cb_used == 0) {
	free(options);
	return (0);
	}

	share_mount_state_t share_mount_state = { 0 };
	share_mount_state.sm_op = op;
	share_mount_state.sm_verbose = verbose;
	share_mount_state.sm_flags = flags;
	share_mount_state.sm_options = options;
	share_mount_state.sm_proto = protocol;
	share_mount_state.sm_total = cb.cb_used;
	pthread_mutex_init(&share_mount_state.sm_lock, NULL);

	/* For a 'zfs share -a' operation start with a clean slate. */
	zfs_truncate_shares(NULL);

	/*
	* libshare isn't mt-safe, so only do the operation in parallel
	* if we're mounting. Additionally, the key-loading option must
	* be serialized so that we can prompt the user for their keys
	* in a consistent manner.
	*/
	zfs_foreach_mountpoint(g_zfs, cb.cb_handles, cb.cb_used,
	share_mount_one_cb, &share_mount_state,
	op == OP_MOUNT && !(flags & MS_CRYPT));
	zfs_commit_shares(NULL);

	ret = share_mount_state.sm_status;

	for (int i = 0; i < cb.cb_used; i++)
	zfs_close(cb.cb_handles[i]);
	free(cb.cb_handles);
	} else if (argc == 0) {
	FILE *mnttab;
	struct mnttab entry;

	if ((op == OP_SHARE) \|\| (options != NULL)) {
	(void) fprintf(stderr, gettext("missing filesystem "
	"argument (specify -a for all)\n"));
	usage(B_FALSE);
	}

	/*
	* When mount is given no arguments, go through
	* /proc/self/mounts and display any active ZFS mounts.
	* We hide any snapshots, since they are controlled
	* automatically.
	*/

	if ((mnttab = fopen(MNTTAB, "re")) == NULL) {
	free(options);
	return (ENOENT);
	}

	while (getmntent(mnttab, &entry) == 0) {
	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0 \|\|
	strchr(entry.mnt_special, '@') != NULL)
	continue;

	(void) printf("%-30s %s\n", entry.mnt_special,
	entry.mnt_mountp);
	}

	(void) fclose(mnttab);
	} else {
	zfs_handle_t *zhp;

	if (argc > 1) {
	(void) fprintf(stderr,
	gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM)) == NULL) {
	ret = 1;
	} else {
	ret = share_mount_one(zhp, op, flags, SA_NO_PROTOCOL,
	B_TRUE, options);
	zfs_commit_shares(NULL);
	zfs_close(zhp);
	}
	}

	free(options);
	return (ret);
	}

	/*
	* zfs mount -a
	* zfs mount filesystem
	*
	* Mount all filesystems, or mount the given filesystem.
	*/
	static int
	zfs_do_mount(int argc, char **argv)
	{
	return (share_mount(OP_MOUNT, argc, argv));
	}

	/*
	* zfs share -a [nfs \| smb]
	* zfs share filesystem
	*
	* Share all filesystems, or share the given filesystem.
	*/
	static int
	zfs_do_share(int argc, char **argv)
	{
	return (share_mount(OP_SHARE, argc, argv));
	}

	typedef struct unshare_unmount_node {
	zfs_handle_t *un_zhp;
	char *un_mountp;
	uu_avl_node_t un_avlnode;
	} unshare_unmount_node_t;

	static int
	unshare_unmount_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const unshare_unmount_node_t *l = larg;
	const unshare_unmount_node_t *r = rarg;

	return (strcmp(l->un_mountp, r->un_mountp));
	}

	/*
	* Convenience routine used by zfs_do_umount() and manual_unmount(). Given an
	* absolute path, find the entry /proc/self/mounts, verify that it's a
	* ZFS filesystem, and unmount it appropriately.
	*/
	static int
	unshare_unmount_path(int op, char *path, int flags, boolean_t is_manual)
	{
	zfs_handle_t *zhp;
	int ret = 0;
	struct stat64 statbuf;
	struct extmnttab entry;
	const char *cmdname = (op == OP_SHARE) ? "unshare" : "unmount";
	ino_t path_inode;

	/*
	* Search for the given (major,minor) pair in the mount table.
	*/

	if (getextmntent(path, &entry, &statbuf) != 0) {
	if (op == OP_SHARE) {
	(void) fprintf(stderr, gettext("cannot %s '%s': not "
	"currently mounted\n"), cmdname, path);
	return (1);
	}
	(void) fprintf(stderr, gettext("warning: %s not in"
	"/proc/self/mounts\n"), path);
	if ((ret = umount2(path, flags)) != 0)
	(void) fprintf(stderr, gettext("%s: %s\n"), path,
	strerror(errno));
	return (ret != 0);
	}
	path_inode = statbuf.st_ino;

	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) {
	(void) fprintf(stderr, gettext("cannot %s '%s': not a ZFS "
	"filesystem\n"), cmdname, path);
	return (1);
	}

	if ((zhp = zfs_open(g_zfs, entry.mnt_special,
	ZFS_TYPE_FILESYSTEM)) == NULL)
	return (1);

	ret = 1;
	if (stat64(entry.mnt_mountp, &statbuf) != 0) {
	(void) fprintf(stderr, gettext("cannot %s '%s': %s\n"),
	cmdname, path, strerror(errno));
	goto out;
	} else if (statbuf.st_ino != path_inode) {
	(void) fprintf(stderr, gettext("cannot "
	"%s '%s': not a mountpoint\n"), cmdname, path);
	goto out;
	}

	if (op == OP_SHARE) {
	char nfs_mnt_prop[ZFS_MAXPROPLEN];
	char smbshare_prop[ZFS_MAXPROPLEN];

	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, nfs_mnt_prop,
	sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshare_prop,
	sizeof (smbshare_prop), NULL, NULL, 0, B_FALSE) == 0);

	if (strcmp(nfs_mnt_prop, "off") == 0 &&
	strcmp(smbshare_prop, "off") == 0) {
	(void) fprintf(stderr, gettext("cannot unshare "
	"'%s': legacy share\n"), path);
	(void) fprintf(stderr, gettext("use exportfs(8) "
	"or smbcontrol(1) to unshare this filesystem\n"));
	} else if (!zfs_is_shared(zhp, NULL, NULL)) {
	(void) fprintf(stderr, gettext("cannot unshare '%s': "
	"not currently shared\n"), path);
	} else {
	ret = zfs_unshare(zhp, path, NULL);
	zfs_commit_shares(NULL);
	}
	} else {
	char mtpt_prop[ZFS_MAXPROPLEN];

	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mtpt_prop,
	sizeof (mtpt_prop), NULL, NULL, 0, B_FALSE) == 0);

	if (is_manual) {
	ret = zfs_unmount(zhp, NULL, flags);
	} else if (strcmp(mtpt_prop, "legacy") == 0) {
	(void) fprintf(stderr, gettext("cannot unmount "
	"'%s': legacy mountpoint\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use umount(8) "
	"to unmount this filesystem\n"));
	} else {
	ret = zfs_unmountall(zhp, flags);
	}
	}

	out:
	zfs_close(zhp);

	return (ret != 0);
	}

	/*
	* Generic callback for unsharing or unmounting a filesystem.
	*/
	static int
	unshare_unmount(int op, int argc, char **argv)
	{
	int do_all = 0;
	int flags = 0;
	int ret = 0;
	int c;
	zfs_handle_t *zhp;
	char nfs_mnt_prop[ZFS_MAXPROPLEN];
	char sharesmb[ZFS_MAXPROPLEN];

	/* check options */
	while ((c = getopt(argc, argv, op == OP_SHARE ? ":a" : "afu")) != -1) {
	switch (c) {
	case 'a':
	do_all = 1;
	break;
	case 'f':
	flags \|= MS_FORCE;
	break;
	case 'u':
	flags \|= MS_CRYPT;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (do_all) {
	/*
	* We could make use of zfs_for_each() to walk all datasets in
	* the system, but this would be very inefficient, especially
	* since we would have to linearly search /proc/self/mounts for
	* each one. Instead, do one pass through /proc/self/mounts
	* looking for zfs entries and call zfs_unmount() for each one.
	*
	* Things get a little tricky if the administrator has created
	* mountpoints beneath other ZFS filesystems. In this case, we
	* have to unmount the deepest filesystems first. To accomplish
	* this, we place all the mountpoints in an AVL tree sorted by
	* the special type (dataset name), and walk the result in
	* reverse to make sure to get any snapshots first.
	*/
	FILE *mnttab;
	struct mnttab entry;
	uu_avl_pool_t *pool;
	uu_avl_t *tree = NULL;
	unshare_unmount_node_t *node;
	uu_avl_index_t idx;
	uu_avl_walk_t *walk;
	enum sa_protocol *protocol = NULL,
	single_protocol[] = {SA_NO_PROTOCOL, SA_NO_PROTOCOL};

	if (op == OP_SHARE && argc > 0) {
	*single_protocol = sa_protocol_decode(argv[0]);
	protocol = single_protocol;
	argc--;
	argv++;
	}

	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (((pool = uu_avl_pool_create("unmount_pool",
	sizeof (unshare_unmount_node_t),
	offsetof(unshare_unmount_node_t, un_avlnode),
	unshare_unmount_compare, UU_DEFAULT)) == NULL) \|\|
	((tree = uu_avl_create(pool, NULL, UU_DEFAULT)) == NULL))
	nomem();

	- if ((mnttab = fopen(MNTTAB, "re")) == NULL)
	+ if ((mnttab = fopen(MNTTAB, "re")) == NULL) {
	+ uu_avl_destroy(tree);
	+ uu_avl_pool_destroy(pool);
	return (ENOENT);
	+ }

	while (getmntent(mnttab, &entry) == 0) {

	/* ignore non-ZFS entries */
	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
	continue;

	/* ignore snapshots */
	if (strchr(entry.mnt_special, '@') != NULL)
	continue;

	if ((zhp = zfs_open(g_zfs, entry.mnt_special,
	ZFS_TYPE_FILESYSTEM)) == NULL) {
	ret = 1;
	continue;
	}

	/*
	* Ignore datasets that are excluded/restricted by
	* parent pool name.
	*/
	if (zpool_skip_pool(zfs_get_pool_name(zhp))) {
	zfs_close(zhp);
	continue;
	}

	switch (op) {
	case OP_SHARE:
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "off") != 0)
	break;
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "off") == 0)
	continue;
	break;
	case OP_MOUNT:
	/* Ignore legacy mounts */
	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "legacy") == 0)
	continue;
	/* Ignore canmount=noauto mounts */
	if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) ==
	ZFS_CANMOUNT_NOAUTO)
	continue;
	break;
	default:
	break;
	}

	node = safe_malloc(sizeof (unshare_unmount_node_t));
	node->un_zhp = zhp;
	node->un_mountp = safe_strdup(entry.mnt_mountp);

	uu_avl_node_init(node, &node->un_avlnode, pool);

	if (uu_avl_find(tree, node, NULL, &idx) == NULL) {
	uu_avl_insert(tree, node, idx);
	} else {
	zfs_close(node->un_zhp);
	free(node->un_mountp);
	free(node);
	}
	}
	(void) fclose(mnttab);

	/*
	* Walk the AVL tree in reverse, unmounting each filesystem and
	* removing it from the AVL tree in the process.
	*/
	if ((walk = uu_avl_walk_start(tree,
	UU_WALK_REVERSE \| UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((node = uu_avl_walk_next(walk)) != NULL) {
	const char *mntarg = NULL;

	uu_avl_remove(tree, node);
	switch (op) {
	case OP_SHARE:
	if (zfs_unshare(node->un_zhp,
	node->un_mountp, protocol) != 0)
	ret = 1;
	break;

	case OP_MOUNT:
	if (zfs_unmount(node->un_zhp,
	mntarg, flags) != 0)
	ret = 1;
	break;
	}

	zfs_close(node->un_zhp);
	free(node->un_mountp);
	free(node);
	}

	if (op == OP_SHARE)
	zfs_commit_shares(protocol);

	uu_avl_walk_end(walk);
	uu_avl_destroy(tree);
	uu_avl_pool_destroy(pool);

	} else {
	if (argc != 1) {
	if (argc == 0)
	(void) fprintf(stderr,
	gettext("missing filesystem argument\n"));
	else
	(void) fprintf(stderr,
	gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/*
	* We have an argument, but it may be a full path or a ZFS
	* filesystem. Pass full paths off to unmount_path() (shared by
	* manual_unmount), otherwise open the filesystem and pass to
	* zfs_unmount().
	*/
	if (argv[0][0] == '/')
	return (unshare_unmount_path(op, argv[0],
	flags, B_FALSE));

	if ((zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM)) == NULL)
	return (1);

	verify(zfs_prop_get(zhp, op == OP_SHARE ?
	ZFS_PROP_SHARENFS : ZFS_PROP_MOUNTPOINT,
	nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL,
	NULL, 0, B_FALSE) == 0);

	switch (op) {
	case OP_SHARE:
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB,
	sharesmb, sizeof (sharesmb), NULL, NULL,
	0, B_FALSE) == 0);

	if (strcmp(nfs_mnt_prop, "off") == 0 &&
	strcmp(sharesmb, "off") == 0) {
	(void) fprintf(stderr, gettext("cannot "
	"unshare '%s': legacy share\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use "
	"exports(5) or smb.conf(5) to unshare "
	"this filesystem\n"));
	ret = 1;
	} else if (!zfs_is_shared(zhp, NULL, NULL)) {
	(void) fprintf(stderr, gettext("cannot "
	"unshare '%s': not currently "
	"shared\n"), zfs_get_name(zhp));
	ret = 1;
	} else if (zfs_unshareall(zhp, NULL) != 0) {
	ret = 1;
	}
	break;

	case OP_MOUNT:
	if (strcmp(nfs_mnt_prop, "legacy") == 0) {
	(void) fprintf(stderr, gettext("cannot "
	"unmount '%s': legacy "
	"mountpoint\n"), zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use "
	"umount(8) to unmount this "
	"filesystem\n"));
	ret = 1;
	} else if (!zfs_is_mounted(zhp, NULL)) {
	(void) fprintf(stderr, gettext("cannot "
	"unmount '%s': not currently "
	"mounted\n"),
	zfs_get_name(zhp));
	ret = 1;
	} else if (zfs_unmountall(zhp, flags) != 0) {
	ret = 1;
	}
	break;
	}

	zfs_close(zhp);
	}

	return (ret);
	}

	/*
	* zfs unmount [-fu] -a
	* zfs unmount [-fu] filesystem
	*
	* Unmount all filesystems, or a specific ZFS filesystem.
	*/
	static int
	zfs_do_unmount(int argc, char **argv)
	{
	return (unshare_unmount(OP_MOUNT, argc, argv));
	}

	/*
	* zfs unshare -a
	* zfs unshare filesystem
	*
	* Unshare all filesystems, or a specific ZFS filesystem.
	*/
	static int
	zfs_do_unshare(int argc, char **argv)
	{
	return (unshare_unmount(OP_SHARE, argc, argv));
	}

	static int
	find_command_idx(const char command, int idx)
	{
	int i;

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	continue;

	if (strcmp(command, command_table[i].name) == 0) {
	*idx = i;
	return (0);
	}
	}
	return (1);
	}

	static int
	zfs_do_diff(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	int flags = 0;
	char *tosnap = NULL;
	char *fromsnap = NULL;
	char atp, copy;
	int err = 0;
	int c;
	struct sigaction sa;

	while ((c = getopt(argc, argv, "FHth")) != -1) {
	switch (c) {
	case 'F':
	flags \|= ZFS_DIFF_CLASSIFY;
	break;
	case 'H':
	flags \|= ZFS_DIFF_PARSEABLE;
	break;
	case 't':
	flags \|= ZFS_DIFF_TIMESTAMP;
	break;
	case 'h':
	flags \|= ZFS_DIFF_NO_MANGLE;
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr,
	gettext("must provide at least one snapshot name\n"));
	usage(B_FALSE);
	}

	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	fromsnap = argv[0];
	tosnap = (argc == 2) ? argv[1] : NULL;

	copy = NULL;
	if (*fromsnap != '@')
	copy = strdup(fromsnap);
	else if (tosnap)
	copy = strdup(tosnap);
	if (copy == NULL)
	usage(B_FALSE);

	if ((atp = strchr(copy, '@')) != NULL)
	*atp = '\0';

	if ((zhp = zfs_open(g_zfs, copy, ZFS_TYPE_FILESYSTEM)) == NULL) {
	free(copy);
	return (1);
	}
	free(copy);

	/*
	* Ignore SIGPIPE so that the library can give us
	* information on any failure
	*/
	if (sigemptyset(&sa.sa_mask) == -1) {
	err = errno;
	goto out;
	}
	sa.sa_flags = 0;
	sa.sa_handler = SIG_IGN;
	if (sigaction(SIGPIPE, &sa, NULL) == -1) {
	err = errno;
	goto out;
	}

	err = zfs_show_diffs(zhp, STDOUT_FILENO, fromsnap, tosnap, flags);
	out:
	zfs_close(zhp);

	return (err != 0);
	}

	/*
	* zfs bookmark <fs@source>\|<fs#source> <fs#bookmark>
	*
	* Creates a bookmark with the given name from the source snapshot
	* or creates a copy of an existing source bookmark.
	*/
	static int
	zfs_do_bookmark(int argc, char **argv)
	{
	char source, bookname;
	char expbuf[ZFS_MAX_DATASET_NAME_LEN];
	int source_type;
	nvlist_t *nvl;
	int ret = 0;
	int c;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source argument\n"));
	goto usage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing bookmark argument\n"));
	goto usage;
	}

	source = argv[0];
	bookname = argv[1];

	if (strchr(source, '@') == NULL && strchr(source, '#') == NULL) {
	(void) fprintf(stderr,
	gettext("invalid source name '%s': "
	"must contain a '@' or '#'\n"), source);
	goto usage;
	}
	if (strchr(bookname, '#') == NULL) {
	(void) fprintf(stderr,
	gettext("invalid bookmark name '%s': "
	"must contain a '#'\n"), bookname);
	goto usage;
	}

	/*
	* expand source or bookname to full path:
	* one of them may be specified as short name
	*/
	{
	char **expand;
	char source_short, bookname_short;
	source_short = strpbrk(source, "@#");
	bookname_short = strpbrk(bookname, "#");
	if (source_short == source &&
	bookname_short == bookname) {
	(void) fprintf(stderr, gettext(
	"either source or bookmark must be specified as "
	"full dataset paths"));
	goto usage;
	} else if (source_short != source &&
	bookname_short != bookname) {
	expand = NULL;
	} else if (source_short != source) {
	strlcpy(expbuf, source, sizeof (expbuf));
	expand = &bookname;
	} else if (bookname_short != bookname) {
	strlcpy(expbuf, bookname, sizeof (expbuf));
	expand = &source;
	} else {
	abort();
	}
	if (expand != NULL) {
	strpbrk(expbuf, "@#") = '\0'; / dataset name in buf */
	(void) strlcat(expbuf, *expand, sizeof (expbuf));
	*expand = expbuf;
	}
	}

	/* determine source type */
	switch (*strpbrk(source, "@#")) {
	case '@': source_type = ZFS_TYPE_SNAPSHOT; break;
	case '#': source_type = ZFS_TYPE_BOOKMARK; break;
	default: abort();
	}

	/* test the source exists */
	zfs_handle_t *zhp;
	zhp = zfs_open(g_zfs, source, source_type);
	if (zhp == NULL)
	goto usage;
	zfs_close(zhp);

	nvl = fnvlist_alloc();
	fnvlist_add_string(nvl, bookname, source);
	ret = lzc_bookmark(nvl, NULL);
	fnvlist_free(nvl);

	if (ret != 0) {
	const char *err_msg = NULL;
	char errbuf[1024];

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot create bookmark '%s'"), bookname);

	switch (ret) {
	case EXDEV:
	err_msg = "bookmark is in a different pool";
	break;
	case ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR:
	err_msg = "source is not an ancestor of the "
	"new bookmark's dataset";
	break;
	case EEXIST:
	err_msg = "bookmark exists";
	break;
	case EINVAL:
	err_msg = "invalid argument";
	break;
	case ENOTSUP:
	err_msg = "bookmark feature not enabled";
	break;
	case ENOSPC:
	err_msg = "out of space";
	break;
	case ENOENT:
	err_msg = "dataset does not exist";
	break;
	default:
	(void) zfs_standard_error(g_zfs, ret, errbuf);
	break;
	}
	if (err_msg != NULL) {
	(void) fprintf(stderr, "%s: %s\n", errbuf,
	dgettext(TEXT_DOMAIN, err_msg));
	}
	}

	return (ret != 0);

	usage:
	usage(B_FALSE);
	return (-1);
	}

	static int
	zfs_do_channel_program(int argc, char **argv)
	{
	int ret, fd, c;
	size_t progsize, progread;
	nvlist_t *outnvl = NULL;
	uint64_t instrlimit = ZCP_DEFAULT_INSTRLIMIT;
	uint64_t memlimit = ZCP_DEFAULT_MEMLIMIT;
	boolean_t sync_flag = B_TRUE, json_output = B_FALSE;
	zpool_handle_t *zhp;

	/* check options */
	while ((c = getopt(argc, argv, "nt:m:j")) != -1) {
	switch (c) {
	case 't':
	case 'm': {
	uint64_t arg;
	char *endp;

	errno = 0;
	arg = strtoull(optarg, &endp, 0);
	if (errno != 0 \|\| *endp != '\0') {
	(void) fprintf(stderr, gettext(
	"invalid argument "
	"'%s': expected integer\n"), optarg);
	goto usage;
	}

	if (c == 't') {
	instrlimit = arg;
	} else {
	ASSERT3U(c, ==, 'm');
	memlimit = arg;
	}
	break;
	}
	case 'n': {
	sync_flag = B_FALSE;
	break;
	}
	case 'j': {
	json_output = B_TRUE;
	break;
	}
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("invalid number of arguments\n"));
	goto usage;
	}

	const char *poolname = argv[0];
	const char *filename = argv[1];
	if (strcmp(filename, "-") == 0) {
	fd = 0;
	filename = "standard input";
	} else if ((fd = open(filename, O_RDONLY)) < 0) {
	(void) fprintf(stderr, gettext("cannot open '%s': %s\n"),
	filename, strerror(errno));
	return (1);
	}

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
	(void) fprintf(stderr, gettext("cannot open pool '%s'\n"),
	poolname);
	if (fd != 0)
	(void) close(fd);
	return (1);
	}
	zpool_close(zhp);

	/*
	* Read in the channel program, expanding the program buffer as
	* necessary.
	*/
	progread = 0;
	progsize = 1024;
	char *progbuf = safe_malloc(progsize);
	do {
	ret = read(fd, progbuf + progread, progsize - progread);
	progread += ret;
	if (progread == progsize && ret > 0) {
	progsize *= 2;
	progbuf = safe_realloc(progbuf, progsize);
	}
	} while (ret > 0);

	if (fd != 0)
	(void) close(fd);
	if (ret < 0) {
	free(progbuf);
	(void) fprintf(stderr,
	gettext("cannot read '%s': %s\n"),
	filename, strerror(errno));
	return (1);
	}
	progbuf[progread] = '\0';

	/*
	* Any remaining arguments are passed as arguments to the lua script as
	* a string array:
	* {
	* "argv" -> [ "arg 1", ... "arg n" ],
	* }
	*/
	nvlist_t *argnvl = fnvlist_alloc();
	fnvlist_add_string_array(argnvl, ZCP_ARG_CLIARGV,
	(const char **)argv + 2, argc - 2);

	if (sync_flag) {
	ret = lzc_channel_program(poolname, progbuf,
	instrlimit, memlimit, argnvl, &outnvl);
	} else {
	ret = lzc_channel_program_nosync(poolname, progbuf,
	instrlimit, memlimit, argnvl, &outnvl);
	}

	if (ret != 0) {
	/*
	* On error, report the error message handed back by lua if one
	* exists. Otherwise, generate an appropriate error message,
	* falling back on strerror() for an unexpected return code.
	*/
	const char *errstring = NULL;
	const char *msg = gettext("Channel program execution failed");
	uint64_t instructions = 0;
	if (outnvl != NULL && nvlist_exists(outnvl, ZCP_RET_ERROR)) {
	char *es = NULL;
	(void) nvlist_lookup_string(outnvl,
	ZCP_RET_ERROR, &es);
	if (es == NULL)
	errstring = strerror(ret);
	else
	errstring = es;
	if (ret == ETIME) {
	(void) nvlist_lookup_uint64(outnvl,
	ZCP_ARG_INSTRLIMIT, &instructions);
	}
	} else {
	switch (ret) {
	case EINVAL:
	errstring =
	"Invalid instruction or memory limit.";
	break;
	case ENOMEM:
	errstring = "Return value too large.";
	break;
	case ENOSPC:
	errstring = "Memory limit exhausted.";
	break;
	case ETIME:
	errstring = "Timed out.";
	break;
	case EPERM:
	errstring = "Permission denied. Channel "
	"programs must be run as root.";
	break;
	default:
	(void) zfs_standard_error(g_zfs, ret, msg);
	}
	}
	if (errstring != NULL)
	(void) fprintf(stderr, "%s:\n%s\n", msg, errstring);

	if (ret == ETIME && instructions != 0)
	(void) fprintf(stderr,
	gettext("%llu Lua instructions\n"),
	(u_longlong_t)instructions);
	} else {
	if (json_output) {
	(void) nvlist_print_json(stdout, outnvl);
	} else if (nvlist_empty(outnvl)) {
	(void) fprintf(stdout, gettext("Channel program fully "
	"executed and did not produce output.\n"));
	} else {
	(void) fprintf(stdout, gettext("Channel program fully "
	"executed and produced output:\n"));
	dump_nvlist(outnvl, 4);
	}
	}

	free(progbuf);
	fnvlist_free(outnvl);
	fnvlist_free(argnvl);
	return (ret != 0);

	usage:
	usage(B_FALSE);
	return (-1);
	}


	typedef struct loadkey_cbdata {
	boolean_t cb_loadkey;
	boolean_t cb_recursive;
	boolean_t cb_noop;
	char *cb_keylocation;
	uint64_t cb_numfailed;
	uint64_t cb_numattempted;
	} loadkey_cbdata_t;

	static int
	load_key_callback(zfs_handle_t zhp, void data)
	{
	int ret;
	boolean_t is_encroot;
	loadkey_cbdata_t *cb = data;
	uint64_t keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS);

	/*
	* If we are working recursively, we want to skip loading / unloading
	* keys for non-encryption roots and datasets whose keys are already
	* in the desired end-state.
	*/
	if (cb->cb_recursive) {
	ret = zfs_crypto_get_encryption_root(zhp, &is_encroot, NULL);
	if (ret != 0)
	return (ret);
	if (!is_encroot)
	return (0);

	if ((cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_AVAILABLE) \|\|
	(!cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_UNAVAILABLE))
	return (0);
	}

	cb->cb_numattempted++;

	if (cb->cb_loadkey)
	ret = zfs_crypto_load_key(zhp, cb->cb_noop, cb->cb_keylocation);
	else
	ret = zfs_crypto_unload_key(zhp);

	if (ret != 0) {
	cb->cb_numfailed++;
	return (ret);
	}

	return (0);
	}

	static int
	load_unload_keys(int argc, char **argv, boolean_t loadkey)
	{
	int c, ret = 0, flags = 0;
	boolean_t do_all = B_FALSE;
	loadkey_cbdata_t cb = { 0 };

	cb.cb_loadkey = loadkey;

	while ((c = getopt(argc, argv, "anrL:")) != -1) {
	/* noop and alternate keylocations only apply to zfs load-key */
	if (loadkey) {
	switch (c) {
	case 'n':
	cb.cb_noop = B_TRUE;
	continue;
	case 'L':
	cb.cb_keylocation = optarg;
	continue;
	default:
	break;
	}
	}

	switch (c) {
	case 'a':
	do_all = B_TRUE;
	cb.cb_recursive = B_TRUE;
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	cb.cb_recursive = B_TRUE;
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (!do_all && argc == 0) {
	(void) fprintf(stderr,
	gettext("Missing dataset argument or -a option\n"));
	usage(B_FALSE);
	}

	if (do_all && argc != 0) {
	(void) fprintf(stderr,
	gettext("Cannot specify dataset with -a option\n"));
	usage(B_FALSE);
	}

	if (cb.cb_recursive && cb.cb_keylocation != NULL &&
	strcmp(cb.cb_keylocation, "prompt") != 0) {
	(void) fprintf(stderr, gettext("alternate keylocation may only "
	"be 'prompt' with -r or -a\n"));
	usage(B_FALSE);
	}

	ret = zfs_for_each(argc, argv, flags,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME, NULL, NULL, 0,
	load_key_callback, &cb);

	if (cb.cb_noop \|\| (cb.cb_recursive && cb.cb_numattempted != 0)) {
	(void) printf(gettext("%llu / %llu key(s) successfully %s\n"),
	(u_longlong_t)(cb.cb_numattempted - cb.cb_numfailed),
	(u_longlong_t)cb.cb_numattempted,
	loadkey ? (cb.cb_noop ? "verified" : "loaded") :
	"unloaded");
	}

	if (cb.cb_numfailed != 0)
	ret = -1;

	return (ret);
	}

	static int
	zfs_do_load_key(int argc, char **argv)
	{
	return (load_unload_keys(argc, argv, B_TRUE));
	}


	static int
	zfs_do_unload_key(int argc, char **argv)
	{
	return (load_unload_keys(argc, argv, B_FALSE));
	}

	static int
	zfs_do_change_key(int argc, char **argv)
	{
	int c, ret;
	uint64_t keystatus;
	boolean_t loadkey = B_FALSE, inheritkey = B_FALSE;
	zfs_handle_t *zhp = NULL;
	nvlist_t *props = fnvlist_alloc();

	while ((c = getopt(argc, argv, "lio:")) != -1) {
	switch (c) {
	case 'l':
	loadkey = B_TRUE;
	break;
	case 'i':
	inheritkey = B_TRUE;
	break;
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	return (1);
	}
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	if (inheritkey && !nvlist_empty(props)) {
	(void) fprintf(stderr,
	gettext("Properties not allowed for inheriting\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("Missing dataset argument\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("Too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[argc - 1],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	usage(B_FALSE);

	if (loadkey) {
	keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS);
	if (keystatus != ZFS_KEYSTATUS_AVAILABLE) {
	ret = zfs_crypto_load_key(zhp, B_FALSE, NULL);
	if (ret != 0) {
	nvlist_free(props);
	zfs_close(zhp);
	return (-1);
	}
	}

	/* refresh the properties so the new keystatus is visible */
	zfs_refresh_properties(zhp);
	}

	ret = zfs_crypto_rewrap(zhp, props, inheritkey);
	if (ret != 0) {
	nvlist_free(props);
	zfs_close(zhp);
	return (-1);
	}

	nvlist_free(props);
	zfs_close(zhp);
	return (0);
	}

	/*
	* 1) zfs project [-d\|-r] <file\|directory ...>
	* List project ID and inherit flag of file(s) or directories.
	* -d: List the directory itself, not its children.
	* -r: List subdirectories recursively.
	*
	* 2) zfs project -C [-k] [-r] <file\|directory ...>
	* Clear project inherit flag and/or ID on the file(s) or directories.
	* -k: Keep the project ID unchanged. If not specified, the project ID
	* will be reset as zero.
	* -r: Clear on subdirectories recursively.
	*
	* 3) zfs project -c [-0] [-d\|-r] [-p id] <file\|directory ...>
	* Check project ID and inherit flag on the file(s) or directories,
	* report the outliers.
	* -0: Print file name followed by a NUL instead of newline.
	* -d: Check the directory itself, not its children.
	* -p: Specify the referenced ID for comparing with the target file(s)
	* or directories' project IDs. If not specified, the target (top)
	* directory's project ID will be used as the referenced one.
	* -r: Check subdirectories recursively.
	*
	* 4) zfs project [-p id] [-r] [-s] <file\|directory ...>
	* Set project ID and/or inherit flag on the file(s) or directories.
	* -p: Set the project ID as the given id.
	* -r: Set on subdirectories recursively. If not specify "-p" option,
	* it will use top-level directory's project ID as the given id,
	* then set both project ID and inherit flag on all descendants
	* of the top-level directory.
	* -s: Set project inherit flag.
	*/
	static int
	zfs_do_project(int argc, char **argv)
	{
	zfs_project_control_t zpc = {
	.zpc_expected_projid = ZFS_INVALID_PROJID,
	.zpc_op = ZFS_PROJECT_OP_DEFAULT,
	.zpc_dironly = B_FALSE,
	.zpc_keep_projid = B_FALSE,
	.zpc_newline = B_TRUE,
	.zpc_recursive = B_FALSE,
	.zpc_set_flag = B_FALSE,
	};
	int ret = 0, c;

	if (argc < 2)
	usage(B_FALSE);

	while ((c = getopt(argc, argv, "0Ccdkp:rs")) != -1) {
	switch (c) {
	case '0':
	zpc.zpc_newline = B_FALSE;
	break;
	case 'C':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_op = ZFS_PROJECT_OP_CLEAR;
	break;
	case 'c':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_op = ZFS_PROJECT_OP_CHECK;
	break;
	case 'd':
	zpc.zpc_dironly = B_TRUE;
	/* overwrite "-r" option */
	zpc.zpc_recursive = B_FALSE;
	break;
	case 'k':
	zpc.zpc_keep_projid = B_TRUE;
	break;
	case 'p': {
	char *endptr;

	errno = 0;
	zpc.zpc_expected_projid = strtoull(optarg, &endptr, 0);
	if (errno != 0 \|\| *endptr != '\0') {
	(void) fprintf(stderr,
	gettext("project ID must be less than "
	"%u\n"), UINT32_MAX);
	usage(B_FALSE);
	}
	if (zpc.zpc_expected_projid >= UINT32_MAX) {
	(void) fprintf(stderr,
	gettext("invalid project ID\n"));
	usage(B_FALSE);
	}
	break;
	}
	case 'r':
	zpc.zpc_recursive = B_TRUE;
	/* overwrite "-d" option */
	zpc.zpc_dironly = B_FALSE;
	break;
	case 's':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_set_flag = B_TRUE;
	zpc.zpc_op = ZFS_PROJECT_OP_SET;
	break;
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (zpc.zpc_op == ZFS_PROJECT_OP_DEFAULT) {
	if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID)
	zpc.zpc_op = ZFS_PROJECT_OP_SET;
	else
	zpc.zpc_op = ZFS_PROJECT_OP_LIST;
	}

	switch (zpc.zpc_op) {
	case ZFS_PROJECT_OP_LIST:
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_CHECK:
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_CLEAR:
	if (zpc.zpc_dironly) {
	(void) fprintf(stderr,
	gettext("'-d' is useless together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID) {
	(void) fprintf(stderr,
	gettext("'-p' is useless together with '-C'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_SET:
	if (zpc.zpc_dironly) {
	(void) fprintf(stderr,
	gettext("'-d' is useless for set project ID and/or "
	"inherit flag\n"));
	usage(B_FALSE);
	}
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	break;
	default:
	ASSERT(0);
	break;
	}

	argv += optind;
	argc -= optind;
	if (argc == 0) {
	(void) fprintf(stderr,
	gettext("missing file or directory target(s)\n"));
	usage(B_FALSE);
	}

	for (int i = 0; i < argc; i++) {
	int err;

	err = zfs_project_handle(argv[i], &zpc);
	if (err && !ret)
	ret = err;
	}

	return (ret);
	}

	static int
	zfs_do_wait(int argc, char **argv)
	{
	boolean_t enabled[ZFS_WAIT_NUM_ACTIVITIES];
	int error, i;
	int c;

	/* By default, wait for all types of activity. */
	for (i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++)
	enabled[i] = B_TRUE;

	while ((c = getopt(argc, argv, "t:")) != -1) {
	switch (c) {
	case 't':
	/* Reset activities array */
	memset(&enabled, 0, sizeof (enabled));

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_subopts[
	ZFS_WAIT_NUM_ACTIVITIES] = { "deleteq" };

	for (i = 0; i < ARRAY_SIZE(col_subopts); ++i)
	if (strcmp(tok, col_subopts[i]) == 0) {
	enabled[i] = B_TRUE;
	goto found;
	}

	(void) fprintf(stderr,
	gettext("invalid activity '%s'\n"), tok);
	usage(B_FALSE);
	found:;
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argv += optind;
	argc -= optind;
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing 'filesystem' "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zfs_handle_t *zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	for (;;) {
	boolean_t missing = B_FALSE;
	boolean_t any_waited = B_FALSE;

	for (int i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++) {
	boolean_t waited;

	if (!enabled[i])
	continue;

	error = zfs_wait_status(zhp, i, &missing, &waited);
	if (error != 0 \|\| missing)
	break;

	any_waited = (any_waited \|\| waited);
	}

	if (error != 0 \|\| missing \|\| !any_waited)
	break;
	}

	zfs_close(zhp);

	return (error);
	}

	/*
	* Display version message
	*/
	static int
	zfs_do_version(int argc, char **argv)
	{
	(void) argc, (void) argv;
	return (zfs_version_print() != 0);
	}

	int
	main(int argc, char **argv)
	{
	int ret = 0;
	int i = 0;
	const char *cmdname;
	char **newargv;

	(void) setlocale(LC_ALL, "");
	(void) setlocale(LC_NUMERIC, "C");
	(void) textdomain(TEXT_DOMAIN);

	opterr = 0;

	/*
	* Make sure the user has specified some command.
	*/
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing command\n"));
	usage(B_FALSE);
	}

	cmdname = argv[1];

	/*
	* The 'umount' command is an alias for 'unmount'
	*/
	if (strcmp(cmdname, "umount") == 0)
	cmdname = "unmount";

	/*
	* The 'recv' command is an alias for 'receive'
	*/
	if (strcmp(cmdname, "recv") == 0)
	cmdname = "receive";

	/*
	* The 'snap' command is an alias for 'snapshot'
	*/
	if (strcmp(cmdname, "snap") == 0)
	cmdname = "snapshot";

	/*
	* Special case '-?'
	*/
	if ((strcmp(cmdname, "-?") == 0) \|\|
	(strcmp(cmdname, "--help") == 0))
	usage(B_TRUE);

	/*
	* Special case '-V\|--version'
	*/
	if ((strcmp(cmdname, "-V") == 0) \|\| (strcmp(cmdname, "--version") == 0))
	return (zfs_do_version(argc, argv));

	if ((g_zfs = libzfs_init()) == NULL) {
	(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
	return (1);
	}

	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));

	libzfs_print_on_error(g_zfs, B_TRUE);

	/*
	* Many commands modify input strings for string parsing reasons.
	* We create a copy to protect the original argv.
	*/
	newargv = safe_malloc((argc + 1) * sizeof (newargv[0]));
	for (i = 0; i < argc; i++)
	newargv[i] = strdup(argv[i]);
	newargv[argc] = NULL;

	/*
	* Run the appropriate command.
	*/
	libzfs_mnttab_cache(g_zfs, B_TRUE);
	if (find_command_idx(cmdname, &i) == 0) {
	current_command = &command_table[i];
	ret = command_table[i].func(argc - 1, newargv + 1);
	} else if (strchr(cmdname, '=') != NULL) {
	verify(find_command_idx("set", &i) == 0);
	current_command = &command_table[i];
	ret = command_table[i].func(argc, newargv);
	} else {
	(void) fprintf(stderr, gettext("unrecognized "
	"command '%s'\n"), cmdname);
	usage(B_FALSE);
	ret = 1;
	}

	for (i = 0; i < argc; i++)
	free(newargv[i]);
	free(newargv);

	if (ret == 0 && log_history)
	(void) zpool_log_history(g_zfs, history_str);

	libzfs_fini(g_zfs);

	/*
	* The 'ZFS_ABORT' environment variable causes us to dump core on exit
	* for the purposes of running ::findleaks.
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	return (ret);
	}

	/*
	* zfs zone nsfile filesystem
	*
	* Add or delete the given dataset to/from the namespace.
	*/
	#ifdef __linux__
	static int
	zfs_do_zone_impl(int argc, char **argv, boolean_t attach)
	{
	zfs_handle_t *zhp;
	int ret;

	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing argument(s)\n"));
	usage(B_FALSE);
	}
	if (argc > 3) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	ret = (zfs_userns(zhp, argv[1], attach) != 0);

	zfs_close(zhp);
	return (ret);
	}

	static int
	zfs_do_zone(int argc, char **argv)
	{
	return (zfs_do_zone_impl(argc, argv, B_TRUE));
	}

	static int
	zfs_do_unzone(int argc, char **argv)
	{
	return (zfs_do_zone_impl(argc, argv, B_FALSE));
	}
	#endif

	#ifdef __FreeBSD__
	#include <sys/jail.h>
	#include <jail.h>
	/*
	* Attach/detach the given dataset to/from the given jail
	*/
	static int
	zfs_do_jail_impl(int argc, char **argv, boolean_t attach)
	{
	zfs_handle_t *zhp;
	int jailid, ret;

	/* check number of arguments */
	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing argument(s)\n"));
	usage(B_FALSE);
	}
	if (argc > 3) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	jailid = jail_getid(argv[1]);
	if (jailid < 0) {
	(void) fprintf(stderr, gettext("invalid jail id or name\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	ret = (zfs_jail(zhp, jailid, attach) != 0);

	zfs_close(zhp);
	return (ret);
	}

	/*
	* zfs jail jailid filesystem
	*
	* Attach the given dataset to the given jail
	*/
	static int
	zfs_do_jail(int argc, char **argv)
	{
	return (zfs_do_jail_impl(argc, argv, B_TRUE));
	}

	/*
	* zfs unjail jailid filesystem
	*
	* Detach the given dataset from the given jail
	*/
	static int
	zfs_do_unjail(int argc, char **argv)
	{
	return (zfs_do_jail_impl(argc, argv, B_FALSE));
	}
	#endif
	diff --git a/sys/contrib/openzfs/cmd/zhack.c b/sys/contrib/openzfs/cmd/zhack.c
	index 75d9c93b3649..8797a53e4763 100644
	--- a/sys/contrib/openzfs/cmd/zhack.c
	+++ b/sys/contrib/openzfs/cmd/zhack.c
	@@ -1,699 +1,701 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2011, 2015 by Delphix. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	*/

	/*
	* zhack is a debugging tool that can write changes to ZFS pool using libzpool
	* for testing purposes. Altering pools with zhack is unsupported and may
	* result in corrupted pools.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	#include <ctype.h>
	#include <sys/stat.h>
	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/zap.h>
	#include <sys/zfs_znode.h>
	#include <sys/dsl_synctask.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/fs/zfs.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_pool.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/zfeature.h>
	#include <sys/dmu_tx.h>
	#include <zfeature_common.h>
	#include <libzutil.h>

	static importargs_t g_importargs;
	static char *g_pool;
	static boolean_t g_readonly;

	static __attribute__((noreturn)) void
	usage(void)
	{
	(void) fprintf(stderr,
	"Usage: zhack [-c cachefile] [-d dir] <subcommand> <args> ...\n"
	"where <subcommand> <args> is one of the following:\n"
	"\n");

	(void) fprintf(stderr,
	" feature stat <pool>\n"
	" print information about enabled features\n"
	" feature enable [-r] [-d desc] <pool> <feature>\n"
	" add a new enabled feature to the pool\n"
	" -d <desc> sets the feature's description\n"
	" -r set read-only compatible flag for feature\n"
	" feature ref [-md] <pool> <feature>\n"
	" change the refcount on the given feature\n"
	" -d decrease instead of increase the refcount\n"
	" -m add the feature to the label if increasing refcount\n"
	"\n"
	" <feature> : should be a feature guid\n"
	"\n"
	" label repair <device>\n"
	" repair corrupted label checksums\n"
	"\n"
	" <device> : path to vdev\n");
	exit(1);
	}


	static __attribute__((format(printf, 3, 4))) __attribute__((noreturn)) void
	fatal(spa_t spa, const void tag, const char *fmt, ...)
	{
	va_list ap;

	if (spa != NULL) {
	spa_close(spa, tag);
	(void) spa_export(g_pool, NULL, B_TRUE, B_FALSE);
	}

	va_start(ap, fmt);
	(void) fputs("zhack: ", stderr);
	(void) vfprintf(stderr, fmt, ap);
	va_end(ap);
	(void) fputc('\n', stderr);

	exit(1);
	}

	static int
	space_delta_cb(dmu_object_type_t bonustype, const void *data,
	zfs_file_info_t *zoi)
	{
	(void) data, (void) zoi;

	/*
	* Is it a valid type of object to track?
	*/
	if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
	return (ENOENT);
	(void) fprintf(stderr, "modifying object that needs user accounting");
	abort();
	}

	/*
	* Target is the dataset whose pool we want to open.
	*/
	static void
	zhack_import(char *target, boolean_t readonly)
	{
	nvlist_t *config;
	nvlist_t *props;
	int error;

	kernel_init(readonly ? SPA_MODE_READ :
	(SPA_MODE_READ \| SPA_MODE_WRITE));

	dmu_objset_register_type(DMU_OST_ZFS, space_delta_cb);

	g_readonly = readonly;
	g_importargs.can_be_active = readonly;
	g_pool = strdup(target);

	error = zpool_find_config(NULL, target, &config, &g_importargs,
	&libzpool_config_ops);
	if (error)
	fatal(NULL, FTAG, "cannot import '%s'", target);

	props = NULL;
	if (readonly) {
	VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0);
	VERIFY(nvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_READONLY), 1) == 0);
	}

	zfeature_checks_disable = B_TRUE;
	error = spa_import(target, config, props,
	(readonly ? ZFS_IMPORT_SKIP_MMP : ZFS_IMPORT_NORMAL));
	fnvlist_free(config);
	zfeature_checks_disable = B_FALSE;
	if (error == EEXIST)
	error = 0;

	if (error)
	fatal(NULL, FTAG, "can't import '%s': %s", target,
	strerror(error));
	}

	static void
	zhack_spa_open(char target, boolean_t readonly, const void tag, spa_t **spa)
	{
	int err;

	zhack_import(target, readonly);

	zfeature_checks_disable = B_TRUE;
	err = spa_open(target, spa, tag);
	zfeature_checks_disable = B_FALSE;

	if (err != 0)
	fatal(*spa, FTAG, "cannot open '%s': %s", target,
	strerror(err));
	if (spa_version(*spa) < SPA_VERSION_FEATURES) {
	fatal(*spa, FTAG, "'%s' has version %d, features not enabled",
	target, (int)spa_version(*spa));
	}
	}

	static void
	dump_obj(objset_t os, uint64_t obj, const char name)
	{
	zap_cursor_t zc;
	zap_attribute_t za;

	(void) printf("%s_obj:\n", name);

	for (zap_cursor_init(&zc, os, obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	if (za.za_integer_length == 8) {
	ASSERT(za.za_num_integers == 1);
	(void) printf("\t%s = %llu\n",
	za.za_name, (u_longlong_t)za.za_first_integer);
	} else {
	ASSERT(za.za_integer_length == 1);
	char val[1024];
	VERIFY(zap_lookup(os, obj, za.za_name,
	1, sizeof (val), val) == 0);
	(void) printf("\t%s = %s\n", za.za_name, val);
	}
	}
	zap_cursor_fini(&zc);
	}

	static void
	dump_mos(spa_t *spa)
	{
	nvlist_t *nv = spa->spa_label_features;
	nvpair_t *pair;

	(void) printf("label config:\n");
	for (pair = nvlist_next_nvpair(nv, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(nv, pair)) {
	(void) printf("\t%s\n", nvpair_name(pair));
	}
	}

	static void
	zhack_do_feature_stat(int argc, char **argv)
	{
	spa_t *spa;
	objset_t *os;
	char *target;

	argc--;
	argv++;

	if (argc < 1) {
	(void) fprintf(stderr, "error: missing pool name\n");
	usage();
	}
	target = argv[0];

	zhack_spa_open(target, B_TRUE, FTAG, &spa);
	os = spa->spa_meta_objset;

	dump_obj(os, spa->spa_feat_for_read_obj, "for_read");
	dump_obj(os, spa->spa_feat_for_write_obj, "for_write");
	dump_obj(os, spa->spa_feat_desc_obj, "descriptions");
	if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
	dump_obj(os, spa->spa_feat_enabled_txg_obj, "enabled_txg");
	}
	dump_mos(spa);

	spa_close(spa, FTAG);
	}

	static void
	zhack_feature_enable_sync(void arg, dmu_tx_t tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	zfeature_info_t *feature = arg;

	feature_enable_sync(spa, feature, tx);

	spa_history_log_internal(spa, "zhack enable feature", tx,
	"name=%s flags=%u",
	feature->fi_guid, feature->fi_flags);
	}

	static void
	zhack_do_feature_enable(int argc, char **argv)
	{
	int c;
	char desc, target;
	spa_t *spa;
	objset_t *mos;
	zfeature_info_t feature;
	const spa_feature_t nodeps[] = { SPA_FEATURE_NONE };

	/*
	* Features are not added to the pool's label until their refcounts
	* are incremented, so fi_mos can just be left as false for now.
	*/
	desc = NULL;
	feature.fi_uname = "zhack";
	feature.fi_flags = 0;
	feature.fi_depends = nodeps;
	feature.fi_feature = SPA_FEATURE_NONE;

	optind = 1;
	while ((c = getopt(argc, argv, "+rd:")) != -1) {
	switch (c) {
	case 'r':
	feature.fi_flags \|= ZFEATURE_FLAG_READONLY_COMPAT;
	break;
	case 'd':
	+ if (desc != NULL)
	+ free(desc);
	desc = strdup(optarg);
	break;
	default:
	usage();
	break;
	}
	}

	if (desc == NULL)
	desc = strdup("zhack injected");
	feature.fi_desc = desc;

	argc -= optind;
	argv += optind;

	if (argc < 2) {
	(void) fprintf(stderr, "error: missing feature or pool name\n");
	usage();
	}
	target = argv[0];
	feature.fi_guid = argv[1];

	if (!zfeature_is_valid_guid(feature.fi_guid))
	fatal(NULL, FTAG, "invalid feature guid: %s", feature.fi_guid);

	zhack_spa_open(target, B_FALSE, FTAG, &spa);
	mos = spa->spa_meta_objset;

	if (zfeature_is_supported(feature.fi_guid))
	fatal(spa, FTAG, "'%s' is a real feature, will not enable",
	feature.fi_guid);
	if (0 == zap_contains(mos, spa->spa_feat_desc_obj, feature.fi_guid))
	fatal(spa, FTAG, "feature already enabled: %s",
	feature.fi_guid);

	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	zhack_feature_enable_sync, &feature, 5, ZFS_SPACE_CHECK_NORMAL));

	spa_close(spa, FTAG);

	free(desc);
	}

	static void
	feature_incr_sync(void arg, dmu_tx_t tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	zfeature_info_t *feature = arg;
	uint64_t refcount;

	VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
	feature_sync(spa, feature, refcount + 1, tx);
	spa_history_log_internal(spa, "zhack feature incr", tx,
	"name=%s", feature->fi_guid);
	}

	static void
	feature_decr_sync(void arg, dmu_tx_t tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	zfeature_info_t *feature = arg;
	uint64_t refcount;

	VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount));
	feature_sync(spa, feature, refcount - 1, tx);
	spa_history_log_internal(spa, "zhack feature decr", tx,
	"name=%s", feature->fi_guid);
	}

	static void
	zhack_do_feature_ref(int argc, char **argv)
	{
	int c;
	char *target;
	boolean_t decr = B_FALSE;
	spa_t *spa;
	objset_t *mos;
	zfeature_info_t feature;
	const spa_feature_t nodeps[] = { SPA_FEATURE_NONE };

	/*
	* fi_desc does not matter here because it was written to disk
	* when the feature was enabled, but we need to properly set the
	* feature for read or write based on the information we read off
	* disk later.
	*/
	feature.fi_uname = "zhack";
	feature.fi_flags = 0;
	feature.fi_desc = NULL;
	feature.fi_depends = nodeps;
	feature.fi_feature = SPA_FEATURE_NONE;

	optind = 1;
	while ((c = getopt(argc, argv, "+md")) != -1) {
	switch (c) {
	case 'm':
	feature.fi_flags \|= ZFEATURE_FLAG_MOS;
	break;
	case 'd':
	decr = B_TRUE;
	break;
	default:
	usage();
	break;
	}
	}
	argc -= optind;
	argv += optind;

	if (argc < 2) {
	(void) fprintf(stderr, "error: missing feature or pool name\n");
	usage();
	}
	target = argv[0];
	feature.fi_guid = argv[1];

	if (!zfeature_is_valid_guid(feature.fi_guid))
	fatal(NULL, FTAG, "invalid feature guid: %s", feature.fi_guid);

	zhack_spa_open(target, B_FALSE, FTAG, &spa);
	mos = spa->spa_meta_objset;

	if (zfeature_is_supported(feature.fi_guid)) {
	fatal(spa, FTAG,
	"'%s' is a real feature, will not change refcount",
	feature.fi_guid);
	}

	if (0 == zap_contains(mos, spa->spa_feat_for_read_obj,
	feature.fi_guid)) {
	feature.fi_flags &= ~ZFEATURE_FLAG_READONLY_COMPAT;
	} else if (0 == zap_contains(mos, spa->spa_feat_for_write_obj,
	feature.fi_guid)) {
	feature.fi_flags \|= ZFEATURE_FLAG_READONLY_COMPAT;
	} else {
	fatal(spa, FTAG, "feature is not enabled: %s", feature.fi_guid);
	}

	if (decr) {
	uint64_t count;
	if (feature_get_refcount_from_disk(spa, &feature,
	&count) == 0 && count == 0) {
	fatal(spa, FTAG, "feature refcount already 0: %s",
	feature.fi_guid);
	}
	}

	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	decr ? feature_decr_sync : feature_incr_sync, &feature,
	5, ZFS_SPACE_CHECK_NORMAL));

	spa_close(spa, FTAG);
	}

	static int
	zhack_do_feature(int argc, char **argv)
	{
	char *subcommand;

	argc--;
	argv++;
	if (argc == 0) {
	(void) fprintf(stderr,
	"error: no feature operation specified\n");
	usage();
	}

	subcommand = argv[0];
	if (strcmp(subcommand, "stat") == 0) {
	zhack_do_feature_stat(argc, argv);
	} else if (strcmp(subcommand, "enable") == 0) {
	zhack_do_feature_enable(argc, argv);
	} else if (strcmp(subcommand, "ref") == 0) {
	zhack_do_feature_ref(argc, argv);
	} else {
	(void) fprintf(stderr, "error: unknown subcommand: %s\n",
	subcommand);
	usage();
	}

	return (0);
	}

	static int
	zhack_repair_label_cksum(int argc, char **argv)
	{
	zio_checksum_info_t *ci = &zio_checksum_table[ZIO_CHECKSUM_LABEL];
	const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION,
	ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID };
	boolean_t labels_repaired[VDEV_LABELS] = {0};
	boolean_t repaired = B_FALSE;
	vdev_label_t labels[VDEV_LABELS] = {{{0}}};
	struct stat st;
	int fd;

	abd_init();

	argc -= 1;
	argv += 1;

	if (argc < 1) {
	(void) fprintf(stderr, "error: missing device\n");
	usage();
	}

	if ((fd = open(argv[0], O_RDWR)) == -1)
	fatal(NULL, FTAG, "cannot open '%s': %s", argv[0],
	strerror(errno));

	if (stat(argv[0], &st) != 0)
	fatal(NULL, FTAG, "cannot stat '%s': %s", argv[0],
	strerror(errno));

	for (int l = 0; l < VDEV_LABELS; l++) {
	uint64_t label_offset, offset;
	zio_cksum_t expected_cksum;
	zio_cksum_t actual_cksum;
	zio_cksum_t verifier;
	zio_eck_t *eck;
	nvlist_t *cfg;
	int byteswap;
	uint64_t val;
	ssize_t err;

	vdev_label_t *vl = &labels[l];

	label_offset = vdev_label_offset(st.st_size, l, 0);
	err = pread64(fd, vl, sizeof (vdev_label_t), label_offset);
	if (err == -1) {
	(void) fprintf(stderr, "error: cannot read "
	"label %d: %s\n", l, strerror(errno));
	continue;
	} else if (err != sizeof (vdev_label_t)) {
	(void) fprintf(stderr, "error: bad label %d read size "
	"\n", l);
	continue;
	}

	err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist,
	VDEV_PHYS_SIZE - sizeof (zio_eck_t), &cfg, 0);
	if (err) {
	(void) fprintf(stderr, "error: cannot unpack nvlist "
	"label %d\n", l);
	continue;
	}

	for (int i = 0; i < ARRAY_SIZE(cfg_keys); i++) {
	err = nvlist_lookup_uint64(cfg, cfg_keys[i], &val);
	if (err) {
	(void) fprintf(stderr, "error: label %d: "
	"cannot find nvlist key %s\n",
	l, cfg_keys[i]);
	continue;
	}
	}

	void data = (char )vl + offsetof(vdev_label_t, vl_vdev_phys);
	eck = (zio_eck_t )((char )(data) + VDEV_PHYS_SIZE) - 1;

	offset = label_offset + offsetof(vdev_label_t, vl_vdev_phys);
	ZIO_SET_CHECKSUM(&verifier, offset, 0, 0, 0);

	byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC));
	if (byteswap)
	byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));

	expected_cksum = eck->zec_cksum;
	eck->zec_cksum = verifier;

	abd_t *abd = abd_get_from_buf(data, VDEV_PHYS_SIZE);
	ci->ci_func[byteswap](abd, VDEV_PHYS_SIZE, NULL, &actual_cksum);
	abd_free(abd);

	if (byteswap)
	byteswap_uint64_array(&expected_cksum,
	sizeof (zio_cksum_t));

	if (ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
	continue;

	eck->zec_cksum = actual_cksum;

	err = pwrite64(fd, data, VDEV_PHYS_SIZE, offset);
	if (err == -1) {
	(void) fprintf(stderr, "error: cannot write "
	"label %d: %s\n", l, strerror(errno));
	continue;
	} else if (err != VDEV_PHYS_SIZE) {
	(void) fprintf(stderr, "error: bad write size "
	"label %d\n", l);
	continue;
	}

	fsync(fd);

	labels_repaired[l] = B_TRUE;
	}

	close(fd);

	abd_fini();

	for (int l = 0; l < VDEV_LABELS; l++) {
	(void) printf("label %d: %s\n", l,
	labels_repaired[l] ? "repaired" : "skipped");
	repaired \|= labels_repaired[l];
	}

	if (repaired)
	return (0);

	return (1);
	}

	static int
	zhack_do_label(int argc, char **argv)
	{
	char *subcommand;
	int err;

	argc--;
	argv++;
	if (argc == 0) {
	(void) fprintf(stderr,
	"error: no label operation specified\n");
	usage();
	}

	subcommand = argv[0];
	if (strcmp(subcommand, "repair") == 0) {
	err = zhack_repair_label_cksum(argc, argv);
	} else {
	(void) fprintf(stderr, "error: unknown subcommand: %s\n",
	subcommand);
	usage();
	}

	return (err);
	}

	#define MAX_NUM_PATHS 1024

	int
	main(int argc, char **argv)
	{
	extern void zfs_prop_init(void);

	char *path[MAX_NUM_PATHS];
	const char *subcommand;
	int rv = 0;
	int c;

	g_importargs.path = path;

	dprintf_setup(&argc, argv);
	zfs_prop_init();

	while ((c = getopt(argc, argv, "+c:d:")) != -1) {
	switch (c) {
	case 'c':
	g_importargs.cachefile = optarg;
	break;
	case 'd':
	assert(g_importargs.paths < MAX_NUM_PATHS);
	g_importargs.path[g_importargs.paths++] = optarg;
	break;
	default:
	usage();
	break;
	}
	}

	argc -= optind;
	argv += optind;
	optind = 1;

	if (argc == 0) {
	(void) fprintf(stderr, "error: no command specified\n");
	usage();
	}

	subcommand = argv[0];

	if (strcmp(subcommand, "feature") == 0) {
	rv = zhack_do_feature(argc, argv);
	} else if (strcmp(subcommand, "label") == 0) {
	return (zhack_do_label(argc, argv));
	} else {
	(void) fprintf(stderr, "error: unknown subcommand: %s\n",
	subcommand);
	usage();
	}

	if (!g_readonly && spa_export(g_pool, NULL, B_TRUE, B_FALSE) != 0) {
	fatal(NULL, FTAG, "pool export failed; "
	"changes may not be committed to disk\n");
	}

	kernel_fini();

	return (rv);
	}
	diff --git a/sys/contrib/openzfs/cmd/zinject/translate.c b/sys/contrib/openzfs/cmd/zinject/translate.c
	index cd4264bdc084..c63caedd1583 100644
	--- a/sys/contrib/openzfs/cmd/zinject/translate.c
	+++ b/sys/contrib/openzfs/cmd/zinject/translate.c
	@@ -1,398 +1,398 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	*/

	#include <libzfs.h>

	#include <errno.h>
	#include <fcntl.h>
	#include <stdarg.h>
	#include <stddef.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <sys/file.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/param.h>
	#include <sys/stat.h>

	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dnode.h>
	#include <sys/vdev_impl.h>

	#include <sys/mkdev.h>

	#include "zinject.h"

	static int debug;

	static void
	ziprintf(const char *fmt, ...)
	{
	va_list ap;

	if (!debug)
	return;

	va_start(ap, fmt);
	(void) vprintf(fmt, ap);
	va_end(ap);
	}

	static void
	compress_slashes(const char src, char dest)
	{
	while (*src != '\0') {
	dest = src++;
	while (dest == '/' && src == '/')
	++src;
	++dest;
	}
	*dest = '\0';
	}

	/*
	* Given a full path to a file, translate into a dataset name and a relative
	* path within the dataset. 'dataset' must be at least MAXNAMELEN characters,
	* and 'relpath' must be at least MAXPATHLEN characters. We also pass a stat64
	* buffer, which we need later to get the object ID.
	*/
	static int
	parse_pathname(const char inpath, char dataset, char *relpath,
	struct stat64 *statbuf)
	{
	struct extmnttab mp;
	const char *rel;
	char fullpath[MAXPATHLEN];

	compress_slashes(inpath, fullpath);

	if (fullpath[0] != '/') {
	(void) fprintf(stderr, "invalid object '%s': must be full "
	"path\n", fullpath);
	usage();
	return (-1);
	}

	if (getextmntent(fullpath, &mp, statbuf) != 0) {
	(void) fprintf(stderr, "cannot find mountpoint for '%s'\n",
	fullpath);
	return (-1);
	}

	if (strcmp(mp.mnt_fstype, MNTTYPE_ZFS) != 0) {
	(void) fprintf(stderr, "invalid path '%s': not a ZFS "
	"filesystem\n", fullpath);
	return (-1);
	}

	if (strncmp(fullpath, mp.mnt_mountp, strlen(mp.mnt_mountp)) != 0) {
	(void) fprintf(stderr, "invalid path '%s': mountpoint "
	"doesn't match path\n", fullpath);
	return (-1);
	}

	- (void) strcpy(dataset, mp.mnt_special);
	+ (void) strlcpy(dataset, mp.mnt_special, MAXNAMELEN);

	rel = fullpath + strlen(mp.mnt_mountp);
	if (rel[0] == '/')
	rel++;
	- (void) strcpy(relpath, rel);
	+ (void) strlcpy(relpath, rel, MAXPATHLEN);

	return (0);
	}

	/*
	* Convert from a dataset to a objset id. Note that
	* we grab the object number from the inode number.
	*/
	static int
	object_from_path(const char dataset, uint64_t object, zinject_record_t record)
	{
	zfs_handle_t *zhp;

	if ((zhp = zfs_open(g_zfs, dataset, ZFS_TYPE_DATASET)) == NULL)
	return (-1);

	record->zi_objset = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	record->zi_object = object;

	zfs_close(zhp);

	return (0);
	}

	/*
	* Initialize the range based on the type, level, and range given.
	*/
	static int
	initialize_range(err_type_t type, int level, char *range,
	zinject_record_t *record)
	{
	/*
	* Determine the numeric range from the string.
	*/
	if (range == NULL) {
	/*
	* If range is unspecified, set the range to [0,-1], which
	* indicates that the whole object should be treated as an
	* error.
	*/
	record->zi_start = 0;
	record->zi_end = -1ULL;
	} else {
	char *end;

	/* XXX add support for suffixes */
	record->zi_start = strtoull(range, &end, 10);


	if (*end == '\0')
	record->zi_end = record->zi_start + 1;
	else if (*end == ',')
	record->zi_end = strtoull(end + 1, &end, 10);

	if (*end != '\0') {
	(void) fprintf(stderr, "invalid range '%s': must be "
	"a numeric range of the form 'start[,end]'\n",
	range);
	return (-1);
	}
	}

	switch (type) {
	default:
	break;

	case TYPE_DATA:
	break;

	case TYPE_DNODE:
	/*
	* If this is a request to inject faults into the dnode, then we
	* must translate the current (objset,object) pair into an
	* offset within the metadnode for the objset. Specifying any
	* kind of range with type 'dnode' is illegal.
	*/
	if (range != NULL) {
	(void) fprintf(stderr, "range cannot be specified when "
	"type is 'dnode'\n");
	return (-1);
	}

	record->zi_start = record->zi_object * sizeof (dnode_phys_t);
	record->zi_end = record->zi_start + sizeof (dnode_phys_t);
	record->zi_object = 0;
	break;
	}

	record->zi_level = level;

	return (0);
	}

	int
	translate_record(err_type_t type, const char object, const char range,
	int level, zinject_record_t record, char poolname, char *dataset)
	{
	char path[MAXPATHLEN];
	char *slash;
	struct stat64 statbuf;
	int ret = -1;

	debug = (getenv("ZINJECT_DEBUG") != NULL);

	ziprintf("translating: %s\n", object);

	if (MOS_TYPE(type)) {
	/*
	* MOS objects are treated specially.
	*/
	switch (type) {
	default:
	break;
	case TYPE_MOS:
	record->zi_type = 0;
	break;
	case TYPE_MOSDIR:
	record->zi_type = DMU_OT_OBJECT_DIRECTORY;
	break;
	case TYPE_METASLAB:
	record->zi_type = DMU_OT_OBJECT_ARRAY;
	break;
	case TYPE_CONFIG:
	record->zi_type = DMU_OT_PACKED_NVLIST;
	break;
	case TYPE_BPOBJ:
	record->zi_type = DMU_OT_BPOBJ;
	break;
	case TYPE_SPACEMAP:
	record->zi_type = DMU_OT_SPACE_MAP;
	break;
	case TYPE_ERRLOG:
	record->zi_type = DMU_OT_ERROR_LOG;
	break;
	}

	dataset[0] = '\0';
	- (void) strcpy(poolname, object);
	+ (void) strlcpy(poolname, object, MAXNAMELEN);
	return (0);
	}

	/*
	* Convert a full path into a (dataset, file) pair.
	*/
	if (parse_pathname(object, dataset, path, &statbuf) != 0)
	goto err;

	ziprintf(" dataset: %s\n", dataset);
	ziprintf(" path: %s\n", path);

	/*
	* Convert (dataset, file) into (objset, object)
	*/
	if (object_from_path(dataset, statbuf.st_ino, record) != 0)
	goto err;

	ziprintf("raw objset: %llu\n", record->zi_objset);
	ziprintf("raw object: %llu\n", record->zi_object);

	/*
	* For the given object, initialize the range in bytes
	*/
	if (initialize_range(type, level, (char *)range, record) != 0)
	goto err;

	ziprintf(" objset: %llu\n", record->zi_objset);
	ziprintf(" object: %llu\n", record->zi_object);
	if (record->zi_start == 0 &&
	record->zi_end == -1ULL)
	ziprintf(" range: all\n");
	else
	ziprintf(" range: [%llu, %llu]\n", record->zi_start,
	record->zi_end);

	/*
	* Copy the pool name
	*/
	- (void) strcpy(poolname, dataset);
	+ (void) strlcpy(poolname, dataset, MAXNAMELEN);
	if ((slash = strchr(poolname, '/')) != NULL)
	*slash = '\0';

	ret = 0;

	err:
	return (ret);
	}

	int
	translate_raw(const char str, zinject_record_t record)
	{
	/*
	* A raw bookmark of the form objset:object:level:blkid, where each
	* number is a hexadecimal value.
	*/
	if (sscanf(str, "%llx:%llx:%x:%llx", (u_longlong_t *)&record->zi_objset,
	(u_longlong_t *)&record->zi_object, &record->zi_level,
	(u_longlong_t *)&record->zi_start) != 4) {
	(void) fprintf(stderr, "bad raw spec '%s': must be of the form "
	"'objset:object:level:blkid'\n", str);
	return (-1);
	}

	record->zi_end = record->zi_start;

	return (0);
	}

	int
	translate_device(const char pool, const char device, err_type_t label_type,
	zinject_record_t *record)
	{
	char *end;
	zpool_handle_t *zhp;
	nvlist_t *tgt;
	boolean_t isspare, iscache;

	/*
	* Given a device name or GUID, create an appropriate injection record
	* with zi_guid set.
	*/
	if ((zhp = zpool_open(g_zfs, pool)) == NULL)
	return (-1);

	record->zi_guid = strtoull(device, &end, 0);
	if (record->zi_guid == 0 \|\| *end != '\0') {
	tgt = zpool_find_vdev(zhp, device, &isspare, &iscache, NULL);

	if (tgt == NULL) {
	(void) fprintf(stderr, "cannot find device '%s' in "
	"pool '%s'\n", device, pool);
	zpool_close(zhp);
	return (-1);
	}

	verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
	&record->zi_guid) == 0);
	}

	/*
	* Device faults can take on three different forms:
	* 1). delayed or hanging I/O
	* 2). zfs label faults
	* 3). generic disk faults
	*/
	if (record->zi_timer != 0) {
	record->zi_cmd = ZINJECT_DELAY_IO;
	} else if (label_type != TYPE_INVAL) {
	record->zi_cmd = ZINJECT_LABEL_FAULT;
	} else {
	record->zi_cmd = ZINJECT_DEVICE_FAULT;
	}

	switch (label_type) {
	default:
	break;
	case TYPE_LABEL_UBERBLOCK:
	record->zi_start = offsetof(vdev_label_t, vl_uberblock[0]);
	record->zi_end = record->zi_start + VDEV_UBERBLOCK_RING - 1;
	break;
	case TYPE_LABEL_NVLIST:
	record->zi_start = offsetof(vdev_label_t, vl_vdev_phys);
	record->zi_end = record->zi_start + VDEV_PHYS_SIZE - 1;
	break;
	case TYPE_LABEL_PAD1:
	record->zi_start = offsetof(vdev_label_t, vl_pad1);
	record->zi_end = record->zi_start + VDEV_PAD_SIZE - 1;
	break;
	case TYPE_LABEL_PAD2:
	record->zi_start = offsetof(vdev_label_t, vl_be);
	record->zi_end = record->zi_start + VDEV_PAD_SIZE - 1;
	break;
	}
	zpool_close(zhp);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_main.c b/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	index b5b0beef5324..39ea615f6319 100644
	--- a/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	+++ b/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	@@ -1,10958 +1,10966 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2012 by Frederik Wessels. All rights reserved.
	* Copyright (c) 2012 by Cyril Plisko. All rights reserved.
	* Copyright (c) 2013 by Prasad Joshi (sTec). All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
	* Copyright (c) 2017 Datto Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	* Copyright (c) 2021, Klara Inc.
	* Copyright [2021] Hewlett Packard Enterprise Development LP
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <dirent.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <getopt.h>
	#include <libgen.h>
	#include <libintl.h>
	#include <libuutil.h>
	#include <locale.h>
	#include <pthread.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <time.h>
	#include <unistd.h>
	#include <pwd.h>
	#include <zone.h>
	#include <sys/wait.h>
	#include <zfs_prop.h>
	#include <sys/fs/zfs.h>
	#include <sys/stat.h>
	#include <sys/systeminfo.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/fm/util.h>
	#include <sys/fm/protocol.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/mount.h>
	#include <sys/sysmacros.h>

	#include <math.h>

	#include <libzfs.h>
	#include <libzutil.h>

	#include "zpool_util.h"
	#include "zfs_comutil.h"
	#include "zfeature_common.h"

	#include "statcommon.h"

	libzfs_handle_t *g_zfs;

	static int zpool_do_create(int, char **);
	static int zpool_do_destroy(int, char **);

	static int zpool_do_add(int, char **);
	static int zpool_do_remove(int, char **);
	static int zpool_do_labelclear(int, char **);

	static int zpool_do_checkpoint(int, char **);

	static int zpool_do_list(int, char **);
	static int zpool_do_iostat(int, char **);
	static int zpool_do_status(int, char **);

	static int zpool_do_online(int, char **);
	static int zpool_do_offline(int, char **);
	static int zpool_do_clear(int, char **);
	static int zpool_do_reopen(int, char **);

	static int zpool_do_reguid(int, char **);

	static int zpool_do_attach(int, char **);
	static int zpool_do_detach(int, char **);
	static int zpool_do_replace(int, char **);
	static int zpool_do_split(int, char **);

	static int zpool_do_initialize(int, char **);
	static int zpool_do_scrub(int, char **);
	static int zpool_do_resilver(int, char **);
	static int zpool_do_trim(int, char **);

	static int zpool_do_import(int, char **);
	static int zpool_do_export(int, char **);

	static int zpool_do_upgrade(int, char **);

	static int zpool_do_history(int, char **);
	static int zpool_do_events(int, char **);

	static int zpool_do_get(int, char **);
	static int zpool_do_set(int, char **);

	static int zpool_do_sync(int, char **);

	static int zpool_do_version(int, char **);

	static int zpool_do_wait(int, char **);

	static zpool_compat_status_t zpool_do_load_compat(
	const char , boolean_t );

	/*
	* These libumem hooks provide a reasonable set of defaults for the allocator's
	* debugging facilities.
	*/

	#ifdef DEBUG
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

	const char *
	_umem_logging_init(void)
	{
	return ("fail,contents"); /* $UMEM_LOGGING setting */
	}
	#endif

	typedef enum {
	HELP_ADD,
	HELP_ATTACH,
	HELP_CLEAR,
	HELP_CREATE,
	HELP_CHECKPOINT,
	HELP_DESTROY,
	HELP_DETACH,
	HELP_EXPORT,
	HELP_HISTORY,
	HELP_IMPORT,
	HELP_IOSTAT,
	HELP_LABELCLEAR,
	HELP_LIST,
	HELP_OFFLINE,
	HELP_ONLINE,
	HELP_REPLACE,
	HELP_REMOVE,
	HELP_INITIALIZE,
	HELP_SCRUB,
	HELP_RESILVER,
	HELP_TRIM,
	HELP_STATUS,
	HELP_UPGRADE,
	HELP_EVENTS,
	HELP_GET,
	HELP_SET,
	HELP_SPLIT,
	HELP_SYNC,
	HELP_REGUID,
	HELP_REOPEN,
	HELP_VERSION,
	HELP_WAIT
	} zpool_help_t;


	/*
	* Flags for stats to display with "zpool iostats"
	*/
	enum iostat_type {
	IOS_DEFAULT = 0,
	IOS_LATENCY = 1,
	IOS_QUEUES = 2,
	IOS_L_HISTO = 3,
	IOS_RQ_HISTO = 4,
	IOS_COUNT, /* always last element */
	};

	/* iostat_type entries as bitmasks */
	#define IOS_DEFAULT_M (1ULL << IOS_DEFAULT)
	#define IOS_LATENCY_M (1ULL << IOS_LATENCY)
	#define IOS_QUEUES_M (1ULL << IOS_QUEUES)
	#define IOS_L_HISTO_M (1ULL << IOS_L_HISTO)
	#define IOS_RQ_HISTO_M (1ULL << IOS_RQ_HISTO)

	/* Mask of all the histo bits */
	#define IOS_ANYHISTO_M (IOS_L_HISTO_M \| IOS_RQ_HISTO_M)

	/*
	* Lookup table for iostat flags to nvlist names. Basically a list
	* of all the nvlists a flag requires. Also specifies the order in
	* which data gets printed in zpool iostat.
	*/
	static const char *vsx_type_to_nvlist[IOS_COUNT][15] = {
	[IOS_L_HISTO] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	NULL},
	[IOS_LATENCY] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	NULL},
	[IOS_QUEUES] = {
	ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE,
	NULL},
	[IOS_RQ_HISTO] = {
	ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
	ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
	ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO,
	ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO,
	NULL},
	};


	/*
	* Given a cb->cb_flags with a histogram bit set, return the iostat_type.
	* Right now, only one histo bit is ever set at one time, so we can
	* just do a highbit64(a)
	*/
	#define IOS_HISTO_IDX(a) (highbit64(a & IOS_ANYHISTO_M) - 1)

	typedef struct zpool_command {
	const char *name;
	int (func)(int, char *);
	zpool_help_t usage;
	} zpool_command_t;

	/*
	* Master command table. Each ZFS command has a name, associated function, and
	* usage message. The usage messages need to be internationalized, so we have
	* to have a function to return the usage message based on a command index.
	*
	* These commands are organized according to how they are displayed in the usage
	* message. An empty command (one with a NULL name) indicates an empty line in
	* the generic usage message.
	*/
	static zpool_command_t command_table[] = {
	{ "version", zpool_do_version, HELP_VERSION },
	{ NULL },
	{ "create", zpool_do_create, HELP_CREATE },
	{ "destroy", zpool_do_destroy, HELP_DESTROY },
	{ NULL },
	{ "add", zpool_do_add, HELP_ADD },
	{ "remove", zpool_do_remove, HELP_REMOVE },
	{ NULL },
	{ "labelclear", zpool_do_labelclear, HELP_LABELCLEAR },
	{ NULL },
	{ "checkpoint", zpool_do_checkpoint, HELP_CHECKPOINT },
	{ NULL },
	{ "list", zpool_do_list, HELP_LIST },
	{ "iostat", zpool_do_iostat, HELP_IOSTAT },
	{ "status", zpool_do_status, HELP_STATUS },
	{ NULL },
	{ "online", zpool_do_online, HELP_ONLINE },
	{ "offline", zpool_do_offline, HELP_OFFLINE },
	{ "clear", zpool_do_clear, HELP_CLEAR },
	{ "reopen", zpool_do_reopen, HELP_REOPEN },
	{ NULL },
	{ "attach", zpool_do_attach, HELP_ATTACH },
	{ "detach", zpool_do_detach, HELP_DETACH },
	{ "replace", zpool_do_replace, HELP_REPLACE },
	{ "split", zpool_do_split, HELP_SPLIT },
	{ NULL },
	{ "initialize", zpool_do_initialize, HELP_INITIALIZE },
	{ "resilver", zpool_do_resilver, HELP_RESILVER },
	{ "scrub", zpool_do_scrub, HELP_SCRUB },
	{ "trim", zpool_do_trim, HELP_TRIM },
	{ NULL },
	{ "import", zpool_do_import, HELP_IMPORT },
	{ "export", zpool_do_export, HELP_EXPORT },
	{ "upgrade", zpool_do_upgrade, HELP_UPGRADE },
	{ "reguid", zpool_do_reguid, HELP_REGUID },
	{ NULL },
	{ "history", zpool_do_history, HELP_HISTORY },
	{ "events", zpool_do_events, HELP_EVENTS },
	{ NULL },
	{ "get", zpool_do_get, HELP_GET },
	{ "set", zpool_do_set, HELP_SET },
	{ "sync", zpool_do_sync, HELP_SYNC },
	{ NULL },
	{ "wait", zpool_do_wait, HELP_WAIT },
	};

	#define NCOMMAND (ARRAY_SIZE(command_table))

	#define VDEV_ALLOC_CLASS_LOGS "logs"

	static zpool_command_t *current_command;
	static zfs_type_t current_prop_type = (ZFS_TYPE_POOL \| ZFS_TYPE_VDEV);
	static char history_str[HIS_MAX_RECORD_LEN];
	static boolean_t log_history = B_TRUE;
	static uint_t timestamp_fmt = NODATE;

	static const char *
	get_usage(zpool_help_t idx)
	{
	switch (idx) {
	case HELP_ADD:
	return (gettext("\tadd [-fgLnP] [-o property=value] "
	"<pool> <vdev> ...\n"));
	case HELP_ATTACH:
	return (gettext("\tattach [-fsw] [-o property=value] "
	"<pool> <device> <new-device>\n"));
	case HELP_CLEAR:
	return (gettext("\tclear [-nF] <pool> [device]\n"));
	case HELP_CREATE:
	return (gettext("\tcreate [-fnd] [-o property=value] ... \n"
	"\t [-O file-system-property=value] ... \n"
	"\t [-m mountpoint] [-R root] <pool> <vdev> ...\n"));
	case HELP_CHECKPOINT:
	return (gettext("\tcheckpoint [-d [-w]] <pool> ...\n"));
	case HELP_DESTROY:
	return (gettext("\tdestroy [-f] <pool>\n"));
	case HELP_DETACH:
	return (gettext("\tdetach <pool> <device>\n"));
	case HELP_EXPORT:
	return (gettext("\texport [-af] <pool> ...\n"));
	case HELP_HISTORY:
	return (gettext("\thistory [-il] [<pool>] ...\n"));
	case HELP_IMPORT:
	return (gettext("\timport [-d dir] [-D]\n"
	"\timport [-o mntopts] [-o property=value] ... \n"
	"\t [-d dir \| -c cachefile] [-D] [-l] [-f] [-m] [-N] "
	"[-R root] [-F [-n]] -a\n"
	"\timport [-o mntopts] [-o property=value] ... \n"
	"\t [-d dir \| -c cachefile] [-D] [-l] [-f] [-m] [-N] "
	"[-R root] [-F [-n]]\n"
	"\t [--rewind-to-checkpoint] <pool \| id> [newpool]\n"));
	case HELP_IOSTAT:
	return (gettext("\tiostat [[[-c [script1,script2,...]"
	"[-lq]]\|[-rw]] [-T d \| u] [-ghHLpPvy]\n"
	"\t [[pool ...]\|[pool vdev ...]\|[vdev ...]]"
	" [[-n] interval [count]]\n"));
	case HELP_LABELCLEAR:
	return (gettext("\tlabelclear [-f] <vdev>\n"));
	case HELP_LIST:
	return (gettext("\tlist [-gHLpPv] [-o property[,...]] "
	"[-T d\|u] [pool] ... \n"
	"\t [interval [count]]\n"));
	case HELP_OFFLINE:
	return (gettext("\toffline [-f] [-t] <pool> <device> ...\n"));
	case HELP_ONLINE:
	return (gettext("\tonline [-e] <pool> <device> ...\n"));
	case HELP_REPLACE:
	return (gettext("\treplace [-fsw] [-o property=value] "
	"<pool> <device> [new-device]\n"));
	case HELP_REMOVE:
	return (gettext("\tremove [-npsw] <pool> <device> ...\n"));
	case HELP_REOPEN:
	return (gettext("\treopen [-n] <pool>\n"));
	case HELP_INITIALIZE:
	return (gettext("\tinitialize [-c \| -s] [-w] <pool> "
	"[<device> ...]\n"));
	case HELP_SCRUB:
	return (gettext("\tscrub [-s \| -p] [-w] <pool> ...\n"));
	case HELP_RESILVER:
	return (gettext("\tresilver <pool> ...\n"));
	case HELP_TRIM:
	return (gettext("\ttrim [-dw] [-r <rate>] [-c \| -s] <pool> "
	"[<device> ...]\n"));
	case HELP_STATUS:
	return (gettext("\tstatus [-c [script1,script2,...]] "
	"[-igLpPstvxD] [-T d\|u] [pool] ... \n"
	"\t [interval [count]]\n"));
	case HELP_UPGRADE:
	return (gettext("\tupgrade\n"
	"\tupgrade -v\n"
	"\tupgrade [-V version] <-a \| pool ...>\n"));
	case HELP_EVENTS:
	return (gettext("\tevents [-vHf [pool] \| -c]\n"));
	case HELP_GET:
	return (gettext("\tget [-Hp] [-o \"all\" \| field[,...]] "
	"<\"all\" \| property[,...]> <pool> ...\n"));
	case HELP_SET:
	return (gettext("\tset <property=value> <pool> \n"));
	case HELP_SPLIT:
	return (gettext("\tsplit [-gLnPl] [-R altroot] [-o mntopts]\n"
	"\t [-o property=value] <pool> <newpool> "
	"[<device> ...]\n"));
	case HELP_REGUID:
	return (gettext("\treguid <pool>\n"));
	case HELP_SYNC:
	return (gettext("\tsync [pool] ...\n"));
	case HELP_VERSION:
	return (gettext("\tversion\n"));
	case HELP_WAIT:
	return (gettext("\twait [-Hp] [-T d\|u] [-t <activity>[,...]] "
	"<pool> [interval]\n"));
	default:
	__builtin_unreachable();
	}
	}

	static void
	zpool_collect_leaves(zpool_handle_t zhp, nvlist_t nvroot, nvlist_t *res)
	{
	uint_t children = 0;
	nvlist_t **child;
	uint_t i;

	(void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children);

	if (children == 0) {
	char *path = zpool_vdev_name(g_zfs, zhp, nvroot,
	VDEV_NAME_PATH);

	if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 &&
	strcmp(path, VDEV_TYPE_HOLE) != 0)
	fnvlist_add_boolean(res, path);

	free(path);
	return;
	}

	for (i = 0; i < children; i++) {
	zpool_collect_leaves(zhp, child[i], res);
	}
	}

	/*
	* Callback routine that will print out a pool property value.
	*/
	static int
	print_pool_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-19s ", zpool_prop_to_name(prop));

	if (zpool_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, " YES ");

	if (zpool_prop_values(prop) == NULL)
	(void) fprintf(fp, "-\n");
	else
	(void) fprintf(fp, "%s\n", zpool_prop_values(prop));

	return (ZPROP_CONT);
	}

	/*
	* Callback routine that will print out a vdev property value.
	*/
	static int
	print_vdev_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-19s ", vdev_prop_to_name(prop));

	if (vdev_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, " YES ");

	if (vdev_prop_values(prop) == NULL)
	(void) fprintf(fp, "-\n");
	else
	(void) fprintf(fp, "%s\n", vdev_prop_values(prop));

	return (ZPROP_CONT);
	}

	/*
	* Display usage message. If we're inside a command, display only the usage for
	* that command. Otherwise, iterate over the entire command table and display
	* a complete usage message.
	*/
	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	FILE *fp = requested ? stdout : stderr;

	if (current_command == NULL) {
	int i;

	(void) fprintf(fp, gettext("usage: zpool command args ...\n"));
	(void) fprintf(fp,
	gettext("where 'command' is one of the following:\n\n"));

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	(void) fprintf(fp, "\n");
	else
	(void) fprintf(fp, "%s",
	get_usage(command_table[i].usage));
	}
	} else {
	(void) fprintf(fp, gettext("usage:\n"));
	(void) fprintf(fp, "%s", get_usage(current_command->usage));
	}

	if (current_command != NULL &&
	current_prop_type != (ZFS_TYPE_POOL \| ZFS_TYPE_VDEV) &&
	((strcmp(current_command->name, "set") == 0) \|\|
	(strcmp(current_command->name, "get") == 0) \|\|
	(strcmp(current_command->name, "list") == 0))) {

	(void) fprintf(fp,
	gettext("\nthe following properties are supported:\n"));

	(void) fprintf(fp, "\n\t%-19s %s %s\n\n",
	"PROPERTY", "EDIT", "VALUES");

	/* Iterate over all properties */
	if (current_prop_type == ZFS_TYPE_POOL) {
	(void) zprop_iter(print_pool_prop_cb, fp, B_FALSE,
	B_TRUE, current_prop_type);

	(void) fprintf(fp, "\t%-19s ", "feature@...");
	(void) fprintf(fp, "YES "
	"disabled \| enabled \| active\n");

	(void) fprintf(fp, gettext("\nThe feature@ properties "
	"must be appended with a feature name.\n"
	"See zpool-features(7).\n"));
	} else if (current_prop_type == ZFS_TYPE_VDEV) {
	(void) zprop_iter(print_vdev_prop_cb, fp, B_FALSE,
	B_TRUE, current_prop_type);
	}
	}

	/*
	* See comments at end of main().
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	exit(requested ? 0 : 2);
	}

	/*
	* zpool initialize [-c \| -s] [-w] <pool> [<vdev> ...]
	* Initialize all unused blocks in the specified vdevs, or all vdevs in the pool
	* if none specified.
	*
	* -c Cancel. Ends active initializing.
	* -s Suspend. Initializing can then be restarted with no flags.
	* -w Wait. Blocks until initializing has completed.
	*/
	int
	zpool_do_initialize(int argc, char **argv)
	{
	int c;
	char *poolname;
	zpool_handle_t *zhp;
	nvlist_t *vdevs;
	int err = 0;
	boolean_t wait = B_FALSE;

	struct option long_options[] = {
	{"cancel", no_argument, NULL, 'c'},
	{"suspend", no_argument, NULL, 's'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	pool_initialize_func_t cmd_type = POOL_INITIALIZE_START;
	while ((c = getopt_long(argc, argv, "csw", long_options, NULL)) != -1) {
	switch (c) {
	case 'c':
	if (cmd_type != POOL_INITIALIZE_START &&
	cmd_type != POOL_INITIALIZE_CANCEL) {
	(void) fprintf(stderr, gettext("-c cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_INITIALIZE_CANCEL;
	break;
	case 's':
	if (cmd_type != POOL_INITIALIZE_START &&
	cmd_type != POOL_INITIALIZE_SUSPEND) {
	(void) fprintf(stderr, gettext("-s cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_INITIALIZE_SUSPEND;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	return (-1);
	}

	if (wait && (cmd_type != POOL_INITIALIZE_START)) {
	(void) fprintf(stderr, gettext("-w cannot be used with -c or "
	"-s\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	zhp = zpool_open(g_zfs, poolname);
	if (zhp == NULL)
	return (-1);

	vdevs = fnvlist_alloc();
	if (argc == 1) {
	/* no individual leaf vdevs specified, so add them all */
	nvlist_t *config = zpool_get_config(zhp, NULL);
	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	zpool_collect_leaves(zhp, nvroot, vdevs);
	} else {
	for (int i = 1; i < argc; i++) {
	fnvlist_add_boolean(vdevs, argv[i]);
	}
	}

	if (wait)
	err = zpool_initialize_wait(zhp, cmd_type, vdevs);
	else
	err = zpool_initialize(zhp, cmd_type, vdevs);

	fnvlist_free(vdevs);
	zpool_close(zhp);

	return (err);
	}

	/*
	* print a pool vdev config for dry runs
	*/
	static void
	print_vdev_tree(zpool_handle_t zhp, const char name, nvlist_t *nv, int indent,
	const char *match, int name_flags)
	{
	nvlist_t **child;
	uint_t c, children;
	char *vname;
	boolean_t printed = B_FALSE;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0) {
	if (name != NULL)
	(void) printf("\t%*s%s\n", indent, "", name);
	return;
	}

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE, is_hole = B_FALSE;
	char class = (char )"";

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&is_hole);

	if (is_hole == B_TRUE) {
	continue;
	}

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);
	if (is_log)
	class = (char *)VDEV_ALLOC_BIAS_LOG;
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &class);
	if (strcmp(match, class) != 0)
	continue;

	if (!printed && name != NULL) {
	(void) printf("\t%*s%s\n", indent, "", name);
	printed = B_TRUE;
	}
	vname = zpool_vdev_name(g_zfs, zhp, child[c], name_flags);
	print_vdev_tree(zhp, vname, child[c], indent + 2, "",
	name_flags);
	free(vname);
	}
	}

	/*
	* Print the list of l2cache devices for dry runs.
	*/
	static void
	print_cache_list(nvlist_t *nv, int indent)
	{
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0 && children > 0) {
	(void) printf("\t%*s%s\n", indent, "", "cache");
	} else {
	return;
	}
	for (c = 0; c < children; c++) {
	char *vname;

	vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
	(void) printf("\t%*s%s\n", indent + 2, "", vname);
	free(vname);
	}
	}

	/*
	* Print the list of spares for dry runs.
	*/
	static void
	print_spare_list(nvlist_t *nv, int indent)
	{
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0 && children > 0) {
	(void) printf("\t%*s%s\n", indent, "", "spares");
	} else {
	return;
	}
	for (c = 0; c < children; c++) {
	char *vname;

	vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
	(void) printf("\t%*s%s\n", indent + 2, "", vname);
	free(vname);
	}
	}

	static boolean_t
	prop_list_contains_feature(nvlist_t *proplist)
	{
	nvpair_t *nvp;
	for (nvp = nvlist_next_nvpair(proplist, NULL); NULL != nvp;
	nvp = nvlist_next_nvpair(proplist, nvp)) {
	if (zpool_prop_feature(nvpair_name(nvp)))
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Add a property pair (name, string-value) into a property nvlist.
	*/
	static int
	add_prop_list(const char propname, const char propval, nvlist_t **props,
	boolean_t poolprop)
	{
	zpool_prop_t prop = ZPOOL_PROP_INVAL;
	nvlist_t *proplist;
	const char *normnm;
	char *strval;

	if (*props == NULL &&
	nvlist_alloc(props, NV_UNIQUE_NAME, 0) != 0) {
	(void) fprintf(stderr,
	gettext("internal error: out of memory\n"));
	return (1);
	}

	proplist = *props;

	if (poolprop) {
	const char *vname = zpool_prop_to_name(ZPOOL_PROP_VERSION);
	const char *cname =
	zpool_prop_to_name(ZPOOL_PROP_COMPATIBILITY);

	if ((prop = zpool_name_to_prop(propname)) == ZPOOL_PROP_INVAL &&
	(!zpool_prop_feature(propname) &&
	!zpool_prop_vdev(propname))) {
	(void) fprintf(stderr, gettext("property '%s' is "
	"not a valid pool or vdev property\n"), propname);
	return (2);
	}

	/*
	* feature@ properties and version should not be specified
	* at the same time.
	*/
	if ((prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname) &&
	nvlist_exists(proplist, vname)) \|\|
	(prop == ZPOOL_PROP_VERSION &&
	prop_list_contains_feature(proplist))) {
	(void) fprintf(stderr, gettext("'feature@' and "
	"'version' properties cannot be specified "
	"together\n"));
	return (2);
	}

	/*
	* if version is specified, only "legacy" compatibility
	* may be requested
	*/
	if ((prop == ZPOOL_PROP_COMPATIBILITY &&
	strcmp(propval, ZPOOL_COMPAT_LEGACY) != 0 &&
	nvlist_exists(proplist, vname)) \|\|
	(prop == ZPOOL_PROP_VERSION &&
	nvlist_exists(proplist, cname) &&
	strcmp(fnvlist_lookup_string(proplist, cname),
	ZPOOL_COMPAT_LEGACY) != 0)) {
	(void) fprintf(stderr, gettext("when 'version' is "
	"specified, the 'compatibility' feature may only "
	"be set to '" ZPOOL_COMPAT_LEGACY "'\n"));
	return (2);
	}

	if (zpool_prop_feature(propname) \|\| zpool_prop_vdev(propname))
	normnm = propname;
	else
	normnm = zpool_prop_to_name(prop);
	} else {
	zfs_prop_t fsprop = zfs_name_to_prop(propname);

	if (zfs_prop_valid_for_type(fsprop, ZFS_TYPE_FILESYSTEM,
	B_FALSE)) {
	normnm = zfs_prop_to_name(fsprop);
	} else if (zfs_prop_user(propname) \|\|
	zfs_prop_userquota(propname)) {
	normnm = propname;
	} else {
	(void) fprintf(stderr, gettext("property '%s' is "
	"not a valid filesystem property\n"), propname);
	return (2);
	}
	}

	if (nvlist_lookup_string(proplist, normnm, &strval) == 0 &&
	prop != ZPOOL_PROP_CACHEFILE) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (2);
	}

	if (nvlist_add_string(proplist, normnm, propval) != 0) {
	(void) fprintf(stderr, gettext("internal "
	"error: out of memory\n"));
	return (1);
	}

	return (0);
	}

	/*
	* Set a default property pair (name, string-value) in a property nvlist
	*/
	static int
	add_prop_list_default(const char propname, const char propval,
	nvlist_t **props)
	{
	char *pval;

	if (nvlist_lookup_string(*props, propname, &pval) == 0)
	return (0);

	return (add_prop_list(propname, propval, props, B_TRUE));
	}

	/*
	* zpool add [-fgLnP] [-o property=value] <pool> <vdev> ...
	*
	* -f Force addition of devices, even if they appear in use
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -n Do not add the devices, but display the resulting layout if
	* they were to be added.
	* -o Set property=value.
	* -P Display full path for vdev name.
	*
	* Adds the given vdevs to 'pool'. As with create, the bulk of this work is
	* handled by make_root_vdev(), which constructs the nvlist needed to pass to
	* libzfs.
	*/
	int
	zpool_do_add(int argc, char **argv)
	{
	boolean_t force = B_FALSE;
	boolean_t dryrun = B_FALSE;
	int name_flags = 0;
	int c;
	nvlist_t *nvroot;
	char *poolname;
	int ret;
	zpool_handle_t *zhp;
	nvlist_t *config;
	nvlist_t *props = NULL;
	char *propval;

	/* check options */
	while ((c = getopt(argc, argv, "fgLno:P")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'g':
	name_flags \|= VDEV_NAME_GUID;
	break;
	case 'L':
	name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	usage(B_FALSE);
	}
	*propval = '\0';
	propval++;

	if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) \|\|
	(add_prop_list(optarg, propval, &props, B_TRUE)))
	usage(B_FALSE);
	break;
	case 'P':
	name_flags \|= VDEV_NAME_PATH;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing vdev specification\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	argc--;
	argv++;

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
	(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
	poolname);
	zpool_close(zhp);
	return (1);
	}

	/* unless manually specified use "ashift" pool property (if set) */
	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
	int intval;
	zprop_source_t src;
	char strval[ZPOOL_MAXPROPLEN];

	intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
	if (src != ZPROP_SRC_DEFAULT) {
	(void) sprintf(strval, "%" PRId32, intval);
	verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
	&props, B_TRUE) == 0);
	}
	}

	/* pass off to make_root_vdev for processing */
	nvroot = make_root_vdev(zhp, props, force, !force, B_FALSE, dryrun,
	argc, argv);
	if (nvroot == NULL) {
	zpool_close(zhp);
	return (1);
	}

	if (dryrun) {
	nvlist_t *poolnvroot;
	nvlist_t l2child, sparechild;
	uint_t l2children, sparechildren, c;
	char *vname;
	boolean_t hadcache = B_FALSE, hadspare = B_FALSE;

	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&poolnvroot) == 0);

	(void) printf(gettext("would update '%s' to the following "
	"configuration:\n\n"), zpool_get_name(zhp));

	/* print original main pool and new tree */
	print_vdev_tree(zhp, poolname, poolnvroot, 0, "",
	name_flags \| VDEV_NAME_TYPE_ID);
	print_vdev_tree(zhp, NULL, nvroot, 0, "", name_flags);

	/* print other classes: 'dedup', 'special', and 'log' */
	if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_DEDUP)) {
	print_vdev_tree(zhp, "dedup", poolnvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_DEDUP)) {
	print_vdev_tree(zhp, "dedup", nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	}

	if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
	print_vdev_tree(zhp, "special", poolnvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
	print_vdev_tree(zhp, "special", nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	}

	if (num_logs(poolnvroot) > 0) {
	print_vdev_tree(zhp, "logs", poolnvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	} else if (num_logs(nvroot) > 0) {
	print_vdev_tree(zhp, "logs", nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	}

	/* Do the same for the caches */
	if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_L2CACHE,
	&l2child, &l2children) == 0 && l2children) {
	hadcache = B_TRUE;
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < l2children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	l2child[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2child, &l2children) == 0 && l2children) {
	if (!hadcache)
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < l2children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	l2child[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	/* And finally the spares */
	if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_SPARES,
	&sparechild, &sparechildren) == 0 && sparechildren > 0) {
	hadspare = B_TRUE;
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < sparechildren; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	sparechild[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&sparechild, &sparechildren) == 0 && sparechildren > 0) {
	if (!hadspare)
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < sparechildren; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	sparechild[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}

	ret = 0;
	} else {
	ret = (zpool_add(zhp, nvroot) != 0);
	}

	nvlist_free(props);
	nvlist_free(nvroot);
	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool remove [-npsw] <pool> <vdev> ...
	*
	* Removes the given vdev from the pool.
	*/
	int
	zpool_do_remove(int argc, char **argv)
	{
	char *poolname;
	int i, ret = 0;
	zpool_handle_t *zhp = NULL;
	boolean_t stop = B_FALSE;
	int c;
	boolean_t noop = B_FALSE;
	boolean_t parsable = B_FALSE;
	boolean_t wait = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "npsw")) != -1) {
	switch (c) {
	case 'n':
	noop = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 's':
	stop = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	if (stop && noop) {
	+ zpool_close(zhp);
	(void) fprintf(stderr, gettext("stop request ignored\n"));
	return (0);
	}

	if (stop) {
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	if (zpool_vdev_remove_cancel(zhp) != 0)
	ret = 1;
	if (wait) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -w cannot be used with -s\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device\n"));
	usage(B_FALSE);
	}

	for (i = 1; i < argc; i++) {
	if (noop) {
	uint64_t size;

	if (zpool_vdev_indirect_size(zhp, argv[i],
	&size) != 0) {
	ret = 1;
	break;
	}
	if (parsable) {
	(void) printf("%s %llu\n",
	argv[i], (unsigned long long)size);
	} else {
	char valstr[32];
	zfs_nicenum(size, valstr,
	sizeof (valstr));
	(void) printf("Memory that will be "
	"used after removing %s: %s\n",
	argv[i], valstr);
	}
	} else {
	if (zpool_vdev_remove(zhp, argv[i]) != 0)
	ret = 1;
	}
	}

	if (ret == 0 && wait)
	ret = zpool_wait(zhp, ZPOOL_WAIT_REMOVE);
	}
	zpool_close(zhp);

	return (ret);
	}

	/*
	* Return 1 if a vdev is active (being used in a pool)
	* Return 0 if a vdev is inactive (offlined or faulted, or not in active pool)
	*
	* This is useful for checking if a disk in an active pool is offlined or
	* faulted.
	*/
	static int
	vdev_is_active(char *vdev_path)
	{
	int fd;
	fd = open(vdev_path, O_EXCL);
	if (fd < 0) {
	return (1); /* cant open O_EXCL - disk is active */
	}

	close(fd);
	return (0); /* disk is inactive in the pool */
	}

	/*
	* zpool labelclear [-f] <vdev>
	*
	* -f Force clearing the label for the vdevs which are members of
	* the exported or foreign pools.
	*
	* Verifies that the vdev is not active and zeros out the label information
	* on the device.
	*/
	int
	zpool_do_labelclear(int argc, char **argv)
	{
	char vdev[MAXPATHLEN];
	char *name = NULL;
	struct stat st;
	int c, fd = -1, ret = 0;
	nvlist_t *config;
	pool_state_t state;
	boolean_t inuse = B_FALSE;
	boolean_t force = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "f")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get vdev name */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing vdev name\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/*
	* Check if we were given absolute path and use it as is.
	* Otherwise if the provided vdev name doesn't point to a file,
	* try prepending expected disk paths and partition numbers.
	*/
	(void) strlcpy(vdev, argv[0], sizeof (vdev));
	if (vdev[0] != '/' && stat(vdev, &st) != 0) {
	int error;

	error = zfs_resolve_shortname(argv[0], vdev, MAXPATHLEN);
	if (error == 0 && zfs_dev_is_whole_disk(vdev)) {
	if (zfs_append_partition(vdev, MAXPATHLEN) == -1)
	error = ENOENT;
	}

	if (error \|\| (stat(vdev, &st) != 0)) {
	(void) fprintf(stderr, gettext(
	"failed to find device %s, try specifying absolute "
	"path instead\n"), argv[0]);
	return (1);
	}
	}

	if ((fd = open(vdev, O_RDWR)) < 0) {
	(void) fprintf(stderr, gettext("failed to open %s: %s\n"),
	vdev, strerror(errno));
	return (1);
	}

	/*
	* Flush all dirty pages for the block device. This should not be
	* fatal when the device does not support BLKFLSBUF as would be the
	* case for a file vdev.
	*/
	if ((zfs_dev_flush(fd) != 0) && (errno != ENOTTY))
	(void) fprintf(stderr, gettext("failed to invalidate "
	"cache for %s: %s\n"), vdev, strerror(errno));

	if (zpool_read_label(fd, &config, NULL) != 0) {
	(void) fprintf(stderr,
	gettext("failed to read label from %s\n"), vdev);
	ret = 1;
	goto errout;
	}
	nvlist_free(config);

	ret = zpool_in_use(g_zfs, fd, &state, &name, &inuse);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to check state for %s\n"), vdev);
	ret = 1;
	goto errout;
	}

	if (!inuse)
	goto wipe_label;

	switch (state) {
	default:
	case POOL_STATE_ACTIVE:
	case POOL_STATE_SPARE:
	case POOL_STATE_L2CACHE:
	/*
	* We allow the user to call 'zpool offline -f'
	* on an offlined disk in an active pool. We can check if
	* the disk is online by calling vdev_is_active().
	*/
	if (force && !vdev_is_active(vdev))
	break;

	(void) fprintf(stderr, gettext(
	"%s is a member (%s) of pool \"%s\""),
	vdev, zpool_pool_state_to_name(state), name);

	if (force) {
	(void) fprintf(stderr, gettext(
	". Offline the disk first to clear its label."));
	}
	printf("\n");
	ret = 1;
	goto errout;

	case POOL_STATE_EXPORTED:
	if (force)
	break;
	(void) fprintf(stderr, gettext(
	"use '-f' to override the following error:\n"
	"%s is a member of exported pool \"%s\"\n"),
	vdev, name);
	ret = 1;
	goto errout;

	case POOL_STATE_POTENTIALLY_ACTIVE:
	if (force)
	break;
	(void) fprintf(stderr, gettext(
	"use '-f' to override the following error:\n"
	"%s is a member of potentially active pool \"%s\"\n"),
	vdev, name);
	ret = 1;
	goto errout;

	case POOL_STATE_DESTROYED:
	/* inuse should never be set for a destroyed pool */
	assert(0);
	break;
	}

	wipe_label:
	ret = zpool_clear_label(fd);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to clear label for %s\n"), vdev);
	}

	errout:
	free(name);
	(void) close(fd);

	return (ret);
	}

	/*
	* zpool create [-fnd] [-o property=value] ...
	* [-O file-system-property=value] ...
	* [-R root] [-m mountpoint] <pool> <dev> ...
	*
	* -f Force creation, even if devices appear in use
	* -n Do not create the pool, but display the resulting layout if it
	* were to be created.
	* -R Create a pool under an alternate root
	* -m Set default mountpoint for the root dataset. By default it's
	* '/<pool>'
	* -o Set property=value.
	* -o Set feature@feature=enabled\|disabled.
	* -d Don't automatically enable all supported pool features
	* (individual features can be enabled with -o).
	* -O Set fsproperty=value in the pool's root file system
	*
	* Creates the named pool according to the given vdev specification. The
	* bulk of the vdev processing is done in make_root_vdev() in zpool_vdev.c.
	* Once we get the nvlist back from make_root_vdev(), we either print out the
	* contents (if '-n' was specified), or pass it to libzfs to do the creation.
	*/
	int
	zpool_do_create(int argc, char **argv)
	{
	boolean_t force = B_FALSE;
	boolean_t dryrun = B_FALSE;
	boolean_t enable_pool_features = B_TRUE;

	int c;
	nvlist_t *nvroot = NULL;
	char *poolname;
	char *tname = NULL;
	int ret = 1;
	char *altroot = NULL;
	char *compat = NULL;
	char *mountpoint = NULL;
	nvlist_t *fsprops = NULL;
	nvlist_t *props = NULL;
	char *propval;

	/* check options */
	while ((c = getopt(argc, argv, ":fndR:m:o:O:t:")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'd':
	enable_pool_features = B_FALSE;
	break;
	case 'R':
	altroot = optarg;
	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
	goto errout;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto errout;
	break;
	case 'm':
	/* Equivalent to -O mountpoint=optarg */
	mountpoint = optarg;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	goto errout;
	}
	*propval = '\0';
	propval++;

	if (add_prop_list(optarg, propval, &props, B_TRUE))
	goto errout;

	/*
	* If the user is creating a pool that doesn't support
	* feature flags, don't enable any features.
	*/
	if (zpool_name_to_prop(optarg) == ZPOOL_PROP_VERSION) {
	char *end;
	u_longlong_t ver;

	ver = strtoull(propval, &end, 10);
	if (*end == '\0' &&
	ver < SPA_VERSION_FEATURES) {
	enable_pool_features = B_FALSE;
	}
	}
	if (zpool_name_to_prop(optarg) == ZPOOL_PROP_ALTROOT)
	altroot = propval;
	if (zpool_name_to_prop(optarg) ==
	ZPOOL_PROP_COMPATIBILITY)
	compat = propval;
	break;
	case 'O':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -O option\n"));
	goto errout;
	}
	*propval = '\0';
	propval++;

	/*
	* Mountpoints are checked and then added later.
	* Uniquely among properties, they can be specified
	* more than once, to avoid conflict with -m.
	*/
	if (0 == strcmp(optarg,
	zfs_prop_to_name(ZFS_PROP_MOUNTPOINT))) {
	mountpoint = propval;
	} else if (add_prop_list(optarg, propval, &fsprops,
	B_FALSE)) {
	goto errout;
	}
	break;
	case 't':
	/*
	* Sanity check temporary pool name.
	*/
	if (strchr(optarg, '/') != NULL) {
	(void) fprintf(stderr, gettext("cannot create "
	"'%s': invalid character '/' in temporary "
	"name\n"), optarg);
	(void) fprintf(stderr, gettext("use 'zfs "
	"create' to create a dataset\n"));
	goto errout;
	}

	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_TNAME), optarg, &props, B_TRUE))
	goto errout;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto errout;
	tname = optarg;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	goto badusage;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto badusage;
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	goto badusage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing vdev specification\n"));
	goto badusage;
	}

	poolname = argv[0];

	/*
	* As a special case, check for use of '/' in the name, and direct the
	* user to use 'zfs create' instead.
	*/
	if (strchr(poolname, '/') != NULL) {
	(void) fprintf(stderr, gettext("cannot create '%s': invalid "
	"character '/' in pool name\n"), poolname);
	(void) fprintf(stderr, gettext("use 'zfs create' to "
	"create a dataset\n"));
	goto errout;
	}

	/* pass off to make_root_vdev for bulk processing */
	nvroot = make_root_vdev(NULL, props, force, !force, B_FALSE, dryrun,
	argc - 1, argv + 1);
	if (nvroot == NULL)
	goto errout;

	/* make_root_vdev() allows 0 toplevel children if there are spares */
	if (!zfs_allocatable_devs(nvroot)) {
	(void) fprintf(stderr, gettext("invalid vdev "
	"specification: at least one toplevel vdev must be "
	"specified\n"));
	goto errout;
	}

	if (altroot != NULL && altroot[0] != '/') {
	(void) fprintf(stderr, gettext("invalid alternate root '%s': "
	"must be an absolute path\n"), altroot);
	goto errout;
	}

	/*
	* Check the validity of the mountpoint and direct the user to use the
	* '-m' mountpoint option if it looks like its in use.
	*/
	if (mountpoint == NULL \|\|
	(strcmp(mountpoint, ZFS_MOUNTPOINT_LEGACY) != 0 &&
	strcmp(mountpoint, ZFS_MOUNTPOINT_NONE) != 0)) {
	char buf[MAXPATHLEN];
	DIR *dirp;

	if (mountpoint && mountpoint[0] != '/') {
	(void) fprintf(stderr, gettext("invalid mountpoint "
	"'%s': must be an absolute path, 'legacy', or "
	"'none'\n"), mountpoint);
	goto errout;
	}

	if (mountpoint == NULL) {
	if (altroot != NULL)
	(void) snprintf(buf, sizeof (buf), "%s/%s",
	altroot, poolname);
	else
	(void) snprintf(buf, sizeof (buf), "/%s",
	poolname);
	} else {
	if (altroot != NULL)
	(void) snprintf(buf, sizeof (buf), "%s%s",
	altroot, mountpoint);
	else
	(void) snprintf(buf, sizeof (buf), "%s",
	mountpoint);
	}

	if ((dirp = opendir(buf)) == NULL && errno != ENOENT) {
	(void) fprintf(stderr, gettext("mountpoint '%s' : "
	"%s\n"), buf, strerror(errno));
	(void) fprintf(stderr, gettext("use '-m' "
	"option to provide a different default\n"));
	goto errout;
	} else if (dirp) {
	int count = 0;

	while (count < 3 && readdir(dirp) != NULL)
	count++;
	(void) closedir(dirp);

	if (count > 2) {
	(void) fprintf(stderr, gettext("mountpoint "
	"'%s' exists and is not empty\n"), buf);
	(void) fprintf(stderr, gettext("use '-m' "
	"option to provide a "
	"different default\n"));
	goto errout;
	}
	}
	}

	/*
	* Now that the mountpoint's validity has been checked, ensure that
	* the property is set appropriately prior to creating the pool.
	*/
	if (mountpoint != NULL) {
	ret = add_prop_list(zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
	mountpoint, &fsprops, B_FALSE);
	if (ret != 0)
	goto errout;
	}

	ret = 1;
	if (dryrun) {
	/*
	* For a dry run invocation, print out a basic message and run
	* through all the vdevs in the list and print out in an
	* appropriate hierarchy.
	*/
	(void) printf(gettext("would create '%s' with the "
	"following layout:\n\n"), poolname);

	print_vdev_tree(NULL, poolname, nvroot, 0, "", 0);
	print_vdev_tree(NULL, "dedup", nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, 0);
	print_vdev_tree(NULL, "special", nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, 0);
	print_vdev_tree(NULL, "logs", nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, 0);
	print_cache_list(nvroot, 0);
	print_spare_list(nvroot, 0);

	ret = 0;
	} else {
	/*
	* Load in feature set.
	* Note: if compatibility property not given, we'll have
	* NULL, which means 'all features'.
	*/
	boolean_t requested_features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, requested_features) !=
	ZPOOL_COMPATIBILITY_OK)
	goto errout;

	/*
	* props contains list of features to enable.
	* For each feature:
	* - remove it if feature@name=disabled
	* - leave it there if feature@name=enabled
	* - add it if:
	* - enable_pool_features (ie: no '-d' or '-o version')
	* - it's supported by the kernel module
	* - it's in the requested feature set
	* - warn if it's enabled but not in compat
	*/
	for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
	char propname[MAXPATHLEN];
	char *propval;
	zfeature_info_t *feat = &spa_feature_table[i];

	(void) snprintf(propname, sizeof (propname),
	"feature@%s", feat->fi_uname);

	if (!nvlist_lookup_string(props, propname, &propval)) {
	if (strcmp(propval,
	ZFS_FEATURE_DISABLED) == 0) {
	(void) nvlist_remove_all(props,
	propname);
	} else if (strcmp(propval,
	ZFS_FEATURE_ENABLED) == 0 &&
	!requested_features[i]) {
	(void) fprintf(stderr, gettext(
	"Warning: feature \"%s\" enabled "
	"but is not in specified "
	"'compatibility' feature set.\n"),
	feat->fi_uname);
	}
	} else if (
	enable_pool_features &&
	feat->fi_zfs_mod_supported &&
	requested_features[i]) {
	ret = add_prop_list(propname,
	ZFS_FEATURE_ENABLED, &props, B_TRUE);
	if (ret != 0)
	goto errout;
	}
	}

	ret = 1;
	if (zpool_create(g_zfs, poolname,
	nvroot, props, fsprops) == 0) {
	zfs_handle_t *pool = zfs_open(g_zfs,
	tname ? tname : poolname, ZFS_TYPE_FILESYSTEM);
	if (pool != NULL) {
	if (zfs_mount(pool, NULL, 0) == 0) {
	ret = zfs_share(pool, NULL);
	zfs_commit_shares(NULL);
	}
	zfs_close(pool);
	}
	} else if (libzfs_errno(g_zfs) == EZFS_INVALIDNAME) {
	(void) fprintf(stderr, gettext("pool name may have "
	"been omitted\n"));
	}
	}

	errout:
	nvlist_free(nvroot);
	nvlist_free(fsprops);
	nvlist_free(props);
	return (ret);
	badusage:
	nvlist_free(fsprops);
	nvlist_free(props);
	usage(B_FALSE);
	return (2);
	}

	/*
	* zpool destroy <pool>
	*
	* -f Forcefully unmount any datasets
	*
	* Destroy the given pool. Automatically unmounts any datasets in the pool.
	*/
	int
	zpool_do_destroy(int argc, char **argv)
	{
	boolean_t force = B_FALSE;
	int c;
	char *pool;
	zpool_handle_t *zhp;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "f")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	pool = argv[0];

	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
	/*
	* As a special case, check for use of '/' in the name, and
	* direct the user to use 'zfs destroy' instead.
	*/
	if (strchr(pool, '/') != NULL)
	(void) fprintf(stderr, gettext("use 'zfs destroy' to "
	"destroy a dataset\n"));
	return (1);
	}

	if (zpool_disable_datasets(zhp, force) != 0) {
	(void) fprintf(stderr, gettext("could not destroy '%s': "
	"could not unmount datasets\n"), zpool_get_name(zhp));
	zpool_close(zhp);
	return (1);
	}

	/* The history must be logged as part of the export */
	log_history = B_FALSE;

	ret = (zpool_destroy(zhp, history_str) != 0);

	zpool_close(zhp);

	return (ret);
	}

	typedef struct export_cbdata {
	boolean_t force;
	boolean_t hardforce;
	} export_cbdata_t;

	/*
	* Export one pool
	*/
	static int
	zpool_export_one(zpool_handle_t zhp, void data)
	{
	export_cbdata_t *cb = data;

	if (zpool_disable_datasets(zhp, cb->force) != 0)
	return (1);

	/* The history must be logged as part of the export */
	log_history = B_FALSE;

	if (cb->hardforce) {
	if (zpool_export_force(zhp, history_str) != 0)
	return (1);
	} else if (zpool_export(zhp, cb->force, history_str) != 0) {
	return (1);
	}

	return (0);
	}

	/*
	* zpool export [-f] <pool> ...
	*
	* -a Export all pools
	* -f Forcefully unmount datasets
	*
	* Export the given pools. By default, the command will attempt to cleanly
	* unmount any active datasets within the pool. If the '-f' flag is specified,
	* then the datasets will be forcefully unmounted.
	*/
	int
	zpool_do_export(int argc, char **argv)
	{
	export_cbdata_t cb;
	boolean_t do_all = B_FALSE;
	boolean_t force = B_FALSE;
	boolean_t hardforce = B_FALSE;
	int c, ret;

	/* check options */
	while ((c = getopt(argc, argv, "afF")) != -1) {
	switch (c) {
	case 'a':
	do_all = B_TRUE;
	break;
	case 'f':
	force = B_TRUE;
	break;
	case 'F':
	hardforce = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	cb.force = force;
	cb.hardforce = hardforce;
	argc -= optind;
	argv += optind;

	if (do_all) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	return (for_each_pool(argc, argv, B_TRUE, NULL,
	ZFS_TYPE_POOL, B_FALSE, zpool_export_one, &cb));
	}

	/* check arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}

	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_export_one, &cb);

	return (ret);
	}

	/*
	* Given a vdev configuration, determine the maximum width needed for the device
	* name column.
	*/
	static int
	max_width(zpool_handle_t zhp, nvlist_t nv, int depth, int max,
	int name_flags)
	{
	static const char *const subtypes[] =
	{ZPOOL_CONFIG_SPARES, ZPOOL_CONFIG_L2CACHE, ZPOOL_CONFIG_CHILDREN};

	char *name = zpool_vdev_name(g_zfs, zhp, nv, name_flags);
	max = MAX(strlen(name) + depth, max);
	free(name);

	nvlist_t **child;
	uint_t children;
	for (size_t i = 0; i < ARRAY_SIZE(subtypes); ++i)
	if (nvlist_lookup_nvlist_array(nv, subtypes[i],
	&child, &children) == 0)
	for (uint_t c = 0; c < children; ++c)
	max = MAX(max_width(zhp, child[c], depth + 2,
	max, name_flags), max);

	return (max);
	}

	typedef struct spare_cbdata {
	uint64_t cb_guid;
	zpool_handle_t *cb_zhp;
	} spare_cbdata_t;

	static boolean_t
	find_vdev(nvlist_t *nv, uint64_t search)
	{
	uint64_t guid;
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
	search == guid)
	return (B_TRUE);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	if (find_vdev(child[c], search))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static int
	find_spare(zpool_handle_t zhp, void data)
	{
	spare_cbdata_t *cbp = data;
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	if (find_vdev(nvroot, cbp->cb_guid)) {
	cbp->cb_zhp = zhp;
	return (1);
	}

	zpool_close(zhp);
	return (0);
	}

	typedef struct status_cbdata {
	int cb_count;
	int cb_name_flags;
	int cb_namewidth;
	boolean_t cb_allpools;
	boolean_t cb_verbose;
	boolean_t cb_literal;
	boolean_t cb_explain;
	boolean_t cb_first;
	boolean_t cb_dedup_stats;
	boolean_t cb_print_status;
	boolean_t cb_print_slow_ios;
	boolean_t cb_print_vdev_init;
	boolean_t cb_print_vdev_trim;
	vdev_cmd_data_list_t *vcdl;
	} status_cbdata_t;

	/* Return 1 if string is NULL, empty, or whitespace; return 0 otherwise. */
	static boolean_t
	is_blank_str(const char *str)
	{
	for (; str != NULL && *str != '\0'; ++str)
	if (!isblank(*str))
	return (B_FALSE);
	return (B_TRUE);
	}

	/* Print command output lines for specific vdev in a specific pool */
	static void
	zpool_print_cmd(vdev_cmd_data_list_t vcdl, const char pool, char *path)
	{
	vdev_cmd_data_t *data;
	int i, j;
	const char *val;

	for (i = 0; i < vcdl->count; i++) {
	if ((strcmp(vcdl->data[i].path, path) != 0) \|\|
	(strcmp(vcdl->data[i].pool, pool) != 0)) {
	/* Not the vdev we're looking for */
	continue;
	}

	data = &vcdl->data[i];
	/* Print out all the output values for this vdev */
	for (j = 0; j < vcdl->uniq_cols_cnt; j++) {
	val = NULL;
	/* Does this vdev have values for this column? */
	for (int k = 0; k < data->cols_cnt; k++) {
	if (strcmp(data->cols[k],
	vcdl->uniq_cols[j]) == 0) {
	/* yes it does, record the value */
	val = data->lines[k];
	break;
	}
	}
	/*
	* Mark empty values with dashes to make output
	* awk-able.
	*/
	if (val == NULL \|\| is_blank_str(val))
	val = "-";

	printf("%*s", vcdl->uniq_cols_width[j], val);
	if (j < vcdl->uniq_cols_cnt - 1)
	fputs(" ", stdout);
	}

	/* Print out any values that aren't in a column at the end */
	for (j = data->cols_cnt; j < data->lines_cnt; j++) {
	/* Did we have any columns? If so print a spacer. */
	if (vcdl->uniq_cols_cnt > 0)
	fputs(" ", stdout);

	val = data->lines[j];
	fputs(val ?: "", stdout);
	}
	break;
	}
	}

	/*
	* Print vdev initialization status for leaves
	*/
	static void
	print_status_initialize(vdev_stat_t *vs, boolean_t verbose)
	{
	if (verbose) {
	if ((vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE \|\|
	vs->vs_initialize_state == VDEV_INITIALIZE_SUSPENDED \|\|
	vs->vs_initialize_state == VDEV_INITIALIZE_COMPLETE) &&
	!vs->vs_scan_removing) {
	char zbuf[1024];
	char tbuf[256];
	struct tm zaction_ts;

	time_t t = vs->vs_initialize_action_time;
	int initialize_pct = 100;
	if (vs->vs_initialize_state !=
	VDEV_INITIALIZE_COMPLETE) {
	initialize_pct = (vs->vs_initialize_bytes_done *
	100 / (vs->vs_initialize_bytes_est + 1));
	}

	(void) localtime_r(&t, &zaction_ts);
	(void) strftime(tbuf, sizeof (tbuf), "%c", &zaction_ts);

	switch (vs->vs_initialize_state) {
	case VDEV_INITIALIZE_SUSPENDED:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("suspended, started at"), tbuf);
	break;
	case VDEV_INITIALIZE_ACTIVE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("started at"), tbuf);
	break;
	case VDEV_INITIALIZE_COMPLETE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("completed at"), tbuf);
	break;
	}

	(void) printf(gettext(" (%d%% initialized%s)"),
	initialize_pct, zbuf);
	} else {
	(void) printf(gettext(" (uninitialized)"));
	}
	} else if (vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) {
	(void) printf(gettext(" (initializing)"));
	}
	}

	/*
	* Print vdev TRIM status for leaves
	*/
	static void
	print_status_trim(vdev_stat_t *vs, boolean_t verbose)
	{
	if (verbose) {
	if ((vs->vs_trim_state == VDEV_TRIM_ACTIVE \|\|
	vs->vs_trim_state == VDEV_TRIM_SUSPENDED \|\|
	vs->vs_trim_state == VDEV_TRIM_COMPLETE) &&
	!vs->vs_scan_removing) {
	char zbuf[1024];
	char tbuf[256];
	struct tm zaction_ts;

	time_t t = vs->vs_trim_action_time;
	int trim_pct = 100;
	if (vs->vs_trim_state != VDEV_TRIM_COMPLETE) {
	trim_pct = (vs->vs_trim_bytes_done *
	100 / (vs->vs_trim_bytes_est + 1));
	}

	(void) localtime_r(&t, &zaction_ts);
	(void) strftime(tbuf, sizeof (tbuf), "%c", &zaction_ts);

	switch (vs->vs_trim_state) {
	case VDEV_TRIM_SUSPENDED:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("suspended, started at"), tbuf);
	break;
	case VDEV_TRIM_ACTIVE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("started at"), tbuf);
	break;
	case VDEV_TRIM_COMPLETE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("completed at"), tbuf);
	break;
	}

	(void) printf(gettext(" (%d%% trimmed%s)"),
	trim_pct, zbuf);
	} else if (vs->vs_trim_notsup) {
	(void) printf(gettext(" (trim unsupported)"));
	} else {
	(void) printf(gettext(" (untrimmed)"));
	}
	} else if (vs->vs_trim_state == VDEV_TRIM_ACTIVE) {
	(void) printf(gettext(" (trimming)"));
	}
	}

	/*
	* Return the color associated with a health string. This includes returning
	* NULL for no color change.
	*/
	static const char *
	health_str_to_color(const char *health)
	{
	if (strcmp(health, gettext("FAULTED")) == 0 \|\|
	strcmp(health, gettext("SUSPENDED")) == 0 \|\|
	strcmp(health, gettext("UNAVAIL")) == 0) {
	return (ANSI_RED);
	}

	if (strcmp(health, gettext("OFFLINE")) == 0 \|\|
	strcmp(health, gettext("DEGRADED")) == 0 \|\|
	strcmp(health, gettext("REMOVED")) == 0) {
	return (ANSI_YELLOW);
	}

	return (NULL);
	}

	/*
	* Print out configuration state as requested by status_callback.
	*/
	static void
	print_status_config(zpool_handle_t zhp, status_cbdata_t cb, const char *name,
	nvlist_t nv, int depth, boolean_t isspare, vdev_rebuild_stat_t vrs)
	{
	nvlist_t *child, root;
	uint_t c, i, vsc, children;
	pool_scan_stat_t *ps = NULL;
	vdev_stat_t *vs;
	char rbuf[6], wbuf[6], cbuf[6];
	char *vname;
	uint64_t notpresent;
	spare_cbdata_t spare_cb;
	const char *state;
	char *type;
	char *path = NULL;
	const char rcolor = NULL, wcolor = NULL, *ccolor = NULL;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &vsc) == 0);

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);

	if (strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	return;

	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);

	if (isspare) {
	/*
	* For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
	* online drives.
	*/
	if (vs->vs_aux == VDEV_AUX_SPARED)
	state = gettext("INUSE");
	else if (vs->vs_state == VDEV_STATE_HEALTHY)
	state = gettext("AVAIL");
	}

	printf_color(health_str_to_color(state),
	"\t%s%-s %-8s", depth, "", cb->cb_namewidth - depth,
	name, state);

	if (!isspare) {
	if (vs->vs_read_errors)
	rcolor = ANSI_RED;

	if (vs->vs_write_errors)
	wcolor = ANSI_RED;

	if (vs->vs_checksum_errors)
	ccolor = ANSI_RED;

	if (cb->cb_literal) {
	fputc(' ', stdout);
	printf_color(rcolor, "%5llu",
	(u_longlong_t)vs->vs_read_errors);
	fputc(' ', stdout);
	printf_color(wcolor, "%5llu",
	(u_longlong_t)vs->vs_write_errors);
	fputc(' ', stdout);
	printf_color(ccolor, "%5llu",
	(u_longlong_t)vs->vs_checksum_errors);
	} else {
	zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
	zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
	zfs_nicenum(vs->vs_checksum_errors, cbuf,
	sizeof (cbuf));
	fputc(' ', stdout);
	printf_color(rcolor, "%5s", rbuf);
	fputc(' ', stdout);
	printf_color(wcolor, "%5s", wbuf);
	fputc(' ', stdout);
	printf_color(ccolor, "%5s", cbuf);
	}
	if (cb->cb_print_slow_ios) {
	if (children == 0) {
	/* Only leafs vdevs have slow IOs */
	zfs_nicenum(vs->vs_slow_ios, rbuf,
	sizeof (rbuf));
	} else {
	snprintf(rbuf, sizeof (rbuf), "-");
	}

	if (cb->cb_literal)
	printf(" %5llu", (u_longlong_t)vs->vs_slow_ios);
	else
	printf(" %5s", rbuf);
	}
	}

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
	&notpresent) == 0) {
	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
	(void) printf(" %s %s", gettext("was"), path);
	} else if (vs->vs_aux != 0) {
	(void) printf(" ");
	color_start(ANSI_RED);
	switch (vs->vs_aux) {
	case VDEV_AUX_OPEN_FAILED:
	(void) printf(gettext("cannot open"));
	break;

	case VDEV_AUX_BAD_GUID_SUM:
	(void) printf(gettext("missing device"));
	break;

	case VDEV_AUX_NO_REPLICAS:
	(void) printf(gettext("insufficient replicas"));
	break;

	case VDEV_AUX_VERSION_NEWER:
	(void) printf(gettext("newer version"));
	break;

	case VDEV_AUX_UNSUP_FEAT:
	(void) printf(gettext("unsupported feature(s)"));
	break;

	case VDEV_AUX_ASHIFT_TOO_BIG:
	(void) printf(gettext("unsupported minimum blocksize"));
	break;

	case VDEV_AUX_SPARED:
	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
	&spare_cb.cb_guid) == 0);
	if (zpool_iter(g_zfs, find_spare, &spare_cb) == 1) {
	if (strcmp(zpool_get_name(spare_cb.cb_zhp),
	zpool_get_name(zhp)) == 0)
	(void) printf(gettext("currently in "
	"use"));
	else
	(void) printf(gettext("in use by "
	"pool '%s'"),
	zpool_get_name(spare_cb.cb_zhp));
	zpool_close(spare_cb.cb_zhp);
	} else {
	(void) printf(gettext("currently in use"));
	}
	break;

	case VDEV_AUX_ERR_EXCEEDED:
	(void) printf(gettext("too many errors"));
	break;

	case VDEV_AUX_IO_FAILURE:
	(void) printf(gettext("experienced I/O failures"));
	break;

	case VDEV_AUX_BAD_LOG:
	(void) printf(gettext("bad intent log"));
	break;

	case VDEV_AUX_EXTERNAL:
	(void) printf(gettext("external device fault"));
	break;

	case VDEV_AUX_SPLIT_POOL:
	(void) printf(gettext("split into new pool"));
	break;

	case VDEV_AUX_ACTIVE:
	(void) printf(gettext("currently in use"));
	break;

	case VDEV_AUX_CHILDREN_OFFLINE:
	(void) printf(gettext("all children offline"));
	break;

	case VDEV_AUX_BAD_LABEL:
	(void) printf(gettext("invalid label"));
	break;

	default:
	(void) printf(gettext("corrupted data"));
	break;
	}
	color_end();
	} else if (children == 0 && !isspare &&
	getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL &&
	VDEV_STAT_VALID(vs_physical_ashift, vsc) &&
	vs->vs_configured_ashift < vs->vs_physical_ashift) {
	(void) printf(
	gettext(" block size: %dB configured, %dB native"),
	1 << vs->vs_configured_ashift, 1 << vs->vs_physical_ashift);
	}

	if (vs->vs_scan_removing != 0) {
	(void) printf(gettext(" (removing)"));
	} else if (VDEV_STAT_VALID(vs_noalloc, vsc) && vs->vs_noalloc != 0) {
	(void) printf(gettext(" (non-allocating)"));
	}

	/* The root vdev has the scrub/resilver stats */
	root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
	ZPOOL_CONFIG_VDEV_TREE);
	(void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS,
	(uint64_t **)&ps, &c);

	- if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0) {
	+ /*
	+ * If you force fault a drive that's resilvering, its scan stats can
	+ * get frozen in time, giving the false impression that it's
	+ * being resilvered. That's why we check the state to see if the vdev
	+ * is healthy before reporting "resilvering" or "repairing".
	+ */
	+ if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0 &&
	+ vs->vs_state == VDEV_STATE_HEALTHY) {
	if (vs->vs_scan_processed != 0) {
	(void) printf(gettext(" (%s)"),
	(ps->pss_func == POOL_SCAN_RESILVER) ?
	"resilvering" : "repairing");
	} else if (vs->vs_resilver_deferred) {
	(void) printf(gettext(" (awaiting resilver)"));
	}
	}

	/* The top-level vdevs have the rebuild stats */
	if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE &&
	- children == 0) {
	+ children == 0 && vs->vs_state == VDEV_STATE_HEALTHY) {
	if (vs->vs_rebuild_processed != 0) {
	(void) printf(gettext(" (resilvering)"));
	}
	}

	if (cb->vcdl != NULL) {
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
	printf(" ");
	zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
	}
	}

	/* Display vdev initialization and trim status for leaves. */
	if (children == 0) {
	print_status_initialize(vs, cb->cb_print_vdev_init);
	print_status_trim(vs, cb->cb_print_vdev_trim);
	}

	(void) printf("\n");

	for (c = 0; c < children; c++) {
	uint64_t islog = B_FALSE, ishole = B_FALSE;

	/* Don't print logs or holes here */
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&islog);
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&ishole);
	if (islog \|\| ishole)
	continue;
	/* Only print normal classes here */
	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	/* Provide vdev_rebuild_stats to children if available */
	if (vrs == NULL) {
	(void) nvlist_lookup_uint64_array(nv,
	ZPOOL_CONFIG_REBUILD_STATS,
	(uint64_t **)&vrs, &i);
	}

	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_status_config(zhp, cb, vname, child[c], depth + 2,
	isspare, vrs);
	free(vname);
	}
	}

	/*
	* Print the configuration of an exported pool. Iterate over all vdevs in the
	* pool, printing out the name and status for each one.
	*/
	static void
	print_import_config(status_cbdata_t cb, const char name, nvlist_t *nv,
	int depth)
	{
	nvlist_t **child;
	uint_t c, children;
	vdev_stat_t *vs;
	char type, vname;

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
	if (strcmp(type, VDEV_TYPE_MISSING) == 0 \|\|
	strcmp(type, VDEV_TYPE_HOLE) == 0)
	return;

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	(void) printf("\t%s%-s", depth, "", cb->cb_namewidth - depth, name);
	(void) printf(" %s", zpool_state_to_name(vs->vs_state, vs->vs_aux));

	if (vs->vs_aux != 0) {
	(void) printf(" ");

	switch (vs->vs_aux) {
	case VDEV_AUX_OPEN_FAILED:
	(void) printf(gettext("cannot open"));
	break;

	case VDEV_AUX_BAD_GUID_SUM:
	(void) printf(gettext("missing device"));
	break;

	case VDEV_AUX_NO_REPLICAS:
	(void) printf(gettext("insufficient replicas"));
	break;

	case VDEV_AUX_VERSION_NEWER:
	(void) printf(gettext("newer version"));
	break;

	case VDEV_AUX_UNSUP_FEAT:
	(void) printf(gettext("unsupported feature(s)"));
	break;

	case VDEV_AUX_ERR_EXCEEDED:
	(void) printf(gettext("too many errors"));
	break;

	case VDEV_AUX_ACTIVE:
	(void) printf(gettext("currently in use"));
	break;

	case VDEV_AUX_CHILDREN_OFFLINE:
	(void) printf(gettext("all children offline"));
	break;

	case VDEV_AUX_BAD_LABEL:
	(void) printf(gettext("invalid label"));
	break;

	default:
	(void) printf(gettext("corrupted data"));
	break;
	}
	}
	(void) printf("\n");

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return;

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);
	if (is_log)
	continue;
	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_import_config(cb, vname, child[c], depth + 2);
	free(vname);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0) {
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0) {
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	}

	/*
	* Print specialized class vdevs.
	*
	* These are recorded as top level vdevs in the main pool child array
	* but with "is_log" set to 1 or an "alloc_bias" string. We use either
	* print_status_config() or print_import_config() to print the top level
	* class vdevs then any of their children (eg mirrored slogs) are printed
	* recursively - which works because only the top level vdev is marked.
	*/
	static void
	print_class_vdevs(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t *nv,
	const char *class)
	{
	uint_t c, children;
	nvlist_t **child;
	boolean_t printed = B_FALSE;

	assert(zhp != NULL \|\| !cb->cb_verbose);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child,
	&children) != 0)
	return;

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE;
	char *bias = NULL;
	char *type = NULL;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);

	if (is_log) {
	bias = (char *)VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_TYPE, &type);
	}

	if (bias == NULL \|\| strcmp(bias, class) != 0)
	continue;
	if (!is_log && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	(void) printf("\t%s\t\n", gettext(class));
	printed = B_TRUE;
	}

	char *name = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	if (cb->cb_print_status)
	print_status_config(zhp, cb, name, child[c], 2,
	B_FALSE, NULL);
	else
	print_import_config(cb, name, child[c], 2);
	free(name);
	}
	}

	/*
	* Display the status for the given pool.
	*/
	static int
	show_import(nvlist_t *config, boolean_t report_error)
	{
	uint64_t pool_state;
	vdev_stat_t *vs;
	char *name;
	uint64_t guid;
	uint64_t hostid = 0;
	const char *msgid;
	const char *hostname = "unknown";
	nvlist_t nvroot, nvinfo;
	zpool_status_t reason;
	zpool_errata_t errata;
	const char *health;
	uint_t vsc;
	char *comment;
	status_cbdata_t cb = { 0 };

	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&name) == 0);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&guid) == 0);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&pool_state) == 0);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &vsc) == 0);
	health = zpool_state_to_name(vs->vs_state, vs->vs_aux);

	reason = zpool_import_status(config, &msgid, &errata);

	/*
	* If we're importing using a cachefile, then we won't report any
	* errors unless we are in the scan phase of the import.
	*/
	if (reason != ZPOOL_STATUS_OK && !report_error)
	return (reason);

	(void) printf(gettext(" pool: %s\n"), name);
	(void) printf(gettext(" id: %llu\n"), (u_longlong_t)guid);
	(void) printf(gettext(" state: %s"), health);
	if (pool_state == POOL_STATE_DESTROYED)
	(void) printf(gettext(" (DESTROYED)"));
	(void) printf("\n");

	switch (reason) {
	case ZPOOL_STATUS_MISSING_DEV_R:
	case ZPOOL_STATUS_MISSING_DEV_NR:
	case ZPOOL_STATUS_BAD_GUID_SUM:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"missing from the system.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_R:
	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices contains"
	" corrupted data.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_DATA:
	(void) printf(
	gettext(" status: The pool data is corrupted.\n"));
	break;

	case ZPOOL_STATUS_OFFLINE_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices "
	"are offlined.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_POOL:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool metadata is "
	"corrupted.\n"));
	break;

	case ZPOOL_STATUS_VERSION_OLDER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"a legacy on-disk version.\n"));
	break;

	case ZPOOL_STATUS_VERSION_NEWER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"an incompatible version.\n"));
	break;

	case ZPOOL_STATUS_FEAT_DISABLED:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Some supported "
	"features are not enabled on the pool.\n\t"
	"(Note that they may be intentionally disabled "
	"if the\n\t'compatibility' property is set.)\n"));
	break;

	case ZPOOL_STATUS_COMPATIBILITY_ERR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Error reading or parsing "
	"the file(s) indicated by the 'compatibility'\n"
	"property.\n"));
	break;

	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more features "
	"are enabled on the pool despite not being\n"
	"requested by the 'compatibility' property.\n"));
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool uses the following "
	"feature(s) not supported on this system:\n"));
	color_start(ANSI_YELLOW);
	zpool_print_unsup_feat(config);
	color_end();
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool can only be "
	"accessed in read-only mode on this system. It\n\tcannot be"
	" accessed in read-write mode because it uses the "
	"following\n\tfeature(s) not supported on this system:\n"));
	color_start(ANSI_YELLOW);
	zpool_print_unsup_feat(config);
	color_end();
	break;

	case ZPOOL_STATUS_HOSTID_ACTIVE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is currently "
	"imported by another system.\n"));
	break;

	case ZPOOL_STATUS_HOSTID_REQUIRED:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool has the "
	"multihost property on. It cannot\n\tbe safely imported "
	"when the system hostid is not set.\n"));
	break;

	case ZPOOL_STATUS_HOSTID_MISMATCH:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool was last accessed "
	"by another system.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_R:
	case ZPOOL_STATUS_FAULTED_DEV_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted.\n"));
	break;

	case ZPOOL_STATUS_BAD_LOG:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("An intent log record cannot "
	"be read.\n"));
	break;

	case ZPOOL_STATUS_RESILVERING:
	case ZPOOL_STATUS_REBUILDING:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices were "
	"being resilvered.\n"));
	break;

	case ZPOOL_STATUS_ERRATA:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
	errata);
	break;

	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"configured to use a non-native block size.\n"
	"\tExpect reduced performance.\n"));
	break;

	default:
	/*
	* No other status can be seen when importing pools.
	*/
	assert(reason == ZPOOL_STATUS_OK);
	}

	/*
	* Print out an action according to the overall state of the pool.
	*/
	if (vs->vs_state == VDEV_STATE_HEALTHY) {
	if (reason == ZPOOL_STATUS_VERSION_OLDER \|\|
	reason == ZPOOL_STATUS_FEAT_DISABLED) {
	(void) printf(gettext(" action: The pool can be "
	"imported using its name or numeric identifier, "
	"though\n\tsome features will not be available "
	"without an explicit 'zpool upgrade'.\n"));
	} else if (reason == ZPOOL_STATUS_COMPATIBILITY_ERR) {
	(void) printf(gettext(" action: The pool can be "
	"imported using its name or numeric\n\tidentifier, "
	"though the file(s) indicated by its "
	"'compatibility'\n\tproperty cannot be parsed at "
	"this time.\n"));
	} else if (reason == ZPOOL_STATUS_HOSTID_MISMATCH) {
	(void) printf(gettext(" action: The pool can be "
	"imported using its name or numeric "
	"identifier and\n\tthe '-f' flag.\n"));
	} else if (reason == ZPOOL_STATUS_ERRATA) {
	switch (errata) {
	case ZPOOL_ERRATA_NONE:
	break;

	case ZPOOL_ERRATA_ZOL_2094_SCRUB:
	(void) printf(gettext(" action: The pool can "
	"be imported using its name or numeric "
	"identifier,\n\thowever there is a compat"
	"ibility issue which should be corrected"
	"\n\tby running 'zpool scrub'\n"));
	break;

	case ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY:
	(void) printf(gettext(" action: The pool can"
	"not be imported with this version of ZFS "
	"due to\n\tan active asynchronous destroy. "
	"Revert to an earlier version\n\tand "
	"allow the destroy to complete before "
	"updating.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
	(void) printf(gettext(" action: Existing "
	"encrypted datasets contain an on-disk "
	"incompatibility, which\n\tneeds to be "
	"corrected. Backup these datasets to new "
	"encrypted datasets\n\tand destroy the "
	"old ones.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
	(void) printf(gettext(" action: Existing "
	"encrypted snapshots and bookmarks contain "
	"an on-disk\n\tincompatibility. This may "
	"cause on-disk corruption if they are used"
	"\n\twith 'zfs recv'. To correct the "
	"issue, enable the bookmark_v2 feature.\n\t"
	"No additional action is needed if there "
	"are no encrypted snapshots or\n\t"
	"bookmarks. If preserving the encrypted "
	"snapshots and bookmarks is\n\trequired, "
	"use a non-raw send to backup and restore "
	"them. Alternately,\n\tthey may be removed"
	" to resolve the incompatibility.\n"));
	break;
	default:
	/*
	* All errata must contain an action message.
	*/
	assert(0);
	}
	} else {
	(void) printf(gettext(" action: The pool can be "
	"imported using its name or numeric "
	"identifier.\n"));
	}
	} else if (vs->vs_state == VDEV_STATE_DEGRADED) {
	(void) printf(gettext(" action: The pool can be imported "
	"despite missing or damaged devices. The\n\tfault "
	"tolerance of the pool may be compromised if imported.\n"));
	} else {
	switch (reason) {
	case ZPOOL_STATUS_VERSION_NEWER:
	(void) printf(gettext(" action: The pool cannot be "
	"imported. Access the pool on a system running "
	"newer\n\tsoftware, or recreate the pool from "
	"backup.\n"));
	break;
	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be "
	"imported. Access the pool on a system that "
	"supports\n\tthe required feature(s), or recreate "
	"the pool from backup.\n"));
	break;
	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be "
	"imported in read-write mode. Import the pool "
	"with\n"
	"\t\"-o readonly=on\", access the pool on a system "
	"that supports the\n\trequired feature(s), or "
	"recreate the pool from backup.\n"));
	break;
	case ZPOOL_STATUS_MISSING_DEV_R:
	case ZPOOL_STATUS_MISSING_DEV_NR:
	case ZPOOL_STATUS_BAD_GUID_SUM:
	(void) printf(gettext(" action: The pool cannot be "
	"imported. Attach the missing\n\tdevices and try "
	"again.\n"));
	break;
	case ZPOOL_STATUS_HOSTID_ACTIVE:
	VERIFY0(nvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_LOAD_INFO, &nvinfo));

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID);

	(void) printf(gettext(" action: The pool must be "
	"exported from %s (hostid=%"PRIx64")\n\tbefore it "
	"can be safely imported.\n"), hostname, hostid);
	break;
	case ZPOOL_STATUS_HOSTID_REQUIRED:
	(void) printf(gettext(" action: Set a unique system "
	"hostid with the zgenhostid(8) command.\n"));
	break;
	default:
	(void) printf(gettext(" action: The pool cannot be "
	"imported due to damaged devices or data.\n"));
	}
	}

	/* Print the comment attached to the pool. */
	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
	(void) printf(gettext("comment: %s\n"), comment);

	/*
	* If the state is "closed" or "can't open", and the aux state
	* is "corrupt data":
	*/
	if (((vs->vs_state == VDEV_STATE_CLOSED) \|\|
	(vs->vs_state == VDEV_STATE_CANT_OPEN)) &&
	(vs->vs_aux == VDEV_AUX_CORRUPT_DATA)) {
	if (pool_state == POOL_STATE_DESTROYED)
	(void) printf(gettext("\tThe pool was destroyed, "
	"but can be imported using the '-Df' flags.\n"));
	else if (pool_state != POOL_STATE_EXPORTED)
	(void) printf(gettext("\tThe pool may be active on "
	"another system, but can be imported using\n\t"
	"the '-f' flag.\n"));
	}

	if (msgid != NULL) {
	(void) printf(gettext(
	" see: https://openzfs.github.io/openzfs-docs/msg/%s\n"),
	msgid);
	}

	(void) printf(gettext(" config:\n\n"));

	cb.cb_namewidth = max_width(NULL, nvroot, 0, strlen(name),
	VDEV_NAME_TYPE_ID);
	if (cb.cb_namewidth < 10)
	cb.cb_namewidth = 10;

	print_import_config(&cb, name, nvroot, 0);

	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_DEDUP);
	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_CLASS_LOGS);

	if (reason == ZPOOL_STATUS_BAD_GUID_SUM) {
	(void) printf(gettext("\n\tAdditional devices are known to "
	"be part of this pool, though their\n\texact "
	"configuration cannot be determined.\n"));
	}
	return (0);
	}

	static boolean_t
	zfs_force_import_required(nvlist_t *config)
	{
	uint64_t state;
	uint64_t hostid = 0;
	nvlist_t *nvinfo;

	state = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE);
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID, &hostid);

	if (state != POOL_STATE_EXPORTED && hostid != get_system_hostid())
	return (B_TRUE);

	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) {
	mmp_state_t mmp_state = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_STATE);

	if (mmp_state != MMP_STATE_INACTIVE)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Perform the import for the given configuration. This passes the heavy
	* lifting off to zpool_import_props(), and then mounts the datasets contained
	* within the pool.
	*/
	static int
	do_import(nvlist_t config, const char newname, const char *mntopts,
	nvlist_t *props, int flags)
	{
	int ret = 0;
	zpool_handle_t *zhp;
	const char *name;
	uint64_t version;

	name = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
	version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);

	if (!SPA_VERSION_IS_SUPPORTED(version)) {
	(void) fprintf(stderr, gettext("cannot import '%s': pool "
	"is formatted using an unsupported ZFS version\n"), name);
	return (1);
	} else if (zfs_force_import_required(config) &&
	!(flags & ZFS_IMPORT_ANY_HOST)) {
	mmp_state_t mmp_state = MMP_STATE_INACTIVE;
	nvlist_t *nvinfo;

	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE))
	mmp_state = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_STATE);

	if (mmp_state == MMP_STATE_ACTIVE) {
	const char *hostname = "<unknown>";
	uint64_t hostid = 0;

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID);

	(void) fprintf(stderr, gettext("cannot import '%s': "
	"pool is imported on %s (hostid: "
	"0x%"PRIx64")\nExport the pool on the other "
	"system, then run 'zpool import'.\n"),
	name, hostname, hostid);
	} else if (mmp_state == MMP_STATE_NO_HOSTID) {
	(void) fprintf(stderr, gettext("Cannot import '%s': "
	"pool has the multihost property on and the\n"
	"system's hostid is not set. Set a unique hostid "
	"with the zgenhostid(8) command.\n"), name);
	} else {
	const char *hostname = "<unknown>";
	time_t timestamp = 0;
	uint64_t hostid = 0;

	if (nvlist_exists(config, ZPOOL_CONFIG_HOSTNAME))
	hostname = fnvlist_lookup_string(config,
	ZPOOL_CONFIG_HOSTNAME);

	if (nvlist_exists(config, ZPOOL_CONFIG_TIMESTAMP))
	timestamp = fnvlist_lookup_uint64(config,
	ZPOOL_CONFIG_TIMESTAMP);

	if (nvlist_exists(config, ZPOOL_CONFIG_HOSTID))
	hostid = fnvlist_lookup_uint64(config,
	ZPOOL_CONFIG_HOSTID);

	(void) fprintf(stderr, gettext("cannot import '%s': "
	"pool was previously in use from another system.\n"
	"Last accessed by %s (hostid=%"PRIx64") at %s"
	"The pool can be imported, use 'zpool import -f' "
	"to import the pool.\n"), name, hostname,
	hostid, ctime(&timestamp));
	}

	return (1);
	}

	if (zpool_import_props(g_zfs, config, newname, props, flags) != 0)
	return (1);

	if (newname != NULL)
	name = newname;

	if ((zhp = zpool_open_canfail(g_zfs, name)) == NULL)
	return (1);

	/*
	* Loading keys is best effort. We don't want to return immediately
	* if it fails but we do want to give the error to the caller.
	*/
	if (flags & ZFS_IMPORT_LOAD_KEYS &&
	zfs_crypto_attempt_load_keys(g_zfs, name) != 0)
	ret = 1;

	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
	!(flags & ZFS_IMPORT_ONLY) &&
	zpool_enable_datasets(zhp, mntopts, 0) != 0) {
	zpool_close(zhp);
	return (1);
	}

	zpool_close(zhp);
	return (ret);
	}

	static int
	import_pools(nvlist_t pools, nvlist_t props, char *mntopts, int flags,
	char orig_name, char new_name,
	boolean_t do_destroyed, boolean_t pool_specified, boolean_t do_all,
	importargs_t *import)
	{
	nvlist_t *config = NULL;
	nvlist_t *found_config = NULL;
	uint64_t pool_state;

	/*
	* At this point we have a list of import candidate configs. Even if
	* we were searching by pool name or guid, we still need to
	* post-process the list to deal with pool state and possible
	* duplicate names.
	*/
	int err = 0;
	nvpair_t *elem = NULL;
	boolean_t first = B_TRUE;
	while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {

	verify(nvpair_value_nvlist(elem, &config) == 0);

	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&pool_state) == 0);
	if (!do_destroyed && pool_state == POOL_STATE_DESTROYED)
	continue;
	if (do_destroyed && pool_state != POOL_STATE_DESTROYED)
	continue;

	verify(nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
	import->policy) == 0);

	if (!pool_specified) {
	if (first)
	first = B_FALSE;
	else if (!do_all)
	(void) fputc('\n', stdout);

	if (do_all) {
	err \|= do_import(config, NULL, mntopts,
	props, flags);
	} else {
	/*
	* If we're importing from cachefile, then
	* we don't want to report errors until we
	* are in the scan phase of the import. If
	* we get an error, then we return that error
	* to invoke the scan phase.
	*/
	if (import->cachefile && !import->scan)
	err = show_import(config, B_FALSE);
	else
	(void) show_import(config, B_TRUE);
	}
	} else if (import->poolname != NULL) {
	char *name;

	/*
	* We are searching for a pool based on name.
	*/
	verify(nvlist_lookup_string(config,
	ZPOOL_CONFIG_POOL_NAME, &name) == 0);

	if (strcmp(name, import->poolname) == 0) {
	if (found_config != NULL) {
	(void) fprintf(stderr, gettext(
	"cannot import '%s': more than "
	"one matching pool\n"),
	import->poolname);
	(void) fprintf(stderr, gettext(
	"import by numeric ID instead\n"));
	err = B_TRUE;
	}
	found_config = config;
	}
	} else {
	uint64_t guid;

	/*
	* Search for a pool by guid.
	*/
	verify(nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_POOL_GUID, &guid) == 0);

	if (guid == import->guid)
	found_config = config;
	}
	}

	/*
	* If we were searching for a specific pool, verify that we found a
	* pool, and then do the import.
	*/
	if (pool_specified && err == 0) {
	if (found_config == NULL) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"no such pool available\n"), orig_name);
	err = B_TRUE;
	} else {
	err \|= do_import(found_config, new_name,
	mntopts, props, flags);
	}
	}

	/*
	* If we were just looking for pools, report an error if none were
	* found.
	*/
	if (!pool_specified && first)
	(void) fprintf(stderr,
	gettext("no pools available to import\n"));
	return (err);
	}

	typedef struct target_exists_args {
	const char *poolname;
	uint64_t poolguid;
	} target_exists_args_t;

	static int
	name_or_guid_exists(zpool_handle_t zhp, void data)
	{
	target_exists_args_t *args = data;
	nvlist_t *config = zpool_get_config(zhp, NULL);
	int found = 0;

	if (config == NULL)
	return (0);

	if (args->poolname != NULL) {
	char *pool_name;

	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&pool_name) == 0);
	if (strcmp(pool_name, args->poolname) == 0)
	found = 1;
	} else {
	uint64_t pool_guid;

	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&pool_guid) == 0);
	if (pool_guid == args->poolguid)
	found = 1;
	}
	zpool_close(zhp);

	return (found);
	}
	/*
	* zpool checkpoint <pool>
	* checkpoint --discard <pool>
	*
	* -d Discard the checkpoint from a checkpointed
	* --discard pool.
	*
	* -w Wait for discarding a checkpoint to complete.
	* --wait
	*
	* Checkpoints the specified pool, by taking a "snapshot" of its
	* current state. A pool can only have one checkpoint at a time.
	*/
	int
	zpool_do_checkpoint(int argc, char **argv)
	{
	boolean_t discard, wait;
	char *pool;
	zpool_handle_t *zhp;
	int c, err;

	struct option long_options[] = {
	{"discard", no_argument, NULL, 'd'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	discard = B_FALSE;
	wait = B_FALSE;
	while ((c = getopt_long(argc, argv, ":dw", long_options, NULL)) != -1) {
	switch (c) {
	case 'd':
	discard = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (wait && !discard) {
	(void) fprintf(stderr, gettext("--wait only valid when "
	"--discard also specified\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	pool = argv[0];

	if ((zhp = zpool_open(g_zfs, pool)) == NULL) {
	/* As a special case, check for use of '/' in the name */
	if (strchr(pool, '/') != NULL)
	(void) fprintf(stderr, gettext("'zpool checkpoint' "
	"doesn't work on datasets. To save the state "
	"of a dataset from a specific point in time "
	"please use 'zfs snapshot'\n"));
	return (1);
	}

	if (discard) {
	err = (zpool_discard_checkpoint(zhp) != 0);
	if (err == 0 && wait)
	err = zpool_wait(zhp, ZPOOL_WAIT_CKPT_DISCARD);
	} else {
	err = (zpool_checkpoint(zhp) != 0);
	}

	zpool_close(zhp);

	return (err);
	}

	#define CHECKPOINT_OPT 1024

	/*
	* zpool import [-d dir] [-D]
	* import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
	* [-d dir \| -c cachefile \| -s] [-f] -a
	* import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
	* [-d dir \| -c cachefile \| -s] [-f] [-n] [-F] <pool \| id>
	* [newpool]
	*
	* -c Read pool information from a cachefile instead of searching
	* devices. If importing from a cachefile config fails, then
	* fallback to searching for devices only in the directories that
	* exist in the cachefile.
	*
	* -d Scan in a specific directory, other than /dev/. More than
	* one directory can be specified using multiple '-d' options.
	*
	* -D Scan for previously destroyed pools or import all or only
	* specified destroyed pools.
	*
	* -R Temporarily import the pool, with all mountpoints relative to
	* the given root. The pool will remain exported when the machine
	* is rebooted.
	*
	* -V Import even in the presence of faulted vdevs. This is an
	* intentionally undocumented option for testing purposes, and
	* treats the pool configuration as complete, leaving any bad
	* vdevs in the FAULTED state. In other words, it does verbatim
	* import.
	*
	* -f Force import, even if it appears that the pool is active.
	*
	* -F Attempt rewind if necessary.
	*
	* -n See if rewind would work, but don't actually rewind.
	*
	* -N Import the pool but don't mount datasets.
	*
	* -T Specify a starting txg to use for import. This option is
	* intentionally undocumented option for testing purposes.
	*
	* -a Import all pools found.
	*
	* -l Load encryption keys while importing.
	*
	* -o Set property=value and/or temporary mount options (without '=').
	*
	* -s Scan using the default search path, the libblkid cache will
	* not be consulted.
	*
	* --rewind-to-checkpoint
	* Import the pool and revert back to the checkpoint.
	*
	* The import command scans for pools to import, and import pools based on pool
	* name and GUID. The pool can also be renamed as part of the import process.
	*/
	int
	zpool_do_import(int argc, char **argv)
	{
	char **searchdirs = NULL;
	char env, envdup = NULL;
	int nsearch = 0;
	int c;
	int err = 0;
	nvlist_t *pools = NULL;
	boolean_t do_all = B_FALSE;
	boolean_t do_destroyed = B_FALSE;
	char *mntopts = NULL;
	uint64_t searchguid = 0;
	char *searchname = NULL;
	char *propval;
	nvlist_t *policy = NULL;
	nvlist_t *props = NULL;
	int flags = ZFS_IMPORT_NORMAL;
	uint32_t rewind_policy = ZPOOL_NO_REWIND;
	boolean_t dryrun = B_FALSE;
	boolean_t do_rewind = B_FALSE;
	boolean_t xtreme_rewind = B_FALSE;
	boolean_t do_scan = B_FALSE;
	boolean_t pool_exists = B_FALSE;
	boolean_t pool_specified = B_FALSE;
	uint64_t txg = -1ULL;
	char *cachefile = NULL;
	importargs_t idata = { 0 };
	char *endptr;

	struct option long_options[] = {
	{"rewind-to-checkpoint", no_argument, NULL, CHECKPOINT_OPT},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, ":aCc:d:DEfFlmnNo:R:stT:VX",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'a':
	do_all = B_TRUE;
	break;
	case 'c':
	cachefile = optarg;
	break;
	case 'd':
	searchdirs = safe_realloc(searchdirs,
	(nsearch + 1) * sizeof (char *));
	searchdirs[nsearch++] = optarg;
	break;
	case 'D':
	do_destroyed = B_TRUE;
	break;
	case 'f':
	flags \|= ZFS_IMPORT_ANY_HOST;
	break;
	case 'F':
	do_rewind = B_TRUE;
	break;
	case 'l':
	flags \|= ZFS_IMPORT_LOAD_KEYS;
	break;
	case 'm':
	flags \|= ZFS_IMPORT_MISSING_LOG;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'N':
	flags \|= ZFS_IMPORT_ONLY;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) != NULL) {
	*propval = '\0';
	propval++;
	if (add_prop_list(optarg, propval,
	&props, B_TRUE))
	goto error;
	} else {
	mntopts = optarg;
	}
	break;
	case 'R':
	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
	goto error;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto error;
	break;
	case 's':
	do_scan = B_TRUE;
	break;
	case 't':
	flags \|= ZFS_IMPORT_TEMP_NAME;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto error;
	break;

	case 'T':
	errno = 0;
	txg = strtoull(optarg, &endptr, 0);
	if (errno != 0 \|\| *endptr != '\0') {
	(void) fprintf(stderr,
	gettext("invalid txg value\n"));
	usage(B_FALSE);
	}
	rewind_policy = ZPOOL_DO_REWIND \| ZPOOL_EXTREME_REWIND;
	break;
	case 'V':
	flags \|= ZFS_IMPORT_VERBATIM;
	break;
	case 'X':
	xtreme_rewind = B_TRUE;
	break;
	case CHECKPOINT_OPT:
	flags \|= ZFS_IMPORT_CHECKPOINT;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (cachefile && nsearch != 0) {
	(void) fprintf(stderr, gettext("-c is incompatible with -d\n"));
	usage(B_FALSE);
	}

	if (cachefile && do_scan) {
	(void) fprintf(stderr, gettext("-c is incompatible with -s\n"));
	usage(B_FALSE);
	}

	if ((flags & ZFS_IMPORT_LOAD_KEYS) && (flags & ZFS_IMPORT_ONLY)) {
	(void) fprintf(stderr, gettext("-l is incompatible with -N\n"));
	usage(B_FALSE);
	}

	if ((flags & ZFS_IMPORT_LOAD_KEYS) && !do_all && argc == 0) {
	(void) fprintf(stderr, gettext("-l is only meaningful during "
	"an import\n"));
	usage(B_FALSE);
	}

	if ((dryrun \|\| xtreme_rewind) && !do_rewind) {
	(void) fprintf(stderr,
	gettext("-n or -X only meaningful with -F\n"));
	usage(B_FALSE);
	}
	if (dryrun)
	rewind_policy = ZPOOL_TRY_REWIND;
	else if (do_rewind)
	rewind_policy = ZPOOL_DO_REWIND;
	if (xtreme_rewind)
	rewind_policy \|= ZPOOL_EXTREME_REWIND;

	/* In the future, we can capture further policy and include it here */
	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 \|\|
	nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, txg) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
	rewind_policy) != 0)
	goto error;

	/* check argument count */
	if (do_all) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	}

	/*
	* Check for the effective uid. We do this explicitly here because
	* otherwise any attempt to discover pools will silently fail.
	*/
	if (argc == 0 && geteuid() != 0) {
	(void) fprintf(stderr, gettext("cannot "
	"discover pools: permission denied\n"));

	free(searchdirs);
	nvlist_free(props);
	nvlist_free(policy);
	return (1);
	}

	/*
	* Depending on the arguments given, we do one of the following:
	*
	* <none> Iterate through all pools and display information about
	* each one.
	*
	* -a Iterate through all pools and try to import each one.
	*
	* <id> Find the pool that corresponds to the given GUID/pool
	* name and import that one.
	*
	* -D Above options applies only to destroyed pools.
	*/
	if (argc != 0) {
	char *endptr;

	errno = 0;
	searchguid = strtoull(argv[0], &endptr, 10);
	if (errno != 0 \|\| *endptr != '\0') {
	searchname = argv[0];
	searchguid = 0;
	}
	pool_specified = B_TRUE;

	/*
	* User specified a name or guid. Ensure it's unique.
	*/
	target_exists_args_t search = {searchname, searchguid};
	pool_exists = zpool_iter(g_zfs, name_or_guid_exists, &search);
	}

	/*
	* Check the environment for the preferred search path.
	*/
	if ((searchdirs == NULL) && (env = getenv("ZPOOL_IMPORT_PATH"))) {
	char dir, tmp = NULL;

	envdup = strdup(env);

	for (dir = strtok_r(envdup, ":", &tmp);
	dir != NULL;
	dir = strtok_r(NULL, ":", &tmp)) {
	searchdirs = safe_realloc(searchdirs,
	(nsearch + 1) * sizeof (char *));
	searchdirs[nsearch++] = dir;
	}
	}

	idata.path = searchdirs;
	idata.paths = nsearch;
	idata.poolname = searchname;
	idata.guid = searchguid;
	idata.cachefile = cachefile;
	idata.scan = do_scan;
	idata.policy = policy;

	pools = zpool_search_import(g_zfs, &idata, &libzfs_config_ops);

	if (pools != NULL && pool_exists &&
	(argc == 1 \|\| strcmp(argv[0], argv[1]) == 0)) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"a pool with that name already exists\n"),
	argv[0]);
	(void) fprintf(stderr, gettext("use the form '%s "
	"<pool \| id> <newpool>' to give it a new name\n"),
	"zpool import");
	err = 1;
	} else if (pools == NULL && pool_exists) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"a pool with that name is already created/imported,\n"),
	argv[0]);
	(void) fprintf(stderr, gettext("and no additional pools "
	"with that name were found\n"));
	err = 1;
	} else if (pools == NULL) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"no such pool available\n"), argv[0]);
	}
	err = 1;
	}

	if (err == 1) {
	free(searchdirs);
	free(envdup);
	nvlist_free(policy);
	nvlist_free(pools);
	nvlist_free(props);
	return (1);
	}

	err = import_pools(pools, props, mntopts, flags,
	argc >= 1 ? argv[0] : NULL,
	argc >= 2 ? argv[1] : NULL,
	do_destroyed, pool_specified, do_all, &idata);

	/*
	* If we're using the cachefile and we failed to import, then
	* fallback to scanning the directory for pools that match
	* those in the cachefile.
	*/
	if (err != 0 && cachefile != NULL) {
	(void) printf(gettext("cachefile import failed, retrying\n"));

	/*
	* We use the scan flag to gather the directories that exist
	* in the cachefile. If we need to fallback to searching for
	* the pool config, we will only search devices in these
	* directories.
	*/
	idata.scan = B_TRUE;
	nvlist_free(pools);
	pools = zpool_search_import(g_zfs, &idata, &libzfs_config_ops);

	err = import_pools(pools, props, mntopts, flags,
	argc >= 1 ? argv[0] : NULL,
	argc >= 2 ? argv[1] : NULL,
	do_destroyed, pool_specified, do_all, &idata);
	}

	error:
	nvlist_free(props);
	nvlist_free(pools);
	nvlist_free(policy);
	free(searchdirs);
	free(envdup);

	return (err ? 1 : 0);
	}

	/*
	* zpool sync [-f] [pool] ...
	*
	* -f (undocumented) force uberblock (and config including zpool cache file)
	* update.
	*
	* Sync the specified pool(s).
	* Without arguments "zpool sync" will sync all pools.
	* This command initiates TXG sync(s) and will return after the TXG(s) commit.
	*
	*/
	static int
	zpool_do_sync(int argc, char **argv)
	{
	int ret;
	boolean_t force = B_FALSE;

	/* check options */
	while ((ret = getopt(argc, argv, "f")) != -1) {
	switch (ret) {
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* if argc == 0 we will execute zpool_sync_one on all pools */
	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_sync_one, &force);

	return (ret);
	}

	typedef struct iostat_cbdata {
	uint64_t cb_flags;
	int cb_namewidth;
	int cb_iteration;
	boolean_t cb_verbose;
	boolean_t cb_literal;
	boolean_t cb_scripted;
	zpool_list_t *cb_list;
	vdev_cmd_data_list_t *vcdl;
	vdev_cbdata_t cb_vdevs;
	} iostat_cbdata_t;

	/* iostat labels */
	typedef struct name_and_columns {
	const char name; / Column name */
	unsigned int columns; /* Center name to this number of columns */
	} name_and_columns_t;

	#define IOSTAT_MAX_LABELS 15 /* Max number of labels on one line */

	static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] =
	{
	[IOS_DEFAULT] = {{"capacity", 2}, {"operations", 2}, {"bandwidth", 2},
	{NULL}},
	[IOS_LATENCY] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
	{"asyncq_wait", 2}, {"scrub", 1}, {"trim", 1}, {"rebuild", 1},
	{NULL}},
	[IOS_QUEUES] = {{"syncq_read", 2}, {"syncq_write", 2},
	{"asyncq_read", 2}, {"asyncq_write", 2}, {"scrubq_read", 2},
	{"trimq_write", 2}, {"rebuildq_write", 2}, {NULL}},
	[IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
	{"asyncq_wait", 2}, {NULL}},
	[IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2},
	{"async_read", 2}, {"async_write", 2}, {"scrub", 2},
	{"trim", 2}, {"rebuild", 2}, {NULL}},
	};

	/* Shorthand - if "columns" field not set, default to 1 column */
	static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] =
	{
	[IOS_DEFAULT] = {{"alloc"}, {"free"}, {"read"}, {"write"}, {"read"},
	{"write"}, {NULL}},
	[IOS_LATENCY] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
	{"write"}, {"read"}, {"write"}, {"wait"}, {"wait"}, {"wait"},
	{NULL}},
	[IOS_QUEUES] = {{"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"},
	{"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"},
	{"pend"}, {"activ"}, {"pend"}, {"activ"}, {NULL}},
	[IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
	{"write"}, {"read"}, {"write"}, {"scrub"}, {"trim"}, {"rebuild"},
	{NULL}},
	[IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
	{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
	{"ind"}, {"agg"}, {NULL}},
	};

	static const char *histo_to_title[] = {
	[IOS_L_HISTO] = "latency",
	[IOS_RQ_HISTO] = "req_size",
	};

	/*
	* Return the number of labels in a null-terminated name_and_columns_t
	* array.
	*
	*/
	static unsigned int
	label_array_len(const name_and_columns_t *labels)
	{
	int i = 0;

	while (labels[i].name)
	i++;

	return (i);
	}

	/*
	* Return the number of strings in a null-terminated string array.
	* For example:
	*
	* const char foo[] = {"bar", "baz", NULL}
	*
	* returns 2
	*/
	static uint64_t
	str_array_len(const char *array[])
	{
	uint64_t i = 0;
	while (array[i])
	i++;

	return (i);
	}


	/*
	* Return a default column width for default/latency/queue columns. This does
	* not include histograms, which have their columns autosized.
	*/
	static unsigned int
	default_column_width(iostat_cbdata_t *cb, enum iostat_type type)
	{
	unsigned long column_width = 5; /* Normal niceprint */
	static unsigned long widths[] = {
	/*
	* Choose some sane default column sizes for printing the
	* raw numbers.
	*/
	[IOS_DEFAULT] = 15, /* 1PB capacity */
	[IOS_LATENCY] = 10, /* 1B ns = 10sec */
	[IOS_QUEUES] = 6, /* 1M queue entries */
	[IOS_L_HISTO] = 10, /* 1B ns = 10sec */
	[IOS_RQ_HISTO] = 6, /* 1M queue entries */
	};

	if (cb->cb_literal)
	column_width = widths[type];

	return (column_width);
	}

	/*
	* Print the column labels, i.e:
	*
	* capacity operations bandwidth
	* alloc free read write read write ...
	*
	* If force_column_width is set, use it for the column width. If not set, use
	* the default column width.
	*/
	static void
	print_iostat_labels(iostat_cbdata_t *cb, unsigned int force_column_width,
	const name_and_columns_t labels[][IOSTAT_MAX_LABELS])
	{
	int i, idx, s;
	int text_start, rw_column_width, spaces_to_end;
	uint64_t flags = cb->cb_flags;
	uint64_t f;
	unsigned int column_width = force_column_width;

	/* For each bit set in flags */
	for (f = flags; f; f &= ~(1ULL << idx)) {
	idx = lowbit64(f) - 1;
	if (!force_column_width)
	column_width = default_column_width(cb, idx);
	/* Print our top labels centered over "read write" label. */
	for (i = 0; i < label_array_len(labels[idx]); i++) {
	const char *name = labels[idx][i].name;
	/*
	* We treat labels[][].columns == 0 as shorthand
	* for one column. It makes writing out the label
	* tables more concise.
	*/
	unsigned int columns = MAX(1, labels[idx][i].columns);
	unsigned int slen = strlen(name);

	rw_column_width = (column_width * columns) +
	(2 * (columns - 1));

	text_start = (int)((rw_column_width) / columns -
	slen / columns);
	if (text_start < 0)
	text_start = 0;

	printf(" "); /* Two spaces between columns */

	/* Space from beginning of column to label */
	for (s = 0; s < text_start; s++)
	printf(" ");

	printf("%s", name);

	/* Print space after label to end of column */
	spaces_to_end = rw_column_width - text_start - slen;
	if (spaces_to_end < 0)
	spaces_to_end = 0;

	for (s = 0; s < spaces_to_end; s++)
	printf(" ");
	}
	}
	}


	/*
	* print_cmd_columns - Print custom column titles from -c
	*
	* If the user specified the "zpool status\|iostat -c" then print their custom
	* column titles in the header. For example, print_cmd_columns() would print
	* the " col1 col2" part of this:
	*
	* $ zpool iostat -vc 'echo col1=val1; echo col2=val2'
	* ...
	* capacity operations bandwidth
	* pool alloc free read write read write col1 col2
	* ---------- ----- ----- ----- ----- ----- ----- ---- ----
	* mypool 269K 1008M 0 0 107 946
	* mirror 269K 1008M 0 0 107 946
	* sdb - - 0 0 102 473 val1 val2
	* sdc - - 0 0 5 473 val1 val2
	* ---------- ----- ----- ----- ----- ----- ----- ---- ----
	*/
	static void
	print_cmd_columns(vdev_cmd_data_list_t *vcdl, int use_dashes)
	{
	int i, j;
	vdev_cmd_data_t *data = &vcdl->data[0];

	if (vcdl->count == 0 \|\| data == NULL)
	return;

	/*
	* Each vdev cmd should have the same column names unless the user did
	* something weird with their cmd. Just take the column names from the
	* first vdev and assume it works for all of them.
	*/
	for (i = 0; i < vcdl->uniq_cols_cnt; i++) {
	printf(" ");
	if (use_dashes) {
	for (j = 0; j < vcdl->uniq_cols_width[i]; j++)
	printf("-");
	} else {
	printf_color(ANSI_BOLD, "%*s", vcdl->uniq_cols_width[i],
	vcdl->uniq_cols[i]);
	}
	}
	}


	/*
	* Utility function to print out a line of dashes like:
	*
	* -------------------------------- ----- ----- ----- ----- -----
	*
	* ...or a dashed named-row line like:
	*
	* logs - - - - -
	*
	* @cb: iostat data
	*
	* @force_column_width If non-zero, use the value as the column width.
	* Otherwise use the default column widths.
	*
	* @name: Print a dashed named-row line starting
	* with @name. Otherwise, print a regular
	* dashed line.
	*/
	static void
	print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width,
	const char *name)
	{
	int i;
	unsigned int namewidth;
	uint64_t flags = cb->cb_flags;
	uint64_t f;
	int idx;
	const name_and_columns_t *labels;
	const char *title;


	if (cb->cb_flags & IOS_ANYHISTO_M) {
	title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
	} else if (cb->cb_vdevs.cb_names_count) {
	title = "vdev";
	} else {
	title = "pool";
	}

	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
	name ? strlen(name) : 0);


	if (name) {
	printf("%-*s", namewidth, name);
	} else {
	for (i = 0; i < namewidth; i++)
	(void) printf("-");
	}

	/* For each bit in flags */
	for (f = flags; f; f &= ~(1ULL << idx)) {
	unsigned int column_width;
	idx = lowbit64(f) - 1;
	if (force_column_width)
	column_width = force_column_width;
	else
	column_width = default_column_width(cb, idx);

	labels = iostat_bottom_labels[idx];
	for (i = 0; i < label_array_len(labels); i++) {
	if (name)
	printf(" %*s-", column_width - 1, " ");
	else
	printf(" %.*s", column_width,
	"--------------------");
	}
	}
	}


	static void
	print_iostat_separator_impl(iostat_cbdata_t *cb,
	unsigned int force_column_width)
	{
	print_iostat_dashes(cb, force_column_width, NULL);
	}

	static void
	print_iostat_separator(iostat_cbdata_t *cb)
	{
	print_iostat_separator_impl(cb, 0);
	}

	static void
	print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width,
	const char *histo_vdev_name)
	{
	unsigned int namewidth;
	const char *title;

	if (cb->cb_flags & IOS_ANYHISTO_M) {
	title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
	} else if (cb->cb_vdevs.cb_names_count) {
	title = "vdev";
	} else {
	title = "pool";
	}

	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
	histo_vdev_name ? strlen(histo_vdev_name) : 0);

	if (histo_vdev_name)
	printf("%-*s", namewidth, histo_vdev_name);
	else
	printf("%*s", namewidth, "");


	print_iostat_labels(cb, force_column_width, iostat_top_labels);
	printf("\n");

	printf("%-*s", namewidth, title);

	print_iostat_labels(cb, force_column_width, iostat_bottom_labels);
	if (cb->vcdl != NULL)
	print_cmd_columns(cb->vcdl, 0);

	printf("\n");

	print_iostat_separator_impl(cb, force_column_width);

	if (cb->vcdl != NULL)
	print_cmd_columns(cb->vcdl, 1);

	printf("\n");
	}

	static void
	print_iostat_header(iostat_cbdata_t *cb)
	{
	print_iostat_header_impl(cb, 0, NULL);
	}


	/*
	* Display a single statistic.
	*/
	static void
	print_one_stat(uint64_t value, enum zfs_nicenum_format format,
	unsigned int column_size, boolean_t scripted)
	{
	char buf[64];

	zfs_nicenum_format(value, buf, sizeof (buf), format);

	if (scripted)
	printf("\t%s", buf);
	else
	printf(" %*s", column_size, buf);
	}

	/*
	* Calculate the default vdev stats
	*
	* Subtract oldvs from newvs, apply a scaling factor, and save the resulting
	* stats into calcvs.
	*/
	static void
	calc_default_iostats(vdev_stat_t oldvs, vdev_stat_t newvs,
	vdev_stat_t *calcvs)
	{
	int i;

	memcpy(calcvs, newvs, sizeof (*calcvs));
	for (i = 0; i < ARRAY_SIZE(calcvs->vs_ops); i++)
	calcvs->vs_ops[i] = (newvs->vs_ops[i] - oldvs->vs_ops[i]);

	for (i = 0; i < ARRAY_SIZE(calcvs->vs_bytes); i++)
	calcvs->vs_bytes[i] = (newvs->vs_bytes[i] - oldvs->vs_bytes[i]);
	}

	/*
	* Internal representation of the extended iostats data.
	*
	* The extended iostat stats are exported in nvlists as either uint64_t arrays
	* or single uint64_t's. We make both look like arrays to make them easier
	* to process. In order to make single uint64_t's look like arrays, we set
	* __data to the stat data, and then set *data = &__data with count = 1. Then,
	* we can just use *data and count.
	*/
	struct stat_array {
	uint64_t *data;
	uint_t count; /* Number of entries in data[] */
	uint64_t __data; /* Only used when data is a single uint64_t */
	};

	static uint64_t
	stat_histo_max(struct stat_array *nva, unsigned int len)
	{
	uint64_t max = 0;
	int i;
	for (i = 0; i < len; i++)
	max = MAX(max, array64_max(nva[i].data, nva[i].count));

	return (max);
	}

	/*
	* Helper function to lookup a uint64_t array or uint64_t value and store its
	* data as a stat_array. If the nvpair is a single uint64_t value, then we make
	* it look like a one element array to make it easier to process.
	*/
	static int
	nvpair64_to_stat_array(nvlist_t nvl, const char name,
	struct stat_array *nva)
	{
	nvpair_t *tmp;
	int ret;

	verify(nvlist_lookup_nvpair(nvl, name, &tmp) == 0);
	switch (nvpair_type(tmp)) {
	case DATA_TYPE_UINT64_ARRAY:
	ret = nvpair_value_uint64_array(tmp, &nva->data, &nva->count);
	break;
	case DATA_TYPE_UINT64:
	ret = nvpair_value_uint64(tmp, &nva->__data);
	nva->data = &nva->__data;
	nva->count = 1;
	break;
	default:
	/* Not a uint64_t */
	ret = EINVAL;
	break;
	}

	return (ret);
	}

	/*
	* Given a list of nvlist names, look up the extended stats in newnv and oldnv,
	* subtract them, and return the results in a newly allocated stat_array.
	* You must free the returned array after you are done with it with
	* free_calc_stats().
	*
	* Additionally, you can set "oldnv" to NULL if you simply want the newnv
	* values.
	*/
	static struct stat_array *
	calc_and_alloc_stats_ex(const char *names, unsigned int len, nvlist_t oldnv,
	nvlist_t *newnv)
	{
	nvlist_t oldnvx = NULL, newnvx;
	struct stat_array oldnva, newnva, *calcnva;
	int i, j;
	unsigned int alloc_size = (sizeof (struct stat_array)) * len;

	/* Extract our extended stats nvlist from the main list */
	verify(nvlist_lookup_nvlist(newnv, ZPOOL_CONFIG_VDEV_STATS_EX,
	&newnvx) == 0);
	if (oldnv) {
	verify(nvlist_lookup_nvlist(oldnv, ZPOOL_CONFIG_VDEV_STATS_EX,
	&oldnvx) == 0);
	}

	newnva = safe_malloc(alloc_size);
	oldnva = safe_malloc(alloc_size);
	calcnva = safe_malloc(alloc_size);

	for (j = 0; j < len; j++) {
	verify(nvpair64_to_stat_array(newnvx, names[j],
	&newnva[j]) == 0);
	calcnva[j].count = newnva[j].count;
	alloc_size = calcnva[j].count * sizeof (calcnva[j].data[0]);
	calcnva[j].data = safe_malloc(alloc_size);
	memcpy(calcnva[j].data, newnva[j].data, alloc_size);

	if (oldnvx) {
	verify(nvpair64_to_stat_array(oldnvx, names[j],
	&oldnva[j]) == 0);
	for (i = 0; i < oldnva[j].count; i++)
	calcnva[j].data[i] -= oldnva[j].data[i];
	}
	}
	free(newnva);
	free(oldnva);
	return (calcnva);
	}

	static void
	free_calc_stats(struct stat_array *nva, unsigned int len)
	{
	int i;
	for (i = 0; i < len; i++)
	free(nva[i].data);

	free(nva);
	}

	static void
	print_iostat_histo(struct stat_array *nva, unsigned int len,
	iostat_cbdata_t *cb, unsigned int column_width, unsigned int namewidth,
	double scale)
	{
	int i, j;
	char buf[6];
	uint64_t val;
	enum zfs_nicenum_format format;
	unsigned int buckets;
	unsigned int start_bucket;

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	/* All these histos are the same size, so just use nva[0].count */
	buckets = nva[0].count;

	if (cb->cb_flags & IOS_RQ_HISTO_M) {
	/* Start at 512 - req size should never be lower than this */
	start_bucket = 9;
	} else {
	start_bucket = 0;
	}

	for (j = start_bucket; j < buckets; j++) {
	/* Print histogram bucket label */
	if (cb->cb_flags & IOS_L_HISTO_M) {
	/* Ending range of this bucket */
	val = (1UL << (j + 1)) - 1;
	zfs_nicetime(val, buf, sizeof (buf));
	} else {
	/* Request size (starting range of bucket) */
	val = (1UL << j);
	zfs_nicenum(val, buf, sizeof (buf));
	}

	if (cb->cb_scripted)
	printf("%llu", (u_longlong_t)val);
	else
	printf("%-*s", namewidth, buf);

	/* Print the values on the line */
	for (i = 0; i < len; i++) {
	print_one_stat(nva[i].data[j] * scale, format,
	column_width, cb->cb_scripted);
	}
	printf("\n");
	}
	}

	static void
	print_solid_separator(unsigned int length)
	{
	while (length--)
	printf("-");
	printf("\n");
	}

	static void
	print_iostat_histos(iostat_cbdata_t cb, nvlist_t oldnv,
	nvlist_t newnv, double scale, const char name)
	{
	unsigned int column_width;
	unsigned int namewidth;
	unsigned int entire_width;
	enum iostat_type type;
	struct stat_array *nva;
	const char **names;
	unsigned int names_len;

	/* What type of histo are we? */
	type = IOS_HISTO_IDX(cb->cb_flags);

	/* Get NULL-terminated array of nvlist names for our histo */
	names = vsx_type_to_nvlist[type];
	names_len = str_array_len(names); /* num of names */

	nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv);

	if (cb->cb_literal) {
	column_width = MAX(5,
	(unsigned int) log10(stat_histo_max(nva, names_len)) + 1);
	} else {
	column_width = 5;
	}

	namewidth = MAX(cb->cb_namewidth,
	strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]));

	/*
	* Calculate the entire line width of what we're printing. The
	* +2 is for the two spaces between columns:
	*/
	/* read write */
	/* ----- ----- */
	/* \|___\| <---------- column_width */
	/* */
	/* \|__________\| <--- entire_width */
	/* */
	entire_width = namewidth + (column_width + 2) *
	label_array_len(iostat_bottom_labels[type]);

	if (cb->cb_scripted)
	printf("%s\n", name);
	else
	print_iostat_header_impl(cb, column_width, name);

	print_iostat_histo(nva, names_len, cb, column_width,
	namewidth, scale);

	free_calc_stats(nva, names_len);
	if (!cb->cb_scripted)
	print_solid_separator(entire_width);
	}

	/*
	* Calculate the average latency of a power-of-two latency histogram
	*/
	static uint64_t
	single_histo_average(uint64_t *histo, unsigned int buckets)
	{
	int i;
	uint64_t count = 0, total = 0;

	for (i = 0; i < buckets; i++) {
	/*
	* Our buckets are power-of-two latency ranges. Use the
	* midpoint latency of each bucket to calculate the average.
	* For example:
	*
	* Bucket Midpoint
	* 8ns-15ns: 12ns
	* 16ns-31ns: 24ns
	* ...
	*/
	if (histo[i] != 0) {
	total += histo[i] * (((1UL << i) + ((1UL << i)/2)));
	count += histo[i];
	}
	}

	/* Prevent divide by zero */
	return (count == 0 ? 0 : total / count);
	}

	static void
	print_iostat_queues(iostat_cbdata_t cb, nvlist_t newnv)
	{
	const char *names[] = {
	ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE,
	};

	struct stat_array *nva;

	unsigned int column_width = default_column_width(cb, IOS_QUEUES);
	enum zfs_nicenum_format format;

	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), NULL, newnv);

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	for (int i = 0; i < ARRAY_SIZE(names); i++) {
	uint64_t val = nva[i].data[0];
	print_one_stat(val, format, column_width, cb->cb_scripted);
	}

	free_calc_stats(nva, ARRAY_SIZE(names));
	}

	static void
	print_iostat_latency(iostat_cbdata_t cb, nvlist_t oldnv,
	nvlist_t *newnv)
	{
	int i;
	uint64_t val;
	const char *names[] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	};
	struct stat_array *nva;

	unsigned int column_width = default_column_width(cb, IOS_LATENCY);
	enum zfs_nicenum_format format;

	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv);

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAWTIME;
	else
	format = ZFS_NICENUM_TIME;

	/* Print our avg latencies on the line */
	for (i = 0; i < ARRAY_SIZE(names); i++) {
	/* Compute average latency for a latency histo */
	val = single_histo_average(nva[i].data, nva[i].count);
	print_one_stat(val, format, column_width, cb->cb_scripted);
	}
	free_calc_stats(nva, ARRAY_SIZE(names));
	}

	/*
	* Print default statistics (capacity/operations/bandwidth)
	*/
	static void
	print_iostat_default(vdev_stat_t vs, iostat_cbdata_t cb, double scale)
	{
	unsigned int column_width = default_column_width(cb, IOS_DEFAULT);
	enum zfs_nicenum_format format;
	char na; /* char to print for "not applicable" values */

	if (cb->cb_literal) {
	format = ZFS_NICENUM_RAW;
	na = '0';
	} else {
	format = ZFS_NICENUM_1024;
	na = '-';
	}

	/* only toplevel vdevs have capacity stats */
	if (vs->vs_space == 0) {
	if (cb->cb_scripted)
	printf("\t%c\t%c", na, na);
	else
	printf(" %c %c", column_width, na, column_width,
	na);
	} else {
	print_one_stat(vs->vs_alloc, format, column_width,
	cb->cb_scripted);
	print_one_stat(vs->vs_space - vs->vs_alloc, format,
	column_width, cb->cb_scripted);
	}

	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_READ] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_WRITE] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_READ] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_WRITE] * scale),
	format, column_width, cb->cb_scripted);
	}

	static const char *const class_name[] = {
	VDEV_ALLOC_BIAS_DEDUP,
	VDEV_ALLOC_BIAS_SPECIAL,
	VDEV_ALLOC_CLASS_LOGS
	};

	/*
	* Print out all the statistics for the given vdev. This can either be the
	* toplevel configuration, or called recursively. If 'name' is NULL, then this
	* is a verbose output, and we don't want to display the toplevel pool stats.
	*
	* Returns the number of stat lines printed.
	*/
	static unsigned int
	print_vdev_stats(zpool_handle_t zhp, const char name, nvlist_t *oldnv,
	nvlist_t newnv, iostat_cbdata_t cb, int depth)
	{
	nvlist_t oldchild, newchild;
	uint_t c, children, oldchildren;
	vdev_stat_t oldvs, newvs, *calcvs;
	vdev_stat_t zerovs = { 0 };
	char *vname;
	int i;
	int ret = 0;
	uint64_t tdelta;
	double scale;

	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
	return (ret);

	calcvs = safe_malloc(sizeof (*calcvs));

	if (oldnv != NULL) {
	verify(nvlist_lookup_uint64_array(oldnv,
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&oldvs, &c) == 0);
	} else {
	oldvs = &zerovs;
	}

	/* Do we only want to see a specific vdev? */
	for (i = 0; i < cb->cb_vdevs.cb_names_count; i++) {
	/* Yes we do. Is this the vdev? */
	if (strcmp(name, cb->cb_vdevs.cb_names[i]) == 0) {
	/*
	* This is our vdev. Since it is the only vdev we
	* will be displaying, make depth = 0 so that it
	* doesn't get indented.
	*/
	depth = 0;
	break;
	}
	}

	if (cb->cb_vdevs.cb_names_count && (i == cb->cb_vdevs.cb_names_count)) {
	/* Couldn't match the name */
	goto children;
	}


	verify(nvlist_lookup_uint64_array(newnv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&newvs, &c) == 0);

	/*
	* Print the vdev name unless it's is a histogram. Histograms
	* display the vdev name in the header itself.
	*/
	if (!(cb->cb_flags & IOS_ANYHISTO_M)) {
	if (cb->cb_scripted) {
	printf("%s", name);
	} else {
	if (strlen(name) + depth > cb->cb_namewidth)
	(void) printf("%*s%s", depth, "", name);
	else
	(void) printf("%s%s%s", depth, "", name,
	(int)(cb->cb_namewidth - strlen(name) -
	depth), "");
	}
	}

	/* Calculate our scaling factor */
	tdelta = newvs->vs_timestamp - oldvs->vs_timestamp;
	if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) {
	/*
	* If we specify printing histograms with no time interval, then
	* print the histogram numbers over the entire lifetime of the
	* vdev.
	*/
	scale = 1;
	} else {
	if (tdelta == 0)
	scale = 1.0;
	else
	scale = (double)NANOSEC / tdelta;
	}

	if (cb->cb_flags & IOS_DEFAULT_M) {
	calc_default_iostats(oldvs, newvs, calcvs);
	print_iostat_default(calcvs, cb, scale);
	}
	if (cb->cb_flags & IOS_LATENCY_M)
	print_iostat_latency(cb, oldnv, newnv);
	if (cb->cb_flags & IOS_QUEUES_M)
	print_iostat_queues(cb, newnv);
	if (cb->cb_flags & IOS_ANYHISTO_M) {
	printf("\n");
	print_iostat_histos(cb, oldnv, newnv, scale, name);
	}

	if (cb->vcdl != NULL) {
	char *path;
	if (nvlist_lookup_string(newnv, ZPOOL_CONFIG_PATH,
	&path) == 0) {
	printf(" ");
	zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
	}
	}

	if (!(cb->cb_flags & IOS_ANYHISTO_M))
	printf("\n");

	ret++;

	children:

	free(calcvs);

	if (!cb->cb_verbose)
	return (ret);

	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_CHILDREN,
	&newchild, &children) != 0)
	return (ret);

	if (oldnv) {
	if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_CHILDREN,
	&oldchild, &oldchildren) != 0)
	return (ret);

	children = MIN(oldchildren, children);
	}

	/*
	* print normal top-level devices
	*/
	for (c = 0; c < children; c++) {
	uint64_t ishole = B_FALSE, islog = B_FALSE;

	(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_HOLE,
	&ishole);

	(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG,
	&islog);

	if (ishole \|\| islog)
	continue;

	if (nvlist_exists(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID);
	ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL,
	newchild[c], cb, depth + 2);
	free(vname);
	}

	/*
	* print all other top-level devices
	*/
	for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) {
	boolean_t printed = B_FALSE;

	for (c = 0; c < children; c++) {
	uint64_t islog = B_FALSE;
	char *bias = NULL;
	char *type = NULL;

	(void) nvlist_lookup_uint64(newchild[c],
	ZPOOL_CONFIG_IS_LOG, &islog);
	if (islog) {
	bias = (char *)VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(newchild[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(newchild[c],
	ZPOOL_CONFIG_TYPE, &type);
	}
	if (bias == NULL \|\| strcmp(bias, class_name[n]) != 0)
	continue;
	if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	if ((!(cb->cb_flags & IOS_ANYHISTO_M)) &&
	!cb->cb_scripted &&
	!cb->cb_vdevs.cb_names) {
	print_iostat_dashes(cb, 0,
	class_name[n]);
	}
	printf("\n");
	printed = B_TRUE;
	}

	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID);
	ret += print_vdev_stats(zhp, vname, oldnv ?
	oldchild[c] : NULL, newchild[c], cb, depth + 2);
	free(vname);
	}
	}

	/*
	* Include level 2 ARC devices in iostat output
	*/
	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_L2CACHE,
	&newchild, &children) != 0)
	return (ret);

	if (oldnv) {
	if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_L2CACHE,
	&oldchild, &oldchildren) != 0)
	return (ret);

	children = MIN(oldchildren, children);
	}

	if (children > 0) {
	if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
	!cb->cb_vdevs.cb_names) {
	print_iostat_dashes(cb, 0, "cache");
	}
	printf("\n");

	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags);
	ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c]
	: NULL, newchild[c], cb, depth + 2);
	free(vname);
	}
	}

	return (ret);
	}

	static int
	refresh_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	boolean_t missing;

	/*
	* If the pool has disappeared, remove it from the list and continue.
	*/
	if (zpool_refresh_stats(zhp, &missing) != 0)
	return (-1);

	if (missing)
	pool_list_remove(cb->cb_list, zhp);

	return (0);
	}

	/*
	* Callback to print out the iostats for the given pool.
	*/
	static int
	print_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	nvlist_t oldconfig, newconfig;
	nvlist_t oldnvroot, newnvroot;
	int ret;

	newconfig = zpool_get_config(zhp, &oldconfig);

	if (cb->cb_iteration == 1)
	oldconfig = NULL;

	verify(nvlist_lookup_nvlist(newconfig, ZPOOL_CONFIG_VDEV_TREE,
	&newnvroot) == 0);

	if (oldconfig == NULL)
	oldnvroot = NULL;
	else
	verify(nvlist_lookup_nvlist(oldconfig, ZPOOL_CONFIG_VDEV_TREE,
	&oldnvroot) == 0);

	ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot,
	cb, 0);
	if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) &&
	!cb->cb_scripted && cb->cb_verbose &&
	!cb->cb_vdevs.cb_names_count) {
	print_iostat_separator(cb);
	if (cb->vcdl != NULL) {
	print_cmd_columns(cb->vcdl, 1);
	}
	printf("\n");
	}

	return (ret);
	}

	static int
	get_columns(void)
	{
	struct winsize ws;
	int columns = 80;
	int error;

	if (isatty(STDOUT_FILENO)) {
	error = ioctl(STDOUT_FILENO, TIOCGWINSZ, &ws);
	if (error == 0)
	columns = ws.ws_col;
	} else {
	columns = 999;
	}

	return (columns);
	}

	/*
	* Return the required length of the pool/vdev name column. The minimum
	* allowed width and output formatting flags must be provided.
	*/
	static int
	get_namewidth(zpool_handle_t *zhp, int min_width, int flags, boolean_t verbose)
	{
	nvlist_t config, nvroot;
	int width = min_width;

	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	size_t poolname_len = strlen(zpool_get_name(zhp));
	if (verbose == B_FALSE) {
	width = MAX(poolname_len, min_width);
	} else {
	width = MAX(poolname_len,
	max_width(zhp, nvroot, 0, min_width, flags));
	}
	}

	return (width);
	}

	/*
	* Parse the input string, get the 'interval' and 'count' value if there is one.
	*/
	static void
	get_interval_count(int argcp, char argv, float iv,
	unsigned long *cnt)
	{
	float interval = 0;
	unsigned long count = 0;
	int argc = *argcp;

	/*
	* Determine if the last argument is an integer or a pool name
	*/
	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
	char *end;

	errno = 0;
	interval = strtof(argv[argc - 1], &end);

	if (*end == '\0' && errno == 0) {
	if (interval == 0) {
	(void) fprintf(stderr, gettext(
	"interval cannot be zero\n"));
	usage(B_FALSE);
	}
	/*
	* Ignore the last parameter
	*/
	argc--;
	} else {
	/*
	* If this is not a valid number, just plow on. The
	* user will get a more informative error message later
	* on.
	*/
	interval = 0;
	}
	}

	/*
	* If the last argument is also an integer, then we have both a count
	* and an interval.
	*/
	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
	char *end;

	errno = 0;
	count = interval;
	interval = strtof(argv[argc - 1], &end);

	if (*end == '\0' && errno == 0) {
	if (interval == 0) {
	(void) fprintf(stderr, gettext(
	"interval cannot be zero\n"));
	usage(B_FALSE);
	}

	/*
	* Ignore the last parameter
	*/
	argc--;
	} else {
	interval = 0;
	}
	}

	*iv = interval;
	*cnt = count;
	*argcp = argc;
	}

	static void
	get_timestamp_arg(char c)
	{
	if (c == 'u')
	timestamp_fmt = UDATE;
	else if (c == 'd')
	timestamp_fmt = DDATE;
	else
	usage(B_FALSE);
	}

	/*
	* Return stat flags that are supported by all pools by both the module and
	* zpool iostat. "*data" should be initialized to all 0xFFs before running.
	* It will get ANDed down until only the flags that are supported on all pools
	* remain.
	*/
	static int
	get_stat_flags_cb(zpool_handle_t zhp, void data)
	{
	uint64_t *mask = data;
	nvlist_t config, nvroot, *nvx;
	uint64_t flags = 0;
	int i, j;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	/* Default stats are always supported, but for completeness.. */
	if (nvlist_exists(nvroot, ZPOOL_CONFIG_VDEV_STATS))
	flags \|= IOS_DEFAULT_M;

	/* Get our extended stats nvlist from the main list */
	if (nvlist_lookup_nvlist(nvroot, ZPOOL_CONFIG_VDEV_STATS_EX,
	&nvx) != 0) {
	/*
	* No extended stats; they're probably running an older
	* module. No big deal, we support that too.
	*/
	goto end;
	}

	/* For each extended stat, make sure all its nvpairs are supported */
	for (j = 0; j < ARRAY_SIZE(vsx_type_to_nvlist); j++) {
	if (!vsx_type_to_nvlist[j][0])
	continue;

	/* Start off by assuming the flag is supported, then check */
	flags \|= (1ULL << j);
	for (i = 0; vsx_type_to_nvlist[j][i]; i++) {
	if (!nvlist_exists(nvx, vsx_type_to_nvlist[j][i])) {
	/* flag isn't supported */
	flags = flags & ~(1ULL << j);
	break;
	}
	}
	}
	end:
	mask = mask & flags;
	return (0);
	}

	/*
	* Return a bitmask of stats that are supported on all pools by both the module
	* and zpool iostat.
	*/
	static uint64_t
	get_stat_flags(zpool_list_t *list)
	{
	uint64_t mask = -1;

	/*
	* get_stat_flags_cb() will lop off bits from "mask" until only the
	* flags that are supported on all pools remain.
	*/
	pool_list_iter(list, B_FALSE, get_stat_flags_cb, &mask);
	return (mask);
	}

	/*
	* Return 1 if cb_data->cb_names[0] is this vdev's name, 0 otherwise.
	*/
	static int
	is_vdev_cb(void zhp_data, nvlist_t nv, void *cb_data)
	{
	vdev_cbdata_t *cb = cb_data;
	char *name = NULL;
	int ret = 1; /* assume match */
	zpool_handle_t *zhp = zhp_data;

	name = zpool_vdev_name(g_zfs, zhp, nv, cb->cb_name_flags);

	if (strcmp(name, cb->cb_names[0])) {
	free(name);
	name = zpool_vdev_name(g_zfs, zhp, nv, VDEV_NAME_GUID);
	ret = (strcmp(name, cb->cb_names[0]) == 0);
	}
	free(name);

	return (ret);
	}

	/*
	* Returns 1 if cb_data->cb_names[0] is a vdev name, 0 otherwise.
	*/
	static int
	is_vdev(zpool_handle_t zhp, void cb_data)
	{
	return (for_each_vdev(zhp, is_vdev_cb, cb_data));
	}

	/*
	* Check if vdevs are in a pool
	*
	* Return 1 if all argv[] strings are vdev names in pool "pool_name". Otherwise
	* return 0. If pool_name is NULL, then search all pools.
	*/
	static int
	are_vdevs_in_pool(int argc, char *argv, char pool_name,
	vdev_cbdata_t *cb)
	{
	char **tmp_name;
	int ret = 0;
	int i;
	int pool_count = 0;

	if ((argc == 0) \|\| !*argv)
	return (0);

	if (pool_name)
	pool_count = 1;

	/* Temporarily hijack cb_names for a second... */
	tmp_name = cb->cb_names;

	/* Go though our list of prospective vdev names */
	for (i = 0; i < argc; i++) {
	cb->cb_names = argv + i;

	/* Is this name a vdev in our pools? */
	ret = for_each_pool(pool_count, &pool_name, B_TRUE, NULL,
	ZFS_TYPE_POOL, B_FALSE, is_vdev, cb);
	if (!ret) {
	/* No match */
	break;
	}
	}

	cb->cb_names = tmp_name;

	return (ret);
	}

	static int
	is_pool_cb(zpool_handle_t zhp, void data)
	{
	char *name = data;
	if (strcmp(name, zpool_get_name(zhp)) == 0)
	return (1);

	return (0);
	}

	/*
	* Do we have a pool named *name? If so, return 1, otherwise 0.
	*/
	static int
	is_pool(char *name)
	{
	return (for_each_pool(0, NULL, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE,
	is_pool_cb, name));
	}

	/* Are all our argv[] strings pool names? If so return 1, 0 otherwise. */
	static int
	are_all_pools(int argc, char **argv)
	{
	if ((argc == 0) \|\| !*argv)
	return (0);

	while (--argc >= 0)
	if (!is_pool(argv[argc]))
	return (0);

	return (1);
	}

	/*
	* Helper function to print out vdev/pool names we can't resolve. Used for an
	* error message.
	*/
	static void
	error_list_unresolved_vdevs(int argc, char *argv, char pool_name,
	vdev_cbdata_t *cb)
	{
	int i;
	char *name;
	char *str;
	for (i = 0; i < argc; i++) {
	name = argv[i];

	if (is_pool(name))
	str = gettext("pool");
	else if (are_vdevs_in_pool(1, &name, pool_name, cb))
	str = gettext("vdev in this pool");
	else if (are_vdevs_in_pool(1, &name, NULL, cb))
	str = gettext("vdev in another pool");
	else
	str = gettext("unknown");

	fprintf(stderr, "\t%s (%s)\n", name, str);
	}
	}

	/*
	* Same as get_interval_count(), but with additional checks to not misinterpret
	* guids as interval/count values. Assumes VDEV_NAME_GUID is set in
	* cb.cb_vdevs.cb_name_flags.
	*/
	static void
	get_interval_count_filter_guids(int argc, char argv, float interval,
	unsigned long count, iostat_cbdata_t cb)
	{
	char **tmpargv = argv;
	int argc_for_interval = 0;

	/* Is the last arg an interval value? Or a guid? */
	if (argc >= 1 && !are_vdevs_in_pool(1, &argv[argc - 1], NULL,
	&cb->cb_vdevs)) {
	/*
	* The last arg is not a guid, so it's probably an
	* interval value.
	*/
	argc_for_interval++;

	if (*argc >= 2 &&
	!are_vdevs_in_pool(1, &argv[*argc - 2], NULL,
	&cb->cb_vdevs)) {
	/*
	* The 2nd to last arg is not a guid, so it's probably
	* an interval value.
	*/
	argc_for_interval++;
	}
	}

	/* Point to our list of possible intervals */
	tmpargv = &argv[*argc - argc_for_interval];

	argc = argc - argc_for_interval;
	get_interval_count(&argc_for_interval, tmpargv,
	interval, count);
	}

	/*
	* Floating point sleep(). Allows you to pass in a floating point value for
	* seconds.
	*/
	static void
	fsleep(float sec)
	{
	struct timespec req;
	req.tv_sec = floor(sec);
	req.tv_nsec = (sec - (float)req.tv_sec) * NANOSEC;
	nanosleep(&req, NULL);
	}

	/*
	* Terminal height, in rows. Returns -1 if stdout is not connected to a TTY or
	* if we were unable to determine its size.
	*/
	static int
	terminal_height(void)
	{
	struct winsize win;

	if (isatty(STDOUT_FILENO) == 0)
	return (-1);

	if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) != -1 && win.ws_row > 0)
	return (win.ws_row);

	return (-1);
	}

	/*
	* Run one of the zpool status/iostat -c scripts with the help (-h) option and
	* print the result.
	*
	* name: Short name of the script ('iostat').
	* path: Full path to the script ('/usr/local/etc/zfs/zpool.d/iostat');
	*/
	static void
	print_zpool_script_help(char name, char path)
	{
	char argv[] = {path, (char )"-h", NULL};
	char **lines = NULL;
	int lines_cnt = 0;
	int rc;

	rc = libzfs_run_process_get_stdout_nopath(path, argv, NULL, &lines,
	&lines_cnt);
	if (rc != 0 \|\| lines == NULL \|\| lines_cnt <= 0) {
	if (lines != NULL)
	libzfs_free_str_array(lines, lines_cnt);
	return;
	}

	for (int i = 0; i < lines_cnt; i++)
	if (!is_blank_str(lines[i]))
	printf(" %-14s %s\n", name, lines[i]);

	libzfs_free_str_array(lines, lines_cnt);
	}

	/*
	* Go though the zpool status/iostat -c scripts in the user's path, run their
	* help option (-h), and print out the results.
	*/
	static void
	print_zpool_dir_scripts(char *dirpath)
	{
	DIR *dir;
	struct dirent *ent;
	char fullpath[MAXPATHLEN];
	struct stat dir_stat;

	if ((dir = opendir(dirpath)) != NULL) {
	/* print all the files and directories within directory */
	while ((ent = readdir(dir)) != NULL) {
	sprintf(fullpath, "%s/%s", dirpath, ent->d_name);

	/* Print the scripts */
	if (stat(fullpath, &dir_stat) == 0)
	if (dir_stat.st_mode & S_IXUSR &&
	S_ISREG(dir_stat.st_mode))
	print_zpool_script_help(ent->d_name,
	fullpath);
	}
	closedir(dir);
	}
	}

	/*
	* Print out help text for all zpool status/iostat -c scripts.
	*/
	static void
	print_zpool_script_list(const char *subcommand)
	{
	char dir, sp, *tmp;

	printf(gettext("Available 'zpool %s -c' commands:\n"), subcommand);

	sp = zpool_get_cmd_search_path();
	if (sp == NULL)
	return;

	for (dir = strtok_r(sp, ":", &tmp);
	dir != NULL;
	dir = strtok_r(NULL, ":", &tmp))
	print_zpool_dir_scripts(dir);

	free(sp);
	}

	/*
	* Set the minimum pool/vdev name column width. The width must be at least 10,
	* but may be as large as the column width - 42 so it still fits on one line.
	* NOTE: 42 is the width of the default capacity/operations/bandwidth output
	*/
	static int
	get_namewidth_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	int width, available_width;

	/*
	* get_namewidth() returns the maximum width of any name in that column
	* for any pool/vdev/device line that will be output.
	*/
	width = get_namewidth(zhp, cb->cb_namewidth,
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID, cb->cb_verbose);

	/*
	* The width we are calculating is the width of the header and also the
	* padding width for names that are less than maximum width. The stats
	* take up 42 characters, so the width available for names is:
	*/
	available_width = get_columns() - 42;

	/*
	* If the maximum width fits on a screen, then great! Make everything
	* line up by justifying all lines to the same width. If that max
	* width is larger than what's available, the name plus stats won't fit
	* on one line, and justifying to that width would cause every line to
	* wrap on the screen. We only want lines with long names to wrap.
	* Limit the padding to what won't wrap.
	*/
	if (width > available_width)
	width = available_width;

	/*
	* And regardless of whatever the screen width is (get_columns can
	* return 0 if the width is not known or less than 42 for a narrow
	* terminal) have the width be a minimum of 10.
	*/
	if (width < 10)
	width = 10;

	/* Save the calculated width */
	cb->cb_namewidth = width;

	return (0);
	}

	/*
	* zpool iostat [[-c [script1,script2,...]] [-lq]\|[-rw]] [-ghHLpPvy] [-n name]
	* [-T d\|u] [[ pool ...]\|[pool vdev ...]\|[vdev ...]]
	* [interval [count]]
	*
	* -c CMD For each vdev, run command CMD
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -P Display full path for vdev name.
	* -v Display statistics for individual vdevs
	* -h Display help
	* -p Display values in parsable (exact) format.
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -l Display average latency
	* -q Display queue depths
	* -w Display latency histograms
	* -r Display request size histogram
	* -T Display a timestamp in date(1) or Unix format
	* -n Only print headers once
	*
	* This command can be tricky because we want to be able to deal with pool
	* creation/destruction as well as vdev configuration changes. The bulk of this
	* processing is handled by the pool_list_* routines in zpool_iter.c. We rely
	* on pool_list_update() to detect the addition of new pools. Configuration
	* changes are all handled within libzfs.
	*/
	int
	zpool_do_iostat(int argc, char **argv)
	{
	int c;
	int ret;
	int npools;
	float interval = 0;
	unsigned long count = 0;
	int winheight = 24;
	zpool_list_t *list;
	boolean_t verbose = B_FALSE;
	boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE;
	boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE;
	boolean_t omit_since_boot = B_FALSE;
	boolean_t guid = B_FALSE;
	boolean_t follow_links = B_FALSE;
	boolean_t full_name = B_FALSE;
	boolean_t headers_once = B_FALSE;
	iostat_cbdata_t cb = { 0 };
	char *cmd = NULL;

	/* Used for printing error message */
	const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q',
	[IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'};

	uint64_t unsupported_flags;

	/* check options */
	while ((c = getopt(argc, argv, "c:gLPT:vyhplqrwnH")) != -1) {
	switch (c) {
	case 'c':
	if (cmd != NULL) {
	fprintf(stderr,
	gettext("Can't set -c flag twice\n"));
	exit(1);
	}

	if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
	fprintf(stderr, gettext(
	"Can't run -c, disabled by "
	"ZPOOL_SCRIPTS_ENABLED.\n"));
	exit(1);
	}

	if ((getuid() <= 0 \|\| geteuid() <= 0) &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
	fprintf(stderr, gettext(
	"Can't run -c with root privileges "
	"unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
	exit(1);
	}
	cmd = optarg;
	verbose = B_TRUE;
	break;
	case 'g':
	guid = B_TRUE;
	break;
	case 'L':
	follow_links = B_TRUE;
	break;
	case 'P':
	full_name = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 'l':
	latency = B_TRUE;
	break;
	case 'q':
	queues = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case 'w':
	l_histo = B_TRUE;
	break;
	case 'r':
	rq_histo = B_TRUE;
	break;
	case 'y':
	omit_since_boot = B_TRUE;
	break;
	case 'n':
	headers_once = B_TRUE;
	break;
	case 'h':
	usage(B_FALSE);
	break;
	case '?':
	if (optopt == 'c') {
	print_zpool_script_list("iostat");
	exit(0);
	} else {
	fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	cb.cb_literal = parsable;
	cb.cb_scripted = scripted;

	if (guid)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_GUID;
	if (follow_links)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	if (full_name)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_PATH;
	cb.cb_iteration = 0;
	cb.cb_namewidth = 0;
	cb.cb_verbose = verbose;

	/* Get our interval and count values (if any) */
	if (guid) {
	get_interval_count_filter_guids(&argc, argv, &interval,
	&count, &cb);
	} else {
	get_interval_count(&argc, argv, &interval, &count);
	}

	if (argc == 0) {
	/* No args, so just print the defaults. */
	} else if (are_all_pools(argc, argv)) {
	/* All the args are pool names */
	} else if (are_vdevs_in_pool(argc, argv, NULL, &cb.cb_vdevs)) {
	/* All the args are vdevs */
	cb.cb_vdevs.cb_names = argv;
	cb.cb_vdevs.cb_names_count = argc;
	argc = 0; /* No pools to process */
	} else if (are_all_pools(1, argv)) {
	/* The first arg is a pool name */
	if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0],
	&cb.cb_vdevs)) {
	/* ...and the rest are vdev names */
	cb.cb_vdevs.cb_names = argv + 1;
	cb.cb_vdevs.cb_names_count = argc - 1;
	argc = 1; /* One pool to process */
	} else {
	fprintf(stderr, gettext("Expected either a list of "));
	fprintf(stderr, gettext("pools, or list of vdevs in"));
	fprintf(stderr, " \"%s\", ", argv[0]);
	fprintf(stderr, gettext("but got:\n"));
	error_list_unresolved_vdevs(argc - 1, argv + 1,
	argv[0], &cb.cb_vdevs);
	fprintf(stderr, "\n");
	usage(B_FALSE);
	return (1);
	}
	} else {
	/*
	* The args don't make sense. The first arg isn't a pool name,
	* nor are all the args vdevs.
	*/
	fprintf(stderr, gettext("Unable to parse pools/vdevs list.\n"));
	fprintf(stderr, "\n");
	return (1);
	}

	if (cb.cb_vdevs.cb_names_count != 0) {
	/*
	* If user specified vdevs, it implies verbose.
	*/
	cb.cb_verbose = B_TRUE;
	}

	/*
	* Construct the list of all interesting pools.
	*/
	ret = 0;
	if ((list = pool_list_get(argc, argv, NULL, ZFS_TYPE_POOL, parsable,
	&ret)) == NULL)
	return (1);

	if (pool_list_count(list) == 0 && argc != 0) {
	pool_list_free(list);
	return (1);
	}

	if (pool_list_count(list) == 0 && interval == 0) {
	pool_list_free(list);
	(void) fprintf(stderr, gettext("no pools available\n"));
	return (1);
	}

	if ((l_histo \|\| rq_histo) && (cmd != NULL \|\| latency \|\| queues)) {
	pool_list_free(list);
	(void) fprintf(stderr,
	gettext("[-r\|-w] isn't allowed with [-c\|-l\|-q]\n"));
	usage(B_FALSE);
	return (1);
	}

	if (l_histo && rq_histo) {
	pool_list_free(list);
	(void) fprintf(stderr,
	gettext("Only one of [-r\|-w] can be passed at a time\n"));
	usage(B_FALSE);
	return (1);
	}

	/*
	* Enter the main iostat loop.
	*/
	cb.cb_list = list;

	if (l_histo) {
	/*
	* Histograms tables look out of place when you try to display
	* them with the other stats, so make a rule that you can only
	* print histograms by themselves.
	*/
	cb.cb_flags = IOS_L_HISTO_M;
	} else if (rq_histo) {
	cb.cb_flags = IOS_RQ_HISTO_M;
	} else {
	cb.cb_flags = IOS_DEFAULT_M;
	if (latency)
	cb.cb_flags \|= IOS_LATENCY_M;
	if (queues)
	cb.cb_flags \|= IOS_QUEUES_M;
	}

	/*
	* See if the module supports all the stats we want to display.
	*/
	unsupported_flags = cb.cb_flags & ~get_stat_flags(list);
	if (unsupported_flags) {
	uint64_t f;
	int idx;
	fprintf(stderr,
	gettext("The loaded zfs module doesn't support:"));

	/* for each bit set in unsupported_flags */
	for (f = unsupported_flags; f; f &= ~(1ULL << idx)) {
	idx = lowbit64(f) - 1;
	fprintf(stderr, " -%c", flag_to_arg[idx]);
	}

	fprintf(stderr, ". Try running a newer module.\n");
	pool_list_free(list);

	return (1);
	}

	for (;;) {
	if ((npools = pool_list_count(list)) == 0)
	(void) fprintf(stderr, gettext("no pools available\n"));
	else {
	/*
	* If this is the first iteration and -y was supplied
	* we skip any printing.
	*/
	boolean_t skip = (omit_since_boot &&
	cb.cb_iteration == 0);

	/*
	* Refresh all statistics. This is done as an
	* explicit step before calculating the maximum name
	* width, so that any * configuration changes are
	* properly accounted for.
	*/
	(void) pool_list_iter(list, B_FALSE, refresh_iostat,
	&cb);

	/*
	* Iterate over all pools to determine the maximum width
	* for the pool / device name column across all pools.
	*/
	cb.cb_namewidth = 0;
	(void) pool_list_iter(list, B_FALSE,
	get_namewidth_iostat, &cb);

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (cmd != NULL && cb.cb_verbose &&
	!(cb.cb_flags & IOS_ANYHISTO_M)) {
	cb.vcdl = all_pools_for_each_vdev_run(argc,
	argv, cmd, g_zfs, cb.cb_vdevs.cb_names,
	cb.cb_vdevs.cb_names_count,
	cb.cb_vdevs.cb_name_flags);
	} else {
	cb.vcdl = NULL;
	}


	/*
	* Check terminal size so we can print headers
	* even when terminal window has its height
	* changed.
	*/
	winheight = terminal_height();
	/*
	* Are we connected to TTY? If not, headers_once
	* should be true, to avoid breaking scripts.
	*/
	if (winheight < 0)
	headers_once = B_TRUE;

	/*
	* If it's the first time and we're not skipping it,
	* or either skip or verbose mode, print the header.
	*
	* The histogram code explicitly prints its header on
	* every vdev, so skip this for histograms.
	*/
	if (((++cb.cb_iteration == 1 && !skip) \|\|
	(skip != verbose) \|\|
	(!headers_once &&
	(cb.cb_iteration % winheight) == 0)) &&
	(!(cb.cb_flags & IOS_ANYHISTO_M)) &&
	!cb.cb_scripted)
	print_iostat_header(&cb);

	if (skip) {
	(void) fsleep(interval);
	continue;
	}

	pool_list_iter(list, B_FALSE, print_iostat, &cb);

	/*
	* If there's more than one pool, and we're not in
	* verbose mode (which prints a separator for us),
	* then print a separator.
	*
	* In addition, if we're printing specific vdevs then
	* we also want an ending separator.
	*/
	if (((npools > 1 && !verbose &&
	!(cb.cb_flags & IOS_ANYHISTO_M)) \|\|
	(!(cb.cb_flags & IOS_ANYHISTO_M) &&
	cb.cb_vdevs.cb_names_count)) &&
	!cb.cb_scripted) {
	print_iostat_separator(&cb);
	if (cb.vcdl != NULL)
	print_cmd_columns(cb.vcdl, 1);
	printf("\n");
	}

	if (cb.vcdl != NULL)
	free_vdev_cmd_data_list(cb.vcdl);

	}

	/*
	* Flush the output so that redirection to a file isn't buffered
	* indefinitely.
	*/
	(void) fflush(stdout);

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	(void) fsleep(interval);
	}

	pool_list_free(list);

	return (ret);
	}

	typedef struct list_cbdata {
	boolean_t cb_verbose;
	int cb_name_flags;
	int cb_namewidth;
	boolean_t cb_scripted;
	zprop_list_t *cb_proplist;
	boolean_t cb_literal;
	} list_cbdata_t;


	/*
	* Given a list of columns to display, output appropriate headers for each one.
	*/
	static void
	print_header(list_cbdata_t *cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	char headerbuf[ZPOOL_MAXPROPLEN];
	const char *header;
	boolean_t first = B_TRUE;
	boolean_t right_justify;
	size_t width = 0;

	for (; pl != NULL; pl = pl->pl_next) {
	width = pl->pl_width;
	if (first && cb->cb_verbose) {
	/*
	* Reset the width to accommodate the verbose listing
	* of devices.
	*/
	width = cb->cb_namewidth;
	}

	if (!first)
	(void) fputs(" ", stdout);
	else
	first = B_FALSE;

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	header = zpool_prop_column_name(pl->pl_prop);
	right_justify = zpool_prop_align_right(pl->pl_prop);
	} else {
	int i;

	for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
	headerbuf[i] = toupper(pl->pl_user_prop[i]);
	headerbuf[i] = '\0';
	header = headerbuf;
	}

	if (pl->pl_next == NULL && !right_justify)
	(void) fputs(header, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)width, header);
	else
	(void) printf("%-*s", (int)width, header);
	}

	(void) fputc('\n', stdout);
	}

	/*
	* Given a pool and a list of properties, print out all the properties according
	* to the described layout. Used by zpool_do_list().
	*/
	static void
	print_pool(zpool_handle_t zhp, list_cbdata_t cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	boolean_t first = B_TRUE;
	char property[ZPOOL_MAXPROPLEN];
	const char *propstr;
	boolean_t right_justify;
	size_t width;

	for (; pl != NULL; pl = pl->pl_next) {

	width = pl->pl_width;
	if (first && cb->cb_verbose) {
	/*
	* Reset the width to accommodate the verbose listing
	* of devices.
	*/
	width = cb->cb_namewidth;
	}

	if (!first) {
	if (cb->cb_scripted)
	(void) fputc('\t', stdout);
	else
	(void) fputs(" ", stdout);
	} else {
	first = B_FALSE;
	}

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	if (zpool_get_prop(zhp, pl->pl_prop, property,
	sizeof (property), NULL, cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;

	right_justify = zpool_prop_align_right(pl->pl_prop);
	} else if ((zpool_prop_feature(pl->pl_user_prop) \|\|
	zpool_prop_unsupported(pl->pl_user_prop)) &&
	zpool_prop_get_feature(zhp, pl->pl_user_prop, property,
	sizeof (property)) == 0) {
	propstr = property;
	} else {
	propstr = "-";
	}


	/*
	* If this is being called in scripted mode, or if this is the
	* last column and it is left-justified, don't include a width
	* format specifier.
	*/
	if (cb->cb_scripted \|\| (pl->pl_next == NULL && !right_justify))
	(void) fputs(propstr, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)width, propstr);
	else
	(void) printf("%-*s", (int)width, propstr);
	}

	(void) fputc('\n', stdout);
	}

	static void
	print_one_column(zpool_prop_t prop, uint64_t value, const char *str,
	boolean_t scripted, boolean_t valid, enum zfs_nicenum_format format)
	{
	char propval[64];
	boolean_t fixed;
	size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL);

	switch (prop) {
	case ZPOOL_PROP_SIZE:
	case ZPOOL_PROP_EXPANDSZ:
	case ZPOOL_PROP_CHECKPOINT:
	case ZPOOL_PROP_DEDUPRATIO:
	if (value == 0)
	(void) strlcpy(propval, "-", sizeof (propval));
	else
	zfs_nicenum_format(value, propval, sizeof (propval),
	format);
	break;
	case ZPOOL_PROP_FRAGMENTATION:
	if (value == ZFS_FRAG_INVALID) {
	(void) strlcpy(propval, "-", sizeof (propval));
	} else if (format == ZFS_NICENUM_RAW) {
	(void) snprintf(propval, sizeof (propval), "%llu",
	(unsigned long long)value);
	} else {
	(void) snprintf(propval, sizeof (propval), "%llu%%",
	(unsigned long long)value);
	}
	break;
	case ZPOOL_PROP_CAPACITY:
	/* capacity value is in parts-per-10,000 (aka permyriad) */
	if (format == ZFS_NICENUM_RAW)
	(void) snprintf(propval, sizeof (propval), "%llu",
	(unsigned long long)value / 100);
	else
	(void) snprintf(propval, sizeof (propval),
	value < 1000 ? "%1.2f%%" : value < 10000 ?
	"%2.1f%%" : "%3.0f%%", value / 100.0);
	break;
	case ZPOOL_PROP_HEALTH:
	width = 8;
	(void) strlcpy(propval, str, sizeof (propval));
	break;
	default:
	zfs_nicenum_format(value, propval, sizeof (propval), format);
	}

	if (!valid)
	(void) strlcpy(propval, "-", sizeof (propval));

	if (scripted)
	(void) printf("\t%s", propval);
	else
	(void) printf(" %*s", (int)width, propval);
	}

	/*
	* print static default line per vdev
	* not compatible with '-o' <proplist> option
	*/
	static void
	print_list_stats(zpool_handle_t zhp, const char name, nvlist_t *nv,
	list_cbdata_t *cb, int depth, boolean_t isspare)
	{
	nvlist_t **child;
	vdev_stat_t *vs;
	uint_t c, children;
	char *vname;
	boolean_t scripted = cb->cb_scripted;
	uint64_t islog = B_FALSE;
	const char dashes = "%-s - - - - "
	"- - - - -\n";

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	if (name != NULL) {
	boolean_t toplevel = (vs->vs_space != 0);
	uint64_t cap;
	enum zfs_nicenum_format format;
	const char *state;

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
	return;

	if (scripted)
	(void) printf("\t%s", name);
	else if (strlen(name) + depth > cb->cb_namewidth)
	(void) printf("%*s%s", depth, "", name);
	else
	(void) printf("%s%s%s", depth, "", name,
	(int)(cb->cb_namewidth - strlen(name) - depth), "");

	/*
	* Print the properties for the individual vdevs. Some
	* properties are only applicable to toplevel vdevs. The
	* 'toplevel' boolean value is passed to the print_one_column()
	* to indicate that the value is valid.
	*/
	if (VDEV_STAT_VALID(vs_pspace, c) && vs->vs_pspace)
	print_one_column(ZPOOL_PROP_SIZE, vs->vs_pspace, NULL,
	scripted, B_TRUE, format);
	else
	print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc,
	NULL, scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_CHECKPOINT,
	vs->vs_checkpoint_space, NULL, scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_EXPANDSZ, vs->vs_esize, NULL,
	scripted, B_TRUE, format);
	print_one_column(ZPOOL_PROP_FRAGMENTATION,
	vs->vs_fragmentation, NULL, scripted,
	(vs->vs_fragmentation != ZFS_FRAG_INVALID && toplevel),
	format);
	cap = (vs->vs_space == 0) ? 0 :
	(vs->vs_alloc * 10000 / vs->vs_space);
	print_one_column(ZPOOL_PROP_CAPACITY, cap, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_DEDUPRATIO, 0, NULL,
	scripted, toplevel, format);
	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
	if (isspare) {
	if (vs->vs_aux == VDEV_AUX_SPARED)
	state = "INUSE";
	else if (vs->vs_state == VDEV_STATE_HEALTHY)
	state = "AVAIL";
	}
	print_one_column(ZPOOL_PROP_HEALTH, 0, state, scripted,
	B_TRUE, format);
	(void) fputc('\n', stdout);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return;

	/* list the normal vdevs first */
	for (c = 0; c < children; c++) {
	uint64_t ishole = B_FALSE;

	if (nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_HOLE, &ishole) == 0 && ishole)
	continue;

	if (nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog)
	continue;

	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE);
	free(vname);
	}

	/* list the classes: 'logs', 'dedup', and 'special' */
	for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) {
	boolean_t printed = B_FALSE;

	for (c = 0; c < children; c++) {
	char *bias = NULL;
	char *type = NULL;

	if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&islog) == 0 && islog) {
	bias = (char *)VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_TYPE, &type);
	}
	if (bias == NULL \|\| strcmp(bias, class_name[n]) != 0)
	continue;
	if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth,
	class_name[n]);
	printed = B_TRUE;
	}
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_FALSE);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0 && children > 0) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth, "cache");
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_FALSE);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child,
	&children) == 0 && children > 0) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth, "spare");
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_TRUE);
	free(vname);
	}
	}
	}

	/*
	* Generic callback function to list a pool.
	*/
	static int
	list_callback(zpool_handle_t zhp, void data)
	{
	list_cbdata_t *cbp = data;

	print_pool(zhp, cbp);

	if (cbp->cb_verbose) {
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	print_list_stats(zhp, NULL, nvroot, cbp, 0, B_FALSE);
	}

	return (0);
	}

	/*
	* Set the minimum pool/vdev name column width. The width must be at least 9,
	* but may be as large as needed.
	*/
	static int
	get_namewidth_list(zpool_handle_t zhp, void data)
	{
	list_cbdata_t *cb = data;
	int width;

	width = get_namewidth(zhp, cb->cb_namewidth,
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID, cb->cb_verbose);

	if (width < 9)
	width = 9;

	cb->cb_namewidth = width;

	return (0);
	}

	/*
	* zpool list [-gHLpP] [-o prop[,prop]*] [-T d\|u] [pool] ... [interval [count]]
	*
	* -g Display guid for individual vdev name.
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -L Follow links when resolving vdev path name.
	* -o List of properties to display. Defaults to
	* "name,size,allocated,free,expandsize,fragmentation,capacity,"
	* "dedupratio,health,altroot"
	* -p Display values in parsable (exact) format.
	* -P Display full path for vdev name.
	* -T Display a timestamp in date(1) or Unix format
	*
	* List all pools in the system, whether or not they're healthy. Output space
	* statistics for each one, as well as health status summary.
	*/
	int
	zpool_do_list(int argc, char **argv)
	{
	int c;
	int ret = 0;
	list_cbdata_t cb = { 0 };
	static char default_props[] =
	"name,size,allocated,free,checkpoint,expandsize,fragmentation,"
	"capacity,dedupratio,health,altroot";
	char *props = default_props;
	float interval = 0;
	unsigned long count = 0;
	zpool_list_t *list;
	boolean_t first = B_TRUE;
	current_prop_type = ZFS_TYPE_POOL;

	/* check options */
	while ((c = getopt(argc, argv, ":gHLo:pPT:v")) != -1) {
	switch (c) {
	case 'g':
	cb.cb_name_flags \|= VDEV_NAME_GUID;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 'L':
	cb.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'o':
	props = optarg;
	break;
	case 'P':
	cb.cb_name_flags \|= VDEV_NAME_PATH;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 'v':
	cb.cb_verbose = B_TRUE;
	cb.cb_namewidth = 8; /* 8 until precalc is avail */
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &interval, &count);

	if (zprop_get_list(g_zfs, props, &cb.cb_proplist, ZFS_TYPE_POOL) != 0)
	usage(B_FALSE);

	for (;;) {
	if ((list = pool_list_get(argc, argv, &cb.cb_proplist,
	ZFS_TYPE_POOL, cb.cb_literal, &ret)) == NULL)
	return (1);

	if (pool_list_count(list) == 0)
	break;

	cb.cb_namewidth = 0;
	(void) pool_list_iter(list, B_FALSE, get_namewidth_list, &cb);

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (!cb.cb_scripted && (first \|\| cb.cb_verbose)) {
	print_header(&cb);
	first = B_FALSE;
	}
	ret = pool_list_iter(list, B_TRUE, list_callback, &cb);

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	pool_list_free(list);
	(void) fsleep(interval);
	}

	if (argc == 0 && !cb.cb_scripted && pool_list_count(list) == 0) {
	(void) printf(gettext("no pools available\n"));
	ret = 0;
	}

	pool_list_free(list);
	zprop_free_list(cb.cb_proplist);
	return (ret);
	}

	static int
	zpool_do_attach_or_replace(int argc, char **argv, int replacing)
	{
	boolean_t force = B_FALSE;
	boolean_t rebuild = B_FALSE;
	boolean_t wait = B_FALSE;
	int c;
	nvlist_t *nvroot;
	char poolname, old_disk, *new_disk;
	zpool_handle_t *zhp;
	nvlist_t *props = NULL;
	char *propval;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "fo:sw")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	usage(B_FALSE);
	}
	*propval = '\0';
	propval++;

	if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) \|\|
	(add_prop_list(optarg, propval, &props, B_TRUE)))
	usage(B_FALSE);
	break;
	case 's':
	rebuild = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("missing <device> specification\n"));
	usage(B_FALSE);
	}

	old_disk = argv[1];

	if (argc < 3) {
	if (!replacing) {
	(void) fprintf(stderr,
	gettext("missing <new_device> specification\n"));
	usage(B_FALSE);
	}
	new_disk = old_disk;
	argc -= 1;
	argv += 1;
	} else {
	new_disk = argv[2];
	argc -= 2;
	argv += 2;
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	if (zpool_get_config(zhp, NULL) == NULL) {
	(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
	poolname);
	zpool_close(zhp);
	nvlist_free(props);
	return (1);
	}

	/* unless manually specified use "ashift" pool property (if set) */
	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
	int intval;
	zprop_source_t src;
	char strval[ZPOOL_MAXPROPLEN];

	intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
	if (src != ZPROP_SRC_DEFAULT) {
	(void) sprintf(strval, "%" PRId32, intval);
	verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
	&props, B_TRUE) == 0);
	}
	}

	nvroot = make_root_vdev(zhp, props, force, B_FALSE, replacing, B_FALSE,
	argc, argv);
	if (nvroot == NULL) {
	zpool_close(zhp);
	nvlist_free(props);
	return (1);
	}

	ret = zpool_vdev_attach(zhp, old_disk, new_disk, nvroot, replacing,
	rebuild);

	if (ret == 0 && wait)
	ret = zpool_wait(zhp,
	replacing ? ZPOOL_WAIT_REPLACE : ZPOOL_WAIT_RESILVER);

	nvlist_free(props);
	nvlist_free(nvroot);
	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool replace [-fsw] [-o property=value] <pool> <device> <new_device>
	*
	* -f Force attach, even if <new_device> appears to be in use.
	* -s Use sequential instead of healing reconstruction for resilver.
	* -o Set property=value.
	* -w Wait for replacing to complete before returning
	*
	* Replace <device> with <new_device>.
	*/
	int
	zpool_do_replace(int argc, char **argv)
	{
	return (zpool_do_attach_or_replace(argc, argv, B_TRUE));
	}

	/*
	* zpool attach [-fsw] [-o property=value] <pool> <device> <new_device>
	*
	* -f Force attach, even if <new_device> appears to be in use.
	* -s Use sequential instead of healing reconstruction for resilver.
	* -o Set property=value.
	* -w Wait for resilvering to complete before returning
	*
	* Attach <new_device> to the mirror containing <device>. If <device> is not
	* part of a mirror, then <device> will be transformed into a mirror of
	* <device> and <new_device>. In either case, <new_device> will begin life
	* with a DTL of [0, now], and will immediately begin to resilver itself.
	*/
	int
	zpool_do_attach(int argc, char **argv)
	{
	return (zpool_do_attach_or_replace(argc, argv, B_FALSE));
	}

	/*
	* zpool detach [-f] <pool> <device>
	*
	* -f Force detach of <device>, even if DTLs argue against it
	* (not supported yet)
	*
	* Detach a device from a mirror. The operation will be refused if <device>
	* is the last device in the mirror, or if the DTLs indicate that this device
	* has the only valid copy of some data.
	*/
	int
	zpool_do_detach(int argc, char **argv)
	{
	int c;
	char poolname, path;
	zpool_handle_t *zhp;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("missing <device> specification\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	path = argv[1];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	ret = zpool_vdev_detach(zhp, path);

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool split [-gLnP] [-o prop=val] ...
	* [-o mntopt] ...
	* [-R altroot] <pool> <newpool> [<device> ...]
	*
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -n Do not split the pool, but display the resulting layout if
	* it were to be split.
	* -o Set property=value, or set mount options.
	* -P Display full path for vdev name.
	* -R Mount the split-off pool under an alternate root.
	* -l Load encryption keys while importing.
	*
	* Splits the named pool and gives it the new pool name. Devices to be split
	* off may be listed, provided that no more than one device is specified
	* per top-level vdev mirror. The newly split pool is left in an exported
	* state unless -R is specified.
	*
	* Restrictions: the top-level of the pool pool must only be made up of
	* mirrors; all devices in the pool must be healthy; no device may be
	* undergoing a resilvering operation.
	*/
	int
	zpool_do_split(int argc, char **argv)
	{
	char srcpool, newpool, *propval;
	char *mntopts = NULL;
	splitflags_t flags;
	int c, ret = 0;
	boolean_t loadkeys = B_FALSE;
	zpool_handle_t *zhp;
	nvlist_t config, props = NULL;

	flags.dryrun = B_FALSE;
	flags.import = B_FALSE;
	flags.name_flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, ":gLR:lno:P")) != -1) {
	switch (c) {
	case 'g':
	flags.name_flags \|= VDEV_NAME_GUID;
	break;
	case 'L':
	flags.name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'R':
	flags.import = B_TRUE;
	if (add_prop_list(
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), optarg,
	&props, B_TRUE) != 0) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'l':
	loadkeys = B_TRUE;
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) != NULL) {
	*propval = '\0';
	propval++;
	if (add_prop_list(optarg, propval,
	&props, B_TRUE) != 0) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	} else {
	mntopts = optarg;
	}
	break;
	case 'P':
	flags.name_flags \|= VDEV_NAME_PATH;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	break;
	}
	}

	if (!flags.import && mntopts != NULL) {
	(void) fprintf(stderr, gettext("setting mntopts is only "
	"valid when importing the pool\n"));
	usage(B_FALSE);
	}

	if (!flags.import && loadkeys) {
	(void) fprintf(stderr, gettext("loading keys is only "
	"valid when importing the pool\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("Missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("Missing new pool name\n"));
	usage(B_FALSE);
	}

	srcpool = argv[0];
	newpool = argv[1];

	argc -= 2;
	argv += 2;

	if ((zhp = zpool_open(g_zfs, srcpool)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	config = split_mirror_vdev(zhp, newpool, props, flags, argc, argv);
	if (config == NULL) {
	ret = 1;
	} else {
	if (flags.dryrun) {
	(void) printf(gettext("would create '%s' with the "
	"following layout:\n\n"), newpool);
	print_vdev_tree(NULL, newpool, config, 0, "",
	flags.name_flags);
	print_vdev_tree(NULL, "dedup", config, 0,
	VDEV_ALLOC_BIAS_DEDUP, 0);
	print_vdev_tree(NULL, "special", config, 0,
	VDEV_ALLOC_BIAS_SPECIAL, 0);
	}
	}

	zpool_close(zhp);

	if (ret != 0 \|\| flags.dryrun \|\| !flags.import) {
	nvlist_free(config);
	nvlist_free(props);
	return (ret);
	}

	/*
	* The split was successful. Now we need to open the new
	* pool and import it.
	*/
	if ((zhp = zpool_open_canfail(g_zfs, newpool)) == NULL) {
	nvlist_free(config);
	nvlist_free(props);
	return (1);
	}

	if (loadkeys) {
	ret = zfs_crypto_attempt_load_keys(g_zfs, newpool);
	if (ret != 0)
	ret = 1;
	}

	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
	zpool_enable_datasets(zhp, mntopts, 0) != 0) {
	ret = 1;
	(void) fprintf(stderr, gettext("Split was successful, but "
	"the datasets could not all be mounted\n"));
	(void) fprintf(stderr, gettext("Try doing '%s' with a "
	"different altroot\n"), "zpool import");
	}
	zpool_close(zhp);
	nvlist_free(config);
	nvlist_free(props);

	return (ret);
	}



	/*
	* zpool online <pool> <device> ...
	*/
	int
	zpool_do_online(int argc, char **argv)
	{
	int c, i;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;
	vdev_state_t newstate;
	int flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, "e")) != -1) {
	switch (c) {
	case 'e':
	flags \|= ZFS_ONLINE_EXPAND;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device name\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	for (i = 1; i < argc; i++) {
	if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
	if (newstate != VDEV_STATE_HEALTHY) {
	(void) printf(gettext("warning: device '%s' "
	"onlined, but remains in faulted state\n"),
	argv[i]);
	if (newstate == VDEV_STATE_FAULTED)
	(void) printf(gettext("use 'zpool "
	"clear' to restore a faulted "
	"device\n"));
	else
	(void) printf(gettext("use 'zpool "
	"replace' to replace devices "
	"that are no longer present\n"));
	}
	} else {
	ret = 1;
	}
	}

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool offline [-ft] <pool> <device> ...
	*
	* -f Force the device into a faulted state.
	*
	* -t Only take the device off-line temporarily. The offline/faulted
	* state will not be persistent across reboots.
	*/
	int
	zpool_do_offline(int argc, char **argv)
	{
	int c, i;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;
	boolean_t istmp = B_FALSE;
	boolean_t fault = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "ft")) != -1) {
	switch (c) {
	case 'f':
	fault = B_TRUE;
	break;
	case 't':
	istmp = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device name\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	for (i = 1; i < argc; i++) {
	if (fault) {
	uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
	vdev_aux_t aux;
	if (istmp == B_FALSE) {
	/* Force the fault to persist across imports */
	aux = VDEV_AUX_EXTERNAL_PERSIST;
	} else {
	aux = VDEV_AUX_EXTERNAL;
	}

	if (guid == 0 \|\| zpool_vdev_fault(zhp, guid, aux) != 0)
	ret = 1;
	} else {
	if (zpool_vdev_offline(zhp, argv[i], istmp) != 0)
	ret = 1;
	}
	}

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool clear <pool> [device]
	*
	* Clear all errors associated with a pool or a particular device.
	*/
	int
	zpool_do_clear(int argc, char **argv)
	{
	int c;
	int ret = 0;
	boolean_t dryrun = B_FALSE;
	boolean_t do_rewind = B_FALSE;
	boolean_t xtreme_rewind = B_FALSE;
	uint32_t rewind_policy = ZPOOL_NO_REWIND;
	nvlist_t *policy = NULL;
	zpool_handle_t *zhp;
	char pool, device;

	/* check options */
	while ((c = getopt(argc, argv, "FnX")) != -1) {
	switch (c) {
	case 'F':
	do_rewind = B_TRUE;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'X':
	xtreme_rewind = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((dryrun \|\| xtreme_rewind) && !do_rewind) {
	(void) fprintf(stderr,
	gettext("-n or -X only meaningful with -F\n"));
	usage(B_FALSE);
	}
	if (dryrun)
	rewind_policy = ZPOOL_TRY_REWIND;
	else if (do_rewind)
	rewind_policy = ZPOOL_DO_REWIND;
	if (xtreme_rewind)
	rewind_policy \|= ZPOOL_EXTREME_REWIND;

	/* In future, further rewind policy choices can be passed along here */
	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
	rewind_policy) != 0) {
	return (1);
	}

	pool = argv[0];
	device = argc == 2 ? argv[1] : NULL;

	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
	nvlist_free(policy);
	return (1);
	}

	if (zpool_clear(zhp, device, policy) != 0)
	ret = 1;

	zpool_close(zhp);

	nvlist_free(policy);

	return (ret);
	}

	/*
	* zpool reguid <pool>
	*/
	int
	zpool_do_reguid(int argc, char **argv)
	{
	int c;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	ret = zpool_reguid(zhp);

	zpool_close(zhp);
	return (ret);
	}


	/*
	* zpool reopen <pool>
	*
	* Reopen the pool so that the kernel can update the sizes of all vdevs.
	*/
	int
	zpool_do_reopen(int argc, char **argv)
	{
	int c;
	int ret = 0;
	boolean_t scrub_restart = B_TRUE;

	/* check options */
	while ((c = getopt(argc, argv, "n")) != -1) {
	switch (c) {
	case 'n':
	scrub_restart = B_FALSE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* if argc == 0 we will execute zpool_reopen_one on all pools */
	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_reopen_one, &scrub_restart);

	return (ret);
	}

	typedef struct scrub_cbdata {
	int cb_type;
	pool_scrub_cmd_t cb_scrub_cmd;
	} scrub_cbdata_t;

	static boolean_t
	zpool_has_checkpoint(zpool_handle_t *zhp)
	{
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);

	if (config != NULL) {
	pool_checkpoint_stat_t *pcs = NULL;
	uint_t c;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);

	if (pcs == NULL \|\| pcs->pcs_state == CS_NONE)
	return (B_FALSE);

	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS \|\|
	pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static int
	scrub_callback(zpool_handle_t zhp, void data)
	{
	scrub_cbdata_t *cb = data;
	int err;

	/*
	* Ignore faulted pools.
	*/
	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
	(void) fprintf(stderr, gettext("cannot scan '%s': pool is "
	"currently unavailable\n"), zpool_get_name(zhp));
	return (1);
	}

	err = zpool_scan(zhp, cb->cb_type, cb->cb_scrub_cmd);

	if (err == 0 && zpool_has_checkpoint(zhp) &&
	cb->cb_type == POOL_SCAN_SCRUB) {
	(void) printf(gettext("warning: will not scrub state that "
	"belongs to the checkpoint of pool '%s'\n"),
	zpool_get_name(zhp));
	}

	return (err != 0);
	}

	static int
	wait_callback(zpool_handle_t zhp, void data)
	{
	zpool_wait_activity_t *act = data;
	return (zpool_wait(zhp, *act));
	}

	/*
	* zpool scrub [-s \| -p] [-w] <pool> ...
	*
	* -s Stop. Stops any in-progress scrub.
	* -p Pause. Pause in-progress scrub.
	* -w Wait. Blocks until scrub has completed.
	*/
	int
	zpool_do_scrub(int argc, char **argv)
	{
	int c;
	scrub_cbdata_t cb;
	boolean_t wait = B_FALSE;
	int error;

	cb.cb_type = POOL_SCAN_SCRUB;
	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;

	/* check options */
	while ((c = getopt(argc, argv, "spw")) != -1) {
	switch (c) {
	case 's':
	cb.cb_type = POOL_SCAN_NONE;
	break;
	case 'p':
	cb.cb_scrub_cmd = POOL_SCRUB_PAUSE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (cb.cb_type == POOL_SCAN_NONE &&
	cb.cb_scrub_cmd == POOL_SCRUB_PAUSE) {
	(void) fprintf(stderr, gettext("invalid option combination: "
	"-s and -p are mutually exclusive\n"));
	usage(B_FALSE);
	}

	if (wait && (cb.cb_type == POOL_SCAN_NONE \|\|
	cb.cb_scrub_cmd == POOL_SCRUB_PAUSE)) {
	(void) fprintf(stderr, gettext("invalid option combination: "
	"-w cannot be used with -p or -s\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, scrub_callback, &cb);

	if (wait && !error) {
	zpool_wait_activity_t act = ZPOOL_WAIT_SCRUB;
	error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, wait_callback, &act);
	}

	return (error);
	}

	/*
	* zpool resilver <pool> ...
	*
	* Restarts any in-progress resilver
	*/
	int
	zpool_do_resilver(int argc, char **argv)
	{
	int c;
	scrub_cbdata_t cb;

	cb.cb_type = POOL_SCAN_RESILVER;
	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	return (for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, scrub_callback, &cb));
	}

	/*
	* zpool trim [-d] [-r <rate>] [-c \| -s] <pool> [<device> ...]
	*
	* -c Cancel. Ends any in-progress trim.
	* -d Secure trim. Requires kernel and device support.
	* -r <rate> Sets the TRIM rate in bytes (per second). Supports
	* adding a multiplier suffix such as 'k' or 'm'.
	* -s Suspend. TRIM can then be restarted with no flags.
	* -w Wait. Blocks until trimming has completed.
	*/
	int
	zpool_do_trim(int argc, char **argv)
	{
	struct option long_options[] = {
	{"cancel", no_argument, NULL, 'c'},
	{"secure", no_argument, NULL, 'd'},
	{"rate", required_argument, NULL, 'r'},
	{"suspend", no_argument, NULL, 's'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	pool_trim_func_t cmd_type = POOL_TRIM_START;
	uint64_t rate = 0;
	boolean_t secure = B_FALSE;
	boolean_t wait = B_FALSE;

	int c;
	while ((c = getopt_long(argc, argv, "cdr:sw", long_options, NULL))
	!= -1) {
	switch (c) {
	case 'c':
	if (cmd_type != POOL_TRIM_START &&
	cmd_type != POOL_TRIM_CANCEL) {
	(void) fprintf(stderr, gettext("-c cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_TRIM_CANCEL;
	break;
	case 'd':
	if (cmd_type != POOL_TRIM_START) {
	(void) fprintf(stderr, gettext("-d cannot be "
	"combined with the -c or -s options\n"));
	usage(B_FALSE);
	}
	secure = B_TRUE;
	break;
	case 'r':
	if (cmd_type != POOL_TRIM_START) {
	(void) fprintf(stderr, gettext("-r cannot be "
	"combined with the -c or -s options\n"));
	usage(B_FALSE);
	}
	if (zfs_nicestrtonum(g_zfs, optarg, &rate) == -1) {
	(void) fprintf(stderr, "%s: %s\n",
	gettext("invalid value for rate"),
	libzfs_error_description(g_zfs));
	usage(B_FALSE);
	}
	break;
	case 's':
	if (cmd_type != POOL_TRIM_START &&
	cmd_type != POOL_TRIM_SUSPEND) {
	(void) fprintf(stderr, gettext("-s cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_TRIM_SUSPEND;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	return (-1);
	}

	if (wait && (cmd_type != POOL_TRIM_START)) {
	(void) fprintf(stderr, gettext("-w cannot be used with -c or "
	"-s\n"));
	usage(B_FALSE);
	}

	char *poolname = argv[0];
	zpool_handle_t *zhp = zpool_open(g_zfs, poolname);
	if (zhp == NULL)
	return (-1);

	trimflags_t trim_flags = {
	.secure = secure,
	.rate = rate,
	.wait = wait,
	};

	nvlist_t *vdevs = fnvlist_alloc();
	if (argc == 1) {
	/* no individual leaf vdevs specified, so add them all */
	nvlist_t *config = zpool_get_config(zhp, NULL);
	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	zpool_collect_leaves(zhp, nvroot, vdevs);
	trim_flags.fullpool = B_TRUE;
	} else {
	trim_flags.fullpool = B_FALSE;
	for (int i = 1; i < argc; i++) {
	fnvlist_add_boolean(vdevs, argv[i]);
	}
	}

	int error = zpool_trim(zhp, cmd_type, vdevs, &trim_flags);

	fnvlist_free(vdevs);
	zpool_close(zhp);

	return (error);
	}

	/*
	* Converts a total number of seconds to a human readable string broken
	* down in to days/hours/minutes/seconds.
	*/
	static void
	secs_to_dhms(uint64_t total, char *buf)
	{
	uint64_t days = total / 60 / 60 / 24;
	uint64_t hours = (total / 60 / 60) % 24;
	uint64_t mins = (total / 60) % 60;
	uint64_t secs = (total % 60);

	if (days > 0) {
	(void) sprintf(buf, "%llu days %02llu:%02llu:%02llu",
	(u_longlong_t)days, (u_longlong_t)hours,
	(u_longlong_t)mins, (u_longlong_t)secs);
	} else {
	(void) sprintf(buf, "%02llu:%02llu:%02llu",
	(u_longlong_t)hours, (u_longlong_t)mins,
	(u_longlong_t)secs);
	}
	}

	/*
	* Print out detailed scrub status.
	*/
	static void
	print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
	{
	time_t start, end, pause;
	uint64_t pass_scanned, scanned, pass_issued, issued, total;
	uint64_t elapsed, scan_rate, issue_rate;
	double fraction_done;
	char processed_buf[7], scanned_buf[7], issued_buf[7], total_buf[7];
	char srate_buf[7], irate_buf[7], time_buf[32];

	printf(" ");
	printf_color(ANSI_BOLD, gettext("scan:"));
	printf(" ");

	/* If there's never been a scan, there's not much to say. */
	if (ps == NULL \|\| ps->pss_func == POOL_SCAN_NONE \|\|
	ps->pss_func >= POOL_SCAN_FUNCS) {
	(void) printf(gettext("none requested\n"));
	return;
	}

	start = ps->pss_start_time;
	end = ps->pss_end_time;
	pause = ps->pss_pass_scrub_pause;

	zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf));

	assert(ps->pss_func == POOL_SCAN_SCRUB \|\|
	ps->pss_func == POOL_SCAN_RESILVER);

	/* Scan is finished or canceled. */
	if (ps->pss_state == DSS_FINISHED) {
	secs_to_dhms(end - start, time_buf);

	if (ps->pss_func == POOL_SCAN_SCRUB) {
	(void) printf(gettext("scrub repaired %s "
	"in %s with %llu errors on %s"), processed_buf,
	time_buf, (u_longlong_t)ps->pss_errors,
	ctime(&end));
	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
	(void) printf(gettext("resilvered %s "
	"in %s with %llu errors on %s"), processed_buf,
	time_buf, (u_longlong_t)ps->pss_errors,
	ctime(&end));
	}
	return;
	} else if (ps->pss_state == DSS_CANCELED) {
	if (ps->pss_func == POOL_SCAN_SCRUB) {
	(void) printf(gettext("scrub canceled on %s"),
	ctime(&end));
	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
	(void) printf(gettext("resilver canceled on %s"),
	ctime(&end));
	}
	return;
	}

	assert(ps->pss_state == DSS_SCANNING);

	/* Scan is in progress. Resilvers can't be paused. */
	if (ps->pss_func == POOL_SCAN_SCRUB) {
	if (pause == 0) {
	(void) printf(gettext("scrub in progress since %s"),
	ctime(&start));
	} else {
	(void) printf(gettext("scrub paused since %s"),
	ctime(&pause));
	(void) printf(gettext("\tscrub started on %s"),
	ctime(&start));
	}
	} else if (ps->pss_func == POOL_SCAN_RESILVER) {
	(void) printf(gettext("resilver in progress since %s"),
	ctime(&start));
	}

	scanned = ps->pss_examined;
	pass_scanned = ps->pss_pass_exam;
	issued = ps->pss_issued;
	pass_issued = ps->pss_pass_issued;
	total = ps->pss_to_examine;

	/* we are only done with a block once we have issued the IO for it */
	fraction_done = (double)issued / total;

	/* elapsed time for this pass, rounding up to 1 if it's 0 */
	elapsed = time(NULL) - ps->pss_pass_start;
	elapsed -= ps->pss_pass_scrub_spent_paused;
	elapsed = (elapsed != 0) ? elapsed : 1;

	scan_rate = pass_scanned / elapsed;
	issue_rate = pass_issued / elapsed;
	uint64_t total_secs_left = (issue_rate != 0 && total >= issued) ?
	((total - issued) / issue_rate) : UINT64_MAX;
	secs_to_dhms(total_secs_left, time_buf);

	/* format all of the numbers we will be reporting */
	zfs_nicebytes(scanned, scanned_buf, sizeof (scanned_buf));
	zfs_nicebytes(issued, issued_buf, sizeof (issued_buf));
	zfs_nicebytes(total, total_buf, sizeof (total_buf));
	zfs_nicebytes(scan_rate, srate_buf, sizeof (srate_buf));
	zfs_nicebytes(issue_rate, irate_buf, sizeof (irate_buf));

	/* do not print estimated time if we have a paused scrub */
	if (pause == 0) {
	(void) printf(gettext("\t%s scanned at %s/s, "
	"%s issued at %s/s, %s total\n"),
	scanned_buf, srate_buf, issued_buf, irate_buf, total_buf);
	} else {
	(void) printf(gettext("\t%s scanned, %s issued, %s total\n"),
	scanned_buf, issued_buf, total_buf);
	}

	if (ps->pss_func == POOL_SCAN_RESILVER) {
	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
	processed_buf, 100 * fraction_done);
	} else if (ps->pss_func == POOL_SCAN_SCRUB) {
	(void) printf(gettext("\t%s repaired, %.2f%% done"),
	processed_buf, 100 * fraction_done);
	}

	if (pause == 0) {
	if (total_secs_left != UINT64_MAX &&
	issue_rate >= 10 * 1024 * 1024) {
	(void) printf(gettext(", %s to go\n"), time_buf);
	} else {
	(void) printf(gettext(", no estimated "
	"completion time\n"));
	}
	} else {
	(void) printf(gettext("\n"));
	}
	}

	static void
	print_rebuild_status_impl(vdev_rebuild_stat_t vrs, char vdev_name)
	{
	if (vrs == NULL \|\| vrs->vrs_state == VDEV_REBUILD_NONE)
	return;

	printf(" ");
	printf_color(ANSI_BOLD, gettext("scan:"));
	printf(" ");

	uint64_t bytes_scanned = vrs->vrs_bytes_scanned;
	uint64_t bytes_issued = vrs->vrs_bytes_issued;
	uint64_t bytes_rebuilt = vrs->vrs_bytes_rebuilt;
	uint64_t bytes_est = vrs->vrs_bytes_est;
	uint64_t scan_rate = (vrs->vrs_pass_bytes_scanned /
	(vrs->vrs_pass_time_ms + 1)) * 1000;
	uint64_t issue_rate = (vrs->vrs_pass_bytes_issued /
	(vrs->vrs_pass_time_ms + 1)) * 1000;
	double scan_pct = MIN((double)bytes_scanned * 100 /
	(bytes_est + 1), 100);

	/* Format all of the numbers we will be reporting */
	char bytes_scanned_buf[7], bytes_issued_buf[7];
	char bytes_rebuilt_buf[7], bytes_est_buf[7];
	char scan_rate_buf[7], issue_rate_buf[7], time_buf[32];
	zfs_nicebytes(bytes_scanned, bytes_scanned_buf,
	sizeof (bytes_scanned_buf));
	zfs_nicebytes(bytes_issued, bytes_issued_buf,
	sizeof (bytes_issued_buf));
	zfs_nicebytes(bytes_rebuilt, bytes_rebuilt_buf,
	sizeof (bytes_rebuilt_buf));
	zfs_nicebytes(bytes_est, bytes_est_buf, sizeof (bytes_est_buf));
	zfs_nicebytes(scan_rate, scan_rate_buf, sizeof (scan_rate_buf));
	zfs_nicebytes(issue_rate, issue_rate_buf, sizeof (issue_rate_buf));

	time_t start = vrs->vrs_start_time;
	time_t end = vrs->vrs_end_time;

	/* Rebuild is finished or canceled. */
	if (vrs->vrs_state == VDEV_REBUILD_COMPLETE) {
	secs_to_dhms(vrs->vrs_scan_time_ms / 1000, time_buf);
	(void) printf(gettext("resilvered (%s) %s in %s "
	"with %llu errors on %s"), vdev_name, bytes_rebuilt_buf,
	time_buf, (u_longlong_t)vrs->vrs_errors, ctime(&end));
	return;
	} else if (vrs->vrs_state == VDEV_REBUILD_CANCELED) {
	(void) printf(gettext("resilver (%s) canceled on %s"),
	vdev_name, ctime(&end));
	return;
	} else if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	(void) printf(gettext("resilver (%s) in progress since %s"),
	vdev_name, ctime(&start));
	}

	assert(vrs->vrs_state == VDEV_REBUILD_ACTIVE);

	secs_to_dhms(MAX((int64_t)bytes_est - (int64_t)bytes_scanned, 0) /
	MAX(scan_rate, 1), time_buf);

	(void) printf(gettext("\t%s scanned at %s/s, %s issued %s/s, "
	"%s total\n"), bytes_scanned_buf, scan_rate_buf,
	bytes_issued_buf, issue_rate_buf, bytes_est_buf);
	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
	bytes_rebuilt_buf, scan_pct);

	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	if (scan_rate >= 10 * 1024 * 1024) {
	(void) printf(gettext(", %s to go\n"), time_buf);
	} else {
	(void) printf(gettext(", no estimated "
	"completion time\n"));
	}
	} else {
	(void) printf(gettext("\n"));
	}
	}

	/*
	* Print rebuild status for top-level vdevs.
	*/
	static void
	print_rebuild_status(zpool_handle_t zhp, nvlist_t nvroot)
	{
	nvlist_t **child;
	uint_t children;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	if (nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
	char *name = zpool_vdev_name(g_zfs, zhp,
	child[c], VDEV_NAME_TYPE_ID);
	print_rebuild_status_impl(vrs, name);
	free(name);
	}
	}
	}

	/*
	* As we don't scrub checkpointed blocks, we want to warn the user that we
	* skipped scanning some blocks if a checkpoint exists or existed at any
	* time during the scan. If a sequential instead of healing reconstruction
	* was performed then the blocks were reconstructed. However, their checksums
	* have not been verified so we still print the warning.
	*/
	static void
	print_checkpoint_scan_warning(pool_scan_stat_t ps, pool_checkpoint_stat_t pcs)
	{
	if (ps == NULL \|\| pcs == NULL)
	return;

	if (pcs->pcs_state == CS_NONE \|\|
	pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
	return;

	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS);

	if (ps->pss_state == DSS_NONE)
	return;

	if ((ps->pss_state == DSS_FINISHED \|\| ps->pss_state == DSS_CANCELED) &&
	ps->pss_end_time < pcs->pcs_start_time)
	return;

	if (ps->pss_state == DSS_FINISHED \|\| ps->pss_state == DSS_CANCELED) {
	(void) printf(gettext(" scan warning: skipped blocks "
	"that are only referenced by the checkpoint.\n"));
	} else {
	assert(ps->pss_state == DSS_SCANNING);
	(void) printf(gettext(" scan warning: skipping blocks "
	"that are only referenced by the checkpoint.\n"));
	}
	}

	/*
	* Returns B_TRUE if there is an active rebuild in progress. Otherwise,
	* B_FALSE is returned and 'rebuild_end_time' is set to the end time for
	* the last completed (or cancelled) rebuild.
	*/
	static boolean_t
	check_rebuilding(nvlist_t nvroot, uint64_t rebuild_end_time)
	{
	nvlist_t **child;
	uint_t children;
	boolean_t rebuilding = B_FALSE;
	uint64_t end_time = 0;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	if (nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {

	if (vrs->vrs_end_time > end_time)
	end_time = vrs->vrs_end_time;

	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	rebuilding = B_TRUE;
	end_time = 0;
	break;
	}
	}
	}

	if (rebuild_end_time != NULL)
	*rebuild_end_time = end_time;

	return (rebuilding);
	}

	/*
	* Print the scan status.
	*/
	static void
	print_scan_status(zpool_handle_t zhp, nvlist_t nvroot)
	{
	uint64_t rebuild_end_time = 0, resilver_end_time = 0;
	boolean_t have_resilver = B_FALSE, have_scrub = B_FALSE;
	boolean_t active_resilver = B_FALSE;
	pool_checkpoint_stat_t *pcs = NULL;
	pool_scan_stat_t *ps = NULL;
	uint_t c;

	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
	(uint64_t **)&ps, &c) == 0) {
	if (ps->pss_func == POOL_SCAN_RESILVER) {
	resilver_end_time = ps->pss_end_time;
	active_resilver = (ps->pss_state == DSS_SCANNING);
	}

	have_resilver = (ps->pss_func == POOL_SCAN_RESILVER);
	have_scrub = (ps->pss_func == POOL_SCAN_SCRUB);
	}

	boolean_t active_rebuild = check_rebuilding(nvroot, &rebuild_end_time);
	boolean_t have_rebuild = (active_rebuild \|\| (rebuild_end_time > 0));

	/* Always print the scrub status when available. */
	if (have_scrub)
	print_scan_scrub_resilver_status(ps);

	/*
	* When there is an active resilver or rebuild print its status.
	* Otherwise print the status of the last resilver or rebuild.
	*/
	if (active_resilver \|\| (!active_rebuild && have_resilver &&
	resilver_end_time && resilver_end_time > rebuild_end_time)) {
	print_scan_scrub_resilver_status(ps);
	} else if (active_rebuild \|\| (!active_resilver && have_rebuild &&
	rebuild_end_time && rebuild_end_time > resilver_end_time)) {
	print_rebuild_status(zhp, nvroot);
	}

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	print_checkpoint_scan_warning(ps, pcs);
	}

	/*
	* Print out detailed removal status.
	*/
	static void
	print_removal_status(zpool_handle_t zhp, pool_removal_stat_t prs)
	{
	char copied_buf[7], examined_buf[7], total_buf[7], rate_buf[7];
	time_t start, end;
	nvlist_t config, nvroot;
	nvlist_t **child;
	uint_t children;
	char *vdev_name;

	if (prs == NULL \|\| prs->prs_state == DSS_NONE)
	return;

	/*
	* Determine name of vdev.
	*/
	config = zpool_get_config(zhp, NULL);
	nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0);
	assert(prs->prs_removing_vdev < children);
	vdev_name = zpool_vdev_name(g_zfs, zhp,
	child[prs->prs_removing_vdev], B_TRUE);

	printf_color(ANSI_BOLD, gettext("remove: "));

	start = prs->prs_start_time;
	end = prs->prs_end_time;
	zfs_nicenum(prs->prs_copied, copied_buf, sizeof (copied_buf));

	/*
	* Removal is finished or canceled.
	*/
	if (prs->prs_state == DSS_FINISHED) {
	uint64_t minutes_taken = (end - start) / 60;

	(void) printf(gettext("Removal of vdev %llu copied %s "
	"in %lluh%um, completed on %s"),
	(longlong_t)prs->prs_removing_vdev,
	copied_buf,
	(u_longlong_t)(minutes_taken / 60),
	(uint_t)(minutes_taken % 60),
	ctime((time_t *)&end));
	} else if (prs->prs_state == DSS_CANCELED) {
	(void) printf(gettext("Removal of %s canceled on %s"),
	vdev_name, ctime(&end));
	} else {
	uint64_t copied, total, elapsed, mins_left, hours_left;
	double fraction_done;
	uint_t rate;

	assert(prs->prs_state == DSS_SCANNING);

	/*
	* Removal is in progress.
	*/
	(void) printf(gettext(
	"Evacuation of %s in progress since %s"),
	vdev_name, ctime(&start));

	copied = prs->prs_copied > 0 ? prs->prs_copied : 1;
	total = prs->prs_to_copy;
	fraction_done = (double)copied / total;

	/* elapsed time for this pass */
	elapsed = time(NULL) - prs->prs_start_time;
	elapsed = elapsed > 0 ? elapsed : 1;
	rate = copied / elapsed;
	rate = rate > 0 ? rate : 1;
	mins_left = ((total - copied) / rate) / 60;
	hours_left = mins_left / 60;

	zfs_nicenum(copied, examined_buf, sizeof (examined_buf));
	zfs_nicenum(total, total_buf, sizeof (total_buf));
	zfs_nicenum(rate, rate_buf, sizeof (rate_buf));

	/*
	* do not print estimated time if hours_left is more than
	* 30 days
	*/
	(void) printf(gettext(
	"\t%s copied out of %s at %s/s, %.2f%% done"),
	examined_buf, total_buf, rate_buf, 100 * fraction_done);
	if (hours_left < (30 * 24)) {
	(void) printf(gettext(", %lluh%um to go\n"),
	(u_longlong_t)hours_left, (uint_t)(mins_left % 60));
	} else {
	(void) printf(gettext(
	", (copy is slow, no estimated time)\n"));
	}
	}
	free(vdev_name);

	if (prs->prs_mapping_memory > 0) {
	char mem_buf[7];
	zfs_nicenum(prs->prs_mapping_memory, mem_buf, sizeof (mem_buf));
	(void) printf(gettext(
	"\t%s memory used for removed device mappings\n"),
	mem_buf);
	}
	}

	static void
	print_checkpoint_status(pool_checkpoint_stat_t *pcs)
	{
	time_t start;
	char space_buf[7];

	if (pcs == NULL \|\| pcs->pcs_state == CS_NONE)
	return;

	(void) printf(gettext("checkpoint: "));

	start = pcs->pcs_start_time;
	zfs_nicenum(pcs->pcs_space, space_buf, sizeof (space_buf));

	if (pcs->pcs_state == CS_CHECKPOINT_EXISTS) {
	char *date = ctime(&start);

	/*
	* ctime() adds a newline at the end of the generated
	* string, thus the weird format specifier and the
	* strlen() call used to chop it off from the output.
	*/
	(void) printf(gettext("created %.*s, consumes %s\n"),
	(int)(strlen(date) - 1), date, space_buf);
	return;
	}

	assert(pcs->pcs_state == CS_CHECKPOINT_DISCARDING);

	(void) printf(gettext("discarding, %s remaining.\n"),
	space_buf);
	}

	static void
	print_error_log(zpool_handle_t *zhp)
	{
	nvlist_t *nverrlist = NULL;
	nvpair_t *elem;
	char *pathname;
	size_t len = MAXPATHLEN * 2;

	if (zpool_get_errlog(zhp, &nverrlist) != 0)
	return;

	(void) printf("errors: Permanent errors have been "
	"detected in the following files:\n\n");

	pathname = safe_malloc(len);
	elem = NULL;
	while ((elem = nvlist_next_nvpair(nverrlist, elem)) != NULL) {
	nvlist_t *nv;
	uint64_t dsobj, obj;

	verify(nvpair_value_nvlist(elem, &nv) == 0);
	verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_DATASET,
	&dsobj) == 0);
	verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_OBJECT,
	&obj) == 0);
	zpool_obj_to_path(zhp, dsobj, obj, pathname, len);
	(void) printf("%7s %s\n", "", pathname);
	}
	free(pathname);
	nvlist_free(nverrlist);
	}

	static void
	print_spares(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t **spares,
	uint_t nspares)
	{
	uint_t i;
	char *name;

	if (nspares == 0)
	return;

	(void) printf(gettext("\tspares\n"));

	for (i = 0; i < nspares; i++) {
	name = zpool_vdev_name(g_zfs, zhp, spares[i],
	cb->cb_name_flags);
	print_status_config(zhp, cb, name, spares[i], 2, B_TRUE, NULL);
	free(name);
	}
	}

	static void
	print_l2cache(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t **l2cache,
	uint_t nl2cache)
	{
	uint_t i;
	char *name;

	if (nl2cache == 0)
	return;

	(void) printf(gettext("\tcache\n"));

	for (i = 0; i < nl2cache; i++) {
	name = zpool_vdev_name(g_zfs, zhp, l2cache[i],
	cb->cb_name_flags);
	print_status_config(zhp, cb, name, l2cache[i], 2,
	B_FALSE, NULL);
	free(name);
	}
	}

	static void
	print_dedup_stats(nvlist_t *config)
	{
	ddt_histogram_t *ddh;
	ddt_stat_t *dds;
	ddt_object_t *ddo;
	uint_t c;
	char dspace[6], mspace[6];

	/*
	* If the pool was faulted then we may not have been able to
	* obtain the config. Otherwise, if we have anything in the dedup
	* table continue processing the stats.
	*/
	if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS,
	(uint64_t **)&ddo, &c) != 0)
	return;

	(void) printf("\n");
	(void) printf(gettext(" dedup: "));
	if (ddo->ddo_count == 0) {
	(void) printf(gettext("no DDT entries\n"));
	return;
	}

	zfs_nicebytes(ddo->ddo_dspace, dspace, sizeof (dspace));
	zfs_nicebytes(ddo->ddo_mspace, mspace, sizeof (mspace));
	(void) printf("DDT entries %llu, size %s on disk, %s in core\n",
	(u_longlong_t)ddo->ddo_count,
	dspace,
	mspace);

	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS,
	(uint64_t **)&dds, &c) == 0);
	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM,
	(uint64_t **)&ddh, &c) == 0);
	zpool_dump_ddt(dds, ddh);
	}

	/*
	* Display a summary of pool status. Displays a summary such as:
	*
	* pool: tank
	* status: DEGRADED
	* reason: One or more devices ...
	* see: https://openzfs.github.io/openzfs-docs/msg/ZFS-xxxx-01
	* config:
	* mirror DEGRADED
	* c1t0d0 OK
	* c2t0d0 UNAVAIL
	*
	* When given the '-v' option, we print out the complete config. If the '-e'
	* option is specified, then we print out error rate information as well.
	*/
	static int
	status_callback(zpool_handle_t zhp, void data)
	{
	status_cbdata_t *cbp = data;
	nvlist_t config, nvroot;
	const char *msgid;
	zpool_status_t reason;
	zpool_errata_t errata;
	const char *health;
	uint_t c;
	vdev_stat_t *vs;

	config = zpool_get_config(zhp, NULL);
	reason = zpool_get_status(zhp, &msgid, &errata);

	cbp->cb_count++;

	/*
	* If we were given 'zpool status -x', only report those pools with
	* problems.
	*/
	if (cbp->cb_explain &&
	(reason == ZPOOL_STATUS_OK \|\|
	reason == ZPOOL_STATUS_VERSION_OLDER \|\|
	reason == ZPOOL_STATUS_FEAT_DISABLED \|\|
	reason == ZPOOL_STATUS_COMPATIBILITY_ERR \|\|
	reason == ZPOOL_STATUS_INCOMPATIBLE_FEAT)) {
	if (!cbp->cb_allpools) {
	(void) printf(gettext("pool '%s' is healthy\n"),
	zpool_get_name(zhp));
	if (cbp->cb_first)
	cbp->cb_first = B_FALSE;
	}
	return (0);
	}

	if (cbp->cb_first)
	cbp->cb_first = B_FALSE;
	else
	(void) printf("\n");

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	health = zpool_get_state_str(zhp);

	printf(" ");
	printf_color(ANSI_BOLD, gettext("pool:"));
	printf(" %s\n", zpool_get_name(zhp));
	fputc(' ', stdout);
	printf_color(ANSI_BOLD, gettext("state: "));

	printf_color(health_str_to_color(health), "%s", health);

	fputc('\n', stdout);

	switch (reason) {
	case ZPOOL_STATUS_MISSING_DEV_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be opened. Sufficient replicas exist for\n\tthe pool "
	"to continue functioning in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Attach the missing device "
	"and online it using 'zpool online'.\n"));
	break;

	case ZPOOL_STATUS_MISSING_DEV_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be opened. There are insufficient\n\treplicas for the"
	" pool to continue functioning.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Attach the missing device "
	"and online it using 'zpool online'.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be used because the label is missing or\n\tinvalid. "
	"Sufficient replicas exist for the pool to continue\n\t"
	"functioning in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Replace the device using "
	"'zpool replace'.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be used because the label is missing \n\tor invalid. "
	"There are insufficient replicas for the pool to "
	"continue\n\tfunctioning.\n"));
	zpool_explain_recover(zpool_get_handle(zhp),
	zpool_get_name(zhp), reason, config);
	break;

	case ZPOOL_STATUS_FAILING_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"experienced an unrecoverable error. An\n\tattempt was "
	"made to correct the error. Applications are "
	"unaffected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Determine if the "
	"device needs to be replaced, and clear the errors\n\tusing"
	" 'zpool clear' or replace the device with 'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_OFFLINE_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"been taken offline by the administrator.\n\tSufficient "
	"replicas exist for the pool to continue functioning in "
	"a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Online the device "
	"using 'zpool online' or replace the device with\n\t'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_REMOVED_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"been removed by the administrator.\n\tSufficient "
	"replicas exist for the pool to continue functioning in "
	"a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Online the device "
	"using zpool online' or replace the device with\n\t'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_RESILVERING:
	case ZPOOL_STATUS_REBUILDING:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices is "
	"currently being resilvered. The pool will\n\tcontinue "
	"to function, possibly in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Wait for the resilver to "
	"complete.\n"));
	break;

	case ZPOOL_STATUS_REBUILD_SCRUB:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices have "
	"been sequentially resilvered, scrubbing\n\tthe pool "
	"is recommended.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Use 'zpool scrub' to "
	"verify all data checksums.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_DATA:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"experienced an error resulting in data\n\tcorruption. "
	"Applications may be affected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Restore the file in question"
	" if possible. Otherwise restore the\n\tentire pool from "
	"backup.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_POOL:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool metadata is "
	"corrupted and the pool cannot be opened.\n"));
	zpool_explain_recover(zpool_get_handle(zhp),
	zpool_get_name(zhp), reason, config);
	break;

	case ZPOOL_STATUS_VERSION_OLDER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"a legacy on-disk format. The pool can\n\tstill be used, "
	"but some features are unavailable.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Upgrade the pool using "
	"'zpool upgrade'. Once this is done, the\n\tpool will no "
	"longer be accessible on software that does not support\n\t"
	"feature flags.\n"));
	break;

	case ZPOOL_STATUS_VERSION_NEWER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool has been upgraded "
	"to a newer, incompatible on-disk version.\n\tThe pool "
	"cannot be accessed on this system.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Access the pool from a "
	"system running more recent software, or\n\trestore the "
	"pool from backup.\n"));
	break;

	case ZPOOL_STATUS_FEAT_DISABLED:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Some supported and "
	"requested features are not enabled on the pool.\n\t"
	"The pool can still be used, but some features are "
	"unavailable.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Enable all features using "
	"'zpool upgrade'. Once this is done,\n\tthe pool may no "
	"longer be accessible by software that does not support\n\t"
	"the features. See zpool-features(7) for details.\n"));
	break;

	case ZPOOL_STATUS_COMPATIBILITY_ERR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("This pool has a "
	"compatibility list specified, but it could not be\n\t"
	"read/parsed at this time. The pool can still be used, "
	"but this\n\tshould be investigated.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Check the value of the "
	"'compatibility' property against the\n\t"
	"appropriate file in " ZPOOL_SYSCONF_COMPAT_D " or "
	ZPOOL_DATA_COMPAT_D ".\n"));
	break;

	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more features "
	"are enabled on the pool despite not being\n\t"
	"requested by the 'compatibility' property.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Consider setting "
	"'compatibility' to an appropriate value, or\n\t"
	"adding needed features to the relevant file in\n\t"
	ZPOOL_SYSCONF_COMPAT_D " or " ZPOOL_DATA_COMPAT_D ".\n"));
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
	"on this system because it uses the\n\tfollowing feature(s)"
	" not supported on this system:\n"));
	zpool_print_unsup_feat(config);
	(void) printf("\n");
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Access the pool from a "
	"system that supports the required feature(s),\n\tor "
	"restore the pool from backup.\n"));
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool can only be "
	"accessed in read-only mode on this system. It\n\tcannot be"
	" accessed in read-write mode because it uses the "
	"following\n\tfeature(s) not supported on this system:\n"));
	zpool_print_unsup_feat(config);
	(void) printf("\n");
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
	"in read-write mode. Import the pool with\n"
	"\t\"-o readonly=on\", access the pool from a system that "
	"supports the\n\trequired feature(s), or restore the "
	"pool from backup.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to persistent errors.\n\tSufficient "
	"replicas exist for the pool to continue functioning "
	"in a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Replace the faulted device, "
	"or use 'zpool clear' to mark the device\n\trepaired.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to persistent errors. There are "
	"insufficient replicas for the pool to\n\tcontinue "
	"functioning.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Destroy and re-create the "
	"pool from a backup source. Manually marking the device\n"
	"\trepaired using 'zpool clear' may allow some data "
	"to be recovered.\n"));
	break;

	case ZPOOL_STATUS_IO_FAILURE_MMP:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is suspended "
	"because multihost writes failed or were delayed;\n\t"
	"another system could import the pool undetected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices"
	" are connected, then reboot your system and\n\timport the "
	"pool.\n"));
	break;

	case ZPOOL_STATUS_IO_FAILURE_WAIT:
	case ZPOOL_STATUS_IO_FAILURE_CONTINUE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to IO failures.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Make sure the affected "
	"devices are connected, then run 'zpool clear'.\n"));
	break;

	case ZPOOL_STATUS_BAD_LOG:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("An intent log record "
	"could not be read.\n"
	"\tWaiting for administrator intervention to fix the "
	"faulted pool.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Either restore the affected "
	"device(s) and run 'zpool online',\n"
	"\tor ignore the intent log records by running "
	"'zpool clear'.\n"));
	break;

	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
	(void) printf(gettext("status: One or more devices are "
	"configured to use a non-native block size.\n"
	"\tExpect reduced performance.\n"));
	(void) printf(gettext("action: Replace affected devices with "
	"devices that support the\n\tconfigured block size, or "
	"migrate data to a properly configured\n\tpool.\n"));
	break;

	case ZPOOL_STATUS_HOSTID_MISMATCH:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Mismatch between pool hostid"
	" and system hostid on imported pool.\n\tThis pool was "
	"previously imported into a system with a different "
	"hostid,\n\tand then was verbatim imported into this "
	"system.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Export this pool on all "
	"systems on which it is imported.\n"
	"\tThen import it to correct the mismatch.\n"));
	break;

	case ZPOOL_STATUS_ERRATA:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
	errata);

	switch (errata) {
	case ZPOOL_ERRATA_NONE:
	break;

	case ZPOOL_ERRATA_ZOL_2094_SCRUB:
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the issue"
	" run 'zpool scrub'.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
	(void) printf(gettext("\tExisting encrypted datasets "
	"contain an on-disk incompatibility\n\twhich "
	"needs to be corrected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the issue"
	" backup existing encrypted datasets to new\n\t"
	"encrypted datasets and destroy the old ones. "
	"'zfs mount -o ro' can\n\tbe used to temporarily "
	"mount existing encrypted datasets readonly.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
	(void) printf(gettext("\tExisting encrypted snapshots "
	"and bookmarks contain an on-disk\n\tincompat"
	"ibility. This may cause on-disk corruption if "
	"they are used\n\twith 'zfs recv'.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the"
	"issue, enable the bookmark_v2 feature. No "
	"additional\n\taction is needed if there are no "
	"encrypted snapshots or bookmarks.\n\tIf preserving"
	"the encrypted snapshots and bookmarks is required,"
	" use\n\ta non-raw send to backup and restore them."
	" Alternately, they may be\n\tremoved to resolve "
	"the incompatibility.\n"));
	break;

	default:
	/*
	* All errata which allow the pool to be imported
	* must contain an action message.
	*/
	assert(0);
	}
	break;

	default:
	/*
	* The remaining errors can't actually be generated, yet.
	*/
	assert(reason == ZPOOL_STATUS_OK);
	}

	if (msgid != NULL) {
	printf(" ");
	printf_color(ANSI_BOLD, gettext("see:"));
	printf(gettext(
	" https://openzfs.github.io/openzfs-docs/msg/%s\n"),
	msgid);
	}

	if (config != NULL) {
	uint64_t nerr;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;
	pool_checkpoint_stat_t *pcs = NULL;
	pool_removal_stat_t *prs = NULL;

	print_scan_status(zhp, nvroot);

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
	print_removal_status(zhp, prs);

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	print_checkpoint_status(pcs);

	cbp->cb_namewidth = max_width(zhp, nvroot, 0, 0,
	cbp->cb_name_flags \| VDEV_NAME_TYPE_ID);
	if (cbp->cb_namewidth < 10)
	cbp->cb_namewidth = 10;

	color_start(ANSI_BOLD);
	(void) printf(gettext("config:\n\n"));
	(void) printf(gettext("\t%-*s %-8s %5s %5s %5s"),
	cbp->cb_namewidth, "NAME", "STATE", "READ", "WRITE",
	"CKSUM");
	color_end();

	if (cbp->cb_print_slow_ios) {
	printf_color(ANSI_BOLD, " %5s", gettext("SLOW"));
	}

	if (cbp->vcdl != NULL)
	print_cmd_columns(cbp->vcdl, 0);

	printf("\n");

	print_status_config(zhp, cbp, zpool_get_name(zhp), nvroot, 0,
	B_FALSE, NULL);

	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_DEDUP);
	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_CLASS_LOGS);

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0)
	print_l2cache(zhp, cbp, l2cache, nl2cache);

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0)
	print_spares(zhp, cbp, spares, nspares);

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
	&nerr) == 0) {
	nvlist_t *nverrlist = NULL;

	/*
	* If the approximate error count is small, get a
	* precise count by fetching the entire log and
	* uniquifying the results.
	*/
	if (nerr > 0 && nerr < 100 && !cbp->cb_verbose &&
	zpool_get_errlog(zhp, &nverrlist) == 0) {
	nvpair_t *elem;

	elem = NULL;
	nerr = 0;
	while ((elem = nvlist_next_nvpair(nverrlist,
	elem)) != NULL) {
	nerr++;
	}
	}
	nvlist_free(nverrlist);

	(void) printf("\n");

	if (nerr == 0)
	(void) printf(gettext("errors: No known data "
	"errors\n"));
	else if (!cbp->cb_verbose)
	(void) printf(gettext("errors: %llu data "
	"errors, use '-v' for a list\n"),
	(u_longlong_t)nerr);
	else
	print_error_log(zhp);
	}

	if (cbp->cb_dedup_stats)
	print_dedup_stats(config);
	} else {
	(void) printf(gettext("config: The configuration cannot be "
	"determined.\n"));
	}

	return (0);
	}

	/*
	* zpool status [-c [script1,script2,...]] [-igLpPstvx] [-T d\|u] [pool] ...
	* [interval [count]]
	*
	* -c CMD For each vdev, run command CMD
	* -i Display vdev initialization status.
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -p Display values in parsable (exact) format.
	* -P Display full path for vdev name.
	* -s Display slow IOs column.
	* -v Display complete error logs
	* -x Display only pools with potential problems
	* -D Display dedup status (undocumented)
	* -t Display vdev TRIM status.
	* -T Display a timestamp in date(1) or Unix format
	*
	* Describes the health status of all pools or some subset.
	*/
	int
	zpool_do_status(int argc, char **argv)
	{
	int c;
	int ret;
	float interval = 0;
	unsigned long count = 0;
	status_cbdata_t cb = { 0 };
	char *cmd = NULL;

	/* check options */
	while ((c = getopt(argc, argv, "c:igLpPsvxDtT:")) != -1) {
	switch (c) {
	case 'c':
	if (cmd != NULL) {
	fprintf(stderr,
	gettext("Can't set -c flag twice\n"));
	exit(1);
	}

	if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
	fprintf(stderr, gettext(
	"Can't run -c, disabled by "
	"ZPOOL_SCRIPTS_ENABLED.\n"));
	exit(1);
	}

	if ((getuid() <= 0 \|\| geteuid() <= 0) &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
	fprintf(stderr, gettext(
	"Can't run -c with root privileges "
	"unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
	exit(1);
	}
	cmd = optarg;
	break;
	case 'i':
	cb.cb_print_vdev_init = B_TRUE;
	break;
	case 'g':
	cb.cb_name_flags \|= VDEV_NAME_GUID;
	break;
	case 'L':
	cb.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'P':
	cb.cb_name_flags \|= VDEV_NAME_PATH;
	break;
	case 's':
	cb.cb_print_slow_ios = B_TRUE;
	break;
	case 'v':
	cb.cb_verbose = B_TRUE;
	break;
	case 'x':
	cb.cb_explain = B_TRUE;
	break;
	case 'D':
	cb.cb_dedup_stats = B_TRUE;
	break;
	case 't':
	cb.cb_print_vdev_trim = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case '?':
	if (optopt == 'c') {
	print_zpool_script_list("status");
	exit(0);
	} else {
	fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &interval, &count);

	if (argc == 0)
	cb.cb_allpools = B_TRUE;

	cb.cb_first = B_TRUE;
	cb.cb_print_status = B_TRUE;

	for (;;) {
	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (cmd != NULL)
	cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd,
	NULL, NULL, 0, 0);

	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	cb.cb_literal, status_callback, &cb);

	if (cb.vcdl != NULL)
	free_vdev_cmd_data_list(cb.vcdl);

	if (argc == 0 && cb.cb_count == 0)
	(void) fprintf(stderr, gettext("no pools available\n"));
	else if (cb.cb_explain && cb.cb_first && cb.cb_allpools)
	(void) printf(gettext("all pools are healthy\n"));

	if (ret != 0)
	return (ret);

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	(void) fsleep(interval);
	}

	return (0);
	}

	typedef struct upgrade_cbdata {
	int cb_first;
	int cb_argc;
	uint64_t cb_version;
	char **cb_argv;
	} upgrade_cbdata_t;

	static int
	check_unsupp_fs(zfs_handle_t zhp, void unsupp_fs)
	{
	int zfs_version = (int)zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
	int count = (int )unsupp_fs;

	if (zfs_version > ZPL_VERSION) {
	(void) printf(gettext("%s (v%d) is not supported by this "
	"implementation of ZFS.\n"),
	zfs_get_name(zhp), zfs_version);
	(*count)++;
	}

	zfs_iter_filesystems(zhp, check_unsupp_fs, unsupp_fs);

	zfs_close(zhp);

	return (0);
	}

	static int
	upgrade_version(zpool_handle_t *zhp, uint64_t version)
	{
	int ret;
	nvlist_t *config;
	uint64_t oldversion;
	int unsupp_fs = 0;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&oldversion) == 0);

	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
	ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	assert(SPA_VERSION_IS_SUPPORTED(oldversion));
	assert(oldversion < version);

	ret = zfs_iter_root(zpool_get_handle(zhp), check_unsupp_fs, &unsupp_fs);
	if (ret != 0)
	return (ret);

	if (unsupp_fs) {
	(void) fprintf(stderr, gettext("Upgrade not performed due "
	"to %d unsupported filesystems (max v%d).\n"),
	unsupp_fs, (int)ZPL_VERSION);
	return (1);
	}

	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
	(void) fprintf(stderr, gettext("Upgrade not performed because "
	"'compatibility' property set to '"
	ZPOOL_COMPAT_LEGACY "'.\n"));
	return (1);
	}

	ret = zpool_upgrade(zhp, version);
	if (ret != 0)
	return (ret);

	if (version >= SPA_VERSION_FEATURES) {
	(void) printf(gettext("Successfully upgraded "
	"'%s' from version %llu to feature flags.\n"),
	zpool_get_name(zhp), (u_longlong_t)oldversion);
	} else {
	(void) printf(gettext("Successfully upgraded "
	"'%s' from version %llu to version %llu.\n"),
	zpool_get_name(zhp), (u_longlong_t)oldversion,
	(u_longlong_t)version);
	}

	return (0);
	}

	static int
	upgrade_enable_all(zpool_handle_t zhp, int countp)
	{
	int i, ret, count;
	boolean_t firstff = B_TRUE;
	nvlist_t *enabled = zpool_get_features(zhp);

	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
	ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	boolean_t requested_features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, requested_features) !=
	ZPOOL_COMPATIBILITY_OK)
	return (-1);

	count = 0;
	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fname = spa_feature_table[i].fi_uname;
	const char *fguid = spa_feature_table[i].fi_guid;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	if (!nvlist_exists(enabled, fguid) && requested_features[i]) {
	char *propname;
	verify(-1 != asprintf(&propname, "feature@%s", fname));
	ret = zpool_set_prop(zhp, propname,
	ZFS_FEATURE_ENABLED);
	if (ret != 0) {
	free(propname);
	return (ret);
	}
	count++;

	if (firstff) {
	(void) printf(gettext("Enabled the "
	"following features on '%s':\n"),
	zpool_get_name(zhp));
	firstff = B_FALSE;
	}
	(void) printf(gettext(" %s\n"), fname);
	free(propname);
	}
	}

	if (countp != NULL)
	*countp = count;
	return (0);
	}

	static int
	upgrade_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;
	boolean_t modified_pool = B_FALSE;
	int ret;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	assert(SPA_VERSION_IS_SUPPORTED(version));

	if (version < cbp->cb_version) {
	cbp->cb_first = B_FALSE;
	ret = upgrade_version(zhp, cbp->cb_version);
	if (ret != 0)
	return (ret);
	modified_pool = B_TRUE;

	/*
	* If they did "zpool upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
	int count;
	ret = upgrade_enable_all(zhp, &count);
	if (ret != 0)
	return (ret);

	if (count > 0) {
	cbp->cb_first = B_FALSE;
	modified_pool = B_TRUE;
	}
	}

	if (modified_pool) {
	(void) printf("\n");
	(void) after_zpool_upgrade(zhp);
	}

	return (0);
	}

	static int
	upgrade_list_older_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	assert(SPA_VERSION_IS_SUPPORTED(version));

	if (version < SPA_VERSION_FEATURES) {
	if (cbp->cb_first) {
	(void) printf(gettext("The following pools are "
	"formatted with legacy version numbers and can\n"
	"be upgraded to use feature flags. After "
	"being upgraded, these pools\nwill no "
	"longer be accessible by software that does not "
	"support feature\nflags.\n\n"
	"Note that setting a pool's 'compatibility' "
	"feature to '" ZPOOL_COMPAT_LEGACY "' will\n"
	"inhibit upgrades.\n\n"));
	(void) printf(gettext("VER POOL\n"));
	(void) printf(gettext("--- ------------\n"));
	cbp->cb_first = B_FALSE;
	}

	(void) printf("%2llu %s\n", (u_longlong_t)version,
	zpool_get_name(zhp));
	}

	return (0);
	}

	static int
	upgrade_list_disabled_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	if (version >= SPA_VERSION_FEATURES) {
	int i;
	boolean_t poolfirst = B_TRUE;
	nvlist_t *enabled = zpool_get_features(zhp);

	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fguid = spa_feature_table[i].fi_guid;
	const char *fname = spa_feature_table[i].fi_uname;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	if (!nvlist_exists(enabled, fguid)) {
	if (cbp->cb_first) {
	(void) printf(gettext("\nSome "
	"supported features are not "
	"enabled on the following pools. "
	"Once a\nfeature is enabled the "
	"pool may become incompatible with "
	"software\nthat does not support "
	"the feature. See "
	"zpool-features(7) for "
	"details.\n\n"
	"Note that the pool "
	"'compatibility' feature can be "
	"used to inhibit\nfeature "
	"upgrades.\n\n"));
	(void) printf(gettext("POOL "
	"FEATURE\n"));
	(void) printf(gettext("------"
	"---------\n"));
	cbp->cb_first = B_FALSE;
	}

	if (poolfirst) {
	(void) printf(gettext("%s\n"),
	zpool_get_name(zhp));
	poolfirst = B_FALSE;
	}

	(void) printf(gettext(" %s\n"), fname);
	}
	/*
	* If they did "zpool upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}
	}

	return (0);
	}

	static int
	upgrade_one(zpool_handle_t zhp, void data)
	{
	boolean_t modified_pool = B_FALSE;
	upgrade_cbdata_t *cbp = data;
	uint64_t cur_version;
	int ret;

	if (strcmp("log", zpool_get_name(zhp)) == 0) {
	(void) fprintf(stderr, gettext("'log' is now a reserved word\n"
	"Pool 'log' must be renamed using export and import"
	" to upgrade.\n"));
	return (1);
	}

	cur_version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
	if (cur_version > cbp->cb_version) {
	(void) printf(gettext("Pool '%s' is already formatted "
	"using more current version '%llu'.\n\n"),
	zpool_get_name(zhp), (u_longlong_t)cur_version);
	return (0);
	}

	if (cbp->cb_version != SPA_VERSION && cur_version == cbp->cb_version) {
	(void) printf(gettext("Pool '%s' is already formatted "
	"using version %llu.\n\n"), zpool_get_name(zhp),
	(u_longlong_t)cbp->cb_version);
	return (0);
	}

	if (cur_version != cbp->cb_version) {
	modified_pool = B_TRUE;
	ret = upgrade_version(zhp, cbp->cb_version);
	if (ret != 0)
	return (ret);
	}

	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
	int count = 0;
	ret = upgrade_enable_all(zhp, &count);
	if (ret != 0)
	return (ret);

	if (count != 0) {
	modified_pool = B_TRUE;
	} else if (cur_version == SPA_VERSION) {
	(void) printf(gettext("Pool '%s' already has all "
	"supported and requested features enabled.\n"),
	zpool_get_name(zhp));
	}
	}

	if (modified_pool) {
	(void) printf("\n");
	(void) after_zpool_upgrade(zhp);
	}

	return (0);
	}

	/*
	* zpool upgrade
	* zpool upgrade -v
	* zpool upgrade [-V version] <-a \| pool ...>
	*
	* With no arguments, display downrev'd ZFS pool available for upgrade.
	* Individual pools can be upgraded by specifying the pool, and '-a' will
	* upgrade all pools.
	*/
	int
	zpool_do_upgrade(int argc, char **argv)
	{
	int c;
	upgrade_cbdata_t cb = { 0 };
	int ret = 0;
	boolean_t showversions = B_FALSE;
	boolean_t upgradeall = B_FALSE;
	char *end;


	/* check options */
	while ((c = getopt(argc, argv, ":avV:")) != -1) {
	switch (c) {
	case 'a':
	upgradeall = B_TRUE;
	break;
	case 'v':
	showversions = B_TRUE;
	break;
	case 'V':
	cb.cb_version = strtoll(optarg, &end, 10);
	if (*end != '\0' \|\|
	!SPA_VERSION_IS_SUPPORTED(cb.cb_version)) {
	(void) fprintf(stderr,
	gettext("invalid version '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	cb.cb_argc = argc;
	cb.cb_argv = argv;
	argc -= optind;
	argv += optind;

	if (cb.cb_version == 0) {
	cb.cb_version = SPA_VERSION;
	} else if (!upgradeall && argc == 0) {
	(void) fprintf(stderr, gettext("-V option is "
	"incompatible with other arguments\n"));
	usage(B_FALSE);
	}

	if (showversions) {
	if (upgradeall \|\| argc != 0) {
	(void) fprintf(stderr, gettext("-v option is "
	"incompatible with other arguments\n"));
	usage(B_FALSE);
	}
	} else if (upgradeall) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("-a option should not "
	"be used along with a pool name\n"));
	usage(B_FALSE);
	}
	}

	(void) printf("%s", gettext("This system supports ZFS pool feature "
	"flags.\n\n"));
	if (showversions) {
	int i;

	(void) printf(gettext("The following features are "
	"supported:\n\n"));
	(void) printf(gettext("FEAT DESCRIPTION\n"));
	(void) printf("----------------------------------------------"
	"---------------\n");
	for (i = 0; i < SPA_FEATURES; i++) {
	zfeature_info_t *fi = &spa_feature_table[i];
	if (!fi->fi_zfs_mod_supported)
	continue;
	const char *ro =
	(fi->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ?
	" (read-only compatible)" : "";

	(void) printf("%-37s%s\n", fi->fi_uname, ro);
	(void) printf(" %s\n", fi->fi_desc);
	}
	(void) printf("\n");

	(void) printf(gettext("The following legacy versions are also "
	"supported:\n\n"));
	(void) printf(gettext("VER DESCRIPTION\n"));
	(void) printf("--- -----------------------------------------"
	"---------------\n");
	(void) printf(gettext(" 1 Initial ZFS version\n"));
	(void) printf(gettext(" 2 Ditto blocks "
	"(replicated metadata)\n"));
	(void) printf(gettext(" 3 Hot spares and double parity "
	"RAID-Z\n"));
	(void) printf(gettext(" 4 zpool history\n"));
	(void) printf(gettext(" 5 Compression using the gzip "
	"algorithm\n"));
	(void) printf(gettext(" 6 bootfs pool property\n"));
	(void) printf(gettext(" 7 Separate intent log devices\n"));
	(void) printf(gettext(" 8 Delegated administration\n"));
	(void) printf(gettext(" 9 refquota and refreservation "
	"properties\n"));
	(void) printf(gettext(" 10 Cache devices\n"));
	(void) printf(gettext(" 11 Improved scrub performance\n"));
	(void) printf(gettext(" 12 Snapshot properties\n"));
	(void) printf(gettext(" 13 snapused property\n"));
	(void) printf(gettext(" 14 passthrough-x aclinherit\n"));
	(void) printf(gettext(" 15 user/group space accounting\n"));
	(void) printf(gettext(" 16 stmf property support\n"));
	(void) printf(gettext(" 17 Triple-parity RAID-Z\n"));
	(void) printf(gettext(" 18 Snapshot user holds\n"));
	(void) printf(gettext(" 19 Log device removal\n"));
	(void) printf(gettext(" 20 Compression using zle "
	"(zero-length encoding)\n"));
	(void) printf(gettext(" 21 Deduplication\n"));
	(void) printf(gettext(" 22 Received properties\n"));
	(void) printf(gettext(" 23 Slim ZIL\n"));
	(void) printf(gettext(" 24 System attributes\n"));
	(void) printf(gettext(" 25 Improved scrub stats\n"));
	(void) printf(gettext(" 26 Improved snapshot deletion "
	"performance\n"));
	(void) printf(gettext(" 27 Improved snapshot creation "
	"performance\n"));
	(void) printf(gettext(" 28 Multiple vdev replacements\n"));
	(void) printf(gettext("\nFor more information on a particular "
	"version, including supported releases,\n"));
	(void) printf(gettext("see the ZFS Administration Guide.\n\n"));
	} else if (argc == 0 && upgradeall) {
	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_cb, &cb);
	if (ret == 0 && cb.cb_first) {
	if (cb.cb_version == SPA_VERSION) {
	(void) printf(gettext("All pools are already "
	"formatted using feature flags.\n\n"));
	(void) printf(gettext("Every feature flags "
	"pool already has all supported and "
	"requested features enabled.\n"));
	} else {
	(void) printf(gettext("All pools are already "
	"formatted with version %llu or higher.\n"),
	(u_longlong_t)cb.cb_version);
	}
	}
	} else if (argc == 0) {
	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_list_older_cb, &cb);
	assert(ret == 0);

	if (cb.cb_first) {
	(void) printf(gettext("All pools are formatted "
	"using feature flags.\n\n"));
	} else {
	(void) printf(gettext("\nUse 'zpool upgrade -v' "
	"for a list of available legacy versions.\n"));
	}

	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_list_disabled_cb, &cb);
	assert(ret == 0);

	if (cb.cb_first) {
	(void) printf(gettext("Every feature flags pool has "
	"all supported and requested features enabled.\n"));
	} else {
	(void) printf(gettext("\n"));
	}
	} else {
	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, upgrade_one, &cb);
	}

	return (ret);
	}

	typedef struct hist_cbdata {
	boolean_t first;
	boolean_t longfmt;
	boolean_t internal;
	} hist_cbdata_t;

	static void
	print_history_records(nvlist_t nvhis, hist_cbdata_t cb)
	{
	nvlist_t **records;
	uint_t numrecords;
	int i;

	verify(nvlist_lookup_nvlist_array(nvhis, ZPOOL_HIST_RECORD,
	&records, &numrecords) == 0);
	for (i = 0; i < numrecords; i++) {
	nvlist_t *rec = records[i];
	char tbuf[64] = "";

	if (nvlist_exists(rec, ZPOOL_HIST_TIME)) {
	time_t tsec;
	struct tm t;

	tsec = fnvlist_lookup_uint64(records[i],
	ZPOOL_HIST_TIME);
	(void) localtime_r(&tsec, &t);
	(void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t);
	}

	if (nvlist_exists(rec, ZPOOL_HIST_ELAPSED_NS)) {
	uint64_t elapsed_ns = fnvlist_lookup_int64(records[i],
	ZPOOL_HIST_ELAPSED_NS);
	(void) snprintf(tbuf + strlen(tbuf),
	sizeof (tbuf) - strlen(tbuf),
	" (%lldms)", (long long)elapsed_ns / 1000 / 1000);
	}

	if (nvlist_exists(rec, ZPOOL_HIST_CMD)) {
	(void) printf("%s %s", tbuf,
	fnvlist_lookup_string(rec, ZPOOL_HIST_CMD));
	} else if (nvlist_exists(rec, ZPOOL_HIST_INT_EVENT)) {
	int ievent =
	fnvlist_lookup_uint64(rec, ZPOOL_HIST_INT_EVENT);
	if (!cb->internal)
	continue;
	if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS) {
	(void) printf("%s unrecognized record:\n",
	tbuf);
	dump_nvlist(rec, 4);
	continue;
	}
	(void) printf("%s [internal %s txg:%lld] %s", tbuf,
	zfs_history_event_names[ievent],
	(longlong_t)fnvlist_lookup_uint64(
	rec, ZPOOL_HIST_TXG),
	fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(rec, ZPOOL_HIST_INT_NAME)) {
	if (!cb->internal)
	continue;
	(void) printf("%s [txg:%lld] %s", tbuf,
	(longlong_t)fnvlist_lookup_uint64(
	rec, ZPOOL_HIST_TXG),
	fnvlist_lookup_string(rec, ZPOOL_HIST_INT_NAME));
	if (nvlist_exists(rec, ZPOOL_HIST_DSNAME)) {
	(void) printf(" %s (%llu)",
	fnvlist_lookup_string(rec,
	ZPOOL_HIST_DSNAME),
	(u_longlong_t)fnvlist_lookup_uint64(rec,
	ZPOOL_HIST_DSID));
	}
	(void) printf(" %s", fnvlist_lookup_string(rec,
	ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(rec, ZPOOL_HIST_IOCTL)) {
	if (!cb->internal)
	continue;
	(void) printf("%s ioctl %s\n", tbuf,
	fnvlist_lookup_string(rec, ZPOOL_HIST_IOCTL));
	if (nvlist_exists(rec, ZPOOL_HIST_INPUT_NVL)) {
	(void) printf(" input:\n");
	dump_nvlist(fnvlist_lookup_nvlist(rec,
	ZPOOL_HIST_INPUT_NVL), 8);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_NVL)) {
	(void) printf(" output:\n");
	dump_nvlist(fnvlist_lookup_nvlist(rec,
	ZPOOL_HIST_OUTPUT_NVL), 8);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_SIZE)) {
	(void) printf(" output nvlist omitted; "
	"original size: %lldKB\n",
	(longlong_t)fnvlist_lookup_int64(rec,
	ZPOOL_HIST_OUTPUT_SIZE) / 1024);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_ERRNO)) {
	(void) printf(" errno: %lld\n",
	(longlong_t)fnvlist_lookup_int64(rec,
	ZPOOL_HIST_ERRNO));
	}
	} else {
	if (!cb->internal)
	continue;
	(void) printf("%s unrecognized record:\n", tbuf);
	dump_nvlist(rec, 4);
	}

	if (!cb->longfmt) {
	(void) printf("\n");
	continue;
	}
	(void) printf(" [");
	if (nvlist_exists(rec, ZPOOL_HIST_WHO)) {
	uid_t who = fnvlist_lookup_uint64(rec, ZPOOL_HIST_WHO);
	struct passwd *pwd = getpwuid(who);
	(void) printf("user %d ", (int)who);
	if (pwd != NULL)
	(void) printf("(%s) ", pwd->pw_name);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_HOST)) {
	(void) printf("on %s",
	fnvlist_lookup_string(rec, ZPOOL_HIST_HOST));
	}
	if (nvlist_exists(rec, ZPOOL_HIST_ZONE)) {
	(void) printf(":%s",
	fnvlist_lookup_string(rec, ZPOOL_HIST_ZONE));
	}

	(void) printf("]");
	(void) printf("\n");
	}
	}

	/*
	* Print out the command history for a specific pool.
	*/
	static int
	get_history_one(zpool_handle_t zhp, void data)
	{
	nvlist_t *nvhis;
	int ret;
	hist_cbdata_t cb = (hist_cbdata_t )data;
	uint64_t off = 0;
	boolean_t eof = B_FALSE;

	cb->first = B_FALSE;

	(void) printf(gettext("History for '%s':\n"), zpool_get_name(zhp));

	while (!eof) {
	if ((ret = zpool_get_history(zhp, &nvhis, &off, &eof)) != 0)
	return (ret);

	print_history_records(nvhis, cb);
	nvlist_free(nvhis);
	}
	(void) printf("\n");

	return (ret);
	}

	/*
	* zpool history <pool>
	*
	* Displays the history of commands that modified pools.
	*/
	int
	zpool_do_history(int argc, char **argv)
	{
	hist_cbdata_t cbdata = { 0 };
	int ret;
	int c;

	cbdata.first = B_TRUE;
	/* check options */
	while ((c = getopt(argc, argv, "li")) != -1) {
	switch (c) {
	case 'l':
	cbdata.longfmt = B_TRUE;
	break;
	case 'i':
	cbdata.internal = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}
	argc -= optind;
	argv += optind;

	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, get_history_one, &cbdata);

	if (argc == 0 && cbdata.first == B_TRUE) {
	(void) fprintf(stderr, gettext("no pools available\n"));
	return (0);
	}

	return (ret);
	}

	typedef struct ev_opts {
	int verbose;
	int scripted;
	int follow;
	int clear;
	char poolname[ZFS_MAX_DATASET_NAME_LEN];
	} ev_opts_t;

	static void
	zpool_do_events_short(nvlist_t nvl, ev_opts_t opts)
	{
	char ctime_str[26], str[32], *ptr;
	int64_t *tv;
	uint_t n;

	verify(nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0);
	memset(str, ' ', 32);
	(void) ctime_r((const time_t *)&tv[0], ctime_str);
	(void) memcpy(str, ctime_str+4, 6); /* 'Jun 30' */
	(void) memcpy(str+7, ctime_str+20, 4); /* '1993' */
	(void) memcpy(str+12, ctime_str+11, 8); /* '21:49:08' */
	(void) sprintf(str+20, ".%09lld", (longlong_t)tv[1]); /* '.123456789' */
	if (opts->scripted)
	(void) printf(gettext("%s\t"), str);
	else
	(void) printf(gettext("%s "), str);

	verify(nvlist_lookup_string(nvl, FM_CLASS, &ptr) == 0);
	(void) printf(gettext("%s\n"), ptr);
	}

	static void
	zpool_do_events_nvprint(nvlist_t *nvl, int depth)
	{
	nvpair_t *nvp;

	for (nvp = nvlist_next_nvpair(nvl, NULL);
	nvp != NULL; nvp = nvlist_next_nvpair(nvl, nvp)) {

	data_type_t type = nvpair_type(nvp);
	const char *name = nvpair_name(nvp);

	boolean_t b;
	uint8_t i8;
	uint16_t i16;
	uint32_t i32;
	uint64_t i64;
	char *str;
	nvlist_t *cnv;

	printf(gettext("%*s%s = "), depth, "", name);

	switch (type) {
	case DATA_TYPE_BOOLEAN:
	printf(gettext("%s"), "1");
	break;

	case DATA_TYPE_BOOLEAN_VALUE:
	(void) nvpair_value_boolean_value(nvp, &b);
	printf(gettext("%s"), b ? "1" : "0");
	break;

	case DATA_TYPE_BYTE:
	(void) nvpair_value_byte(nvp, &i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_INT8:
	(void) nvpair_value_int8(nvp, (void *)&i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_UINT8:
	(void) nvpair_value_uint8(nvp, &i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_INT16:
	(void) nvpair_value_int16(nvp, (void *)&i16);
	printf(gettext("0x%x"), i16);
	break;

	case DATA_TYPE_UINT16:
	(void) nvpair_value_uint16(nvp, &i16);
	printf(gettext("0x%x"), i16);
	break;

	case DATA_TYPE_INT32:
	(void) nvpair_value_int32(nvp, (void *)&i32);
	printf(gettext("0x%x"), i32);
	break;

	case DATA_TYPE_UINT32:
	(void) nvpair_value_uint32(nvp, &i32);
	printf(gettext("0x%x"), i32);
	break;

	case DATA_TYPE_INT64:
	(void) nvpair_value_int64(nvp, (void *)&i64);
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	break;

	case DATA_TYPE_UINT64:
	(void) nvpair_value_uint64(nvp, &i64);
	/*
	* translate vdev state values to readable
	* strings to aide zpool events consumers
	*/
	if (strcmp(name,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE) == 0 \|\|
	strcmp(name,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE) == 0) {
	printf(gettext("\"%s\" (0x%llx)"),
	zpool_state_to_name(i64, VDEV_AUX_NONE),
	(u_longlong_t)i64);
	} else {
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	}
	break;

	case DATA_TYPE_HRTIME:
	(void) nvpair_value_hrtime(nvp, (void *)&i64);
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	break;

	case DATA_TYPE_STRING:
	(void) nvpair_value_string(nvp, &str);
	printf(gettext("\"%s\""), str ? str : "<NULL>");
	break;

	case DATA_TYPE_NVLIST:
	printf(gettext("(embedded nvlist)\n"));
	(void) nvpair_value_nvlist(nvp, &cnv);
	zpool_do_events_nvprint(cnv, depth + 8);
	printf(gettext("%*s(end %s)"), depth, "", name);
	break;

	case DATA_TYPE_NVLIST_ARRAY: {
	nvlist_t **val;
	uint_t i, nelem;

	(void) nvpair_value_nvlist_array(nvp, &val, &nelem);
	printf(gettext("(%d embedded nvlists)\n"), nelem);
	for (i = 0; i < nelem; i++) {
	printf(gettext("%*s%s[%d] = %s\n"),
	depth, "", name, i, "(embedded nvlist)");
	zpool_do_events_nvprint(val[i], depth + 8);
	printf(gettext("%*s(end %s[%i])\n"),
	depth, "", name, i);
	}
	printf(gettext("%*s(end %s)\n"), depth, "", name);
	}
	break;

	case DATA_TYPE_INT8_ARRAY: {
	int8_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int8_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT8_ARRAY: {
	uint8_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint8_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT16_ARRAY: {
	int16_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int16_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT16_ARRAY: {
	uint16_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint16_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT32_ARRAY: {
	int32_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int32_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT32_ARRAY: {
	uint32_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint32_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT64_ARRAY: {
	int64_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int64_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%llx "),
	(u_longlong_t)val[i]);

	break;
	}

	case DATA_TYPE_UINT64_ARRAY: {
	uint64_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint64_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%llx "),
	(u_longlong_t)val[i]);

	break;
	}

	case DATA_TYPE_STRING_ARRAY: {
	char **str;
	uint_t i, nelem;

	(void) nvpair_value_string_array(nvp, &str, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("\"%s\" "),
	str[i] ? str[i] : "<NULL>");

	break;
	}

	case DATA_TYPE_BOOLEAN_ARRAY:
	case DATA_TYPE_BYTE_ARRAY:
	case DATA_TYPE_DOUBLE:
	case DATA_TYPE_DONTCARE:
	case DATA_TYPE_UNKNOWN:
	printf(gettext("<unknown>"));
	break;
	}

	printf(gettext("\n"));
	}
	}

	static int
	zpool_do_events_next(ev_opts_t *opts)
	{
	nvlist_t *nvl;
	int zevent_fd, ret, dropped;
	char *pool;

	zevent_fd = open(ZFS_DEV, O_RDWR);
	VERIFY(zevent_fd >= 0);

	if (!opts->scripted)
	(void) printf(gettext("%-30s %s\n"), "TIME", "CLASS");

	while (1) {
	ret = zpool_events_next(g_zfs, &nvl, &dropped,
	(opts->follow ? ZEVENT_NONE : ZEVENT_NONBLOCK), zevent_fd);
	if (ret \|\| nvl == NULL)
	break;

	if (dropped > 0)
	(void) printf(gettext("dropped %d events\n"), dropped);

	if (strlen(opts->poolname) > 0 &&
	nvlist_lookup_string(nvl, FM_FMRI_ZFS_POOL, &pool) == 0 &&
	strcmp(opts->poolname, pool) != 0)
	continue;

	zpool_do_events_short(nvl, opts);

	if (opts->verbose) {
	zpool_do_events_nvprint(nvl, 8);
	printf(gettext("\n"));
	}
	(void) fflush(stdout);

	nvlist_free(nvl);
	}

	VERIFY(0 == close(zevent_fd));

	return (ret);
	}

	static int
	zpool_do_events_clear(void)
	{
	int count, ret;

	ret = zpool_events_clear(g_zfs, &count);
	if (!ret)
	(void) printf(gettext("cleared %d events\n"), count);

	return (ret);
	}

	/*
	* zpool events [-vHf [pool] \| -c]
	*
	* Displays events logs by ZFS.
	*/
	int
	zpool_do_events(int argc, char **argv)
	{
	ev_opts_t opts = { 0 };
	int ret;
	int c;

	/* check options */
	while ((c = getopt(argc, argv, "vHfc")) != -1) {
	switch (c) {
	case 'v':
	opts.verbose = 1;
	break;
	case 'H':
	opts.scripted = 1;
	break;
	case 'f':
	opts.follow = 1;
	break;
	case 'c':
	opts.clear = 1;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}
	argc -= optind;
	argv += optind;

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	} else if (argc == 1) {
	(void) strlcpy(opts.poolname, argv[0], sizeof (opts.poolname));
	if (!zfs_name_valid(opts.poolname, ZFS_TYPE_POOL)) {
	(void) fprintf(stderr,
	gettext("invalid pool name '%s'\n"), opts.poolname);
	usage(B_FALSE);
	}
	}

	if ((argc == 1 \|\| opts.verbose \|\| opts.scripted \|\| opts.follow) &&
	opts.clear) {
	(void) fprintf(stderr,
	gettext("invalid options combined with -c\n"));
	usage(B_FALSE);
	}

	if (opts.clear)
	ret = zpool_do_events_clear();
	else
	ret = zpool_do_events_next(&opts);

	return (ret);
	}

	static int
	get_callback_vdev(zpool_handle_t zhp, char vdevname, void *data)
	{
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char value[ZFS_MAXPROPLEN];
	zprop_source_t srctype;

	for (zprop_list_t *pl = cbp->cb_proplist; pl != NULL;
	pl = pl->pl_next) {
	char *prop_name;
	/*
	* If the first property is pool name, it is a special
	* placeholder that we can skip. This will also skip
	* over the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZPOOL_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop == ZPROP_INVAL) {
	prop_name = pl->pl_user_prop;
	} else {
	prop_name = (char *)vdev_prop_to_name(pl->pl_prop);
	}
	if (zpool_get_vdev_prop(zhp, vdevname, pl->pl_prop,
	prop_name, value, sizeof (value), &srctype,
	cbp->cb_literal) == 0) {
	zprop_print_one_property(vdevname, cbp, prop_name,
	value, srctype, NULL, NULL);
	}
	}

	return (0);
	}

	static int
	get_callback_vdev_width_cb(void zhp_data, nvlist_t nv, void *data)
	{
	zpool_handle_t *zhp = zhp_data;
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char *vdevname = zpool_vdev_name(g_zfs, zhp, nv,
	cbp->cb_vdevs.cb_name_flags);
	int ret;

	/* Adjust the column widths for the vdev properties */
	ret = vdev_expand_proplist(zhp, vdevname, &cbp->cb_proplist);

	return (ret);
	}

	static int
	get_callback_vdev_cb(void zhp_data, nvlist_t nv, void *data)
	{
	zpool_handle_t *zhp = zhp_data;
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char *vdevname = zpool_vdev_name(g_zfs, zhp, nv,
	cbp->cb_vdevs.cb_name_flags);
	int ret;

	/* Display the properties */
	ret = get_callback_vdev(zhp, vdevname, data);

	return (ret);
	}

	static int
	get_callback(zpool_handle_t zhp, void data)
	{
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char value[MAXNAMELEN];
	zprop_source_t srctype;
	zprop_list_t *pl;
	int vid;

	if (cbp->cb_type == ZFS_TYPE_VDEV) {
	if (strcmp(cbp->cb_vdevs.cb_names[0], "all-vdevs") == 0) {
	for_each_vdev(zhp, get_callback_vdev_width_cb, data);
	for_each_vdev(zhp, get_callback_vdev_cb, data);
	} else {
	/* Adjust column widths for vdev properties */
	for (vid = 0; vid < cbp->cb_vdevs.cb_names_count;
	vid++) {
	vdev_expand_proplist(zhp,
	cbp->cb_vdevs.cb_names[vid],
	&cbp->cb_proplist);
	}
	/* Display the properties */
	for (vid = 0; vid < cbp->cb_vdevs.cb_names_count;
	vid++) {
	get_callback_vdev(zhp,
	cbp->cb_vdevs.cb_names[vid], data);
	}
	}
	} else {
	assert(cbp->cb_type == ZFS_TYPE_POOL);
	for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) {
	/*
	* Skip the special fake placeholder. This will also
	* skip over the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZPOOL_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop == ZPROP_INVAL &&
	(zpool_prop_feature(pl->pl_user_prop) \|\|
	zpool_prop_unsupported(pl->pl_user_prop))) {
	srctype = ZPROP_SRC_LOCAL;

	if (zpool_prop_get_feature(zhp,
	pl->pl_user_prop, value,
	sizeof (value)) == 0) {
	zprop_print_one_property(
	zpool_get_name(zhp), cbp,
	pl->pl_user_prop, value, srctype,
	NULL, NULL);
	}
	} else {
	if (zpool_get_prop(zhp, pl->pl_prop, value,
	sizeof (value), &srctype,
	cbp->cb_literal) != 0)
	continue;

	zprop_print_one_property(zpool_get_name(zhp),
	cbp, zpool_prop_to_name(pl->pl_prop),
	value, srctype, NULL, NULL);
	}
	}
	}

	return (0);
	}

	/*
	* zpool get [-Hp] [-o "all" \| field[,...]] <"all" \| property[,...]> <pool> ...
	*
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -o List of columns to display. Defaults to
	* "name,property,value,source".
	* -p Display values in parsable (exact) format.
	*
	* Get properties of pools in the system. Output space statistics
	* for each one as well as other attributes.
	*/
	int
	zpool_do_get(int argc, char **argv)
	{
	zprop_get_cbdata_t cb = { 0 };
	zprop_list_t fake_name = { 0 };
	int ret;
	int c, i;
	char *propstr = NULL;

	cb.cb_first = B_TRUE;

	/*
	* Set up default columns and sources.
	*/
	cb.cb_sources = ZPROP_SRC_ALL;
	cb.cb_columns[0] = GET_COL_NAME;
	cb.cb_columns[1] = GET_COL_PROPERTY;
	cb.cb_columns[2] = GET_COL_VALUE;
	cb.cb_columns[3] = GET_COL_SOURCE;
	cb.cb_type = ZFS_TYPE_POOL;
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_TYPE_ID;
	current_prop_type = cb.cb_type;

	/* check options */
	while ((c = getopt(argc, argv, ":Hpo:")) != -1) {
	switch (c) {
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 'o':
	memset(&cb.cb_columns, 0, sizeof (cb.cb_columns));
	i = 0;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_opts[] =
	{ "name", "property", "value", "source",
	"all" };
	static const zfs_get_column_t col_cols[] =
	{ GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE,
	GET_COL_SOURCE };

	if (i == ZFS_GET_NCOLS - 1) {
	(void) fprintf(stderr, gettext("too "
	"many fields given to -o "
	"option\n"));
	usage(B_FALSE);
	}

	for (c = 0; c < ARRAY_SIZE(col_opts); ++c)
	if (strcmp(tok, col_opts[c]) == 0)
	goto found;

	(void) fprintf(stderr,
	gettext("invalid column name '%s'\n"), tok);
	usage(B_FALSE);

	found:
	if (c >= 4) {
	if (i > 0) {
	(void) fprintf(stderr,
	gettext("\"all\" conflicts "
	"with specific fields "
	"given to -o option\n"));
	usage(B_FALSE);
	}

	memcpy(cb.cb_columns, col_cols,
	sizeof (col_cols));
	i = ZFS_GET_NCOLS - 1;
	} else
	cb.cb_columns[i++] = col_cols[c];
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property "
	"argument\n"));
	usage(B_FALSE);
	}

	/* Properties list is needed later by zprop_get_list() */
	propstr = argv[0];

	argc--;
	argv++;

	if (argc == 0) {
	/* No args, so just print the defaults. */
	} else if (are_all_pools(argc, argv)) {
	/* All the args are pool names */
	} else if (are_all_pools(1, argv)) {
	/* The first arg is a pool name */
	if ((argc == 2 && strcmp(argv[1], "all-vdevs") == 0) \|\|
	are_vdevs_in_pool(argc - 1, argv + 1, argv[0],
	&cb.cb_vdevs)) {
	/* ... and the rest are vdev names */
	cb.cb_vdevs.cb_names = argv + 1;
	cb.cb_vdevs.cb_names_count = argc - 1;
	cb.cb_type = ZFS_TYPE_VDEV;
	argc = 1; /* One pool to process */
	} else {
	fprintf(stderr, gettext("Expected a list of vdevs in"
	" \"%s\", but got:\n"), argv[0]);
	error_list_unresolved_vdevs(argc - 1, argv + 1,
	argv[0], &cb.cb_vdevs);
	fprintf(stderr, "\n");
	usage(B_FALSE);
	return (1);
	}
	} else {
	/*
	* The first arg isn't a pool name,
	*/
	fprintf(stderr, gettext("missing pool name.\n"));
	fprintf(stderr, "\n");
	usage(B_FALSE);
	return (1);
	}

	if (zprop_get_list(g_zfs, propstr, &cb.cb_proplist,
	cb.cb_type) != 0) {
	/* Use correct list of valid properties (pool or vdev) */
	current_prop_type = cb.cb_type;
	usage(B_FALSE);
	}

	if (cb.cb_proplist != NULL) {
	fake_name.pl_prop = ZPOOL_PROP_NAME;
	fake_name.pl_width = strlen(gettext("NAME"));
	fake_name.pl_next = cb.cb_proplist;
	cb.cb_proplist = &fake_name;
	}

	ret = for_each_pool(argc, argv, B_TRUE, &cb.cb_proplist, cb.cb_type,
	cb.cb_literal, get_callback, &cb);

	if (cb.cb_proplist == &fake_name)
	zprop_free_list(fake_name.pl_next);
	else
	zprop_free_list(cb.cb_proplist);

	return (ret);
	}

	typedef struct set_cbdata {
	char *cb_propname;
	char *cb_value;
	zfs_type_t cb_type;
	vdev_cbdata_t cb_vdevs;
	boolean_t cb_any_successful;
	} set_cbdata_t;

	static int
	set_pool_callback(zpool_handle_t zhp, set_cbdata_t cb)
	{
	int error;

	/* Check if we have out-of-bounds features */
	if (strcmp(cb->cb_propname, ZPOOL_CONFIG_COMPATIBILITY) == 0) {
	boolean_t features[SPA_FEATURES];
	if (zpool_do_load_compat(cb->cb_value, features) !=
	ZPOOL_COMPATIBILITY_OK)
	return (-1);

	nvlist_t *enabled = zpool_get_features(zhp);
	spa_feature_t i;
	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fguid = spa_feature_table[i].fi_guid;
	if (nvlist_exists(enabled, fguid) && !features[i])
	break;
	}
	if (i < SPA_FEATURES)
	(void) fprintf(stderr, gettext("Warning: one or "
	"more features already enabled on pool '%s'\n"
	"are not present in this compatibility set.\n"),
	zpool_get_name(zhp));
	}

	/* if we're setting a feature, check it's in compatibility set */
	if (zpool_prop_feature(cb->cb_propname) &&
	strcmp(cb->cb_value, ZFS_FEATURE_ENABLED) == 0) {
	char *fname = strchr(cb->cb_propname, '@') + 1;
	spa_feature_t f;

	if (zfeature_lookup_name(fname, &f) == 0) {
	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY,
	compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	boolean_t features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, features) !=
	ZPOOL_COMPATIBILITY_OK) {
	(void) fprintf(stderr, gettext("Error: "
	"cannot enable feature '%s' on pool '%s'\n"
	"because the pool's 'compatibility' "
	"property cannot be parsed.\n"),
	fname, zpool_get_name(zhp));
	return (-1);
	}

	if (!features[f]) {
	(void) fprintf(stderr, gettext("Error: "
	"cannot enable feature '%s' on pool '%s'\n"
	"as it is not specified in this pool's "
	"current compatibility set.\n"
	"Consider setting 'compatibility' to a "
	"less restrictive set, or to 'off'.\n"),
	fname, zpool_get_name(zhp));
	return (-1);
	}
	}
	}

	error = zpool_set_prop(zhp, cb->cb_propname, cb->cb_value);

	return (error);
	}

	static int
	set_callback(zpool_handle_t zhp, void data)
	{
	int error;
	set_cbdata_t cb = (set_cbdata_t )data;

	if (cb->cb_type == ZFS_TYPE_VDEV) {
	error = zpool_set_vdev_prop(zhp, *cb->cb_vdevs.cb_names,
	cb->cb_propname, cb->cb_value);
	} else {
	assert(cb->cb_type == ZFS_TYPE_POOL);
	error = set_pool_callback(zhp, cb);
	}

	cb->cb_any_successful = !error;
	return (error);
	}

	int
	zpool_do_set(int argc, char **argv)
	{
	set_cbdata_t cb = { 0 };
	int error;

	current_prop_type = ZFS_TYPE_POOL;
	if (argc > 1 && argv[1][0] == '-') {
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	argv[1][1]);
	usage(B_FALSE);
	}

	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing property=value "
	"argument\n"));
	usage(B_FALSE);
	}

	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 4) {
	(void) fprintf(stderr, gettext("too many pool names\n"));
	usage(B_FALSE);
	}

	cb.cb_propname = argv[1];
	cb.cb_type = ZFS_TYPE_POOL;
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_TYPE_ID;
	cb.cb_value = strchr(cb.cb_propname, '=');
	if (cb.cb_value == NULL) {
	(void) fprintf(stderr, gettext("missing value in "
	"property=value argument\n"));
	usage(B_FALSE);
	}

	*(cb.cb_value) = '\0';
	cb.cb_value++;
	argc -= 2;
	argv += 2;

	if (are_vdevs_in_pool(argc, argv, NULL, &cb.cb_vdevs)) {
	/* Argument is a vdev */
	cb.cb_vdevs.cb_names = argv;
	cb.cb_vdevs.cb_names_count = 1;
	cb.cb_type = ZFS_TYPE_VDEV;
	argc = 0; /* No pools to process */
	} else if (are_all_pools(1, argv)) {
	/* The first arg is a pool name */
	if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0],
	&cb.cb_vdevs)) {
	/* 2nd argument is a vdev */
	cb.cb_vdevs.cb_names = argv + 1;
	cb.cb_vdevs.cb_names_count = 1;
	cb.cb_type = ZFS_TYPE_VDEV;
	argc = 1; /* One pool to process */
	} else if (argc > 1) {
	(void) fprintf(stderr,
	gettext("too many pool names\n"));
	usage(B_FALSE);
	}
	}

	error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, set_callback, &cb);

	return (error);
	}

	/* Add up the total number of bytes left to initialize/trim across all vdevs */
	static uint64_t
	vdev_activity_remaining(nvlist_t *nv, zpool_wait_activity_t activity)
	{
	uint64_t bytes_remaining;
	nvlist_t **child;
	uint_t c, children;
	vdev_stat_t *vs;

	assert(activity == ZPOOL_WAIT_INITIALIZE \|\|
	activity == ZPOOL_WAIT_TRIM);

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	if (activity == ZPOOL_WAIT_INITIALIZE &&
	vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE)
	bytes_remaining = vs->vs_initialize_bytes_est -
	vs->vs_initialize_bytes_done;
	else if (activity == ZPOOL_WAIT_TRIM &&
	vs->vs_trim_state == VDEV_TRIM_ACTIVE)
	bytes_remaining = vs->vs_trim_bytes_est -
	vs->vs_trim_bytes_done;
	else
	bytes_remaining = 0;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (c = 0; c < children; c++)
	bytes_remaining += vdev_activity_remaining(child[c], activity);

	return (bytes_remaining);
	}

	/* Add up the total number of bytes left to rebuild across top-level vdevs */
	static uint64_t
	vdev_activity_top_remaining(nvlist_t *nv)
	{
	uint64_t bytes_remaining = 0;
	nvlist_t **child;
	uint_t children;
	int error;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	error = nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i);
	if (error == 0) {
	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	bytes_remaining += (vrs->vrs_bytes_est -
	vrs->vrs_bytes_rebuilt);
	}
	}
	}

	return (bytes_remaining);
	}

	/* Whether any vdevs are 'spare' or 'replacing' vdevs */
	static boolean_t
	vdev_any_spare_replacing(nvlist_t *nv)
	{
	nvlist_t **child;
	uint_t c, children;
	char *vdev_type;

	(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &vdev_type);

	if (strcmp(vdev_type, VDEV_TYPE_REPLACING) == 0 \|\|
	strcmp(vdev_type, VDEV_TYPE_SPARE) == 0 \|\|
	strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0) {
	return (B_TRUE);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (c = 0; c < children; c++) {
	if (vdev_any_spare_replacing(child[c]))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	typedef struct wait_data {
	char *wd_poolname;
	boolean_t wd_scripted;
	boolean_t wd_exact;
	boolean_t wd_headers_once;
	boolean_t wd_should_exit;
	/* Which activities to wait for */
	boolean_t wd_enabled[ZPOOL_WAIT_NUM_ACTIVITIES];
	float wd_interval;
	pthread_cond_t wd_cv;
	pthread_mutex_t wd_mutex;
	} wait_data_t;

	/*
	* Print to stdout a single line, containing one column for each activity that
	* we are waiting for specifying how many bytes of work are left for that
	* activity.
	*/
	static void
	print_wait_status_row(wait_data_t wd, zpool_handle_t zhp, int row)
	{
	nvlist_t config, nvroot;
	uint_t c;
	int i;
	pool_checkpoint_stat_t *pcs = NULL;
	pool_scan_stat_t *pss = NULL;
	pool_removal_stat_t *prs = NULL;
	const char *const headers[] = {"DISCARD", "FREE", "INITIALIZE",
	"REPLACE", "REMOVE", "RESILVER", "SCRUB", "TRIM"};
	int col_widths[ZPOOL_WAIT_NUM_ACTIVITIES];

	/* Calculate the width of each column */
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	/*
	* Make sure we have enough space in the col for pretty-printed
	* numbers and for the column header, and then leave a couple
	* spaces between cols for readability.
	*/
	col_widths[i] = MAX(strlen(headers[i]), 6) + 2;
	}

	/* Print header if appropriate */
	int term_height = terminal_height();
	boolean_t reprint_header = (!wd->wd_headers_once && term_height > 0 &&
	row % (term_height-1) == 0);
	if (!wd->wd_scripted && (row == 0 \|\| reprint_header)) {
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	if (wd->wd_enabled[i])
	(void) printf("%*s", col_widths[i], headers[i]);
	}
	(void) fputc('\n', stdout);
	}

	/* Bytes of work remaining in each activity */
	int64_t bytes_rem[ZPOOL_WAIT_NUM_ACTIVITIES] = {0};

	bytes_rem[ZPOOL_WAIT_FREE] =
	zpool_get_prop_int(zhp, ZPOOL_PROP_FREEING, NULL);

	config = zpool_get_config(zhp, NULL);
	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	if (pcs != NULL && pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
	bytes_rem[ZPOOL_WAIT_CKPT_DISCARD] = pcs->pcs_space;

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
	if (prs != NULL && prs->prs_state == DSS_SCANNING)
	bytes_rem[ZPOOL_WAIT_REMOVE] = prs->prs_to_copy -
	prs->prs_copied;

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&pss, &c);
	if (pss != NULL && pss->pss_state == DSS_SCANNING &&
	pss->pss_pass_scrub_pause == 0) {
	int64_t rem = pss->pss_to_examine - pss->pss_issued;
	if (pss->pss_func == POOL_SCAN_SCRUB)
	bytes_rem[ZPOOL_WAIT_SCRUB] = rem;
	else
	bytes_rem[ZPOOL_WAIT_RESILVER] = rem;
	} else if (check_rebuilding(nvroot, NULL)) {
	bytes_rem[ZPOOL_WAIT_RESILVER] =
	vdev_activity_top_remaining(nvroot);
	}

	bytes_rem[ZPOOL_WAIT_INITIALIZE] =
	vdev_activity_remaining(nvroot, ZPOOL_WAIT_INITIALIZE);
	bytes_rem[ZPOOL_WAIT_TRIM] =
	vdev_activity_remaining(nvroot, ZPOOL_WAIT_TRIM);

	/*
	* A replace finishes after resilvering finishes, so the amount of work
	* left for a replace is the same as for resilvering.
	*
	* It isn't quite correct to say that if we have any 'spare' or
	* 'replacing' vdevs and a resilver is happening, then a replace is in
	* progress, like we do here. When a hot spare is used, the faulted vdev
	* is not removed after the hot spare is resilvered, so parent 'spare'
	* vdev is not removed either. So we could have a 'spare' vdev, but be
	* resilvering for a different reason. However, we use it as a heuristic
	* because we don't have access to the DTLs, which could tell us whether
	* or not we have really finished resilvering a hot spare.
	*/
	if (vdev_any_spare_replacing(nvroot))
	bytes_rem[ZPOOL_WAIT_REPLACE] = bytes_rem[ZPOOL_WAIT_RESILVER];

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	char buf[64];
	if (!wd->wd_enabled[i])
	continue;

	if (wd->wd_exact)
	(void) snprintf(buf, sizeof (buf), "%" PRIi64,
	bytes_rem[i]);
	else
	zfs_nicenum(bytes_rem[i], buf, sizeof (buf));

	if (wd->wd_scripted)
	(void) printf(i == 0 ? "%s" : "\t%s", buf);
	else
	(void) printf(" %*s", col_widths[i] - 1, buf);
	}
	(void) printf("\n");
	(void) fflush(stdout);
	}

	static void *
	wait_status_thread(void *arg)
	{
	wait_data_t wd = (wait_data_t )arg;
	zpool_handle_t *zhp;

	if ((zhp = zpool_open(g_zfs, wd->wd_poolname)) == NULL)
	return (void *)(1);

	for (int row = 0; ; row++) {
	boolean_t missing;
	struct timespec timeout;
	int ret = 0;
	(void) clock_gettime(CLOCK_REALTIME, &timeout);

	if (zpool_refresh_stats(zhp, &missing) != 0 \|\| missing \|\|
	zpool_props_refresh(zhp) != 0) {
	zpool_close(zhp);
	return (void *)(uintptr_t)(missing ? 0 : 1);
	}

	print_wait_status_row(wd, zhp, row);

	timeout.tv_sec += floor(wd->wd_interval);
	long nanos = timeout.tv_nsec +
	(wd->wd_interval - floor(wd->wd_interval)) * NANOSEC;
	if (nanos >= NANOSEC) {
	timeout.tv_sec++;
	timeout.tv_nsec = nanos - NANOSEC;
	} else {
	timeout.tv_nsec = nanos;
	}
	pthread_mutex_lock(&wd->wd_mutex);
	if (!wd->wd_should_exit)
	ret = pthread_cond_timedwait(&wd->wd_cv, &wd->wd_mutex,
	&timeout);
	pthread_mutex_unlock(&wd->wd_mutex);
	if (ret == 0) {
	break; /* signaled by main thread */
	} else if (ret != ETIMEDOUT) {
	(void) fprintf(stderr, gettext("pthread_cond_timedwait "
	"failed: %s\n"), strerror(ret));
	zpool_close(zhp);
	return (void *)(uintptr_t)(1);
	}
	}

	zpool_close(zhp);
	return (void *)(0);
	}

	int
	zpool_do_wait(int argc, char **argv)
	{
	boolean_t verbose = B_FALSE;
	int c, i;
	unsigned long count;
	pthread_t status_thr;
	int error = 0;
	zpool_handle_t *zhp;

	wait_data_t wd;
	wd.wd_scripted = B_FALSE;
	wd.wd_exact = B_FALSE;
	wd.wd_headers_once = B_FALSE;
	wd.wd_should_exit = B_FALSE;

	pthread_mutex_init(&wd.wd_mutex, NULL);
	pthread_cond_init(&wd.wd_cv, NULL);

	/* By default, wait for all types of activity. */
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++)
	wd.wd_enabled[i] = B_TRUE;

	while ((c = getopt(argc, argv, "HpT:t:")) != -1) {
	switch (c) {
	case 'H':
	wd.wd_scripted = B_TRUE;
	break;
	case 'n':
	wd.wd_headers_once = B_TRUE;
	break;
	case 'p':
	wd.wd_exact = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 't':
	/* Reset activities array */
	memset(&wd.wd_enabled, 0, sizeof (wd.wd_enabled));

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_opts[] = {
	"discard", "free", "initialize", "replace",
	"remove", "resilver", "scrub", "trim" };

	for (i = 0; i < ARRAY_SIZE(col_opts); ++i)
	if (strcmp(tok, col_opts[i]) == 0) {
	wd.wd_enabled[i] = B_TRUE;
	goto found;
	}

	(void) fprintf(stderr,
	gettext("invalid activity '%s'\n"), tok);
	usage(B_FALSE);
	found:;
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &wd.wd_interval, &count);
	if (count != 0) {
	/* This subcmd only accepts an interval, not a count */
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (wd.wd_interval != 0)
	verbose = B_TRUE;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing 'pool' argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	wd.wd_poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, wd.wd_poolname)) == NULL)
	return (1);

	if (verbose) {
	/*
	* We use a separate thread for printing status updates because
	* the main thread will call lzc_wait(), which blocks as long
	* as an activity is in progress, which can be a long time.
	*/
	if (pthread_create(&status_thr, NULL, wait_status_thread, &wd)
	!= 0) {
	(void) fprintf(stderr, gettext("failed to create status"
	"thread: %s\n"), strerror(errno));
	zpool_close(zhp);
	return (1);
	}
	}

	/*
	* Loop over all activities that we are supposed to wait for until none
	* of them are in progress. Note that this means we can end up waiting
	* for more activities to complete than just those that were in progress
	* when we began waiting; if an activity we are interested in begins
	* while we are waiting for another activity, we will wait for both to
	* complete before exiting.
	*/
	for (;;) {
	boolean_t missing = B_FALSE;
	boolean_t any_waited = B_FALSE;

	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	boolean_t waited;

	if (!wd.wd_enabled[i])
	continue;

	error = zpool_wait_status(zhp, i, &missing, &waited);
	if (error != 0 \|\| missing)
	break;

	any_waited = (any_waited \|\| waited);
	}

	if (error != 0 \|\| missing \|\| !any_waited)
	break;
	}

	zpool_close(zhp);

	if (verbose) {
	uintptr_t status;
	pthread_mutex_lock(&wd.wd_mutex);
	wd.wd_should_exit = B_TRUE;
	pthread_cond_signal(&wd.wd_cv);
	pthread_mutex_unlock(&wd.wd_mutex);
	(void) pthread_join(status_thr, (void *)&status);
	if (status != 0)
	error = status;
	}

	pthread_mutex_destroy(&wd.wd_mutex);
	pthread_cond_destroy(&wd.wd_cv);
	return (error);
	}

	static int
	find_command_idx(const char command, int idx)
	{
	for (int i = 0; i < NCOMMAND; ++i) {
	if (command_table[i].name == NULL)
	continue;

	if (strcmp(command, command_table[i].name) == 0) {
	*idx = i;
	return (0);
	}
	}
	return (1);
	}

	/*
	* Display version message
	*/
	static int
	zpool_do_version(int argc, char **argv)
	{
	(void) argc, (void) argv;
	return (zfs_version_print() != 0);
	}

	/*
	* Do zpool_load_compat() and print error message on failure
	*/
	static zpool_compat_status_t
	zpool_do_load_compat(const char compat, boolean_t list)
	{
	char report[1024];

	zpool_compat_status_t ret;

	ret = zpool_load_compat(compat, list, report, 1024);
	switch (ret) {

	case ZPOOL_COMPATIBILITY_OK:
	break;

	case ZPOOL_COMPATIBILITY_NOFILES:
	case ZPOOL_COMPATIBILITY_BADFILE:
	case ZPOOL_COMPATIBILITY_BADTOKEN:
	(void) fprintf(stderr, "Error: %s\n", report);
	break;

	case ZPOOL_COMPATIBILITY_WARNTOKEN:
	(void) fprintf(stderr, "Warning: %s\n", report);
	ret = ZPOOL_COMPATIBILITY_OK;
	break;
	}
	return (ret);
	}

	int
	main(int argc, char **argv)
	{
	int ret = 0;
	int i = 0;
	char *cmdname;
	char **newargv;

	(void) setlocale(LC_ALL, "");
	(void) setlocale(LC_NUMERIC, "C");
	(void) textdomain(TEXT_DOMAIN);
	srand(time(NULL));

	opterr = 0;

	/*
	* Make sure the user has specified some command.
	*/
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing command\n"));
	usage(B_FALSE);
	}

	cmdname = argv[1];

	/*
	* Special case '-?'
	*/
	if ((strcmp(cmdname, "-?") == 0) \|\| strcmp(cmdname, "--help") == 0)
	usage(B_TRUE);

	/*
	* Special case '-V\|--version'
	*/
	if ((strcmp(cmdname, "-V") == 0) \|\| (strcmp(cmdname, "--version") == 0))
	return (zpool_do_version(argc, argv));

	if ((g_zfs = libzfs_init()) == NULL) {
	(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
	return (1);
	}

	libzfs_print_on_error(g_zfs, B_TRUE);

	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));

	/*
	* Many commands modify input strings for string parsing reasons.
	* We create a copy to protect the original argv.
	*/
	newargv = safe_malloc((argc + 1) * sizeof (newargv[0]));
	for (i = 0; i < argc; i++)
	newargv[i] = strdup(argv[i]);
	newargv[argc] = NULL;

	/*
	* Run the appropriate command.
	*/
	if (find_command_idx(cmdname, &i) == 0) {
	current_command = &command_table[i];
	ret = command_table[i].func(argc - 1, newargv + 1);
	} else if (strchr(cmdname, '=')) {
	verify(find_command_idx("set", &i) == 0);
	current_command = &command_table[i];
	ret = command_table[i].func(argc, newargv);
	} else if (strcmp(cmdname, "freeze") == 0 && argc == 3) {
	/*
	* 'freeze' is a vile debugging abomination, so we treat
	* it as such.
	*/
	zfs_cmd_t zc = {"\0"};

	(void) strlcpy(zc.zc_name, argv[2], sizeof (zc.zc_name));
	ret = zfs_ioctl(g_zfs, ZFS_IOC_POOL_FREEZE, &zc);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to freeze pool: %d\n"), errno);
	ret = 1;
	}

	log_history = 0;
	} else {
	(void) fprintf(stderr, gettext("unrecognized "
	"command '%s'\n"), cmdname);
	usage(B_FALSE);
	ret = 1;
	}

	for (i = 0; i < argc; i++)
	free(newargv[i]);
	free(newargv);

	if (ret == 0 && log_history)
	(void) zpool_log_history(g_zfs, history_str);

	libzfs_fini(g_zfs);

	/*
	* The 'ZFS_ABORT' environment variable causes us to dump core on exit
	* for the purposes of running ::findleaks.
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	return (ret);
	}
	diff --git a/sys/contrib/openzfs/cmd/zpool_influxdb/zpool_influxdb.c b/sys/contrib/openzfs/cmd/zpool_influxdb/zpool_influxdb.c
	index 251d588d832a..73a9bedb8164 100644
	--- a/sys/contrib/openzfs/cmd/zpool_influxdb/zpool_influxdb.c
	+++ b/sys/contrib/openzfs/cmd/zpool_influxdb/zpool_influxdb.c
	@@ -1,844 +1,846 @@
	/*
	* Gather top-level ZFS pool and resilver/scan statistics and print using
	* influxdb line protocol
	* usage: [options] [pool_name]
	* where options are:
	* --execd, -e run in telegraf execd input plugin mode, [CR] on
	* stdin causes a sample to be printed and wait for
	* the next [CR]
	* --no-histograms, -n don't print histogram data (reduces cardinality
	* if you don't care about histograms)
	* --sum-histogram-buckets, -s sum histogram bucket values
	*
	* To integrate into telegraf use one of:
	* 1. the `inputs.execd` plugin with the `--execd` option
	* 2. the `inputs.exec` plugin to simply run with no options
	*
	* NOTE: libzfs is an unstable interface. YMMV.
	*
	* The design goals of this software include:
	* + be as lightweight as possible
	* + reduce the number of external dependencies as far as possible, hence
	* there is no dependency on a client library for managing the metric
	* collection -- info is printed, KISS
	* + broken pools or kernel bugs can cause this process to hang in an
	* unkillable state. For this reason, it is best to keep the damage limited
	* to a small process like zpool_influxdb rather than a larger collector.
	*
	* Copyright 2018-2020 Richard Elling
	*
	* This software is dual-licensed MIT and CDDL.
	*
	* The MIT License (MIT)
	*
	* Permission is hereby granted, free of charge, to any person obtaining a copy
	* of this software and associated documentation files (the "Software"), to deal
	* in the Software without restriction, including without limitation the rights
	* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
	* copies of the Software, and to permit persons to whom the Software is
	* furnished to do so, subject to the following conditions:
	*
	* The above copyright notice and this permission notice shall be included in
	* all copies or substantial portions of the Software.
	*
	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
	* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
	* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	* SOFTWARE.
	*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License Version 1.0 (CDDL-1.0).
	* You can obtain a copy of the license from the top-level file
	* "OPENSOLARIS.LICENSE" or at <http://opensource.org/licenses/CDDL-1.0>.
	* You may not use this file except in compliance with the license.
	*
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* CDDL HEADER END
	*/
	#include <string.h>
	#include <getopt.h>
	#include <stdio.h>
	#include <stdint.h>
	#include <inttypes.h>
	#include <libzfs.h>

	#define POOL_MEASUREMENT "zpool_stats"
	#define SCAN_MEASUREMENT "zpool_scan_stats"
	#define VDEV_MEASUREMENT "zpool_vdev_stats"
	#define POOL_LATENCY_MEASUREMENT "zpool_latency"
	#define POOL_QUEUE_MEASUREMENT "zpool_vdev_queue"
	#define MIN_LAT_INDEX 10 /* minimum latency index 10 = 1024ns */
	#define POOL_IO_SIZE_MEASUREMENT "zpool_io_size"
	#define MIN_SIZE_INDEX 9 /* minimum size index 9 = 512 bytes */

	/* global options */
	int execd_mode = 0;
	int no_histograms = 0;
	int sum_histogram_buckets = 0;
	char metric_data_type = 'u';
	uint64_t metric_value_mask = UINT64_MAX;
	uint64_t timestamp = 0;
	int complained_about_sync = 0;
	const char *tags = "";

	typedef int (stat_printer_f)(nvlist_t , const char , const char );

	/*
	* influxdb line protocol rules for escaping are important because the
	* zpool name can include characters that need to be escaped
	*
	* caller is responsible for freeing result
	*/
	static char *
	escape_string(const char *s)
	{
	const char *c;
	char *d;
	char t = (char )malloc(ZFS_MAX_DATASET_NAME_LEN * 2);
	if (t == NULL) {
	fprintf(stderr, "error: cannot allocate memory\n");
	exit(1);
	}

	for (c = s, d = t; *c != '\0'; c++, d++) {
	switch (*c) {
	case ' ':
	case ',':
	case '=':
	case '\\':
	*d++ = '\\';
	zfs_fallthrough;
	default:
	d = c;
	}
	}
	*d = '\0';
	return (t);
	}

	/*
	* print key=value where value is a uint64_t
	*/
	static void
	print_kv(const char *key, uint64_t value)
	{
	printf("%s=%llu%c", key,
	(u_longlong_t)value & metric_value_mask, metric_data_type);
	}

	/*
	* print_scan_status() prints the details as often seen in the "zpool status"
	* output. However, unlike the zpool command, which is intended for humans,
	* this output is suitable for long-term tracking in influxdb.
	* TODO: update to include issued scan data
	*/
	static int
	print_scan_status(nvlist_t nvroot, const char pool_name)
	{
	uint_t c;
	int64_t elapsed;
	uint64_t examined, pass_exam, paused_time, paused_ts, rate;
	uint64_t remaining_time;
	pool_scan_stat_t *ps = NULL;
	double pct_done;
	const char *const state[DSS_NUM_STATES] = {
	"none", "scanning", "finished", "canceled"};
	const char *func;

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_SCAN_STATS,
	(uint64_t **)&ps, &c);

	/*
	* ignore if there are no stats
	*/
	if (ps == NULL)
	return (0);

	/*
	* return error if state is bogus
	*/
	if (ps->pss_state >= DSS_NUM_STATES \|\|
	ps->pss_func >= POOL_SCAN_FUNCS) {
	if (complained_about_sync % 1000 == 0) {
	fprintf(stderr, "error: cannot decode scan stats: "
	"ZFS is out of sync with compiled zpool_influxdb");
	complained_about_sync++;
	}
	return (1);
	}

	switch (ps->pss_func) {
	case POOL_SCAN_NONE:
	func = "none_requested";
	break;
	case POOL_SCAN_SCRUB:
	func = "scrub";
	break;
	case POOL_SCAN_RESILVER:
	func = "resilver";
	break;
	#ifdef POOL_SCAN_REBUILD
	case POOL_SCAN_REBUILD:
	func = "rebuild";
	break;
	#endif
	default:
	func = "scan";
	}

	/* overall progress */
	examined = ps->pss_examined ? ps->pss_examined : 1;
	pct_done = 0.0;
	if (ps->pss_to_examine > 0)
	pct_done = 100.0 * examined / ps->pss_to_examine;

	#ifdef EZFS_SCRUB_PAUSED
	paused_ts = ps->pss_pass_scrub_pause;
	paused_time = ps->pss_pass_scrub_spent_paused;
	#else
	paused_ts = 0;
	paused_time = 0;
	#endif

	/* calculations for this pass */
	if (ps->pss_state == DSS_SCANNING) {
	elapsed = (int64_t)time(NULL) - (int64_t)ps->pss_pass_start -
	(int64_t)paused_time;
	elapsed = (elapsed > 0) ? elapsed : 1;
	pass_exam = ps->pss_pass_exam ? ps->pss_pass_exam : 1;
	rate = pass_exam / elapsed;
	rate = (rate > 0) ? rate : 1;
	remaining_time = ps->pss_to_examine - examined / rate;
	} else {
	elapsed =
	(int64_t)ps->pss_end_time - (int64_t)ps->pss_pass_start -
	(int64_t)paused_time;
	elapsed = (elapsed > 0) ? elapsed : 1;
	pass_exam = ps->pss_pass_exam ? ps->pss_pass_exam : 1;
	rate = pass_exam / elapsed;
	remaining_time = 0;
	}
	rate = rate ? rate : 1;

	/* influxdb line protocol format: "tags metrics timestamp" */
	printf("%s%s,function=%s,name=%s,state=%s ",
	SCAN_MEASUREMENT, tags, func, pool_name, state[ps->pss_state]);
	print_kv("end_ts", ps->pss_end_time);
	print_kv(",errors", ps->pss_errors);
	print_kv(",examined", examined);
	print_kv(",issued", ps->pss_issued);
	print_kv(",pass_examined", pass_exam);
	print_kv(",pass_issued", ps->pss_pass_issued);
	print_kv(",paused_ts", paused_ts);
	print_kv(",paused_t", paused_time);
	printf(",pct_done=%.2f", pct_done);
	print_kv(",processed", ps->pss_processed);
	print_kv(",rate", rate);
	print_kv(",remaining_t", remaining_time);
	print_kv(",start_ts", ps->pss_start_time);
	print_kv(",to_examine", ps->pss_to_examine);
	print_kv(",to_process", ps->pss_to_process);
	printf(" %llu\n", (u_longlong_t)timestamp);
	return (0);
	}

	/*
	* get a vdev name that corresponds to the top-level vdev names
	* printed by `zpool status`
	*/
	static char *
	get_vdev_name(nvlist_t nvroot, const char parent_name)
	{
	static char vdev_name[256];
	uint64_t vdev_id = 0;

	char vdev_type = (char )"unknown";
	- nvlist_lookup_string(nvroot, ZPOOL_CONFIG_TYPE, &vdev_type);
	+ (void) nvlist_lookup_string(nvroot, ZPOOL_CONFIG_TYPE, &vdev_type);

	if (nvlist_lookup_uint64(
	nvroot, ZPOOL_CONFIG_ID, &vdev_id) != 0)
	vdev_id = UINT64_MAX;

	if (parent_name == NULL) {
	(void) snprintf(vdev_name, sizeof (vdev_name), "%s",
	vdev_type);
	} else {
	(void) snprintf(vdev_name, sizeof (vdev_name),
	"%.220s/%s-%llu",
	parent_name, vdev_type, (u_longlong_t)vdev_id);
	}
	return (vdev_name);
	}

	/*
	* get a string suitable for an influxdb tag that describes this vdev
	*
	* By default only the vdev hierarchical name is shown, separated by '/'
	* If the vdev has an associated path, which is typical of leaf vdevs,
	* then the path is added.
	* It would be nice to have the devid instead of the path, but under
	* Linux we cannot be sure a devid will exist and we'd rather have
	* something than nothing, so we'll use path instead.
	*/
	static char *
	get_vdev_desc(nvlist_t nvroot, const char parent_name)
	{
	static char vdev_desc[2 * MAXPATHLEN];
	char vdev_value[MAXPATHLEN];
	char s, t;

	char vdev_type = (char )"unknown";
	uint64_t vdev_id = UINT64_MAX;
	char *vdev_path = NULL;
	- nvlist_lookup_string(nvroot, ZPOOL_CONFIG_TYPE, &vdev_type);
	- nvlist_lookup_uint64(nvroot, ZPOOL_CONFIG_ID, &vdev_id);
	- nvlist_lookup_string(nvroot, ZPOOL_CONFIG_PATH, &vdev_path);
	+ (void) nvlist_lookup_string(nvroot, ZPOOL_CONFIG_TYPE, &vdev_type);
	+ (void) nvlist_lookup_uint64(nvroot, ZPOOL_CONFIG_ID, &vdev_id);
	+ (void) nvlist_lookup_string(nvroot, ZPOOL_CONFIG_PATH, &vdev_path);

	if (parent_name == NULL) {
	s = escape_string(vdev_type);
	(void) snprintf(vdev_value, sizeof (vdev_value), "vdev=%s", s);
	free(s);
	} else {
	s = escape_string((char *)parent_name);
	t = escape_string(vdev_type);
	(void) snprintf(vdev_value, sizeof (vdev_value),
	"vdev=%s/%s-%llu", s, t, (u_longlong_t)vdev_id);
	free(s);
	free(t);
	}
	if (vdev_path == NULL) {
	(void) snprintf(vdev_desc, sizeof (vdev_desc), "%s",
	vdev_value);
	} else {
	s = escape_string(vdev_path);
	(void) snprintf(vdev_desc, sizeof (vdev_desc), "path=%s,%s",
	s, vdev_value);
	free(s);
	}
	return (vdev_desc);
	}

	/*
	* vdev summary stats are a combination of the data shown by
	* `zpool status` and `zpool list -v`
	*/
	static int
	print_summary_stats(nvlist_t nvroot, const char pool_name,
	const char *parent_name)
	{
	uint_t c;
	vdev_stat_t *vs;
	char *vdev_desc = NULL;
	vdev_desc = get_vdev_desc(nvroot, parent_name);
	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) != 0) {
	return (1);
	}
	printf("%s%s,name=%s,state=%s,%s ", POOL_MEASUREMENT, tags,
	pool_name, zpool_state_to_name((vdev_state_t)vs->vs_state,
	(vdev_aux_t)vs->vs_aux), vdev_desc);
	print_kv("alloc", vs->vs_alloc);
	print_kv(",free", vs->vs_space - vs->vs_alloc);
	print_kv(",size", vs->vs_space);
	print_kv(",read_bytes", vs->vs_bytes[ZIO_TYPE_READ]);
	print_kv(",read_errors", vs->vs_read_errors);
	print_kv(",read_ops", vs->vs_ops[ZIO_TYPE_READ]);
	print_kv(",write_bytes", vs->vs_bytes[ZIO_TYPE_WRITE]);
	print_kv(",write_errors", vs->vs_write_errors);
	print_kv(",write_ops", vs->vs_ops[ZIO_TYPE_WRITE]);
	print_kv(",checksum_errors", vs->vs_checksum_errors);
	print_kv(",fragmentation", vs->vs_fragmentation);
	printf(" %llu\n", (u_longlong_t)timestamp);
	return (0);
	}

	/*
	* vdev latency stats are histograms stored as nvlist arrays of uint64.
	* Latency stats include the ZIO scheduler classes plus lower-level
	* vdev latencies.
	*
	* In many cases, the top-level "root" view obscures the underlying
	* top-level vdev operations. For example, if a pool has a log, special,
	* or cache device, then each can behave very differently. It is useful
	* to see how each is responding.
	*/
	static int
	print_vdev_latency_stats(nvlist_t nvroot, const char pool_name,
	const char *parent_name)
	{
	uint_t c, end = 0;
	nvlist_t *nv_ex;
	char *vdev_desc = NULL;

	/* short_names become part of the metric name and are influxdb-ready */
	struct lat_lookup {
	const char *name;
	const char *short_name;
	uint64_t sum;
	uint64_t *array;
	};
	struct lat_lookup lat_type[] = {
	{ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO, "total_read", 0},
	{ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO, "total_write", 0},
	{ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO, "disk_read", 0},
	{ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO, "disk_write", 0},
	{ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO, "sync_read", 0},
	{ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO, "sync_write", 0},
	{ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO, "async_read", 0},
	{ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO, "async_write", 0},
	{ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO, "scrub", 0},
	#ifdef ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO
	{ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO, "trim", 0},
	#endif
	{ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO, "rebuild", 0},
	{NULL, NULL}
	};

	if (nvlist_lookup_nvlist(nvroot,
	ZPOOL_CONFIG_VDEV_STATS_EX, &nv_ex) != 0) {
	return (6);
	}

	vdev_desc = get_vdev_desc(nvroot, parent_name);

	for (int i = 0; lat_type[i].name; i++) {
	if (nvlist_lookup_uint64_array(nv_ex,
	lat_type[i].name, &lat_type[i].array, &c) != 0) {
	fprintf(stderr, "error: can't get %s\n",
	lat_type[i].name);
	return (3);
	}
	/* end count count, all of the arrays are the same size */
	end = c - 1;
	}

	for (int bucket = 0; bucket <= end; bucket++) {
	if (bucket < MIN_LAT_INDEX) {
	/* don't print, but collect the sum */
	for (int i = 0; lat_type[i].name; i++) {
	lat_type[i].sum += lat_type[i].array[bucket];
	}
	continue;
	}
	if (bucket < end) {
	printf("%s%s,le=%0.6f,name=%s,%s ",
	POOL_LATENCY_MEASUREMENT, tags,
	(float)(1ULL << bucket) * 1e-9,
	pool_name, vdev_desc);
	} else {
	printf("%s%s,le=+Inf,name=%s,%s ",
	POOL_LATENCY_MEASUREMENT, tags, pool_name,
	vdev_desc);
	}
	for (int i = 0; lat_type[i].name; i++) {
	if (bucket <= MIN_LAT_INDEX \|\| sum_histogram_buckets) {
	lat_type[i].sum += lat_type[i].array[bucket];
	} else {
	lat_type[i].sum = lat_type[i].array[bucket];
	}
	print_kv(lat_type[i].short_name, lat_type[i].sum);
	if (lat_type[i + 1].name != NULL) {
	printf(",");
	}
	}
	printf(" %llu\n", (u_longlong_t)timestamp);
	}
	return (0);
	}

	/*
	* vdev request size stats are histograms stored as nvlist arrays of uint64.
	* Request size stats include the ZIO scheduler classes plus lower-level
	* vdev sizes. Both independent (ind) and aggregated (agg) sizes are reported.
	*
	* In many cases, the top-level "root" view obscures the underlying
	* top-level vdev operations. For example, if a pool has a log, special,
	* or cache device, then each can behave very differently. It is useful
	* to see how each is responding.
	*/
	static int
	print_vdev_size_stats(nvlist_t nvroot, const char pool_name,
	const char *parent_name)
	{
	uint_t c, end = 0;
	nvlist_t *nv_ex;
	char *vdev_desc = NULL;

	/* short_names become the field name */
	struct size_lookup {
	const char *name;
	const char *short_name;
	uint64_t sum;
	uint64_t *array;
	};
	struct size_lookup size_type[] = {
	{ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO, "sync_read_ind"},
	{ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO, "sync_write_ind"},
	{ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO, "async_read_ind"},
	{ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO, "async_write_ind"},
	{ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO, "scrub_read_ind"},
	{ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO, "sync_read_agg"},
	{ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO, "sync_write_agg"},
	{ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO, "async_read_agg"},
	{ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO, "async_write_agg"},
	{ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO, "scrub_read_agg"},
	#ifdef ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO
	{ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO, "trim_write_ind"},
	{ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO, "trim_write_agg"},
	#endif
	{ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO, "rebuild_write_ind"},
	{ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO, "rebuild_write_agg"},
	{NULL, NULL}
	};

	if (nvlist_lookup_nvlist(nvroot,
	ZPOOL_CONFIG_VDEV_STATS_EX, &nv_ex) != 0) {
	return (6);
	}

	vdev_desc = get_vdev_desc(nvroot, parent_name);

	for (int i = 0; size_type[i].name; i++) {
	if (nvlist_lookup_uint64_array(nv_ex, size_type[i].name,
	&size_type[i].array, &c) != 0) {
	fprintf(stderr, "error: can't get %s\n",
	size_type[i].name);
	return (3);
	}
	/* end count count, all of the arrays are the same size */
	end = c - 1;
	}

	for (int bucket = 0; bucket <= end; bucket++) {
	if (bucket < MIN_SIZE_INDEX) {
	/* don't print, but collect the sum */
	for (int i = 0; size_type[i].name; i++) {
	size_type[i].sum += size_type[i].array[bucket];
	}
	continue;
	}

	if (bucket < end) {
	printf("%s%s,le=%llu,name=%s,%s ",
	POOL_IO_SIZE_MEASUREMENT, tags, 1ULL << bucket,
	pool_name, vdev_desc);
	} else {
	printf("%s%s,le=+Inf,name=%s,%s ",
	POOL_IO_SIZE_MEASUREMENT, tags, pool_name,
	vdev_desc);
	}
	for (int i = 0; size_type[i].name; i++) {
	if (bucket <= MIN_SIZE_INDEX \|\| sum_histogram_buckets) {
	size_type[i].sum += size_type[i].array[bucket];
	} else {
	size_type[i].sum = size_type[i].array[bucket];
	}
	print_kv(size_type[i].short_name, size_type[i].sum);
	if (size_type[i + 1].name != NULL) {
	printf(",");
	}
	}
	printf(" %llu\n", (u_longlong_t)timestamp);
	}
	return (0);
	}

	/*
	* ZIO scheduler queue stats are stored as gauges. This is unfortunate
	* because the values can change very rapidly and any point-in-time
	* value will quickly be obsoleted. It is also not easy to downsample.
	* Thus only the top-level queue stats might be beneficial... maybe.
	*/
	static int
	print_queue_stats(nvlist_t nvroot, const char pool_name,
	const char *parent_name)
	{
	nvlist_t *nv_ex;
	uint64_t value;

	/* short_names are used for the field name */
	struct queue_lookup {
	const char *name;
	const char *short_name;
	};
	struct queue_lookup queue_type[] = {
	{ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, "sync_r_active"},
	{ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, "sync_w_active"},
	{ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, "async_r_active"},
	{ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, "async_w_active"},
	{ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, "async_scrub_active"},
	{ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE, "rebuild_active"},
	{ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, "sync_r_pend"},
	{ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, "sync_w_pend"},
	{ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, "async_r_pend"},
	{ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, "async_w_pend"},
	{ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, "async_scrub_pend"},
	{ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE, "rebuild_pend"},
	{NULL, NULL}
	};

	if (nvlist_lookup_nvlist(nvroot,
	ZPOOL_CONFIG_VDEV_STATS_EX, &nv_ex) != 0) {
	return (6);
	}

	printf("%s%s,name=%s,%s ", POOL_QUEUE_MEASUREMENT, tags, pool_name,
	get_vdev_desc(nvroot, parent_name));
	for (int i = 0; queue_type[i].name; i++) {
	if (nvlist_lookup_uint64(nv_ex,
	queue_type[i].name, &value) != 0) {
	fprintf(stderr, "error: can't get %s\n",
	queue_type[i].name);
	return (3);
	}
	print_kv(queue_type[i].short_name, value);
	if (queue_type[i + 1].name != NULL) {
	printf(",");
	}
	}
	printf(" %llu\n", (u_longlong_t)timestamp);
	return (0);
	}

	/*
	* top-level vdev stats are at the pool level
	*/
	static int
	print_top_level_vdev_stats(nvlist_t nvroot, const char pool_name)
	{
	nvlist_t *nv_ex;
	uint64_t value;

	/* short_names become part of the metric name */
	struct queue_lookup {
	const char *name;
	const char *short_name;
	};
	struct queue_lookup queue_type[] = {
	{ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE, "sync_r_active_queue"},
	{ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE, "sync_w_active_queue"},
	{ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE, "async_r_active_queue"},
	{ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE, "async_w_active_queue"},
	{ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE, "async_scrub_active_queue"},
	{ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE, "rebuild_active_queue"},
	{ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE, "sync_r_pend_queue"},
	{ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE, "sync_w_pend_queue"},
	{ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE, "async_r_pend_queue"},
	{ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE, "async_w_pend_queue"},
	{ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE, "async_scrub_pend_queue"},
	{ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE, "rebuild_pend_queue"},
	{NULL, NULL}
	};

	if (nvlist_lookup_nvlist(nvroot,
	ZPOOL_CONFIG_VDEV_STATS_EX, &nv_ex) != 0) {
	return (6);
	}

	printf("%s%s,name=%s,vdev=root ", VDEV_MEASUREMENT, tags,
	pool_name);
	for (int i = 0; queue_type[i].name; i++) {
	if (nvlist_lookup_uint64(nv_ex,
	queue_type[i].name, &value) != 0) {
	fprintf(stderr, "error: can't get %s\n",
	queue_type[i].name);
	return (3);
	}
	if (i > 0)
	printf(",");
	print_kv(queue_type[i].short_name, value);
	}

	printf(" %llu\n", (u_longlong_t)timestamp);
	return (0);
	}

	/*
	* recursive stats printer
	*/
	static int
	print_recursive_stats(stat_printer_f func, nvlist_t *nvroot,
	const char pool_name, const char parent_name, int descend)
	{
	uint_t c, children;
	nvlist_t **child;
	char vdev_name[256];
	int err;

	err = func(nvroot, pool_name, parent_name);
	if (err)
	return (err);

	if (descend && nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	(void) strlcpy(vdev_name, get_vdev_name(nvroot, parent_name),
	sizeof (vdev_name));

	for (c = 0; c < children; c++) {
	- print_recursive_stats(func, child[c], pool_name,
	+ err = print_recursive_stats(func, child[c], pool_name,
	vdev_name, descend);
	+ if (err)
	+ return (err);
	}
	}
	return (0);
	}

	/*
	* call-back to print the stats from the pool config
	*
	* Note: if the pool is broken, this can hang indefinitely and perhaps in an
	* unkillable state.
	*/
	static int
	print_stats(zpool_handle_t zhp, void data)
	{
	uint_t c;
	int err;
	boolean_t missing;
	nvlist_t config, nvroot;
	vdev_stat_t *vs;
	struct timespec tv;
	char *pool_name;

	/* if not this pool return quickly */
	if (data &&
	strncmp(data, zpool_get_name(zhp), ZFS_MAX_DATASET_NAME_LEN) != 0) {
	zpool_close(zhp);
	return (0);
	}

	if (zpool_refresh_stats(zhp, &missing) != 0) {
	zpool_close(zhp);
	return (1);
	}

	config = zpool_get_config(zhp, NULL);
	if (clock_gettime(CLOCK_REALTIME, &tv) != 0)
	timestamp = (uint64_t)time(NULL) * 1000000000;
	else
	timestamp =
	((uint64_t)tv.tv_sec * 1000000000) + (uint64_t)tv.tv_nsec;

	if (nvlist_lookup_nvlist(
	config, ZPOOL_CONFIG_VDEV_TREE, &nvroot) != 0) {
	zpool_close(zhp);
	return (2);
	}
	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) != 0) {
	zpool_close(zhp);
	return (3);
	}

	pool_name = escape_string(zpool_get_name(zhp));
	err = print_recursive_stats(print_summary_stats, nvroot,
	pool_name, NULL, 1);
	/* if any of these return an error, skip the rest */
	if (err == 0)
	err = print_top_level_vdev_stats(nvroot, pool_name);

	if (no_histograms == 0) {
	if (err == 0)
	err = print_recursive_stats(print_vdev_latency_stats, nvroot,
	pool_name, NULL, 1);
	if (err == 0)
	err = print_recursive_stats(print_vdev_size_stats, nvroot,
	pool_name, NULL, 1);
	if (err == 0)
	err = print_recursive_stats(print_queue_stats, nvroot,
	pool_name, NULL, 0);
	}
	if (err == 0)
	err = print_scan_status(nvroot, pool_name);

	free(pool_name);
	zpool_close(zhp);
	return (err);
	}

	static void
	usage(char *name)
	{
	fprintf(stderr, "usage: %s [--execd][--no-histograms]"
	"[--sum-histogram-buckets] [--signed-int] [poolname]\n", name);
	exit(EXIT_FAILURE);
	}

	int
	main(int argc, char *argv[])
	{
	int opt;
	int ret = 8;
	char line = NULL, ttags = NULL;
	size_t len, tagslen = 0;
	struct option long_options[] = {
	{"execd", no_argument, NULL, 'e'},
	{"help", no_argument, NULL, 'h'},
	{"no-histograms", no_argument, NULL, 'n'},
	{"signed-int", no_argument, NULL, 'i'},
	{"sum-histogram-buckets", no_argument, NULL, 's'},
	{"tags", required_argument, NULL, 't'},
	{0, 0, 0, 0}
	};
	while ((opt = getopt_long(
	argc, argv, "ehinst:", long_options, NULL)) != -1) {
	switch (opt) {
	case 'e':
	execd_mode = 1;
	break;
	case 'i':
	metric_data_type = 'i';
	metric_value_mask = INT64_MAX;
	break;
	case 'n':
	no_histograms = 1;
	break;
	case 's':
	sum_histogram_buckets = 1;
	break;
	case 't':
	free(ttags);
	tagslen = strlen(optarg) + 2;
	ttags = calloc(1, tagslen);
	if (ttags == NULL) {
	fprintf(stderr,
	"error: cannot allocate memory "
	"for tags\n");
	exit(1);
	}
	(void) snprintf(ttags, tagslen, ",%s", optarg);
	tags = ttags;
	break;
	default:
	usage(argv[0]);
	}
	}

	libzfs_handle_t *g_zfs;
	if ((g_zfs = libzfs_init()) == NULL) {
	fprintf(stderr,
	"error: cannot initialize libzfs. "
	"Is the zfs module loaded or zrepl running?\n");
	exit(EXIT_FAILURE);
	}
	if (execd_mode == 0) {
	ret = zpool_iter(g_zfs, print_stats, argv[optind]);
	return (ret);
	}
	while (getline(&line, &len, stdin) != -1) {
	ret = zpool_iter(g_zfs, print_stats, argv[optind]);
	fflush(stdout);
	}
	return (ret);
	}
	diff --git a/sys/contrib/openzfs/cmd/zstream/zstream_dump.c b/sys/contrib/openzfs/cmd/zstream/zstream_dump.c
	index 0e3954678509..4b562c237373 100644
	--- a/sys/contrib/openzfs/cmd/zstream/zstream_dump.c
	+++ b/sys/contrib/openzfs/cmd/zstream/zstream_dump.c
	@@ -1,812 +1,812 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*
	* Portions Copyright 2012 Martin Matuska <martin@matuska.org>
	*/

	/*
	* Copyright (c) 2013, 2015 by Delphix. All rights reserved.
	*/

	#include <ctype.h>
	#include <libnvpair.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <stddef.h>

	#include <sys/dmu.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zio.h>
	#include <zfs_fletcher.h>
	#include "zstream.h"

	/*
	* If dump mode is enabled, the number of bytes to print per line
	*/
	#define BYTES_PER_LINE 16
	/*
	* If dump mode is enabled, the number of bytes to group together, separated
	* by newlines or spaces
	*/
	#define DUMP_GROUPING 4

	uint64_t total_stream_len = 0;
	FILE *send_stream = 0;
	boolean_t do_byteswap = B_FALSE;
	boolean_t do_cksum = B_TRUE;

	void *
	safe_malloc(size_t size)
	{
	void *rv = malloc(size);
	if (rv == NULL) {
	(void) fprintf(stderr, "ERROR; failed to allocate %zu bytes\n",
	size);
	abort();
	}
	return (rv);
	}

	/*
	* ssread - send stream read.
	*
	* Read while computing incremental checksum
	*/
	static size_t
	ssread(void buf, size_t len, zio_cksum_t cksum)
	{
	size_t outlen;

	if ((outlen = fread(buf, len, 1, send_stream)) == 0)
	return (0);

	if (do_cksum) {
	if (do_byteswap)
	fletcher_4_incremental_byteswap(buf, len, cksum);
	else
	fletcher_4_incremental_native(buf, len, cksum);
	}
	total_stream_len += len;
	return (outlen);
	}

	static size_t
	read_hdr(dmu_replay_record_t drr, zio_cksum_t cksum)
	{
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	size_t r = ssread(drr, sizeof (*drr) - sizeof (zio_cksum_t), cksum);
	if (r == 0)
	return (0);
	zio_cksum_t saved_cksum = *cksum;
	r = ssread(&drr->drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), cksum);
	if (r == 0)
	return (0);
	if (do_cksum &&
	!ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.drr_checksum.drr_checksum) &&
	!ZIO_CHECKSUM_EQUAL(saved_cksum,
	drr->drr_u.drr_checksum.drr_checksum)) {
	fprintf(stderr, "invalid checksum\n");
	(void) printf("Incorrect checksum in record header.\n");
	(void) printf("Expected checksum = %llx/%llx/%llx/%llx\n",
	(longlong_t)saved_cksum.zc_word[0],
	(longlong_t)saved_cksum.zc_word[1],
	(longlong_t)saved_cksum.zc_word[2],
	(longlong_t)saved_cksum.zc_word[3]);
	return (0);
	}
	return (sizeof (*drr));
	}

	/*
	* Print part of a block in ASCII characters
	*/
	static void
	print_ascii_block(char *subbuf, int length)
	{
	int i;

	for (i = 0; i < length; i++) {
	char char_print = isprint(subbuf[i]) ? subbuf[i] : '.';
	if (i != 0 && i % DUMP_GROUPING == 0) {
	(void) printf(" ");
	}
	(void) printf("%c", char_print);
	}
	(void) printf("\n");
	}

	/*
	* print_block - Dump the contents of a modified block to STDOUT
	*
	* Assume that buf has capacity evenly divisible by BYTES_PER_LINE
	*/
	static void
	print_block(char *buf, int length)
	{
	int i;
	/*
	* Start printing ASCII characters at a constant offset, after
	* the hex prints. Leave 3 characters per byte on a line (2 digit
	* hex number plus 1 space) plus spaces between characters and
	* groupings.
	*/
	int ascii_start = BYTES_PER_LINE * 3 +
	BYTES_PER_LINE / DUMP_GROUPING + 2;

	for (i = 0; i < length; i += BYTES_PER_LINE) {
	int j;
	int this_line_length = MIN(BYTES_PER_LINE, length - i);
	int print_offset = 0;

	for (j = 0; j < this_line_length; j++) {
	int buf_offset = i + j;

	/*
	* Separate every DUMP_GROUPING bytes by a space.
	*/
	if (buf_offset % DUMP_GROUPING == 0) {
	print_offset += printf(" ");
	}

	/*
	* Print the two-digit hex value for this byte.
	*/
	unsigned char hex_print = buf[buf_offset];
	print_offset += printf("%02x ", hex_print);
	}

	(void) printf("%*s", ascii_start - print_offset, " ");

	print_ascii_block(buf + i, this_line_length);
	}
	}

	/*
	* Print an array of bytes to stdout as hexadecimal characters. str must
	* have buf_len * 2 + 1 bytes of space.
	*/
	static void
	sprintf_bytes(char str, uint8_t buf, uint_t buf_len)
	{
	int i, n;

	for (i = 0; i < buf_len; i++) {
	n = sprintf(str, "%02x", buf[i] & 0xff);
	str += n;
	}

	str[0] = '\0';
	}

	int
	zstream_do_dump(int argc, char *argv[])
	{
	char *buf = safe_malloc(SPA_MAXBLOCKSIZE);
	uint64_t drr_record_count[DRR_NUMTYPES] = { 0 };
	uint64_t total_payload_size = 0;
	uint64_t total_overhead_size = 0;
	uint64_t drr_byte_count[DRR_NUMTYPES] = { 0 };
	char salt[ZIO_DATA_SALT_LEN * 2 + 1];
	char iv[ZIO_DATA_IV_LEN * 2 + 1];
	char mac[ZIO_DATA_MAC_LEN * 2 + 1];
	uint64_t total_records = 0;
	uint64_t payload_size;
	dmu_replay_record_t thedrr;
	dmu_replay_record_t *drr = &thedrr;
	struct drr_begin *drrb = &thedrr.drr_u.drr_begin;
	struct drr_end *drre = &thedrr.drr_u.drr_end;
	struct drr_object *drro = &thedrr.drr_u.drr_object;
	struct drr_freeobjects *drrfo = &thedrr.drr_u.drr_freeobjects;
	struct drr_write *drrw = &thedrr.drr_u.drr_write;
	struct drr_write_byref *drrwbr = &thedrr.drr_u.drr_write_byref;
	struct drr_free *drrf = &thedrr.drr_u.drr_free;
	struct drr_spill *drrs = &thedrr.drr_u.drr_spill;
	struct drr_write_embedded *drrwe = &thedrr.drr_u.drr_write_embedded;
	struct drr_object_range *drror = &thedrr.drr_u.drr_object_range;
	struct drr_redact *drrr = &thedrr.drr_u.drr_redact;
	struct drr_checksum *drrc = &thedrr.drr_u.drr_checksum;
	int c;
	boolean_t verbose = B_FALSE;
	boolean_t very_verbose = B_FALSE;
	boolean_t first = B_TRUE;
	/*
	* dump flag controls whether the contents of any modified data blocks
	* are printed to the console during processing of the stream. Warning:
	* for large streams, this can obviously lead to massive prints.
	*/
	boolean_t dump = B_FALSE;
	int err;
	zio_cksum_t zc = { { 0 } };
	zio_cksum_t pcksum = { { 0 } };

	while ((c = getopt(argc, argv, ":vCd")) != -1) {
	switch (c) {
	case 'C':
	do_cksum = B_FALSE;
	break;
	case 'v':
	if (verbose)
	very_verbose = B_TRUE;
	verbose = B_TRUE;
	break;
	case 'd':
	dump = B_TRUE;
	verbose = B_TRUE;
	very_verbose = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr,
	"missing argument for '%c' option\n", optopt);
	zstream_usage();
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	zstream_usage();
	break;
	}
	}

	if (argc > optind) {
	const char *filename = argv[optind];
	send_stream = fopen(filename, "r");
	if (send_stream == NULL) {
	(void) fprintf(stderr,
	"Error while opening file '%s': %s\n",
	filename, strerror(errno));
	exit(1);
	}
	} else {
	if (isatty(STDIN_FILENO)) {
	(void) fprintf(stderr,
	"Error: The send stream is a binary format "
	"and can not be read from a\n"
	"terminal. Standard input must be redirected, "
	"or a file must be\n"
	"specified as a command-line argument.\n");
	exit(1);
	}
	send_stream = stdin;
	}

	fletcher_4_init();
	while (read_hdr(drr, &zc)) {
	uint64_t featureflags = 0;

	/*
	* If this is the first DMU record being processed, check for
	* the magic bytes and figure out the endian-ness based on them.
	*/
	if (first) {
	if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
	do_byteswap = B_TRUE;
	if (do_cksum) {
	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	/*
	* recalculate header checksum now
	* that we know it needs to be
	* byteswapped.
	*/
	fletcher_4_incremental_byteswap(drr,
	sizeof (dmu_replay_record_t), &zc);
	}
	} else if (drrb->drr_magic != DMU_BACKUP_MAGIC) {
	(void) fprintf(stderr, "Invalid stream "
	"(bad magic number)\n");
	exit(1);
	}
	first = B_FALSE;
	}
	if (do_byteswap) {
	drr->drr_type = BSWAP_32(drr->drr_type);
	drr->drr_payloadlen =
	BSWAP_32(drr->drr_payloadlen);
	}

	/*
	* At this point, the leading fields of the replay record
	* (drr_type and drr_payloadlen) have been byte-swapped if
	* necessary, but the rest of the data structure (the
	* union of type-specific structures) is still in its
	* original state.
	*/
	if (drr->drr_type >= DRR_NUMTYPES) {
	(void) printf("INVALID record found: type 0x%x\n",
	drr->drr_type);
	(void) printf("Aborting.\n");
	exit(1);
	}

	drr_record_count[drr->drr_type]++;
	total_overhead_size += sizeof (*drr);
	total_records++;
	payload_size = 0;

	switch (drr->drr_type) {
	case DRR_BEGIN:
	if (do_byteswap) {
	drrb->drr_magic = BSWAP_64(drrb->drr_magic);
	drrb->drr_versioninfo =
	BSWAP_64(drrb->drr_versioninfo);
	drrb->drr_creation_time =
	BSWAP_64(drrb->drr_creation_time);
	drrb->drr_type = BSWAP_32(drrb->drr_type);
	drrb->drr_flags = BSWAP_32(drrb->drr_flags);
	drrb->drr_toguid = BSWAP_64(drrb->drr_toguid);
	drrb->drr_fromguid =
	BSWAP_64(drrb->drr_fromguid);
	}

	- featureflags =
	- DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	-
	(void) printf("BEGIN record\n");
	(void) printf("\thdrtype = %lld\n",
	DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo));
	(void) printf("\tfeatures = %llx\n",
	DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo));
	(void) printf("\tmagic = %llx\n",
	(u_longlong_t)drrb->drr_magic);
	(void) printf("\tcreation_time = %llx\n",
	(u_longlong_t)drrb->drr_creation_time);
	(void) printf("\ttype = %u\n", drrb->drr_type);
	(void) printf("\tflags = 0x%x\n", drrb->drr_flags);
	(void) printf("\ttoguid = %llx\n",
	(u_longlong_t)drrb->drr_toguid);
	(void) printf("\tfromguid = %llx\n",
	(u_longlong_t)drrb->drr_fromguid);
	(void) printf("\ttoname = %s\n", drrb->drr_toname);
	(void) printf("\tpayloadlen = %u\n",
	drr->drr_payloadlen);
	if (verbose)
	(void) printf("\n");

	if (drr->drr_payloadlen != 0) {
	nvlist_t *nv;
	int sz = drr->drr_payloadlen;

	if (sz > SPA_MAXBLOCKSIZE) {
	free(buf);
	buf = safe_malloc(sz);
	}
	(void) ssread(buf, sz, &zc);
	if (ferror(send_stream))
	perror("fread");
	err = nvlist_unpack(buf, sz, &nv, 0);
	if (err) {
	perror(strerror(err));
	} else {
	nvlist_print(stdout, nv);
	nvlist_free(nv);
	}
	payload_size = sz;
	}
	break;

	case DRR_END:
	if (do_byteswap) {
	drre->drr_checksum.zc_word[0] =
	BSWAP_64(drre->drr_checksum.zc_word[0]);
	drre->drr_checksum.zc_word[1] =
	BSWAP_64(drre->drr_checksum.zc_word[1]);
	drre->drr_checksum.zc_word[2] =
	BSWAP_64(drre->drr_checksum.zc_word[2]);
	drre->drr_checksum.zc_word[3] =
	BSWAP_64(drre->drr_checksum.zc_word[3]);
	}
	/*
	* We compare against the previous checksum
	* value, because the stored checksum is of
	* everything before the DRR_END record.
	*/
	if (do_cksum && !ZIO_CHECKSUM_EQUAL(drre->drr_checksum,
	pcksum)) {
	(void) printf("Expected checksum differs from "
	"checksum in stream.\n");
	(void) printf("Expected checksum = "
	"%llx/%llx/%llx/%llx\n",
	(long long unsigned int)pcksum.zc_word[0],
	(long long unsigned int)pcksum.zc_word[1],
	(long long unsigned int)pcksum.zc_word[2],
	(long long unsigned int)pcksum.zc_word[3]);
	}
	(void) printf("END checksum = %llx/%llx/%llx/%llx\n",
	(long long unsigned int)
	drre->drr_checksum.zc_word[0],
	(long long unsigned int)
	drre->drr_checksum.zc_word[1],
	(long long unsigned int)
	drre->drr_checksum.zc_word[2],
	(long long unsigned int)
	drre->drr_checksum.zc_word[3]);

	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	break;

	case DRR_OBJECT:
	if (do_byteswap) {
	drro->drr_object = BSWAP_64(drro->drr_object);
	drro->drr_type = BSWAP_32(drro->drr_type);
	drro->drr_bonustype =
	BSWAP_32(drro->drr_bonustype);
	drro->drr_blksz = BSWAP_32(drro->drr_blksz);
	drro->drr_bonuslen =
	BSWAP_32(drro->drr_bonuslen);
	drro->drr_raw_bonuslen =
	BSWAP_32(drro->drr_raw_bonuslen);
	drro->drr_toguid = BSWAP_64(drro->drr_toguid);
	drro->drr_maxblkid =
	BSWAP_64(drro->drr_maxblkid);
	}

	+ featureflags =
	+ DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	+
	if (featureflags & DMU_BACKUP_FEATURE_RAW &&
	drro->drr_bonuslen > drro->drr_raw_bonuslen) {
	(void) fprintf(stderr,
	"Warning: Object %llu has bonuslen = "
	"%u > raw_bonuslen = %u\n\n",
	(u_longlong_t)drro->drr_object,
	drro->drr_bonuslen, drro->drr_raw_bonuslen);
	}

	payload_size = DRR_OBJECT_PAYLOAD_SIZE(drro);

	if (verbose) {
	(void) printf("OBJECT object = %llu type = %u "
	"bonustype = %u blksz = %u bonuslen = %u "
	"dn_slots = %u raw_bonuslen = %u "
	"flags = %u maxblkid = %llu "
	"indblkshift = %u nlevels = %u "
	"nblkptr = %u\n",
	(u_longlong_t)drro->drr_object,
	drro->drr_type,
	drro->drr_bonustype,
	drro->drr_blksz,
	drro->drr_bonuslen,
	drro->drr_dn_slots,
	drro->drr_raw_bonuslen,
	drro->drr_flags,
	(u_longlong_t)drro->drr_maxblkid,
	drro->drr_indblkshift,
	drro->drr_nlevels,
	drro->drr_nblkptr);
	}
	if (drro->drr_bonuslen > 0) {
	(void) ssread(buf, payload_size, &zc);
	if (dump)
	print_block(buf, payload_size);
	}
	break;

	case DRR_FREEOBJECTS:
	if (do_byteswap) {
	drrfo->drr_firstobj =
	BSWAP_64(drrfo->drr_firstobj);
	drrfo->drr_numobjs =
	BSWAP_64(drrfo->drr_numobjs);
	drrfo->drr_toguid = BSWAP_64(drrfo->drr_toguid);
	}
	if (verbose) {
	(void) printf("FREEOBJECTS firstobj = %llu "
	"numobjs = %llu\n",
	(u_longlong_t)drrfo->drr_firstobj,
	(u_longlong_t)drrfo->drr_numobjs);
	}
	break;

	case DRR_WRITE:
	if (do_byteswap) {
	drrw->drr_object = BSWAP_64(drrw->drr_object);
	drrw->drr_type = BSWAP_32(drrw->drr_type);
	drrw->drr_offset = BSWAP_64(drrw->drr_offset);
	drrw->drr_logical_size =
	BSWAP_64(drrw->drr_logical_size);
	drrw->drr_toguid = BSWAP_64(drrw->drr_toguid);
	drrw->drr_key.ddk_prop =
	BSWAP_64(drrw->drr_key.ddk_prop);
	drrw->drr_compressed_size =
	BSWAP_64(drrw->drr_compressed_size);
	}

	payload_size = DRR_WRITE_PAYLOAD_SIZE(drrw);

	/*
	* If this is verbose and/or dump output,
	* print info on the modified block
	*/
	if (verbose) {
	sprintf_bytes(salt, drrw->drr_salt,
	ZIO_DATA_SALT_LEN);
	sprintf_bytes(iv, drrw->drr_iv,
	ZIO_DATA_IV_LEN);
	sprintf_bytes(mac, drrw->drr_mac,
	ZIO_DATA_MAC_LEN);

	(void) printf("WRITE object = %llu type = %u "
	"checksum type = %u compression type = %u "
	"flags = %u offset = %llu "
	"logical_size = %llu "
	"compressed_size = %llu "
	"payload_size = %llu props = %llx "
	"salt = %s iv = %s mac = %s\n",
	(u_longlong_t)drrw->drr_object,
	drrw->drr_type,
	drrw->drr_checksumtype,
	drrw->drr_compressiontype,
	drrw->drr_flags,
	(u_longlong_t)drrw->drr_offset,
	(u_longlong_t)drrw->drr_logical_size,
	(u_longlong_t)drrw->drr_compressed_size,
	(u_longlong_t)payload_size,
	(u_longlong_t)drrw->drr_key.ddk_prop,
	salt,
	iv,
	mac);
	}

	/*
	* Read the contents of the block in from STDIN to buf
	*/
	(void) ssread(buf, payload_size, &zc);
	/*
	* If in dump mode
	*/
	if (dump) {
	print_block(buf, payload_size);
	}
	break;

	case DRR_WRITE_BYREF:
	if (do_byteswap) {
	drrwbr->drr_object =
	BSWAP_64(drrwbr->drr_object);
	drrwbr->drr_offset =
	BSWAP_64(drrwbr->drr_offset);
	drrwbr->drr_length =
	BSWAP_64(drrwbr->drr_length);
	drrwbr->drr_toguid =
	BSWAP_64(drrwbr->drr_toguid);
	drrwbr->drr_refguid =
	BSWAP_64(drrwbr->drr_refguid);
	drrwbr->drr_refobject =
	BSWAP_64(drrwbr->drr_refobject);
	drrwbr->drr_refoffset =
	BSWAP_64(drrwbr->drr_refoffset);
	drrwbr->drr_key.ddk_prop =
	BSWAP_64(drrwbr->drr_key.ddk_prop);
	}
	if (verbose) {
	(void) printf("WRITE_BYREF object = %llu "
	"checksum type = %u props = %llx "
	"offset = %llu length = %llu "
	"toguid = %llx refguid = %llx "
	"refobject = %llu refoffset = %llu\n",
	(u_longlong_t)drrwbr->drr_object,
	drrwbr->drr_checksumtype,
	(u_longlong_t)drrwbr->drr_key.ddk_prop,
	(u_longlong_t)drrwbr->drr_offset,
	(u_longlong_t)drrwbr->drr_length,
	(u_longlong_t)drrwbr->drr_toguid,
	(u_longlong_t)drrwbr->drr_refguid,
	(u_longlong_t)drrwbr->drr_refobject,
	(u_longlong_t)drrwbr->drr_refoffset);
	}
	break;

	case DRR_FREE:
	if (do_byteswap) {
	drrf->drr_object = BSWAP_64(drrf->drr_object);
	drrf->drr_offset = BSWAP_64(drrf->drr_offset);
	drrf->drr_length = BSWAP_64(drrf->drr_length);
	}
	if (verbose) {
	(void) printf("FREE object = %llu "
	"offset = %llu length = %lld\n",
	(u_longlong_t)drrf->drr_object,
	(u_longlong_t)drrf->drr_offset,
	(longlong_t)drrf->drr_length);
	}
	break;
	case DRR_SPILL:
	if (do_byteswap) {
	drrs->drr_object = BSWAP_64(drrs->drr_object);
	drrs->drr_length = BSWAP_64(drrs->drr_length);
	drrs->drr_compressed_size =
	BSWAP_64(drrs->drr_compressed_size);
	drrs->drr_type = BSWAP_32(drrs->drr_type);
	}

	payload_size = DRR_SPILL_PAYLOAD_SIZE(drrs);

	if (verbose) {
	sprintf_bytes(salt, drrs->drr_salt,
	ZIO_DATA_SALT_LEN);
	sprintf_bytes(iv, drrs->drr_iv,
	ZIO_DATA_IV_LEN);
	sprintf_bytes(mac, drrs->drr_mac,
	ZIO_DATA_MAC_LEN);

	(void) printf("SPILL block for object = %llu "
	"length = %llu flags = %u "
	"compression type = %u "
	"compressed_size = %llu "
	"payload_size = %llu "
	"salt = %s iv = %s mac = %s\n",
	(u_longlong_t)drrs->drr_object,
	(u_longlong_t)drrs->drr_length,
	drrs->drr_flags,
	drrs->drr_compressiontype,
	(u_longlong_t)drrs->drr_compressed_size,
	(u_longlong_t)payload_size,
	salt,
	iv,
	mac);
	}
	(void) ssread(buf, payload_size, &zc);
	if (dump) {
	print_block(buf, payload_size);
	}
	break;
	case DRR_WRITE_EMBEDDED:
	if (do_byteswap) {
	drrwe->drr_object =
	BSWAP_64(drrwe->drr_object);
	drrwe->drr_offset =
	BSWAP_64(drrwe->drr_offset);
	drrwe->drr_length =
	BSWAP_64(drrwe->drr_length);
	drrwe->drr_toguid =
	BSWAP_64(drrwe->drr_toguid);
	drrwe->drr_lsize =
	BSWAP_32(drrwe->drr_lsize);
	drrwe->drr_psize =
	BSWAP_32(drrwe->drr_psize);
	}
	if (verbose) {
	(void) printf("WRITE_EMBEDDED object = %llu "
	"offset = %llu length = %llu "
	"toguid = %llx comp = %u etype = %u "
	"lsize = %u psize = %u\n",
	(u_longlong_t)drrwe->drr_object,
	(u_longlong_t)drrwe->drr_offset,
	(u_longlong_t)drrwe->drr_length,
	(u_longlong_t)drrwe->drr_toguid,
	drrwe->drr_compression,
	drrwe->drr_etype,
	drrwe->drr_lsize,
	drrwe->drr_psize);
	}
	(void) ssread(buf,
	P2ROUNDUP(drrwe->drr_psize, 8), &zc);
	if (dump) {
	print_block(buf,
	P2ROUNDUP(drrwe->drr_psize, 8));
	}
	payload_size = P2ROUNDUP(drrwe->drr_psize, 8);
	break;
	case DRR_OBJECT_RANGE:
	if (do_byteswap) {
	drror->drr_firstobj =
	BSWAP_64(drror->drr_firstobj);
	drror->drr_numslots =
	BSWAP_64(drror->drr_numslots);
	drror->drr_toguid = BSWAP_64(drror->drr_toguid);
	}
	if (verbose) {
	sprintf_bytes(salt, drror->drr_salt,
	ZIO_DATA_SALT_LEN);
	sprintf_bytes(iv, drror->drr_iv,
	ZIO_DATA_IV_LEN);
	sprintf_bytes(mac, drror->drr_mac,
	ZIO_DATA_MAC_LEN);

	(void) printf("OBJECT_RANGE firstobj = %llu "
	"numslots = %llu flags = %u "
	"salt = %s iv = %s mac = %s\n",
	(u_longlong_t)drror->drr_firstobj,
	(u_longlong_t)drror->drr_numslots,
	drror->drr_flags,
	salt,
	iv,
	mac);
	}
	break;
	case DRR_REDACT:
	if (do_byteswap) {
	drrr->drr_object = BSWAP_64(drrr->drr_object);
	drrr->drr_offset = BSWAP_64(drrr->drr_offset);
	drrr->drr_length = BSWAP_64(drrr->drr_length);
	drrr->drr_toguid = BSWAP_64(drrr->drr_toguid);
	}
	if (verbose) {
	(void) printf("REDACT object = %llu offset = "
	"%llu length = %llu\n",
	(u_longlong_t)drrr->drr_object,
	(u_longlong_t)drrr->drr_offset,
	(u_longlong_t)drrr->drr_length);
	}
	break;
	case DRR_NUMTYPES:
	/* should never be reached */
	exit(1);
	}
	if (drr->drr_type != DRR_BEGIN && very_verbose) {
	(void) printf(" checksum = %llx/%llx/%llx/%llx\n",
	(longlong_t)drrc->drr_checksum.zc_word[0],
	(longlong_t)drrc->drr_checksum.zc_word[1],
	(longlong_t)drrc->drr_checksum.zc_word[2],
	(longlong_t)drrc->drr_checksum.zc_word[3]);
	}
	pcksum = zc;
	drr_byte_count[drr->drr_type] += payload_size;
	total_payload_size += payload_size;
	}
	free(buf);
	fletcher_4_fini();

	/* Print final summary */

	(void) printf("SUMMARY:\n");
	(void) printf("\tTotal DRR_BEGIN records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_BEGIN],
	(u_longlong_t)drr_byte_count[DRR_BEGIN]);
	(void) printf("\tTotal DRR_END records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_END],
	(u_longlong_t)drr_byte_count[DRR_END]);
	(void) printf("\tTotal DRR_OBJECT records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_OBJECT],
	(u_longlong_t)drr_byte_count[DRR_OBJECT]);
	(void) printf("\tTotal DRR_FREEOBJECTS records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_FREEOBJECTS],
	(u_longlong_t)drr_byte_count[DRR_FREEOBJECTS]);
	(void) printf("\tTotal DRR_WRITE records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_WRITE],
	(u_longlong_t)drr_byte_count[DRR_WRITE]);
	(void) printf("\tTotal DRR_WRITE_BYREF records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_WRITE_BYREF],
	(u_longlong_t)drr_byte_count[DRR_WRITE_BYREF]);
	(void) printf("\tTotal DRR_WRITE_EMBEDDED records = %lld (%llu "
	"bytes)\n", (u_longlong_t)drr_record_count[DRR_WRITE_EMBEDDED],
	(u_longlong_t)drr_byte_count[DRR_WRITE_EMBEDDED]);
	(void) printf("\tTotal DRR_FREE records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_FREE],
	(u_longlong_t)drr_byte_count[DRR_FREE]);
	(void) printf("\tTotal DRR_SPILL records = %lld (%llu bytes)\n",
	(u_longlong_t)drr_record_count[DRR_SPILL],
	(u_longlong_t)drr_byte_count[DRR_SPILL]);
	(void) printf("\tTotal records = %lld\n",
	(u_longlong_t)total_records);
	(void) printf("\tTotal payload size = %lld (0x%llx)\n",
	(u_longlong_t)total_payload_size, (u_longlong_t)total_payload_size);
	(void) printf("\tTotal header overhead = %lld (0x%llx)\n",
	(u_longlong_t)total_overhead_size,
	(u_longlong_t)total_overhead_size);
	(void) printf("\tTotal stream length = %lld (0x%llx)\n",
	(u_longlong_t)total_stream_len, (u_longlong_t)total_stream_len);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/cmd/zstream/zstream_token.c b/sys/contrib/openzfs/cmd/zstream/zstream_token.c
	index 23cc56dce37c..795a372633ef 100644
	--- a/sys/contrib/openzfs/cmd/zstream/zstream_token.c
	+++ b/sys/contrib/openzfs/cmd/zstream/zstream_token.c
	@@ -1,78 +1,82 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*
	* Portions Copyright 2012 Martin Matuska <martin@matuska.org>
	*/

	/*
	* Copyright (c) 2020 by Datto Inc. All rights reserved.
	*/

	#include <ctype.h>
	#include <libnvpair.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <stddef.h>

	#include <libzfs.h>
	#include <libzfs_core.h>

	#include <sys/dmu.h>
	#include <sys/zfs_ioctl.h>
	#include "zstream.h"

	int
	zstream_do_token(int argc, char *argv[])
	{
	char *resume_token = NULL;
	+ libzfs_handle_t *hdl;

	if (argc < 2) {
	(void) fprintf(stderr, "Need to pass the resume token\n");
	zstream_usage();
	}

	resume_token = argv[1];

	- libzfs_handle_t *hdl = libzfs_init();
	+ if ((hdl = libzfs_init()) == NULL) {
	+ (void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
	+ return (1);
	+ }

	nvlist_t *resume_nvl =
	zfs_send_resume_token_to_nvlist(hdl, resume_token);

	if (resume_nvl == NULL) {
	(void) fprintf(stderr,
	"Unable to parse resume token: %s\n",
	libzfs_error_description(hdl));
	libzfs_fini(hdl);
	return (1);
	}

	dump_nvlist(resume_nvl, 5);
	nvlist_free(resume_nvl);

	libzfs_fini(hdl);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/cmd/ztest.c b/sys/contrib/openzfs/cmd/ztest.c
	index 0712f286bf66..e630d3353187 100644
	--- a/sys/contrib/openzfs/cmd/ztest.c
	+++ b/sys/contrib/openzfs/cmd/ztest.c
	@@ -1,8288 +1,8295 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	*/

	/*
	* The objective of this program is to provide a DMU/ZAP/SPA stress test
	* that runs entirely in userland, is easy to use, and easy to extend.
	*
	* The overall design of the ztest program is as follows:
	*
	* (1) For each major functional area (e.g. adding vdevs to a pool,
	* creating and destroying datasets, reading and writing objects, etc)
	* we have a simple routine to test that functionality. These
	* individual routines do not have to do anything "stressful".
	*
	* (2) We turn these simple functionality tests into a stress test by
	* running them all in parallel, with as many threads as desired,
	* and spread across as many datasets, objects, and vdevs as desired.
	*
	* (3) While all this is happening, we inject faults into the pool to
	* verify that self-healing data really works.
	*
	* (4) Every time we open a dataset, we change its checksum and compression
	* functions. Thus even individual objects vary from block to block
	* in which checksum they use and whether they're compressed.
	*
	* (5) To verify that we never lose on-disk consistency after a crash,
	* we run the entire test in a child of the main process.
	* At random times, the child self-immolates with a SIGKILL.
	* This is the software equivalent of pulling the power cord.
	* The parent then runs the test again, using the existing
	* storage pool, as many times as desired. If backwards compatibility
	* testing is enabled ztest will sometimes run the "older" version
	* of ztest after a SIGKILL.
	*
	* (6) To verify that we don't have future leaks or temporal incursions,
	* many of the functional tests record the transaction group number
	* as part of their data. When reading old data, they verify that
	* the transaction group number is less than the current, open txg.
	* If you add a new test, please do this if applicable.
	*
	* (7) Threads are created with a reduced stack size, for sanity checking.
	* Therefore, it's important not to allocate huge buffers on the stack.
	*
	* When run with no arguments, ztest runs for about five minutes and
	* produces no output if successful. To get a little bit of information,
	* specify -V. To get more information, specify -VV, and so on.
	*
	* To turn this into an overnight stress test, use -T to specify run time.
	*
	* You can ask more vdevs [-v], datasets [-d], or threads [-t]
	* to increase the pool capacity, fanout, and overall stress level.
	*
	* Use the -k option to set the desired frequency of kills.
	*
	* When ztest invokes itself it passes all relevant information through a
	* temporary file which is mmap-ed in the child process. This allows shared
	* memory to survive the exec syscall. The ztest_shared_hdr_t struct is always
	* stored at offset 0 of this file and contains information on the size and
	* number of shared structures in the file. The information stored in this file
	* must remain backwards compatible with older versions of ztest so that
	* ztest can invoke them during backwards compatibility testing (-B).
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/dmu.h>
	#include <sys/txg.h>
	#include <sys/dbuf.h>
	#include <sys/zap.h>
	#include <sys/dmu_objset.h>
	#include <sys/poll.h>
	#include <sys/stat.h>
	#include <sys/time.h>
	#include <sys/wait.h>
	#include <sys/mman.h>
	#include <sys/resource.h>
	#include <sys/zio.h>
	#include <sys/zil.h>
	#include <sys/zil_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_file.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_raidz.h>
	#include <sys/vdev_trim.h>
	#include <sys/spa_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_scan.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_refcount.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_userhold.h>
	#include <sys/abd.h>
	#include <sys/blake3.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <getopt.h>
	#include <signal.h>
	#include <umem.h>
	#include <ctype.h>
	#include <math.h>
	#include <sys/fs/zfs.h>
	#include <zfs_fletcher.h>
	#include <libnvpair.h>
	#include <libzutil.h>
	#include <sys/crypto/icp.h>
	#if (__GLIBC__ && !__UCLIBC__)
	#include <execinfo.h> /* for backtrace() */
	#endif

	static int ztest_fd_data = -1;
	static int ztest_fd_rand = -1;

	typedef struct ztest_shared_hdr {
	uint64_t zh_hdr_size;
	uint64_t zh_opts_size;
	uint64_t zh_size;
	uint64_t zh_stats_size;
	uint64_t zh_stats_count;
	uint64_t zh_ds_size;
	uint64_t zh_ds_count;
	} ztest_shared_hdr_t;

	static ztest_shared_hdr_t *ztest_shared_hdr;

	enum ztest_class_state {
	ZTEST_VDEV_CLASS_OFF,
	ZTEST_VDEV_CLASS_ON,
	ZTEST_VDEV_CLASS_RND
	};

	#define ZO_GVARS_MAX_ARGLEN ((size_t)64)
	#define ZO_GVARS_MAX_COUNT ((size_t)10)

	typedef struct ztest_shared_opts {
	char zo_pool[ZFS_MAX_DATASET_NAME_LEN];
	char zo_dir[ZFS_MAX_DATASET_NAME_LEN];
	char zo_alt_ztest[MAXNAMELEN];
	char zo_alt_libpath[MAXNAMELEN];
	uint64_t zo_vdevs;
	uint64_t zo_vdevtime;
	size_t zo_vdev_size;
	int zo_ashift;
	int zo_mirrors;
	int zo_raid_children;
	int zo_raid_parity;
	char zo_raid_type[8];
	int zo_draid_data;
	int zo_draid_spares;
	int zo_datasets;
	int zo_threads;
	uint64_t zo_passtime;
	uint64_t zo_killrate;
	int zo_verbose;
	int zo_init;
	uint64_t zo_time;
	uint64_t zo_maxloops;
	uint64_t zo_metaslab_force_ganging;
	int zo_mmp_test;
	int zo_special_vdevs;
	int zo_dump_dbgmsg;
	int zo_gvars_count;
	char zo_gvars[ZO_GVARS_MAX_COUNT][ZO_GVARS_MAX_ARGLEN];
	} ztest_shared_opts_t;

	/* Default values for command line options. */
	#define DEFAULT_POOL "ztest"
	#define DEFAULT_VDEV_DIR "/tmp"
	#define DEFAULT_VDEV_COUNT 5
	#define DEFAULT_VDEV_SIZE (SPA_MINDEVSIZE * 4) /* 256m default size */
	#define DEFAULT_VDEV_SIZE_STR "256M"
	#define DEFAULT_ASHIFT SPA_MINBLOCKSHIFT
	#define DEFAULT_MIRRORS 2
	#define DEFAULT_RAID_CHILDREN 4
	#define DEFAULT_RAID_PARITY 1
	#define DEFAULT_DRAID_DATA 4
	#define DEFAULT_DRAID_SPARES 1
	#define DEFAULT_DATASETS_COUNT 7
	#define DEFAULT_THREADS 23
	#define DEFAULT_RUN_TIME 300 /* 300 seconds */
	#define DEFAULT_RUN_TIME_STR "300 sec"
	#define DEFAULT_PASS_TIME 60 /* 60 seconds */
	#define DEFAULT_PASS_TIME_STR "60 sec"
	#define DEFAULT_KILL_RATE 70 /* 70% kill rate */
	#define DEFAULT_KILLRATE_STR "70%"
	#define DEFAULT_INITS 1
	#define DEFAULT_MAX_LOOPS 50 /* 5 minutes */
	#define DEFAULT_FORCE_GANGING (64 << 10)
	#define DEFAULT_FORCE_GANGING_STR "64K"

	/* Simplifying assumption: -1 is not a valid default. */
	#define NO_DEFAULT -1

	static const ztest_shared_opts_t ztest_opts_defaults = {
	.zo_pool = DEFAULT_POOL,
	.zo_dir = DEFAULT_VDEV_DIR,
	.zo_alt_ztest = { '\0' },
	.zo_alt_libpath = { '\0' },
	.zo_vdevs = DEFAULT_VDEV_COUNT,
	.zo_ashift = DEFAULT_ASHIFT,
	.zo_mirrors = DEFAULT_MIRRORS,
	.zo_raid_children = DEFAULT_RAID_CHILDREN,
	.zo_raid_parity = DEFAULT_RAID_PARITY,
	.zo_raid_type = VDEV_TYPE_RAIDZ,
	.zo_vdev_size = DEFAULT_VDEV_SIZE,
	.zo_draid_data = DEFAULT_DRAID_DATA, /* data drives */
	.zo_draid_spares = DEFAULT_DRAID_SPARES, /* distributed spares */
	.zo_datasets = DEFAULT_DATASETS_COUNT,
	.zo_threads = DEFAULT_THREADS,
	.zo_passtime = DEFAULT_PASS_TIME,
	.zo_killrate = DEFAULT_KILL_RATE,
	.zo_verbose = 0,
	.zo_mmp_test = 0,
	.zo_init = DEFAULT_INITS,
	.zo_time = DEFAULT_RUN_TIME,
	.zo_maxloops = DEFAULT_MAX_LOOPS, /* max loops during spa_freeze() */
	.zo_metaslab_force_ganging = DEFAULT_FORCE_GANGING,
	.zo_special_vdevs = ZTEST_VDEV_CLASS_RND,
	.zo_gvars_count = 0,
	};

	extern uint64_t metaslab_force_ganging;
	extern uint64_t metaslab_df_alloc_threshold;
	extern unsigned long zfs_deadman_synctime_ms;
	-extern int metaslab_preload_limit;
	+extern uint_t metaslab_preload_limit;
	extern int zfs_compressed_arc_enabled;
	extern int zfs_abd_scatter_enabled;
	-extern int dmu_object_alloc_chunk_shift;
	+extern uint_t dmu_object_alloc_chunk_shift;
	extern boolean_t zfs_force_some_double_word_sm_entries;
	extern unsigned long zio_decompress_fail_fraction;
	extern unsigned long zfs_reconstruct_indirect_damage_fraction;


	static ztest_shared_opts_t *ztest_shared_opts;
	static ztest_shared_opts_t ztest_opts;
	static const char *const ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345";

	typedef struct ztest_shared_ds {
	uint64_t zd_seq;
	} ztest_shared_ds_t;

	static ztest_shared_ds_t *ztest_shared_ds;
	#define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d])

	#define BT_MAGIC 0x123456789abcdefULL
	#define MAXFAULTS(zs) \
	(MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raid_parity + 1) - 1)

	enum ztest_io_type {
	ZTEST_IO_WRITE_TAG,
	ZTEST_IO_WRITE_PATTERN,
	ZTEST_IO_WRITE_ZEROES,
	ZTEST_IO_TRUNCATE,
	ZTEST_IO_SETATTR,
	ZTEST_IO_REWRITE,
	ZTEST_IO_TYPES
	};

	typedef struct ztest_block_tag {
	uint64_t bt_magic;
	uint64_t bt_objset;
	uint64_t bt_object;
	uint64_t bt_dnodesize;
	uint64_t bt_offset;
	uint64_t bt_gen;
	uint64_t bt_txg;
	uint64_t bt_crtxg;
	} ztest_block_tag_t;

	typedef struct bufwad {
	uint64_t bw_index;
	uint64_t bw_txg;
	uint64_t bw_data;
	} bufwad_t;

	/*
	* It would be better to use a rangelock_t per object. Unfortunately
	* the rangelock_t is not a drop-in replacement for rl_t, because we
	* still need to map from object ID to rangelock_t.
	*/
	typedef enum {
	RL_READER,
	RL_WRITER,
	RL_APPEND
	} rl_type_t;

	typedef struct rll {
	void *rll_writer;
	int rll_readers;
	kmutex_t rll_lock;
	kcondvar_t rll_cv;
	} rll_t;

	typedef struct rl {
	uint64_t rl_object;
	uint64_t rl_offset;
	uint64_t rl_size;
	rll_t *rl_lock;
	} rl_t;

	#define ZTEST_RANGE_LOCKS 64
	#define ZTEST_OBJECT_LOCKS 64

	/*
	* Object descriptor. Used as a template for object lookup/create/remove.
	*/
	typedef struct ztest_od {
	uint64_t od_dir;
	uint64_t od_object;
	dmu_object_type_t od_type;
	dmu_object_type_t od_crtype;
	uint64_t od_blocksize;
	uint64_t od_crblocksize;
	uint64_t od_crdnodesize;
	uint64_t od_gen;
	uint64_t od_crgen;
	char od_name[ZFS_MAX_DATASET_NAME_LEN];
	} ztest_od_t;

	/*
	* Per-dataset state.
	*/
	typedef struct ztest_ds {
	ztest_shared_ds_t *zd_shared;
	objset_t *zd_os;
	pthread_rwlock_t zd_zilog_lock;
	zilog_t *zd_zilog;
	ztest_od_t zd_od; / debugging aid */
	char zd_name[ZFS_MAX_DATASET_NAME_LEN];
	kmutex_t zd_dirobj_lock;
	rll_t zd_object_lock[ZTEST_OBJECT_LOCKS];
	rll_t zd_range_lock[ZTEST_RANGE_LOCKS];
	} ztest_ds_t;

	/*
	* Per-iteration state.
	*/
	typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id);

	typedef struct ztest_info {
	ztest_func_t zi_func; / test function */
	uint64_t zi_iters; /* iterations per execution */
	uint64_t zi_interval; / execute every <interval> seconds */
	const char zi_funcname; / name of test function */
	} ztest_info_t;

	typedef struct ztest_shared_callstate {
	uint64_t zc_count; /* per-pass count */
	uint64_t zc_time; /* per-pass time */
	uint64_t zc_next; /* next time to call this function */
	} ztest_shared_callstate_t;

	static ztest_shared_callstate_t *ztest_shared_callstate;
	#define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c])

	ztest_func_t ztest_dmu_read_write;
	ztest_func_t ztest_dmu_write_parallel;
	ztest_func_t ztest_dmu_object_alloc_free;
	ztest_func_t ztest_dmu_object_next_chunk;
	ztest_func_t ztest_dmu_commit_callbacks;
	ztest_func_t ztest_zap;
	ztest_func_t ztest_zap_parallel;
	ztest_func_t ztest_zil_commit;
	ztest_func_t ztest_zil_remount;
	ztest_func_t ztest_dmu_read_write_zcopy;
	ztest_func_t ztest_dmu_objset_create_destroy;
	ztest_func_t ztest_dmu_prealloc;
	ztest_func_t ztest_fzap;
	ztest_func_t ztest_dmu_snapshot_create_destroy;
	ztest_func_t ztest_dsl_prop_get_set;
	ztest_func_t ztest_spa_prop_get_set;
	ztest_func_t ztest_spa_create_destroy;
	ztest_func_t ztest_fault_inject;
	ztest_func_t ztest_dmu_snapshot_hold;
	ztest_func_t ztest_mmp_enable_disable;
	ztest_func_t ztest_scrub;
	ztest_func_t ztest_dsl_dataset_promote_busy;
	ztest_func_t ztest_vdev_attach_detach;
	ztest_func_t ztest_vdev_LUN_growth;
	ztest_func_t ztest_vdev_add_remove;
	ztest_func_t ztest_vdev_class_add;
	ztest_func_t ztest_vdev_aux_add_remove;
	ztest_func_t ztest_split_pool;
	ztest_func_t ztest_reguid;
	ztest_func_t ztest_spa_upgrade;
	ztest_func_t ztest_device_removal;
	ztest_func_t ztest_spa_checkpoint_create_discard;
	ztest_func_t ztest_initialize;
	ztest_func_t ztest_trim;
	ztest_func_t ztest_blake3;
	ztest_func_t ztest_fletcher;
	ztest_func_t ztest_fletcher_incr;
	ztest_func_t ztest_verify_dnode_bt;

	uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */
	uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */
	uint64_t zopt_often = 1ULL * NANOSEC; /* every second */
	uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */
	uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */

	#define ZTI_INIT(func, iters, interval) \
	{ .zi_func = (func), \
	.zi_iters = (iters), \
	.zi_interval = (interval), \
	.zi_funcname = # func }

	ztest_info_t ztest_info[] = {
	ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always),
	ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always),
	ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always),
	ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always),
	ZTI_INIT(ztest_zap, 30, &zopt_always),
	ZTI_INIT(ztest_zap_parallel, 100, &zopt_always),
	ZTI_INIT(ztest_split_pool, 1, &zopt_always),
	ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant),
	ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often),
	ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often),
	ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often),
	ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes),
	#if 0
	ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes),
	#endif
	ZTI_INIT(ztest_fzap, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes),
	ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes),
	ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes),
	ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes),
	ZTI_INIT(ztest_reguid, 1, &zopt_rarely),
	ZTI_INIT(ztest_scrub, 1, &zopt_rarely),
	ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely),
	ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely),
	ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes),
	ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely),
	ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes),
	ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely),
	ZTI_INIT(ztest_initialize, 1, &zopt_sometimes),
	ZTI_INIT(ztest_trim, 1, &zopt_sometimes),
	ZTI_INIT(ztest_blake3, 1, &zopt_rarely),
	ZTI_INIT(ztest_fletcher, 1, &zopt_rarely),
	ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely),
	ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes),
	};

	#define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t))

	/*
	* The following struct is used to hold a list of uncalled commit callbacks.
	* The callbacks are ordered by txg number.
	*/
	typedef struct ztest_cb_list {
	kmutex_t zcl_callbacks_lock;
	list_t zcl_callbacks;
	} ztest_cb_list_t;

	/*
	* Stuff we need to share writably between parent and child.
	*/
	typedef struct ztest_shared {
	boolean_t zs_do_init;
	hrtime_t zs_proc_start;
	hrtime_t zs_proc_stop;
	hrtime_t zs_thread_start;
	hrtime_t zs_thread_stop;
	hrtime_t zs_thread_kill;
	uint64_t zs_enospc_count;
	uint64_t zs_vdev_next_leaf;
	uint64_t zs_vdev_aux;
	uint64_t zs_alloc;
	uint64_t zs_space;
	uint64_t zs_splits;
	uint64_t zs_mirrors;
	uint64_t zs_metaslab_sz;
	uint64_t zs_metaslab_df_alloc_threshold;
	uint64_t zs_guid;
	} ztest_shared_t;

	#define ID_PARALLEL -1ULL

	static char ztest_dev_template[] = "%s/%s.%llua";
	static char ztest_aux_template[] = "%s/%s.%s.%llu";
	ztest_shared_t *ztest_shared;

	static spa_t *ztest_spa = NULL;
	static ztest_ds_t *ztest_ds;

	static kmutex_t ztest_vdev_lock;
	static boolean_t ztest_device_removal_active = B_FALSE;
	static boolean_t ztest_pool_scrubbed = B_FALSE;
	static kmutex_t ztest_checkpoint_lock;

	/*
	* The ztest_name_lock protects the pool and dataset namespace used by
	* the individual tests. To modify the namespace, consumers must grab
	* this lock as writer. Grabbing the lock as reader will ensure that the
	* namespace does not change while the lock is held.
	*/
	static pthread_rwlock_t ztest_name_lock;

	static boolean_t ztest_dump_core = B_TRUE;
	static boolean_t ztest_exiting;

	/* Global commit callback list */
	static ztest_cb_list_t zcl;
	/* Commit cb delay */
	static uint64_t zc_min_txg_delay = UINT64_MAX;
	static int zc_cb_counter = 0;

	/*
	* Minimum number of commit callbacks that need to be registered for us to check
	* whether the minimum txg delay is acceptable.
	*/
	#define ZTEST_COMMIT_CB_MIN_REG 100

	/*
	* If a number of txgs equal to this threshold have been created after a commit
	* callback has been registered but not called, then we assume there is an
	* implementation bug.
	*/
	#define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000)

	enum ztest_object {
	ZTEST_META_DNODE = 0,
	ZTEST_DIROBJ,
	ZTEST_OBJECTS
	};

	static __attribute__((noreturn)) void usage(boolean_t requested);
	static int ztest_scrub_impl(spa_t *spa);

	/*
	* These libumem hooks provide a reasonable set of defaults for the allocator's
	* debugging facilities.
	*/
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

	const char *
	_umem_logging_init(void)
	{
	return ("fail,contents"); /* $UMEM_LOGGING setting */
	}

	static void
	dump_debug_buffer(void)
	{
	ssize_t ret __attribute__((unused));

	if (!ztest_opts.zo_dump_dbgmsg)
	return;

	/*
	* We use write() instead of printf() so that this function
	* is safe to call from a signal handler.
	*/
	ret = write(STDOUT_FILENO, "\n", 1);
	zfs_dbgmsg_print("ztest");
	}

	#define BACKTRACE_SZ 100

	static void sig_handler(int signo)
	{
	struct sigaction action;
	#if (__GLIBC__ && !__UCLIBC__) /* backtrace() is a GNU extension */
	int nptrs;
	void *buffer[BACKTRACE_SZ];

	nptrs = backtrace(buffer, BACKTRACE_SZ);
	backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO);
	#endif
	dump_debug_buffer();

	/*
	* Restore default action and re-raise signal so SIGSEGV and
	* SIGABRT can trigger a core dump.
	*/
	action.sa_handler = SIG_DFL;
	sigemptyset(&action.sa_mask);
	action.sa_flags = 0;
	(void) sigaction(signo, &action, NULL);
	raise(signo);
	}

	#define FATAL_MSG_SZ 1024

	static const char *fatal_msg;

	static __attribute__((format(printf, 2, 3))) __attribute__((noreturn)) void
	fatal(int do_perror, const char *message, ...)
	{
	va_list args;
	int save_errno = errno;
	char *buf;

	(void) fflush(stdout);
	buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL);
	if (buf == NULL)
	goto out;

	va_start(args, message);
	(void) sprintf(buf, "ztest: ");
	/* LINTED */
	(void) vsprintf(buf + strlen(buf), message, args);
	va_end(args);
	if (do_perror) {
	(void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf),
	": %s", strerror(save_errno));
	}
	(void) fprintf(stderr, "%s\n", buf);
	fatal_msg = buf; /* to ease debugging */

	out:
	if (ztest_dump_core)
	abort();
	else
	dump_debug_buffer();

	exit(3);
	}

	static int
	str2shift(const char *buf)
	{
	const char *ends = "BKMGTPEZ";
	int i;

	if (buf[0] == '\0')
	return (0);
	for (i = 0; i < strlen(ends); i++) {
	if (toupper(buf[0]) == ends[i])
	break;
	}
	if (i == strlen(ends)) {
	(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n",
	buf);
	usage(B_FALSE);
	}
	if (buf[1] == '\0' \|\| (toupper(buf[1]) == 'B' && buf[2] == '\0')) {
	return (10*i);
	}
	(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf);
	usage(B_FALSE);
	}

	static uint64_t
	nicenumtoull(const char *buf)
	{
	char *end;
	uint64_t val;

	val = strtoull(buf, &end, 0);
	if (end == buf) {
	(void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf);
	usage(B_FALSE);
	} else if (end[0] == '.') {
	double fval = strtod(buf, &end);
	fval *= pow(2, str2shift(end));
	/*
	* UINT64_MAX is not exactly representable as a double.
	* The closest representation is UINT64_MAX + 1, so we
	* use a >= comparison instead of > for the bounds check.
	*/
	if (fval >= (double)UINT64_MAX) {
	(void) fprintf(stderr, "ztest: value too large: %s\n",
	buf);
	usage(B_FALSE);
	}
	val = (uint64_t)fval;
	} else {
	int shift = str2shift(end);
	if (shift >= 64 \|\| (val << shift) >> shift != val) {
	(void) fprintf(stderr, "ztest: value too large: %s\n",
	buf);
	usage(B_FALSE);
	}
	val <<= shift;
	}
	return (val);
	}

	typedef struct ztest_option {
	const char short_opt;
	const char *long_opt;
	const char *long_opt_param;
	const char *comment;
	unsigned int default_int;
	const char *default_str;
	} ztest_option_t;

	/*
	* The following option_table is used for generating the usage info as well as
	* the long and short option information for calling getopt_long().
	*/
	static ztest_option_t option_table[] = {
	{ 'v', "vdevs", "INTEGER", "Number of vdevs", DEFAULT_VDEV_COUNT,
	NULL},
	{ 's', "vdev-size", "INTEGER", "Size of each vdev",
	NO_DEFAULT, DEFAULT_VDEV_SIZE_STR},
	{ 'a', "alignment-shift", "INTEGER",
	"Alignment shift; use 0 for random", DEFAULT_ASHIFT, NULL},
	{ 'm', "mirror-copies", "INTEGER", "Number of mirror copies",
	DEFAULT_MIRRORS, NULL},
	{ 'r', "raid-disks", "INTEGER", "Number of raidz/draid disks",
	DEFAULT_RAID_CHILDREN, NULL},
	{ 'R', "raid-parity", "INTEGER", "Raid parity",
	DEFAULT_RAID_PARITY, NULL},
	{ 'K', "raid-kind", "raidz\|draid\|random", "Raid kind",
	NO_DEFAULT, "random"},
	{ 'D', "draid-data", "INTEGER", "Number of draid data drives",
	DEFAULT_DRAID_DATA, NULL},
	{ 'S', "draid-spares", "INTEGER", "Number of draid spares",
	DEFAULT_DRAID_SPARES, NULL},
	{ 'd', "datasets", "INTEGER", "Number of datasets",
	DEFAULT_DATASETS_COUNT, NULL},
	{ 't', "threads", "INTEGER", "Number of ztest threads",
	DEFAULT_THREADS, NULL},
	{ 'g', "gang-block-threshold", "INTEGER",
	"Metaslab gang block threshold",
	NO_DEFAULT, DEFAULT_FORCE_GANGING_STR},
	{ 'i', "init-count", "INTEGER", "Number of times to initialize pool",
	DEFAULT_INITS, NULL},
	{ 'k', "kill-percentage", "INTEGER", "Kill percentage",
	NO_DEFAULT, DEFAULT_KILLRATE_STR},
	{ 'p', "pool-name", "STRING", "Pool name",
	NO_DEFAULT, DEFAULT_POOL},
	{ 'f', "vdev-file-directory", "PATH", "File directory for vdev files",
	NO_DEFAULT, DEFAULT_VDEV_DIR},
	{ 'M', "multi-host", NULL,
	"Multi-host; simulate pool imported on remote host",
	NO_DEFAULT, NULL},
	{ 'E', "use-existing-pool", NULL,
	"Use existing pool instead of creating new one", NO_DEFAULT, NULL},
	{ 'T', "run-time", "INTEGER", "Total run time",
	NO_DEFAULT, DEFAULT_RUN_TIME_STR},
	{ 'P', "pass-time", "INTEGER", "Time per pass",
	NO_DEFAULT, DEFAULT_PASS_TIME_STR},
	{ 'F', "freeze-loops", "INTEGER", "Max loops in spa_freeze()",
	DEFAULT_MAX_LOOPS, NULL},
	{ 'B', "alt-ztest", "PATH", "Alternate ztest path",
	NO_DEFAULT, NULL},
	{ 'C', "vdev-class-state", "on\|off\|random", "vdev class state",
	NO_DEFAULT, "random"},
	{ 'o', "option", "\"OPTION=INTEGER\"",
	"Set global variable to an unsigned 32-bit integer value",
	NO_DEFAULT, NULL},
	{ 'G', "dump-debug-msg", NULL,
	"Dump zfs_dbgmsg buffer before exiting due to an error",
	NO_DEFAULT, NULL},
	{ 'V', "verbose", NULL,
	"Verbose (use multiple times for ever more verbosity)",
	NO_DEFAULT, NULL},
	{ 'h', "help", NULL, "Show this help",
	NO_DEFAULT, NULL},
	{0, 0, 0, 0, 0, 0}
	};

	static struct option *long_opts = NULL;
	static char *short_opts = NULL;

	static void
	init_options(void)
	{
	ASSERT3P(long_opts, ==, NULL);
	ASSERT3P(short_opts, ==, NULL);

	int count = sizeof (option_table) / sizeof (option_table[0]);
	long_opts = umem_alloc(sizeof (struct option) * count, UMEM_NOFAIL);

	short_opts = umem_alloc(sizeof (char) * 2 * count, UMEM_NOFAIL);
	int short_opt_index = 0;

	for (int i = 0; i < count; i++) {
	long_opts[i].val = option_table[i].short_opt;
	long_opts[i].name = option_table[i].long_opt;
	long_opts[i].has_arg = option_table[i].long_opt_param != NULL
	? required_argument : no_argument;
	long_opts[i].flag = NULL;
	short_opts[short_opt_index++] = option_table[i].short_opt;
	if (option_table[i].long_opt_param != NULL) {
	short_opts[short_opt_index++] = ':';
	}
	}
	}

	static void
	fini_options(void)
	{
	int count = sizeof (option_table) / sizeof (option_table[0]);

	umem_free(long_opts, sizeof (struct option) * count);
	umem_free(short_opts, sizeof (char) * 2 * count);

	long_opts = NULL;
	short_opts = NULL;
	}

	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	char option[80];
	FILE *fp = requested ? stdout : stderr;

	(void) fprintf(fp, "Usage: %s [OPTIONS...]\n", DEFAULT_POOL);
	for (int i = 0; option_table[i].short_opt != 0; i++) {
	if (option_table[i].long_opt_param != NULL) {
	(void) sprintf(option, " -%c --%s=%s",
	option_table[i].short_opt,
	option_table[i].long_opt,
	option_table[i].long_opt_param);
	} else {
	(void) sprintf(option, " -%c --%s",
	option_table[i].short_opt,
	option_table[i].long_opt);
	}
	(void) fprintf(fp, " %-40s%s", option,
	option_table[i].comment);

	if (option_table[i].long_opt_param != NULL) {
	if (option_table[i].default_str != NULL) {
	(void) fprintf(fp, " (default: %s)",
	option_table[i].default_str);
	} else if (option_table[i].default_int != NO_DEFAULT) {
	(void) fprintf(fp, " (default: %u)",
	option_table[i].default_int);
	}
	}
	(void) fprintf(fp, "\n");
	}
	exit(requested ? 0 : 1);
	}

	static uint64_t
	ztest_random(uint64_t range)
	{
	uint64_t r;

	ASSERT3S(ztest_fd_rand, >=, 0);

	if (range == 0)
	return (0);

	if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r))
	fatal(B_TRUE, "short read from /dev/urandom");

	return (r % range);
	}

	static void
	ztest_parse_name_value(const char input, ztest_shared_opts_t zo)
	{
	char name[32];
	char *value;
	int state = ZTEST_VDEV_CLASS_RND;

	(void) strlcpy(name, input, sizeof (name));

	value = strchr(name, '=');
	if (value == NULL) {
	(void) fprintf(stderr, "missing value in property=value "
	"'-C' argument (%s)\n", input);
	usage(B_FALSE);
	}
	*(value) = '\0';
	value++;

	if (strcmp(value, "on") == 0) {
	state = ZTEST_VDEV_CLASS_ON;
	} else if (strcmp(value, "off") == 0) {
	state = ZTEST_VDEV_CLASS_OFF;
	} else if (strcmp(value, "random") == 0) {
	state = ZTEST_VDEV_CLASS_RND;
	} else {
	(void) fprintf(stderr, "invalid property value '%s'\n", value);
	usage(B_FALSE);
	}

	if (strcmp(name, "special") == 0) {
	zo->zo_special_vdevs = state;
	} else {
	(void) fprintf(stderr, "invalid property name '%s'\n", name);
	usage(B_FALSE);
	}
	if (zo->zo_verbose >= 3)
	(void) printf("%s vdev state is '%s'\n", name, value);
	}

	static void
	process_options(int argc, char **argv)
	{
	char *path;
	ztest_shared_opts_t *zo = &ztest_opts;

	int opt;
	uint64_t value;
	const char *raid_kind = "random";

	memcpy(zo, &ztest_opts_defaults, sizeof (*zo));

	init_options();

	while ((opt = getopt_long(argc, argv, short_opts, long_opts,
	NULL)) != EOF) {
	value = 0;
	switch (opt) {
	case 'v':
	case 's':
	case 'a':
	case 'm':
	case 'r':
	case 'R':
	case 'D':
	case 'S':
	case 'd':
	case 't':
	case 'g':
	case 'i':
	case 'k':
	case 'T':
	case 'P':
	case 'F':
	value = nicenumtoull(optarg);
	}
	switch (opt) {
	case 'v':
	zo->zo_vdevs = value;
	break;
	case 's':
	zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value);
	break;
	case 'a':
	zo->zo_ashift = value;
	break;
	case 'm':
	zo->zo_mirrors = value;
	break;
	case 'r':
	zo->zo_raid_children = MAX(1, value);
	break;
	case 'R':
	zo->zo_raid_parity = MIN(MAX(value, 1), 3);
	break;
	case 'K':
	raid_kind = optarg;
	break;
	case 'D':
	zo->zo_draid_data = MAX(1, value);
	break;
	case 'S':
	zo->zo_draid_spares = MAX(1, value);
	break;
	case 'd':
	zo->zo_datasets = MAX(1, value);
	break;
	case 't':
	zo->zo_threads = MAX(1, value);
	break;
	case 'g':
	zo->zo_metaslab_force_ganging =
	MAX(SPA_MINBLOCKSIZE << 1, value);
	break;
	case 'i':
	zo->zo_init = value;
	break;
	case 'k':
	zo->zo_killrate = value;
	break;
	case 'p':
	(void) strlcpy(zo->zo_pool, optarg,
	sizeof (zo->zo_pool));
	break;
	case 'f':
	path = realpath(optarg, NULL);
	if (path == NULL) {
	(void) fprintf(stderr, "error: %s: %s\n",
	optarg, strerror(errno));
	usage(B_FALSE);
	} else {
	(void) strlcpy(zo->zo_dir, path,
	sizeof (zo->zo_dir));
	free(path);
	}
	break;
	case 'M':
	zo->zo_mmp_test = 1;
	break;
	case 'V':
	zo->zo_verbose++;
	break;
	case 'E':
	zo->zo_init = 0;
	break;
	case 'T':
	zo->zo_time = value;
	break;
	case 'P':
	zo->zo_passtime = MAX(1, value);
	break;
	case 'F':
	zo->zo_maxloops = MAX(1, value);
	break;
	case 'B':
	(void) strlcpy(zo->zo_alt_ztest, optarg,
	sizeof (zo->zo_alt_ztest));
	break;
	case 'C':
	ztest_parse_name_value(optarg, zo);
	break;
	case 'o':
	if (zo->zo_gvars_count >= ZO_GVARS_MAX_COUNT) {
	(void) fprintf(stderr,
	"max global var count (%zu) exceeded\n",
	ZO_GVARS_MAX_COUNT);
	usage(B_FALSE);
	}
	char *v = zo->zo_gvars[zo->zo_gvars_count];
	if (strlcpy(v, optarg, ZO_GVARS_MAX_ARGLEN) >=
	ZO_GVARS_MAX_ARGLEN) {
	(void) fprintf(stderr,
	"global var option '%s' is too long\n",
	optarg);
	usage(B_FALSE);
	}
	zo->zo_gvars_count++;
	break;
	case 'G':
	zo->zo_dump_dbgmsg = 1;
	break;
	case 'h':
	usage(B_TRUE);
	break;
	case '?':
	default:
	usage(B_FALSE);
	break;
	}
	}

	fini_options();

	/* When raid choice is 'random' add a draid pool 50% of the time */
	if (strcmp(raid_kind, "random") == 0) {
	raid_kind = (ztest_random(2) == 0) ? "draid" : "raidz";

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("choosing RAID type '%s'\n", raid_kind);
	}

	if (strcmp(raid_kind, "draid") == 0) {
	uint64_t min_devsize;

	/* With fewer disk use 256M, otherwise 128M is OK */
	min_devsize = (ztest_opts.zo_raid_children < 16) ?
	(256ULL << 20) : (128ULL << 20);

	/* No top-level mirrors with dRAID for now */
	zo->zo_mirrors = 0;

	/* Use more appropriate defaults for dRAID */
	if (zo->zo_vdevs == ztest_opts_defaults.zo_vdevs)
	zo->zo_vdevs = 1;
	if (zo->zo_raid_children ==
	ztest_opts_defaults.zo_raid_children)
	zo->zo_raid_children = 16;
	if (zo->zo_ashift < 12)
	zo->zo_ashift = 12;
	if (zo->zo_vdev_size < min_devsize)
	zo->zo_vdev_size = min_devsize;

	if (zo->zo_draid_data + zo->zo_raid_parity >
	zo->zo_raid_children - zo->zo_draid_spares) {
	(void) fprintf(stderr, "error: too few draid "
	"children (%d) for stripe width (%d)\n",
	zo->zo_raid_children,
	zo->zo_draid_data + zo->zo_raid_parity);
	usage(B_FALSE);
	}

	(void) strlcpy(zo->zo_raid_type, VDEV_TYPE_DRAID,
	sizeof (zo->zo_raid_type));

	} else /* using raidz */ {
	ASSERT0(strcmp(raid_kind, "raidz"));

	zo->zo_raid_parity = MIN(zo->zo_raid_parity,
	zo->zo_raid_children - 1);
	}

	zo->zo_vdevtime =
	(zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs :
	UINT64_MAX >> 2);

	if (*zo->zo_alt_ztest) {
	const char *invalid_what = "ztest";
	char *val = zo->zo_alt_ztest;
	if (0 != access(val, X_OK) \|\|
	(strrchr(val, '/') == NULL && (errno = EINVAL)))
	goto invalid;

	int dirlen = strrchr(val, '/') - val;
	- strncpy(zo->zo_alt_libpath, val, dirlen);
	+ strlcpy(zo->zo_alt_libpath, val,
	+ MIN(sizeof (zo->zo_alt_libpath), dirlen + 1));
	invalid_what = "library path", val = zo->zo_alt_libpath;
	if (strrchr(val, '/') == NULL && (errno = EINVAL))
	goto invalid;
	*strrchr(val, '/') = '\0';
	strlcat(val, "/lib", sizeof (zo->zo_alt_libpath));

	if (0 != access(zo->zo_alt_libpath, X_OK))
	goto invalid;
	return;

	invalid:
	ztest_dump_core = B_FALSE;
	fatal(B_TRUE, "invalid alternate %s %s", invalid_what, val);
	}
	}

	static void
	ztest_kill(ztest_shared_t *zs)
	{
	zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa));
	zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa));

	/*
	* Before we kill ourselves, make sure that the config is updated.
	* See comment above spa_write_cachefile().
	*/
	mutex_enter(&spa_namespace_lock);
	- spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE);
	+ spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE, B_FALSE);
	mutex_exit(&spa_namespace_lock);

	(void) raise(SIGKILL);
	}

	static void
	ztest_record_enospc(const char *s)
	{
	(void) s;
	ztest_shared->zs_enospc_count++;
	}

	static uint64_t
	ztest_get_ashift(void)
	{
	if (ztest_opts.zo_ashift == 0)
	return (SPA_MINBLOCKSHIFT + ztest_random(5));
	return (ztest_opts.zo_ashift);
	}

	static boolean_t
	ztest_is_draid_spare(const char *name)
	{
	uint64_t spare_id = 0, parity = 0, vdev_id = 0;

	if (sscanf(name, VDEV_TYPE_DRAID "%"PRIu64"-%"PRIu64"-%"PRIu64"",
	&parity, &vdev_id, &spare_id) == 3) {
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static nvlist_t *
	make_vdev_file(const char path, const char aux, const char *pool,
	size_t size, uint64_t ashift)
	{
	char *pathbuf = NULL;
	uint64_t vdev;
	nvlist_t *file;
	boolean_t draid_spare = B_FALSE;


	if (ashift == 0)
	ashift = ztest_get_ashift();

	if (path == NULL) {
	pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	path = pathbuf;

	if (aux != NULL) {
	vdev = ztest_shared->zs_vdev_aux;
	(void) snprintf(pathbuf, MAXPATHLEN,
	ztest_aux_template, ztest_opts.zo_dir,
	pool == NULL ? ztest_opts.zo_pool : pool,
	aux, vdev);
	} else {
	vdev = ztest_shared->zs_vdev_next_leaf++;
	(void) snprintf(pathbuf, MAXPATHLEN,
	ztest_dev_template, ztest_opts.zo_dir,
	pool == NULL ? ztest_opts.zo_pool : pool, vdev);
	}
	} else {
	draid_spare = ztest_is_draid_spare(path);
	}

	if (size != 0 && !draid_spare) {
	int fd = open(path, O_RDWR \| O_CREAT \| O_TRUNC, 0666);
	if (fd == -1)
	fatal(B_TRUE, "can't open %s", path);
	if (ftruncate(fd, size) != 0)
	fatal(B_TRUE, "can't ftruncate %s", path);
	(void) close(fd);
	}

	file = fnvlist_alloc();
	fnvlist_add_string(file, ZPOOL_CONFIG_TYPE,
	draid_spare ? VDEV_TYPE_DRAID_SPARE : VDEV_TYPE_FILE);
	fnvlist_add_string(file, ZPOOL_CONFIG_PATH, path);
	fnvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift);
	umem_free(pathbuf, MAXPATHLEN);

	return (file);
	}

	static nvlist_t *
	make_vdev_raid(const char path, const char aux, const char *pool, size_t size,
	uint64_t ashift, int r)
	{
	nvlist_t raid, *child;
	int c;

	if (r < 2)
	return (make_vdev_file(path, aux, pool, size, ashift));
	child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < r; c++)
	child[c] = make_vdev_file(path, aux, pool, size, ashift);

	raid = fnvlist_alloc();
	fnvlist_add_string(raid, ZPOOL_CONFIG_TYPE,
	ztest_opts.zo_raid_type);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_NPARITY,
	ztest_opts.zo_raid_parity);
	fnvlist_add_nvlist_array(raid, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, r);

	if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) {
	uint64_t ndata = ztest_opts.zo_draid_data;
	uint64_t nparity = ztest_opts.zo_raid_parity;
	uint64_t nspares = ztest_opts.zo_draid_spares;
	uint64_t children = ztest_opts.zo_raid_children;
	uint64_t ngroups = 1;

	/*
	* Calculate the minimum number of groups required to fill a
	* slice. This is the LCM of the stripe width (data + parity)
	* and the number of data drives (children - spares).
	*/
	while (ngroups * (ndata + nparity) % (children - nspares) != 0)
	ngroups++;

	/* Store the basic dRAID configuration. */
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NDATA, ndata);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NSPARES, nspares);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NGROUPS, ngroups);
	}

	for (c = 0; c < r; c++)
	fnvlist_free(child[c]);

	umem_free(child, r * sizeof (nvlist_t *));

	return (raid);
	}

	static nvlist_t *
	make_vdev_mirror(const char path, const char aux, const char *pool,
	size_t size, uint64_t ashift, int r, int m)
	{
	nvlist_t mirror, *child;
	int c;

	if (m < 1)
	return (make_vdev_raid(path, aux, pool, size, ashift, r));

	child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < m; c++)
	child[c] = make_vdev_raid(path, aux, pool, size, ashift, r);

	mirror = fnvlist_alloc();
	fnvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, VDEV_TYPE_MIRROR);
	fnvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, m);

	for (c = 0; c < m; c++)
	fnvlist_free(child[c]);

	umem_free(child, m * sizeof (nvlist_t *));

	return (mirror);
	}

	static nvlist_t *
	make_vdev_root(const char path, const char aux, const char *pool, size_t size,
	uint64_t ashift, const char *class, int r, int m, int t)
	{
	nvlist_t root, *child;
	int c;
	boolean_t log;

	ASSERT3S(t, >, 0);

	log = (class != NULL && strcmp(class, "log") == 0);

	child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < t; c++) {
	child[c] = make_vdev_mirror(path, aux, pool, size, ashift,
	r, m);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, log);

	if (class != NULL && class[0] != '\0') {
	ASSERT(m > 1 \|\| log); /* expecting a mirror */
	fnvlist_add_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, class);
	}
	}

	root = fnvlist_alloc();
	fnvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
	fnvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, t);

	for (c = 0; c < t; c++)
	fnvlist_free(child[c]);

	umem_free(child, t * sizeof (nvlist_t *));

	return (root);
	}

	/*
	* Find a random spa version. Returns back a random spa version in the
	* range [initial_version, SPA_VERSION_FEATURES].
	*/
	static uint64_t
	ztest_random_spa_version(uint64_t initial_version)
	{
	uint64_t version = initial_version;

	if (version <= SPA_VERSION_BEFORE_FEATURES) {
	version = version +
	ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1);
	}

	if (version > SPA_VERSION_BEFORE_FEATURES)
	version = SPA_VERSION_FEATURES;

	ASSERT(SPA_VERSION_IS_SUPPORTED(version));
	return (version);
	}

	static int
	ztest_random_blocksize(void)
	{
	ASSERT3U(ztest_spa->spa_max_ashift, !=, 0);

	/*
	* Choose a block size >= the ashift.
	* If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
	*/
	int maxbs = SPA_OLD_MAXBLOCKSHIFT;
	if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE)
	maxbs = 20;
	uint64_t block_shift =
	ztest_random(maxbs - ztest_spa->spa_max_ashift + 1);
	return (1 << (SPA_MINBLOCKSHIFT + block_shift));
	}

	static int
	ztest_random_dnodesize(void)
	{
	int slots;
	int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT;

	if (max_slots == DNODE_MIN_SLOTS)
	return (DNODE_MIN_SIZE);

	/*
	* Weight the random distribution more heavily toward smaller
	* dnode sizes since that is more likely to reflect real-world
	* usage.
	*/
	ASSERT3U(max_slots, >, 4);
	switch (ztest_random(10)) {
	case 0:
	slots = 5 + ztest_random(max_slots - 4);
	break;
	case 1 ... 4:
	slots = 2 + ztest_random(3);
	break;
	default:
	slots = 1;
	break;
	}

	return (slots << DNODE_SHIFT);
	}

	static int
	ztest_random_ibshift(void)
	{
	return (DN_MIN_INDBLKSHIFT +
	ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1));
	}

	static uint64_t
	ztest_random_vdev_top(spa_t *spa, boolean_t log_ok)
	{
	uint64_t top;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t *tvd;

	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	do {
	top = ztest_random(rvd->vdev_children);
	tvd = rvd->vdev_child[top];
	} while (!vdev_is_concrete(tvd) \|\| (tvd->vdev_islog && !log_ok) \|\|
	tvd->vdev_mg == NULL \|\| tvd->vdev_mg->mg_class == NULL);

	return (top);
	}

	static uint64_t
	ztest_random_dsl_prop(zfs_prop_t prop)
	{
	uint64_t value;

	do {
	value = zfs_prop_random_value(prop, ztest_random(-1ULL));
	} while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF);

	return (value);
	}

	static int
	ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value,
	boolean_t inherit)
	{
	const char *propname = zfs_prop_to_name(prop);
	const char *valname;
	char *setpoint;
	uint64_t curval;
	int error;

	error = dsl_prop_set_int(osname, propname,
	(inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value);

	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT0(error);

	setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint));

	if (ztest_opts.zo_verbose >= 6) {
	int err;

	err = zfs_prop_index_to_string(prop, curval, &valname);
	if (err)
	(void) printf("%s %s = %llu at '%s'\n", osname,
	propname, (unsigned long long)curval, setpoint);
	else
	(void) printf("%s %s = %s at '%s'\n",
	osname, propname, valname, setpoint);
	}
	umem_free(setpoint, MAXPATHLEN);

	return (error);
	}

	static int
	ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value)
	{
	spa_t *spa = ztest_spa;
	nvlist_t *props = NULL;
	int error;

	props = fnvlist_alloc();
	fnvlist_add_uint64(props, zpool_prop_to_name(prop), value);

	error = spa_prop_set(spa, props);

	fnvlist_free(props);

	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT0(error);

	return (error);
	}

	static int
	ztest_dmu_objset_own(const char *name, dmu_objset_type_t type,
	boolean_t readonly, boolean_t decrypt, const void tag, objset_t *osp)
	{
	int err;
	char *cp = NULL;
	char ddname[ZFS_MAX_DATASET_NAME_LEN];

	- strcpy(ddname, name);
	+ strlcpy(ddname, name, sizeof (ddname));
	cp = strchr(ddname, '@');
	if (cp != NULL)
	*cp = '\0';

	err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
	while (decrypt && err == EACCES) {
	dsl_crypto_params_t *dcp;
	nvlist_t *crypto_args = fnvlist_alloc();

	fnvlist_add_uint8_array(crypto_args, "wkeydata",
	(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);
	VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
	crypto_args, &dcp));
	err = spa_keystore_load_wkey(ddname, dcp, B_FALSE);
	/*
	* Note: if there was an error loading, the wkey was not
	* consumed, and needs to be freed.
	*/
	dsl_crypto_params_free(dcp, (err != 0));
	fnvlist_free(crypto_args);

	if (err == EINVAL) {
	/*
	* We couldn't load a key for this dataset so try
	* the parent. This loop will eventually hit the
	* encryption root since ztest only makes clones
	* as children of their origin datasets.
	*/
	cp = strrchr(ddname, '/');
	if (cp == NULL)
	return (err);

	*cp = '\0';
	err = EACCES;
	continue;
	} else if (err != 0) {
	break;
	}

	err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
	break;
	}

	return (err);
	}

	static void
	ztest_rll_init(rll_t *rll)
	{
	rll->rll_writer = NULL;
	rll->rll_readers = 0;
	mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL);
	}

	static void
	ztest_rll_destroy(rll_t *rll)
	{
	ASSERT3P(rll->rll_writer, ==, NULL);
	ASSERT0(rll->rll_readers);
	mutex_destroy(&rll->rll_lock);
	cv_destroy(&rll->rll_cv);
	}

	static void
	ztest_rll_lock(rll_t *rll, rl_type_t type)
	{
	mutex_enter(&rll->rll_lock);

	if (type == RL_READER) {
	while (rll->rll_writer != NULL)
	(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
	rll->rll_readers++;
	} else {
	while (rll->rll_writer != NULL \|\| rll->rll_readers)
	(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
	rll->rll_writer = curthread;
	}

	mutex_exit(&rll->rll_lock);
	}

	static void
	ztest_rll_unlock(rll_t *rll)
	{
	mutex_enter(&rll->rll_lock);

	if (rll->rll_writer) {
	ASSERT0(rll->rll_readers);
	rll->rll_writer = NULL;
	} else {
	ASSERT3S(rll->rll_readers, >, 0);
	ASSERT3P(rll->rll_writer, ==, NULL);
	rll->rll_readers--;
	}

	if (rll->rll_writer == NULL && rll->rll_readers == 0)
	cv_broadcast(&rll->rll_cv);

	mutex_exit(&rll->rll_lock);
	}

	static void
	ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type)
	{
	rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];

	ztest_rll_lock(rll, type);
	}

	static void
	ztest_object_unlock(ztest_ds_t *zd, uint64_t object)
	{
	rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];

	ztest_rll_unlock(rll);
	}

	static rl_t *
	ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset,
	uint64_t size, rl_type_t type)
	{
	uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1));
	rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)];
	rl_t *rl;

	rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL);
	rl->rl_object = object;
	rl->rl_offset = offset;
	rl->rl_size = size;
	rl->rl_lock = rll;

	ztest_rll_lock(rll, type);

	return (rl);
	}

	static void
	ztest_range_unlock(rl_t *rl)
	{
	rll_t *rll = rl->rl_lock;

	ztest_rll_unlock(rll);

	umem_free(rl, sizeof (*rl));
	}

	static void
	ztest_zd_init(ztest_ds_t zd, ztest_shared_ds_t szd, objset_t *os)
	{
	zd->zd_os = os;
	zd->zd_zilog = dmu_objset_zil(os);
	zd->zd_shared = szd;
	dmu_objset_name(os, zd->zd_name);
	int l;

	if (zd->zd_shared != NULL)
	zd->zd_shared->zd_seq = 0;

	VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL));
	mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL);

	for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
	ztest_rll_init(&zd->zd_object_lock[l]);

	for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
	ztest_rll_init(&zd->zd_range_lock[l]);
	}

	static void
	ztest_zd_fini(ztest_ds_t *zd)
	{
	int l;

	mutex_destroy(&zd->zd_dirobj_lock);
	(void) pthread_rwlock_destroy(&zd->zd_zilog_lock);

	for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
	ztest_rll_destroy(&zd->zd_object_lock[l]);

	for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
	ztest_rll_destroy(&zd->zd_range_lock[l]);
	}

	#define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)

	static uint64_t
	ztest_tx_assign(dmu_tx_t tx, uint64_t txg_how, const char tag)
	{
	uint64_t txg;
	int error;

	/*
	* Attempt to assign tx to some transaction group.
	*/
	error = dmu_tx_assign(tx, txg_how);
	if (error) {
	if (error == ERESTART) {
	ASSERT3U(txg_how, ==, TXG_NOWAIT);
	dmu_tx_wait(tx);
	} else {
	ASSERT3U(error, ==, ENOSPC);
	ztest_record_enospc(tag);
	}
	dmu_tx_abort(tx);
	return (0);
	}
	txg = dmu_tx_get_txg(tx);
	ASSERT3U(txg, !=, 0);
	return (txg);
	}

	static void
	ztest_bt_generate(ztest_block_tag_t bt, objset_t os, uint64_t object,
	uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
	uint64_t crtxg)
	{
	bt->bt_magic = BT_MAGIC;
	bt->bt_objset = dmu_objset_id(os);
	bt->bt_object = object;
	bt->bt_dnodesize = dnodesize;
	bt->bt_offset = offset;
	bt->bt_gen = gen;
	bt->bt_txg = txg;
	bt->bt_crtxg = crtxg;
	}

	static void
	ztest_bt_verify(ztest_block_tag_t bt, objset_t os, uint64_t object,
	uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
	uint64_t crtxg)
	{
	ASSERT3U(bt->bt_magic, ==, BT_MAGIC);
	ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os));
	ASSERT3U(bt->bt_object, ==, object);
	ASSERT3U(bt->bt_dnodesize, ==, dnodesize);
	ASSERT3U(bt->bt_offset, ==, offset);
	ASSERT3U(bt->bt_gen, <=, gen);
	ASSERT3U(bt->bt_txg, <=, txg);
	ASSERT3U(bt->bt_crtxg, ==, crtxg);
	}

	static ztest_block_tag_t *
	ztest_bt_bonus(dmu_buf_t *db)
	{
	dmu_object_info_t doi;
	ztest_block_tag_t *bt;

	dmu_object_info_from_db(db, &doi);
	ASSERT3U(doi.doi_bonus_size, <=, db->db_size);
	ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt));
	bt = (void )((char )db->db_data + doi.doi_bonus_size - sizeof (*bt));

	return (bt);
	}

	/*
	* Generate a token to fill up unused bonus buffer space. Try to make
	* it unique to the object, generation, and offset to verify that data
	* is not getting overwritten by data from other dnodes.
	*/
	#define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \
	(((ds) << 48) \| ((gen) << 32) \| ((obj) << 8) \| (offset))

	/*
	* Fill up the unused bonus buffer region before the block tag with a
	* verifiable pattern. Filling the whole bonus area with non-zero data
	* helps ensure that all dnode traversal code properly skips the
	* interior regions of large dnodes.
	*/
	static void
	ztest_fill_unused_bonus(dmu_buf_t db, void end, uint64_t obj,
	objset_t *os, uint64_t gen)
	{
	uint64_t *bonusp;

	ASSERT(IS_P2ALIGNED((char )end - (char )db->db_data, 8));

	for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
	uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
	gen, bonusp - (uint64_t *)db->db_data);
	*bonusp = token;
	}
	}

	/*
	* Verify that the unused area of a bonus buffer is filled with the
	* expected tokens.
	*/
	static void
	ztest_verify_unused_bonus(dmu_buf_t db, void end, uint64_t obj,
	objset_t *os, uint64_t gen)
	{
	uint64_t *bonusp;

	for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
	uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
	gen, bonusp - (uint64_t *)db->db_data);
	VERIFY3U(*bonusp, ==, token);
	}
	}

	/*
	* ZIL logging ops
	*/

	#define lrz_type lr_mode
	#define lrz_blocksize lr_uid
	#define lrz_ibshift lr_gid
	#define lrz_bonustype lr_rdev
	#define lrz_dnodesize lr_crtime[1]

	static void
	ztest_log_create(ztest_ds_t zd, dmu_tx_t tx, lr_create_t *lr)
	{
	char name = (void )(lr + 1); /* name follows lr */
	size_t namesize = strlen(name) + 1;
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize);
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) + namesize - sizeof (lr_t));

	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_remove(ztest_ds_t zd, dmu_tx_t tx, lr_remove_t *lr, uint64_t object)
	{
	char name = (void )(lr + 1); /* name follows lr */
	size_t namesize = strlen(name) + 1;
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize);
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) + namesize - sizeof (lr_t));

	itx->itx_oid = object;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_write(ztest_ds_t zd, dmu_tx_t tx, lr_write_t *lr)
	{
	itx_t *itx;
	itx_wr_state_t write_state = ztest_random(WR_NUM_STATES);

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	if (lr->lr_length > zil_max_log_data(zd->zd_zilog))
	write_state = WR_INDIRECT;

	itx = zil_itx_create(TX_WRITE,
	sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0));

	if (write_state == WR_COPIED &&
	dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length,
	((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) {
	zil_itx_destroy(itx);
	itx = zil_itx_create(TX_WRITE, sizeof (*lr));
	write_state = WR_NEED_COPY;
	}
	itx->itx_private = zd;
	itx->itx_wr_state = write_state;
	itx->itx_sync = (ztest_random(8) == 0);

	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_truncate(ztest_ds_t zd, dmu_tx_t tx, lr_truncate_t *lr)
	{
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	itx->itx_sync = B_FALSE;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_setattr(ztest_ds_t zd, dmu_tx_t tx, lr_setattr_t *lr)
	{
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_SETATTR, sizeof (*lr));
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	itx->itx_sync = B_FALSE;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	/*
	* ZIL replay ops
	*/
	static int
	ztest_replay_create(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_create_t *lr = arg2;
	char name = (void )(lr + 1); /* name follows lr */
	objset_t *os = zd->zd_os;
	ztest_block_tag_t *bbt;
	dmu_buf_t *db;
	dmu_tx_t *tx;
	uint64_t txg;
	int error = 0;
	int bonuslen;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
	ASSERT3S(name[0], !=, '\0');

	tx = dmu_tx_create(os);

	dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name);

	if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
	} else {
	dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
	}

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0)
	return (ENOSPC);

	ASSERT3U(dmu_objset_zil(os)->zl_replay, ==, !!lr->lr_foid);
	bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize);

	if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
	if (lr->lr_foid == 0) {
	lr->lr_foid = zap_create_dnsize(os,
	lr->lrz_type, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	} else {
	error = zap_create_claim_dnsize(os, lr->lr_foid,
	lr->lrz_type, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	}
	} else {
	if (lr->lr_foid == 0) {
	lr->lr_foid = dmu_object_alloc_dnsize(os,
	lr->lrz_type, 0, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	} else {
	error = dmu_object_claim_dnsize(os, lr->lr_foid,
	lr->lrz_type, 0, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	}
	}

	if (error) {
	ASSERT3U(error, ==, EEXIST);
	ASSERT(zd->zd_zilog->zl_replay);
	dmu_tx_commit(tx);
	return (error);
	}

	ASSERT3U(lr->lr_foid, !=, 0);

	if (lr->lrz_type != DMU_OT_ZAP_OTHER)
	VERIFY0(dmu_object_set_blocksize(os, lr->lr_foid,
	lr->lrz_blocksize, lr->lrz_ibshift, tx));

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
	bbt = ztest_bt_bonus(db);
	dmu_buf_will_dirty(db, tx);
	ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL,
	lr->lr_gen, txg, txg);
	ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen);
	dmu_buf_rele(db, FTAG);

	VERIFY0(zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1,
	&lr->lr_foid, tx));

	(void) ztest_log_create(zd, tx, lr);

	dmu_tx_commit(tx);

	return (0);
	}

	static int
	ztest_replay_remove(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_remove_t *lr = arg2;
	char name = (void )(lr + 1); /* name follows lr */
	objset_t *os = zd->zd_os;
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	uint64_t object, txg;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
	ASSERT3S(name[0], !=, '\0');

	VERIFY0(
	zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object));
	ASSERT3U(object, !=, 0);

	ztest_object_lock(zd, object, RL_WRITER);

	VERIFY0(dmu_object_info(os, object, &doi));

	tx = dmu_tx_create(os);

	dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name);
	dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	ztest_object_unlock(zd, object);
	return (ENOSPC);
	}

	if (doi.doi_type == DMU_OT_ZAP_OTHER) {
	VERIFY0(zap_destroy(os, object, tx));
	} else {
	VERIFY0(dmu_object_free(os, object, tx));
	}

	VERIFY0(zap_remove(os, lr->lr_doid, name, tx));

	(void) ztest_log_remove(zd, tx, lr, object);

	dmu_tx_commit(tx);

	ztest_object_unlock(zd, object);

	return (0);
	}

	static int
	ztest_replay_write(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_write_t *lr = arg2;
	objset_t *os = zd->zd_os;
	void data = lr + 1; / data follows lr */
	uint64_t offset, length;
	ztest_block_tag_t *bt = data;
	ztest_block_tag_t *bbt;
	uint64_t gen, txg, lrtxg, crtxg;
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	dmu_buf_t *db;
	arc_buf_t *abuf = NULL;
	rl_t *rl;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	offset = lr->lr_offset;
	length = lr->lr_length;

	/* If it's a dmu_sync() block, write the whole block */
	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
	uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
	if (length < blocksize) {
	offset -= offset % blocksize;
	length = blocksize;
	}
	}

	if (bt->bt_magic == BSWAP_64(BT_MAGIC))
	byteswap_uint64_array(bt, sizeof (*bt));

	if (bt->bt_magic != BT_MAGIC)
	bt = NULL;

	ztest_object_lock(zd, lr->lr_foid, RL_READER);
	rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER);

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));

	dmu_object_info_from_db(db, &doi);

	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	gen = bbt->bt_gen;
	crtxg = bbt->bt_crtxg;
	lrtxg = lr->lr_common.lrc_txg;

	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, lr->lr_foid, offset, length);

	if (ztest_random(8) == 0 && length == doi.doi_data_block_size &&
	P2PHASE(offset, length) == 0)
	abuf = dmu_request_arcbuf(db, length);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	if (abuf != NULL)
	dmu_return_arcbuf(abuf);
	dmu_buf_rele(db, FTAG);
	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	if (bt != NULL) {
	/*
	* Usually, verify the old data before writing new data --
	* but not always, because we also want to verify correct
	* behavior when the data was not recently read into cache.
	*/
	ASSERT0(offset % doi.doi_data_block_size);
	if (ztest_random(4) != 0) {
	int prefetch = ztest_random(2) ?
	DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH;
	ztest_block_tag_t rbt;

	VERIFY(dmu_read(os, lr->lr_foid, offset,
	sizeof (rbt), &rbt, prefetch) == 0);
	if (rbt.bt_magic == BT_MAGIC) {
	ztest_bt_verify(&rbt, os, lr->lr_foid, 0,
	offset, gen, txg, crtxg);
	}
	}

	/*
	* Writes can appear to be newer than the bonus buffer because
	* the ztest_get_data() callback does a dmu_read() of the
	* open-context data, which may be different than the data
	* as it was when the write was generated.
	*/
	if (zd->zd_zilog->zl_replay) {
	ztest_bt_verify(bt, os, lr->lr_foid, 0, offset,
	MAX(gen, bt->bt_gen), MAX(txg, lrtxg),
	bt->bt_crtxg);
	}

	/*
	* Set the bt's gen/txg to the bonus buffer's gen/txg
	* so that all of the usual ASSERTs will work.
	*/
	ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg,
	crtxg);
	}

	if (abuf == NULL) {
	dmu_write(os, lr->lr_foid, offset, length, data, tx);
	} else {
	memcpy(abuf->b_data, data, length);
	- dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx);
	+ VERIFY0(dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx));
	}

	(void) ztest_log_write(zd, tx, lr);

	dmu_buf_rele(db, FTAG);

	dmu_tx_commit(tx);

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	static int
	ztest_replay_truncate(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_truncate_t *lr = arg2;
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	uint64_t txg;
	rl_t *rl;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ztest_object_lock(zd, lr->lr_foid, RL_READER);
	rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length,
	RL_WRITER);

	tx = dmu_tx_create(os);

	dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	VERIFY0(dmu_free_range(os, lr->lr_foid, lr->lr_offset,
	lr->lr_length, tx));

	(void) ztest_log_truncate(zd, tx, lr);

	dmu_tx_commit(tx);

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	static int
	ztest_replay_setattr(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_setattr_t *lr = arg2;
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	dmu_buf_t *db;
	ztest_block_tag_t *bbt;
	uint64_t txg, lrtxg, crtxg, dnodesize;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ztest_object_lock(zd, lr->lr_foid, RL_WRITER);

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));

	tx = dmu_tx_create(os);
	dmu_tx_hold_bonus(tx, lr->lr_foid);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	crtxg = bbt->bt_crtxg;
	lrtxg = lr->lr_common.lrc_txg;
	dnodesize = bbt->bt_dnodesize;

	if (zd->zd_zilog->zl_replay) {
	ASSERT3U(lr->lr_size, !=, 0);
	ASSERT3U(lr->lr_mode, !=, 0);
	ASSERT3U(lrtxg, !=, 0);
	} else {
	/*
	* Randomly change the size and increment the generation.
	*/
	lr->lr_size = (ztest_random(db->db_size / sizeof (bbt)) + 1)
	sizeof (*bbt);
	lr->lr_mode = bbt->bt_gen + 1;
	ASSERT0(lrtxg);
	}

	/*
	* Verify that the current bonus buffer is not newer than our txg.
	*/
	ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
	MAX(txg, lrtxg), crtxg);

	dmu_buf_will_dirty(db, tx);

	ASSERT3U(lr->lr_size, >=, sizeof (*bbt));
	ASSERT3U(lr->lr_size, <=, db->db_size);
	VERIFY0(dmu_set_bonus(db, lr->lr_size, tx));
	bbt = ztest_bt_bonus(db);

	ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
	txg, crtxg);
	ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen);
	dmu_buf_rele(db, FTAG);

	(void) ztest_log_setattr(zd, tx, lr);

	dmu_tx_commit(tx);

	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = {
	NULL, /* 0 no such transaction type */
	ztest_replay_create, /* TX_CREATE */
	NULL, /* TX_MKDIR */
	NULL, /* TX_MKXATTR */
	NULL, /* TX_SYMLINK */
	ztest_replay_remove, /* TX_REMOVE */
	NULL, /* TX_RMDIR */
	NULL, /* TX_LINK */
	NULL, /* TX_RENAME */
	ztest_replay_write, /* TX_WRITE */
	ztest_replay_truncate, /* TX_TRUNCATE */
	ztest_replay_setattr, /* TX_SETATTR */
	NULL, /* TX_ACL */
	NULL, /* TX_CREATE_ACL */
	NULL, /* TX_CREATE_ATTR */
	NULL, /* TX_CREATE_ACL_ATTR */
	NULL, /* TX_MKDIR_ACL */
	NULL, /* TX_MKDIR_ATTR */
	NULL, /* TX_MKDIR_ACL_ATTR */
	NULL, /* TX_WRITE2 */
	NULL, /* TX_SETSAXATTR */
	};

	/*
	* ZIL get_data callbacks
	*/

	static void
	ztest_get_done(zgd_t *zgd, int error)
	{
	(void) error;
	ztest_ds_t *zd = zgd->zgd_private;
	uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object;

	if (zgd->zgd_db)
	dmu_buf_rele(zgd->zgd_db, zgd);

	ztest_range_unlock((rl_t *)zgd->zgd_lr);
	ztest_object_unlock(zd, object);

	umem_free(zgd, sizeof (*zgd));
	}

	static int
	ztest_get_data(void arg, uint64_t arg2, lr_write_t lr, char *buf,
	struct lwb lwb, zio_t zio)
	{
	(void) arg2;
	ztest_ds_t *zd = arg;
	objset_t *os = zd->zd_os;
	uint64_t object = lr->lr_foid;
	uint64_t offset = lr->lr_offset;
	uint64_t size = lr->lr_length;
	uint64_t txg = lr->lr_common.lrc_txg;
	uint64_t crtxg;
	dmu_object_info_t doi;
	dmu_buf_t *db;
	zgd_t *zgd;
	int error;

	ASSERT3P(lwb, !=, NULL);
	ASSERT3P(zio, !=, NULL);
	ASSERT3U(size, !=, 0);

	ztest_object_lock(zd, object, RL_READER);
	error = dmu_bonus_hold(os, object, FTAG, &db);
	if (error) {
	ztest_object_unlock(zd, object);
	return (error);
	}

	crtxg = ztest_bt_bonus(db)->bt_crtxg;

	if (crtxg == 0 \|\| crtxg > txg) {
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, object);
	return (ENOENT);
	}

	dmu_object_info_from_db(db, &doi);
	dmu_buf_rele(db, FTAG);
	db = NULL;

	zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL);
	zgd->zgd_lwb = lwb;
	zgd->zgd_private = zd;

	if (buf != NULL) { /* immediate write */
	zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
	object, offset, size, RL_READER);

	error = dmu_read(os, object, offset, size, buf,
	DMU_READ_NO_PREFETCH);
	ASSERT0(error);
	} else {
	size = doi.doi_data_block_size;
	if (ISP2(size)) {
	offset = P2ALIGN(offset, size);
	} else {
	ASSERT3U(offset, <, size);
	offset = 0;
	}

	zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
	object, offset, size, RL_READER);

	error = dmu_buf_hold(os, object, offset, zgd, &db,
	DMU_READ_NO_PREFETCH);

	if (error == 0) {
	blkptr_t *bp = &lr->lr_blkptr;

	zgd->zgd_db = db;
	zgd->zgd_bp = bp;

	ASSERT3U(db->db_offset, ==, offset);
	ASSERT3U(db->db_size, ==, size);

	error = dmu_sync(zio, lr->lr_common.lrc_txg,
	ztest_get_done, zgd);

	if (error == 0)
	return (0);
	}
	}

	ztest_get_done(zgd, error);

	return (error);
	}

	static void *
	ztest_lr_alloc(size_t lrsize, char *name)
	{
	char *lr;
	size_t namesize = name ? strlen(name) + 1 : 0;

	lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL);

	if (name)
	memcpy(lr + lrsize, name, namesize);

	return (lr);
	}

	static void
	ztest_lr_free(void lr, size_t lrsize, char name)
	{
	size_t namesize = name ? strlen(name) + 1 : 0;

	umem_free(lr, lrsize + namesize);
	}

	/*
	* Lookup a bunch of objects. Returns the number of objects not found.
	*/
	static int
	ztest_lookup(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int error;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	for (i = 0; i < count; i++, od++) {
	od->od_object = 0;
	error = zap_lookup(zd->zd_os, od->od_dir, od->od_name,
	sizeof (uint64_t), 1, &od->od_object);
	if (error) {
	ASSERT3S(error, ==, ENOENT);
	ASSERT0(od->od_object);
	missing++;
	} else {
	dmu_buf_t *db;
	ztest_block_tag_t *bbt;
	dmu_object_info_t doi;

	ASSERT3U(od->od_object, !=, 0);
	ASSERT0(missing); /* there should be no gaps */

	ztest_object_lock(zd, od->od_object, RL_READER);
	VERIFY0(dmu_bonus_hold(zd->zd_os, od->od_object,
	FTAG, &db));
	dmu_object_info_from_db(db, &doi);
	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	od->od_type = doi.doi_type;
	od->od_blocksize = doi.doi_data_block_size;
	od->od_gen = bbt->bt_gen;
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, od->od_object);
	}
	}

	return (missing);
	}

	static int
	ztest_create(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	for (i = 0; i < count; i++, od++) {
	if (missing) {
	od->od_object = 0;
	missing++;
	continue;
	}

	lr_create_t lr = ztest_lr_alloc(sizeof (lr), od->od_name);

	lr->lr_doid = od->od_dir;
	lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */
	lr->lrz_type = od->od_crtype;
	lr->lrz_blocksize = od->od_crblocksize;
	lr->lrz_ibshift = ztest_random_ibshift();
	lr->lrz_bonustype = DMU_OT_UINT64_OTHER;
	lr->lrz_dnodesize = od->od_crdnodesize;
	lr->lr_gen = od->od_crgen;
	lr->lr_crtime[0] = time(NULL);

	if (ztest_replay_create(zd, lr, B_FALSE) != 0) {
	ASSERT0(missing);
	od->od_object = 0;
	missing++;
	} else {
	od->od_object = lr->lr_foid;
	od->od_type = od->od_crtype;
	od->od_blocksize = od->od_crblocksize;
	od->od_gen = od->od_crgen;
	ASSERT3U(od->od_object, !=, 0);
	}

	ztest_lr_free(lr, sizeof (*lr), od->od_name);
	}

	return (missing);
	}

	static int
	ztest_remove(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int error;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	od += count - 1;

	for (i = count - 1; i >= 0; i--, od--) {
	if (missing) {
	missing++;
	continue;
	}

	/*
	* No object was found.
	*/
	if (od->od_object == 0)
	continue;

	lr_remove_t lr = ztest_lr_alloc(sizeof (lr), od->od_name);

	lr->lr_doid = od->od_dir;

	if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) {
	ASSERT3U(error, ==, ENOSPC);
	missing++;
	} else {
	od->od_object = 0;
	}
	ztest_lr_free(lr, sizeof (*lr), od->od_name);
	}

	return (missing);
	}

	static int
	ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size,
	void *data)
	{
	lr_write_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr) + size, NULL);

	lr->lr_foid = object;
	lr->lr_offset = offset;
	lr->lr_length = size;
	lr->lr_blkoff = 0;
	BP_ZERO(&lr->lr_blkptr);

	memcpy(lr + 1, data, size);

	error = ztest_replay_write(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr) + size, NULL);

	return (error);
	}

	static int
	ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
	{
	lr_truncate_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr), NULL);

	lr->lr_foid = object;
	lr->lr_offset = offset;
	lr->lr_length = size;

	error = ztest_replay_truncate(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr), NULL);

	return (error);
	}

	static int
	ztest_setattr(ztest_ds_t *zd, uint64_t object)
	{
	lr_setattr_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr), NULL);

	lr->lr_foid = object;
	lr->lr_size = 0;
	lr->lr_mode = 0;

	error = ztest_replay_setattr(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr), NULL);

	return (error);
	}

	static void
	ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
	{
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	uint64_t txg;
	rl_t *rl;

	txg_wait_synced(dmu_objset_pool(os), 0);

	ztest_object_lock(zd, object, RL_READER);
	rl = ztest_range_lock(zd, object, offset, size, RL_WRITER);

	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, object, offset, size);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);

	if (txg != 0) {
	dmu_prealloc(os, object, offset, size, tx);
	dmu_tx_commit(tx);
	txg_wait_synced(dmu_objset_pool(os), txg);
	} else {
	(void) dmu_free_long_range(os, object, offset, size);
	}

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, object);
	}

	static void
	ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
	{
	int err;
	ztest_block_tag_t wbt;
	dmu_object_info_t doi;
	enum ztest_io_type io_type;
	uint64_t blocksize;
	void *data;

	VERIFY0(dmu_object_info(zd->zd_os, object, &doi));
	blocksize = doi.doi_data_block_size;
	data = umem_alloc(blocksize, UMEM_NOFAIL);

	/*
	* Pick an i/o type at random, biased toward writing block tags.
	*/
	io_type = ztest_random(ZTEST_IO_TYPES);
	if (ztest_random(2) == 0)
	io_type = ZTEST_IO_WRITE_TAG;

	(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);

	switch (io_type) {

	case ZTEST_IO_WRITE_TAG:
	ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize,
	offset, 0, 0, 0);
	(void) ztest_write(zd, object, offset, sizeof (wbt), &wbt);
	break;

	case ZTEST_IO_WRITE_PATTERN:
	(void) memset(data, 'a' + (object + offset) % 5, blocksize);
	if (ztest_random(2) == 0) {
	/*
	* Induce fletcher2 collisions to ensure that
	* zio_ddt_collision() detects and resolves them
	* when using fletcher2-verify for deduplication.
	*/
	((uint64_t *)data)[0] ^= 1ULL << 63;
	((uint64_t *)data)[4] ^= 1ULL << 63;
	}
	(void) ztest_write(zd, object, offset, blocksize, data);
	break;

	case ZTEST_IO_WRITE_ZEROES:
	memset(data, 0, blocksize);
	(void) ztest_write(zd, object, offset, blocksize, data);
	break;

	case ZTEST_IO_TRUNCATE:
	(void) ztest_truncate(zd, object, offset, blocksize);
	break;

	case ZTEST_IO_SETATTR:
	(void) ztest_setattr(zd, object);
	break;
	default:
	break;

	case ZTEST_IO_REWRITE:
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	err = ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa),
	B_FALSE);
	VERIFY(err == 0 \|\| err == ENOSPC);
	err = ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_COMPRESSION,
	ztest_random_dsl_prop(ZFS_PROP_COMPRESSION),
	B_FALSE);
	VERIFY(err == 0 \|\| err == ENOSPC);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data,
	DMU_READ_NO_PREFETCH));

	(void) ztest_write(zd, object, offset, blocksize, data);
	break;
	}

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);

	umem_free(data, blocksize);
	}

	/*
	* Initialize an object description template.
	*/
	static void
	ztest_od_init(ztest_od_t od, uint64_t id, const char tag, uint64_t index,
	dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize,
	uint64_t gen)
	{
	od->od_dir = ZTEST_DIROBJ;
	od->od_object = 0;

	od->od_crtype = type;
	od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize();
	od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize();
	od->od_crgen = gen;

	od->od_type = DMU_OT_NONE;
	od->od_blocksize = 0;
	od->od_gen = 0;

	(void) snprintf(od->od_name, sizeof (od->od_name),
	"%s(%"PRId64")[%"PRIu64"]",
	tag, id, index);
	}

	/*
	* Lookup or create the objects for a test using the od template.
	* If the objects do not all exist, or if 'remove' is specified,
	* remove any existing objects and create new ones. Otherwise,
	* use the existing objects.
	*/
	static int
	ztest_object_init(ztest_ds_t zd, ztest_od_t od, size_t size, boolean_t remove)
	{
	int count = size / sizeof (*od);
	int rv = 0;

	mutex_enter(&zd->zd_dirobj_lock);
	if ((ztest_lookup(zd, od, count) != 0 \|\| remove) &&
	(ztest_remove(zd, od, count) != 0 \|\|
	ztest_create(zd, od, count) != 0))
	rv = -1;
	zd->zd_od = od;
	mutex_exit(&zd->zd_dirobj_lock);

	return (rv);
	}

	void
	ztest_zil_commit(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	zilog_t *zilog = zd->zd_zilog;

	(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);

	zil_commit(zilog, ztest_random(ZTEST_OBJECTS));

	/*
	* Remember the committed values in zd, which is in parent/child
	* shared memory. If we die, the next iteration of ztest_run()
	* will verify that the log really does contain this record.
	*/
	mutex_enter(&zilog->zl_lock);
	ASSERT3P(zd->zd_shared, !=, NULL);
	ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq);
	zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq;
	mutex_exit(&zilog->zl_lock);

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
	}

	/*
	* This function is designed to simulate the operations that occur during a
	* mount/unmount operation. We hold the dataset across these operations in an
	* attempt to expose any implicit assumptions about ZIL management.
	*/
	void
	ztest_zil_remount(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;

	/*
	* We hold the ztest_vdev_lock so we don't cause problems with
	* other threads that wish to remove a log device, such as
	* ztest_device_removal().
	*/
	mutex_enter(&ztest_vdev_lock);

	/*
	* We grab the zd_dirobj_lock to ensure that no other thread is
	* updating the zil (i.e. adding in-memory log records) and the
	* zd_zilog_lock to block any I/O.
	*/
	mutex_enter(&zd->zd_dirobj_lock);
	(void) pthread_rwlock_wrlock(&zd->zd_zilog_lock);

	/* zfsvfs_teardown() */
	zil_close(zd->zd_zilog);

	/* zfsvfs_setup() */
	VERIFY3P(zil_open(os, ztest_get_data, NULL), ==, zd->zd_zilog);
	zil_replay(os, zd, ztest_replay_vector);

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
	mutex_exit(&zd->zd_dirobj_lock);
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify that we can't destroy an active pool, create an existing pool,
	* or create a pool with a bad vdev spec.
	*/
	void
	ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_opts_t *zo = &ztest_opts;
	spa_t *spa;
	nvlist_t *nvroot;

	if (zo->zo_mmp_test)
	return;

	/*
	* Attempt to create using a bad file.
	*/
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
	VERIFY3U(ENOENT, ==,
	spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* Attempt to create using a bad mirror.
	*/
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1);
	VERIFY3U(ENOENT, ==,
	spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* Attempt to create an existing pool. It shouldn't matter
	* what's in the nvroot; we should fail with EEXIST.
	*/
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
	VERIFY3U(EEXIST, ==,
	spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* We open a reference to the spa and then we try to export it
	* expecting one of the following errors:
	*
	* EBUSY
	* Because of the reference we just opened.
	*
	* ZFS_ERR_EXPORT_IN_PROGRESS
	* For the case that there is another ztest thread doing
	* an export concurrently.
	*/
	VERIFY0(spa_open(zo->zo_pool, &spa, FTAG));
	int error = spa_destroy(zo->zo_pool);
	if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) {
	fatal(B_FALSE, "spa_destroy(%s) returned unexpected value %d",
	spa->spa_name, error);
	}
	spa_close(spa, FTAG);

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Start and then stop the MMP threads to ensure the startup and shutdown code
	* works properly. Actual protection and property-related code tested via ZTS.
	*/
	void
	ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_opts_t *zo = &ztest_opts;
	spa_t *spa = ztest_spa;

	if (zo->zo_mmp_test)
	return;

	/*
	* Since enabling MMP involves setting a property, it could not be done
	* while the pool is suspended.
	*/
	if (spa_suspended(spa))
	return;

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	mutex_enter(&spa->spa_props_lock);

	zfs_multihost_fail_intervals = 0;

	if (!spa_multihost(spa)) {
	spa->spa_multihost = B_TRUE;
	mmp_thread_start(spa);
	}

	mutex_exit(&spa->spa_props_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	txg_wait_synced(spa_get_dsl(spa), 0);
	mmp_signal_all_threads();
	txg_wait_synced(spa_get_dsl(spa), 0);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	mutex_enter(&spa->spa_props_lock);

	if (spa_multihost(spa)) {
	mmp_thread_stop(spa);
	spa->spa_multihost = B_FALSE;
	}

	mutex_exit(&spa->spa_props_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	void
	ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa;
	uint64_t initial_version = SPA_VERSION_INITIAL;
	uint64_t version, newversion;
	nvlist_t nvroot, props;
	char *name;

	if (ztest_opts.zo_mmp_test)
	return;

	/* dRAID added after feature flags, skip upgrade test. */
	if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0)
	return;

	mutex_enter(&ztest_vdev_lock);
	name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool);

	/*
	* Clean up from previous runs.
	*/
	(void) spa_destroy(name);

	nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0,
	NULL, ztest_opts.zo_raid_children, ztest_opts.zo_mirrors, 1);

	/*
	* If we're configuring a RAIDZ device then make sure that the
	* initial version is capable of supporting that feature.
	*/
	switch (ztest_opts.zo_raid_parity) {
	case 0:
	case 1:
	initial_version = SPA_VERSION_INITIAL;
	break;
	case 2:
	initial_version = SPA_VERSION_RAIDZ2;
	break;
	case 3:
	initial_version = SPA_VERSION_RAIDZ3;
	break;
	}

	/*
	* Create a pool with a spa version that can be upgraded. Pick
	* a value between initial_version and SPA_VERSION_BEFORE_FEATURES.
	*/
	do {
	version = ztest_random_spa_version(initial_version);
	} while (version > SPA_VERSION_BEFORE_FEATURES);

	props = fnvlist_alloc();
	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), version);
	VERIFY0(spa_create(name, nvroot, props, NULL, NULL));
	fnvlist_free(nvroot);
	fnvlist_free(props);

	VERIFY0(spa_open(name, &spa, FTAG));
	VERIFY3U(spa_version(spa), ==, version);
	newversion = ztest_random_spa_version(version + 1);

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("upgrading spa version from "
	"%"PRIu64" to %"PRIu64"\n",
	version, newversion);
	}

	spa_upgrade(spa, newversion);
	VERIFY3U(spa_version(spa), >, version);
	VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config,
	zpool_prop_to_name(ZPOOL_PROP_VERSION)));
	spa_close(spa, FTAG);

	kmem_strfree(name);
	mutex_exit(&ztest_vdev_lock);
	}

	static void
	ztest_spa_checkpoint(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));

	int error = spa_checkpoint(spa->spa_name);

	switch (error) {
	case 0:
	case ZFS_ERR_DEVRM_IN_PROGRESS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	case ZFS_ERR_CHECKPOINT_EXISTS:
	break;
	case ENOSPC:
	ztest_record_enospc(FTAG);
	break;
	default:
	fatal(B_FALSE, "spa_checkpoint(%s) = %d", spa->spa_name, error);
	}
	}

	static void
	ztest_spa_discard_checkpoint(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));

	int error = spa_checkpoint_discard(spa->spa_name);

	switch (error) {
	case 0:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	case ZFS_ERR_NO_CHECKPOINT:
	break;
	default:
	fatal(B_FALSE, "spa_discard_checkpoint(%s) = %d",
	spa->spa_name, error);
	}

	}

	void
	ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;

	mutex_enter(&ztest_checkpoint_lock);
	if (ztest_random(2) == 0) {
	ztest_spa_checkpoint(spa);
	} else {
	ztest_spa_discard_checkpoint(spa);
	}
	mutex_exit(&ztest_checkpoint_lock);
	}


	static vdev_t *
	vdev_lookup_by_path(vdev_t vd, const char path)
	{
	vdev_t *mvd;
	int c;

	if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0)
	return (vd);

	for (c = 0; c < vd->vdev_children; c++)
	if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) !=
	NULL)
	return (mvd);

	return (NULL);
	}

	static int
	spa_num_top_vdevs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV);
	return (rvd->vdev_children);
	}

	/*
	* Verify that vdev_add() works as expected.
	*/
	void
	ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	uint64_t leaves;
	uint64_t guid;
	nvlist_t *nvroot;
	int error;

	if (ztest_opts.zo_mmp_test)
	return;

	mutex_enter(&ztest_vdev_lock);
	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) *
	ztest_opts.zo_raid_children;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;

	/*
	* If we have slogs then remove them 1/4 of the time.
	*/
	if (spa_has_slogs(spa) && ztest_random(4) == 0) {
	metaslab_group_t *mg;

	/*
	* find the first real slog in log allocation class
	*/
	mg = spa_log_class(spa)->mc_allocator[0].mca_rotor;
	while (!mg->mg_vd->vdev_islog)
	mg = mg->mg_next;

	guid = mg->mg_vd->vdev_guid;

	spa_config_exit(spa, SCL_VDEV, FTAG);

	/*
	* We have to grab the zs_name_lock as writer to
	* prevent a race between removing a slog (dmu_objset_find)
	* and destroying a dataset. Removing the slog will
	* grab a reference on the dataset which may cause
	* dsl_destroy_head() to fail with EBUSY thus
	* leaving the dataset in an inconsistent state.
	*/
	pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_vdev_remove(spa, guid, B_FALSE);
	pthread_rwlock_unlock(&ztest_name_lock);

	switch (error) {
	case 0:
	case EEXIST: /* Generic zil_reset() error */
	case EBUSY: /* Replay required */
	case EACCES: /* Crypto key not loaded */
	case ZFS_ERR_CHECKPOINT_EXISTS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	break;
	default:
	fatal(B_FALSE, "spa_vdev_remove() = %d", error);
	}
	} else {
	spa_config_exit(spa, SCL_VDEV, FTAG);

	/*
	* Make 1/4 of the devices be log devices
	*/
	nvroot = make_vdev_root(NULL, NULL, NULL,
	ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ?
	"log" : NULL, ztest_opts.zo_raid_children, zs->zs_mirrors,
	1);

	error = spa_vdev_add(spa, nvroot);
	fnvlist_free(nvroot);

	switch (error) {
	case 0:
	break;
	case ENOSPC:
	ztest_record_enospc("spa_vdev_add");
	break;
	default:
	fatal(B_FALSE, "spa_vdev_add() = %d", error);
	}
	}

	mutex_exit(&ztest_vdev_lock);
	}

	void
	ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	uint64_t leaves;
	nvlist_t *nvroot;
	const char *class = (ztest_random(2) == 0) ?
	VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP;
	int error;

	/*
	* By default add a special vdev 50% of the time
	*/
	if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) \|\|
	(ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND &&
	ztest_random(2) == 0)) {
	return;
	}

	mutex_enter(&ztest_vdev_lock);

	/* Only test with mirrors */
	if (zs->zs_mirrors < 2) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/* requires feature@allocation_classes */
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) *
	ztest_opts.zo_raid_children;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;
	spa_config_exit(spa, SCL_VDEV, FTAG);

	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
	class, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);

	error = spa_vdev_add(spa, nvroot);
	fnvlist_free(nvroot);

	if (error == ENOSPC)
	ztest_record_enospc("spa_vdev_add");
	else if (error != 0)
	fatal(B_FALSE, "spa_vdev_add() = %d", error);

	/*
	* 50% of the time allow small blocks in the special class
	*/
	if (error == 0 &&
	spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) {
	if (ztest_opts.zo_verbose >= 3)
	(void) printf("Enabling special VDEV small blocks\n");
	(void) ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE);
	}

	mutex_exit(&ztest_vdev_lock);

	if (ztest_opts.zo_verbose >= 3) {
	metaslab_class_t *mc;

	if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0)
	mc = spa_special_class(spa);
	else
	mc = spa_dedup_class(spa);
	(void) printf("Added a %s mirrored vdev (of %d)\n",
	class, (int)mc->mc_groups);
	}
	}

	/*
	* Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
	*/
	void
	ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	spa_aux_vdev_t *sav;
	const char *aux;
	char *path;
	uint64_t guid = 0;
	int error, ignore_err = 0;

	if (ztest_opts.zo_mmp_test)
	return;

	path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	if (ztest_random(2) == 0) {
	sav = &spa->spa_spares;
	aux = ZPOOL_CONFIG_SPARES;
	} else {
	sav = &spa->spa_l2cache;
	aux = ZPOOL_CONFIG_L2CACHE;
	}

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	if (sav->sav_count != 0 && ztest_random(4) == 0) {
	/*
	* Pick a random device to remove.
	*/
	vdev_t *svd = sav->sav_vdevs[ztest_random(sav->sav_count)];

	/* dRAID spares cannot be removed; try anyways to see ENOTSUP */
	if (strstr(svd->vdev_path, VDEV_TYPE_DRAID) != NULL)
	ignore_err = ENOTSUP;

	guid = svd->vdev_guid;
	} else {
	/*
	* Find an unused device we can add.
	*/
	zs->zs_vdev_aux = 0;
	for (;;) {
	int c;
	(void) snprintf(path, MAXPATHLEN, ztest_aux_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool, aux,
	zs->zs_vdev_aux);
	for (c = 0; c < sav->sav_count; c++)
	if (strcmp(sav->sav_vdevs[c]->vdev_path,
	path) == 0)
	break;
	if (c == sav->sav_count &&
	vdev_lookup_by_path(rvd, path) == NULL)
	break;
	zs->zs_vdev_aux++;
	}
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);

	if (guid == 0) {
	/*
	* Add a new device.
	*/
	nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL,
	(ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1);
	error = spa_vdev_add(spa, nvroot);

	switch (error) {
	case 0:
	break;
	default:
	fatal(B_FALSE, "spa_vdev_add(%p) = %d", nvroot, error);
	}
	fnvlist_free(nvroot);
	} else {
	/*
	* Remove an existing device. Sometimes, dirty its
	* vdev state first to make sure we handle removal
	* of devices that have pending state changes.
	*/
	if (ztest_random(2) == 0)
	(void) vdev_online(spa, guid, 0, NULL);

	error = spa_vdev_remove(spa, guid, B_FALSE);

	switch (error) {
	case 0:
	case EBUSY:
	case ZFS_ERR_CHECKPOINT_EXISTS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	break;
	default:
	if (error != ignore_err)
	fatal(B_FALSE,
	"spa_vdev_remove(%"PRIu64") = %d",
	guid, error);
	}
	}

	mutex_exit(&ztest_vdev_lock);

	umem_free(path, MAXPATHLEN);
	}

	/*
	* split a pool if it has mirror tlvdevs
	*/
	void
	ztest_split_pool(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	nvlist_t tree, child, config, split, *schild;
	uint_t c, children, schildren = 0, lastlogid = 0;
	int error = 0;

	if (ztest_opts.zo_mmp_test)
	return;

	mutex_enter(&ztest_vdev_lock);

	/* ensure we have a usable config; mirrors of raidz aren't supported */
	if (zs->zs_mirrors < 3 \|\| ztest_opts.zo_raid_children > 1) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/* clean up the old pool, if any */
	(void) spa_destroy("splitp");

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* generate a config from the existing config */
	mutex_enter(&spa->spa_props_lock);
	tree = fnvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE);
	mutex_exit(&spa->spa_props_lock);

	VERIFY0(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
	&child, &children));

	schild = malloc(rvd->vdev_children * sizeof (nvlist_t *));
	for (c = 0; c < children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	nvlist_t **mchild;
	uint_t mchildren;

	if (tvd->vdev_islog \|\| tvd->vdev_ops == &vdev_hole_ops) {
	schild[schildren] = fnvlist_alloc();
	fnvlist_add_string(schild[schildren],
	ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE);
	fnvlist_add_uint64(schild[schildren],
	ZPOOL_CONFIG_IS_HOLE, 1);
	if (lastlogid == 0)
	lastlogid = schildren;
	++schildren;
	continue;
	}
	lastlogid = 0;
	VERIFY0(nvlist_lookup_nvlist_array(child[c],
	ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren));
	schild[schildren++] = fnvlist_dup(mchild[0]);
	}

	/* OK, create a config that can be used to split */
	split = fnvlist_alloc();
	fnvlist_add_string(split, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
	fnvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)schild, lastlogid != 0 ? lastlogid : schildren);

	config = fnvlist_alloc();
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split);

	for (c = 0; c < schildren; c++)
	fnvlist_free(schild[c]);
	free(schild);
	fnvlist_free(split);

	spa_config_exit(spa, SCL_VDEV, FTAG);

	(void) pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	fnvlist_free(config);

	if (error == 0) {
	(void) printf("successful split - results:\n");
	mutex_enter(&spa_namespace_lock);
	show_pool_stats(spa);
	show_pool_stats(spa_lookup("splitp"));
	mutex_exit(&spa_namespace_lock);
	++zs->zs_splits;
	--zs->zs_mirrors;
	}
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify that we can attach and detach devices.
	*/
	void
	ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	spa_aux_vdev_t *sav = &spa->spa_spares;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t oldvd, newvd, *pvd;
	nvlist_t *root;
	uint64_t leaves;
	uint64_t leaf, top;
	uint64_t ashift = ztest_get_ashift();
	uint64_t oldguid, pguid;
	uint64_t oldsize, newsize;
	char oldpath, newpath;
	int replacing;
	int oldvd_has_siblings = B_FALSE;
	int newvd_is_spare = B_FALSE;
	int newvd_is_dspare = B_FALSE;
	int oldvd_is_log;
	int error, expected_error;

	if (ztest_opts.zo_mmp_test)
	return;

	oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	mutex_enter(&ztest_vdev_lock);
	leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* If a vdev is in the process of being removed, its removal may
	* finish while we are in progress, leading to an unexpected error
	* value. Don't bother trying to attach while we are in the middle
	* of removal.
	*/
	if (ztest_device_removal_active) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	/*
	* Decide whether to do an attach or a replace.
	*/
	replacing = ztest_random(2);

	/*
	* Pick a random top-level vdev.
	*/
	top = ztest_random_vdev_top(spa, B_TRUE);

	/*
	* Pick a random leaf within it.
	*/
	leaf = ztest_random(leaves);

	/*
	* Locate this vdev.
	*/
	oldvd = rvd->vdev_child[top];

	/* pick a child from the mirror */
	if (zs->zs_mirrors >= 1) {
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_mirror_ops);
	ASSERT3U(oldvd->vdev_children, >=, zs->zs_mirrors);
	oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raid_children];
	}

	/* pick a child out of the raidz group */
	if (ztest_opts.zo_raid_children > 1) {
	if (strcmp(oldvd->vdev_ops->vdev_op_type, "raidz") == 0)
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_raidz_ops);
	else
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_draid_ops);
	ASSERT3U(oldvd->vdev_children, ==, ztest_opts.zo_raid_children);
	oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raid_children];
	}

	/*
	* If we're already doing an attach or replace, oldvd may be a
	* mirror vdev -- in which case, pick a random child.
	*/
	while (oldvd->vdev_children != 0) {
	oldvd_has_siblings = B_TRUE;
	ASSERT3U(oldvd->vdev_children, >=, 2);
	oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)];
	}

	oldguid = oldvd->vdev_guid;
	oldsize = vdev_get_min_asize(oldvd);
	oldvd_is_log = oldvd->vdev_top->vdev_islog;
	- (void) strcpy(oldpath, oldvd->vdev_path);
	+ (void) strlcpy(oldpath, oldvd->vdev_path, MAXPATHLEN);
	pvd = oldvd->vdev_parent;
	pguid = pvd->vdev_guid;

	/*
	* If oldvd has siblings, then half of the time, detach it. Prior
	* to the detach the pool is scrubbed in order to prevent creating
	* unrepairable blocks as a result of the data corruption injection.
	*/
	if (oldvd_has_siblings && ztest_random(2) == 0) {
	spa_config_exit(spa, SCL_ALL, FTAG);

	error = ztest_scrub_impl(spa);
	if (error)
	goto out;

	error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE);
	if (error != 0 && error != ENODEV && error != EBUSY &&
	error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS &&
	error != ZFS_ERR_DISCARDING_CHECKPOINT)
	fatal(B_FALSE, "detach (%s) returned %d",
	oldpath, error);
	goto out;
	}

	/*
	* For the new vdev, choose with equal probability between the two
	* standard paths (ending in either 'a' or 'b') or a random hot spare.
	*/
	if (sav->sav_count != 0 && ztest_random(3) == 0) {
	newvd = sav->sav_vdevs[ztest_random(sav->sav_count)];
	newvd_is_spare = B_TRUE;

	if (newvd->vdev_ops == &vdev_draid_spare_ops)
	newvd_is_dspare = B_TRUE;

	- (void) strcpy(newpath, newvd->vdev_path);
	+ (void) strlcpy(newpath, newvd->vdev_path, MAXPATHLEN);
	} else {
	(void) snprintf(newpath, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + leaf);
	if (ztest_random(2) == 0)
	newpath[strlen(newpath) - 1] = 'b';
	newvd = vdev_lookup_by_path(rvd, newpath);
	}

	if (newvd) {
	/*
	* Reopen to ensure the vdev's asize field isn't stale.
	*/
	vdev_reopen(newvd);
	newsize = vdev_get_min_asize(newvd);
	} else {
	/*
	* Make newsize a little bigger or smaller than oldsize.
	* If it's smaller, the attach should fail.
	* If it's larger, and we're doing a replace,
	* we should get dynamic LUN growth when we're done.
	*/
	newsize = 10 * oldsize / (9 + ztest_random(3));
	}

	/*
	* If pvd is not a mirror or root, the attach should fail with ENOTSUP,
	* unless it's a replace; in that case any non-replacing parent is OK.
	*
	* If newvd is already part of the pool, it should fail with EBUSY.
	*
	* If newvd is too small, it should fail with EOVERFLOW.
	*
	* If newvd is a distributed spare and it's being attached to a
	* dRAID which is not its parent it should fail with EINVAL.
	*/
	if (pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_root_ops && (!replacing \|\|
	pvd->vdev_ops == &vdev_replacing_ops \|\|
	pvd->vdev_ops == &vdev_spare_ops))
	expected_error = ENOTSUP;
	else if (newvd_is_spare && (!replacing \|\| oldvd_is_log))
	expected_error = ENOTSUP;
	else if (newvd == oldvd)
	expected_error = replacing ? 0 : EBUSY;
	else if (vdev_lookup_by_path(rvd, newpath) != NULL)
	expected_error = EBUSY;
	else if (!newvd_is_dspare && newsize < oldsize)
	expected_error = EOVERFLOW;
	else if (ashift > oldvd->vdev_top->vdev_ashift)
	expected_error = EDOM;
	else if (newvd_is_dspare && pvd != vdev_draid_spare_get_parent(newvd))
	expected_error = ENOTSUP;
	else
	expected_error = 0;

	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* Build the nvlist describing newpath.
	*/
	root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0,
	ashift, NULL, 0, 0, 1);

	/*
	* When supported select either a healing or sequential resilver.
	*/
	boolean_t rebuilding = B_FALSE;
	if (pvd->vdev_ops == &vdev_mirror_ops \|\|
	pvd->vdev_ops == &vdev_root_ops) {
	rebuilding = !!ztest_random(2);
	}

	error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding);

	fnvlist_free(root);

	/*
	* If our parent was the replacing vdev, but the replace completed,
	* then instead of failing with ENOTSUP we may either succeed,
	* fail with ENODEV, or fail with EOVERFLOW.
	*/
	if (expected_error == ENOTSUP &&
	(error == 0 \|\| error == ENODEV \|\| error == EOVERFLOW))
	expected_error = error;

	/*
	* If someone grew the LUN, the replacement may be too small.
	*/
	if (error == EOVERFLOW \|\| error == EBUSY)
	expected_error = error;

	if (error == ZFS_ERR_CHECKPOINT_EXISTS \|\|
	error == ZFS_ERR_DISCARDING_CHECKPOINT \|\|
	error == ZFS_ERR_RESILVER_IN_PROGRESS \|\|
	error == ZFS_ERR_REBUILD_IN_PROGRESS)
	expected_error = error;

	if (error != expected_error && expected_error != EBUSY) {
	fatal(B_FALSE, "attach (%s %"PRIu64", %s %"PRIu64", %d) "
	"returned %d, expected %d",
	oldpath, oldsize, newpath,
	newsize, replacing, error, expected_error);
	}
	out:
	mutex_exit(&ztest_vdev_lock);

	umem_free(oldpath, MAXPATHLEN);
	umem_free(newpath, MAXPATHLEN);
	}

	void
	ztest_device_removal(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	vdev_t *vd;
	uint64_t guid;
	int error;

	mutex_enter(&ztest_vdev_lock);

	if (ztest_device_removal_active) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* Remove a random top-level vdev and wait for removal to finish.
	*/
	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE));
	guid = vd->vdev_guid;
	spa_config_exit(spa, SCL_VDEV, FTAG);

	error = spa_vdev_remove(spa, guid, B_FALSE);
	if (error == 0) {
	ztest_device_removal_active = B_TRUE;
	mutex_exit(&ztest_vdev_lock);

	/*
	* spa->spa_vdev_removal is created in a sync task that
	* is initiated via dsl_sync_task_nowait(). Since the
	* task may not run before spa_vdev_remove() returns, we
	* must wait at least 1 txg to ensure that the removal
	* struct has been created.
	*/
	txg_wait_synced(spa_get_dsl(spa), 0);

	while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
	txg_wait_synced(spa_get_dsl(spa), 0);
	} else {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The pool needs to be scrubbed after completing device removal.
	* Failure to do so may result in checksum errors due to the
	* strategy employed by ztest_fault_inject() when selecting which
	* offset are redundant and can be damaged.
	*/
	error = spa_scan(spa, POOL_SCAN_SCRUB);
	if (error == 0) {
	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	mutex_enter(&ztest_vdev_lock);
	ztest_device_removal_active = B_FALSE;
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Callback function which expands the physical size of the vdev.
	*/
	static vdev_t *
	grow_vdev(vdev_t vd, void arg)
	{
	spa_t *spa __maybe_unused = vd->vdev_spa;
	size_t *newsize = arg;
	size_t fsize;
	int fd;

	ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	if ((fd = open(vd->vdev_path, O_RDWR)) == -1)
	return (vd);

	fsize = lseek(fd, 0, SEEK_END);
	VERIFY0(ftruncate(fd, *newsize));

	if (ztest_opts.zo_verbose >= 6) {
	(void) printf("%s grew from %lu to %lu bytes\n",
	vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize);
	}
	(void) close(fd);
	return (NULL);
	}

	/*
	* Callback function which expands a given vdev by calling vdev_online().
	*/
	static vdev_t *
	online_vdev(vdev_t vd, void arg)
	{
	(void) arg;
	spa_t *spa = vd->vdev_spa;
	vdev_t *tvd = vd->vdev_top;
	uint64_t guid = vd->vdev_guid;
	uint64_t generation = spa->spa_config_generation + 1;
	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
	int error;

	ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	/* Calling vdev_online will initialize the new metaslabs */
	spa_config_exit(spa, SCL_STATE, spa);
	error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate);
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	/*
	* If vdev_online returned an error or the underlying vdev_open
	* failed then we abort the expand. The only way to know that
	* vdev_open fails is by checking the returned newstate.
	*/
	if (error \|\| newstate != VDEV_STATE_HEALTHY) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Unable to expand vdev, state %u, "
	"error %d\n", newstate, error);
	}
	return (vd);
	}
	ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY);

	/*
	* Since we dropped the lock we need to ensure that we're
	* still talking to the original vdev. It's possible this
	* vdev may have been detached/replaced while we were
	* trying to online it.
	*/
	if (generation != spa->spa_config_generation) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("vdev configuration has changed, "
	"guid %"PRIu64", state %"PRIu64", "
	"expected gen %"PRIu64", got gen %"PRIu64"\n",
	guid,
	tvd->vdev_state,
	generation,
	spa->spa_config_generation);
	}
	return (vd);
	}
	return (NULL);
	}

	/*
	* Traverse the vdev tree calling the supplied function.
	* We continue to walk the tree until we either have walked all
	* children or we receive a non-NULL return from the callback.
	* If a NULL callback is passed, then we just return back the first
	* leaf vdev we encounter.
	*/
	static vdev_t *
	vdev_walk_tree(vdev_t vd, vdev_t (func)(vdev_t , void ), void arg)
	{
	uint_t c;

	if (vd->vdev_ops->vdev_op_leaf) {
	if (func == NULL)
	return (vd);
	else
	return (func(vd, arg));
	}

	for (c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL)
	return (cvd);
	}
	return (NULL);
	}

	/*
	* Verify that dynamic LUN growth works as expected.
	*/
	void
	ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	vdev_t vd, tvd;
	metaslab_class_t *mc;
	metaslab_group_t *mg;
	size_t psize, newsize;
	uint64_t top;
	uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count;

	mutex_enter(&ztest_checkpoint_lock);
	mutex_enter(&ztest_vdev_lock);
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	/*
	* If there is a vdev removal in progress, it could complete while
	* we are running, in which case we would not be able to verify
	* that the metaslab_class space increased (because it decreases
	* when the device removal completes).
	*/
	if (ztest_device_removal_active) {
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	top = ztest_random_vdev_top(spa, B_TRUE);

	tvd = spa->spa_root_vdev->vdev_child[top];
	mg = tvd->vdev_mg;
	mc = mg->mg_class;
	old_ms_count = tvd->vdev_ms_count;
	old_class_space = metaslab_class_get_space(mc);

	/*
	* Determine the size of the first leaf vdev associated with
	* our top-level device.
	*/
	vd = vdev_walk_tree(tvd, NULL, NULL);
	ASSERT3P(vd, !=, NULL);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	psize = vd->vdev_psize;

	/*
	* We only try to expand the vdev if it's healthy, less than 4x its
	* original size, and it has a valid psize.
	*/
	if (tvd->vdev_state != VDEV_STATE_HEALTHY \|\|
	psize == 0 \|\| psize >= 4 * ztest_opts.zo_vdev_size) {
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}
	ASSERT3U(psize, >, 0);
	newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE);
	ASSERT3U(newsize, >, psize);

	if (ztest_opts.zo_verbose >= 6) {
	(void) printf("Expanding LUN %s from %lu to %lu\n",
	vd->vdev_path, (ulong_t)psize, (ulong_t)newsize);
	}

	/*
	* Growing the vdev is a two step process:
	* 1). expand the physical size (i.e. relabel)
	* 2). online the vdev to create the new metaslabs
	*/
	if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL \|\|
	vdev_walk_tree(tvd, online_vdev, NULL) != NULL \|\|
	tvd->vdev_state != VDEV_STATE_HEALTHY) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Could not expand LUN because "
	"the vdev configuration changed.\n");
	}
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	spa_config_exit(spa, SCL_STATE, spa);

	/*
	* Expanding the LUN will update the config asynchronously,
	* thus we must wait for the async thread to complete any
	* pending tasks before proceeding.
	*/
	for (;;) {
	boolean_t done;
	mutex_enter(&spa->spa_async_lock);
	done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks);
	mutex_exit(&spa->spa_async_lock);
	if (done)
	break;
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 100);
	}

	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	tvd = spa->spa_root_vdev->vdev_child[top];
	new_ms_count = tvd->vdev_ms_count;
	new_class_space = metaslab_class_get_space(mc);

	if (tvd->vdev_mg != mg \|\| mg->mg_class != mc) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Could not verify LUN expansion due to "
	"intervening vdev offline or remove.\n");
	}
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	/*
	* Make sure we were able to grow the vdev.
	*/
	if (new_ms_count <= old_ms_count) {
	fatal(B_FALSE,
	"LUN expansion failed: ms_count %"PRIu64" < %"PRIu64"\n",
	old_ms_count, new_ms_count);
	}

	/*
	* Make sure we were able to grow the pool.
	*/
	if (new_class_space <= old_class_space) {
	fatal(B_FALSE,
	"LUN expansion failed: class_space %"PRIu64" < %"PRIu64"\n",
	old_class_space, new_class_space);
	}

	if (ztest_opts.zo_verbose >= 5) {
	char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ];

	nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf));
	nicenum(new_class_space, newnumbuf, sizeof (newnumbuf));
	(void) printf("%s grew from %s to %s\n",
	spa->spa_name, oldnumbuf, newnumbuf);
	}

	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	}

	/*
	* Verify that dmu_objset_{create,destroy,open,close} work as expected.
	*/
	static void
	ztest_objset_create_cb(objset_t os, void arg, cred_t cr, dmu_tx_t tx)
	{
	(void) arg, (void) cr;

	/*
	* Create the objects common to all ztest datasets.
	*/
	VERIFY0(zap_create_claim(os, ZTEST_DIROBJ,
	DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx));
	}

	static int
	ztest_dataset_create(char *dsname)
	{
	int err;
	uint64_t rand;
	dsl_crypto_params_t *dcp = NULL;

	/*
	* 50% of the time, we create encrypted datasets
	* using a random cipher suite and a hard-coded
	* wrapping key.
	*/
	rand = ztest_random(2);
	if (rand != 0) {
	nvlist_t *crypto_args = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	/* slight bias towards the default cipher suite */
	rand = ztest_random(ZIO_CRYPT_FUNCTIONS);
	if (rand < ZIO_CRYPT_AES_128_CCM)
	rand = ZIO_CRYPT_ON;

	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand);
	fnvlist_add_uint8_array(crypto_args, "wkeydata",
	(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);

	/*
	* These parameters aren't really used by the kernel. They
	* are simply stored so that userspace knows how to load
	* the wrapping key.
	*/
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW);
	fnvlist_add_string(props,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt");
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL);
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL);

	VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props,
	crypto_args, &dcp));

	/*
	* Cycle through all available encryption implementations
	* to verify interoperability.
	*/
	VERIFY0(gcm_impl_set("cycle"));
	VERIFY0(aes_impl_set("cycle"));

	fnvlist_free(crypto_args);
	fnvlist_free(props);
	}

	err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp,
	ztest_objset_create_cb, NULL);
	dsl_crypto_params_free(dcp, !!err);

	rand = ztest_random(100);
	if (err \|\| rand < 80)
	return (err);

	if (ztest_opts.zo_verbose >= 5)
	(void) printf("Setting dataset %s to sync always\n", dsname);
	return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC,
	ZFS_SYNC_ALWAYS, B_FALSE));
	}

	static int
	ztest_objset_destroy_cb(const char name, void arg)
	{
	(void) arg;
	objset_t *os;
	dmu_object_info_t doi;
	int error;

	/*
	* Verify that the dataset contains a directory object.
	*/
	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
	B_TRUE, FTAG, &os));
	error = dmu_object_info(os, ZTEST_DIROBJ, &doi);
	if (error != ENOENT) {
	/* We could have crashed in the middle of destroying it */
	ASSERT0(error);
	ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER);
	ASSERT3S(doi.doi_physical_blocks_512, >=, 0);
	}
	dmu_objset_disown(os, B_TRUE, FTAG);

	/*
	* Destroy the dataset.
	*/
	if (strchr(name, '@') != NULL) {
	error = dsl_destroy_snapshot(name, B_TRUE);
	if (error != ECHRNG) {
	/*
	* The program was executed, but encountered a runtime
	* error, such as insufficient slop, or a hold on the
	* dataset.
	*/
	ASSERT0(error);
	}
	} else {
	error = dsl_destroy_head(name);
	if (error == ENOSPC) {
	/* There could be checkpoint or insufficient slop */
	ztest_record_enospc(FTAG);
	} else if (error != EBUSY) {
	/* There could be a hold on this dataset */
	ASSERT0(error);
	}
	}
	return (0);
	}

	static boolean_t
	ztest_snapshot_create(char *osname, uint64_t id)
	{
	char snapname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	(void) snprintf(snapname, sizeof (snapname), "%"PRIu64"", id);

	error = dmu_objset_snapshot_one(osname, snapname);
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (B_FALSE);
	}
	if (error != 0 && error != EEXIST) {
	fatal(B_FALSE, "ztest_snapshot_create(%s@%s) = %d", osname,
	snapname, error);
	}
	return (B_TRUE);
	}

	static boolean_t
	ztest_snapshot_destroy(char *osname, uint64_t id)
	{
	char snapname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	(void) snprintf(snapname, sizeof (snapname), "%s@%"PRIu64"",
	osname, id);

	error = dsl_destroy_snapshot(snapname, B_FALSE);
	if (error != 0 && error != ENOENT)
	fatal(B_FALSE, "ztest_snapshot_destroy(%s) = %d",
	snapname, error);
	return (B_TRUE);
	}

	void
	ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd;
	ztest_ds_t *zdtmp;
	int iters;
	int error;
	objset_t os, os2;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	zilog_t *zilog;
	int i;

	zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	(void) snprintf(name, sizeof (name), "%s/temp_%"PRIu64"",
	ztest_opts.zo_pool, id);

	/*
	* If this dataset exists from a previous run, process its replay log
	* half of the time. If we don't replay it, then dsl_destroy_head()
	* (invoked from ztest_objset_destroy_cb()) should just throw it away.
	*/
	if (ztest_random(2) == 0 &&
	ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
	B_TRUE, FTAG, &os) == 0) {
	ztest_zd_init(zdtmp, NULL, os);
	zil_replay(os, zdtmp, ztest_replay_vector);
	ztest_zd_fini(zdtmp);
	dmu_objset_disown(os, B_TRUE, FTAG);
	}

	/*
	* There may be an old instance of the dataset we're about to
	* create lying around from a previous run. If so, destroy it
	* and all of its snapshots.
	*/
	(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
	DS_FIND_CHILDREN \| DS_FIND_SNAPSHOTS);

	/*
	* Verify that the destroyed dataset is no longer in the namespace.
	*/
	VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
	B_TRUE, FTAG, &os));

	/*
	* Verify that we can create a new dataset.
	*/
	error = ztest_dataset_create(name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", name, error);
	}

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE,
	FTAG, &os));

	ztest_zd_init(zdtmp, NULL, os);

	/*
	* Open the intent log for it.
	*/
	zilog = zil_open(os, ztest_get_data, NULL);

	/*
	* Put some objects in there, do a little I/O to them,
	* and randomly take a couple of snapshots along the way.
	*/
	iters = ztest_random(5);
	for (i = 0; i < iters; i++) {
	ztest_dmu_object_alloc_free(zdtmp, id);
	if (ztest_random(iters) == 0)
	(void) ztest_snapshot_create(name, i);
	}

	/*
	* Verify that we cannot create an existing dataset.
	*/
	VERIFY3U(EEXIST, ==,
	dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL));

	/*
	* Verify that we can hold an objset that is also owned.
	*/
	VERIFY0(dmu_objset_hold(name, FTAG, &os2));
	dmu_objset_rele(os2, FTAG);

	/*
	* Verify that we cannot own an objset that is already owned.
	*/
	VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER,
	B_FALSE, B_TRUE, FTAG, &os2));

	zil_close(zilog);
	dmu_objset_disown(os, B_TRUE, FTAG);
	ztest_zd_fini(zdtmp);
	out:
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	umem_free(zdtmp, sizeof (ztest_ds_t));
	}

	/*
	* Verify that dmu_snapshot_{create,destroy,open,close} work as expected.
	*/
	void
	ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	(void) ztest_snapshot_destroy(zd->zd_name, id);
	(void) ztest_snapshot_create(zd->zd_name, id);
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Cleanup non-standard snapshots and clones.
	*/
	static void
	ztest_dsl_dataset_cleanup(char *osname, uint64_t id)
	{
	char *snap1name;
	char *clone1name;
	char *snap2name;
	char *clone2name;
	char *snap3name;
	int error;

	snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);

	(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
	osname, id);
	(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
	osname, id);
	(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
	clone1name, id);
	(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
	osname, id);
	(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
	clone1name, id);

	error = dsl_destroy_head(clone2name);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone2name, error);
	error = dsl_destroy_snapshot(snap3name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap3name, error);
	error = dsl_destroy_snapshot(snap2name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap2name, error);
	error = dsl_destroy_head(clone1name);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone1name, error);
	error = dsl_destroy_snapshot(snap1name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap1name, error);

	umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
	}

	/*
	* Verify dsl_dataset_promote handles EBUSY
	*/
	void
	ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os;
	char *snap1name;
	char *clone1name;
	char *snap2name;
	char *clone2name;
	char *snap3name;
	char *osname = zd->zd_name;
	int error;

	snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	ztest_dsl_dataset_cleanup(osname, id);

	(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
	osname, id);
	(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
	osname, id);
	(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
	clone1name, id);
	(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
	osname, id);
	(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
	clone1name, id);

	error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_take_snapshot(%s) = %d", snap1name, error);
	}

	error = dmu_objset_clone(clone1name, snap1name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone1name, error);
	}

	error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap2name, error);
	}

	error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap3name, error);
	}

	error = dmu_objset_clone(clone2name, snap3name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone2name, error);
	}

	error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE,
	FTAG, &os);
	if (error)
	fatal(B_FALSE, "dmu_objset_own(%s) = %d", snap2name, error);
	error = dsl_dataset_promote(clone2name, NULL);
	if (error == ENOSPC) {
	dmu_objset_disown(os, B_TRUE, FTAG);
	ztest_record_enospc(FTAG);
	goto out;
	}
	if (error != EBUSY)
	fatal(B_FALSE, "dsl_dataset_promote(%s), %d, not EBUSY",
	clone2name, error);
	dmu_objset_disown(os, B_TRUE, FTAG);

	out:
	ztest_dsl_dataset_cleanup(osname, id);

	(void) pthread_rwlock_unlock(&ztest_name_lock);

	umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 4

	/*
	* Verify that dmu_object_{alloc,free} work as expected.
	*/
	void
	ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id)
	{
	ztest_od_t *od;
	int batchsize;
	int size;
	int b;

	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);
	batchsize = OD_ARRAY_SIZE;

	for (b = 0; b < batchsize; b++)
	ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER,
	0, 0, 0);

	/*
	* Destroy the previous batch of objects, create a new batch,
	* and do some I/O on the new objects.
	*/
	if (ztest_object_init(zd, od, size, B_TRUE) != 0)
	return;

	while (ztest_random(4 * batchsize) != 0)
	ztest_io(zd, od[ztest_random(batchsize)].od_object,
	ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);

	umem_free(od, size);
	}

	/*
	* Rewind the global allocator to verify object allocation backfilling.
	*/
	void
	ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	- int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	+ uint_t dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	uint64_t object;

	/*
	* Rewind the global allocator randomly back to a lower object number
	* to force backfilling and reclamation of recently freed dnodes.
	*/
	mutex_enter(&os->os_obj_lock);
	object = ztest_random(os->os_obj_next_chunk);
	os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk);
	mutex_exit(&os->os_obj_lock);
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 2

	/*
	* Verify that dmu_{read,write} work as expected.
	*/
	void
	ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id)
	{
	int size;
	ztest_od_t *od;

	objset_t *os = zd->zd_os;
	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);
	dmu_tx_t *tx;
	int freeit, error;
	uint64_t i, n, s, txg;
	bufwad_t packbuf, bigbuf, pack, bigH, *bigT;
	uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
	uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t);
	uint64_t regions = 997;
	uint64_t stride = 123456789ULL;
	uint64_t width = 40;
	int free_percent = 5;

	/*
	* This test uses two objects, packobj and bigobj, that are always
	* updated together (i.e. in the same tx) so that their contents are
	* in sync and can be compared. Their contents relate to each other
	* in a simple way: packobj is a dense array of 'bufwad' structures,
	* while bigobj is a sparse array of the same bufwads. Specifically,
	* for any index n, there are three bufwads that should be identical:
	*
	* packobj, at offset n * sizeof (bufwad_t)
	* bigobj, at the head of the nth chunk
	* bigobj, at the tail of the nth chunk
	*
	* The chunk size is arbitrary. It doesn't have to be a power of two,
	* and it doesn't have any relation to the object blocksize.
	* The only requirement is that it can hold at least two bufwads.
	*
	* Normally, we write the bufwad to each of these locations.
	* However, free_percent of the time we instead write zeroes to
	* packobj and perform a dmu_free_range() on bigobj. By comparing
	* bigobj to packobj, we can verify that the DMU is correctly
	* tracking which parts of an object are allocated and free,
	* and that the contents of the allocated blocks are correct.
	*/

	/*
	* Read the directory info. If it's the first time, set things up.
	*/
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize);
	ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
	chunksize);

	if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
	umem_free(od, size);
	return;
	}

	bigobj = od[0].od_object;
	packobj = od[1].od_object;
	chunksize = od[0].od_gen;
	ASSERT3U(chunksize, ==, od[1].od_gen);

	/*
	* Prefetch a random chunk of the big object.
	* Our aim here is to get some async reads in flight
	* for blocks that we may free below; the DMU should
	* handle this race correctly.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(2 * width - 1);
	dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize,
	ZIO_PRIORITY_SYNC_READ);

	/*
	* Pick a random index and compute the offsets into packobj and bigobj.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(width - 1);

	packoff = n * sizeof (bufwad_t);
	packsize = s * sizeof (bufwad_t);

	bigoff = n * chunksize;
	bigsize = s * chunksize;

	packbuf = umem_alloc(packsize, UMEM_NOFAIL);
	bigbuf = umem_alloc(bigsize, UMEM_NOFAIL);

	/*
	* free_percent of the time, free a range of bigobj rather than
	* overwriting it.
	*/
	freeit = (ztest_random(100) < free_percent);

	/*
	* Read the current contents of our objects.
	*/
	error = dmu_read(os, packobj, packoff, packsize, packbuf,
	DMU_READ_PREFETCH);
	ASSERT0(error);
	error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf,
	DMU_READ_PREFETCH);
	ASSERT0(error);

	/*
	* Get a tx for the mods to both packobj and bigobj.
	*/
	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, packobj, packoff, packsize);

	if (freeit)
	dmu_tx_hold_free(tx, bigobj, bigoff, bigsize);
	else
	dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);

	/* This accounts for setting the checksum/compression. */
	dmu_tx_hold_bonus(tx, bigobj);

	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(od, size);
	return;
	}

	enum zio_checksum cksum;
	do {
	cksum = (enum zio_checksum)
	ztest_random_dsl_prop(ZFS_PROP_CHECKSUM);
	} while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS);
	dmu_object_set_checksum(os, bigobj, cksum, tx);

	enum zio_compress comp;
	do {
	comp = (enum zio_compress)
	ztest_random_dsl_prop(ZFS_PROP_COMPRESSION);
	} while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS);
	dmu_object_set_compress(os, bigobj, comp, tx);

	/*
	* For each index from n to n + s, verify that the existing bufwad
	* in packobj matches the bufwads at the head and tail of the
	* corresponding chunk in bigobj. Then update all three bufwads
	* with the new values we want to write out.
	*/
	for (i = 0; i < s; i++) {
	/* LINTED */
	pack = (bufwad_t )((char )packbuf + i * sizeof (bufwad_t));
	/* LINTED */
	bigH = (bufwad_t )((char )bigbuf + i * chunksize);
	/* LINTED */
	bigT = (bufwad_t )((char )bigH + chunksize) - 1;

	ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
	ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);

	if (pack->bw_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIx64", open txg is %"PRIx64"",
	pack->bw_txg, txg);

	if (pack->bw_data != 0 && pack->bw_index != n + i)
	fatal(B_FALSE, "wrong index: "
	"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
	pack->bw_index, n, i);

	if (memcmp(pack, bigH, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
	pack, bigH);

	if (memcmp(pack, bigT, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
	pack, bigT);

	if (freeit) {
	memset(pack, 0, sizeof (bufwad_t));
	} else {
	pack->bw_index = n + i;
	pack->bw_txg = txg;
	pack->bw_data = 1 + ztest_random(-2ULL);
	}
	bigH = pack;
	bigT = pack;
	}

	/*
	* We've verified all the old bufwads, and made new ones.
	* Now write them out.
	*/
	dmu_write(os, packobj, packoff, packsize, packbuf, tx);

	if (freeit) {
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("freeing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	VERIFY0(dmu_free_range(os, bigobj, bigoff, bigsize, tx));
	} else {
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("writing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx);
	}

	dmu_tx_commit(tx);

	/*
	* Sanity check the stuff we just wrote.
	*/
	{
	void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
	void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_read(os, packobj, packoff,
	packsize, packcheck, DMU_READ_PREFETCH));
	VERIFY0(dmu_read(os, bigobj, bigoff,
	bigsize, bigcheck, DMU_READ_PREFETCH));

	ASSERT0(memcmp(packbuf, packcheck, packsize));
	ASSERT0(memcmp(bigbuf, bigcheck, bigsize));

	umem_free(packcheck, packsize);
	umem_free(bigcheck, bigsize);
	}

	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(od, size);
	}

	static void
	compare_and_update_pbbufs(uint64_t s, bufwad_t packbuf, bufwad_t bigbuf,
	uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg)
	{
	uint64_t i;
	bufwad_t *pack;
	bufwad_t *bigH;
	bufwad_t *bigT;

	/*
	* For each index from n to n + s, verify that the existing bufwad
	* in packobj matches the bufwads at the head and tail of the
	* corresponding chunk in bigobj. Then update all three bufwads
	* with the new values we want to write out.
	*/
	for (i = 0; i < s; i++) {
	/* LINTED */
	pack = (bufwad_t )((char )packbuf + i * sizeof (bufwad_t));
	/* LINTED */
	bigH = (bufwad_t )((char )bigbuf + i * chunksize);
	/* LINTED */
	bigT = (bufwad_t )((char )bigH + chunksize) - 1;

	ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
	ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);

	if (pack->bw_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIx64", open txg is %"PRIx64"",
	pack->bw_txg, txg);

	if (pack->bw_data != 0 && pack->bw_index != n + i)
	fatal(B_FALSE, "wrong index: "
	"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
	pack->bw_index, n, i);

	if (memcmp(pack, bigH, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
	pack, bigH);

	if (memcmp(pack, bigT, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
	pack, bigT);

	pack->bw_index = n + i;
	pack->bw_txg = txg;
	pack->bw_data = 1 + ztest_random(-2ULL);

	bigH = pack;
	bigT = pack;
	}
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 2

	void
	ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	dmu_tx_t *tx;
	uint64_t i;
	int error;
	int size;
	uint64_t n, s, txg;
	bufwad_t packbuf, bigbuf;
	uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
	uint64_t blocksize = ztest_random_blocksize();
	uint64_t chunksize = blocksize;
	uint64_t regions = 997;
	uint64_t stride = 123456789ULL;
	uint64_t width = 9;
	dmu_buf_t *bonus_db;
	arc_buf_t **bigbuf_arcbufs;
	dmu_object_info_t doi;

	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);

	/*
	* This test uses two objects, packobj and bigobj, that are always
	* updated together (i.e. in the same tx) so that their contents are
	* in sync and can be compared. Their contents relate to each other
	* in a simple way: packobj is a dense array of 'bufwad' structures,
	* while bigobj is a sparse array of the same bufwads. Specifically,
	* for any index n, there are three bufwads that should be identical:
	*
	* packobj, at offset n * sizeof (bufwad_t)
	* bigobj, at the head of the nth chunk
	* bigobj, at the tail of the nth chunk
	*
	* The chunk size is set equal to bigobj block size so that
	* dmu_assign_arcbuf_by_dbuf() can be tested for object updates.
	*/

	/*
	* Read the directory info. If it's the first time, set things up.
	*/
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);
	ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
	chunksize);


	if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
	umem_free(od, size);
	return;
	}

	bigobj = od[0].od_object;
	packobj = od[1].od_object;
	blocksize = od[0].od_blocksize;
	chunksize = blocksize;
	ASSERT3U(chunksize, ==, od[1].od_gen);

	VERIFY0(dmu_object_info(os, bigobj, &doi));
	VERIFY(ISP2(doi.doi_data_block_size));
	VERIFY3U(chunksize, ==, doi.doi_data_block_size);
	VERIFY3U(chunksize, >=, 2 * sizeof (bufwad_t));

	/*
	* Pick a random index and compute the offsets into packobj and bigobj.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(width - 1);

	packoff = n * sizeof (bufwad_t);
	packsize = s * sizeof (bufwad_t);

	bigoff = n * chunksize;
	bigsize = s * chunksize;

	packbuf = umem_zalloc(packsize, UMEM_NOFAIL);
	bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_bonus_hold(os, bigobj, FTAG, &bonus_db));

	bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL);

	/*
	* Iteration 0 test zcopy for DB_UNCACHED dbufs.
	* Iteration 1 test zcopy to already referenced dbufs.
	* Iteration 2 test zcopy to dirty dbuf in the same txg.
	* Iteration 3 test zcopy to dbuf dirty in previous txg.
	* Iteration 4 test zcopy when dbuf is no longer dirty.
	* Iteration 5 test zcopy when it can't be done.
	* Iteration 6 one more zcopy write.
	*/
	for (i = 0; i < 7; i++) {
	uint64_t j;
	uint64_t off;

	/*
	* In iteration 5 (i == 5) use arcbufs
	* that don't match bigobj blksz to test
	* dmu_assign_arcbuf_by_dbuf() when it can't directly
	* assign an arcbuf to a dbuf.
	*/
	for (j = 0; j < s; j++) {
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	bigbuf_arcbufs[j] =
	dmu_request_arcbuf(bonus_db, chunksize);
	} else {
	bigbuf_arcbufs[2 * j] =
	dmu_request_arcbuf(bonus_db, chunksize / 2);
	bigbuf_arcbufs[2 * j + 1] =
	dmu_request_arcbuf(bonus_db, chunksize / 2);
	}
	}

	/*
	* Get a tx for the mods to both packobj and bigobj.
	*/
	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, packobj, packoff, packsize);
	dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);

	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	for (j = 0; j < s; j++) {
	if (i != 5 \|\|
	chunksize < (SPA_MINBLOCKSIZE * 2)) {
	dmu_return_arcbuf(bigbuf_arcbufs[j]);
	} else {
	dmu_return_arcbuf(
	bigbuf_arcbufs[2 * j]);
	dmu_return_arcbuf(
	bigbuf_arcbufs[2 * j + 1]);
	}
	}
	umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
	umem_free(od, size);
	dmu_buf_rele(bonus_db, FTAG);
	return;
	}

	/*
	* 50% of the time don't read objects in the 1st iteration to
	* test dmu_assign_arcbuf_by_dbuf() for the case when there are
	* no existing dbufs for the specified offsets.
	*/
	if (i != 0 \|\| ztest_random(2) != 0) {
	error = dmu_read(os, packobj, packoff,
	packsize, packbuf, DMU_READ_PREFETCH);
	ASSERT0(error);
	error = dmu_read(os, bigobj, bigoff, bigsize,
	bigbuf, DMU_READ_PREFETCH);
	ASSERT0(error);
	}
	compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize,
	n, chunksize, txg);

	/*
	* We've verified all the old bufwads, and made new ones.
	* Now write them out.
	*/
	dmu_write(os, packobj, packoff, packsize, packbuf, tx);
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("writing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	for (off = bigoff, j = 0; j < s; j++, off += chunksize) {
	dmu_buf_t *dbt;
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	memcpy(bigbuf_arcbufs[j]->b_data,
	(caddr_t)bigbuf + (off - bigoff),
	chunksize);
	} else {
	memcpy(bigbuf_arcbufs[2 * j]->b_data,
	(caddr_t)bigbuf + (off - bigoff),
	chunksize / 2);
	memcpy(bigbuf_arcbufs[2 * j + 1]->b_data,
	(caddr_t)bigbuf + (off - bigoff) +
	chunksize / 2,
	chunksize / 2);
	}

	if (i == 1) {
	VERIFY(dmu_buf_hold(os, bigobj, off,
	FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0);
	}
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off, bigbuf_arcbufs[j], tx));
	} else {
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off, bigbuf_arcbufs[2 * j], tx));
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off + chunksize / 2,
	bigbuf_arcbufs[2 * j + 1], tx));
	}
	if (i == 1) {
	dmu_buf_rele(dbt, FTAG);
	}
	}
	dmu_tx_commit(tx);

	/*
	* Sanity check the stuff we just wrote.
	*/
	{
	void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
	void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_read(os, packobj, packoff,
	packsize, packcheck, DMU_READ_PREFETCH));
	VERIFY0(dmu_read(os, bigobj, bigoff,
	bigsize, bigcheck, DMU_READ_PREFETCH));

	ASSERT0(memcmp(packbuf, packcheck, packsize));
	ASSERT0(memcmp(bigbuf, bigcheck, bigsize));

	umem_free(packcheck, packsize);
	umem_free(bigcheck, bigsize);
	}
	if (i == 2) {
	txg_wait_open(dmu_objset_pool(os), 0, B_TRUE);
	} else if (i == 3) {
	txg_wait_synced(dmu_objset_pool(os), 0);
	}
	}

	dmu_buf_rele(bonus_db, FTAG);
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
	umem_free(od, size);
	}

	void
	ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	ztest_od_t *od;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	uint64_t offset = (1ULL << (ztest_random(20) + 43)) +
	(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);

	/*
	* Have multiple threads write to large offsets in an object
	* to verify that parallel writes to an object -- even to the
	* same blocks within the object -- doesn't cause any trouble.
	*/
	ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0)
	return;

	while (ztest_random(10) != 0)
	ztest_io(zd, od->od_object, offset);

	umem_free(od, sizeof (ztest_od_t));
	}

	void
	ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id)
	{
	ztest_od_t *od;
	uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) +
	(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
	uint64_t count = ztest_random(20) + 1;
	uint64_t blocksize = ztest_random_blocksize();
	void *data;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);

	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	ztest_prealloc(zd, od->od_object, offset, count * blocksize);

	data = umem_zalloc(blocksize, UMEM_NOFAIL);

	while (ztest_random(count) != 0) {
	uint64_t randoff = offset + (ztest_random(count) * blocksize);
	if (ztest_write(zd, od->od_object, randoff, blocksize,
	data) != 0)
	break;
	while (ztest_random(4) != 0)
	ztest_io(zd, od->od_object, randoff);
	}

	umem_free(data, blocksize);
	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Verify that zap_{create,destroy,add,remove,update} work as expected.
	*/
	#define ZTEST_ZAP_MIN_INTS 1
	#define ZTEST_ZAP_MAX_INTS 4
	#define ZTEST_ZAP_MAX_PROPS 1000

	void
	ztest_zap(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t object;
	uint64_t txg, last_txg;
	uint64_t value[ZTEST_ZAP_MAX_INTS];
	uint64_t zl_ints, zl_intsize, prop;
	int i, ints;
	dmu_tx_t *tx;
	char propname[100], txgname[100];
	int error;
	const char *const hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" };

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0)
	goto out;

	object = od->od_object;

	/*
	* Generate a known hash collision, and verify that
	* we can lookup and remove both entries.
	*/
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	for (i = 0; i < 2; i++) {
	value[i] = i;
	VERIFY0(zap_add(os, object, hc[i], sizeof (uint64_t),
	1, &value[i], tx));
	}
	for (i = 0; i < 2; i++) {
	VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i],
	sizeof (uint64_t), 1, &value[i], tx));
	VERIFY0(
	zap_length(os, object, hc[i], &zl_intsize, &zl_ints));
	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, 1);
	}
	for (i = 0; i < 2; i++) {
	VERIFY0(zap_remove(os, object, hc[i], tx));
	}
	dmu_tx_commit(tx);

	/*
	* Generate a bunch of random entries.
	*/
	ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS);

	prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
	(void) sprintf(propname, "prop_%"PRIu64"", prop);
	(void) sprintf(txgname, "txg_%"PRIu64"", prop);
	memset(value, 0, sizeof (value));
	last_txg = 0;

	/*
	* If these zap entries already exist, validate their contents.
	*/
	error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
	if (error == 0) {
	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, 1);

	VERIFY0(zap_lookup(os, object, txgname, zl_intsize,
	zl_ints, &last_txg));

	VERIFY0(zap_length(os, object, propname, &zl_intsize,
	&zl_ints));

	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, ints);

	VERIFY0(zap_lookup(os, object, propname, zl_intsize,
	zl_ints, value));

	for (i = 0; i < ints; i++) {
	ASSERT3U(value[i], ==, last_txg + object + i);
	}
	} else {
	ASSERT3U(error, ==, ENOENT);
	}

	/*
	* Atomically update two entries in our zap object.
	* The first is named txg_%llu, and contains the txg
	* in which the property was last updated. The second
	* is named prop_%llu, and the nth element of its value
	* should be txg + object + n.
	*/
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;

	if (last_txg > txg)
	fatal(B_FALSE, "zap future leak: old %"PRIu64" new %"PRIu64"",
	last_txg, txg);

	for (i = 0; i < ints; i++)
	value[i] = txg + object + i;

	VERIFY0(zap_update(os, object, txgname, sizeof (uint64_t),
	1, &txg, tx));
	VERIFY0(zap_update(os, object, propname, sizeof (uint64_t),
	ints, value, tx));

	dmu_tx_commit(tx);

	/*
	* Remove a random pair of entries.
	*/
	prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
	(void) sprintf(propname, "prop_%"PRIu64"", prop);
	(void) sprintf(txgname, "txg_%"PRIu64"", prop);

	error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);

	if (error == ENOENT)
	goto out;

	ASSERT0(error);

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	VERIFY0(zap_remove(os, object, txgname, tx));
	VERIFY0(zap_remove(os, object, propname, tx));
	dmu_tx_commit(tx);
	out:
	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Test case to test the upgrading of a microzap to fatzap.
	*/
	void
	ztest_fzap(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t object, txg, value;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0)
	goto out;
	object = od->od_object;

	/*
	* Add entries to this ZAP and make sure it spills over
	* and gets upgraded to a fatzap. Also, since we are adding
	* 2050 entries we should see ptrtbl growth and leaf-block split.
	*/
	for (value = 0; value < 2050; value++) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	dmu_tx_t *tx;
	int error;

	(void) snprintf(name, sizeof (name), "fzap-%"PRIu64"-%"PRIu64"",
	id, value);

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, name);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	error = zap_add(os, object, name, sizeof (uint64_t), 1,
	&value, tx);
	ASSERT(error == 0 \|\| error == EEXIST);
	dmu_tx_commit(tx);
	}
	out:
	umem_free(od, sizeof (ztest_od_t));
	}

	void
	ztest_zap_parallel(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc;
	dmu_tx_t *tx;
	int i, namelen, error;
	int micro = ztest_random(2);
	char name[20], string_value[20];
	void *data;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	object = od->od_object;

	/*
	* Generate a random name of the form 'xxx.....' where each
	* x is a random printable character and the dots are dots.
	* There are 94 such characters, and the name length goes from
	* 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names.
	*/
	namelen = ztest_random(sizeof (name) - 5) + 5 + 1;

	for (i = 0; i < 3; i++)
	name[i] = '!' + ztest_random('~' - '!' + 1);
	for (; i < namelen - 1; i++)
	name[i] = '.';
	name[i] = '\0';

	if ((namelen & 1) \|\| micro) {
	wsize = sizeof (txg);
	wc = 1;
	data = &txg;
	} else {
	wsize = 1;
	wc = namelen;
	data = string_value;
	}

	count = -1ULL;
	VERIFY0(zap_count(os, object, &count));
	ASSERT3S(count, !=, -1ULL);

	/*
	* Select an operation: length, lookup, add, update, remove.
	*/
	i = ztest_random(5);

	if (i >= 2) {
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}
	memcpy(string_value, name, namelen);
	} else {
	tx = NULL;
	txg = 0;
	memset(string_value, 0, namelen);
	}

	switch (i) {

	case 0:
	error = zap_length(os, object, name, &zl_wsize, &zl_wc);
	if (error == 0) {
	ASSERT3U(wsize, ==, zl_wsize);
	ASSERT3U(wc, ==, zl_wc);
	} else {
	ASSERT3U(error, ==, ENOENT);
	}
	break;

	case 1:
	error = zap_lookup(os, object, name, wsize, wc, data);
	if (error == 0) {
	if (data == string_value &&
	memcmp(name, data, namelen) != 0)
	fatal(B_FALSE, "name '%s' != val '%s' len %d",
	name, (char *)data, namelen);
	} else {
	ASSERT3U(error, ==, ENOENT);
	}
	break;

	case 2:
	error = zap_add(os, object, name, wsize, wc, data, tx);
	ASSERT(error == 0 \|\| error == EEXIST);
	break;

	case 3:
	VERIFY0(zap_update(os, object, name, wsize, wc, data, tx));
	break;

	case 4:
	error = zap_remove(os, object, name, tx);
	ASSERT(error == 0 \|\| error == ENOENT);
	break;
	}

	if (tx != NULL)
	dmu_tx_commit(tx);

	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Commit callback data.
	*/
	typedef struct ztest_cb_data {
	list_node_t zcd_node;
	uint64_t zcd_txg;
	int zcd_expected_err;
	boolean_t zcd_added;
	boolean_t zcd_called;
	spa_t *zcd_spa;
	} ztest_cb_data_t;

	/* This is the actual commit callback function */
	static void
	ztest_commit_callback(void *arg, int error)
	{
	ztest_cb_data_t *data = arg;
	uint64_t synced_txg;

	VERIFY3P(data, !=, NULL);
	VERIFY3S(data->zcd_expected_err, ==, error);
	VERIFY(!data->zcd_called);

	synced_txg = spa_last_synced_txg(data->zcd_spa);
	if (data->zcd_txg > synced_txg)
	fatal(B_FALSE,
	"commit callback of txg %"PRIu64" called prematurely, "
	"last synced txg = %"PRIu64"\n",
	data->zcd_txg, synced_txg);

	data->zcd_called = B_TRUE;

	if (error == ECANCELED) {
	ASSERT0(data->zcd_txg);
	ASSERT(!data->zcd_added);

	/*
	* The private callback data should be destroyed here, but
	* since we are going to check the zcd_called field after
	* dmu_tx_abort(), we will destroy it there.
	*/
	return;
	}

	ASSERT(data->zcd_added);
	ASSERT3U(data->zcd_txg, !=, 0);

	(void) mutex_enter(&zcl.zcl_callbacks_lock);

	/* See if this cb was called more quickly */
	if ((synced_txg - data->zcd_txg) < zc_min_txg_delay)
	zc_min_txg_delay = synced_txg - data->zcd_txg;

	/* Remove our callback from the list */
	list_remove(&zcl.zcl_callbacks, data);

	(void) mutex_exit(&zcl.zcl_callbacks_lock);

	umem_free(data, sizeof (ztest_cb_data_t));
	}

	/* Allocate and initialize callback data structure */
	static ztest_cb_data_t *
	ztest_create_cb_data(objset_t *os, uint64_t txg)
	{
	ztest_cb_data_t *cb_data;

	cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL);

	cb_data->zcd_txg = txg;
	cb_data->zcd_spa = dmu_objset_spa(os);
	list_link_init(&cb_data->zcd_node);

	return (cb_data);
	}

	/*
	* Commit callback test.
	*/
	void
	ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	dmu_tx_t *tx;
	ztest_cb_data_t cb_data[3], tmp_cb;
	uint64_t old_txg, txg;
	int i, error = 0;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	tx = dmu_tx_create(os);

	cb_data[0] = ztest_create_cb_data(os, 0);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]);

	dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t));

	/* Every once in a while, abort the transaction on purpose */
	if (ztest_random(100) == 0)
	error = -1;

	if (!error)
	error = dmu_tx_assign(tx, TXG_NOWAIT);

	txg = error ? 0 : dmu_tx_get_txg(tx);

	cb_data[0]->zcd_txg = txg;
	cb_data[1] = ztest_create_cb_data(os, txg);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]);

	if (error) {
	/*
	* It's not a strict requirement to call the registered
	* callbacks from inside dmu_tx_abort(), but that's what
	* it's supposed to happen in the current implementation
	* so we will check for that.
	*/
	for (i = 0; i < 2; i++) {
	cb_data[i]->zcd_expected_err = ECANCELED;
	VERIFY(!cb_data[i]->zcd_called);
	}

	dmu_tx_abort(tx);

	for (i = 0; i < 2; i++) {
	VERIFY(cb_data[i]->zcd_called);
	umem_free(cb_data[i], sizeof (ztest_cb_data_t));
	}

	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	cb_data[2] = ztest_create_cb_data(os, txg);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]);

	/*
	* Read existing data to make sure there isn't a future leak.
	*/
	VERIFY0(dmu_read(os, od->od_object, 0, sizeof (uint64_t),
	&old_txg, DMU_READ_PREFETCH));

	if (old_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIu64", open txg is %"PRIu64"",
	old_txg, txg);

	dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx);

	(void) mutex_enter(&zcl.zcl_callbacks_lock);

	/*
	* Since commit callbacks don't have any ordering requirement and since
	* it is theoretically possible for a commit callback to be called
	* after an arbitrary amount of time has elapsed since its txg has been
	* synced, it is difficult to reliably determine whether a commit
	* callback hasn't been called due to high load or due to a flawed
	* implementation.
	*
	* In practice, we will assume that if after a certain number of txgs a
	* commit callback hasn't been called, then most likely there's an
	* implementation bug..
	*/
	tmp_cb = list_head(&zcl.zcl_callbacks);
	if (tmp_cb != NULL &&
	tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) {
	fatal(B_FALSE,
	"Commit callback threshold exceeded, "
	"oldest txg: %"PRIu64", open txg: %"PRIu64"\n",
	tmp_cb->zcd_txg, txg);
	}

	/*
	* Let's find the place to insert our callbacks.
	*
	* Even though the list is ordered by txg, it is possible for the
	* insertion point to not be the end because our txg may already be
	* quiescing at this point and other callbacks in the open txg
	* (from other objsets) may have sneaked in.
	*/
	tmp_cb = list_tail(&zcl.zcl_callbacks);
	while (tmp_cb != NULL && tmp_cb->zcd_txg > txg)
	tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb);

	/* Add the 3 callbacks to the list */
	for (i = 0; i < 3; i++) {
	if (tmp_cb == NULL)
	list_insert_head(&zcl.zcl_callbacks, cb_data[i]);
	else
	list_insert_after(&zcl.zcl_callbacks, tmp_cb,
	cb_data[i]);

	cb_data[i]->zcd_added = B_TRUE;
	VERIFY(!cb_data[i]->zcd_called);

	tmp_cb = cb_data[i];
	}

	zc_cb_counter += 3;

	(void) mutex_exit(&zcl.zcl_callbacks_lock);

	dmu_tx_commit(tx);

	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Visit each object in the dataset. Verify that its properties
	* are consistent what was stored in the block tag when it was created,
	* and that its unused bonus buffer space has not been overwritten.
	*/
	void
	ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	uint64_t obj;
	int err = 0;

	for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
	ztest_block_tag_t *bt = NULL;
	dmu_object_info_t doi;
	dmu_buf_t *db;

	ztest_object_lock(zd, obj, RL_READER);
	if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) {
	ztest_object_unlock(zd, obj);
	continue;
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_size >= sizeof (*bt))
	bt = ztest_bt_bonus(db);

	if (bt && bt->bt_magic == BT_MAGIC) {
	ztest_bt_verify(bt, os, obj, doi.doi_dnodesize,
	bt->bt_offset, bt->bt_gen, bt->bt_txg,
	bt->bt_crtxg);
	ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen);
	}

	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, obj);
	}
	}

	void
	ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	zfs_prop_t proplist[] = {
	ZFS_PROP_CHECKSUM,
	ZFS_PROP_COMPRESSION,
	ZFS_PROP_COPIES,
	ZFS_PROP_DEDUP
	};

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	for (int p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++)
	(void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p],
	ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2));

	VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE,
	ztest_random_blocksize(), (int)ztest_random(2)));

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	void
	ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	nvlist_t *props = NULL;

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	(void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2));

	VERIFY0(spa_prop_get(ztest_spa, &props));

	if (ztest_opts.zo_verbose >= 6)
	dump_nvlist(props, 4);

	fnvlist_free(props);

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	static int
	user_release_one(const char snapname, const char holdname)
	{
	nvlist_t snaps, holds;
	int error;

	snaps = fnvlist_alloc();
	holds = fnvlist_alloc();
	fnvlist_add_boolean(holds, holdname);
	fnvlist_add_nvlist(snaps, snapname, holds);
	fnvlist_free(holds);
	error = dsl_dataset_user_release(snaps, NULL);
	fnvlist_free(snaps);
	return (error);
	}

	/*
	* Test snapshot hold/release and deferred destroy.
	*/
	void
	ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id)
	{
	int error;
	objset_t *os = zd->zd_os;
	objset_t *origin;
	char snapname[100];
	char fullname[100];
	char clonename[100];
	char tag[100];
	char osname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *holds;

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	dmu_objset_name(os, osname);

	(void) snprintf(snapname, sizeof (snapname), "sh1_%"PRIu64"", id);
	(void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname);
	(void) snprintf(clonename, sizeof (clonename), "%s/ch1_%"PRIu64"",
	osname, id);
	(void) snprintf(tag, sizeof (tag), "tag_%"PRIu64"", id);

	/*
	* Clean up from any previous run.
	*/
	error = dsl_destroy_head(clonename);
	if (error != ENOENT)
	ASSERT0(error);
	error = user_release_one(fullname, tag);
	if (error != ESRCH && error != ENOENT)
	ASSERT0(error);
	error = dsl_destroy_snapshot(fullname, B_FALSE);
	if (error != ENOENT)
	ASSERT0(error);

	/*
	* Create snapshot, clone it, mark snap for deferred destroy,
	* destroy clone, verify snap was also destroyed.
	*/
	error = dmu_objset_snapshot_one(osname, snapname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_snapshot");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
	}

	error = dmu_objset_clone(clonename, fullname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_clone");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_clone(%s) = %d", clonename, error);
	}

	error = dsl_destroy_snapshot(fullname, B_TRUE);
	if (error) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
	fullname, error);
	}

	error = dsl_destroy_head(clonename);
	if (error)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clonename, error);

	error = dmu_objset_hold(fullname, FTAG, &origin);
	if (error != ENOENT)
	fatal(B_FALSE, "dmu_objset_hold(%s) = %d", fullname, error);

	/*
	* Create snapshot, add temporary hold, verify that we can't
	* destroy a held snapshot, mark for deferred destroy,
	* release hold, verify snapshot was destroyed.
	*/
	error = dmu_objset_snapshot_one(osname, snapname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_snapshot");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
	}

	holds = fnvlist_alloc();
	fnvlist_add_string(holds, fullname, tag);
	error = dsl_dataset_user_hold(holds, 0, NULL);
	fnvlist_free(holds);

	if (error == ENOSPC) {
	ztest_record_enospc("dsl_dataset_user_hold");
	goto out;
	} else if (error) {
	fatal(B_FALSE, "dsl_dataset_user_hold(%s, %s) = %u",
	fullname, tag, error);
	}

	error = dsl_destroy_snapshot(fullname, B_FALSE);
	if (error != EBUSY) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_FALSE) = %d",
	fullname, error);
	}

	error = dsl_destroy_snapshot(fullname, B_TRUE);
	if (error) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
	fullname, error);
	}

	error = user_release_one(fullname, tag);
	if (error)
	fatal(B_FALSE, "user_release_one(%s, %s) = %d",
	fullname, tag, error);

	VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT);

	out:
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Inject random faults into the on-disk data.
	*/
	void
	ztest_fault_inject(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	int fd;
	uint64_t offset;
	uint64_t leaves;
	uint64_t bad = 0x1990c0ffeedecadeull;
	uint64_t top, leaf;
	char *path0;
	char *pathrand;
	size_t fsize;
	int bshift = SPA_MAXBLOCKSHIFT + 2;
	int iters = 1000;
	int maxfaults;
	int mirror_save;
	vdev_t *vd0 = NULL;
	uint64_t guid0 = 0;
	boolean_t islog = B_FALSE;

	path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	mutex_enter(&ztest_vdev_lock);

	/*
	* Device removal is in progress, fault injection must be disabled
	* until it completes and the pool is scrubbed. The fault injection
	* strategy for damaging blocks does not take in to account evacuated
	* blocks which may have already been damaged.
	*/
	if (ztest_device_removal_active) {
	mutex_exit(&ztest_vdev_lock);
	goto out;
	}

	maxfaults = MAXFAULTS(zs);
	leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children;
	mirror_save = zs->zs_mirrors;
	mutex_exit(&ztest_vdev_lock);

	ASSERT3U(leaves, >=, 1);

	/*
	* While ztest is running the number of leaves will not change. This
	* is critical for the fault injection logic as it determines where
	* errors can be safely injected such that they are always repairable.
	*
	* When restarting ztest a different number of leaves may be requested
	* which will shift the regions to be damaged. This is fine as long
	* as the pool has been scrubbed prior to using the new mapping.
	* Failure to do can result in non-repairable damage being injected.
	*/
	if (ztest_pool_scrubbed == B_FALSE)
	goto out;

	/*
	* Grab the name lock as reader. There are some operations
	* which don't like to have their vdevs changed while
	* they are in progress (i.e. spa_change_guid). Those
	* operations will have grabbed the name lock as writer.
	*/
	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	/*
	* We need SCL_STATE here because we're going to look at vd0->vdev_tsd.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	if (ztest_random(2) == 0) {
	/*
	* Inject errors on a normal data device or slog device.
	*/
	top = ztest_random_vdev_top(spa, B_TRUE);
	leaf = ztest_random(leaves) + zs->zs_splits;

	/*
	* Generate paths to the first leaf in this top-level vdev,
	* and to the random leaf we selected. We'll induce transient
	* write failures and random online/offline activity on leaf 0,
	* and we'll write random garbage to the randomly chosen leaf.
	*/
	(void) snprintf(path0, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + zs->zs_splits);
	(void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + leaf);

	vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0);
	if (vd0 != NULL && vd0->vdev_top->vdev_islog)
	islog = B_TRUE;

	/*
	* If the top-level vdev needs to be resilvered
	* then we only allow faults on the device that is
	* resilvering.
	*/
	if (vd0 != NULL && maxfaults != 1 &&
	(!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) \|\|
	vd0->vdev_resilver_txg != 0)) {
	/*
	* Make vd0 explicitly claim to be unreadable,
	* or unwritable, or reach behind its back
	* and close the underlying fd. We can do this if
	* maxfaults == 0 because we'll fail and reexecute,
	* and we can do it if maxfaults >= 2 because we'll
	* have enough redundancy. If maxfaults == 1, the
	* combination of this with injection of random data
	* corruption below exceeds the pool's fault tolerance.
	*/
	vdev_file_t *vf = vd0->vdev_tsd;

	zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d",
	(long long)vd0->vdev_id, (int)maxfaults);

	if (vf != NULL && ztest_random(3) == 0) {
	(void) close(vf->vf_file->f_fd);
	vf->vf_file->f_fd = -1;
	} else if (ztest_random(2) == 0) {
	vd0->vdev_cant_read = B_TRUE;
	} else {
	vd0->vdev_cant_write = B_TRUE;
	}
	guid0 = vd0->vdev_guid;
	}
	} else {
	/*
	* Inject errors on an l2cache device.
	*/
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	if (sav->sav_count == 0) {
	spa_config_exit(spa, SCL_STATE, FTAG);
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	goto out;
	}
	vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)];
	guid0 = vd0->vdev_guid;
	- (void) strcpy(path0, vd0->vdev_path);
	- (void) strcpy(pathrand, vd0->vdev_path);
	+ (void) strlcpy(path0, vd0->vdev_path, MAXPATHLEN);
	+ (void) strlcpy(pathrand, vd0->vdev_path, MAXPATHLEN);

	leaf = 0;
	leaves = 1;
	maxfaults = INT_MAX; /* no limit on cache devices */
	}

	spa_config_exit(spa, SCL_STATE, FTAG);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	/*
	* If we can tolerate two or more faults, or we're dealing
	* with a slog, randomly online/offline vd0.
	*/
	if ((maxfaults >= 2 \|\| islog) && guid0 != 0) {
	if (ztest_random(10) < 6) {
	int flags = (ztest_random(2) == 0 ?
	ZFS_OFFLINE_TEMPORARY : 0);

	/*
	* We have to grab the zs_name_lock as writer to
	* prevent a race between offlining a slog and
	* destroying a dataset. Offlining the slog will
	* grab a reference on the dataset which may cause
	* dsl_destroy_head() to fail with EBUSY thus
	* leaving the dataset in an inconsistent state.
	*/
	if (islog)
	(void) pthread_rwlock_wrlock(&ztest_name_lock);

	VERIFY3U(vdev_offline(spa, guid0, flags), !=, EBUSY);

	if (islog)
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	} else {
	/*
	* Ideally we would like to be able to randomly
	* call vdev_[on\|off]line without holding locks
	* to force unpredictable failures but the side
	* effects of vdev_[on\|off]line prevent us from
	* doing so. We grab the ztest_vdev_lock here to
	* prevent a race between injection testing and
	* aux_vdev removal.
	*/
	mutex_enter(&ztest_vdev_lock);
	(void) vdev_online(spa, guid0, 0, NULL);
	mutex_exit(&ztest_vdev_lock);
	}
	}

	if (maxfaults == 0)
	goto out;

	/*
	* We have at least single-fault tolerance, so inject data corruption.
	*/
	fd = open(pathrand, O_RDWR);

	if (fd == -1) /* we hit a gap in the device namespace */
	goto out;

	fsize = lseek(fd, 0, SEEK_END);

	while (--iters != 0) {
	/*
	* The offset must be chosen carefully to ensure that
	* we do not inject a given logical block with errors
	* on two different leaf devices, because ZFS can not
	* tolerate that (if maxfaults==1).
	*
	* To achieve this we divide each leaf device into
	* chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4).
	* Each chunk is further divided into error-injection
	* ranges (can accept errors) and clear ranges (we do
	* not inject errors in those). Each error-injection
	* range can accept errors only for a single leaf vdev.
	* Error-injection ranges are separated by clear ranges.
	*
	* For example, with 3 leaves, each chunk looks like:
	* 0 to 32M: injection range for leaf 0
	* 32M to 64M: clear range - no injection allowed
	* 64M to 96M: injection range for leaf 1
	* 96M to 128M: clear range - no injection allowed
	* 128M to 160M: injection range for leaf 2
	* 160M to 192M: clear range - no injection allowed
	*
	* Each clear range must be large enough such that a
	* single block cannot straddle it. This way a block
	* can't be a target in two different injection ranges
	* (on different leaf vdevs).
	*/
	offset = ztest_random(fsize / (leaves << bshift)) *
	(leaves << bshift) + (leaf << bshift) +
	(ztest_random(1ULL << (bshift - 1)) & -8ULL);

	/*
	* Only allow damage to the labels at one end of the vdev.
	*
	* If all labels are damaged, the device will be totally
	* inaccessible, which will result in loss of data,
	* because we also damage (parts of) the other side of
	* the mirror/raidz.
	*
	* Additionally, we will always have both an even and an
	* odd label, so that we can handle crashes in the
	* middle of vdev_config_sync().
	*/
	if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE)
	continue;

	/*
	* The two end labels are stored at the "end" of the disk, but
	* the end of the disk (vdev_psize) is aligned to
	* sizeof (vdev_label_t).
	*/
	uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t));
	if ((leaf & 1) == 1 &&
	offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE)
	continue;

	mutex_enter(&ztest_vdev_lock);
	if (mirror_save != zs->zs_mirrors) {
	mutex_exit(&ztest_vdev_lock);
	(void) close(fd);
	goto out;
	}

	if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad))
	fatal(B_TRUE,
	"can't inject bad word at 0x%"PRIx64" in %s",
	offset, pathrand);

	mutex_exit(&ztest_vdev_lock);

	if (ztest_opts.zo_verbose >= 7)
	(void) printf("injected bad word into %s,"
	" offset 0x%"PRIx64"\n", pathrand, offset);
	}

	(void) close(fd);
	out:
	umem_free(path0, MAXPATHLEN);
	umem_free(pathrand, MAXPATHLEN);
	}

	/*
	* By design ztest will never inject uncorrectable damage in to the pool.
	* Issue a scrub, wait for it to complete, and verify there is never any
	* persistent damage.
	*
	* Only after a full scrub has been completed is it safe to start injecting
	* data corruption. See the comment in zfs_fault_inject().
	*/
	static int
	ztest_scrub_impl(spa_t *spa)
	{
	int error = spa_scan(spa, POOL_SCAN_SCRUB);
	if (error)
	return (error);

	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
	txg_wait_synced(spa_get_dsl(spa), 0);

	if (spa_get_errlog_size(spa) > 0)
	return (ECKSUM);

	ztest_pool_scrubbed = B_TRUE;

	return (0);
	}

	/*
	* Scrub the pool.
	*/
	void
	ztest_scrub(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error;

	/*
	* Scrub in progress by device removal.
	*/
	if (ztest_device_removal_active)
	return;

	/*
	* Start a scrub, wait a moment, then force a restart.
	*/
	(void) spa_scan(spa, POOL_SCAN_SCRUB);
	(void) poll(NULL, 0, 100);

	error = ztest_scrub_impl(spa);
	if (error == EBUSY)
	error = 0;
	ASSERT0(error);
	}

	/*
	* Change the guid for the pool.
	*/
	void
	ztest_reguid(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	uint64_t orig, load;
	int error;

	if (ztest_opts.zo_mmp_test)
	return;

	orig = spa_guid(spa);
	load = spa_load_guid(spa);

	(void) pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_change_guid(spa);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	if (error != 0)
	return;

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Changed guid old %"PRIu64" -> %"PRIu64"\n",
	orig, spa_guid(spa));
	}

	VERIFY3U(orig, !=, spa_guid(spa));
	VERIFY3U(load, ==, spa_load_guid(spa));
	}

	void
	ztest_blake3(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	hrtime_t end = gethrtime() + NANOSEC;
	zio_cksum_salt_t salt;
	void *salt_ptr = &salt.zcs_bytes;
	struct abd abd_data, abd_meta;
	void buf, templ;
	int i, *ptr;
	uint32_t size;
	BLAKE3_CTX ctx;

	size = ztest_random_blocksize();
	buf = umem_alloc(size, UMEM_NOFAIL);
	abd_data = abd_alloc(size, B_FALSE);
	abd_meta = abd_alloc(size, B_TRUE);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);
	memset(salt_ptr, 'A', 32);

	abd_copy_from_buf_off(abd_data, buf, 0, size);
	abd_copy_from_buf_off(abd_meta, buf, 0, size);

	while (gethrtime() <= end) {
	int run_count = 100;
	zio_cksum_t zc_ref1, zc_ref2;
	zio_cksum_t zc_res1, zc_res2;

	void *ref1 = &zc_ref1;
	void *ref2 = &zc_ref2;
	void *res1 = &zc_res1;
	void *res2 = &zc_res2;

	/* BLAKE3_KEY_LEN = 32 */
	VERIFY0(blake3_impl_setname("generic"));
	templ = abd_checksum_blake3_tmpl_init(&salt);
	Blake3_InitKeyed(&ctx, salt_ptr);
	Blake3_Update(&ctx, buf, size);
	Blake3_Final(&ctx, ref1);
	zc_ref2 = zc_ref1;
	ZIO_CHECKSUM_BSWAP(&zc_ref2);
	abd_checksum_blake3_tmpl_free(templ);

	VERIFY0(blake3_impl_setname("cycle"));
	while (run_count-- > 0) {

	/* Test current implementation */
	Blake3_InitKeyed(&ctx, salt_ptr);
	Blake3_Update(&ctx, buf, size);
	Blake3_Final(&ctx, res1);
	zc_res2 = zc_res1;
	ZIO_CHECKSUM_BSWAP(&zc_res2);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	/* Test ABD - data */
	templ = abd_checksum_blake3_tmpl_init(&salt);
	abd_checksum_blake3_native(abd_data, size,
	templ, &zc_res1);
	abd_checksum_blake3_byteswap(abd_data, size,
	templ, &zc_res2);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	/* Test ABD - metadata */
	abd_checksum_blake3_native(abd_meta, size,
	templ, &zc_res1);
	abd_checksum_blake3_byteswap(abd_meta, size,
	templ, &zc_res2);
	abd_checksum_blake3_tmpl_free(templ);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	}
	}

	abd_free(abd_data);
	abd_free(abd_meta);
	umem_free(buf, size);
	}

	void
	ztest_fletcher(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	hrtime_t end = gethrtime() + NANOSEC;

	while (gethrtime() <= end) {
	int run_count = 100;
	void *buf;
	struct abd abd_data, abd_meta;
	uint32_t size;
	int *ptr;
	int i;
	zio_cksum_t zc_ref;
	zio_cksum_t zc_ref_byteswap;

	size = ztest_random_blocksize();

	buf = umem_alloc(size, UMEM_NOFAIL);
	abd_data = abd_alloc(size, B_FALSE);
	abd_meta = abd_alloc(size, B_TRUE);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);

	abd_copy_from_buf_off(abd_data, buf, 0, size);
	abd_copy_from_buf_off(abd_meta, buf, 0, size);

	VERIFY0(fletcher_4_impl_set("scalar"));
	fletcher_4_native(buf, size, NULL, &zc_ref);
	fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap);

	VERIFY0(fletcher_4_impl_set("cycle"));
	while (run_count-- > 0) {
	zio_cksum_t zc;
	zio_cksum_t zc_byteswap;

	fletcher_4_byteswap(buf, size, NULL, &zc_byteswap);
	fletcher_4_native(buf, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	/* Test ABD - data */
	abd_fletcher_4_byteswap(abd_data, size, NULL,
	&zc_byteswap);
	abd_fletcher_4_native(abd_data, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	/* Test ABD - metadata */
	abd_fletcher_4_byteswap(abd_meta, size, NULL,
	&zc_byteswap);
	abd_fletcher_4_native(abd_meta, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	}

	umem_free(buf, size);
	abd_free(abd_data);
	abd_free(abd_meta);
	}
	}

	void
	ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	void *buf;
	size_t size;
	int *ptr;
	int i;
	zio_cksum_t zc_ref;
	zio_cksum_t zc_ref_bswap;

	hrtime_t end = gethrtime() + NANOSEC;

	while (gethrtime() <= end) {
	int run_count = 100;

	size = ztest_random_blocksize();
	buf = umem_alloc(size, UMEM_NOFAIL);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);

	VERIFY0(fletcher_4_impl_set("scalar"));
	fletcher_4_native(buf, size, NULL, &zc_ref);
	fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap);

	VERIFY0(fletcher_4_impl_set("cycle"));

	while (run_count-- > 0) {
	zio_cksum_t zc;
	zio_cksum_t zc_bswap;
	size_t pos = 0;

	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);

	while (pos < size) {
	size_t inc = 64 * ztest_random(size / 67);
	/* sometimes add few bytes to test non-simd */
	if (ztest_random(100) < 10)
	inc += P2ALIGN(ztest_random(64),
	sizeof (uint32_t));

	if (inc > (size - pos))
	inc = size - pos;

	fletcher_4_incremental_native(buf + pos, inc,
	&zc);
	fletcher_4_incremental_byteswap(buf + pos, inc,
	&zc_bswap);

	pos += inc;
	}

	VERIFY3U(pos, ==, size);

	VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
	VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));

	/*
	* verify if incremental on the whole buffer is
	* equivalent to non-incremental version
	*/
	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);

	fletcher_4_incremental_native(buf, size, &zc);
	fletcher_4_incremental_byteswap(buf, size, &zc_bswap);

	VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
	VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));
	}

	umem_free(buf, size);
	}
	}

	static int
	ztest_set_global_vars(void)
	{
	for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
	char *kv = ztest_opts.zo_gvars[i];
	VERIFY3U(strlen(kv), <=, ZO_GVARS_MAX_ARGLEN);
	VERIFY3U(strlen(kv), >, 0);
	int err = set_global_var(kv);
	if (ztest_opts.zo_verbose > 0) {
	(void) printf("setting global var %s ... %s\n", kv,
	err ? "failed" : "ok");
	}
	if (err != 0) {
	(void) fprintf(stderr,
	"failed to set global var '%s'\n", kv);
	return (err);
	}
	}
	return (0);
	}

	static char **
	ztest_global_vars_to_zdb_args(void)
	{
	char *args = calloc(2ztest_opts.zo_gvars_count + 1, sizeof (char *));
	char **cur = args;
	+ if (args == NULL)
	+ return (NULL);
	for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
	cur++ = (char )"-o";
	*cur++ = ztest_opts.zo_gvars[i];
	}
	ASSERT3P(cur, ==, &args[2*ztest_opts.zo_gvars_count]);
	*cur = NULL;
	return (args);
	}

	/* The end of strings is indicated by a NULL element */
	static char *
	join_strings(char *strings, const char sep)
	{
	size_t totallen = 0;
	for (char *sp = strings; sp != NULL; sp++) {
	totallen += strlen(*sp);
	totallen += strlen(sep);
	}
	if (totallen > 0) {
	ASSERT(totallen >= strlen(sep));
	totallen -= strlen(sep);
	}

	size_t buflen = totallen + 1;
	char o = malloc(buflen); / trailing 0 byte */
	o[0] = '\0';
	for (char *sp = strings; sp != NULL; sp++) {
	size_t would;
	would = strlcat(o, *sp, buflen);
	VERIFY3U(would, <, buflen);
	if (*(sp+1) == NULL) {
	break;
	}
	would = strlcat(o, sep, buflen);
	VERIFY3U(would, <, buflen);
	}
	ASSERT3S(strlen(o), ==, totallen);
	return (o);
	}

	static int
	ztest_check_path(char *path)
	{
	struct stat s;
	/* return true on success */
	return (!stat(path, &s));
	}

	static void
	ztest_get_zdb_bin(char *bin, int len)
	{
	char *zdb_path;
	/*
	* Try to use $ZDB and in-tree zdb path. If not successful, just
	* let popen to search through PATH.
	*/
	if ((zdb_path = getenv("ZDB"))) {
	strlcpy(bin, zdb_path, len); /* In env */
	if (!ztest_check_path(bin)) {
	ztest_dump_core = 0;
	fatal(B_TRUE, "invalid ZDB '%s'", bin);
	}
	return;
	}

	VERIFY3P(realpath(getexecname(), bin), !=, NULL);
	if (strstr(bin, ".libs/ztest")) {
	strstr(bin, ".libs/ztest")[0] = '\0'; /* In-tree */
	strcat(bin, "zdb");
	if (ztest_check_path(bin))
	return;
	}
	strcpy(bin, "zdb");
	}

	static vdev_t *
	ztest_random_concrete_vdev_leaf(vdev_t *vd)
	{
	if (vd == NULL)
	return (NULL);

	if (vd->vdev_children == 0)
	return (vd);

	vdev_t *eligible[vd->vdev_children];
	int eligible_idx = 0, i;
	for (i = 0; i < vd->vdev_children; i++) {
	vdev_t *cvd = vd->vdev_child[i];
	if (cvd->vdev_top->vdev_removing)
	continue;
	if (cvd->vdev_children > 0 \|\|
	(vdev_is_concrete(cvd) && !cvd->vdev_detached)) {
	eligible[eligible_idx++] = cvd;
	}
	}
	VERIFY3S(eligible_idx, >, 0);

	uint64_t child_no = ztest_random(eligible_idx);
	return (ztest_random_concrete_vdev_leaf(eligible[child_no]));
	}

	void
	ztest_initialize(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error = 0;

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* Random leaf vdev */
	vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
	if (rand_vd == NULL) {
	spa_config_exit(spa, SCL_VDEV, FTAG);
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The random vdev we've selected may change as soon as we
	* drop the spa_config_lock. We create local copies of things
	* we're interested in.
	*/
	uint64_t guid = rand_vd->vdev_guid;
	char *path = strdup(rand_vd->vdev_path);
	boolean_t active = rand_vd->vdev_initialize_thread != NULL;

	zfs_dbgmsg("vd %px, guid %llu", rand_vd, (u_longlong_t)guid);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS);

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *vdev_errlist = fnvlist_alloc();
	fnvlist_add_uint64(vdev_guids, path, guid);
	error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist);
	fnvlist_free(vdev_guids);
	fnvlist_free(vdev_errlist);

	switch (cmd) {
	case POOL_INITIALIZE_CANCEL:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Cancel initialize %s", path);
	if (!active)
	(void) printf(" failed (no initialize active)");
	(void) printf("\n");
	}
	break;
	case POOL_INITIALIZE_START:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Start initialize %s", path);
	if (active && error == 0)
	(void) printf(" failed (already active)");
	else if (error != 0)
	(void) printf(" failed (error %d)", error);
	(void) printf("\n");
	}
	break;
	case POOL_INITIALIZE_SUSPEND:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Suspend initialize %s", path);
	if (!active)
	(void) printf(" failed (no initialize active)");
	(void) printf("\n");
	}
	break;
	}
	free(path);
	mutex_exit(&ztest_vdev_lock);
	}

	void
	ztest_trim(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error = 0;

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* Random leaf vdev */
	vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
	if (rand_vd == NULL) {
	spa_config_exit(spa, SCL_VDEV, FTAG);
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The random vdev we've selected may change as soon as we
	* drop the spa_config_lock. We create local copies of things
	* we're interested in.
	*/
	uint64_t guid = rand_vd->vdev_guid;
	char *path = strdup(rand_vd->vdev_path);
	boolean_t active = rand_vd->vdev_trim_thread != NULL;

	zfs_dbgmsg("vd %p, guid %llu", rand_vd, (u_longlong_t)guid);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	uint64_t cmd = ztest_random(POOL_TRIM_FUNCS);
	uint64_t rate = 1 << ztest_random(30);
	boolean_t partial = (ztest_random(5) > 0);
	boolean_t secure = (ztest_random(5) > 0);

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *vdev_errlist = fnvlist_alloc();
	fnvlist_add_uint64(vdev_guids, path, guid);
	error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial,
	secure, vdev_errlist);
	fnvlist_free(vdev_guids);
	fnvlist_free(vdev_errlist);

	switch (cmd) {
	case POOL_TRIM_CANCEL:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Cancel TRIM %s", path);
	if (!active)
	(void) printf(" failed (no TRIM active)");
	(void) printf("\n");
	}
	break;
	case POOL_TRIM_START:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Start TRIM %s", path);
	if (active && error == 0)
	(void) printf(" failed (already active)");
	else if (error != 0)
	(void) printf(" failed (error %d)", error);
	(void) printf("\n");
	}
	break;
	case POOL_TRIM_SUSPEND:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Suspend TRIM %s", path);
	if (!active)
	(void) printf(" failed (no TRIM active)");
	(void) printf("\n");
	}
	break;
	}
	free(path);
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify pool integrity by running zdb.
	*/
	static void
	ztest_run_zdb(const char *pool)
	{
	int status;
	char *bin;
	char *zdb;
	char *zbuf;
	const int len = MAXPATHLEN + MAXNAMELEN + 20;
	FILE *fp;

	bin = umem_alloc(len, UMEM_NOFAIL);
	zdb = umem_alloc(len, UMEM_NOFAIL);
	zbuf = umem_alloc(1024, UMEM_NOFAIL);

	ztest_get_zdb_bin(bin, len);

	char **set_gvars_args = ztest_global_vars_to_zdb_args();
	+ if (set_gvars_args == NULL) {
	+ fatal(B_FALSE, "Failed to allocate memory in "
	+ "ztest_global_vars_to_zdb_args(). Cannot run zdb.\n");
	+ }
	char *set_gvars_args_joined = join_strings(set_gvars_args, " ");
	free(set_gvars_args);

	size_t would = snprintf(zdb, len,
	"%s -bcc%s%s -G -d -Y -e -y %s -p %s %s",
	bin,
	ztest_opts.zo_verbose >= 3 ? "s" : "",
	ztest_opts.zo_verbose >= 4 ? "v" : "",
	set_gvars_args_joined,
	ztest_opts.zo_dir,
	pool);
	ASSERT3U(would, <, len);

	free(set_gvars_args_joined);

	if (ztest_opts.zo_verbose >= 5)
	(void) printf("Executing %s\n", zdb);

	fp = popen(zdb, "r");

	while (fgets(zbuf, 1024, fp) != NULL)
	if (ztest_opts.zo_verbose >= 3)
	(void) printf("%s", zbuf);

	status = pclose(fp);

	if (status == 0)
	goto out;

	ztest_dump_core = 0;
	if (WIFEXITED(status))
	fatal(B_FALSE, "'%s' exit code %d", zdb, WEXITSTATUS(status));
	else
	fatal(B_FALSE, "'%s' died with signal %d",
	zdb, WTERMSIG(status));
	out:
	umem_free(bin, len);
	umem_free(zdb, len);
	umem_free(zbuf, 1024);
	}

	static void
	ztest_walk_pool_directory(const char *header)
	{
	spa_t *spa = NULL;

	if (ztest_opts.zo_verbose >= 6)
	(void) puts(header);

	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	if (ztest_opts.zo_verbose >= 6)
	(void) printf("\t%s\n", spa_name(spa));
	mutex_exit(&spa_namespace_lock);
	}

	static void
	ztest_spa_import_export(char oldname, char newname)
	{
	nvlist_t config, newconfig;
	uint64_t pool_guid;
	spa_t *spa;
	int error;

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("import/export: old = %s, new = %s\n",
	oldname, newname);
	}

	/*
	* Clean up from previous runs.
	*/
	(void) spa_destroy(newname);

	/*
	* Get the pool's configuration and guid.
	*/
	VERIFY0(spa_open(oldname, &spa, FTAG));

	/*
	* Kick off a scrub to tickle scrub/export races.
	*/
	if (ztest_random(2) == 0)
	(void) spa_scan(spa, POOL_SCAN_SCRUB);

	pool_guid = spa_guid(spa);
	spa_close(spa, FTAG);

	ztest_walk_pool_directory("pools before export");

	/*
	* Export it.
	*/
	VERIFY0(spa_export(oldname, &config, B_FALSE, B_FALSE));

	ztest_walk_pool_directory("pools after export");

	/*
	* Try to import it.
	*/
	newconfig = spa_tryimport(config);
	ASSERT3P(newconfig, !=, NULL);
	fnvlist_free(newconfig);

	/*
	* Import it under the new name.
	*/
	error = spa_import(newname, config, NULL, 0);
	if (error != 0) {
	dump_nvlist(config, 0);
	fatal(B_FALSE, "couldn't import pool %s as %s: error %u",
	oldname, newname, error);
	}

	ztest_walk_pool_directory("pools after import");

	/*
	* Try to import it again -- should fail with EEXIST.
	*/
	VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0));

	/*
	* Try to import it under a different name -- should fail with EEXIST.
	*/
	VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0));

	/*
	* Verify that the pool is no longer visible under the old name.
	*/
	VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));

	/*
	* Verify that we can open and close the pool using the new name.
	*/
	VERIFY0(spa_open(newname, &spa, FTAG));
	ASSERT3U(pool_guid, ==, spa_guid(spa));
	spa_close(spa, FTAG);

	fnvlist_free(config);
	}

	static void
	ztest_resume(spa_t *spa)
	{
	if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6)
	(void) printf("resuming from suspended state\n");
	spa_vdev_state_enter(spa, SCL_NONE);
	vdev_clear(spa, NULL);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	(void) zio_resume(spa);
	}

	static __attribute__((noreturn)) void
	ztest_resume_thread(void *arg)
	{
	spa_t *spa = arg;

	while (!ztest_exiting) {
	if (spa_suspended(spa))
	ztest_resume(spa);
	(void) poll(NULL, 0, 100);

	/*
	* Periodically change the zfs_compressed_arc_enabled setting.
	*/
	if (ztest_random(10) == 0)
	zfs_compressed_arc_enabled = ztest_random(2);

	/*
	* Periodically change the zfs_abd_scatter_enabled setting.
	*/
	if (ztest_random(10) == 0)
	zfs_abd_scatter_enabled = ztest_random(2);
	}

	thread_exit();
	}

	static __attribute__((noreturn)) void
	ztest_deadman_thread(void *arg)
	{
	ztest_shared_t *zs = arg;
	spa_t *spa = ztest_spa;
	hrtime_t delay, overdue, last_run = gethrtime();

	delay = (zs->zs_thread_stop - zs->zs_thread_start) +
	MSEC2NSEC(zfs_deadman_synctime_ms);

	while (!ztest_exiting) {
	/*
	* Wait for the delay timer while checking occasionally
	* if we should stop.
	*/
	if (gethrtime() < last_run + delay) {
	(void) poll(NULL, 0, 1000);
	continue;
	}

	/*
	* If the pool is suspended then fail immediately. Otherwise,
	* check to see if the pool is making any progress. If
	* vdev_deadman() discovers that there hasn't been any recent
	* I/Os then it will end up aborting the tests.
	*/
	if (spa_suspended(spa) \|\| spa->spa_root_vdev == NULL) {
	fatal(B_FALSE,
	"aborting test after %lu seconds because "
	"pool has transitioned to a suspended state.",
	zfs_deadman_synctime_ms / 1000);
	}
	vdev_deadman(spa->spa_root_vdev, FTAG);

	/*
	* If the process doesn't complete within a grace period of
	* zfs_deadman_synctime_ms over the expected finish time,
	* then it may be hung and is terminated.
	*/
	overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms);
	if (gethrtime() > overdue) {
	fatal(B_FALSE,
	"aborting test after %llu seconds because "
	"the process is overdue for termination.",
	(gethrtime() - zs->zs_proc_start) / NANOSEC);
	}

	(void) printf("ztest has been running for %lld seconds\n",
	(gethrtime() - zs->zs_proc_start) / NANOSEC);

	last_run = gethrtime();
	delay = MSEC2NSEC(zfs_deadman_checktime_ms);
	}

	thread_exit();
	}

	static void
	ztest_execute(int test, ztest_info_t *zi, uint64_t id)
	{
	ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets];
	ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test);
	hrtime_t functime = gethrtime();
	int i;

	for (i = 0; i < zi->zi_iters; i++)
	zi->zi_func(zd, id);

	functime = gethrtime() - functime;

	atomic_add_64(&zc->zc_count, 1);
	atomic_add_64(&zc->zc_time, functime);

	if (ztest_opts.zo_verbose >= 4)
	(void) printf("%6.2f sec in %s\n",
	(double)functime / NANOSEC, zi->zi_funcname);
	}

	static __attribute__((noreturn)) void
	ztest_thread(void *arg)
	{
	int rand;
	uint64_t id = (uintptr_t)arg;
	ztest_shared_t *zs = ztest_shared;
	uint64_t call_next;
	hrtime_t now;
	ztest_info_t *zi;
	ztest_shared_callstate_t *zc;

	while ((now = gethrtime()) < zs->zs_thread_stop) {
	/*
	* See if it's time to force a crash.
	*/
	if (now > zs->zs_thread_kill)
	ztest_kill(zs);

	/*
	* If we're getting ENOSPC with some regularity, stop.
	*/
	if (zs->zs_enospc_count > 10)
	break;

	/*
	* Pick a random function to execute.
	*/
	rand = ztest_random(ZTEST_FUNCS);
	zi = &ztest_info[rand];
	zc = ZTEST_GET_SHARED_CALLSTATE(rand);
	call_next = zc->zc_next;

	if (now >= call_next &&
	atomic_cas_64(&zc->zc_next, call_next, call_next +
	ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) {
	ztest_execute(rand, zi, id);
	}
	}

	thread_exit();
	}

	static void
	ztest_dataset_name(char dsname, const char pool, int d)
	{
	(void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d);
	}

	static void
	ztest_dataset_destroy(int d)
	{
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int t;

	ztest_dataset_name(name, ztest_opts.zo_pool, d);

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("Destroying %s to free up space\n", name);

	/*
	* Cleanup any non-standard clones and snapshots. In general,
	* ztest thread t operates on dataset (t % zopt_datasets),
	* so there may be more than one thing to clean up.
	*/
	for (t = d; t < ztest_opts.zo_threads;
	t += ztest_opts.zo_datasets)
	ztest_dsl_dataset_cleanup(name, t);

	(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
	DS_FIND_SNAPSHOTS \| DS_FIND_CHILDREN);
	}

	static void
	ztest_dataset_dirobj_verify(ztest_ds_t *zd)
	{
	uint64_t usedobjs, dirobjs, scratch;

	/*
	* ZTEST_DIROBJ is the object directory for the entire dataset.
	* Therefore, the number of objects in use should equal the
	* number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself.
	* If not, we have an object leak.
	*
	* Note that we can only check this in ztest_dataset_open(),
	* when the open-context and syncing-context values agree.
	* That's because zap_count() returns the open-context value,
	* while dmu_objset_space() returns the rootbp fill count.
	*/
	VERIFY0(zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs));
	dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch);
	ASSERT3U(dirobjs + 1, ==, usedobjs);
	}

	static int
	ztest_dataset_open(int d)
	{
	ztest_ds_t *zd = &ztest_ds[d];
	uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq;
	objset_t *os;
	zilog_t *zilog;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	ztest_dataset_name(name, ztest_opts.zo_pool, d);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	error = ztest_dataset_create(name);
	if (error == ENOSPC) {
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT(error == 0 \|\| error == EEXIST);

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
	B_TRUE, zd, &os));
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os);

	zilog = zd->zd_zilog;

	if (zilog->zl_header->zh_claim_lr_seq != 0 &&
	zilog->zl_header->zh_claim_lr_seq < committed_seq)
	fatal(B_FALSE, "missing log records: "
	"claimed %"PRIu64" < committed %"PRIu64"",
	zilog->zl_header->zh_claim_lr_seq, committed_seq);

	ztest_dataset_dirobj_verify(zd);

	zil_replay(os, zd, ztest_replay_vector);

	ztest_dataset_dirobj_verify(zd);

	if (ztest_opts.zo_verbose >= 6)
	(void) printf("%s replay %"PRIu64" blocks, "
	"%"PRIu64" records, seq %"PRIu64"\n",
	zd->zd_name,
	zilog->zl_parse_blk_count,
	zilog->zl_parse_lr_count,
	zilog->zl_replaying_seq);

	zilog = zil_open(os, ztest_get_data, NULL);

	if (zilog->zl_replaying_seq != 0 &&
	zilog->zl_replaying_seq < committed_seq)
	fatal(B_FALSE, "missing log records: "
	"replayed %"PRIu64" < committed %"PRIu64"",
	zilog->zl_replaying_seq, committed_seq);

	return (0);
	}

	static void
	ztest_dataset_close(int d)
	{
	ztest_ds_t *zd = &ztest_ds[d];

	zil_close(zd->zd_zilog);
	dmu_objset_disown(zd->zd_os, B_TRUE, zd);

	ztest_zd_fini(zd);
	}

	static int
	ztest_replay_zil_cb(const char name, void arg)
	{
	(void) arg;
	objset_t *os;
	ztest_ds_t *zdtmp;

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE,
	B_TRUE, FTAG, &os));

	zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);

	ztest_zd_init(zdtmp, NULL, os);
	zil_replay(os, zdtmp, ztest_replay_vector);
	ztest_zd_fini(zdtmp);

	if (dmu_objset_zil(os)->zl_parse_lr_count != 0 &&
	ztest_opts.zo_verbose >= 6) {
	zilog_t *zilog = dmu_objset_zil(os);

	(void) printf("%s replay %"PRIu64" blocks, "
	"%"PRIu64" records, seq %"PRIu64"\n",
	name,
	zilog->zl_parse_blk_count,
	zilog->zl_parse_lr_count,
	zilog->zl_replaying_seq);
	}

	umem_free(zdtmp, sizeof (ztest_ds_t));

	dmu_objset_disown(os, B_TRUE, FTAG);
	return (0);
	}

	static void
	ztest_freeze(void)
	{
	ztest_ds_t *zd = &ztest_ds[0];
	spa_t *spa;
	int numloops = 0;

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("testing spa_freeze()...\n");

	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	VERIFY0(ztest_dataset_open(0));
	ztest_spa = spa;

	/*
	* Force the first log block to be transactionally allocated.
	* We have to do this before we freeze the pool -- otherwise
	* the log chain won't be anchored.
	*/
	while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) {
	ztest_dmu_object_alloc_free(zd, 0);
	zil_commit(zd->zd_zilog, 0);
	}

	txg_wait_synced(spa_get_dsl(spa), 0);

	/*
	* Freeze the pool. This stops spa_sync() from doing anything,
	* so that the only way to record changes from now on is the ZIL.
	*/
	spa_freeze(spa);

	/*
	* Because it is hard to predict how much space a write will actually
	* require beforehand, we leave ourselves some fudge space to write over
	* capacity.
	*/
	uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2;

	/*
	* Run tests that generate log records but don't alter the pool config
	* or depend on DSL sync tasks (snapshots, objset create/destroy, etc).
	* We do a txg_wait_synced() after each iteration to force the txg
	* to increase well beyond the last synced value in the uberblock.
	* The ZIL should be OK with that.
	*
	* Run a random number of times less than zo_maxloops and ensure we do
	* not run out of space on the pool.
	*/
	while (ztest_random(10) != 0 &&
	numloops++ < ztest_opts.zo_maxloops &&
	metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) {
	ztest_od_t od;
	ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);
	VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE));
	ztest_io(zd, od.od_object,
	ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	/*
	* Commit all of the changes we just generated.
	*/
	zil_commit(zd->zd_zilog, 0);
	txg_wait_synced(spa_get_dsl(spa), 0);

	/*
	* Close our dataset and close the pool.
	*/
	ztest_dataset_close(0);
	spa_close(spa, FTAG);
	kernel_fini();

	/*
	* Open and close the pool and dataset to induce log replay.
	*/
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	ASSERT3U(spa_freeze_txg(spa), ==, UINT64_MAX);
	VERIFY0(ztest_dataset_open(0));
	ztest_spa = spa;
	txg_wait_synced(spa_get_dsl(spa), 0);
	ztest_dataset_close(0);
	ztest_reguid(NULL, 0);

	spa_close(spa, FTAG);
	kernel_fini();
	}

	static void
	ztest_import_impl(void)
	{
	importargs_t args = { 0 };
	nvlist_t *cfg = NULL;
	int nsearch = 1;
	char *searchdirs[nsearch];
	int flags = ZFS_IMPORT_MISSING_LOG;

	searchdirs[0] = ztest_opts.zo_dir;
	args.paths = nsearch;
	args.path = searchdirs;
	args.can_be_active = B_FALSE;

	VERIFY0(zpool_find_config(NULL, ztest_opts.zo_pool, &cfg, &args,
	&libzpool_config_ops));
	VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags));
	fnvlist_free(cfg);
	}

	/*
	* Import a storage pool with the given name.
	*/
	static void
	ztest_import(ztest_shared_t *zs)
	{
	spa_t *spa;

	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);

	ztest_import_impl();

	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	spa_close(spa, FTAG);

	kernel_fini();

	if (!ztest_opts.zo_mmp_test) {
	ztest_run_zdb(ztest_opts.zo_pool);
	ztest_freeze();
	ztest_run_zdb(ztest_opts.zo_pool);
	}

	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	/*
	* Kick off threads to run tests on all datasets in parallel.
	*/
	static void
	ztest_run(ztest_shared_t *zs)
	{
	spa_t *spa;
	objset_t *os;
	kthread_t resume_thread, deadman_thread;
	kthread_t **run_threads;
	uint64_t object;
	int error;
	int t, d;

	ztest_exiting = B_FALSE;

	/*
	* Initialize parent/child shared state.
	*/
	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	zs->zs_thread_start = gethrtime();
	zs->zs_thread_stop =
	zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC;
	zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop);
	zs->zs_thread_kill = zs->zs_thread_stop;
	if (ztest_random(100) < ztest_opts.zo_killrate) {
	zs->zs_thread_kill -=
	ztest_random(ztest_opts.zo_passtime * NANOSEC);
	}

	mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL);

	list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t),
	offsetof(ztest_cb_data_t, zcd_node));

	/*
	* Open our pool. It may need to be imported first depending on
	* what tests were running when the previous pass was terminated.
	*/
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	error = spa_open(ztest_opts.zo_pool, &spa, FTAG);
	if (error) {
	VERIFY3S(error, ==, ENOENT);
	ztest_import_impl();
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	}

	metaslab_preload_limit = ztest_random(20) + 1;
	ztest_spa = spa;

	VERIFY0(vdev_raidz_impl_set("cycle"));

	dmu_objset_stats_t dds;
	VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool,
	DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os));
	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
	dmu_objset_fast_stat(os, &dds);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	zs->zs_guid = dds.dds_guid;
	dmu_objset_disown(os, B_TRUE, FTAG);

	/*
	* Create a thread to periodically resume suspended I/O.
	*/
	resume_thread = thread_create(NULL, 0, ztest_resume_thread,
	spa, 0, NULL, TS_RUN \| TS_JOINABLE, defclsyspri);

	/*
	* Create a deadman thread and set to panic if we hang.
	*/
	deadman_thread = thread_create(NULL, 0, ztest_deadman_thread,
	zs, 0, NULL, TS_RUN \| TS_JOINABLE, defclsyspri);

	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;

	/*
	* Verify that we can safely inquire about any object,
	* whether it's allocated or not. To make it interesting,
	* we probe a 5-wide window around each power of two.
	* This hits all edge cases, including zero and the max.
	*/
	for (t = 0; t < 64; t++) {
	for (d = -5; d <= 5; d++) {
	error = dmu_object_info(spa->spa_meta_objset,
	(1ULL << t) + d, NULL);
	ASSERT(error == 0 \|\| error == ENOENT \|\|
	error == EINVAL);
	}
	}

	/*
	* If we got any ENOSPC errors on the previous run, destroy something.
	*/
	if (zs->zs_enospc_count != 0) {
	int d = ztest_random(ztest_opts.zo_datasets);
	ztest_dataset_destroy(d);
	}
	zs->zs_enospc_count = 0;

	/*
	* If we were in the middle of ztest_device_removal() and were killed
	* we need to ensure the removal and scrub complete before running
	* any tests that check ztest_device_removal_active. The removal will
	* be restarted automatically when the spa is opened, but we need to
	* initiate the scrub manually if it is not already in progress. Note
	* that we always run the scrub whenever an indirect vdev exists
	* because we have no way of knowing for sure if ztest_device_removal()
	* fully completed its scrub before the pool was reimported.
	*/
	if (spa->spa_removing_phys.sr_state == DSS_SCANNING \|\|
	spa->spa_removing_phys.sr_prev_indirect_vdev != -1) {
	while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
	txg_wait_synced(spa_get_dsl(spa), 0);

	error = ztest_scrub_impl(spa);
	if (error == EBUSY)
	error = 0;
	ASSERT0(error);
	}

	run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *),
	UMEM_NOFAIL);

	if (ztest_opts.zo_verbose >= 4)
	(void) printf("starting main threads...\n");

	/*
	* Replay all logs of all datasets in the pool. This is primarily for
	* temporary datasets which wouldn't otherwise get replayed, which
	* can trigger failures when attempting to offline a SLOG in
	* ztest_fault_inject().
	*/
	(void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb,
	NULL, DS_FIND_CHILDREN);

	/*
	* Kick off all the tests that run in parallel.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++) {
	if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) {
	umem_free(run_threads, ztest_opts.zo_threads *
	sizeof (kthread_t *));
	return;
	}

	run_threads[t] = thread_create(NULL, 0, ztest_thread,
	(void *)(uintptr_t)t, 0, NULL, TS_RUN \| TS_JOINABLE,
	defclsyspri);
	}

	/*
	* Wait for all of the tests to complete.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++)
	VERIFY0(thread_join(run_threads[t]));

	/*
	* Close all datasets. This must be done after all the threads
	* are joined so we can be sure none of the datasets are in-use
	* by any of the threads.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++) {
	if (t < ztest_opts.zo_datasets)
	ztest_dataset_close(t);
	}

	txg_wait_synced(spa_get_dsl(spa), 0);

	zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	zs->zs_space = metaslab_class_get_space(spa_normal_class(spa));

	umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *));

	/* Kill the resume and deadman threads */
	ztest_exiting = B_TRUE;
	VERIFY0(thread_join(resume_thread));
	VERIFY0(thread_join(deadman_thread));
	ztest_resume(spa);

	/*
	* Right before closing the pool, kick off a bunch of async I/O;
	* spa_close() should wait for it to complete.
	*/
	for (object = 1; object < 50; object++) {
	dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20,
	ZIO_PRIORITY_SYNC_READ);
	}

	/* Verify that at least one commit cb was called in a timely fashion */
	if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG)
	VERIFY0(zc_min_txg_delay);

	spa_close(spa, FTAG);

	/*
	* Verify that we can loop over all pools.
	*/
	mutex_enter(&spa_namespace_lock);
	for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa))
	if (ztest_opts.zo_verbose > 3)
	(void) printf("spa_next: found %s\n", spa_name(spa));
	mutex_exit(&spa_namespace_lock);

	/*
	* Verify that we can export the pool and reimport it under a
	* different name.
	*/
	if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	(void) snprintf(name, sizeof (name), "%s_import",
	ztest_opts.zo_pool);
	ztest_spa_import_export(ztest_opts.zo_pool, name);
	ztest_spa_import_export(name, ztest_opts.zo_pool);
	}

	kernel_fini();

	list_destroy(&zcl.zcl_callbacks);
	mutex_destroy(&zcl.zcl_callbacks_lock);
	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	static void
	print_time(hrtime_t t, char *timebuf)
	{
	hrtime_t s = t / NANOSEC;
	hrtime_t m = s / 60;
	hrtime_t h = m / 60;
	hrtime_t d = h / 24;

	s -= m * 60;
	m -= h * 60;
	h -= d * 24;

	timebuf[0] = '\0';

	if (d)
	(void) sprintf(timebuf,
	"%llud%02lluh%02llum%02llus", d, h, m, s);
	else if (h)
	(void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s);
	else if (m)
	(void) sprintf(timebuf, "%llum%02llus", m, s);
	else
	(void) sprintf(timebuf, "%llus", s);
	}

	static nvlist_t *
	make_random_props(void)
	{
	nvlist_t *props;

	props = fnvlist_alloc();

	if (ztest_random(2) == 0)
	return (props);

	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1);

	return (props);
	}

	/*
	* Create a storage pool with the given name and initial vdev size.
	* Then test spa_freeze() functionality.
	*/
	static void
	ztest_init(ztest_shared_t *zs)
	{
	spa_t *spa;
	nvlist_t nvroot, props;
	int i;

	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);

	/*
	* Create the storage pool.
	*/
	(void) spa_destroy(ztest_opts.zo_pool);
	ztest_shared->zs_vdev_next_leaf = 0;
	zs->zs_splits = 0;
	zs->zs_mirrors = ztest_opts.zo_mirrors;
	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
	NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);
	props = make_random_props();

	/*
	* We don't expect the pool to suspend unless maxfaults == 0,
	* in which case ztest_fault_inject() temporarily takes away
	* the only valid replica.
	*/
	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE),
	MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT);

	for (i = 0; i < SPA_FEATURES; i++) {
	char *buf;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	/*
	* 75% chance of using the log space map feature. We want ztest
	* to exercise both the code paths that use the log space map
	* feature and the ones that don't.
	*/
	if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0)
	continue;

	VERIFY3S(-1, !=, asprintf(&buf, "feature@%s",
	spa_feature_table[i].fi_uname));
	fnvlist_add_uint64(props, buf, 0);
	free(buf);
	}

	VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL));
	fnvlist_free(nvroot);
	fnvlist_free(props);

	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	spa_close(spa, FTAG);

	kernel_fini();

	if (!ztest_opts.zo_mmp_test) {
	ztest_run_zdb(ztest_opts.zo_pool);
	ztest_freeze();
	ztest_run_zdb(ztest_opts.zo_pool);
	}

	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	static void
	setup_data_fd(void)
	{
	static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX";

	ztest_fd_data = mkstemp(ztest_name_data);
	ASSERT3S(ztest_fd_data, >=, 0);
	(void) unlink(ztest_name_data);
	}

	static int
	shared_data_size(ztest_shared_hdr_t *hdr)
	{
	int size;

	size = hdr->zh_hdr_size;
	size += hdr->zh_opts_size;
	size += hdr->zh_size;
	size += hdr->zh_stats_size * hdr->zh_stats_count;
	size += hdr->zh_ds_size * hdr->zh_ds_count;

	return (size);
	}

	static void
	setup_hdr(void)
	{
	int size;
	ztest_shared_hdr_t *hdr;

	hdr = (void )mmap(0, P2ROUNDUP(sizeof (hdr), getpagesize()),
	PROT_READ \| PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);

	VERIFY0(ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t)));

	hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t);
	hdr->zh_opts_size = sizeof (ztest_shared_opts_t);
	hdr->zh_size = sizeof (ztest_shared_t);
	hdr->zh_stats_size = sizeof (ztest_shared_callstate_t);
	hdr->zh_stats_count = ZTEST_FUNCS;
	hdr->zh_ds_size = sizeof (ztest_shared_ds_t);
	hdr->zh_ds_count = ztest_opts.zo_datasets;

	size = shared_data_size(hdr);
	VERIFY0(ftruncate(ztest_fd_data, size));

	(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
	}

	static void
	setup_data(void)
	{
	int size, offset;
	ztest_shared_hdr_t *hdr;
	uint8_t *buf;

	hdr = (void )mmap(0, P2ROUNDUP(sizeof (hdr), getpagesize()),
	PROT_READ, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);

	size = shared_data_size(hdr);

	(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
	hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()),
	PROT_READ \| PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);
	buf = (uint8_t *)hdr;

	offset = hdr->zh_hdr_size;
	ztest_shared_opts = (void *)&buf[offset];
	offset += hdr->zh_opts_size;
	ztest_shared = (void *)&buf[offset];
	offset += hdr->zh_size;
	ztest_shared_callstate = (void *)&buf[offset];
	offset += hdr->zh_stats_size * hdr->zh_stats_count;
	ztest_shared_ds = (void *)&buf[offset];
	}

	static boolean_t
	exec_child(char cmd, char libpath, boolean_t ignorekill, int *statusp)
	{
	pid_t pid;
	int status;
	char *cmdbuf = NULL;

	pid = fork();

	if (cmd == NULL) {
	cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	(void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN);
	cmd = cmdbuf;
	}

	if (pid == -1)
	fatal(B_TRUE, "fork failed");

	if (pid == 0) { /* child */
	char fd_data_str[12];

	VERIFY3S(11, >=,
	snprintf(fd_data_str, 12, "%d", ztest_fd_data));
	VERIFY0(setenv("ZTEST_FD_DATA", fd_data_str, 1));

	if (libpath != NULL) {
	const char *curlp = getenv("LD_LIBRARY_PATH");
	if (curlp == NULL)
	VERIFY0(setenv("LD_LIBRARY_PATH", libpath, 1));
	else {
	char *newlp = NULL;
	VERIFY3S(-1, !=,
	asprintf(&newlp, "%s:%s", libpath, curlp));
	VERIFY0(setenv("LD_LIBRARY_PATH", newlp, 1));
	free(newlp);
	}
	}
	(void) execl(cmd, cmd, (char *)NULL);
	ztest_dump_core = B_FALSE;
	fatal(B_TRUE, "exec failed: %s", cmd);
	}

	if (cmdbuf != NULL) {
	umem_free(cmdbuf, MAXPATHLEN);
	cmd = NULL;
	}

	while (waitpid(pid, &status, 0) != pid)
	continue;
	if (statusp != NULL)
	*statusp = status;

	if (WIFEXITED(status)) {
	if (WEXITSTATUS(status) != 0) {
	(void) fprintf(stderr, "child exited with code %d\n",
	WEXITSTATUS(status));
	exit(2);
	}
	return (B_FALSE);
	} else if (WIFSIGNALED(status)) {
	if (!ignorekill \|\| WTERMSIG(status) != SIGKILL) {
	(void) fprintf(stderr, "child died with signal %d\n",
	WTERMSIG(status));
	exit(3);
	}
	return (B_TRUE);
	} else {
	(void) fprintf(stderr, "something strange happened to child\n");
	exit(4);
	}
	}

	static void
	ztest_run_init(void)
	{
	int i;

	ztest_shared_t *zs = ztest_shared;

	/*
	* Blow away any existing copy of zpool.cache
	*/
	(void) remove(spa_config_path);

	if (ztest_opts.zo_init == 0) {
	if (ztest_opts.zo_verbose >= 1)
	(void) printf("Importing pool %s\n",
	ztest_opts.zo_pool);
	ztest_import(zs);
	return;
	}

	/*
	* Create and initialize our storage pool.
	*/
	for (i = 1; i <= ztest_opts.zo_init; i++) {
	memset(zs, 0, sizeof (*zs));
	if (ztest_opts.zo_verbose >= 3 &&
	ztest_opts.zo_init != 1) {
	(void) printf("ztest_init(), pass %d\n", i);
	}
	ztest_init(zs);
	}
	}

	int
	main(int argc, char **argv)
	{
	int kills = 0;
	int iters = 0;
	int older = 0;
	int newer = 0;
	ztest_shared_t *zs;
	ztest_info_t *zi;
	ztest_shared_callstate_t *zc;
	char timebuf[100];
	char numbuf[NN_NUMBUF_SZ];
	char *cmd;
	boolean_t hasalt;
	int f, err;
	char *fd_data_str = getenv("ZTEST_FD_DATA");
	struct sigaction action;

	(void) setvbuf(stdout, NULL, _IOLBF, 0);

	dprintf_setup(&argc, argv);
	zfs_deadman_synctime_ms = 300000;
	zfs_deadman_checktime_ms = 30000;
	/*
	* As two-word space map entries may not come up often (especially
	* if pool and vdev sizes are small) we want to force at least some
	* of them so the feature get tested.
	*/
	zfs_force_some_double_word_sm_entries = B_TRUE;

	/*
	* Verify that even extensively damaged split blocks with many
	* segments can be reconstructed in a reasonable amount of time
	* when reconstruction is known to be possible.
	*
	* Note: the lower this value is, the more damage we inflict, and
	* the more time ztest spends in recovering that damage. We chose
	* to induce damage 1/100th of the time so recovery is tested but
	* not so frequently that ztest doesn't get to test other code paths.
	*/
	zfs_reconstruct_indirect_damage_fraction = 100;

	action.sa_handler = sig_handler;
	sigemptyset(&action.sa_mask);
	action.sa_flags = 0;

	if (sigaction(SIGSEGV, &action, NULL) < 0) {
	(void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n",
	strerror(errno));
	exit(EXIT_FAILURE);
	}

	if (sigaction(SIGABRT, &action, NULL) < 0) {
	(void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n",
	strerror(errno));
	exit(EXIT_FAILURE);
	}

	/*
	* Force random_get_bytes() to use /dev/urandom in order to prevent
	* ztest from needlessly depleting the system entropy pool.
	*/
	random_path = "/dev/urandom";
	ztest_fd_rand = open(random_path, O_RDONLY \| O_CLOEXEC);
	ASSERT3S(ztest_fd_rand, >=, 0);

	if (!fd_data_str) {
	process_options(argc, argv);

	setup_data_fd();
	setup_hdr();
	setup_data();
	memcpy(ztest_shared_opts, &ztest_opts,
	sizeof (*ztest_shared_opts));
	} else {
	ztest_fd_data = atoi(fd_data_str);
	setup_data();
	memcpy(&ztest_opts, ztest_shared_opts, sizeof (ztest_opts));
	}
	ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count);

	err = ztest_set_global_vars();
	if (err != 0 && !fd_data_str) {
	/* error message done by ztest_set_global_vars */
	exit(EXIT_FAILURE);
	} else {
	/* children should not be spawned if setting gvars fails */
	VERIFY3S(err, ==, 0);
	}

	/* Override location of zpool.cache */
	VERIFY3S(asprintf((char **)&spa_config_path, "%s/zpool.cache",
	ztest_opts.zo_dir), !=, -1);

	ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t),
	UMEM_NOFAIL);
	zs = ztest_shared;

	if (fd_data_str) {
	metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging;
	metaslab_df_alloc_threshold =
	zs->zs_metaslab_df_alloc_threshold;

	if (zs->zs_do_init)
	ztest_run_init();
	else
	ztest_run(zs);
	exit(0);
	}

	hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0);

	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("%"PRIu64" vdevs, %d datasets, %d threads,"
	"%d %s disks, %"PRIu64" seconds...\n\n",
	ztest_opts.zo_vdevs,
	ztest_opts.zo_datasets,
	ztest_opts.zo_threads,
	ztest_opts.zo_raid_children,
	ztest_opts.zo_raid_type,
	ztest_opts.zo_time);
	}

	cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL);
	(void) strlcpy(cmd, getexecname(), MAXNAMELEN);

	zs->zs_do_init = B_TRUE;
	if (strlen(ztest_opts.zo_alt_ztest) != 0) {
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing older ztest for "
	"initialization: %s\n", ztest_opts.zo_alt_ztest);
	}
	VERIFY(!exec_child(ztest_opts.zo_alt_ztest,
	ztest_opts.zo_alt_libpath, B_FALSE, NULL));
	} else {
	VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL));
	}
	zs->zs_do_init = B_FALSE;

	zs->zs_proc_start = gethrtime();
	zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC;

	for (f = 0; f < ZTEST_FUNCS; f++) {
	zi = &ztest_info[f];
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop)
	zc->zc_next = UINT64_MAX;
	else
	zc->zc_next = zs->zs_proc_start +
	ztest_random(2 * zi->zi_interval[0] + 1);
	}

	/*
	* Run the tests in a loop. These tests include fault injection
	* to verify that self-healing data works, and forced crashes
	* to verify that we never lose on-disk consistency.
	*/
	while (gethrtime() < zs->zs_proc_stop) {
	int status;
	boolean_t killed;

	/*
	* Initialize the workload counters for each function.
	*/
	for (f = 0; f < ZTEST_FUNCS; f++) {
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	zc->zc_count = 0;
	zc->zc_time = 0;
	}

	/* Set the allocation switch size */
	zs->zs_metaslab_df_alloc_threshold =
	ztest_random(zs->zs_metaslab_sz / 4) + 1;

	if (!hasalt \|\| ztest_random(2) == 0) {
	if (hasalt && ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing newer ztest: %s\n",
	cmd);
	}
	newer++;
	killed = exec_child(cmd, NULL, B_TRUE, &status);
	} else {
	if (hasalt && ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing older ztest: %s\n",
	ztest_opts.zo_alt_ztest);
	}
	older++;
	killed = exec_child(ztest_opts.zo_alt_ztest,
	ztest_opts.zo_alt_libpath, B_TRUE, &status);
	}

	if (killed)
	kills++;
	iters++;

	if (ztest_opts.zo_verbose >= 1) {
	hrtime_t now = gethrtime();

	now = MIN(now, zs->zs_proc_stop);
	print_time(zs->zs_proc_stop - now, timebuf);
	nicenum(zs->zs_space, numbuf, sizeof (numbuf));

	(void) printf("Pass %3d, %8s, %3"PRIu64" ENOSPC, "
	"%4.1f%% of %5s used, %3.0f%% done, %8s to go\n",
	iters,
	WIFEXITED(status) ? "Complete" : "SIGKILL",
	zs->zs_enospc_count,
	100.0 * zs->zs_alloc / zs->zs_space,
	numbuf,
	100.0 * (now - zs->zs_proc_start) /
	(ztest_opts.zo_time * NANOSEC), timebuf);
	}

	if (ztest_opts.zo_verbose >= 2) {
	(void) printf("\nWorkload summary:\n\n");
	(void) printf("%7s %9s %s\n",
	"Calls", "Time", "Function");
	(void) printf("%7s %9s %s\n",
	"-----", "----", "--------");
	for (f = 0; f < ZTEST_FUNCS; f++) {
	zi = &ztest_info[f];
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	print_time(zc->zc_time, timebuf);
	(void) printf("%7"PRIu64" %9s %s\n",
	zc->zc_count, timebuf,
	zi->zi_funcname);
	}
	(void) printf("\n");
	}

	if (!ztest_opts.zo_mmp_test)
	ztest_run_zdb(ztest_opts.zo_pool);
	}

	if (ztest_opts.zo_verbose >= 1) {
	if (hasalt) {
	(void) printf("%d runs of older ztest: %s\n", older,
	ztest_opts.zo_alt_ztest);
	(void) printf("%d runs of newer ztest: %s\n", newer,
	cmd);
	}
	(void) printf("%d killed, %d completed, %.0f%% kill rate\n",
	kills, iters - kills, (100.0 * kills) / MAX(1, iters));
	}

	umem_free(cmd, MAXNAMELEN);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/config/Rules.am b/sys/contrib/openzfs/config/Rules.am
	index 7162b771869d..abb4ced33233 100644
	--- a/sys/contrib/openzfs/config/Rules.am
	+++ b/sys/contrib/openzfs/config/Rules.am
	@@ -1,83 +1,85 @@
	#
	# Default build rules for all user space components, every Makefile.am
	# should include these rules and override or extend them as needed.
	#

	PHONY =
	AM_CPPFLAGS = \
	-include $(top_builddir)/zfs_config.h \
	-I$(top_builddir)/include \
	-I$(top_srcdir)/include \
	-I$(top_srcdir)/module/icp/include \
	-I$(top_srcdir)/lib/libspl/include \
	-I$(top_srcdir)/lib/libspl/include/os/@ac_system_l@

	AM_LIBTOOLFLAGS = --silent

	AM_CFLAGS = -std=gnu99 -Wall -Wextra -Wstrict-prototypes -Wmissing-prototypes -Wwrite-strings -Wno-sign-compare -Wno-missing-field-initializers
	AM_CFLAGS += -fno-strict-aliasing
	AM_CFLAGS += $(NO_OMIT_FRAME_POINTER)
	AM_CFLAGS += $(IMPLICIT_FALLTHROUGH)
	AM_CFLAGS += $(DEBUG_CFLAGS)
	AM_CFLAGS += $(ASAN_CFLAGS)
	AM_CFLAGS += $(UBSAN_CFLAGS)
	AM_CFLAGS += $(CODE_COVERAGE_CFLAGS) $(NO_FORMAT_ZERO_LENGTH)
	if BUILD_FREEBSD
	AM_CFLAGS += -fPIC -Werror -Wno-unknown-pragmas -Wno-enum-conversion
	AM_CFLAGS += -include $(top_srcdir)/include/os/freebsd/spl/sys/ccompile.h
	AM_CFLAGS += -I/usr/include -I/usr/local/include
	endif

	AM_CPPFLAGS += -D_GNU_SOURCE
	AM_CPPFLAGS += -D_REENTRANT
	AM_CPPFLAGS += -D_FILE_OFFSET_BITS=64
	AM_CPPFLAGS += -D_LARGEFILE64_SOURCE
	AM_CPPFLAGS += -DLIBEXECDIR=\"$(libexecdir)\"
	AM_CPPFLAGS += -DRUNSTATEDIR=\"$(runstatedir)\"
	AM_CPPFLAGS += -DSBINDIR=\"$(sbindir)\"
	AM_CPPFLAGS += -DSYSCONFDIR=\"$(sysconfdir)\"
	AM_CPPFLAGS += -DPKGDATADIR=\"$(pkgdatadir)\"
	AM_CPPFLAGS += $(DEBUG_CPPFLAGS)
	AM_CPPFLAGS += $(CODE_COVERAGE_CPPFLAGS)
	AM_CPPFLAGS += -DTEXT_DOMAIN=\"zfs-@ac_system_l@-user\"

	AM_CPPFLAGS_NOCHECK = -D"strtok(...)=strtok(__VA_ARGS__) __attribute__((deprecated(\"Use strtok_r(3) instead!\")))"
	AM_CPPFLAGS_NOCHECK += -D"__xpg_basename(...)=__xpg_basename(__VA_ARGS__) __attribute__((deprecated(\"basename(3) is underspecified. Use zfs_basename() instead!\")))"
	AM_CPPFLAGS_NOCHECK += -D"basename(...)=basename(__VA_ARGS__) __attribute__((deprecated(\"basename(3) is underspecified. Use zfs_basename() instead!\")))"
	AM_CPPFLAGS_NOCHECK += -D"dirname(...)=dirname(__VA_ARGS__) __attribute__((deprecated(\"dirname(3) is underspecified. Use zfs_dirnamelen() instead!\")))"
	AM_CPPFLAGS_NOCHECK += -D"bcopy(...)=__attribute__((deprecated(\"bcopy(3) is deprecated. Use memcpy(3)/memmove(3) instead!\"))) bcopy(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"bcmp(...)=__attribute__((deprecated(\"bcmp(3) is deprecated. Use memcmp(3) instead!\"))) bcmp(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"bzero(...)=__attribute__((deprecated(\"bzero(3) is deprecated. Use memset(3) instead!\"))) bzero(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"asctime(...)=__attribute__((deprecated(\"Use strftime(3) instead!\"))) asctime(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"asctime_r(...)=__attribute__((deprecated(\"Use strftime(3) instead!\"))) asctime_r(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"gmtime(...)=__attribute__((deprecated(\"gmtime(3) isn't thread-safe. Use gmtime_r(3) instead!\"))) gmtime(__VA_ARGS__)"
	AM_CPPFLAGS_NOCHECK += -D"localtime(...)=__attribute__((deprecated(\"localtime(3) isn't thread-safe. Use localtime_r(3) instead!\"))) localtime(__VA_ARGS__)"
	+AM_CPPFLAGS_NOCHECK += -D"strncpy(...)=__attribute__((deprecated(\"strncpy(3) is deprecated. Use strlcpy(3) instead!\"))) strncpy(__VA_ARGS__)"
	+
	AM_CPPFLAGS += $(AM_CPPFLAGS_NOCHECK)

	if ASAN_ENABLED
	AM_CPPFLAGS += -DZFS_ASAN_ENABLED
	endif

	if UBSAN_ENABLED
	AM_CPPFLAGS += -DZFS_UBSAN_ENABLED
	endif

	AM_LDFLAGS = $(DEBUG_LDFLAGS)
	AM_LDFLAGS += $(ASAN_LDFLAGS)
	AM_LDFLAGS += $(UBSAN_LDFLAGS)

	if BUILD_FREEBSD
	AM_LDFLAGS += -fstack-protector-strong -shared
	AM_LDFLAGS += -Wl,-x -Wl,--fatal-warnings -Wl,--warn-shared-textrel
	AM_LDFLAGS += -lm
	endif


	# If a target includes kernel code, generate warnings for large stack frames
	KERNEL_CFLAGS = $(FRAME_LARGER_THAN)

	# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
	LIBRARY_CFLAGS = -no-suppress

	# Forcibly enable asserts/debugging for libzpool &al.
	FORCEDEBUG_CPPFLAGS = -DDEBUG -UNDEBUG -DZFS_DEBUG
	diff --git a/sys/contrib/openzfs/config/kernel-blkdev.m4 b/sys/contrib/openzfs/config/kernel-blkdev.m4
	index 462d6c6efa8e..28e5364581ea 100644
	--- a/sys/contrib/openzfs/config/kernel-blkdev.m4
	+++ b/sys/contrib/openzfs/config/kernel-blkdev.m4
	@@ -1,424 +1,479 @@
	dnl #
	dnl # 2.6.38 API change,
	dnl # Added blkdev_get_by_path()
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH], [
	ZFS_LINUX_TEST_SRC([blkdev_get_by_path], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev __attribute__ ((unused)) = NULL;
	const char *path = "path";
	fmode_t mode = 0;
	void *holder = NULL;

	bdev = blkdev_get_by_path(path, mode, holder);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_GET_BY_PATH], [
	AC_MSG_CHECKING([whether blkdev_get_by_path() exists])
	ZFS_LINUX_TEST_RESULT([blkdev_get_by_path], [
	AC_MSG_RESULT(yes)
	], [
	ZFS_LINUX_TEST_ERROR([blkdev_get_by_path()])
	])
	])

	dnl #
	dnl # 2.6.38 API change,
	dnl # Added blkdev_put()
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PUT], [
	ZFS_LINUX_TEST_SRC([blkdev_put], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev = NULL;
	fmode_t mode = 0;

	blkdev_put(bdev, mode);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_PUT], [
	AC_MSG_CHECKING([whether blkdev_put() exists])
	ZFS_LINUX_TEST_RESULT([blkdev_put], [
	AC_MSG_RESULT(yes)
	], [
	ZFS_LINUX_TEST_ERROR([blkdev_put()])
	])
	])

	dnl #
	dnl # 4.1 API, exported blkdev_reread_part() symbol, back ported to the
	dnl # 3.10.0 CentOS 7.x enterprise kernels.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_REREAD_PART], [
	ZFS_LINUX_TEST_SRC([blkdev_reread_part], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev = NULL;
	int error;

	error = blkdev_reread_part(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_REREAD_PART], [
	AC_MSG_CHECKING([whether blkdev_reread_part() exists])
	ZFS_LINUX_TEST_RESULT([blkdev_reread_part], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BLKDEV_REREAD_PART, 1,
	[blkdev_reread_part() exists])
	], [
	AC_MSG_RESULT(no)
	])
	])

	dnl #
	dnl # check_disk_change() was removed in 5.10
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_CHECK_DISK_CHANGE], [
	ZFS_LINUX_TEST_SRC([check_disk_change], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev = NULL;
	bool error;

	error = check_disk_change(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE], [
	AC_MSG_CHECKING([whether check_disk_change() exists])
	ZFS_LINUX_TEST_RESULT([check_disk_change], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_CHECK_DISK_CHANGE, 1,
	[check_disk_change() exists])
	], [
	AC_MSG_RESULT(no)
	])
	])

	+dnl #
	+dnl # bdev_kobj() is introduced from 5.12
	+dnl #
	+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ], [
	+ ZFS_LINUX_TEST_SRC([bdev_kobj], [
	+ #include <linux/fs.h>
	+ #include <linux/blkdev.h>
	+ #include <linux/kobject.h>
	+ ], [
	+ struct block_device *bdev = NULL;
	+ struct kobject *disk_kobj;
	+ disk_kobj = bdev_kobj(bdev);
	+ ])
	+])
	+
	+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ], [
	+ AC_MSG_CHECKING([whether bdev_kobj() exists])
	+ ZFS_LINUX_TEST_RESULT([bdev_kobj], [
	+ AC_MSG_RESULT(yes)
	+ AC_DEFINE(HAVE_BDEV_KOBJ, 1,
	+ [bdev_kobj() exists])
	+ ], [
	+ AC_MSG_RESULT(no)
	+ ])
	+])
	+
	+dnl #
	+dnl # part_to_dev() was removed in 5.12
	+dnl #
	+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV], [
	+ ZFS_LINUX_TEST_SRC([part_to_dev], [
	+ #include <linux/fs.h>
	+ #include <linux/blkdev.h>
	+ ], [
	+ struct hd_struct *p = NULL;
	+ struct device *pdev;
	+ pdev = part_to_dev(p);
	+ ])
	+])
	+
	+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV], [
	+ AC_MSG_CHECKING([whether part_to_dev() exists])
	+ ZFS_LINUX_TEST_RESULT([part_to_dev], [
	+ AC_MSG_RESULT(yes)
	+ AC_DEFINE(HAVE_PART_TO_DEV, 1,
	+ [part_to_dev() exists])
	+ ], [
	+ AC_MSG_RESULT(no)
	+ ])
	+])
	+
	dnl #
	dnl # 5.10 API, check_disk_change() is removed, in favor of
	dnl # bdev_check_media_change(), which doesn't force revalidation
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_CHECK_MEDIA_CHANGE], [
	ZFS_LINUX_TEST_SRC([bdev_check_media_change], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev = NULL;
	int error;

	error = bdev_check_media_change(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_CHECK_MEDIA_CHANGE], [
	AC_MSG_CHECKING([whether bdev_check_media_change() exists])
	ZFS_LINUX_TEST_RESULT([bdev_check_media_change], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BDEV_CHECK_MEDIA_CHANGE, 1,
	[bdev_check_media_change() exists])
	], [
	AC_MSG_RESULT(no)
	])
	])

	dnl #
	dnl # 2.6.22 API change
	dnl # Single argument invalidate_bdev()
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_INVALIDATE_BDEV], [
	ZFS_LINUX_TEST_SRC([invalidate_bdev], [
	#include <linux/buffer_head.h>
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev = NULL;
	invalidate_bdev(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_INVALIDATE_BDEV], [
	AC_MSG_CHECKING([whether invalidate_bdev() exists])
	ZFS_LINUX_TEST_RESULT([invalidate_bdev], [
	AC_MSG_RESULT(yes)
	],[
	ZFS_LINUX_TEST_ERROR([invalidate_bdev()])
	])
	])

	dnl #
	dnl # 5.11 API, lookup_bdev() takes dev_t argument.
	dnl # 2.6.27 API, lookup_bdev() was first exported.
	dnl # 4.4.0-6.21 API, lookup_bdev() on Ubuntu takes mode argument.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_LOOKUP_BDEV], [
	ZFS_LINUX_TEST_SRC([lookup_bdev_devt], [
	#include <linux/blkdev.h>
	], [
	int error __attribute__ ((unused));
	const char path[] = "/example/path";
	dev_t dev;

	error = lookup_bdev(path, &dev);
	])

	ZFS_LINUX_TEST_SRC([lookup_bdev_1arg], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev __attribute__ ((unused));
	const char path[] = "/example/path";

	bdev = lookup_bdev(path);
	])

	ZFS_LINUX_TEST_SRC([lookup_bdev_mode], [
	#include <linux/fs.h>
	], [
	struct block_device *bdev __attribute__ ((unused));
	const char path[] = "/example/path";

	bdev = lookup_bdev(path, FMODE_READ);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_LOOKUP_BDEV], [
	AC_MSG_CHECKING([whether lookup_bdev() wants dev_t arg])
	ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_devt],
	[lookup_bdev], [fs/block_dev.c], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_DEVT_LOOKUP_BDEV, 1,
	[lookup_bdev() wants dev_t arg])
	], [
	AC_MSG_RESULT(no)

	AC_MSG_CHECKING([whether lookup_bdev() wants 1 arg])
	ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_1arg],
	[lookup_bdev], [fs/block_dev.c], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_1ARG_LOOKUP_BDEV, 1,
	[lookup_bdev() wants 1 arg])
	], [
	AC_MSG_RESULT(no)

	AC_MSG_CHECKING([whether lookup_bdev() wants mode arg])
	ZFS_LINUX_TEST_RESULT_SYMBOL([lookup_bdev_mode],
	[lookup_bdev], [fs/block_dev.c], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_MODE_LOOKUP_BDEV, 1,
	[lookup_bdev() wants mode arg])
	], [
	ZFS_LINUX_TEST_ERROR([lookup_bdev()])
	])
	])
	])
	])

	dnl #
	dnl # 2.6.30 API change
	dnl #
	dnl # The bdev_physical_block_size() interface was added to provide a way
	dnl # to determine the smallest write which can be performed without a
	dnl # read-modify-write operation.
	dnl #
	dnl # Unfortunately, this interface isn't entirely reliable because
	dnl # drives are sometimes known to misreport this value.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE], [
	ZFS_LINUX_TEST_SRC([bdev_physical_block_size], [
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev __attribute__ ((unused)) = NULL;
	bdev_physical_block_size(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE], [
	AC_MSG_CHECKING([whether bdev_physical_block_size() is available])
	ZFS_LINUX_TEST_RESULT([bdev_physical_block_size], [
	AC_MSG_RESULT(yes)
	],[
	ZFS_LINUX_TEST_ERROR([bdev_physical_block_size()])
	])
	])

	dnl #
	dnl # 2.6.30 API change
	dnl # Added bdev_logical_block_size().
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE], [
	ZFS_LINUX_TEST_SRC([bdev_logical_block_size], [
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev __attribute__ ((unused)) = NULL;
	bdev_logical_block_size(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE], [
	AC_MSG_CHECKING([whether bdev_logical_block_size() is available])
	ZFS_LINUX_TEST_RESULT([bdev_logical_block_size], [
	AC_MSG_RESULT(yes)
	],[
	ZFS_LINUX_TEST_ERROR([bdev_logical_block_size()])
	])
	])

	dnl #
	dnl # 5.11 API change
	dnl # Added bdev_whole() helper.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE], [
	ZFS_LINUX_TEST_SRC([bdev_whole], [
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev = NULL;
	bdev = bdev_whole(bdev);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE], [
	AC_MSG_CHECKING([whether bdev_whole() is available])
	ZFS_LINUX_TEST_RESULT([bdev_whole], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BDEV_WHOLE, 1, [bdev_whole() is available])
	],[
	AC_MSG_RESULT(no)
	])
	])

	dnl #
	dnl # 5.20 API change,
	dnl # Removed bdevname(), snprintf(.., %pg) should be used.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME], [
	ZFS_LINUX_TEST_SRC([bdevname], [
	#include <linux/fs.h>
	#include <linux/blkdev.h>
	], [
	struct block_device *bdev __attribute__ ((unused)) = NULL;
	char path[BDEVNAME_SIZE];

	(void) bdevname(bdev, path);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEVNAME], [
	AC_MSG_CHECKING([whether bdevname() exists])
	ZFS_LINUX_TEST_RESULT([bdevname], [
	AC_DEFINE(HAVE_BDEVNAME, 1, [bdevname() is available])
	AC_MSG_RESULT(yes)
	], [
	AC_MSG_RESULT(no)
	])
	])

	dnl #
	dnl # 5.19 API: blkdev_issue_secure_erase()
	dnl # 3.10 API: blkdev_issue_discard(..., BLKDEV_DISCARD_SECURE)
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE], [
	ZFS_LINUX_TEST_SRC([blkdev_issue_secure_erase], [
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev = NULL;
	sector_t sector = 0;
	sector_t nr_sects = 0;
	int error __attribute__ ((unused));

	error = blkdev_issue_secure_erase(bdev,
	sector, nr_sects, GFP_KERNEL);
	])

	ZFS_LINUX_TEST_SRC([blkdev_issue_discard_flags], [
	#include <linux/blkdev.h>
	],[
	struct block_device *bdev = NULL;
	sector_t sector = 0;
	sector_t nr_sects = 0;
	unsigned long flags = 0;
	int error __attribute__ ((unused));

	error = blkdev_issue_discard(bdev,
	sector, nr_sects, GFP_KERNEL, flags);
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE], [
	AC_MSG_CHECKING([whether blkdev_issue_secure_erase() is available])
	ZFS_LINUX_TEST_RESULT([blkdev_issue_secure_erase], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BLKDEV_ISSUE_SECURE_ERASE, 1,
	[blkdev_issue_secure_erase() is available])
	],[
	AC_MSG_RESULT(no)

	AC_MSG_CHECKING([whether blkdev_issue_discard() is available])
	ZFS_LINUX_TEST_RESULT([blkdev_issue_discard_flags], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BLKDEV_ISSUE_DISCARD, 1,
	[blkdev_issue_discard() is available])
	],[
	ZFS_LINUX_TEST_ERROR([blkdev_issue_discard()])
	])
	])
	])

	dnl #
	dnl # 5.13 API change
	dnl # blkdev_get_by_path() no longer handles ERESTARTSYS
	dnl #
	dnl # Unfortunately we're forced to rely solely on the kernel version
	dnl # number in order to determine the expected behavior. This was an
	dnl # internal change to blkdev_get_by_dev(), see commit a8ed1a0607.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS], [
	AC_MSG_CHECKING([whether blkdev_get_by_path() handles ERESTARTSYS])
	AS_VERSION_COMPARE([$LINUX_VERSION], [5.13.0], [
	AC_MSG_RESULT(yes)
	AC_DEFINE(HAVE_BLKDEV_GET_ERESTARTSYS, 1,
	[blkdev_get_by_path() handles ERESTARTSYS])
	],[
	AC_MSG_RESULT(no)
	],[
	AC_MSG_RESULT(no)
	])
	])

	AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [
	ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH
	ZFS_AC_KERNEL_SRC_BLKDEV_PUT
	ZFS_AC_KERNEL_SRC_BLKDEV_REREAD_PART
	ZFS_AC_KERNEL_SRC_BLKDEV_INVALIDATE_BDEV
	ZFS_AC_KERNEL_SRC_BLKDEV_LOOKUP_BDEV
	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE
	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE
	ZFS_AC_KERNEL_SRC_BLKDEV_CHECK_DISK_CHANGE
	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_CHECK_MEDIA_CHANGE
	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE
	ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME
	ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE
	+ ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ
	+ ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV
	])

	AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
	ZFS_AC_KERNEL_BLKDEV_GET_BY_PATH
	ZFS_AC_KERNEL_BLKDEV_PUT
	ZFS_AC_KERNEL_BLKDEV_REREAD_PART
	ZFS_AC_KERNEL_BLKDEV_INVALIDATE_BDEV
	ZFS_AC_KERNEL_BLKDEV_LOOKUP_BDEV
	ZFS_AC_KERNEL_BLKDEV_BDEV_LOGICAL_BLOCK_SIZE
	ZFS_AC_KERNEL_BLKDEV_BDEV_PHYSICAL_BLOCK_SIZE
	ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE
	ZFS_AC_KERNEL_BLKDEV_BDEV_CHECK_MEDIA_CHANGE
	ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE
	ZFS_AC_KERNEL_BLKDEV_BDEVNAME
	ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS
	ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE
	+ ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ
	+ ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV
	])
	diff --git a/sys/contrib/openzfs/contrib/coverity/model.c b/sys/contrib/openzfs/contrib/coverity/model.c
	index d27abd038767..8baa3a7e2d31 100644
	--- a/sys/contrib/openzfs/contrib/coverity/model.c
	+++ b/sys/contrib/openzfs/contrib/coverity/model.c
	@@ -1,422 +1,359 @@
	/*
	* Coverity Scan model
	* https://scan.coverity.com/models
	*
	* This is a modeling file for Coverity Scan.
	* Modeling helps to avoid false positives.
	*
	* - Modeling doesn't need full structs and typedefs. Rudimentary structs
	* and similar types are sufficient.
	* - An uninitialized local pointer is not an error. It signifies that the
	* variable could be either NULL or have some data.
	*
	* Coverity Scan doesn't pick up modifications automatically. The model file
	* must be uploaded by an admin in the analysis settings.
	*
	* Some of this initially cribbed from:
	*
	* https://github.com/kees/coverity-linux/blob/trunk/model.c
	*
	* The below model was based on the original model by Brian Behlendorf for the
	* original zfsonlinux/zfs repository. Some inspiration was taken from
	* kees/coverity-linux, specifically involving memory copies.
	*/

	#include <stdarg.h>

	#define UMEM_DEFAULT 0x0000 /* normal -- may fail */
	#define UMEM_NOFAIL 0x0100 /* Never fails */

	#define NULL (0)

	int condition0, condition1;

	-void
	-abort()
	-{
	- __coverity_panic__();
	-}
	-
	-void
	-exit(int status)
	-{
	- (void) status;
	-
	- __coverity_panic__();
	-}
	-
	-void
	-_exit(int status)
	-{
	- (void) status;
	-
	- __coverity_panic__();
	-}
	-
	-void
	-zed_log_die(const char *fmt, ...)
	-{
	- __coverity_format_string_sink__(fmt);
	- __coverity_panic__();
	-}
	-
	-void
	-panic(const char *fmt, ...)
	-{
	- __coverity_format_string_sink__(fmt);
	- __coverity_panic__();
	-}
	-
	-void
	-vpanic(const char *fmt, va_list adx)
	-{
	- (void) adx;
	-
	- __coverity_format_string_sink__(fmt);
	- __coverity_panic__();
	-}
	-
	-void
	-uu_panic(const char *format, ...)
	-{
	- __coverity_format_string_sink__(format);
	- __coverity_panic__();
	-}
	-
	-int
	-libspl_assertf(const char file, const char func, int line,
	- const char *format, ...)
	-{
	- __coverity_format_string_sink__(format);
	- __coverity_panic__();
	-}
	-
	int
	ddi_copyin(const void from, void to, size_t len, int flags)
	{
	__coverity_tainted_data_argument__(from);
	__coverity_tainted_data_argument__(to);
	__coverity_writeall__(to);
	}

	void *
	memset(void *dst, int c, size_t len)
	{
	__coverity_writeall__(dst);
	return (dst);
	}

	void *
	memmove(void dst, void src, size_t len)
	{
	__coverity_writeall__(dst);
	return (dst);
	}

	void *
	memcpy(void dst, void src, size_t len)
	{
	__coverity_writeall__(dst);
	return (dst);
	}

	void *
	umem_alloc_aligned(size_t size, size_t align, int kmflags)
	{
	(void) align;

	- if (UMEM_NOFAIL & kmflags == UMEM_NOFAIL)
	+ if ((UMEM_NOFAIL & kmflags) == UMEM_NOFAIL)
	return (__coverity_alloc__(size));
	else if (condition0)
	return (__coverity_alloc__(size));
	else
	return (NULL);
	}

	void *
	umem_alloc(size_t size, int kmflags)
	{
	- if (UMEM_NOFAIL & kmflags == UMEM_NOFAIL)
	+ if ((UMEM_NOFAIL & kmflags) == UMEM_NOFAIL)
	return (__coverity_alloc__(size));
	else if (condition0)
	return (__coverity_alloc__(size));
	else
	return (NULL);
	}

	void *
	umem_zalloc(size_t size, int kmflags)
	{
	- if (UMEM_NOFAIL & kmflags == UMEM_NOFAIL)
	+ if ((UMEM_NOFAIL & kmflags) == UMEM_NOFAIL)
	return (__coverity_alloc__(size));
	else if (condition0)
	return (__coverity_alloc__(size));
	else
	return (NULL);
	}

	void
	umem_free(void *buf, size_t size)
	{
	(void) size;

	__coverity_free__(buf);
	}

	+typedef struct {} umem_cache_t;
	+
	+void *
	+umem_cache_alloc(umem_cache_t *skc, int flags)
	+{
	+ (void) skc;
	+
	+ if (condition1)
	+ __coverity_sleep__();
	+
	+ if ((UMEM_NOFAIL & flags) == UMEM_NOFAIL)
	+ return (__coverity_alloc_nosize__());
	+ else if (condition0)
	+ return (__coverity_alloc_nosize__());
	+ else
	+ return (NULL);
	+}
	+
	+void
	+umem_cache_free(umem_cache_t skc, void obj)
	+{
	+ (void) skc;
	+
	+ __coverity_free__(obj);
	+}
	+
	void *
	spl_kmem_alloc(size_t sz, int fl, const char *func, int line)
	{
	(void) func;
	(void) line;

	if (condition1)
	__coverity_sleep__();

	if (fl == 0) {
	return (__coverity_alloc__(sz));
	} else if (condition0)
	return (__coverity_alloc__(sz));
	else
	return (NULL);
	}

	void *
	spl_kmem_zalloc(size_t sz, int fl, const char *func, int line)
	{
	(void) func;
	(void) line;

	if (condition1)
	__coverity_sleep__();

	if (fl == 0) {
	return (__coverity_alloc__(sz));
	} else if (condition0)
	return (__coverity_alloc__(sz));
	else
	return (NULL);
	}

	void
	spl_kmem_free(const void *ptr, size_t sz)
	{
	(void) sz;

	__coverity_free__(ptr);
	}

	typedef struct {} spl_kmem_cache_t;

	void *
	spl_kmem_cache_alloc(spl_kmem_cache_t *skc, int flags)
	{
	(void) skc;

	if (condition1)
	__coverity_sleep__();

	if (flags == 0) {
	return (__coverity_alloc_nosize__());
	} else if (condition0)
	return (__coverity_alloc_nosize__());
	else
	return (NULL);
	}

	void
	spl_kmem_cache_free(spl_kmem_cache_t skc, void obj)
	{
	(void) skc;

	__coverity_free__(obj);
	}

	void
	malloc(size_t size)
	{
	__coverity_alloc__(size);
	}

	void
	free(void *buf)
	{
	__coverity_free__(buf);
	}

	-int
	-spl_panic(const char file, const char func, int line, const char *fmt, ...)
	-{
	- __coverity_format_string_sink__(fmt);
	- __coverity_panic__();
	-}
	-
	int
	sched_yield(void)
	{
	__coverity_sleep__();
	}

	typedef struct {} kmutex_t;
	typedef struct {} krwlock_t;
	typedef int krw_t;

	/*
	* Coverty reportedly does not support macros, so this only works for
	* userspace.
	*/

	void
	mutex_enter(kmutex_t *mp)
	{
	if (condition0)
	__coverity_sleep__();

	__coverity_exclusive_lock_acquire__(mp);
	}

	int
	mutex_tryenter(kmutex_t *mp)
	{
	if (condition0) {
	__coverity_exclusive_lock_acquire__(mp);
	return (1);
	}

	return (0);
	}

	void
	mutex_exit(kmutex_t *mp)
	{
	__coverity_exclusive_lock_release__(mp);
	}

	void
	rw_enter(krwlock_t *rwlp, krw_t rw)
	{
	(void) rw;

	if (condition0)
	__coverity_sleep__();

	__coverity_recursive_lock_acquire__(rwlp);
	}

	void
	rw_exit(krwlock_t *rwlp)
	{
	__coverity_recursive_lock_release__(rwlp);

	}

	int
	rw_tryenter(krwlock_t *rwlp, krw_t rw)
	{
	if (condition0) {
	__coverity_recursive_lock_acquire__(rwlp);
	return (1);
	}

	return (0);
	}

	/* Thus, we fallback to the Linux kernel locks */
	struct {} mutex;
	struct {} rw_semaphore;

	void
	mutex_lock(struct mutex *lock)
	{
	if (condition0) {
	__coverity_sleep__();
	}
	__coverity_exclusive_lock_acquire__(lock);
	}

	void
	mutex_unlock(struct mutex *lock)
	{
	__coverity_exclusive_lock_release__(lock);
	}

	void
	down_read(struct rw_semaphore *sem)
	{
	if (condition0) {
	__coverity_sleep__();
	}
	__coverity_recursive_lock_acquire__(sem);
	}

	void
	down_write(struct rw_semaphore *sem)
	{
	if (condition0) {
	__coverity_sleep__();
	}
	__coverity_recursive_lock_acquire__(sem);
	}

	int
	down_read_trylock(struct rw_semaphore *sem)
	{
	if (condition0) {
	__coverity_recursive_lock_acquire__(sem);
	return (1);
	}

	return (0);
	}

	int
	down_write_trylock(struct rw_semaphore *sem)
	{
	if (condition0) {
	__coverity_recursive_lock_acquire__(sem);
	return (1);
	}

	return (0);
	}

	void
	up_read(struct rw_semaphore *sem)
	{
	__coverity_recursive_lock_release__(sem);
	}

	void
	up_write(struct rw_semaphore *sem)
	{
	__coverity_recursive_lock_release__(sem);
	}

	int
	__cond_resched(void)
	{
	if (condition0) {
	__coverity_sleep__();
	}
	}
	-
	-/*
	- * An endian-independent filesystem must support doing byte swaps on data. We
	- * attempt to suppress taint warnings, which are false positives for us.
	- */
	-void
	-byteswap_uint64_array(void *vbuf, size_t size)
	-{
	- __coverity_tainted_data_sanitize__(vbuf);
	-}
	-
	-void
	-byteswap_uint32_array(void *vbuf, size_t size)
	-{
	- __coverity_tainted_data_sanitize__(vbuf);
	-}
	-
	-void
	-byteswap_uint16_array(void *vbuf, size_t size)
	-{
	- __coverity_tainted_data_sanitize__(vbuf);
	-}
	diff --git a/sys/contrib/openzfs/contrib/pam_zfs_key/pam_zfs_key.c b/sys/contrib/openzfs/contrib/pam_zfs_key/pam_zfs_key.c
	index c1001e6b81c2..8c59fc7eb716 100644
	--- a/sys/contrib/openzfs/contrib/pam_zfs_key/pam_zfs_key.c
	+++ b/sys/contrib/openzfs/contrib/pam_zfs_key/pam_zfs_key.c
	@@ -1,848 +1,850 @@
	/*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions are met:
	* * Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* * Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	* * Neither the name of the <organization> nor the
	* names of its contributors may be used to endorse or promote products
	* derived from this software without specific prior written permission.
	*
	* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
	* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL <COPYRIGHT HOLDER> BE LIABLE FOR ANY
	* DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
	* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
	* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
	* ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
	* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
	* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
	*
	* Copyright (c) 2020, Felix Dörre
	* All rights reserved.
	*/

	#include <sys/dsl_crypt.h>
	#include <sys/byteorder.h>
	#include <libzfs.h>

	#include <syslog.h>

	#include <sys/zio_crypt.h>
	#include <openssl/evp.h>

	#define PAM_SM_AUTH
	#define PAM_SM_PASSWORD
	#define PAM_SM_SESSION
	#include <security/pam_modules.h>

	#if defined(__linux__)
	#include <security/pam_ext.h>
	#define MAP_FLAGS MAP_PRIVATE \| MAP_ANONYMOUS
	#elif defined(__FreeBSD__)
	#include <security/pam_appl.h>
	static void
	pam_syslog(pam_handle_t pamh, int loglevel, const char fmt, ...)
	{
	(void) pamh;
	va_list args;
	va_start(args, fmt);
	vsyslog(loglevel, fmt, args);
	va_end(args);
	}
	#define MAP_FLAGS MAP_PRIVATE \| MAP_ANON \| MAP_NOCORE
	#endif

	#include <string.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <sys/stat.h>
	#include <sys/file.h>
	#include <sys/wait.h>
	#include <pwd.h>

	#include <sys/mman.h>

	static const char PASSWORD_VAR_NAME[] = "pam_zfs_key_authtok";

	static libzfs_handle_t *g_zfs;

	static void destroy_pw(pam_handle_t pamh, void data, int errcode);

	typedef int (mlock_func_t) (const void , size_t);

	typedef struct {
	size_t len;
	char *value;
	} pw_password_t;

	/*
	* Try to mlock(2) or munlock(2) addr while handling EAGAIN by retrying ten
	* times and sleeping 10 milliseconds in between for a total of 0.1
	* seconds. lock_func must point to either mlock(2) or munlock(2).
	*/
	static int
	try_lock(mlock_func_t lock_func, const void *addr, size_t len)
	{
	int err;
	int retries = 10;
	useconds_t sleep_dur = 10 * 1000;

	if ((err = (*lock_func)(addr, len)) != EAGAIN) {
	return (err);
	}
	for (int i = retries; i > 0; --i) {
	(void) usleep(sleep_dur);
	if ((err = (*lock_func)(addr, len)) != EAGAIN) {
	break;
	}
	}
	return (err);
	}


	static pw_password_t *
	alloc_pw_size(size_t len)
	{
	pw_password_t *pw = malloc(sizeof (pw_password_t));
	if (!pw) {
	return (NULL);
	}
	pw->len = len;
	/*
	* We use mmap(2) rather than malloc(3) since later on we mlock(2) the
	* memory region. Since mlock(2) and munlock(2) operate on whole memory
	* pages we should allocate a whole page here as mmap(2) does. Further
	* this ensures that the addresses passed to mlock(2) an munlock(2) are
	* on a page boundary as suggested by FreeBSD and required by some
	* other implementations. Finally we avoid inadvertently munlocking
	* memory mlocked by an concurrently running instance of us.
	*/
	pw->value = mmap(NULL, pw->len, PROT_READ \| PROT_WRITE, MAP_FLAGS,
	-1, 0);

	if (pw->value == MAP_FAILED) {
	free(pw);
	return (NULL);
	}
	if (try_lock(mlock, pw->value, pw->len) != 0) {
	(void) munmap(pw->value, pw->len);
	free(pw);
	return (NULL);
	}
	return (pw);
	}

	static pw_password_t *
	alloc_pw_string(const char *source)
	{
	size_t len = strlen(source) + 1;
	pw_password_t *pw = alloc_pw_size(len);

	if (!pw) {
	return (NULL);
	}
	memcpy(pw->value, source, pw->len);
	return (pw);
	}

	static void
	pw_free(pw_password_t *pw)
	{
	memset(pw->value, 0, pw->len);
	if (try_lock(munlock, pw->value, pw->len) == 0) {
	(void) munmap(pw->value, pw->len);
	}
	free(pw);
	}

	static pw_password_t *
	pw_fetch(pam_handle_t *pamh)
	{
	const char *token;
	if (pam_get_authtok(pamh, PAM_AUTHTOK, &token, NULL) != PAM_SUCCESS) {
	pam_syslog(pamh, LOG_ERR,
	"couldn't get password from PAM stack");
	return (NULL);
	}
	if (!token) {
	pam_syslog(pamh, LOG_ERR,
	"token from PAM stack is null");
	return (NULL);
	}
	return (alloc_pw_string(token));
	}

	static const pw_password_t *
	pw_fetch_lazy(pam_handle_t *pamh)
	{
	pw_password_t *pw = pw_fetch(pamh);
	if (pw == NULL) {
	return (NULL);
	}
	int ret = pam_set_data(pamh, PASSWORD_VAR_NAME, pw, destroy_pw);
	if (ret != PAM_SUCCESS) {
	pw_free(pw);
	pam_syslog(pamh, LOG_ERR, "pam_set_data failed");
	return (NULL);
	}
	return (pw);
	}

	static const pw_password_t *
	pw_get(pam_handle_t *pamh)
	{
	const pw_password_t *authtok = NULL;
	int ret = pam_get_data(pamh, PASSWORD_VAR_NAME,
	(const void**)(&authtok));
	if (ret == PAM_SUCCESS)
	return (authtok);
	if (ret == PAM_NO_MODULE_DATA)
	return (pw_fetch_lazy(pamh));
	pam_syslog(pamh, LOG_ERR, "password not available");
	return (NULL);
	}

	static int
	pw_clear(pam_handle_t *pamh)
	{
	int ret = pam_set_data(pamh, PASSWORD_VAR_NAME, NULL, NULL);
	if (ret != PAM_SUCCESS) {
	pam_syslog(pamh, LOG_ERR, "clearing password failed");
	return (-1);
	}
	return (0);
	}

	static void
	destroy_pw(pam_handle_t pamh, void data, int errcode)
	{
	(void) pamh, (void) errcode;

	if (data != NULL) {
	pw_free((pw_password_t *)data);
	}
	}

	static int
	pam_zfs_init(pam_handle_t *pamh)
	{
	int error = 0;
	if ((g_zfs = libzfs_init()) == NULL) {
	error = errno;
	pam_syslog(pamh, LOG_ERR, "Zfs initialization error: %s",
	libzfs_error_init(error));
	}
	return (error);
	}

	static void
	pam_zfs_free(void)
	{
	libzfs_fini(g_zfs);
	}

	static pw_password_t *
	prepare_passphrase(pam_handle_t pamh, zfs_handle_t ds,
	const char passphrase, nvlist_t nvlist)
	{
	pw_password_t *key = alloc_pw_size(WRAPPING_KEY_LEN);
	if (!key) {
	return (NULL);
	}
	uint64_t salt;
	uint64_t iters;
	if (nvlist != NULL) {
	int fd = open("/dev/urandom", O_RDONLY);
	if (fd < 0) {
	pw_free(key);
	return (NULL);
	}
	int bytes_read = 0;
	char buf = (char )&salt;
	size_t bytes = sizeof (uint64_t);
	while (bytes_read < bytes) {
	ssize_t len = read(fd, buf + bytes_read, bytes
	- bytes_read);
	if (len < 0) {
	close(fd);
	pw_free(key);
	return (NULL);
	}
	bytes_read += len;
	}
	close(fd);

	if (nvlist_add_uint64(nvlist,
	zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), salt)) {
	pam_syslog(pamh, LOG_ERR,
	"failed to add salt to nvlist");
	pw_free(key);
	return (NULL);
	}
	iters = DEFAULT_PBKDF2_ITERATIONS;
	if (nvlist_add_uint64(nvlist, zfs_prop_to_name(
	ZFS_PROP_PBKDF2_ITERS), iters)) {
	pam_syslog(pamh, LOG_ERR,
	"failed to add iters to nvlist");
	pw_free(key);
	return (NULL);
	}
	} else {
	salt = zfs_prop_get_int(ds, ZFS_PROP_PBKDF2_SALT);
	iters = zfs_prop_get_int(ds, ZFS_PROP_PBKDF2_ITERS);
	}

	salt = LE_64(salt);
	if (!PKCS5_PBKDF2_HMAC_SHA1((char *)passphrase,
	strlen(passphrase), (uint8_t *)&salt,
	sizeof (uint64_t), iters, WRAPPING_KEY_LEN,
	(uint8_t *)key->value)) {
	pam_syslog(pamh, LOG_ERR, "pbkdf failed");
	pw_free(key);
	return (NULL);
	}
	return (key);
	}

	static int
	is_key_loaded(pam_handle_t pamh, const char ds_name)
	{
	zfs_handle_t *ds = zfs_open(g_zfs, ds_name, ZFS_TYPE_FILESYSTEM);
	if (ds == NULL) {
	pam_syslog(pamh, LOG_ERR, "dataset %s not found", ds_name);
	return (-1);
	}
	int keystatus = zfs_prop_get_int(ds, ZFS_PROP_KEYSTATUS);
	zfs_close(ds);
	return (keystatus != ZFS_KEYSTATUS_UNAVAILABLE);
	}

	static int
	change_key(pam_handle_t pamh, const char ds_name,
	const char *passphrase)
	{
	zfs_handle_t *ds = zfs_open(g_zfs, ds_name, ZFS_TYPE_FILESYSTEM);
	if (ds == NULL) {
	pam_syslog(pamh, LOG_ERR, "dataset %s not found", ds_name);
	return (-1);
	}
	nvlist_t *nvlist = fnvlist_alloc();
	pw_password_t *key = prepare_passphrase(pamh, ds, passphrase, nvlist);
	if (key == NULL) {
	nvlist_free(nvlist);
	zfs_close(ds);
	return (-1);
	}
	if (nvlist_add_string(nvlist,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION),
	"prompt")) {
	pam_syslog(pamh, LOG_ERR, "nvlist_add failed for keylocation");
	pw_free(key);
	nvlist_free(nvlist);
	zfs_close(ds);
	return (-1);
	}
	if (nvlist_add_uint64(nvlist,
	zfs_prop_to_name(ZFS_PROP_KEYFORMAT),
	ZFS_KEYFORMAT_PASSPHRASE)) {
	pam_syslog(pamh, LOG_ERR, "nvlist_add failed for keyformat");
	pw_free(key);
	nvlist_free(nvlist);
	zfs_close(ds);
	return (-1);
	}
	int ret = lzc_change_key(ds_name, DCP_CMD_NEW_KEY, nvlist,
	(uint8_t *)key->value, WRAPPING_KEY_LEN);
	pw_free(key);
	if (ret) {
	pam_syslog(pamh, LOG_ERR, "change_key failed: %d", ret);
	nvlist_free(nvlist);
	zfs_close(ds);
	return (-1);
	}
	nvlist_free(nvlist);
	zfs_close(ds);
	return (0);
	}

	static int
	decrypt_mount(pam_handle_t pamh, const char ds_name,
	const char *passphrase)
	{
	zfs_handle_t *ds = zfs_open(g_zfs, ds_name, ZFS_TYPE_FILESYSTEM);
	if (ds == NULL) {
	pam_syslog(pamh, LOG_ERR, "dataset %s not found", ds_name);
	return (-1);
	}
	pw_password_t *key = prepare_passphrase(pamh, ds, passphrase, NULL);
	if (key == NULL) {
	zfs_close(ds);
	return (-1);
	}
	int ret = lzc_load_key(ds_name, B_FALSE, (uint8_t *)key->value,
	WRAPPING_KEY_LEN);
	pw_free(key);
	if (ret) {
	pam_syslog(pamh, LOG_ERR, "load_key failed: %d", ret);
	zfs_close(ds);
	return (-1);
	}
	ret = zfs_mount(ds, NULL, 0);
	if (ret) {
	pam_syslog(pamh, LOG_ERR, "mount failed: %d", ret);
	zfs_close(ds);
	return (-1);
	}
	zfs_close(ds);
	return (0);
	}

	static int
	unmount_unload(pam_handle_t pamh, const char ds_name)
	{
	zfs_handle_t *ds = zfs_open(g_zfs, ds_name, ZFS_TYPE_FILESYSTEM);
	if (ds == NULL) {
	pam_syslog(pamh, LOG_ERR, "dataset %s not found", ds_name);
	return (-1);
	}
	int ret = zfs_unmount(ds, NULL, 0);
	if (ret) {
	pam_syslog(pamh, LOG_ERR, "zfs_unmount failed with: %d", ret);
	zfs_close(ds);
	return (-1);
	}

	ret = lzc_unload_key(ds_name);
	if (ret) {
	pam_syslog(pamh, LOG_ERR, "unload_key failed with: %d", ret);
	zfs_close(ds);
	return (-1);
	}
	zfs_close(ds);
	return (0);
	}

	typedef struct {
	char *homes_prefix;
	char *runstatedir;
	char *homedir;
	char *dsname;
	uid_t uid;
	const char *username;
	int unmount_and_unload;
	} zfs_key_config_t;

	static int
	zfs_key_config_load(pam_handle_t pamh, zfs_key_config_t config,
	int argc, const char **argv)
	{
	config->homes_prefix = strdup("rpool/home");
	if (config->homes_prefix == NULL) {
	pam_syslog(pamh, LOG_ERR, "strdup failure");
	return (-1);
	}
	config->runstatedir = strdup(RUNSTATEDIR "/pam_zfs_key");
	if (config->runstatedir == NULL) {
	pam_syslog(pamh, LOG_ERR, "strdup failure");
	free(config->homes_prefix);
	return (-1);
	}
	const char *name;
	if (pam_get_user(pamh, &name, NULL) != PAM_SUCCESS) {
	pam_syslog(pamh, LOG_ERR,
	"couldn't get username from PAM stack");
	free(config->runstatedir);
	free(config->homes_prefix);
	return (-1);
	}
	struct passwd *entry = getpwnam(name);
	if (!entry) {
	free(config->runstatedir);
	free(config->homes_prefix);
	return (-1);
	}
	config->uid = entry->pw_uid;
	config->username = name;
	config->unmount_and_unload = 1;
	config->dsname = NULL;
	config->homedir = NULL;
	for (int c = 0; c < argc; c++) {
	if (strncmp(argv[c], "homes=", 6) == 0) {
	free(config->homes_prefix);
	config->homes_prefix = strdup(argv[c] + 6);
	} else if (strncmp(argv[c], "runstatedir=", 12) == 0) {
	free(config->runstatedir);
	config->runstatedir = strdup(argv[c] + 12);
	} else if (strcmp(argv[c], "nounmount") == 0) {
	config->unmount_and_unload = 0;
	} else if (strcmp(argv[c], "prop_mountpoint") == 0) {
	- config->homedir = strdup(entry->pw_dir);
	+ if (config->homedir == NULL)
	+ config->homedir = strdup(entry->pw_dir);
	}
	}
	return (0);
	}

	static void
	zfs_key_config_free(zfs_key_config_t *config)
	{
	free(config->homes_prefix);
	free(config->runstatedir);
	free(config->homedir);
	free(config->dsname);
	}

	static int
	find_dsname_by_prop_value(zfs_handle_t zhp, void data)
	{
	zfs_type_t type = zfs_get_type(zhp);
	zfs_key_config_t *target = data;
	char mountpoint[ZFS_MAXPROPLEN];

	/* Skip any datasets whose type does not match */
	if ((type & ZFS_TYPE_FILESYSTEM) == 0) {
	zfs_close(zhp);
	return (0);
	}

	/* Skip any datasets whose mountpoint does not match */
	(void) zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mountpoint,
	sizeof (mountpoint), NULL, NULL, 0, B_FALSE);
	if (strcmp(target->homedir, mountpoint) != 0) {
	zfs_close(zhp);
	return (0);
	}

	target->dsname = strdup(zfs_get_name(zhp));
	zfs_close(zhp);
	return (1);
	}

	static char *
	zfs_key_config_get_dataset(zfs_key_config_t *config)
	{
	if (config->homedir != NULL &&
	config->homes_prefix != NULL) {
	zfs_handle_t *zhp = zfs_open(g_zfs, config->homes_prefix,
	ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL) {
	pam_syslog(NULL, LOG_ERR, "dataset %s not found",
	config->homes_prefix);
	return (NULL);
	}

	(void) zfs_iter_filesystems(zhp, find_dsname_by_prop_value,
	config);
	zfs_close(zhp);
	char *dsname = config->dsname;
	config->dsname = NULL;
	return (dsname);
	}

	if (config->homes_prefix == NULL) {
	return (NULL);
	}

	size_t len = ZFS_MAX_DATASET_NAME_LEN;
	size_t total_len = strlen(config->homes_prefix) + 1
	+ strlen(config->username);
	if (total_len > len) {
	return (NULL);
	}
	char *ret = malloc(len + 1);
	if (!ret) {
	return (NULL);
	}
	ret[0] = 0;
	- strcat(ret, config->homes_prefix);
	- strcat(ret, "/");
	- strcat(ret, config->username);
	+ (void) snprintf(ret, len + 1, "%s/%s", config->homes_prefix,
	+ config->username);
	return (ret);
	}

	static int
	zfs_key_config_modify_session_counter(pam_handle_t *pamh,
	zfs_key_config_t *config, int delta)
	{
	const char *runtime_path = config->runstatedir;
	if (mkdir(runtime_path, S_IRWXU) != 0 && errno != EEXIST) {
	pam_syslog(pamh, LOG_ERR, "Can't create runtime path: %d",
	errno);
	return (-1);
	}
	if (chown(runtime_path, 0, 0) != 0) {
	pam_syslog(pamh, LOG_ERR, "Can't chown runtime path: %d",
	errno);
	return (-1);
	}
	if (chmod(runtime_path, S_IRWXU) != 0) {
	pam_syslog(pamh, LOG_ERR, "Can't chmod runtime path: %d",
	errno);
	return (-1);
	}
	size_t runtime_path_len = strlen(runtime_path);
	size_t counter_path_len = runtime_path_len + 1 + 10;
	char *counter_path = malloc(counter_path_len + 1);
	if (!counter_path) {
	return (-1);
	}
	counter_path[0] = 0;
	strcat(counter_path, runtime_path);
	snprintf(counter_path + runtime_path_len, counter_path_len, "/%d",
	config->uid);
	const int fd = open(counter_path,
	O_RDWR \| O_CLOEXEC \| O_CREAT \| O_NOFOLLOW,
	S_IRUSR \| S_IWUSR);
	free(counter_path);
	if (fd < 0) {
	pam_syslog(pamh, LOG_ERR, "Can't open counter file: %d", errno);
	return (-1);
	}
	if (flock(fd, LOCK_EX) != 0) {
	pam_syslog(pamh, LOG_ERR, "Can't lock counter file: %d", errno);
	close(fd);
	return (-1);
	}
	char counter[20];
	char *pos = counter;
	int remaining = sizeof (counter) - 1;
	int ret;
	counter[sizeof (counter) - 1] = 0;
	while (remaining > 0 && (ret = read(fd, pos, remaining)) > 0) {
	remaining -= ret;
	pos += ret;
	}
	*pos = 0;
	long int counter_value = strtol(counter, NULL, 10);
	counter_value += delta;
	if (counter_value < 0) {
	counter_value = 0;
	}
	lseek(fd, 0, SEEK_SET);
	if (ftruncate(fd, 0) != 0) {
	pam_syslog(pamh, LOG_ERR, "Can't truncate counter file: %d",
	errno);
	close(fd);
	return (-1);
	}
	snprintf(counter, sizeof (counter), "%ld", counter_value);
	remaining = strlen(counter);
	pos = counter;
	while (remaining > 0 && (ret = write(fd, pos, remaining)) > 0) {
	remaining -= ret;
	pos += ret;
	}
	close(fd);
	return (counter_value);
	}

	__attribute__((visibility("default")))
	PAM_EXTERN int
	pam_sm_authenticate(pam_handle_t *pamh, int flags,
	int argc, const char **argv)
	{
	(void) flags, (void) argc, (void) argv;

	if (pw_fetch_lazy(pamh) == NULL) {
	return (PAM_AUTH_ERR);
	}

	return (PAM_SUCCESS);
	}

	__attribute__((visibility("default")))
	PAM_EXTERN int
	pam_sm_setcred(pam_handle_t *pamh, int flags,
	int argc, const char **argv)
	{
	(void) pamh, (void) flags, (void) argc, (void) argv;
	return (PAM_SUCCESS);
	}

	__attribute__((visibility("default")))
	PAM_EXTERN int
	pam_sm_chauthtok(pam_handle_t *pamh, int flags,
	int argc, const char **argv)
	{
	if (geteuid() != 0) {
	pam_syslog(pamh, LOG_ERR,
	"Cannot zfs_mount when not being root.");
	return (PAM_PERM_DENIED);
	}
	zfs_key_config_t config;
	if (zfs_key_config_load(pamh, &config, argc, argv) == -1) {
	return (PAM_SERVICE_ERR);
	}
	if (config.uid < 1000) {
	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}
	{
	if (pam_zfs_init(pamh) != 0) {
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	char *dataset = zfs_key_config_get_dataset(&config);
	if (!dataset) {
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	int key_loaded = is_key_loaded(pamh, dataset);
	if (key_loaded == -1) {
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	free(dataset);
	pam_zfs_free();
	if (! key_loaded) {
	pam_syslog(pamh, LOG_ERR,
	"key not loaded, returning try_again");
	zfs_key_config_free(&config);
	return (PAM_PERM_DENIED);
	}
	}

	if ((flags & PAM_UPDATE_AUTHTOK) != 0) {
	const pw_password_t *token = pw_get(pamh);
	if (token == NULL) {
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	if (pam_zfs_init(pamh) != 0) {
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	char *dataset = zfs_key_config_get_dataset(&config);
	if (!dataset) {
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	if (change_key(pamh, dataset, token->value) == -1) {
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	if (pw_clear(pamh) == -1) {
	return (PAM_SERVICE_ERR);
	}
	} else {
	zfs_key_config_free(&config);
	}
	return (PAM_SUCCESS);
	}

	PAM_EXTERN int
	pam_sm_open_session(pam_handle_t *pamh, int flags,
	int argc, const char **argv)
	{
	(void) flags;

	if (geteuid() != 0) {
	pam_syslog(pamh, LOG_ERR,
	"Cannot zfs_mount when not being root.");
	return (PAM_SUCCESS);
	}
	zfs_key_config_t config;
	zfs_key_config_load(pamh, &config, argc, argv);
	if (config.uid < 1000) {
	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}

	int counter = zfs_key_config_modify_session_counter(pamh, &config, 1);
	if (counter != 1) {
	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}

	const pw_password_t *token = pw_get(pamh);
	if (token == NULL) {
	zfs_key_config_free(&config);
	return (PAM_SESSION_ERR);
	}
	if (pam_zfs_init(pamh) != 0) {
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	char *dataset = zfs_key_config_get_dataset(&config);
	if (!dataset) {
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	if (decrypt_mount(pamh, dataset, token->value) == -1) {
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	if (pw_clear(pamh) == -1) {
	return (PAM_SERVICE_ERR);
	}
	return (PAM_SUCCESS);

	}

	__attribute__((visibility("default")))
	PAM_EXTERN int
	pam_sm_close_session(pam_handle_t *pamh, int flags,
	int argc, const char **argv)
	{
	(void) flags;

	if (geteuid() != 0) {
	pam_syslog(pamh, LOG_ERR,
	"Cannot zfs_mount when not being root.");
	return (PAM_SUCCESS);
	}
	zfs_key_config_t config;
	- zfs_key_config_load(pamh, &config, argc, argv);
	+ if (zfs_key_config_load(pamh, &config, argc, argv) != 0) {
	+ return (PAM_SESSION_ERR);
	+ }
	if (config.uid < 1000) {
	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}

	int counter = zfs_key_config_modify_session_counter(pamh, &config, -1);
	if (counter != 0) {
	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}

	if (config.unmount_and_unload) {
	if (pam_zfs_init(pamh) != 0) {
	zfs_key_config_free(&config);
	return (PAM_SERVICE_ERR);
	}
	char *dataset = zfs_key_config_get_dataset(&config);
	if (!dataset) {
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SESSION_ERR);
	}
	if (unmount_unload(pamh, dataset) == -1) {
	free(dataset);
	pam_zfs_free();
	zfs_key_config_free(&config);
	return (PAM_SESSION_ERR);
	}
	free(dataset);
	pam_zfs_free();
	}

	zfs_key_config_free(&config);
	return (PAM_SUCCESS);
	}
	diff --git a/sys/contrib/openzfs/contrib/pyzfs/libzfs_core/_constants.py b/sys/contrib/openzfs/contrib/pyzfs/libzfs_core/_constants.py
	index 2db2bba8a824..4db1de8d9a6c 100644
	--- a/sys/contrib/openzfs/contrib/pyzfs/libzfs_core/_constants.py
	+++ b/sys/contrib/openzfs/contrib/pyzfs/libzfs_core/_constants.py
	@@ -1,117 +1,118 @@
	#
	# Copyright 2015 ClusterHQ
	#
	# Licensed under the Apache License, Version 2.0 (the "License");
	# you may not use this file except in compliance with the License.
	# You may obtain a copy of the License at
	#
	# http://www.apache.org/licenses/LICENSE-2.0
	#
	# Unless required by applicable law or agreed to in writing, software
	# distributed under the License is distributed on an "AS IS" BASIS,
	# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
	# See the License for the specific language governing permissions and
	# limitations under the License.
	#

	"""
	Important `libzfs_core` constants.
	"""

	from __future__ import absolute_import, division, print_function
	import errno
	import sys


	# Compat for platform-specific errnos
	if sys.platform.startswith('freebsd'):
	ECHRNG = errno.ENXIO
	ECKSUM = 97 # EINTEGRITY
	ETIME = errno.ETIMEDOUT
	else:
	ECHRNG = errno.ECHRNG
	ECKSUM = errno.EBADE
	ETIME = errno.ETIME


	# https://stackoverflow.com/a/1695250
	def enum_with_offset(offset, sequential, named):
	enums = dict(((b, a + offset) for a, b in enumerate(sequential)), **named)
	return type('Enum', (), enums)


	def enum(sequential, *named):
	return enum_with_offset(0, sequential, named)


	#: Maximum length of any ZFS name.
	MAXNAMELEN = 255
	#: Default channel program limits
	ZCP_DEFAULT_INSTRLIMIT = 10 * 1000 * 1000
	ZCP_DEFAULT_MEMLIMIT = 10 * 1024 * 1024
	#: Encryption wrapping key length
	WRAPPING_KEY_LEN = 32
	#: Encryption key location enum
	zfs_key_location = enum(
	'ZFS_KEYLOCATION_NONE',
	'ZFS_KEYLOCATION_PROMPT',
	'ZFS_KEYLOCATION_URI'
	)
	#: Encryption key format enum
	zfs_keyformat = enum(
	'ZFS_KEYFORMAT_NONE',
	'ZFS_KEYFORMAT_RAW',
	'ZFS_KEYFORMAT_HEX',
	'ZFS_KEYFORMAT_PASSPHRASE'
	)
	# Encryption algorithms enum
	zio_encrypt = enum(
	'ZIO_CRYPT_INHERIT',
	'ZIO_CRYPT_ON',
	'ZIO_CRYPT_OFF',
	'ZIO_CRYPT_AES_128_CCM',
	'ZIO_CRYPT_AES_192_CCM',
	'ZIO_CRYPT_AES_256_CCM',
	'ZIO_CRYPT_AES_128_GCM',
	'ZIO_CRYPT_AES_192_GCM',
	'ZIO_CRYPT_AES_256_GCM'
	)
	# ZFS-specific error codes
	zfs_errno = enum_with_offset(1024, [
	'ZFS_ERR_CHECKPOINT_EXISTS',
	'ZFS_ERR_DISCARDING_CHECKPOINT',
	'ZFS_ERR_NO_CHECKPOINT',
	'ZFS_ERR_DEVRM_IN_PROGRESS',
	'ZFS_ERR_VDEV_TOO_BIG',
	'ZFS_ERR_IOC_CMD_UNAVAIL',
	'ZFS_ERR_IOC_ARG_UNAVAIL',
	'ZFS_ERR_IOC_ARG_REQUIRED',
	'ZFS_ERR_IOC_ARG_BADTYPE',
	'ZFS_ERR_WRONG_PARENT',
	'ZFS_ERR_FROM_IVSET_GUID_MISSING',
	'ZFS_ERR_FROM_IVSET_GUID_MISMATCH',
	'ZFS_ERR_SPILL_BLOCK_FLAG_MISSING',
	'ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE',
	'ZFS_ERR_EXPORT_IN_PROGRESS',
	'ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR',
	'ZFS_ERR_STREAM_TRUNCATED',
	'ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH',
	'ZFS_ERR_RESILVER_IN_PROGRESS',
	'ZFS_ERR_REBUILD_IN_PROGRESS',
	'ZFS_ERR_BADPROP',
	'ZFS_ERR_VDEV_NOTSUP',
	'ZFS_ERR_NOT_USER_NAMESPACE',
	+ 'ZFS_ERR_RESUME_EXISTS',
	],
	{}
	)
	# compat before we used the enum helper for these values
	ZFS_ERR_CHECKPOINT_EXISTS = zfs_errno.ZFS_ERR_CHECKPOINT_EXISTS
	assert (ZFS_ERR_CHECKPOINT_EXISTS == 1024)
	ZFS_ERR_DISCARDING_CHECKPOINT = zfs_errno.ZFS_ERR_DISCARDING_CHECKPOINT
	ZFS_ERR_NO_CHECKPOINT = zfs_errno.ZFS_ERR_NO_CHECKPOINT
	ZFS_ERR_DEVRM_IN_PROGRESS = zfs_errno.ZFS_ERR_DEVRM_IN_PROGRESS
	ZFS_ERR_VDEV_TOO_BIG = zfs_errno.ZFS_ERR_VDEV_TOO_BIG
	ZFS_ERR_WRONG_PARENT = zfs_errno.ZFS_ERR_WRONG_PARENT
	ZFS_ERR_VDEV_NOTSUP = zfs_errno.ZFS_ERR_VDEV_NOTSUP

	# vim: softtabstop=4 tabstop=4 expandtab shiftwidth=4
	diff --git a/sys/contrib/openzfs/etc/init.d/README.md b/sys/contrib/openzfs/etc/init.d/README.md
	index f417b24c5923..2de05042ce63 100644
	--- a/sys/contrib/openzfs/etc/init.d/README.md
	+++ b/sys/contrib/openzfs/etc/init.d/README.md
	@@ -1,75 +1,75 @@
	DESCRIPTION
	These script were written with the primary intention of being portable and
	usable on as many systems as possible.

	This is, in practice, usually not possible. But the intention is there.
	And it is a good one.

	They have been tested successfully on:

	* Debian GNU/Linux Wheezy
	* Debian GNU/Linux Jessie
	* Ubuntu Trusty
	* CentOS 6.0
	* CentOS 6.6
	* Gentoo

	SUPPORT
	If you find that they don't work for your platform, please report this
	at the OpenZFS issue tracker at https://github.com/openzfs/zfs/issues.

	Please include:

	* Distribution name
	* Distribution version
	* Where to find an install CD image
	* Architecture

	If you have code to share that fixes the problem, that is much better.
	But please remember to try your best keep portability in mind. If you
	suspect that what you're writing/modifying won't work on anything else
	than your distribution, please make sure to put that code in appropriate
	if/else/fi code.

	It currently MUST be bash (or fully compatible) for this to work.

	If you're making your own distribution and you want the scripts to
	work on that, the biggest problem you'll (probably) have is the part
	at the beginning of the "zfs-functions" file which sets up the
	logging output.

	INSTALLING INIT SCRIPT LINKS
	To setup the init script links in /etc/rc?.d manually on a Debian GNU/Linux
	(or derived) system, run the following commands (the order is important!):

	update-rc.d zfs-import start 07 S . stop 07 0 1 6 .
	update-rc.d zfs-load-key start 02 2 3 4 5 . stop 06 0 1 6 .
	- update-rc.d zfs-mount start 02 2 3 4 5 . stop 06 0 1 6 .
	+ update-rc.d zfs-mount start 02 S . stop 06 0 1 6 .
	update-rc.d zfs-zed start 07 2 3 4 5 . stop 08 0 1 6 .
	update-rc.d zfs-share start 27 2 3 4 5 . stop 05 0 1 6 .

	To do the same on RedHat, Fedora and/or CentOS:

	chkconfig zfs-import
	chkconfig zfs-load-key
	chkconfig zfs-mount
	chkconfig zfs-zed
	chkconfig zfs-share

	On Gentoo:

	rc-update add zfs-import boot
	rc-update add zfs-load-key boot
	rc-update add zfs-mount boot
	rc-update add zfs-zed default
	rc-update add zfs-share default

	The idea here is to make sure all of the ZFS filesystems, including possibly
	separate datasets like /var, are mounted before anything else is started.

	Then, ZED, which depends on /var, can be started. It will consume and act
	on events that occurred before it started. ZED may also play a role in
	sharing filesystems in the future, so it is important to start before the
	'share' service.

	Finally, we share filesystems configured with the share\* property.
	diff --git a/sys/contrib/openzfs/etc/init.d/zfs-mount.in b/sys/contrib/openzfs/etc/init.d/zfs-mount.in
	index df28c6c951c5..a0825f19fcdd 100755
	--- a/sys/contrib/openzfs/etc/init.d/zfs-mount.in
	+++ b/sys/contrib/openzfs/etc/init.d/zfs-mount.in
	@@ -1,141 +1,142 @@
	#!@DEFAULT_INIT_SHELL@
	# shellcheck disable=SC2154
	#
	# zfs-mount This script will mount/umount the zfs filesystems.
	#
	# chkconfig: 2345 06 99
	# description: This script will mount/umount the zfs filesystems during
	# system boot/shutdown. Configuration of which filesystems
	# should be mounted is handled by the zfs 'mountpoint' and
	# 'canmount' properties. See the zfs(8) man page for details.
	# It is also responsible for all userspace zfs services.
	# probe: true
	#
	### BEGIN INIT INFO
	# Provides: zfs-mount
	-# Required-Start: $local_fs zfs-import
	+# Required-Start: zfs-import
	# Required-Stop: $local_fs zfs-import
	-# Default-Start: 2 3 4 5
	+# Default-Start: S
	# Default-Stop: 0 1 6
	+# X-Start-Before: mountall
	# X-Stop-After: zfs-zed
	# Short-Description: Mount ZFS filesystems and volumes
	# Description: Run the `zfs mount -a` or `zfs umount -a` commands.
	### END INIT INFO
	#
	# Released under the 2-clause BSD license.
	#
	# This script is based on debian/zfsutils.zfs.init from the
	# Debian GNU/kFreeBSD zfsutils 8.1-3 package, written by Aurelien Jarno.

	# Source the common init script
	. @sysconfdir@/zfs/zfs-functions

	# ----------------------------------------------------

	chkroot() {
	while read -r _ mp _; do
	if [ "$mp" = "/" ]; then
	return 0
	fi
	done < /proc/self/mounts

	return 1
	}

	do_depend()
	{
	# Try to allow people to mix and match fstab with ZFS in a way that makes sense.
	if [ "$(mountinfo -s /)" = 'zfs' ]
	then
	before localmount
	else
	after localmount
	fi

	# bootmisc will log to /var which may be a different zfs than root.
	before bootmisc logger

	after zfs-import sysfs
	use mtab
	keyword -lxc -openvz -prefix -vserver
	}

	# Mount all datasets/filesystems
	do_mount()
	{
	local verbose overlay

	check_boolean "$VERBOSE_MOUNT" && verbose=v
	check_boolean "$DO_OVERLAY_MOUNTS" && overlay=O

	zfs_action "Mounting ZFS filesystem(s)" \
	"$ZFS" mount "-a$verbose$overlay" "$MOUNT_EXTRA_OPTIONS"

	return 0
	}

	# Unmount all filesystems
	do_unmount()
	{
	# This shouldn't really be necessary, as long as one can get
	# zfs-import to run sufficiently late in the shutdown/reboot process
	# - after unmounting local filesystems. This is just here in case/if
	# this isn't possible.
	zfs_action "Unmounting ZFS filesystems" "$ZFS" unmount -a

	return 0
	}

	do_start()
	{
	check_boolean "$ZFS_MOUNT" \|\| exit 0

	check_module_loaded "zfs" \|\| exit 0

	# Ensure / exists in /proc/self/mounts.
	# This should be handled by rc.sysinit but lets be paranoid.
	if ! chkroot
	then
	mount -f /
	fi

	do_mount
	}

	do_stop()
	{
	check_boolean "$ZFS_UNMOUNT" \|\| exit 0

	check_module_loaded "zfs" \|\| exit 0

	do_unmount
	}

	# ----------------------------------------------------

	if [ ! -e /sbin/openrc-run ]
	then
	case "$1" in
	start)
	do_start
	;;
	stop)
	do_stop
	;;
	force-reload\|condrestart\|reload\|restart\|status)
	# no-op
	;;
	*)
	[ -n "$1" ] && echo "Error: Unknown command $1."
	echo "Usage: $0 {start\|stop}"
	exit 3
	;;
	esac

	exit $?
	else
	# Create wrapper functions since Gentoo don't use the case part.
	depend() { do_depend; }
	start() { do_start; }
	stop() { do_stop; }
	fi
	diff --git a/sys/contrib/openzfs/include/libuutil_impl.h b/sys/contrib/openzfs/include/libuutil_impl.h
	index b9907447afd9..7aa8f0cedf15 100644
	--- a/sys/contrib/openzfs/include/libuutil_impl.h
	+++ b/sys/contrib/openzfs/include/libuutil_impl.h
	@@ -1,174 +1,175 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2005 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _LIBUUTIL_IMPL_H
	#define _LIBUUTIL_IMPL_H



	#include <libuutil.h>
	#include <pthread.h>

	#include <sys/avl_impl.h>
	#include <sys/byteorder.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	void uu_set_error(uint_t);


	-void uu_panic(const char *format, ...) __attribute__((format(printf, 1, 2)));
	+__attribute__((format(printf, 1, 2), __noreturn__))
	+void uu_panic(const char *format, ...);


	/*
	* For debugging purposes, libuutil keeps around linked lists of all uu_lists
	* and uu_avls, along with pointers to their parents. These can cause false
	* negatives when looking for memory leaks, so we encode the pointers by
	* storing them with swapped endianness; this is not perfect, but it's about
	* the best we can do without wasting a lot of space.
	*/
	#ifdef _LP64
	#define UU_PTR_ENCODE(ptr) BSWAP_64((uintptr_t)(void *)(ptr))
	#else
	#define UU_PTR_ENCODE(ptr) BSWAP_32((uintptr_t)(void *)(ptr))
	#endif

	#define UU_PTR_DECODE(ptr) ((void *)UU_PTR_ENCODE(ptr))

	/*
	* uu_list structures
	*/
	typedef struct uu_list_node_impl {
	struct uu_list_node_impl *uln_next;
	struct uu_list_node_impl *uln_prev;
	} uu_list_node_impl_t;

	struct uu_list_walk {
	uu_list_walk_t *ulw_next;
	uu_list_walk_t *ulw_prev;

	uu_list_t *ulw_list;
	int8_t ulw_dir;
	uint8_t ulw_robust;
	uu_list_node_impl_t *ulw_next_result;
	};

	struct uu_list {
	uintptr_t ul_next_enc;
	uintptr_t ul_prev_enc;

	uu_list_pool_t *ul_pool;
	uintptr_t ul_parent_enc; /* encoded parent pointer */
	size_t ul_offset;
	size_t ul_numnodes;
	uint8_t ul_debug;
	uint8_t ul_sorted;
	uint8_t ul_index; /* mark for uu_list_index_ts */

	uu_list_node_impl_t ul_null_node;
	uu_list_walk_t ul_null_walk; /* for robust walkers */
	};

	#define UU_LIST_PTR(ptr) ((uu_list_t *)UU_PTR_DECODE(ptr))

	#define UU_LIST_POOL_MAXNAME 64

	struct uu_list_pool {
	uu_list_pool_t *ulp_next;
	uu_list_pool_t *ulp_prev;

	char ulp_name[UU_LIST_POOL_MAXNAME];
	size_t ulp_nodeoffset;
	size_t ulp_objsize;
	uu_compare_fn_t *ulp_cmp;
	uint8_t ulp_debug;
	uint8_t ulp_last_index;
	pthread_mutex_t ulp_lock; /* protects null_list */
	uu_list_t ulp_null_list;
	};

	/*
	* uu_avl structures
	*/
	typedef struct avl_node uu_avl_node_impl_t;

	struct uu_avl_walk {
	uu_avl_walk_t *uaw_next;
	uu_avl_walk_t *uaw_prev;

	uu_avl_t *uaw_avl;
	void *uaw_next_result;
	int8_t uaw_dir;
	uint8_t uaw_robust;
	};

	struct uu_avl {
	uintptr_t ua_next_enc;
	uintptr_t ua_prev_enc;

	uu_avl_pool_t *ua_pool;
	uintptr_t ua_parent_enc;
	uint8_t ua_debug;
	uint8_t ua_index; /* mark for uu_avl_index_ts */

	struct avl_tree ua_tree;
	uu_avl_walk_t ua_null_walk;
	};

	#define UU_AVL_PTR(x) ((uu_avl_t *)UU_PTR_DECODE(x))

	#define UU_AVL_POOL_MAXNAME 64

	struct uu_avl_pool {
	uu_avl_pool_t *uap_next;
	uu_avl_pool_t *uap_prev;

	char uap_name[UU_AVL_POOL_MAXNAME];
	size_t uap_nodeoffset;
	size_t uap_objsize;
	uu_compare_fn_t *uap_cmp;
	uint8_t uap_debug;
	uint8_t uap_last_index;
	pthread_mutex_t uap_lock; /* protects null_avl */
	uu_avl_t uap_null_avl;
	};

	/*
	* atfork() handlers
	*/
	void uu_avl_lockup(void);
	void uu_avl_release(void);

	void uu_list_lockup(void);
	void uu_list_release(void);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _LIBUUTIL_IMPL_H */
	diff --git a/sys/contrib/openzfs/include/libzfs.h b/sys/contrib/openzfs/include/libzfs.h
	index 4fc776122596..df17873369ad 100644
	--- a/sys/contrib/openzfs/include/libzfs.h
	+++ b/sys/contrib/openzfs/include/libzfs.h
	@@ -1,1001 +1,1003 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2022 by Delphix. All rights reserved.
	* Copyright Joyent, Inc.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2016, Intel Corporation.
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	#ifndef _LIBZFS_H
	#define _LIBZFS_H extern __attribute__((visibility("default")))

	#include <assert.h>
	#include <libshare.h>
	#include <libnvpair.h>
	#include <sys/mnttab.h>
	#include <sys/param.h>
	#include <sys/types.h>
	#include <sys/fs/zfs.h>
	#include <sys/avl.h>
	#include <libzfs_core.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Miscellaneous ZFS constants
	*/
	#define ZFS_MAXPROPLEN MAXPATHLEN
	#define ZPOOL_MAXPROPLEN MAXPATHLEN

	/*
	* libzfs errors
	*/
	typedef enum zfs_error {
	EZFS_SUCCESS = 0, /* no error -- success */
	EZFS_NOMEM = 2000, /* out of memory */
	EZFS_BADPROP, /* invalid property value */
	EZFS_PROPREADONLY, /* cannot set readonly property */
	EZFS_PROPTYPE, /* property does not apply to dataset type */
	EZFS_PROPNONINHERIT, /* property is not inheritable */
	EZFS_PROPSPACE, /* bad quota or reservation */
	EZFS_BADTYPE, /* dataset is not of appropriate type */
	EZFS_BUSY, /* pool or dataset is busy */
	EZFS_EXISTS, /* pool or dataset already exists */
	EZFS_NOENT, /* no such pool or dataset */
	EZFS_BADSTREAM, /* bad backup stream */
	EZFS_DSREADONLY, /* dataset is readonly */
	EZFS_VOLTOOBIG, /* volume is too large for 32-bit system */
	EZFS_INVALIDNAME, /* invalid dataset name */
	EZFS_BADRESTORE, /* unable to restore to destination */
	EZFS_BADBACKUP, /* backup failed */
	EZFS_BADTARGET, /* bad attach/detach/replace target */
	EZFS_NODEVICE, /* no such device in pool */
	EZFS_BADDEV, /* invalid device to add */
	EZFS_NOREPLICAS, /* no valid replicas */
	EZFS_RESILVERING, /* resilvering (healing reconstruction) */
	EZFS_BADVERSION, /* unsupported version */
	EZFS_POOLUNAVAIL, /* pool is currently unavailable */
	EZFS_DEVOVERFLOW, /* too many devices in one vdev */
	EZFS_BADPATH, /* must be an absolute path */
	EZFS_CROSSTARGET, /* rename or clone across pool or dataset */
	EZFS_ZONED, /* used improperly in local zone */
	EZFS_MOUNTFAILED, /* failed to mount dataset */
	EZFS_UMOUNTFAILED, /* failed to unmount dataset */
	EZFS_UNSHARENFSFAILED, /* failed to unshare over nfs */
	EZFS_SHARENFSFAILED, /* failed to share over nfs */
	EZFS_PERM, /* permission denied */
	EZFS_NOSPC, /* out of space */
	EZFS_FAULT, /* bad address */
	EZFS_IO, /* I/O error */
	EZFS_INTR, /* signal received */
	EZFS_ISSPARE, /* device is a hot spare */
	EZFS_INVALCONFIG, /* invalid vdev configuration */
	EZFS_RECURSIVE, /* recursive dependency */
	EZFS_NOHISTORY, /* no history object */
	EZFS_POOLPROPS, /* couldn't retrieve pool props */
	EZFS_POOL_NOTSUP, /* ops not supported for this type of pool */
	EZFS_POOL_INVALARG, /* invalid argument for this pool operation */
	EZFS_NAMETOOLONG, /* dataset name is too long */
	EZFS_OPENFAILED, /* open of device failed */
	EZFS_NOCAP, /* couldn't get capacity */
	EZFS_LABELFAILED, /* write of label failed */
	EZFS_BADWHO, /* invalid permission who */
	EZFS_BADPERM, /* invalid permission */
	EZFS_BADPERMSET, /* invalid permission set name */
	EZFS_NODELEGATION, /* delegated administration is disabled */
	EZFS_UNSHARESMBFAILED, /* failed to unshare over smb */
	EZFS_SHARESMBFAILED, /* failed to share over smb */
	EZFS_BADCACHE, /* bad cache file */
	EZFS_ISL2CACHE, /* device is for the level 2 ARC */
	EZFS_VDEVNOTSUP, /* unsupported vdev type */
	EZFS_NOTSUP, /* ops not supported on this dataset */
	EZFS_ACTIVE_SPARE, /* pool has active shared spare devices */
	EZFS_UNPLAYED_LOGS, /* log device has unplayed logs */
	EZFS_REFTAG_RELE, /* snapshot release: tag not found */
	EZFS_REFTAG_HOLD, /* snapshot hold: tag already exists */
	EZFS_TAGTOOLONG, /* snapshot hold/rele: tag too long */
	EZFS_PIPEFAILED, /* pipe create failed */
	EZFS_THREADCREATEFAILED, /* thread create failed */
	EZFS_POSTSPLIT_ONLINE, /* onlining a disk after splitting it */
	EZFS_SCRUBBING, /* currently scrubbing */
	EZFS_NO_SCRUB, /* no active scrub */
	EZFS_DIFF, /* general failure of zfs diff */
	EZFS_DIFFDATA, /* bad zfs diff data */
	EZFS_POOLREADONLY, /* pool is in read-only mode */
	EZFS_SCRUB_PAUSED, /* scrub currently paused */
	EZFS_ACTIVE_POOL, /* pool is imported on a different system */
	EZFS_CRYPTOFAILED, /* failed to setup encryption */
	EZFS_NO_PENDING, /* cannot cancel, no operation is pending */
	EZFS_CHECKPOINT_EXISTS, /* checkpoint exists */
	EZFS_DISCARDING_CHECKPOINT, /* currently discarding a checkpoint */
	EZFS_NO_CHECKPOINT, /* pool has no checkpoint */
	EZFS_DEVRM_IN_PROGRESS, /* a device is currently being removed */
	EZFS_VDEV_TOO_BIG, /* a device is too big to be used */
	EZFS_IOC_NOTSUPPORTED, /* operation not supported by zfs module */
	EZFS_TOOMANY, /* argument list too long */
	EZFS_INITIALIZING, /* currently initializing */
	EZFS_NO_INITIALIZE, /* no active initialize */
	EZFS_WRONG_PARENT, /* invalid parent dataset (e.g ZVOL) */
	EZFS_TRIMMING, /* currently trimming */
	EZFS_NO_TRIM, /* no active trim */
	EZFS_TRIM_NOTSUP, /* device does not support trim */
	EZFS_NO_RESILVER_DEFER, /* pool doesn't support resilver_defer */
	EZFS_EXPORT_IN_PROGRESS, /* currently exporting the pool */
	EZFS_REBUILDING, /* resilvering (sequential reconstrution) */
	EZFS_VDEV_NOTSUP, /* ops not supported for this type of vdev */
	EZFS_NOT_USER_NAMESPACE, /* a file is not a user namespace */
	EZFS_CKSUM, /* insufficient replicas */
	+ EZFS_RESUME_EXISTS, /* Resume on existing dataset without force */
	EZFS_UNKNOWN
	} zfs_error_t;

	/*
	* The following data structures are all part
	* of the zfs_allow_t data structure which is
	* used for printing 'allow' permissions.
	* It is a linked list of zfs_allow_t's which
	* then contain avl tree's for user/group/sets/...
	* and each one of the entries in those trees have
	* avl tree's for the permissions they belong to and
	* whether they are local,descendent or local+descendent
	* permissions. The AVL trees are used primarily for
	* sorting purposes, but also so that we can quickly find
	* a given user and or permission.
	*/
	typedef struct zfs_perm_node {
	avl_node_t z_node;
	char z_pname[MAXPATHLEN];
	} zfs_perm_node_t;

	typedef struct zfs_allow_node {
	avl_node_t z_node;
	char z_key[MAXPATHLEN]; /* name, such as joe */
	avl_tree_t z_localdescend; /* local+descendent perms */
	avl_tree_t z_local; /* local permissions */
	avl_tree_t z_descend; /* descendent permissions */
	} zfs_allow_node_t;

	typedef struct zfs_allow {
	struct zfs_allow *z_next;
	char z_setpoint[MAXPATHLEN];
	avl_tree_t z_sets;
	avl_tree_t z_crperms;
	avl_tree_t z_user;
	avl_tree_t z_group;
	avl_tree_t z_everyone;
	} zfs_allow_t;

	/*
	* Basic handle types
	*/
	typedef struct zfs_handle zfs_handle_t;
	typedef struct zpool_handle zpool_handle_t;
	typedef struct libzfs_handle libzfs_handle_t;

	_LIBZFS_H int zpool_wait(zpool_handle_t *, zpool_wait_activity_t);
	_LIBZFS_H int zpool_wait_status(zpool_handle_t *, zpool_wait_activity_t,
	boolean_t , boolean_t );

	/*
	* Library initialization
	*/
	_LIBZFS_H libzfs_handle_t *libzfs_init(void);
	_LIBZFS_H void libzfs_fini(libzfs_handle_t *);

	_LIBZFS_H libzfs_handle_t zpool_get_handle(zpool_handle_t );
	_LIBZFS_H libzfs_handle_t zfs_get_handle(zfs_handle_t );

	_LIBZFS_H void libzfs_print_on_error(libzfs_handle_t *, boolean_t);

	_LIBZFS_H void zfs_save_arguments(int argc, char *, char , int);
	_LIBZFS_H int zpool_log_history(libzfs_handle_t , const char );

	_LIBZFS_H int libzfs_errno(libzfs_handle_t *);
	_LIBZFS_H const char *libzfs_error_init(int);
	_LIBZFS_H const char libzfs_error_action(libzfs_handle_t );
	_LIBZFS_H const char libzfs_error_description(libzfs_handle_t );
	_LIBZFS_H int zfs_standard_error(libzfs_handle_t , int, const char );
	_LIBZFS_H void libzfs_mnttab_init(libzfs_handle_t *);
	_LIBZFS_H void libzfs_mnttab_fini(libzfs_handle_t *);
	_LIBZFS_H void libzfs_mnttab_cache(libzfs_handle_t *, boolean_t);
	_LIBZFS_H int libzfs_mnttab_find(libzfs_handle_t , const char ,
	struct mnttab *);
	_LIBZFS_H void libzfs_mnttab_add(libzfs_handle_t , const char ,
	const char , const char );
	_LIBZFS_H void libzfs_mnttab_remove(libzfs_handle_t , const char );

	/*
	* Basic handle functions
	*/
	_LIBZFS_H zpool_handle_t zpool_open(libzfs_handle_t , const char *);
	_LIBZFS_H zpool_handle_t zpool_open_canfail(libzfs_handle_t , const char *);
	_LIBZFS_H void zpool_close(zpool_handle_t *);
	_LIBZFS_H const char zpool_get_name(zpool_handle_t );
	_LIBZFS_H int zpool_get_state(zpool_handle_t *);
	_LIBZFS_H const char *zpool_state_to_name(vdev_state_t, vdev_aux_t);
	_LIBZFS_H const char *zpool_pool_state_to_name(pool_state_t);
	_LIBZFS_H void zpool_free_handles(libzfs_handle_t *);

	/*
	* Iterate over all active pools in the system.
	*/
	typedef int (zpool_iter_f)(zpool_handle_t , void *);
	_LIBZFS_H int zpool_iter(libzfs_handle_t , zpool_iter_f, void );
	_LIBZFS_H boolean_t zpool_skip_pool(const char *);

	/*
	* Functions to create and destroy pools
	*/
	_LIBZFS_H int zpool_create(libzfs_handle_t , const char , nvlist_t *,
	nvlist_t , nvlist_t );
	_LIBZFS_H int zpool_destroy(zpool_handle_t , const char );
	_LIBZFS_H int zpool_add(zpool_handle_t , nvlist_t );

	typedef struct splitflags {
	/* do not split, but return the config that would be split off */
	int dryrun : 1;

	/* after splitting, import the pool */
	int import : 1;
	int name_flags;
	} splitflags_t;

	typedef struct trimflags {
	/* requested vdevs are for the entire pool */
	boolean_t fullpool;

	/* request a secure trim, requires support from device */
	boolean_t secure;

	/* after starting trim, block until trim completes */
	boolean_t wait;

	/* trim at the requested rate in bytes/second */
	uint64_t rate;
	} trimflags_t;

	/*
	* Functions to manipulate pool and vdev state
	*/
	_LIBZFS_H int zpool_scan(zpool_handle_t *, pool_scan_func_t, pool_scrub_cmd_t);
	_LIBZFS_H int zpool_initialize(zpool_handle_t *, pool_initialize_func_t,
	nvlist_t *);
	_LIBZFS_H int zpool_initialize_wait(zpool_handle_t *, pool_initialize_func_t,
	nvlist_t *);
	_LIBZFS_H int zpool_trim(zpool_handle_t , pool_trim_func_t, nvlist_t ,
	trimflags_t *);

	_LIBZFS_H int zpool_clear(zpool_handle_t , const char , nvlist_t *);
	_LIBZFS_H int zpool_reguid(zpool_handle_t *);
	_LIBZFS_H int zpool_reopen_one(zpool_handle_t , void );

	_LIBZFS_H int zpool_sync_one(zpool_handle_t , void );

	_LIBZFS_H int zpool_vdev_online(zpool_handle_t , const char , int,
	vdev_state_t *);
	_LIBZFS_H int zpool_vdev_offline(zpool_handle_t , const char , boolean_t);
	_LIBZFS_H int zpool_vdev_attach(zpool_handle_t , const char ,
	const char , nvlist_t , int, boolean_t);
	_LIBZFS_H int zpool_vdev_detach(zpool_handle_t , const char );
	_LIBZFS_H int zpool_vdev_remove(zpool_handle_t , const char );
	_LIBZFS_H int zpool_vdev_remove_cancel(zpool_handle_t *);
	_LIBZFS_H int zpool_vdev_indirect_size(zpool_handle_t , const char ,
	uint64_t *);
	_LIBZFS_H int zpool_vdev_split(zpool_handle_t , char , nvlist_t **,
	nvlist_t *, splitflags_t);
	+_LIBZFS_H int zpool_vdev_remove_wanted(zpool_handle_t , const char );

	_LIBZFS_H int zpool_vdev_fault(zpool_handle_t *, uint64_t, vdev_aux_t);
	_LIBZFS_H int zpool_vdev_degrade(zpool_handle_t *, uint64_t, vdev_aux_t);
	_LIBZFS_H int zpool_vdev_clear(zpool_handle_t *, uint64_t);

	_LIBZFS_H nvlist_t zpool_find_vdev(zpool_handle_t , const char , boolean_t ,
	boolean_t , boolean_t );
	_LIBZFS_H nvlist_t zpool_find_vdev_by_physpath(zpool_handle_t , const char *,
	boolean_t , boolean_t , boolean_t *);
	_LIBZFS_H int zpool_label_disk(libzfs_handle_t , zpool_handle_t ,
	const char *);
	_LIBZFS_H uint64_t zpool_vdev_path_to_guid(zpool_handle_t *zhp,
	const char *path);

	_LIBZFS_H const char zpool_get_state_str(zpool_handle_t );

	/*
	* Functions to manage pool properties
	*/
	_LIBZFS_H int zpool_set_prop(zpool_handle_t , const char , const char *);
	_LIBZFS_H int zpool_get_prop(zpool_handle_t , zpool_prop_t, char ,
	size_t proplen, zprop_source_t *, boolean_t literal);
	_LIBZFS_H uint64_t zpool_get_prop_int(zpool_handle_t *, zpool_prop_t,
	zprop_source_t *);
	_LIBZFS_H int zpool_props_refresh(zpool_handle_t *);

	_LIBZFS_H const char *zpool_prop_to_name(zpool_prop_t);
	_LIBZFS_H const char *zpool_prop_values(zpool_prop_t);

	/*
	* Functions to manage vdev properties
	*/
	_LIBZFS_H int zpool_get_vdev_prop_value(nvlist_t , vdev_prop_t, char , char *,
	size_t, zprop_source_t *, boolean_t);
	_LIBZFS_H int zpool_get_vdev_prop(zpool_handle_t , const char , vdev_prop_t,
	char , char , size_t, zprop_source_t *, boolean_t);
	_LIBZFS_H int zpool_get_all_vdev_props(zpool_handle_t , const char ,
	nvlist_t **);
	_LIBZFS_H int zpool_set_vdev_prop(zpool_handle_t , const char , const char *,
	const char *);

	_LIBZFS_H const char *vdev_prop_to_name(vdev_prop_t);
	_LIBZFS_H const char *vdev_prop_values(vdev_prop_t);
	_LIBZFS_H boolean_t vdev_prop_user(const char *name);
	_LIBZFS_H const char *vdev_prop_column_name(vdev_prop_t);
	_LIBZFS_H boolean_t vdev_prop_align_right(vdev_prop_t);

	/*
	* Pool health statistics.
	*/
	typedef enum {
	/*
	* The following correspond to faults as defined in the (fault.fs.zfs.*)
	* event namespace. Each is associated with a corresponding message ID.
	* This must be kept in sync with the zfs_msgid_table in
	* lib/libzfs/libzfs_status.c.
	*/
	ZPOOL_STATUS_CORRUPT_CACHE, /* corrupt /kernel/drv/zpool.cache */
	ZPOOL_STATUS_MISSING_DEV_R, /* missing device with replicas */
	ZPOOL_STATUS_MISSING_DEV_NR, /* missing device with no replicas */
	ZPOOL_STATUS_CORRUPT_LABEL_R, /* bad device label with replicas */
	ZPOOL_STATUS_CORRUPT_LABEL_NR, /* bad device label with no replicas */
	ZPOOL_STATUS_BAD_GUID_SUM, /* sum of device guids didn't match */
	ZPOOL_STATUS_CORRUPT_POOL, /* pool metadata is corrupted */
	ZPOOL_STATUS_CORRUPT_DATA, /* data errors in user (meta)data */
	ZPOOL_STATUS_FAILING_DEV, /* device experiencing errors */
	ZPOOL_STATUS_VERSION_NEWER, /* newer on-disk version */
	ZPOOL_STATUS_HOSTID_MISMATCH, /* last accessed by another system */
	ZPOOL_STATUS_HOSTID_ACTIVE, /* currently active on another system */
	ZPOOL_STATUS_HOSTID_REQUIRED, /* multihost=on and hostid=0 */
	ZPOOL_STATUS_IO_FAILURE_WAIT, /* failed I/O, failmode 'wait' */
	ZPOOL_STATUS_IO_FAILURE_CONTINUE, /* failed I/O, failmode 'continue' */
	ZPOOL_STATUS_IO_FAILURE_MMP, /* failed MMP, failmode not 'panic' */
	ZPOOL_STATUS_BAD_LOG, /* cannot read log chain(s) */
	ZPOOL_STATUS_ERRATA, /* informational errata available */

	/*
	* If the pool has unsupported features but can still be opened in
	* read-only mode, its status is ZPOOL_STATUS_UNSUP_FEAT_WRITE. If the
	* pool has unsupported features but cannot be opened at all, its
	* status is ZPOOL_STATUS_UNSUP_FEAT_READ.
	*/
	ZPOOL_STATUS_UNSUP_FEAT_READ, /* unsupported features for read */
	ZPOOL_STATUS_UNSUP_FEAT_WRITE, /* unsupported features for write */

	/*
	* These faults have no corresponding message ID. At the time we are
	* checking the status, the original reason for the FMA fault (I/O or
	* checksum errors) has been lost.
	*/
	ZPOOL_STATUS_FAULTED_DEV_R, /* faulted device with replicas */
	ZPOOL_STATUS_FAULTED_DEV_NR, /* faulted device with no replicas */

	/*
	* The following are not faults per se, but still an error possibly
	* requiring administrative attention. There is no corresponding
	* message ID.
	*/
	ZPOOL_STATUS_VERSION_OLDER, /* older legacy on-disk version */
	ZPOOL_STATUS_FEAT_DISABLED, /* supported features are disabled */
	ZPOOL_STATUS_RESILVERING, /* device being resilvered */
	ZPOOL_STATUS_OFFLINE_DEV, /* device offline */
	ZPOOL_STATUS_REMOVED_DEV, /* removed device */
	ZPOOL_STATUS_REBUILDING, /* device being rebuilt */
	ZPOOL_STATUS_REBUILD_SCRUB, /* recommend scrubbing the pool */
	ZPOOL_STATUS_NON_NATIVE_ASHIFT, /* (e.g. 512e dev with ashift of 9) */
	ZPOOL_STATUS_COMPATIBILITY_ERR, /* bad 'compatibility' property */
	ZPOOL_STATUS_INCOMPATIBLE_FEAT, /* feature set outside compatibility */

	/*
	* Finally, the following indicates a healthy pool.
	*/
	ZPOOL_STATUS_OK
	} zpool_status_t;

	_LIBZFS_H zpool_status_t zpool_get_status(zpool_handle_t , const char *,
	zpool_errata_t *);
	_LIBZFS_H zpool_status_t zpool_import_status(nvlist_t , const char *,
	zpool_errata_t *);

	/*
	* Statistics and configuration functions.
	*/
	_LIBZFS_H nvlist_t zpool_get_config(zpool_handle_t , nvlist_t **);
	_LIBZFS_H nvlist_t zpool_get_features(zpool_handle_t );
	_LIBZFS_H int zpool_refresh_stats(zpool_handle_t , boolean_t );
	_LIBZFS_H int zpool_get_errlog(zpool_handle_t , nvlist_t *);

	/*
	* Import and export functions
	*/
	_LIBZFS_H int zpool_export(zpool_handle_t , boolean_t, const char );
	_LIBZFS_H int zpool_export_force(zpool_handle_t , const char );
	_LIBZFS_H int zpool_import(libzfs_handle_t , nvlist_t , const char *,
	char *altroot);
	_LIBZFS_H int zpool_import_props(libzfs_handle_t , nvlist_t , const char *,
	nvlist_t *, int);
	_LIBZFS_H void zpool_print_unsup_feat(nvlist_t *config);

	/*
	* Miscellaneous pool functions
	*/
	struct zfs_cmd;

	_LIBZFS_H const char *const zfs_history_event_names[];

	typedef enum {
	VDEV_NAME_PATH = 1 << 0,
	VDEV_NAME_GUID = 1 << 1,
	VDEV_NAME_FOLLOW_LINKS = 1 << 2,
	VDEV_NAME_TYPE_ID = 1 << 3,
	} vdev_name_t;

	_LIBZFS_H char zpool_vdev_name(libzfs_handle_t , zpool_handle_t , nvlist_t ,
	int name_flags);
	_LIBZFS_H int zpool_upgrade(zpool_handle_t *, uint64_t);
	_LIBZFS_H int zpool_get_history(zpool_handle_t , nvlist_t , uint64_t ,
	boolean_t *);
	_LIBZFS_H int zpool_events_next(libzfs_handle_t , nvlist_t , int , unsigned,
	int);
	_LIBZFS_H int zpool_events_clear(libzfs_handle_t , int );
	_LIBZFS_H int zpool_events_seek(libzfs_handle_t *, uint64_t, int);
	_LIBZFS_H void zpool_obj_to_path_ds(zpool_handle_t *, uint64_t, uint64_t,
	char *, size_t);
	_LIBZFS_H void zpool_obj_to_path(zpool_handle_t , uint64_t, uint64_t, char ,
	size_t);
	_LIBZFS_H int zfs_ioctl(libzfs_handle_t , int, struct zfs_cmd );
	_LIBZFS_H void zpool_explain_recover(libzfs_handle_t , const char , int,
	nvlist_t *);
	_LIBZFS_H int zpool_checkpoint(zpool_handle_t *);
	_LIBZFS_H int zpool_discard_checkpoint(zpool_handle_t *);
	_LIBZFS_H boolean_t zpool_is_draid_spare(const char *);

	/*
	* Basic handle manipulations. These functions do not create or destroy the
	* underlying datasets, only the references to them.
	*/
	_LIBZFS_H zfs_handle_t zfs_open(libzfs_handle_t , const char *, int);
	_LIBZFS_H zfs_handle_t zfs_handle_dup(zfs_handle_t );
	_LIBZFS_H void zfs_close(zfs_handle_t *);
	_LIBZFS_H zfs_type_t zfs_get_type(const zfs_handle_t *);
	_LIBZFS_H zfs_type_t zfs_get_underlying_type(const zfs_handle_t *);
	_LIBZFS_H const char zfs_get_name(const zfs_handle_t );
	_LIBZFS_H zpool_handle_t zfs_get_pool_handle(const zfs_handle_t );
	_LIBZFS_H const char zfs_get_pool_name(const zfs_handle_t );

	/*
	* Property management functions. Some functions are shared with the kernel,
	* and are found in sys/fs/zfs.h.
	*/

	/*
	* zfs dataset property management
	*/
	_LIBZFS_H const char *zfs_prop_default_string(zfs_prop_t);
	_LIBZFS_H uint64_t zfs_prop_default_numeric(zfs_prop_t);
	_LIBZFS_H const char *zfs_prop_column_name(zfs_prop_t);
	_LIBZFS_H boolean_t zfs_prop_align_right(zfs_prop_t);

	_LIBZFS_H nvlist_t zfs_valid_proplist(libzfs_handle_t , zfs_type_t,
	nvlist_t , uint64_t, zfs_handle_t , zpool_handle_t *, boolean_t,
	const char *);

	_LIBZFS_H const char *zfs_prop_to_name(zfs_prop_t);
	_LIBZFS_H int zfs_prop_set(zfs_handle_t , const char , const char *);
	_LIBZFS_H int zfs_prop_set_list(zfs_handle_t , nvlist_t );
	_LIBZFS_H int zfs_prop_get(zfs_handle_t , zfs_prop_t, char , size_t,
	zprop_source_t , char , size_t, boolean_t);
	_LIBZFS_H int zfs_prop_get_recvd(zfs_handle_t , const char , char *, size_t,
	boolean_t);
	_LIBZFS_H int zfs_prop_get_numeric(zfs_handle_t , zfs_prop_t, uint64_t ,
	zprop_source_t , char , size_t);
	_LIBZFS_H int zfs_prop_get_userquota_int(zfs_handle_t *zhp,
	const char propname, uint64_t propvalue);
	_LIBZFS_H int zfs_prop_get_userquota(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal);
	_LIBZFS_H int zfs_prop_get_written_int(zfs_handle_t zhp, const char propname,
	uint64_t *propvalue);
	_LIBZFS_H int zfs_prop_get_written(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal);
	_LIBZFS_H int zfs_prop_get_feature(zfs_handle_t zhp, const char propname,
	char *buf, size_t len);
	_LIBZFS_H uint64_t getprop_uint64(zfs_handle_t , zfs_prop_t, char *);
	_LIBZFS_H uint64_t zfs_prop_get_int(zfs_handle_t *, zfs_prop_t);
	_LIBZFS_H int zfs_prop_inherit(zfs_handle_t , const char , boolean_t);
	_LIBZFS_H const char *zfs_prop_values(zfs_prop_t);
	_LIBZFS_H int zfs_prop_is_string(zfs_prop_t prop);
	_LIBZFS_H nvlist_t zfs_get_all_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_user_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_recvd_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_clones_nvl(zfs_handle_t );

	_LIBZFS_H int zfs_wait_status(zfs_handle_t *, zfs_wait_activity_t,
	boolean_t , boolean_t );

	/*
	* zfs encryption management
	*/
	_LIBZFS_H int zfs_crypto_get_encryption_root(zfs_handle_t , boolean_t ,
	char *);
	_LIBZFS_H int zfs_crypto_create(libzfs_handle_t , char , nvlist_t *,
	nvlist_t , boolean_t stdin_available, uint8_t , uint_t );
	_LIBZFS_H int zfs_crypto_clone_check(libzfs_handle_t , zfs_handle_t , char *,
	nvlist_t *);
	_LIBZFS_H int zfs_crypto_attempt_load_keys(libzfs_handle_t , const char );
	_LIBZFS_H int zfs_crypto_load_key(zfs_handle_t , boolean_t, const char );
	_LIBZFS_H int zfs_crypto_unload_key(zfs_handle_t *);
	_LIBZFS_H int zfs_crypto_rewrap(zfs_handle_t , nvlist_t , boolean_t);

	typedef struct zprop_list {
	int pl_prop;
	char *pl_user_prop;
	struct zprop_list *pl_next;
	boolean_t pl_all;
	size_t pl_width;
	size_t pl_recvd_width;
	boolean_t pl_fixed;
	} zprop_list_t;

	_LIBZFS_H int zfs_expand_proplist(zfs_handle_t , zprop_list_t *, boolean_t,
	boolean_t);
	_LIBZFS_H void zfs_prune_proplist(zfs_handle_t , uint8_t );
	_LIBZFS_H int vdev_expand_proplist(zpool_handle_t , const char ,
	zprop_list_t **);

	#define ZFS_MOUNTPOINT_NONE "none"
	#define ZFS_MOUNTPOINT_LEGACY "legacy"

	#define ZFS_FEATURE_DISABLED "disabled"
	#define ZFS_FEATURE_ENABLED "enabled"
	#define ZFS_FEATURE_ACTIVE "active"

	#define ZFS_UNSUPPORTED_INACTIVE "inactive"
	#define ZFS_UNSUPPORTED_READONLY "readonly"

	/*
	* zpool property management
	*/
	_LIBZFS_H int zpool_expand_proplist(zpool_handle_t , zprop_list_t *,
	zfs_type_t, boolean_t);
	_LIBZFS_H int zpool_prop_get_feature(zpool_handle_t , const char , char *,
	size_t);
	_LIBZFS_H const char *zpool_prop_default_string(zpool_prop_t);
	_LIBZFS_H uint64_t zpool_prop_default_numeric(zpool_prop_t);
	_LIBZFS_H const char *zpool_prop_column_name(zpool_prop_t);
	_LIBZFS_H boolean_t zpool_prop_align_right(zpool_prop_t);

	/*
	* Functions shared by zfs and zpool property management.
	*/
	_LIBZFS_H int zprop_iter(zprop_func func, void *cb, boolean_t show_all,
	boolean_t ordered, zfs_type_t type);
	_LIBZFS_H int zprop_get_list(libzfs_handle_t , char , zprop_list_t **,
	zfs_type_t);
	_LIBZFS_H void zprop_free_list(zprop_list_t *);

	#define ZFS_GET_NCOLS 5

	typedef enum {
	GET_COL_NONE,
	GET_COL_NAME,
	GET_COL_PROPERTY,
	GET_COL_VALUE,
	GET_COL_RECVD,
	GET_COL_SOURCE
	} zfs_get_column_t;

	/*
	* Functions for printing zfs or zpool properties
	*/
	typedef struct vdev_cbdata {
	int cb_name_flags;
	char **cb_names;
	unsigned int cb_names_count;
	} vdev_cbdata_t;

	typedef struct zprop_get_cbdata {
	int cb_sources;
	zfs_get_column_t cb_columns[ZFS_GET_NCOLS];
	int cb_colwidths[ZFS_GET_NCOLS + 1];
	boolean_t cb_scripted;
	boolean_t cb_literal;
	boolean_t cb_first;
	zprop_list_t *cb_proplist;
	zfs_type_t cb_type;
	vdev_cbdata_t cb_vdevs;
	} zprop_get_cbdata_t;

	_LIBZFS_H void zprop_print_one_property(const char , zprop_get_cbdata_t ,
	const char , const char , zprop_source_t, const char *,
	const char *);

	/*
	* Iterator functions.
	*/
	typedef int (zfs_iter_f)(zfs_handle_t , void *);
	_LIBZFS_H int zfs_iter_root(libzfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_children(zfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_dependents(zfs_handle_t *, boolean_t, zfs_iter_f,
	void *);
	_LIBZFS_H int zfs_iter_filesystems(zfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_snapshots(zfs_handle_t , boolean_t, zfs_iter_f, void ,
	uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapshots_sorted(zfs_handle_t , zfs_iter_f, void ,
	uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapspec(zfs_handle_t , const char , zfs_iter_f,
	void *);
	_LIBZFS_H int zfs_iter_bookmarks(zfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_mounted(zfs_handle_t , zfs_iter_f, void );

	typedef struct get_all_cb {
	zfs_handle_t **cb_handles;
	size_t cb_alloc;
	size_t cb_used;
	} get_all_cb_t;

	_LIBZFS_H void zfs_foreach_mountpoint(libzfs_handle_t , zfs_handle_t *,
	size_t, zfs_iter_f, void *, boolean_t);
	_LIBZFS_H void libzfs_add_handle(get_all_cb_t , zfs_handle_t );

	/*
	* Functions to create and destroy datasets.
	*/
	_LIBZFS_H int zfs_create(libzfs_handle_t , const char , zfs_type_t,
	nvlist_t *);
	_LIBZFS_H int zfs_create_ancestors(libzfs_handle_t , const char );
	_LIBZFS_H int zfs_destroy(zfs_handle_t *, boolean_t);
	_LIBZFS_H int zfs_destroy_snaps(zfs_handle_t , char , boolean_t);
	_LIBZFS_H int zfs_destroy_snaps_nvl(libzfs_handle_t , nvlist_t , boolean_t);
	_LIBZFS_H int zfs_destroy_snaps_nvl_os(libzfs_handle_t , nvlist_t );
	_LIBZFS_H int zfs_clone(zfs_handle_t , const char , nvlist_t *);
	_LIBZFS_H int zfs_snapshot(libzfs_handle_t , const char , boolean_t,
	nvlist_t *);
	_LIBZFS_H int zfs_snapshot_nvl(libzfs_handle_t hdl, nvlist_t snaps,
	nvlist_t *props);
	_LIBZFS_H int zfs_rollback(zfs_handle_t , zfs_handle_t , boolean_t);

	typedef struct renameflags {
	/* recursive rename */
	int recursive : 1;

	/* don't unmount file systems */
	int nounmount : 1;

	/* force unmount file systems */
	int forceunmount : 1;
	} renameflags_t;

	_LIBZFS_H int zfs_rename(zfs_handle_t , const char , renameflags_t);

	typedef struct sendflags {
	/* Amount of extra information to print. */
	int verbosity;

	/* recursive send (ie, -R) */
	boolean_t replicate;

	/* for recursive send, skip sending missing snapshots */
	boolean_t skipmissing;

	/* for incrementals, do all intermediate snapshots */
	boolean_t doall;

	/* if dataset is a clone, do incremental from its origin */
	boolean_t fromorigin;

	/* field no longer used, maintained for backwards compatibility */
	boolean_t pad;

	/* send properties (ie, -p) */
	boolean_t props;

	/* do not send (no-op, ie. -n) */
	boolean_t dryrun;

	/* parsable verbose output (ie. -P) */
	boolean_t parsable;

	/* show progress (ie. -v) */
	boolean_t progress;

	/* large blocks (>128K) are permitted */
	boolean_t largeblock;

	/* WRITE_EMBEDDED records of type DATA are permitted */
	boolean_t embed_data;

	/* compressed WRITE records are permitted */
	boolean_t compress;

	/* raw encrypted records are permitted */
	boolean_t raw;

	/* only send received properties (ie. -b) */
	boolean_t backup;

	/* include snapshot holds in send stream */
	boolean_t holds;

	/* stream represents a partially received dataset */
	boolean_t saved;
	} sendflags_t;

	typedef boolean_t (snapfilter_cb_t)(zfs_handle_t , void );

	_LIBZFS_H int zfs_send(zfs_handle_t , const char , const char *,
	sendflags_t , int, snapfilter_cb_t, void , nvlist_t **);
	_LIBZFS_H int zfs_send_one(zfs_handle_t , const char , int, sendflags_t *,
	const char *);
	_LIBZFS_H int zfs_send_progress(zfs_handle_t , int, uint64_t , uint64_t *);
	_LIBZFS_H int zfs_send_resume(libzfs_handle_t , sendflags_t , int outfd,
	const char *);
	_LIBZFS_H int zfs_send_saved(zfs_handle_t , sendflags_t , int, const char *);
	_LIBZFS_H nvlist_t zfs_send_resume_token_to_nvlist(libzfs_handle_t hdl,
	const char *token);

	_LIBZFS_H int zfs_promote(zfs_handle_t *);
	_LIBZFS_H int zfs_hold(zfs_handle_t , const char , const char *,
	boolean_t, int);
	_LIBZFS_H int zfs_hold_nvl(zfs_handle_t , int, nvlist_t );
	_LIBZFS_H int zfs_release(zfs_handle_t , const char , const char *,
	boolean_t);
	_LIBZFS_H int zfs_get_holds(zfs_handle_t , nvlist_t *);
	_LIBZFS_H uint64_t zvol_volsize_to_reservation(zpool_handle_t *, uint64_t,
	nvlist_t *);

	typedef int (zfs_userspace_cb_t)(void arg, const char *domain,
	uid_t rid, uint64_t space);

	_LIBZFS_H int zfs_userspace(zfs_handle_t *, zfs_userquota_prop_t,
	zfs_userspace_cb_t, void *);

	_LIBZFS_H int zfs_get_fsacl(zfs_handle_t , nvlist_t *);
	_LIBZFS_H int zfs_set_fsacl(zfs_handle_t , boolean_t, nvlist_t );

	typedef struct recvflags {
	/* print informational messages (ie, -v was specified) */
	boolean_t verbose;

	/* the destination is a prefix, not the exact fs (ie, -d) */
	boolean_t isprefix;

	/*
	* Only the tail of the sent snapshot path is appended to the
	* destination to determine the received snapshot name (ie, -e).
	*/
	boolean_t istail;

	/* do not actually do the recv, just check if it would work (ie, -n) */
	boolean_t dryrun;

	/* rollback/destroy filesystems as necessary (eg, -F) */
	boolean_t force;

	/* set "canmount=off" on all modified filesystems */
	boolean_t canmountoff;

	/*
	* Mark the file systems as "resumable" and do not destroy them if the
	* receive is interrupted
	*/
	boolean_t resumable;

	/* byteswap flag is used internally; callers need not specify */
	boolean_t byteswap;

	/* do not mount file systems as they are extracted (private) */
	boolean_t nomount;

	/* Was holds flag set in the compound header? */
	boolean_t holds;

	/* skip receive of snapshot holds */
	boolean_t skipholds;

	/* mount the filesystem unless nomount is specified */
	boolean_t domount;

	/* force unmount while recv snapshot (private) */
	boolean_t forceunmount;

	/* use this recv to check (and heal if needed) an existing snapshot */
	boolean_t heal;
	} recvflags_t;

	_LIBZFS_H int zfs_receive(libzfs_handle_t , const char , nvlist_t *,
	recvflags_t , int, avl_tree_t );

	typedef enum diff_flags {
	ZFS_DIFF_PARSEABLE = 1 << 0,
	ZFS_DIFF_TIMESTAMP = 1 << 1,
	ZFS_DIFF_CLASSIFY = 1 << 2,
	ZFS_DIFF_NO_MANGLE = 1 << 3
	} diff_flags_t;

	_LIBZFS_H int zfs_show_diffs(zfs_handle_t , int, const char , const char *,
	int);

	/*
	* Miscellaneous functions.
	*/
	_LIBZFS_H const char *zfs_type_to_name(zfs_type_t);
	_LIBZFS_H void zfs_refresh_properties(zfs_handle_t *);
	_LIBZFS_H int zfs_name_valid(const char *, zfs_type_t);
	_LIBZFS_H zfs_handle_t zfs_path_to_zhandle(libzfs_handle_t , const char *,
	zfs_type_t);
	_LIBZFS_H int zfs_parent_name(zfs_handle_t , char , size_t);
	_LIBZFS_H boolean_t zfs_dataset_exists(libzfs_handle_t , const char ,
	zfs_type_t);
	_LIBZFS_H int zfs_spa_version(zfs_handle_t , int );
	_LIBZFS_H boolean_t zfs_bookmark_exists(const char *path);

	/*
	* Mount support functions.
	*/
	_LIBZFS_H boolean_t is_mounted(libzfs_handle_t , const char special, char **);
	_LIBZFS_H boolean_t zfs_is_mounted(zfs_handle_t , char *);
	_LIBZFS_H int zfs_mount(zfs_handle_t , const char , int);
	_LIBZFS_H int zfs_mount_at(zfs_handle_t , const char , int, const char *);
	_LIBZFS_H int zfs_unmount(zfs_handle_t , const char , int);
	_LIBZFS_H int zfs_unmountall(zfs_handle_t *, int);
	_LIBZFS_H int zfs_mount_delegation_check(void);

	#if defined(__linux__) \|\| defined(__APPLE__)
	_LIBZFS_H int zfs_parse_mount_options(const char *mntopts,
	unsigned long mntflags, unsigned long zfsflags, int sloppy, char *badopt,
	char *mtabopt);
	_LIBZFS_H void zfs_adjust_mount_options(zfs_handle_t zhp, const char mntpoint,
	char mntopts, char mtabopt);
	#endif

	/*
	* Share support functions.
	*
	* enum sa_protocol * lists are terminated with SA_NO_PROTOCOL,
	* NULL means "all/any known to this libzfs".
	*/
	#define SA_NO_PROTOCOL -1

	_LIBZFS_H boolean_t zfs_is_shared(zfs_handle_t zhp, char *where,
	const enum sa_protocol *proto);
	_LIBZFS_H int zfs_share(zfs_handle_t zhp, const enum sa_protocol proto);
	_LIBZFS_H int zfs_unshare(zfs_handle_t zhp, const char mountpoint,
	const enum sa_protocol *proto);
	_LIBZFS_H int zfs_unshareall(zfs_handle_t *zhp,
	const enum sa_protocol *proto);
	_LIBZFS_H void zfs_commit_shares(const enum sa_protocol *proto);
	_LIBZFS_H void zfs_truncate_shares(const enum sa_protocol *proto);

	_LIBZFS_H int zfs_nicestrtonum(libzfs_handle_t , const char , uint64_t *);

	/*
	* Utility functions to run an external process.
	*/
	#define STDOUT_VERBOSE 0x01
	#define STDERR_VERBOSE 0x02
	#define NO_DEFAULT_PATH 0x04 /* Don't use $PATH to lookup the command */

	_LIBZFS_H int libzfs_run_process(const char , char *, int);
	_LIBZFS_H int libzfs_run_process_get_stdout(const char , char [], char *[],
	char *[], int );
	_LIBZFS_H int libzfs_run_process_get_stdout_nopath(const char , char [],
	char [], char [], int );

	_LIBZFS_H void libzfs_free_str_array(char **, int);

	_LIBZFS_H boolean_t libzfs_envvar_is_set(const char *);

	/*
	* Utility functions for zfs version
	*/
	_LIBZFS_H const char *zfs_version_userland(void);
	_LIBZFS_H char *zfs_version_kernel(void);
	_LIBZFS_H int zfs_version_print(void);

	/*
	* Given a device or file, determine if it is part of a pool.
	*/
	_LIBZFS_H int zpool_in_use(libzfs_handle_t , int, pool_state_t , char **,
	boolean_t *);

	/*
	* Label manipulation.
	*/
	_LIBZFS_H int zpool_clear_label(int);
	_LIBZFS_H int zpool_set_bootenv(zpool_handle_t , const nvlist_t );
	_LIBZFS_H int zpool_get_bootenv(zpool_handle_t , nvlist_t *);

	/*
	* Management interfaces for SMB ACL files
	*/

	_LIBZFS_H int zfs_smb_acl_add(libzfs_handle_t , char , char , char );
	_LIBZFS_H int zfs_smb_acl_remove(libzfs_handle_t , char , char , char );
	_LIBZFS_H int zfs_smb_acl_purge(libzfs_handle_t , char , char *);
	_LIBZFS_H int zfs_smb_acl_rename(libzfs_handle_t , char , char , char ,
	char *);

	/*
	* Enable and disable datasets within a pool by mounting/unmounting and
	* sharing/unsharing them.
	*/
	_LIBZFS_H int zpool_enable_datasets(zpool_handle_t , const char , int);
	_LIBZFS_H int zpool_disable_datasets(zpool_handle_t *, boolean_t);
	_LIBZFS_H void zpool_disable_datasets_os(zpool_handle_t *, boolean_t);
	_LIBZFS_H void zpool_disable_volume_os(const char *);

	/*
	* Parse a features file for -o compatibility
	*/
	typedef enum {
	ZPOOL_COMPATIBILITY_OK,
	ZPOOL_COMPATIBILITY_WARNTOKEN,
	ZPOOL_COMPATIBILITY_BADTOKEN,
	ZPOOL_COMPATIBILITY_BADFILE,
	ZPOOL_COMPATIBILITY_NOFILES
	} zpool_compat_status_t;

	_LIBZFS_H zpool_compat_status_t zpool_load_compat(const char *,
	boolean_t , char , size_t);

	#ifdef __FreeBSD__

	/*
	* Attach/detach the given filesystem to/from the given jail.
	*/
	_LIBZFS_H int zfs_jail(zfs_handle_t *zhp, int jailid, int attach);

	/*
	* Set loader options for next boot.
	*/
	_LIBZFS_H int zpool_nextboot(libzfs_handle_t *, uint64_t, uint64_t,
	const char *);

	#endif /* __FreeBSD__ */

	#ifdef __linux__

	/*
	* Add or delete the given filesystem to/from the given user namespace.
	*/
	_LIBZFS_H int zfs_userns(zfs_handle_t zhp, const char nspath, int attach);

	#endif

	#ifdef __cplusplus
	}
	#endif

	#endif /* _LIBZFS_H */
	diff --git a/sys/contrib/openzfs/include/libzutil.h b/sys/contrib/openzfs/include/libzutil.h
	index 9b86c351bd2d..0b4075c16016 100644
	--- a/sys/contrib/openzfs/include/libzutil.h
	+++ b/sys/contrib/openzfs/include/libzutil.h
	@@ -1,179 +1,179 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#ifndef _LIBZUTIL_H
	#define _LIBZUTIL_H extern __attribute__((visibility("default")))

	#include <sys/nvpair.h>
	#include <sys/fs/zfs.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Default wait time for a device name to be created.
	*/
	#define DISK_LABEL_WAIT (30 * 1000) /* 30 seconds */


	/*
	* Pool Config Operations
	*
	* These are specific to the library libzfs or libzpool instance.
	*/
	typedef nvlist_t refresh_config_func_t(void , nvlist_t *);

	typedef int pool_active_func_t(void , const char , uint64_t, boolean_t *);

	typedef const struct pool_config_ops {
	refresh_config_func_t *pco_refresh_config;
	pool_active_func_t *pco_pool_active;
	} pool_config_ops_t;

	/*
	* An instance of pool_config_ops_t is expected in the caller's binary.
	*/
	-_LIBZUTIL_H const pool_config_ops_t libzfs_config_ops;
	-_LIBZUTIL_H const pool_config_ops_t libzpool_config_ops;
	+_LIBZUTIL_H pool_config_ops_t libzfs_config_ops;
	+_LIBZUTIL_H pool_config_ops_t libzpool_config_ops;

	typedef struct importargs {
	char *path; / a list of paths to search */
	int paths; /* number of paths to search */
	const char poolname; / name of a pool to find */
	uint64_t guid; /* guid of a pool to find */
	const char cachefile; / cachefile to use for import */
	boolean_t can_be_active; /* can the pool be active? */
	boolean_t scan; /* prefer scanning to libblkid cache */
	nvlist_t policy; / load policy (max txg, rewind, etc.) */
	} importargs_t;

	_LIBZUTIL_H nvlist_t zpool_search_import(void , importargs_t *,
	- const pool_config_ops_t *);
	+ pool_config_ops_t *);
	_LIBZUTIL_H int zpool_find_config(void , const char , nvlist_t **,
	- importargs_t , const pool_config_ops_t );
	+ importargs_t , pool_config_ops_t );

	_LIBZUTIL_H const char * const * zpool_default_search_paths(size_t *count);
	_LIBZUTIL_H int zpool_read_label(int, nvlist_t *, int );
	_LIBZUTIL_H int zpool_label_disk_wait(const char *, int);

	struct udev_device;

	_LIBZUTIL_H int zfs_device_get_devid(struct udev_device , char , size_t);
	_LIBZUTIL_H int zfs_device_get_physical(struct udev_device , char , size_t);

	_LIBZUTIL_H void update_vdev_config_dev_strs(nvlist_t *);

	/*
	* Default device paths
	*/
	#define DISK_ROOT "/dev"
	#define UDISK_ROOT "/dev/disk"
	#define ZVOL_ROOT "/dev/zvol"

	_LIBZUTIL_H int zfs_append_partition(char *path, size_t max_len);
	_LIBZUTIL_H int zfs_resolve_shortname(const char name, char path,
	size_t pathlen);

	_LIBZUTIL_H char zfs_strip_partition(const char );
	_LIBZUTIL_H const char zfs_strip_path(const char );

	_LIBZUTIL_H int zfs_strcmp_pathname(const char , const char , int);

	_LIBZUTIL_H boolean_t zfs_dev_is_dm(const char *);
	_LIBZUTIL_H boolean_t zfs_dev_is_whole_disk(const char *);
	_LIBZUTIL_H int zfs_dev_flush(int);
	_LIBZUTIL_H char zfs_get_underlying_path(const char );
	_LIBZUTIL_H char zfs_get_enclosure_sysfs_path(const char );

	_LIBZUTIL_H boolean_t is_mpath_whole_disk(const char *);

	_LIBZUTIL_H boolean_t zfs_isnumber(const char *);

	/*
	* Formats for iostat numbers. Examples: "12K", "30ms", "4B", "2321234", "-".
	*
	* ZFS_NICENUM_1024: Print kilo, mega, tera, peta, exa..
	* ZFS_NICENUM_BYTES: Print single bytes ("13B"), kilo, mega, tera...
	* ZFS_NICENUM_TIME: Print nanosecs, microsecs, millisecs, seconds...
	* ZFS_NICENUM_RAW: Print the raw number without any formatting
	* ZFS_NICENUM_RAWTIME: Same as RAW, but print dashes ('-') for zero.
	*/
	enum zfs_nicenum_format {
	ZFS_NICENUM_1024 = 0,
	ZFS_NICENUM_BYTES = 1,
	ZFS_NICENUM_TIME = 2,
	ZFS_NICENUM_RAW = 3,
	ZFS_NICENUM_RAWTIME = 4
	};

	/*
	* Convert a number to a human-readable form.
	*/
	_LIBZUTIL_H void zfs_nicebytes(uint64_t, char *, size_t);
	_LIBZUTIL_H void zfs_nicenum(uint64_t, char *, size_t);
	_LIBZUTIL_H void zfs_nicenum_format(uint64_t, char *, size_t,
	enum zfs_nicenum_format);
	_LIBZUTIL_H void zfs_nicetime(uint64_t, char *, size_t);
	_LIBZUTIL_H void zfs_niceraw(uint64_t, char *, size_t);

	#define nicenum(num, buf, size) zfs_nicenum(num, buf, size)

	_LIBZUTIL_H void zpool_dump_ddt(const ddt_stat_t , const ddt_histogram_t );
	_LIBZUTIL_H int zpool_history_unpack(char , uint64_t, uint64_t , nvlist_t ***,
	uint_t *);

	struct zfs_cmd;

	/*
	* List of colors to use
	*/
	#define ANSI_RED "\033[0;31m"
	#define ANSI_YELLOW "\033[0;33m"
	#define ANSI_RESET "\033[0m"
	#define ANSI_BOLD "\033[1m"

	_LIBZUTIL_H void color_start(const char *color);
	_LIBZUTIL_H void color_end(void);
	_LIBZUTIL_H int printf_color(const char color, const char format, ...);

	_LIBZUTIL_H const char zfs_basename(const char path);
	_LIBZUTIL_H ssize_t zfs_dirnamelen(const char *path);

	/*
	* These functions are used by the ZFS libraries and cmd/zpool code, but are
	* not exported in the ABI.
	*/
	typedef int (pool_vdev_iter_f)(void , nvlist_t , void );
	int for_each_vdev_cb(void zhp, nvlist_t nv, pool_vdev_iter_f func,
	void *data);
	int for_each_vdev_in_nvlist(nvlist_t *nvroot, pool_vdev_iter_f func,
	void *data);
	void update_vdevs_config_dev_sysfs_path(nvlist_t *config);
	#ifdef __cplusplus
	}
	#endif

	#endif /* _LIBZUTIL_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/byteorder.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/byteorder.h
	index 4ea56094cea5..0c09757c8e15 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/byteorder.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/byteorder.h
	@@ -1,109 +1,123 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*
	* $FreeBSD$
	*/

	/*
	* Copyright 2005 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */
	/* All Rights Reserved */

	/*
	* University Copyright- Copyright (c) 1982, 1986, 1988
	* The Regents of the University of California
	* All Rights Reserved
	*
	* University Acknowledgment- Portions of this document are derived from
	* software developed by the University of California, Berkeley, and its
	* contributors.
	*/

	#ifndef _OPENSOLARIS_SYS_BYTEORDER_H_
	#define _OPENSOLARIS_SYS_BYTEORDER_H_

	#include <sys/endian.h>

	+#ifdef __COVERITY__
	+/*
	+ * Coverity's taint warnings from byteswapping are false positives for us.
	+ * Suppress them by hiding byteswapping from Coverity.
	+ */
	+#define BSWAP_8(x) ((x) & 0xff)
	+#define BSWAP_16(x) ((x) & 0xffff)
	+#define BSWAP_32(x) ((x) & 0xffffffff)
	+#define BSWAP_64(x) (x)
	+
	+#else /* __COVERITY__ */
	+
	/*
	* Macros to reverse byte order
	*/
	#define BSWAP_8(x) ((x) & 0xff)
	#define BSWAP_16(x) ((BSWAP_8(x) << 8) \| BSWAP_8((x) >> 8))
	#define BSWAP_32(x) ((BSWAP_16(x) << 16) \| BSWAP_16((x) >> 16))
	#define BSWAP_64(x) ((BSWAP_32(x) << 32) \| BSWAP_32((x) >> 32))

	+#endif /* __COVERITY__ */
	+
	#define BMASK_8(x) ((x) & 0xff)
	#define BMASK_16(x) ((x) & 0xffff)
	#define BMASK_32(x) ((x) & 0xffffffff)
	#define BMASK_64(x) (x)

	/*
	* Macros to convert from a specific byte order to/from native byte order
	*/
	#if BYTE_ORDER == _BIG_ENDIAN
	#define BE_8(x) BMASK_8(x)
	#define BE_16(x) BMASK_16(x)
	#define BE_32(x) BMASK_32(x)
	#define BE_64(x) BMASK_64(x)
	#define LE_8(x) BSWAP_8(x)
	#define LE_16(x) BSWAP_16(x)
	#define LE_32(x) BSWAP_32(x)
	#define LE_64(x) BSWAP_64(x)
	#else
	#define LE_8(x) BMASK_8(x)
	#define LE_16(x) BMASK_16(x)
	#define LE_32(x) BMASK_32(x)
	#define LE_64(x) BMASK_64(x)
	#define BE_8(x) BSWAP_8(x)
	#define BE_16(x) BSWAP_16(x)
	#define BE_32(x) BSWAP_32(x)
	#define BE_64(x) BSWAP_64(x)
	#endif

	#if !defined(_STANDALONE)
	#if BYTE_ORDER == _BIG_ENDIAN
	#define htonll(x) BMASK_64(x)
	#define ntohll(x) BMASK_64(x)
	#else /* BYTE_ORDER == _LITTLE_ENDIAN */
	#ifndef __LP64__
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return ((((uint64_t)htonl(n)) << 32) + htonl(n >> 32));
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return ((((uint64_t)ntohl(n)) << 32) + ntohl(n >> 32));
	}
	#else /* !__LP64__ */
	#define htonll(x) BSWAP_64(x)
	#define ntohll(x) BSWAP_64(x)
	#endif /* __LP64__ */
	#endif /* BYTE_ORDER */
	#endif /* _STANDALONE */

	#define BE_IN32(xa) htonl(((uint32_t )(void *)(xa)))

	#endif /* _OPENSOLARIS_SYS_BYTEORDER_H_ */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/cmn_err.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/cmn_err.h
	index 88132337f038..a8f9a88247cd 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/cmn_err.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/cmn_err.h
	@@ -1,82 +1,82 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
	/* All Rights Reserved */


	/*
	* Copyright 2004 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _SYS_CMN_ERR_H
	#define _SYS_CMN_ERR_H

	#if !defined(_ASM)
	#include <sys/_stdarg.h>
	#endif

	#ifdef __cplusplus
	extern "C" {
	#endif

	/* Common error handling severity levels */

	#define CE_CONT 0 /* continuation */
	#define CE_NOTE 1 /* notice */
	#define CE_WARN 2 /* warning */
	#define CE_PANIC 3 /* panic */
	#define CE_IGNORE 4 /* print nothing */

	#ifndef _ASM

	extern void cmn_err(int, const char *, ...)
	__attribute__((format(printf, 2, 3)));

	extern void vzcmn_err(zoneid_t, int, const char *, __va_list)
	__attribute__((format(printf, 3, 0)));

	extern void vcmn_err(int, const char *, __va_list)
	__attribute__((format(printf, 2, 0)));

	extern void zcmn_err(zoneid_t, int, const char *, ...)
	__attribute__((format(printf, 3, 4)));

	extern void vzprintf(zoneid_t, const char *, __va_list)
	__attribute__((format(printf, 2, 0)));

	extern void zprintf(zoneid_t, const char *, ...)
	__attribute__((format(printf, 2, 3)));

	extern void vuprintf(const char *, __va_list)
	__attribute__((format(printf, 1, 0)));

	extern void panic(const char *, ...)
	- __attribute__((format(printf, 1, 2)));
	+ __attribute__((format(printf, 1, 2), __noreturn__));

	#endif /* !_ASM */

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_CMN_ERR_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/debug.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/debug.h
	index 86fad18e17d2..a5bd677b16d4 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/debug.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/debug.h
	@@ -1,171 +1,179 @@
	/*
	* Copyright (c) 2020 iXsystems, Inc.
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*
	* $FreeBSD$
	*/

	/*
	* Available Solaris debug functions. All of the ASSERT() macros will be
	* compiled out when NDEBUG is defined, this is the default behavior for
	* the SPL. To enable assertions use the --enable-debug with configure.
	* The VERIFY() functions are never compiled out and cannot be disabled.
	*
	* PANIC() - Panic the node and print message.
	* ASSERT() - Assert X is true, if not panic.
	* ASSERT3B() - Assert boolean X OP Y is true, if not panic.
	* ASSERT3S() - Assert signed X OP Y is true, if not panic.
	* ASSERT3U() - Assert unsigned X OP Y is true, if not panic.
	* ASSERT3P() - Assert pointer X OP Y is true, if not panic.
	* ASSERT0() - Assert value is zero, if not panic.
	* VERIFY() - Verify X is true, if not panic.
	* VERIFY3B() - Verify boolean X OP Y is true, if not panic.
	* VERIFY3S() - Verify signed X OP Y is true, if not panic.
	* VERIFY3U() - Verify unsigned X OP Y is true, if not panic.
	* VERIFY3P() - Verify pointer X OP Y is true, if not panic.
	* VERIFY0() - Verify value is zero, if not panic.
	*/

	#ifndef _SPL_DEBUG_H
	#define _SPL_DEBUG_H


	/*
	* Common DEBUG functionality.
	*/
	-int spl_panic(const char file, const char func, int line,
	- const char *fmt, ...);
	-void spl_dumpstack(void);
	+extern void spl_panic(const char file, const char func, int line,
	+ const char *fmt, ...) __attribute__((__noreturn__));
	+extern void spl_dumpstack(void);
	+
	+static inline int
	+spl_assert(const char buf, const char file, const char *func, int line)
	+{
	+ spl_panic(file, func, line, "%s", buf);
	+ return (0);
	+}

	#ifndef expect
	#define expect(expr, value) (__builtin_expect((expr), (value)))
	#endif
	#define likely(expr) expect((expr) != 0, 1)
	#define unlikely(expr) expect((expr) != 0, 0)

	#define PANIC(fmt, a...) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, fmt, ## a)

	#define VERIFY(cond) \
	(void) (unlikely(!(cond)) && \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "%s", "VERIFY(" #cond ") failed\n"))
	+ spl_assert("VERIFY(" #cond ") failed\n", \
	+ __FILE__, __FUNCTION__, __LINE__))

	#define VERIFY3B(LEFT, OP, RIGHT) do { \
	const boolean_t _verify3_left = (boolean_t)(LEFT); \
	const boolean_t _verify3_right = (boolean_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%d " #OP " %d)\n", \
	(boolean_t)(_verify3_left), \
	(boolean_t)(_verify3_right)); \
	} while (0)

	#define VERIFY3S(LEFT, OP, RIGHT) do { \
	const int64_t _verify3_left = (int64_t)(LEFT); \
	const int64_t _verify3_right = (int64_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%lld " #OP " %lld)\n", \
	(long long) (_verify3_left), \
	(long long) (_verify3_right)); \
	} while (0)

	#define VERIFY3U(LEFT, OP, RIGHT) do { \
	const uint64_t _verify3_left = (uint64_t)(LEFT); \
	const uint64_t _verify3_right = (uint64_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%llu " #OP " %llu)\n", \
	(unsigned long long) (_verify3_left), \
	(unsigned long long) (_verify3_right)); \
	} while (0)

	#define VERIFY3P(LEFT, OP, RIGHT) do { \
	const uintptr_t _verify3_left = (uintptr_t)(LEFT); \
	const uintptr_t _verify3_right = (uintptr_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%px " #OP " %px)\n", \
	(void *) (_verify3_left), \
	(void *) (_verify3_right)); \
	} while (0)

	#define VERIFY0(RIGHT) do { \
	const int64_t _verify3_left = (int64_t)(0); \
	const int64_t _verify3_right = (int64_t)(RIGHT); \
	if (unlikely(!(_verify3_left == _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(0 == " #RIGHT ") " \
	"failed (0 == %lld)\n", \
	(long long) (_verify3_right)); \
	} while (0)

	/*
	* Debugging disabled (--disable-debug)
	*/
	#ifdef NDEBUG

	#define ASSERT(x) ((void) sizeof ((uintptr_t)(x)))
	#define ASSERT3B(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3S(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3U(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3P(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT0(x) ((void) sizeof ((uintptr_t)(x)))
	#define IMPLY(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))
	#define EQUIV(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))

	/*
	* Debugging enabled (--enable-debug)
	*/
	#else

	#define ASSERT3B VERIFY3B
	#define ASSERT3S VERIFY3S
	#define ASSERT3U VERIFY3U
	#define ASSERT3P VERIFY3P
	#define ASSERT0 VERIFY0
	#define ASSERT VERIFY
	#define IMPLY(A, B) \
	- ((void)(likely((!(A)) \|\| (B)) \|\| \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "(" #A ") implies (" #B ")")))
	+ ((void)(likely((!(A)) \|\| (B)) \|\| \
	+ spl_assert("(" #A ") implies (" #B ")", \
	+ __FILE__, __FUNCTION__, __LINE__)))
	#define EQUIV(A, B) \
	- ((void)(likely(!!(A) == !!(B)) \|\| \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "(" #A ") is equivalent to (" #B ")")))
	+ ((void)(likely(!!(A) == !!(B)) \|\| \
	+ spl_assert("(" #A ") is equivalent to (" #B ")", \
	+ __FILE__, __FUNCTION__, __LINE__)))
	+

	#endif /* NDEBUG */

	#endif /* SPL_DEBUG_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/kmem.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/kmem.h
	index a81cb1fb521d..ae2941b80912 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/kmem.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/kmem.h
	@@ -1,107 +1,110 @@
	/*
	* Copyright (c) 2007 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*
	* $FreeBSD$
	*/

	#ifndef _OPENSOLARIS_SYS_KMEM_H_
	#define _OPENSOLARIS_SYS_KMEM_H_

	#ifdef _KERNEL
	#include <sys/param.h>
	#include <sys/malloc.h>
	#include <sys/vmem.h>
	#include <sys/counter.h>

	#include <vm/uma.h>
	#include <vm/vm.h>
	#include <vm/vm_extern.h>

	MALLOC_DECLARE(M_SOLARIS);

	#define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3))
	#define POINTER_INVALIDATE(pp) ((pp) = (void )((uintptr_t)(*(pp)) \| 0x1))

	#define KM_SLEEP M_WAITOK
	#define KM_PUSHPAGE M_WAITOK
	#define KM_NOSLEEP M_NOWAIT
	#define KM_NORMALPRI 0
	#define KMC_NODEBUG UMA_ZONE_NODUMP

	typedef struct vmem vmem_t;

	-extern char kmem_asprintf(const char , ...);
	-extern char kmem_vasprintf(const char fmt, va_list ap);
	+extern char kmem_asprintf(const char , ...)
	+ __attribute__((format(printf, 1, 2)));
	+extern char kmem_vasprintf(const char fmt, va_list ap)
	+ __attribute__((format(printf, 1, 0)));

	typedef struct kmem_cache {
	char kc_name[32];
	#if !defined(KMEM_DEBUG)
	uma_zone_t kc_zone;
	#else
	size_t kc_size;
	#endif
	int (kc_constructor)(void , void *, int);
	void (kc_destructor)(void , void *);
	void *kc_private;
	} kmem_cache_t;

	extern uint64_t spl_kmem_cache_inuse(kmem_cache_t *cache);
	extern uint64_t spl_kmem_cache_entry_size(kmem_cache_t *cache);

	+__attribute__((alloc_size(1)))
	void *zfs_kmem_alloc(size_t size, int kmflags);
	void zfs_kmem_free(void *buf, size_t size);
	uint64_t kmem_size(void);
	kmem_cache_t kmem_cache_create(const char name, size_t bufsize, size_t align,
	int (constructor)(void , void , int), void (destructor)(void , void ),
	void (reclaim)(void ) __unused, void private, vmem_t vmp, int cflags);
	void kmem_cache_destroy(kmem_cache_t *cache);
	void kmem_cache_alloc(kmem_cache_t cache, int flags);
	void kmem_cache_free(kmem_cache_t cache, void buf);
	boolean_t kmem_cache_reap_active(void);
	void kmem_cache_reap_soon(kmem_cache_t *);
	void kmem_reap(void);
	int kmem_debugging(void);
	void *calloc(size_t n, size_t s);


	#define kmem_cache_reap_now kmem_cache_reap_soon
	#define freemem vm_free_count()
	#define minfree vm_cnt.v_free_min
	#define kmem_alloc(size, kmflags) zfs_kmem_alloc((size), (kmflags))
	#define kmem_zalloc(size, kmflags) \
	zfs_kmem_alloc((size), (kmflags) \| M_ZERO)
	#define kmem_free(buf, size) zfs_kmem_free((buf), (size))

	#endif /* _KERNEL */

	#ifdef _STANDALONE
	/*
	* At the moment, we just need it for the type. We redirect the alloc/free
	* routines to the usual Free and Malloc in that environment.
	*/
	typedef int kmem_cache_t;
	#endif /* _STANDALONE */

	#endif /* _OPENSOLARIS_SYS_KMEM_H_ */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/kstat.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/kstat.h
	index 947dfee62393..7dc2c4753b02 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/kstat.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/kstat.h
	@@ -1,230 +1,230 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_KSTAT_H
	#define _SPL_KSTAT_H

	#include <sys/types.h>
	#ifndef _STANDALONE
	#include <sys/sysctl.h>
	#endif
	struct list_head {};
	#include <sys/mutex.h>
	#include <sys/proc.h>

	#define KSTAT_STRLEN 255
	#define KSTAT_RAW_MAX (128*1024)

	/*
	* For reference valid classes are:
	* disk, tape, net, controller, vm, kvm, hat, streams, kstat, misc
	*/

	#define KSTAT_TYPE_RAW 0 /* can be anything; ks_ndata >= 1 */
	#define KSTAT_TYPE_NAMED 1 /* name/value pair; ks_ndata >= 1 */
	#define KSTAT_TYPE_INTR 2 /* interrupt stats; ks_ndata == 1 */
	#define KSTAT_TYPE_IO 3 /* I/O stats; ks_ndata == 1 */
	#define KSTAT_TYPE_TIMER 4 /* event timer; ks_ndata >= 1 */
	#define KSTAT_NUM_TYPES 5

	#define KSTAT_DATA_CHAR 0
	#define KSTAT_DATA_INT32 1
	#define KSTAT_DATA_UINT32 2
	#define KSTAT_DATA_INT64 3
	#define KSTAT_DATA_UINT64 4
	#define KSTAT_DATA_LONG 5
	#define KSTAT_DATA_ULONG 6
	#define KSTAT_DATA_STRING 7
	#define KSTAT_NUM_DATAS 8

	#define KSTAT_INTR_HARD 0
	#define KSTAT_INTR_SOFT 1
	#define KSTAT_INTR_WATCHDOG 2
	#define KSTAT_INTR_SPURIOUS 3
	#define KSTAT_INTR_MULTSVC 4
	#define KSTAT_NUM_INTRS 5

	#define KSTAT_FLAG_VIRTUAL 0x01
	#define KSTAT_FLAG_VAR_SIZE 0x02
	#define KSTAT_FLAG_WRITABLE 0x04
	#define KSTAT_FLAG_PERSISTENT 0x08
	#define KSTAT_FLAG_DORMANT 0x10
	#define KSTAT_FLAG_INVALID 0x20
	#define KSTAT_FLAG_LONGSTRINGS 0x40
	#define KSTAT_FLAG_NO_HEADERS 0x80

	#define KS_MAGIC 0x9d9d9d9d

	/* Dynamic updates */
	#define KSTAT_READ 0
	#define KSTAT_WRITE 1

	struct kstat_s;
	typedef struct kstat_s kstat_t;

	typedef int kid_t; /* unique kstat id */
	typedef int kstat_update_t(struct kstat_s , int); / dynamic update cb */

	struct seq_file {
	char *sf_buf;
	size_t sf_size;
	};

	void seq_printf(struct seq_file m, const char fmt, ...);


	typedef struct kstat_module {
	- char ksm_name[KSTAT_STRLEN+1]; /* module name */
	+ char ksm_name[KSTAT_STRLEN]; /* module name */
	struct list_head ksm_module_list; /* module linkage */
	struct list_head ksm_kstat_list; /* list of kstat entries */
	struct proc_dir_entry ksm_proc; / proc entry */
	} kstat_module_t;

	typedef struct kstat_raw_ops {
	int (headers)(char buf, size_t size);
	int (seq_headers)(struct seq_file );
	int (data)(char buf, size_t size, void *data);
	void (addr)(kstat_t *ksp, loff_t index);
	} kstat_raw_ops_t;

	struct kstat_s {
	int ks_magic; /* magic value */
	kid_t ks_kid; /* unique kstat ID */
	hrtime_t ks_crtime; /* creation time */
	hrtime_t ks_snaptime; /* last access time */
	- char ks_module[KSTAT_STRLEN+1]; /* provider module name */
	+ char ks_module[KSTAT_STRLEN]; /* provider module name */
	int ks_instance; /* provider module instance */
	- char ks_name[KSTAT_STRLEN+1]; /* kstat name */
	- char ks_class[KSTAT_STRLEN+1]; /* kstat class */
	+ char ks_name[KSTAT_STRLEN]; /* kstat name */
	+ char ks_class[KSTAT_STRLEN]; /* kstat class */
	uchar_t ks_type; /* kstat data type */
	uchar_t ks_flags; /* kstat flags */
	void ks_data; / kstat type-specific data */
	uint_t ks_ndata; /* # of data records */
	size_t ks_data_size; /* size of kstat data section */
	kstat_update_t ks_update; / dynamic updates */
	void ks_private; / private data */
	void ks_private1; / private data */
	kmutex_t ks_private_lock; /* kstat private data lock */
	kmutex_t ks_lock; / kstat data lock */
	struct list_head ks_list; /* kstat linkage */
	kstat_module_t ks_owner; / kstat module linkage */
	kstat_raw_ops_t ks_raw_ops; /* ops table for raw type */
	char ks_raw_buf; / buf used for raw ops */
	size_t ks_raw_bufsize; /* size of raw ops buffer */
	#ifndef _STANDALONE
	struct sysctl_ctx_list ks_sysctl_ctx;
	struct sysctl_oid *ks_sysctl_root;
	#endif /* _STANDALONE */
	};

	typedef struct kstat_named_s {
	char name[KSTAT_STRLEN]; /* name of counter */
	uchar_t data_type; /* data type */
	union {
	char c[16]; /* 128-bit int */
	int32_t i32; /* 32-bit signed int */
	uint32_t ui32; /* 32-bit unsigned int */
	int64_t i64; /* 64-bit signed int */
	uint64_t ui64; /* 64-bit unsigned int */
	long l; /* native signed long */
	ulong_t ul; /* native unsigned long */
	struct {
	union {
	char ptr; / NULL-term string */
	char __pad[8]; /* 64-bit padding */
	} addr;
	uint32_t len; /* # bytes for strlen + '\0' */
	} string;
	} value;
	} kstat_named_t;

	#define KSTAT_NAMED_STR_PTR(knptr) ((knptr)->value.string.addr.ptr)
	#define KSTAT_NAMED_STR_BUFLEN(knptr) ((knptr)->value.string.len)

	typedef struct kstat_intr {
	uint_t intrs[KSTAT_NUM_INTRS];
	} kstat_intr_t;

	typedef struct kstat_io {
	u_longlong_t nread; /* number of bytes read */
	u_longlong_t nwritten; /* number of bytes written */
	uint_t reads; /* number of read operations */
	uint_t writes; /* number of write operations */
	hrtime_t wtime; /* cumulative wait (pre-service) time */
	hrtime_t wlentime; /* cumulative wait lentime product /
	hrtime_t wlastupdate; /* last time wait queue changed */
	hrtime_t rtime; /* cumulative run (service) time */
	hrtime_t rlentime; /* cumulative run lengthtime product /
	hrtime_t rlastupdate; /* last time run queue changed */
	uint_t wcnt; /* count of elements in wait state */
	uint_t rcnt; /* count of elements in run state */
	} kstat_io_t;

	typedef struct kstat_timer {
	- char name[KSTAT_STRLEN+1]; /* event name */
	+ char name[KSTAT_STRLEN]; /* event name */
	u_longlong_t num_events; /* number of events */
	hrtime_t elapsed_time; /* cumulative elapsed time */
	hrtime_t min_time; /* shortest event duration */
	hrtime_t max_time; /* longest event duration */
	hrtime_t start_time; /* previous event start time */
	hrtime_t stop_time; /* previous event stop time */
	} kstat_timer_t;

	int spl_kstat_init(void);
	void spl_kstat_fini(void);

	extern void __kstat_set_raw_ops(kstat_t *ksp,
	int (headers)(char buf, size_t size),
	int (data)(char buf, size_t size, void *data),
	void* (addr)(kstat_t ksp, loff_t index));

	extern void __kstat_set_seq_raw_ops(kstat_t *ksp,
	int (headers)(struct seq_file ),
	int (data)(char buf, size_t size, void *data),
	void* (addr)(kstat_t ksp, loff_t index));


	extern kstat_t __kstat_create(const char ks_module, int ks_instance,
	const char ks_name, const char ks_class, uchar_t ks_type,
	uint_t ks_ndata, uchar_t ks_flags);

	extern void __kstat_install(kstat_t *ksp);
	extern void __kstat_delete(kstat_t *ksp);

	#define kstat_set_seq_raw_ops(k, h, d, a) \
	__kstat_set_seq_raw_ops(k, h, d, a)
	#define kstat_set_raw_ops(k, h, d, a) \
	__kstat_set_raw_ops(k, h, d, a)
	#ifndef _STANDALONE
	#define kstat_create(m, i, n, c, t, s, f) \
	__kstat_create(m, i, n, c, t, s, f)

	#define kstat_install(k) __kstat_install(k)
	#define kstat_delete(k) __kstat_delete(k)
	#else
	#define kstat_create(m, i, n, c, t, s, f) ((kstat_t *)0)
	#define kstat_install(k)
	#define kstat_delete(k)
	#endif

	#endif /* _SPL_KSTAT_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/simd_powerpc.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/simd_powerpc.h
	index f101e3192fac..34d5e23e2fbb 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/simd_powerpc.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/simd_powerpc.h
	@@ -1,91 +1,92 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (C) 2022 Tino Reichardt <milky-zfs@mcmilk.de>
	*/

	/*
	* USER API:
	*
	* Kernel fpu methods:
	* kfpu_allowed()
	* kfpu_begin()
	* kfpu_end()
	* kfpu_init()
	* kfpu_fini()
	*
	* SIMD support:
	*
	* Following functions should be called to determine whether CPU feature
	* is supported. All functions are usable in kernel and user space.
	* If a SIMD algorithm is using more than one instruction set
	* all relevant feature test functions should be called.
	*
	* Supported features:
	* zfs_altivec_available()
	* zfs_vsx_available()
	* zfs_isa207_available()
	*/

	#ifndef _FREEBSD_SIMD_POWERPC_H
	#define _FREEBSD_SIMD_POWERPC_H

	#include <sys/types.h>
	#include <sys/cdefs.h>

	#include <machine/pcb.h>
	#include <machine/cpu.h>

	#define kfpu_allowed() 1
	#define kfpu_initialize(tsk) do {} while (0)
	#define kfpu_begin() do {} while (0)
	#define kfpu_end() do {} while (0)
	#define kfpu_init() (0)
	#define kfpu_fini() do {} while (0)

	/*
	* Check if Altivec is available
	*/
	static inline boolean_t
	zfs_altivec_available(void)
	{
	return ((cpu_features & PPC_FEATURE_HAS_ALTIVEC) != 0);
	}

	/*
	* Check if VSX is available
	*/
	static inline boolean_t
	zfs_vsx_available(void)
	{
	return ((cpu_features & PPC_FEATURE_HAS_VSX) != 0);
	}

	/*
	* Check if POWER ISA 2.07 is available (SHA2)
	*/
	static inline boolean_t
	zfs_isa207_available(void)
	{
	return ((cpu_features2 & PPC_FEATURE2_ARCH_2_07) != 0);
	}
	+
	#endif
	diff --git a/sys/contrib/openzfs/include/os/linux/Makefile.am b/sys/contrib/openzfs/include/os/linux/Makefile.am
	index 4d6901c694c8..13ba8060c62d 100644
	--- a/sys/contrib/openzfs/include/os/linux/Makefile.am
	+++ b/sys/contrib/openzfs/include/os/linux/Makefile.am
	@@ -1,111 +1,112 @@
	if CONFIG_KERNEL
	kernel_linuxdir = $(kerneldir)/linux
	kernel_linux_HEADERS = \
	%D%/kernel/linux/blkdev_compat.h \
	%D%/kernel/linux/compiler_compat.h \
	%D%/kernel/linux/dcache_compat.h \
	%D%/kernel/linux/kmap_compat.h \
	%D%/kernel/linux/mod_compat.h \
	%D%/kernel/linux/page_compat.h \
	%D%/kernel/linux/percpu_compat.h \
	%D%/kernel/linux/simd.h \
	%D%/kernel/linux/simd_aarch64.h \
	%D%/kernel/linux/simd_powerpc.h \
	%D%/kernel/linux/simd_x86.h \
	%D%/kernel/linux/utsname_compat.h \
	%D%/kernel/linux/vfs_compat.h \
	%D%/kernel/linux/xattr_compat.h

	kernel_sysdir = $(kerneldir)/sys
	kernel_sys_HEADERS = \
	%D%/zfs/sys/policy.h \
	%D%/zfs/sys/sha2.h \
	%D%/zfs/sys/trace_acl.h \
	%D%/zfs/sys/trace_arc.h \
	%D%/zfs/sys/trace_common.h \
	%D%/zfs/sys/trace_dbgmsg.h \
	%D%/zfs/sys/trace_dbuf.h \
	%D%/zfs/sys/trace_dmu.h \
	%D%/zfs/sys/trace_dnode.h \
	%D%/zfs/sys/trace_multilist.h \
	%D%/zfs/sys/trace_rrwlock.h \
	%D%/zfs/sys/trace_txg.h \
	%D%/zfs/sys/trace_vdev.h \
	%D%/zfs/sys/trace_zfs.h \
	%D%/zfs/sys/trace_zil.h \
	%D%/zfs/sys/trace_zio.h \
	%D%/zfs/sys/trace_zrlock.h \
	%D%/zfs/sys/zfs_bootenv_os.h \
	%D%/zfs/sys/zfs_context_os.h \
	%D%/zfs/sys/zfs_ctldir.h \
	%D%/zfs/sys/zfs_dir.h \
	%D%/zfs/sys/zfs_vfsops_os.h \
	%D%/zfs/sys/zfs_vnops_os.h \
	%D%/zfs/sys/zfs_znode_impl.h \
	%D%/zfs/sys/zpl.h

	kernel_spl_rpcdir = $(kerneldir)/spl/rpc
	kernel_spl_rpc_HEADERS = \
	%D%/spl/rpc/xdr.h

	kernel_spl_sysdir = $(kerneldir)/spl/sys
	kernel_spl_sys_HEADERS = \
	%D%/spl/sys/acl.h \
	%D%/spl/sys/atomic.h \
	%D%/spl/sys/byteorder.h \
	%D%/spl/sys/callb.h \
	%D%/spl/sys/callo.h \
	%D%/spl/sys/cmn_err.h \
	%D%/spl/sys/condvar.h \
	%D%/spl/sys/cred.h \
	%D%/spl/sys/ctype.h \
	%D%/spl/sys/debug.h \
	%D%/spl/sys/disp.h \
	%D%/spl/sys/dkio.h \
	%D%/spl/sys/errno.h \
	%D%/spl/sys/fcntl.h \
	%D%/spl/sys/file.h \
	%D%/spl/sys/inttypes.h \
	%D%/spl/sys/isa_defs.h \
	%D%/spl/sys/kmem.h \
	%D%/spl/sys/kmem_cache.h \
	%D%/spl/sys/kstat.h \
	%D%/spl/sys/list.h \
	+ %D%/spl/sys/misc.h \
	%D%/spl/sys/mod_os.h \
	%D%/spl/sys/mutex.h \
	%D%/spl/sys/param.h \
	%D%/spl/sys/proc.h \
	%D%/spl/sys/processor.h \
	%D%/spl/sys/procfs_list.h \
	%D%/spl/sys/random.h \
	%D%/spl/sys/rwlock.h \
	%D%/spl/sys/shrinker.h \
	%D%/spl/sys/sid.h \
	%D%/spl/sys/signal.h \
	%D%/spl/sys/simd.h \
	%D%/spl/sys/stat.h \
	%D%/spl/sys/string.h \
	%D%/spl/sys/sunddi.h \
	%D%/spl/sys/sysmacros.h \
	%D%/spl/sys/systeminfo.h \
	%D%/spl/sys/taskq.h \
	%D%/spl/sys/thread.h \
	%D%/spl/sys/time.h \
	%D%/spl/sys/timer.h \
	%D%/spl/sys/trace.h \
	%D%/spl/sys/trace_spl.h \
	%D%/spl/sys/trace_taskq.h \
	%D%/spl/sys/tsd.h \
	%D%/spl/sys/types.h \
	%D%/spl/sys/types32.h \
	%D%/spl/sys/uio.h \
	%D%/spl/sys/user.h \
	%D%/spl/sys/vfs.h \
	%D%/spl/sys/vmem.h \
	%D%/spl/sys/vmsystm.h \
	%D%/spl/sys/vnode.h \
	%D%/spl/sys/wait.h \
	%D%/spl/sys/wmsum.h \
	%D%/spl/sys/zmod.h \
	%D%/spl/sys/zone.h
	endif
	diff --git a/sys/contrib/openzfs/include/os/linux/kernel/linux/blkdev_compat.h b/sys/contrib/openzfs/include/os/linux/kernel/linux/blkdev_compat.h
	index cdcea166903d..3276796537a4 100644
	--- a/sys/contrib/openzfs/include/os/linux/kernel/linux/blkdev_compat.h
	+++ b/sys/contrib/openzfs/include/os/linux/kernel/linux/blkdev_compat.h
	@@ -1,722 +1,748 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (C) 2011 Lawrence Livermore National Security, LLC.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* LLNL-CODE-403049.
	*/

	#ifndef _ZFS_BLKDEV_H
	#define _ZFS_BLKDEV_H

	#include <linux/blkdev.h>
	#include <linux/backing-dev.h>
	#include <linux/hdreg.h>
	#include <linux/major.h>
	#include <linux/msdos_fs.h> /* for SECTOR_* */
	#include <linux/bio.h>

	#ifdef HAVE_BLK_MQ
	#include <linux/blk-mq.h>
	#endif

	#ifndef HAVE_BLK_QUEUE_FLAG_SET
	static inline void
	blk_queue_flag_set(unsigned int flag, struct request_queue *q)
	{
	queue_flag_set(flag, q);
	}
	#endif

	#ifndef HAVE_BLK_QUEUE_FLAG_CLEAR
	static inline void
	blk_queue_flag_clear(unsigned int flag, struct request_queue *q)
	{
	queue_flag_clear(flag, q);
	}
	#endif

	/*
	* 4.7 API,
	* The blk_queue_write_cache() interface has replaced blk_queue_flush()
	* interface. However, the new interface is GPL-only thus we implement
	* our own trivial wrapper when the GPL-only version is detected.
	*
	* 2.6.36 - 4.6 API,
	* The blk_queue_flush() interface has replaced blk_queue_ordered()
	* interface. However, while the old interface was available to all the
	* new one is GPL-only. Thus if the GPL-only version is detected we
	* implement our own trivial helper.
	*/
	static inline void
	blk_queue_set_write_cache(struct request_queue *q, bool wc, bool fua)
	{
	#if defined(HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY)
	if (wc)
	blk_queue_flag_set(QUEUE_FLAG_WC, q);
	else
	blk_queue_flag_clear(QUEUE_FLAG_WC, q);
	if (fua)
	blk_queue_flag_set(QUEUE_FLAG_FUA, q);
	else
	blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
	#elif defined(HAVE_BLK_QUEUE_WRITE_CACHE)
	blk_queue_write_cache(q, wc, fua);
	#elif defined(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY)
	if (wc)
	q->flush_flags \|= REQ_FLUSH;
	if (fua)
	q->flush_flags \|= REQ_FUA;
	#elif defined(HAVE_BLK_QUEUE_FLUSH)
	blk_queue_flush(q, (wc ? REQ_FLUSH : 0) \| (fua ? REQ_FUA : 0));
	#else
	#error "Unsupported kernel"
	#endif
	}

	static inline void
	blk_queue_set_read_ahead(struct request_queue *q, unsigned long ra_pages)
	{
	#if !defined(HAVE_BLK_QUEUE_UPDATE_READAHEAD) && \
	!defined(HAVE_DISK_UPDATE_READAHEAD)
	#ifdef HAVE_BLK_QUEUE_BDI_DYNAMIC
	q->backing_dev_info->ra_pages = ra_pages;
	#else
	q->backing_dev_info.ra_pages = ra_pages;
	#endif
	#endif
	}

	#ifdef HAVE_BIO_BVEC_ITER
	#define BIO_BI_SECTOR(bio) (bio)->bi_iter.bi_sector
	#define BIO_BI_SIZE(bio) (bio)->bi_iter.bi_size
	#define BIO_BI_IDX(bio) (bio)->bi_iter.bi_idx
	#define BIO_BI_SKIP(bio) (bio)->bi_iter.bi_bvec_done
	#define bio_for_each_segment4(bv, bvp, b, i) \
	bio_for_each_segment((bv), (b), (i))
	typedef struct bvec_iter bvec_iterator_t;
	#else
	#define BIO_BI_SECTOR(bio) (bio)->bi_sector
	#define BIO_BI_SIZE(bio) (bio)->bi_size
	#define BIO_BI_IDX(bio) (bio)->bi_idx
	#define BIO_BI_SKIP(bio) (0)
	#define bio_for_each_segment4(bv, bvp, b, i) \
	bio_for_each_segment((bvp), (b), (i))
	typedef int bvec_iterator_t;
	#endif

	static inline void
	bio_set_flags_failfast(struct block_device bdev, int flags)
	{
	#ifdef CONFIG_BUG
	/*
	* Disable FAILFAST for loopback devices because of the
	* following incorrect BUG_ON() in loop_make_request().
	* This support is also disabled for md devices because the
	* test suite layers md devices on top of loopback devices.
	* This may be removed when the loopback driver is fixed.
	*
	* BUG_ON(!lo \|\| (rw != READ && rw != WRITE));
	*/
	if ((MAJOR(bdev->bd_dev) == LOOP_MAJOR) \|\|
	(MAJOR(bdev->bd_dev) == MD_MAJOR))
	return;

	#ifdef BLOCK_EXT_MAJOR
	if (MAJOR(bdev->bd_dev) == BLOCK_EXT_MAJOR)
	return;
	#endif /* BLOCK_EXT_MAJOR */
	#endif /* CONFIG_BUG */

	*flags \|= REQ_FAILFAST_MASK;
	}

	/*
	* Maximum disk label length, it may be undefined for some kernels.
	*/
	#if !defined(DISK_NAME_LEN)
	#define DISK_NAME_LEN 32
	#endif /* DISK_NAME_LEN */

	#ifdef HAVE_BIO_BI_STATUS
	static inline int
	bi_status_to_errno(blk_status_t status)
	{
	switch (status) {
	case BLK_STS_OK:
	return (0);
	case BLK_STS_NOTSUPP:
	return (EOPNOTSUPP);
	case BLK_STS_TIMEOUT:
	return (ETIMEDOUT);
	case BLK_STS_NOSPC:
	return (ENOSPC);
	case BLK_STS_TRANSPORT:
	return (ENOLINK);
	case BLK_STS_TARGET:
	return (EREMOTEIO);
	case BLK_STS_NEXUS:
	return (EBADE);
	case BLK_STS_MEDIUM:
	return (ENODATA);
	case BLK_STS_PROTECTION:
	return (EILSEQ);
	case BLK_STS_RESOURCE:
	return (ENOMEM);
	case BLK_STS_AGAIN:
	return (EAGAIN);
	case BLK_STS_IOERR:
	return (EIO);
	default:
	return (EIO);
	}
	}

	static inline blk_status_t
	errno_to_bi_status(int error)
	{
	switch (error) {
	case 0:
	return (BLK_STS_OK);
	case EOPNOTSUPP:
	return (BLK_STS_NOTSUPP);
	case ETIMEDOUT:
	return (BLK_STS_TIMEOUT);
	case ENOSPC:
	return (BLK_STS_NOSPC);
	case ENOLINK:
	return (BLK_STS_TRANSPORT);
	case EREMOTEIO:
	return (BLK_STS_TARGET);
	case EBADE:
	return (BLK_STS_NEXUS);
	case ENODATA:
	return (BLK_STS_MEDIUM);
	case EILSEQ:
	return (BLK_STS_PROTECTION);
	case ENOMEM:
	return (BLK_STS_RESOURCE);
	case EAGAIN:
	return (BLK_STS_AGAIN);
	case EIO:
	return (BLK_STS_IOERR);
	default:
	return (BLK_STS_IOERR);
	}
	}
	#endif /* HAVE_BIO_BI_STATUS */

	/*
	* 4.3 API change
	* The bio_endio() prototype changed slightly. These are helper
	* macro's to ensure the prototype and invocation are handled.
	*/
	#ifdef HAVE_1ARG_BIO_END_IO_T
	#ifdef HAVE_BIO_BI_STATUS
	#define BIO_END_IO_ERROR(bio) bi_status_to_errno(bio->bi_status)
	#define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x)
	#define BIO_END_IO(bio, error) bio_set_bi_status(bio, error)
	static inline void
	bio_set_bi_status(struct bio *bio, int error)
	{
	ASSERT3S(error, <=, 0);
	bio->bi_status = errno_to_bi_status(-error);
	bio_endio(bio);
	}
	#else
	#define BIO_END_IO_ERROR(bio) (-(bio->bi_error))
	#define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x)
	#define BIO_END_IO(bio, error) bio_set_bi_error(bio, error)
	static inline void
	bio_set_bi_error(struct bio *bio, int error)
	{
	ASSERT3S(error, <=, 0);
	bio->bi_error = error;
	bio_endio(bio);
	}
	#endif /* HAVE_BIO_BI_STATUS */

	#else
	#define BIO_END_IO_PROTO(fn, x, z) static void fn(struct bio *x, int z)
	#define BIO_END_IO(bio, error) bio_endio(bio, error);
	#endif /* HAVE_1ARG_BIO_END_IO_T */

	+/*
	+ * 5.15 MACRO,
	+ * GD_DEAD
	+ *
	+ * 2.6.36 - 5.14 MACRO,
	+ * GENHD_FL_UP
	+ *
	+ * Check the disk status and return B_TRUE if alive
	+ * otherwise B_FALSE
	+ */
	+static inline boolean_t
	+zfs_check_disk_status(struct block_device *bdev)
	+{
	+#if defined(GENHD_FL_UP)
	+ return (!!(bdev->bd_disk->flags & GENHD_FL_UP));
	+#elif defined(GD_DEAD)
	+ return (!test_bit(GD_DEAD, &bdev->bd_disk->state));
	+#else
	+/*
	+ * This is encountered if neither GENHD_FL_UP nor GD_DEAD is available in
	+ * the kernel - likely due to an MACRO change that needs to be chased down.
	+ */
	+#error "Unsupported kernel: no usable disk status check"
	+#endif
	+}
	+
	/*
	* 4.1 API,
	* 3.10.0 CentOS 7.x API,
	* blkdev_reread_part()
	*
	* For older kernels trigger a re-reading of the partition table by calling
	* check_disk_change() which calls flush_disk() to invalidate the device.
	*
	* For newer kernels (as of 5.10), bdev_check_media_change is used, in favor of
	* check_disk_change(), with the modification that invalidation is no longer
	* forced.
	*/
	#ifdef HAVE_CHECK_DISK_CHANGE
	#define zfs_check_media_change(bdev) check_disk_change(bdev)
	#ifdef HAVE_BLKDEV_REREAD_PART
	#define vdev_bdev_reread_part(bdev) blkdev_reread_part(bdev)
	#else
	#define vdev_bdev_reread_part(bdev) check_disk_change(bdev)
	#endif /* HAVE_BLKDEV_REREAD_PART */
	#else
	#ifdef HAVE_BDEV_CHECK_MEDIA_CHANGE
	static inline int
	zfs_check_media_change(struct block_device *bdev)
	{
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	struct gendisk *gd = bdev->bd_disk;
	const struct block_device_operations *bdo = gd->fops;
	#endif

	if (!bdev_check_media_change(bdev))
	return (0);

	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	/*
	* Force revalidation, to mimic the old behavior of
	* check_disk_change()
	*/
	if (bdo->revalidate_disk)
	bdo->revalidate_disk(gd);
	#endif

	return (0);
	}
	#define vdev_bdev_reread_part(bdev) zfs_check_media_change(bdev)
	#else
	/*
	* This is encountered if check_disk_change() and bdev_check_media_change()
	* are not available in the kernel - likely due to an API change that needs
	* to be chased down.
	*/
	#error "Unsupported kernel: no usable disk change check"
	#endif /* HAVE_BDEV_CHECK_MEDIA_CHANGE */
	#endif /* HAVE_CHECK_DISK_CHANGE */

	/*
	* 2.6.27 API change
	* The function was exported for use, prior to this it existed but the
	* symbol was not exported.
	*
	* 4.4.0-6.21 API change for Ubuntu
	* lookup_bdev() gained a second argument, FMODE_*, to check inode permissions.
	*
	* 5.11 API change
	* Changed to take a dev_t argument which is set on success and return a
	* non-zero error code on failure.
	*/
	static inline int
	vdev_lookup_bdev(const char path, dev_t dev)
	{
	#if defined(HAVE_DEVT_LOOKUP_BDEV)
	return (lookup_bdev(path, dev));
	#elif defined(HAVE_1ARG_LOOKUP_BDEV)
	struct block_device *bdev = lookup_bdev(path);
	if (IS_ERR(bdev))
	return (PTR_ERR(bdev));

	*dev = bdev->bd_dev;
	bdput(bdev);

	return (0);
	#elif defined(HAVE_MODE_LOOKUP_BDEV)
	struct block_device *bdev = lookup_bdev(path, FMODE_READ);
	if (IS_ERR(bdev))
	return (PTR_ERR(bdev));

	*dev = bdev->bd_dev;
	bdput(bdev);

	return (0);
	#else
	#error "Unsupported kernel"
	#endif
	}

	/*
	* Kernels without bio_set_op_attrs use bi_rw for the bio flags.
	*/
	#if !defined(HAVE_BIO_SET_OP_ATTRS)
	static inline void
	bio_set_op_attrs(struct bio *bio, unsigned rw, unsigned flags)
	{
	bio->bi_rw \|= rw \| flags;
	}
	#endif

	/*
	* bio_set_flush - Set the appropriate flags in a bio to guarantee
	* data are on non-volatile media on completion.
	*
	* 2.6.37 - 4.8 API,
	* Introduce WRITE_FLUSH, WRITE_FUA, and WRITE_FLUSH_FUA flags as a
	* replacement for WRITE_BARRIER to allow expressing richer semantics
	* to the block layer. It's up to the block layer to implement the
	* semantics correctly. Use the WRITE_FLUSH_FUA flag combination.
	*
	* 4.8 - 4.9 API,
	* REQ_FLUSH was renamed to REQ_PREFLUSH. For consistency with previous
	* OpenZFS releases, prefer the WRITE_FLUSH_FUA flag set if it's available.
	*
	* 4.10 API,
	* The read/write flags and their modifiers, including WRITE_FLUSH,
	* WRITE_FUA and WRITE_FLUSH_FUA were removed from fs.h in
	* torvalds/linux@70fd7614 and replaced by direct flag modification
	* of the REQ_ flags in bio->bi_opf. Use REQ_PREFLUSH.
	*/
	static inline void
	bio_set_flush(struct bio *bio)
	{
	#if defined(HAVE_REQ_PREFLUSH) /* >= 4.10 */
	bio_set_op_attrs(bio, 0, REQ_PREFLUSH);
	#elif defined(WRITE_FLUSH_FUA) /* >= 2.6.37 and <= 4.9 */
	bio_set_op_attrs(bio, 0, WRITE_FLUSH_FUA);
	#else
	#error "Allowing the build will cause bio_set_flush requests to be ignored."
	#endif
	}

	/*
	* 4.8 API,
	* REQ_OP_FLUSH
	*
	* 4.8-rc0 - 4.8-rc1,
	* REQ_PREFLUSH
	*
	* 2.6.36 - 4.7 API,
	* REQ_FLUSH
	*
	* in all cases but may have a performance impact for some kernels. It
	* has the advantage of minimizing kernel specific changes in the zvol code.
	*
	*/
	static inline boolean_t
	bio_is_flush(struct bio *bio)
	{
	#if defined(HAVE_REQ_OP_FLUSH) && defined(HAVE_BIO_BI_OPF)
	return ((bio_op(bio) == REQ_OP_FLUSH) \|\| (bio->bi_opf & REQ_PREFLUSH));
	#elif defined(HAVE_REQ_PREFLUSH) && defined(HAVE_BIO_BI_OPF)
	return (bio->bi_opf & REQ_PREFLUSH);
	#elif defined(HAVE_REQ_PREFLUSH) && !defined(HAVE_BIO_BI_OPF)
	return (bio->bi_rw & REQ_PREFLUSH);
	#elif defined(HAVE_REQ_FLUSH)
	return (bio->bi_rw & REQ_FLUSH);
	#else
	#error "Unsupported kernel"
	#endif
	}

	/*
	* 4.8 API,
	* REQ_FUA flag moved to bio->bi_opf
	*
	* 2.6.x - 4.7 API,
	* REQ_FUA
	*/
	static inline boolean_t
	bio_is_fua(struct bio *bio)
	{
	#if defined(HAVE_BIO_BI_OPF)
	return (bio->bi_opf & REQ_FUA);
	#elif defined(REQ_FUA)
	return (bio->bi_rw & REQ_FUA);
	#else
	#error "Allowing the build will cause fua requests to be ignored."
	#endif
	}

	/*
	* 4.8 API,
	* REQ_OP_DISCARD
	*
	* 2.6.36 - 4.7 API,
	* REQ_DISCARD
	*
	* In all cases the normal I/O path is used for discards. The only
	* difference is how the kernel tags individual I/Os as discards.
	*/
	static inline boolean_t
	bio_is_discard(struct bio *bio)
	{
	#if defined(HAVE_REQ_OP_DISCARD)
	return (bio_op(bio) == REQ_OP_DISCARD);
	#elif defined(HAVE_REQ_DISCARD)
	return (bio->bi_rw & REQ_DISCARD);
	#else
	#error "Unsupported kernel"
	#endif
	}

	/*
	* 4.8 API,
	* REQ_OP_SECURE_ERASE
	*
	* 2.6.36 - 4.7 API,
	* REQ_SECURE
	*/
	static inline boolean_t
	bio_is_secure_erase(struct bio *bio)
	{
	#if defined(HAVE_REQ_OP_SECURE_ERASE)
	return (bio_op(bio) == REQ_OP_SECURE_ERASE);
	#elif defined(REQ_SECURE)
	return (bio->bi_rw & REQ_SECURE);
	#else
	return (0);
	#endif
	}

	/*
	* 2.6.33 API change
	* Discard granularity and alignment restrictions may now be set. For
	* older kernels which do not support this it is safe to skip it.
	*/
	static inline void
	blk_queue_discard_granularity(struct request_queue *q, unsigned int dg)
	{
	q->limits.discard_granularity = dg;
	}

	/*
	* 5.19 API,
	* bdev_max_discard_sectors()
	*
	* 2.6.32 API,
	* blk_queue_discard()
	*/
	static inline boolean_t
	bdev_discard_supported(struct block_device *bdev)
	{
	#if defined(HAVE_BDEV_MAX_DISCARD_SECTORS)
	return (!!bdev_max_discard_sectors(bdev));
	#elif defined(HAVE_BLK_QUEUE_DISCARD)
	return (!!blk_queue_discard(bdev_get_queue(bdev)));
	#else
	#error "Unsupported kernel"
	#endif
	}

	/*
	* 5.19 API,
	* bdev_max_secure_erase_sectors()
	*
	* 4.8 API,
	* blk_queue_secure_erase()
	*
	* 2.6.36 - 4.7 API,
	* blk_queue_secdiscard()
	*/
	static inline boolean_t
	bdev_secure_discard_supported(struct block_device *bdev)
	{
	#if defined(HAVE_BDEV_MAX_SECURE_ERASE_SECTORS)
	return (!!bdev_max_secure_erase_sectors(bdev));
	#elif defined(HAVE_BLK_QUEUE_SECURE_ERASE)
	return (!!blk_queue_secure_erase(bdev_get_queue(bdev)));
	#elif defined(HAVE_BLK_QUEUE_SECDISCARD)
	return (!!blk_queue_secdiscard(bdev_get_queue(bdev)));
	#else
	#error "Unsupported kernel"
	#endif
	}

	/*
	* A common holder for vdev_bdev_open() is used to relax the exclusive open
	* semantics slightly. Internal vdev disk callers may pass VDEV_HOLDER to
	* allow them to open the device multiple times. Other kernel callers and
	* user space processes which don't pass this value will get EBUSY. This is
	* currently required for the correct operation of hot spares.
	*/
	#define VDEV_HOLDER ((void *)0x2401de7)

	static inline unsigned long
	blk_generic_start_io_acct(struct request_queue *q __attribute__((unused)),
	struct gendisk *disk __attribute__((unused)),
	int rw __attribute__((unused)), struct bio *bio)
	{
	#if defined(HAVE_BDEV_IO_ACCT)
	return (bdev_start_io_acct(bio->bi_bdev, bio_sectors(bio),
	bio_op(bio), jiffies));
	#elif defined(HAVE_DISK_IO_ACCT)
	return (disk_start_io_acct(disk, bio_sectors(bio), bio_op(bio)));
	#elif defined(HAVE_BIO_IO_ACCT)
	return (bio_start_io_acct(bio));
	#elif defined(HAVE_GENERIC_IO_ACCT_3ARG)
	unsigned long start_time = jiffies;
	generic_start_io_acct(rw, bio_sectors(bio), &disk->part0);
	return (start_time);
	#elif defined(HAVE_GENERIC_IO_ACCT_4ARG)
	unsigned long start_time = jiffies;
	generic_start_io_acct(q, rw, bio_sectors(bio), &disk->part0);
	return (start_time);
	#else
	/* Unsupported */
	return (0);
	#endif
	}

	static inline void
	blk_generic_end_io_acct(struct request_queue *q __attribute__((unused)),
	struct gendisk *disk __attribute__((unused)),
	int rw __attribute__((unused)), struct bio *bio, unsigned long start_time)
	{
	#if defined(HAVE_BDEV_IO_ACCT)
	bdev_end_io_acct(bio->bi_bdev, bio_op(bio), start_time);
	#elif defined(HAVE_DISK_IO_ACCT)
	disk_end_io_acct(disk, bio_op(bio), start_time);
	#elif defined(HAVE_BIO_IO_ACCT)
	bio_end_io_acct(bio, start_time);
	#elif defined(HAVE_GENERIC_IO_ACCT_3ARG)
	generic_end_io_acct(rw, &disk->part0, start_time);
	#elif defined(HAVE_GENERIC_IO_ACCT_4ARG)
	generic_end_io_acct(q, rw, &disk->part0, start_time);
	#endif
	}

	#ifndef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	static inline struct request_queue *
	blk_generic_alloc_queue(make_request_fn make_request, int node_id)
	{
	#if defined(HAVE_BLK_ALLOC_QUEUE_REQUEST_FN)
	return (blk_alloc_queue(make_request, node_id));
	#elif defined(HAVE_BLK_ALLOC_QUEUE_REQUEST_FN_RH)
	return (blk_alloc_queue_rh(make_request, node_id));
	#else
	struct request_queue *q = blk_alloc_queue(GFP_KERNEL);
	if (q != NULL)
	blk_queue_make_request(q, make_request);

	return (q);
	#endif
	}
	#endif /* !HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */

	/*
	* All the io_*() helper functions below can operate on a bio, or a rq, but
	* not both. The older submit_bio() codepath will pass a bio, and the
	* newer blk-mq codepath will pass a rq.
	*/
	static inline int
	io_data_dir(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL) {
	if (op_is_write(req_op(rq))) {
	return (WRITE);
	} else {
	return (READ);
	}
	}
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_data_dir(bio));
	}

	static inline int
	io_is_flush(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (req_op(rq) == REQ_OP_FLUSH);
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_is_flush(bio));
	}

	static inline int
	io_is_discard(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (req_op(rq) == REQ_OP_DISCARD);
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_is_discard(bio));
	}

	static inline int
	io_is_secure_erase(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (req_op(rq) == REQ_OP_SECURE_ERASE);
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_is_secure_erase(bio));
	}

	static inline int
	io_is_fua(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (rq->cmd_flags & REQ_FUA);
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_is_fua(bio));
	}


	static inline uint64_t
	io_offset(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (blk_rq_pos(rq) << 9);
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (BIO_BI_SECTOR(bio) << 9);
	}

	static inline uint64_t
	io_size(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (blk_rq_bytes(rq));
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (BIO_BI_SIZE(bio));
	}

	static inline int
	io_has_data(struct bio bio, struct request rq)
	{
	#ifdef HAVE_BLK_MQ
	if (rq != NULL)
	return (bio_has_data(rq->bio));
	#else
	ASSERT3P(rq, ==, NULL);
	#endif
	return (bio_has_data(bio));
	}
	#endif /* _ZFS_BLKDEV_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/byteorder.h b/sys/contrib/openzfs/include/os/linux/spl/sys/byteorder.h
	index bb5e173ce5e4..e9406b4c0b86 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/byteorder.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/byteorder.h
	@@ -1,86 +1,101 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_BYTEORDER_H
	#define _SPL_BYTEORDER_H

	#include <asm/byteorder.h>

	#if defined(__BIG_ENDIAN) && !defined(_ZFS_BIG_ENDIAN)
	#define _ZFS_BIG_ENDIAN
	#endif

	#if defined(__LITTLE_ENDIAN) && !defined(_ZFS_LITTLE_ENDIAN)
	#define _ZFS_LITTLE_ENDIAN
	#endif

	#include <sys/isa_defs.h>

	+#ifdef __COVERITY__
	+/*
	+ * Coverity's taint warnings from byteswapping are false positives for us.
	+ * Suppress them by hiding byteswapping from Coverity.
	+ */
	+
	+#define BSWAP_8(x) ((x) & 0xff)
	+#define BSWAP_16(x) ((x) & 0xffff)
	+#define BSWAP_32(x) ((x) & 0xffffffff)
	+#define BSWAP_64(x) (x)
	+
	+#else /* __COVERITY__ */
	+
	#define BSWAP_8(x) ((x) & 0xff)
	#define BSWAP_16(x) ((BSWAP_8(x) << 8) \| BSWAP_8((x) >> 8))
	#define BSWAP_32(x) ((BSWAP_16(x) << 16) \| BSWAP_16((x) >> 16))
	#define BSWAP_64(x) ((BSWAP_32(x) << 32) \| BSWAP_32((x) >> 32))

	+#endif /* __COVERITY__ */
	+
	#define LE_16(x) cpu_to_le16(x)
	#define LE_32(x) cpu_to_le32(x)
	#define LE_64(x) cpu_to_le64(x)
	#define BE_16(x) cpu_to_be16(x)
	#define BE_32(x) cpu_to_be32(x)
	#define BE_64(x) cpu_to_be64(x)

	#define BE_IN8(xa) \
	((uint8_t )(xa))

	#define BE_IN16(xa) \
	(((uint16_t)BE_IN8(xa) << 8) \| BE_IN8((uint8_t *)(xa)+1))

	#define BE_IN32(xa) \
	(((uint32_t)BE_IN16(xa) << 16) \| BE_IN16((uint8_t *)(xa)+2))

	#ifdef _ZFS_BIG_ENDIAN
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return (n);
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return (n);
	}
	#else
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return ((((uint64_t)htonl(n)) << 32) + htonl(n >> 32));
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return ((((uint64_t)ntohl(n)) << 32) + ntohl(n >> 32));
	}
	#endif

	#endif /* SPL_BYTEORDER_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/cmn_err.h b/sys/contrib/openzfs/include/os/linux/spl/sys/cmn_err.h
	index d2088371c6bc..161bcf9b3a46 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/cmn_err.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/cmn_err.h
	@@ -1,48 +1,48 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_CMN_ERR_H
	#define _SPL_CMN_ERR_H

	#if defined(_KERNEL) && defined(HAVE_STANDALONE_LINUX_STDARG)
	#include <linux/stdarg.h>
	#else
	#include <stdarg.h>
	#endif

	#define CE_CONT 0 /* continuation */
	#define CE_NOTE 1 /* notice */
	#define CE_WARN 2 /* warning */
	#define CE_PANIC 3 /* panic */
	#define CE_IGNORE 4 /* print nothing */

	extern void cmn_err(int, const char *, ...)
	__attribute__((format(printf, 2, 3)));
	extern void vcmn_err(int, const char *, va_list)
	__attribute__((format(printf, 2, 0)));
	extern void vpanic(const char *, va_list)
	- __attribute__((format(printf, 1, 0)));
	+ __attribute__((format(printf, 1, 0), __noreturn__));

	#define fm_panic panic

	#endif /* SPL_CMN_ERR_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/debug.h b/sys/contrib/openzfs/include/os/linux/spl/sys/debug.h
	index 9c013370c5f0..8ad199808633 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/debug.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/debug.h
	@@ -1,165 +1,172 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	/*
	* Available Solaris debug functions. All of the ASSERT() macros will be
	* compiled out when NDEBUG is defined, this is the default behavior for
	* the SPL. To enable assertions use the --enable-debug with configure.
	* The VERIFY() functions are never compiled out and cannot be disabled.
	*
	* PANIC() - Panic the node and print message.
	* ASSERT() - Assert X is true, if not panic.
	* ASSERT3B() - Assert boolean X OP Y is true, if not panic.
	* ASSERT3S() - Assert signed X OP Y is true, if not panic.
	* ASSERT3U() - Assert unsigned X OP Y is true, if not panic.
	* ASSERT3P() - Assert pointer X OP Y is true, if not panic.
	* ASSERT0() - Assert value is zero, if not panic.
	* VERIFY() - Verify X is true, if not panic.
	* VERIFY3B() - Verify boolean X OP Y is true, if not panic.
	* VERIFY3S() - Verify signed X OP Y is true, if not panic.
	* VERIFY3U() - Verify unsigned X OP Y is true, if not panic.
	* VERIFY3P() - Verify pointer X OP Y is true, if not panic.
	* VERIFY0() - Verify value is zero, if not panic.
	*/

	#ifndef _SPL_DEBUG_H
	#define _SPL_DEBUG_H

	/*
	* Common DEBUG functionality.
	*/
	#define __printflike(a, b) __printf(a, b)

	#ifndef __maybe_unused
	#define __maybe_unused __attribute__((unused))
	#endif

	-int spl_panic(const char file, const char func, int line,
	- const char *fmt, ...);
	-void spl_dumpstack(void);
	+extern void spl_panic(const char file, const char func, int line,
	+ const char *fmt, ...) __attribute__((__noreturn__));
	+extern void spl_dumpstack(void);
	+
	+static inline int
	+spl_assert(const char buf, const char file, const char *func, int line)
	+{
	+ spl_panic(file, func, line, "%s", buf);
	+ return (0);
	+}

	#define PANIC(fmt, a...) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, fmt, ## a)

	#define VERIFY(cond) \
	(void) (unlikely(!(cond)) && \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "%s", "VERIFY(" #cond ") failed\n"))
	+ spl_assert("VERIFY(" #cond ") failed\n", \
	+ __FILE__, __FUNCTION__, __LINE__))

	#define VERIFY3B(LEFT, OP, RIGHT) do { \
	const boolean_t _verify3_left = (boolean_t)(LEFT); \
	const boolean_t _verify3_right = (boolean_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%d " #OP " %d)\n", \
	(boolean_t)(_verify3_left), \
	(boolean_t)(_verify3_right)); \
	} while (0)

	#define VERIFY3S(LEFT, OP, RIGHT) do { \
	const int64_t _verify3_left = (int64_t)(LEFT); \
	const int64_t _verify3_right = (int64_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%lld " #OP " %lld)\n", \
	(long long)(_verify3_left), \
	(long long)(_verify3_right)); \
	} while (0)

	#define VERIFY3U(LEFT, OP, RIGHT) do { \
	const uint64_t _verify3_left = (uint64_t)(LEFT); \
	const uint64_t _verify3_right = (uint64_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%llu " #OP " %llu)\n", \
	(unsigned long long)(_verify3_left), \
	(unsigned long long)(_verify3_right)); \
	} while (0)

	#define VERIFY3P(LEFT, OP, RIGHT) do { \
	const uintptr_t _verify3_left = (uintptr_t)(LEFT); \
	const uintptr_t _verify3_right = (uintptr_t)(RIGHT); \
	if (unlikely(!(_verify3_left OP _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(" #LEFT " " #OP " " #RIGHT ") " \
	"failed (%px " #OP " %px)\n", \
	(void *) (_verify3_left), \
	(void *) (_verify3_right)); \
	} while (0)

	#define VERIFY0(RIGHT) do { \
	const int64_t _verify3_left = (int64_t)(0); \
	const int64_t _verify3_right = (int64_t)(RIGHT); \
	if (unlikely(!(_verify3_left == _verify3_right))) \
	spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	"VERIFY3(0 == " #RIGHT ") " \
	"failed (0 == %lld)\n", \
	(long long) (_verify3_right)); \
	} while (0)

	/*
	* Debugging disabled (--disable-debug)
	*/
	#ifdef NDEBUG

	#define ASSERT(x) ((void) sizeof ((uintptr_t)(x)))
	#define ASSERT3B(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3S(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3U(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3P(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT0(x) ((void) sizeof ((uintptr_t)(x)))
	#define IMPLY(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))
	#define EQUIV(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))

	/*
	* Debugging enabled (--enable-debug)
	*/
	#else

	#define ASSERT3B VERIFY3B
	#define ASSERT3S VERIFY3S
	#define ASSERT3U VERIFY3U
	#define ASSERT3P VERIFY3P
	#define ASSERT0 VERIFY0
	#define ASSERT VERIFY
	#define IMPLY(A, B) \
	- ((void)(likely((!(A)) \|\| (B)) \|\| \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "(" #A ") implies (" #B ")")))
	+ ((void)(likely((!(A)) \|\| (B)) \|\| \
	+ spl_assert("(" #A ") implies (" #B ")", \
	+ __FILE__, __FUNCTION__, __LINE__)))
	#define EQUIV(A, B) \
	- ((void)(likely(!!(A) == !!(B)) \|\| \
	- spl_panic(__FILE__, __FUNCTION__, __LINE__, \
	- "(" #A ") is equivalent to (" #B ")")))
	+ ((void)(likely(!!(A) == !!(B)) \|\| \
	+ spl_assert("(" #A ") is equivalent to (" #B ")", \
	+ __FILE__, __FUNCTION__, __LINE__)))

	#endif /* NDEBUG */

	#endif /* SPL_DEBUG_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/kmem.h b/sys/contrib/openzfs/include/os/linux/spl/sys/kmem.h
	index a93e87df8069..86e872fa2492 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/kmem.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/kmem.h
	@@ -1,210 +1,214 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_KMEM_H
	#define _SPL_KMEM_H

	#include <sys/debug.h>
	#include <linux/slab.h>
	#include <linux/sched.h>
	#include <linux/mm.h>
	#include <linux/vmalloc.h>

	extern int kmem_debugging(void);
	-extern char kmem_vasprintf(const char fmt, va_list ap);
	-extern char kmem_asprintf(const char fmt, ...);
	+extern char kmem_vasprintf(const char fmt, va_list ap)
	+ __attribute__((format(printf, 1, 0)));
	+extern char kmem_asprintf(const char fmt, ...)
	+ __attribute__((format(printf, 1, 2)));
	extern char kmem_strdup(const char str);
	extern void kmem_strfree(char *str);

	/*
	* Memory allocation interfaces
	*/
	#define KM_SLEEP 0x0000 /* can block for memory; success guaranteed */
	#define KM_NOSLEEP 0x0001 /* cannot block for memory; may fail */
	#define KM_PUSHPAGE 0x0004 /* can block for memory; may use reserve */
	#define KM_ZERO 0x1000 /* zero the allocation */
	#define KM_VMEM 0x2000 /* caller is vmem_* wrapper */

	#define KM_PUBLIC_MASK (KM_SLEEP \| KM_NOSLEEP \| KM_PUSHPAGE)

	static int spl_fstrans_check(void);
	void *spl_kvmalloc(size_t size, gfp_t flags);

	/*
	* Convert a KM_* flags mask to its Linux GFP_* counterpart. The conversion
	* function is context aware which means that KM_SLEEP allocations can be
	* safely used in syncing contexts which have set PF_FSTRANS.
	*/
	static inline gfp_t
	kmem_flags_convert(int flags)
	{
	gfp_t lflags = __GFP_NOWARN \| __GFP_COMP;

	if (flags & KM_NOSLEEP) {
	lflags \|= GFP_ATOMIC \| __GFP_NORETRY;
	} else {
	lflags \|= GFP_KERNEL;
	if (spl_fstrans_check())
	lflags &= ~(__GFP_IO\|__GFP_FS);
	}

	if (flags & KM_PUSHPAGE)
	lflags \|= __GFP_HIGH;

	if (flags & KM_ZERO)
	lflags \|= __GFP_ZERO;

	return (lflags);
	}

	typedef struct {
	struct task_struct *fstrans_thread;
	unsigned int saved_flags;
	} fstrans_cookie_t;

	/*
	* Introduced in Linux 3.9, however this cannot be solely relied on before
	* Linux 3.18 as it doesn't turn off __GFP_FS as it should.
	*/
	#ifdef PF_MEMALLOC_NOIO
	#define __SPL_PF_MEMALLOC_NOIO (PF_MEMALLOC_NOIO)
	#else
	#define __SPL_PF_MEMALLOC_NOIO (0)
	#endif

	/*
	* PF_FSTRANS is removed from Linux 4.12
	*/
	#ifdef PF_FSTRANS
	#define __SPL_PF_FSTRANS (PF_FSTRANS)
	#else
	#define __SPL_PF_FSTRANS (0)
	#endif

	#define SPL_FSTRANS (__SPL_PF_FSTRANS\|__SPL_PF_MEMALLOC_NOIO)

	static inline fstrans_cookie_t
	spl_fstrans_mark(void)
	{
	fstrans_cookie_t cookie;

	BUILD_BUG_ON(SPL_FSTRANS == 0);

	cookie.fstrans_thread = current;
	cookie.saved_flags = current->flags & SPL_FSTRANS;
	current->flags \|= SPL_FSTRANS;

	return (cookie);
	}

	static inline void
	spl_fstrans_unmark(fstrans_cookie_t cookie)
	{
	ASSERT3P(cookie.fstrans_thread, ==, current);
	ASSERT((current->flags & SPL_FSTRANS) == SPL_FSTRANS);

	current->flags &= ~SPL_FSTRANS;
	current->flags \|= cookie.saved_flags;
	}

	static inline int
	spl_fstrans_check(void)
	{
	return (current->flags & SPL_FSTRANS);
	}

	/*
	* specifically used to check PF_FSTRANS flag, cannot be relied on for
	* checking spl_fstrans_mark().
	*/
	static inline int
	__spl_pf_fstrans_check(void)
	{
	return (current->flags & __SPL_PF_FSTRANS);
	}

	/*
	* Kernel compatibility for GFP flags
	*/
	/* < 4.13 */
	#ifndef __GFP_RETRY_MAYFAIL
	#define __GFP_RETRY_MAYFAIL __GFP_REPEAT
	#endif
	/* < 4.4 */
	#ifndef __GFP_RECLAIM
	#define __GFP_RECLAIM __GFP_WAIT
	#endif

	#ifdef HAVE_ATOMIC64_T
	#define kmem_alloc_used_add(size) atomic64_add(size, &kmem_alloc_used)
	#define kmem_alloc_used_sub(size) atomic64_sub(size, &kmem_alloc_used)
	#define kmem_alloc_used_read() atomic64_read(&kmem_alloc_used)
	#define kmem_alloc_used_set(size) atomic64_set(&kmem_alloc_used, size)
	extern atomic64_t kmem_alloc_used;
	extern unsigned long long kmem_alloc_max;
	#else /* HAVE_ATOMIC64_T */
	#define kmem_alloc_used_add(size) atomic_add(size, &kmem_alloc_used)
	#define kmem_alloc_used_sub(size) atomic_sub(size, &kmem_alloc_used)
	#define kmem_alloc_used_read() atomic_read(&kmem_alloc_used)
	#define kmem_alloc_used_set(size) atomic_set(&kmem_alloc_used, size)
	extern atomic_t kmem_alloc_used;
	extern unsigned long long kmem_alloc_max;
	#endif /* HAVE_ATOMIC64_T */

	extern unsigned int spl_kmem_alloc_warn;
	extern unsigned int spl_kmem_alloc_max;

	#define kmem_alloc(sz, fl) spl_kmem_alloc((sz), (fl), __func__, __LINE__)
	#define kmem_zalloc(sz, fl) spl_kmem_zalloc((sz), (fl), __func__, __LINE__)
	#define kmem_free(ptr, sz) spl_kmem_free((ptr), (sz))
	#define kmem_cache_reap_active spl_kmem_cache_reap_active

	-extern void spl_kmem_alloc(size_t sz, int fl, const char func, int line);
	-extern void spl_kmem_zalloc(size_t sz, int fl, const char func, int line);
	+extern void spl_kmem_alloc(size_t sz, int fl, const char func, int line)
	+ __attribute__((alloc_size(1)));
	+extern void spl_kmem_zalloc(size_t sz, int fl, const char func, int line)
	+ __attribute__((alloc_size(1)));
	extern void spl_kmem_free(const void *ptr, size_t sz);

	/*
	* 5.8 API change, pgprot_t argument removed.
	*/
	#ifdef HAVE_VMALLOC_PAGE_KERNEL
	#define spl_vmalloc(size, flags) __vmalloc(size, flags, PAGE_KERNEL)
	#else
	#define spl_vmalloc(size, flags) __vmalloc(size, flags)
	#endif

	/*
	* The following functions are only available for internal use.
	*/
	extern void *spl_kmem_alloc_impl(size_t size, int flags, int node);
	extern void *spl_kmem_alloc_debug(size_t size, int flags, int node);
	extern void *spl_kmem_alloc_track(size_t size, int flags,
	const char *func, int line, int node);
	extern void spl_kmem_free_impl(const void *buf, size_t size);
	extern void spl_kmem_free_debug(const void *buf, size_t size);
	extern void spl_kmem_free_track(const void *buf, size_t size);

	extern int spl_kmem_init(void);
	extern void spl_kmem_fini(void);
	extern int spl_kmem_cache_reap_active(void);

	#endif /* _SPL_KMEM_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/kstat.h b/sys/contrib/openzfs/include/os/linux/spl/sys/kstat.h
	index 928f70757545..305c411ddfa0 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/kstat.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/kstat.h
	@@ -1,218 +1,218 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_KSTAT_H
	#define _SPL_KSTAT_H

	#include <linux/module.h>
	#include <sys/types.h>
	#include <sys/time.h>
	#include <sys/kmem.h>
	#include <sys/mutex.h>
	#include <sys/proc.h>

	#define KSTAT_STRLEN 255
	#define KSTAT_RAW_MAX (128*1024)

	/*
	* For reference valid classes are:
	* disk, tape, net, controller, vm, kvm, hat, streams, kstat, misc
	*/

	#define KSTAT_TYPE_RAW 0 /* can be anything; ks_ndata >= 1 */
	#define KSTAT_TYPE_NAMED 1 /* name/value pair; ks_ndata >= 1 */
	#define KSTAT_TYPE_INTR 2 /* interrupt stats; ks_ndata == 1 */
	#define KSTAT_TYPE_IO 3 /* I/O stats; ks_ndata == 1 */
	#define KSTAT_TYPE_TIMER 4 /* event timer; ks_ndata >= 1 */
	#define KSTAT_NUM_TYPES 5

	#define KSTAT_DATA_CHAR 0
	#define KSTAT_DATA_INT32 1
	#define KSTAT_DATA_UINT32 2
	#define KSTAT_DATA_INT64 3
	#define KSTAT_DATA_UINT64 4
	#define KSTAT_DATA_LONG 5
	#define KSTAT_DATA_ULONG 6
	#define KSTAT_DATA_STRING 7
	#define KSTAT_NUM_DATAS 8

	#define KSTAT_INTR_HARD 0
	#define KSTAT_INTR_SOFT 1
	#define KSTAT_INTR_WATCHDOG 2
	#define KSTAT_INTR_SPURIOUS 3
	#define KSTAT_INTR_MULTSVC 4
	#define KSTAT_NUM_INTRS 5

	#define KSTAT_FLAG_VIRTUAL 0x01
	#define KSTAT_FLAG_VAR_SIZE 0x02
	#define KSTAT_FLAG_WRITABLE 0x04
	#define KSTAT_FLAG_PERSISTENT 0x08
	#define KSTAT_FLAG_DORMANT 0x10
	#define KSTAT_FLAG_INVALID 0x20
	#define KSTAT_FLAG_LONGSTRINGS 0x40
	#define KSTAT_FLAG_NO_HEADERS 0x80

	#define KS_MAGIC 0x9d9d9d9d

	/* Dynamic updates */
	#define KSTAT_READ 0
	#define KSTAT_WRITE 1

	struct kstat_s;
	typedef struct kstat_s kstat_t;

	typedef int kid_t; /* unique kstat id */
	typedef int kstat_update_t(struct kstat_s , int); / dynamic update cb */

	typedef struct kstat_module {
	- char ksm_name[KSTAT_STRLEN+1]; /* module name */
	+ char ksm_name[KSTAT_STRLEN]; /* module name */
	struct list_head ksm_module_list; /* module linkage */
	struct list_head ksm_kstat_list; /* list of kstat entries */
	struct proc_dir_entry ksm_proc; / proc entry */
	} kstat_module_t;

	typedef struct kstat_raw_ops {
	int (headers)(char buf, size_t size);
	int (data)(char buf, size_t size, void *data);
	void (addr)(kstat_t *ksp, loff_t index);
	} kstat_raw_ops_t;

	typedef struct kstat_proc_entry {
	- char kpe_name[KSTAT_STRLEN+1]; /* kstat name */
	- char kpe_module[KSTAT_STRLEN+1]; /* provider module name */
	+ char kpe_name[KSTAT_STRLEN]; /* kstat name */
	+ char kpe_module[KSTAT_STRLEN]; /* provider module name */
	kstat_module_t kpe_owner; / kstat module linkage */
	struct list_head kpe_list; /* kstat linkage */
	struct proc_dir_entry kpe_proc; / procfs entry */
	} kstat_proc_entry_t;

	struct kstat_s {
	int ks_magic; /* magic value */
	kid_t ks_kid; /* unique kstat ID */
	hrtime_t ks_crtime; /* creation time */
	hrtime_t ks_snaptime; /* last access time */
	int ks_instance; /* provider module instance */
	- char ks_class[KSTAT_STRLEN+1]; /* kstat class */
	+ char ks_class[KSTAT_STRLEN]; /* kstat class */
	uchar_t ks_type; /* kstat data type */
	uchar_t ks_flags; /* kstat flags */
	void ks_data; / kstat type-specific data */
	uint_t ks_ndata; /* # of data records */
	size_t ks_data_size; /* size of kstat data section */
	kstat_update_t ks_update; / dynamic updates */
	void ks_private; / private data */
	kmutex_t ks_private_lock; /* kstat private data lock */
	kmutex_t ks_lock; / kstat data lock */
	kstat_raw_ops_t ks_raw_ops; /* ops table for raw type */
	char ks_raw_buf; / buf used for raw ops */
	size_t ks_raw_bufsize; /* size of raw ops buffer */
	kstat_proc_entry_t ks_proc; /* data for procfs entry */
	};

	typedef struct kstat_named_s {
	char name[KSTAT_STRLEN]; /* name of counter */
	uchar_t data_type; /* data type */
	union {
	char c[16]; /* 128-bit int */
	int32_t i32; /* 32-bit signed int */
	uint32_t ui32; /* 32-bit unsigned int */
	int64_t i64; /* 64-bit signed int */
	uint64_t ui64; /* 64-bit unsigned int */
	long l; /* native signed long */
	ulong_t ul; /* native unsigned long */
	struct {
	union {
	char ptr; / NULL-term string */
	char __pad[8]; /* 64-bit padding */
	} addr;
	uint32_t len; /* # bytes for strlen + '\0' */
	} string;
	} value;
	} kstat_named_t;

	#define KSTAT_NAMED_STR_PTR(knptr) ((knptr)->value.string.addr.ptr)
	#define KSTAT_NAMED_STR_BUFLEN(knptr) ((knptr)->value.string.len)

	#ifdef HAVE_PROC_OPS_STRUCT
	typedef struct proc_ops kstat_proc_op_t;
	#else
	typedef struct file_operations kstat_proc_op_t;
	#endif

	typedef struct kstat_intr {
	uint_t intrs[KSTAT_NUM_INTRS];
	} kstat_intr_t;

	typedef struct kstat_io {
	u_longlong_t nread; /* number of bytes read */
	u_longlong_t nwritten; /* number of bytes written */
	uint_t reads; /* number of read operations */
	uint_t writes; /* number of write operations */
	hrtime_t wtime; /* cumulative wait (pre-service) time */
	hrtime_t wlentime; /* cumulative wait lentime product /
	hrtime_t wlastupdate; /* last time wait queue changed */
	hrtime_t rtime; /* cumulative run (service) time */
	hrtime_t rlentime; /* cumulative run lengthtime product /
	hrtime_t rlastupdate; /* last time run queue changed */
	uint_t wcnt; /* count of elements in wait state */
	uint_t rcnt; /* count of elements in run state */
	} kstat_io_t;

	typedef struct kstat_timer {
	- char name[KSTAT_STRLEN+1]; /* event name */
	+ char name[KSTAT_STRLEN]; /* event name */
	u_longlong_t num_events; /* number of events */
	hrtime_t elapsed_time; /* cumulative elapsed time */
	hrtime_t min_time; /* shortest event duration */
	hrtime_t max_time; /* longest event duration */
	hrtime_t start_time; /* previous event start time */
	hrtime_t stop_time; /* previous event stop time */
	} kstat_timer_t;

	int spl_kstat_init(void);
	void spl_kstat_fini(void);

	extern void __kstat_set_raw_ops(kstat_t *ksp,
	int (headers)(char buf, size_t size),
	int (data)(char buf, size_t size, void *data),
	void* (addr)(kstat_t ksp, loff_t index));

	extern kstat_t __kstat_create(const char ks_module, int ks_instance,
	const char ks_name, const char ks_class, uchar_t ks_type,
	uint_t ks_ndata, uchar_t ks_flags);

	extern void kstat_proc_entry_init(kstat_proc_entry_t *kpep,
	const char module, const char name);
	extern void kstat_proc_entry_delete(kstat_proc_entry_t *kpep);
	extern void kstat_proc_entry_install(kstat_proc_entry_t *kpep, mode_t mode,
	const kstat_proc_op_t file_ops, void data);

	extern void __kstat_install(kstat_t *ksp);
	extern void __kstat_delete(kstat_t *ksp);

	#define kstat_set_raw_ops(k, h, d, a) \
	__kstat_set_raw_ops(k, h, d, a)
	#define kstat_create(m, i, n, c, t, s, f) \
	__kstat_create(m, i, n, c, t, s, f)

	#define kstat_install(k) __kstat_install(k)
	#define kstat_delete(k) __kstat_delete(k)

	#endif /* _SPL_KSTAT_H */
	diff --git a/sys/contrib/openzfs/lib/libspl/assert.c b/sys/contrib/openzfs/include/os/linux/spl/sys/misc.h
	similarity index 57%
	copy from sys/contrib/openzfs/lib/libspl/assert.c
	copy to sys/contrib/openzfs/include/os/linux/spl/sys/misc.h
	index 207da3a1d860..299fe9c1ab07 100644
	--- a/sys/contrib/openzfs/lib/libspl/assert.c
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/misc.h
	@@ -1,52 +1,29 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	-/*
	- * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	- * Use is subject to license terms.
	- */
	-
	-#include <assert.h>

	-static boolean_t libspl_assert_ok = B_FALSE;
	+#ifndef _OS_LINUX_SPL_MISC_H
	+#define _OS_LINUX_SPL_MISC_H

	-void
	-libspl_set_assert_ok(boolean_t val)
	-{
	- libspl_assert_ok = val;
	-}
	+#include <linux/kobject.h>

	-/* printf version of libspl_assert */
	-void
	-libspl_assertf(const char file, const char func, int line,
	- const char *format, ...)
	-{
	- va_list args;
	+extern void spl_signal_kobj_evt(struct block_device *bdev);

	- va_start(args, format);
	- vfprintf(stderr, format, args);
	- fprintf(stderr, "\n");
	- fprintf(stderr, "ASSERT at %s:%d:%s()", file, line, func);
	- va_end(args);
	- if (libspl_assert_ok) {
	- return;
	- }
	- abort();
	-}
	+#endif
	diff --git a/sys/contrib/openzfs/include/sys/arc.h b/sys/contrib/openzfs/include/sys/arc.h
	index fc8605122434..532a2fe4bc03 100644
	--- a/sys/contrib/openzfs/include/sys/arc.h
	+++ b/sys/contrib/openzfs/include/sys/arc.h
	@@ -1,351 +1,351 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2016 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2019, Klara Inc.
	*/

	#ifndef _SYS_ARC_H
	#define _SYS_ARC_H

	#include <sys/zfs_context.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	#include <sys/zio.h>
	#include <sys/dmu.h>
	#include <sys/spa.h>
	#include <sys/zfs_refcount.h>

	/*
	* Used by arc_flush() to inform arc_evict_state() that it should evict
	* all available buffers from the arc state being passed in.
	*/
	#define ARC_EVICT_ALL UINT64_MAX

	/*
	* ZFS gets very unhappy when the maximum ARC size is smaller than the maximum
	* block size and a larger block is written. To leave some safety margin, we
	* limit the minimum for zfs_arc_max to the maximium transaction size.
	*/
	#define MIN_ARC_MAX DMU_MAX_ACCESS

	#define HDR_SET_LSIZE(hdr, x) do { \
	ASSERT(IS_P2ALIGNED(x, 1U << SPA_MINBLOCKSHIFT)); \
	(hdr)->b_lsize = ((x) >> SPA_MINBLOCKSHIFT); \
	} while (0)

	#define HDR_SET_PSIZE(hdr, x) do { \
	ASSERT(IS_P2ALIGNED((x), 1U << SPA_MINBLOCKSHIFT)); \
	(hdr)->b_psize = ((x) >> SPA_MINBLOCKSHIFT); \
	} while (0)

	#define HDR_GET_LSIZE(hdr) ((hdr)->b_lsize << SPA_MINBLOCKSHIFT)
	#define HDR_GET_PSIZE(hdr) ((hdr)->b_psize << SPA_MINBLOCKSHIFT)

	typedef struct arc_buf_hdr arc_buf_hdr_t;
	typedef struct arc_buf arc_buf_t;
	typedef struct arc_prune arc_prune_t;

	/*
	* Because the ARC can store encrypted data, errors (not due to bugs) may arise
	* while transforming data into its desired format - specifically, when
	* decrypting, the key may not be present, or the HMAC may not be correct
	* which signifies deliberate tampering with the on-disk state
	* (assuming that the checksum was correct). If any error occurs, the "buf"
	* parameter will be NULL.
	*/
	typedef void arc_read_done_func_t(zio_t zio, const zbookmark_phys_t zb,
	const blkptr_t bp, arc_buf_t buf, void *priv);
	typedef void arc_write_done_func_t(zio_t zio, arc_buf_t buf, void *priv);
	typedef void arc_prune_func_t(int64_t bytes, void *priv);

	/* Shared module parameters */
	-extern int zfs_arc_average_blocksize;
	+extern uint_t zfs_arc_average_blocksize;
	extern int l2arc_exclude_special;

	/* generic arc_done_func_t's which you can use */
	arc_read_done_func_t arc_bcopy_func;
	arc_read_done_func_t arc_getbuf_func;

	/* generic arc_prune_func_t wrapper for callbacks */
	struct arc_prune {
	arc_prune_func_t *p_pfunc;
	void *p_private;
	uint64_t p_adjust;
	list_node_t p_node;
	zfs_refcount_t p_refcnt;
	};

	typedef enum arc_strategy {
	ARC_STRATEGY_META_ONLY = 0, /* Evict only meta data buffers */
	ARC_STRATEGY_META_BALANCED = 1, /* Evict data buffers if needed */
	} arc_strategy_t;

	typedef enum arc_flags
	{
	/*
	* Public flags that can be passed into the ARC by external consumers.
	*/
	ARC_FLAG_WAIT = 1 << 0, /* perform sync I/O */
	ARC_FLAG_NOWAIT = 1 << 1, /* perform async I/O */
	ARC_FLAG_PREFETCH = 1 << 2, /* I/O is a prefetch */
	ARC_FLAG_CACHED = 1 << 3, /* I/O was in cache */
	ARC_FLAG_L2CACHE = 1 << 4, /* cache in L2ARC */
	ARC_FLAG_PREDICTIVE_PREFETCH = 1 << 5, /* I/O from zfetch */
	ARC_FLAG_PRESCIENT_PREFETCH = 1 << 6, /* long min lifespan */

	/*
	* Private ARC flags. These flags are private ARC only flags that
	* will show up in b_flags in the arc_hdr_buf_t. These flags should
	* only be set by ARC code.
	*/
	ARC_FLAG_IN_HASH_TABLE = 1 << 7, /* buffer is hashed */
	ARC_FLAG_IO_IN_PROGRESS = 1 << 8, /* I/O in progress */
	ARC_FLAG_IO_ERROR = 1 << 9, /* I/O failed for buf */
	ARC_FLAG_INDIRECT = 1 << 10, /* indirect block */
	/* Indicates that block was read with ASYNC priority. */
	ARC_FLAG_PRIO_ASYNC_READ = 1 << 11,
	ARC_FLAG_L2_WRITING = 1 << 12, /* write in progress */
	ARC_FLAG_L2_EVICTED = 1 << 13, /* evicted during I/O */
	ARC_FLAG_L2_WRITE_HEAD = 1 << 14, /* head of write list */
	/*
	* Encrypted or authenticated on disk (may be plaintext in memory).
	* This header has b_crypt_hdr allocated. Does not include indirect
	* blocks with checksums of MACs which will also have their X
	* (encrypted) bit set in the bp.
	*/
	ARC_FLAG_PROTECTED = 1 << 15,
	/* data has not been authenticated yet */
	ARC_FLAG_NOAUTH = 1 << 16,
	/* indicates that the buffer contains metadata (otherwise, data) */
	ARC_FLAG_BUFC_METADATA = 1 << 17,

	/* Flags specifying whether optional hdr struct fields are defined */
	ARC_FLAG_HAS_L1HDR = 1 << 18,
	ARC_FLAG_HAS_L2HDR = 1 << 19,

	/*
	* Indicates the arc_buf_hdr_t's b_pdata matches the on-disk data.
	* This allows the l2arc to use the blkptr's checksum to verify
	* the data without having to store the checksum in the hdr.
	*/
	ARC_FLAG_COMPRESSED_ARC = 1 << 20,
	ARC_FLAG_SHARED_DATA = 1 << 21,

	/*
	* Fail this arc_read() (with ENOENT) if the data is not already present
	* in cache.
	*/
	ARC_FLAG_CACHED_ONLY = 1 << 22,

	/*
	* Don't instantiate an arc_buf_t for arc_read_done.
	*/
	ARC_FLAG_NO_BUF = 1 << 23,

	/*
	* The arc buffer's compression mode is stored in the top 7 bits of the
	* flags field, so these dummy flags are included so that MDB can
	* interpret the enum properly.
	*/
	ARC_FLAG_COMPRESS_0 = 1 << 24,
	ARC_FLAG_COMPRESS_1 = 1 << 25,
	ARC_FLAG_COMPRESS_2 = 1 << 26,
	ARC_FLAG_COMPRESS_3 = 1 << 27,
	ARC_FLAG_COMPRESS_4 = 1 << 28,
	ARC_FLAG_COMPRESS_5 = 1 << 29,
	ARC_FLAG_COMPRESS_6 = 1 << 30

	} arc_flags_t;

	typedef enum arc_buf_flags {
	ARC_BUF_FLAG_SHARED = 1 << 0,
	ARC_BUF_FLAG_COMPRESSED = 1 << 1,
	/*
	* indicates whether this arc_buf_t is encrypted, regardless of
	* state on-disk
	*/
	ARC_BUF_FLAG_ENCRYPTED = 1 << 2
	} arc_buf_flags_t;

	struct arc_buf {
	arc_buf_hdr_t *b_hdr;
	arc_buf_t *b_next;
	kmutex_t b_evict_lock;
	void *b_data;
	arc_buf_flags_t b_flags;
	};

	typedef enum arc_buf_contents {
	ARC_BUFC_INVALID, /* invalid type */
	ARC_BUFC_DATA, /* buffer contains data */
	ARC_BUFC_METADATA, /* buffer contains metadata */
	ARC_BUFC_NUMTYPES
	} arc_buf_contents_t;

	/*
	* The following breakdowns of arc_size exist for kstat only.
	*/
	typedef enum arc_space_type {
	ARC_SPACE_DATA,
	ARC_SPACE_META,
	ARC_SPACE_HDRS,
	ARC_SPACE_L2HDRS,
	ARC_SPACE_DBUF,
	ARC_SPACE_DNODE,
	ARC_SPACE_BONUS,
	ARC_SPACE_ABD_CHUNK_WASTE,
	ARC_SPACE_NUMTYPES
	} arc_space_type_t;

	typedef enum arc_state_type {
	ARC_STATE_ANON,
	ARC_STATE_MRU,
	ARC_STATE_MRU_GHOST,
	ARC_STATE_MFU,
	ARC_STATE_MFU_GHOST,
	ARC_STATE_L2C_ONLY,
	ARC_STATE_NUMTYPES
	} arc_state_type_t;

	typedef struct arc_buf_info {
	arc_state_type_t abi_state_type;
	arc_buf_contents_t abi_state_contents;
	uint32_t abi_flags;
	uint32_t abi_bufcnt;
	uint64_t abi_size;
	uint64_t abi_spa;
	uint64_t abi_access;
	uint32_t abi_mru_hits;
	uint32_t abi_mru_ghost_hits;
	uint32_t abi_mfu_hits;
	uint32_t abi_mfu_ghost_hits;
	uint32_t abi_l2arc_hits;
	uint32_t abi_holds;
	uint64_t abi_l2arc_dattr;
	uint64_t abi_l2arc_asize;
	enum zio_compress abi_l2arc_compress;
	} arc_buf_info_t;

	void arc_space_consume(uint64_t space, arc_space_type_t type);
	void arc_space_return(uint64_t space, arc_space_type_t type);
	boolean_t arc_is_metadata(arc_buf_t *buf);
	boolean_t arc_is_encrypted(arc_buf_t *buf);
	boolean_t arc_is_unauthenticated(arc_buf_t *buf);
	enum zio_compress arc_get_compression(arc_buf_t *buf);
	void arc_get_raw_params(arc_buf_t buf, boolean_t byteorder, uint8_t *salt,
	uint8_t iv, uint8_t mac);
	int arc_untransform(arc_buf_t buf, spa_t spa, const zbookmark_phys_t *zb,
	boolean_t in_place);
	void arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder,
	dmu_object_type_t ot, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac);
	arc_buf_t arc_alloc_buf(spa_t spa, const void *tag, arc_buf_contents_t type,
	int32_t size);
	arc_buf_t arc_alloc_compressed_buf(spa_t spa, const void *tag,
	uint64_t psize, uint64_t lsize, enum zio_compress compression_type,
	uint8_t complevel);
	arc_buf_t arc_alloc_raw_buf(spa_t spa, const void *tag, uint64_t dsobj,
	boolean_t byteorder, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel);
	uint8_t arc_get_complevel(arc_buf_t *buf);
	arc_buf_t arc_loan_buf(spa_t spa, boolean_t is_metadata, int size);
	arc_buf_t arc_loan_compressed_buf(spa_t spa, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel);
	arc_buf_t arc_loan_raw_buf(spa_t spa, uint64_t dsobj, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t *mac,
	dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel);
	void arc_return_buf(arc_buf_t buf, const void tag);
	void arc_loan_inuse_buf(arc_buf_t buf, const void tag);
	void arc_buf_destroy(arc_buf_t buf, const void tag);
	void arc_buf_info(arc_buf_t buf, arc_buf_info_t abi, int state_index);
	uint64_t arc_buf_size(arc_buf_t *buf);
	uint64_t arc_buf_lsize(arc_buf_t *buf);
	void arc_buf_access(arc_buf_t *buf);
	void arc_release(arc_buf_t buf, const void tag);
	int arc_released(arc_buf_t *buf);
	void arc_buf_sigsegv(int sig, siginfo_t si, void unused);
	void arc_buf_freeze(arc_buf_t *buf);
	void arc_buf_thaw(arc_buf_t *buf);
	#ifdef ZFS_DEBUG
	int arc_referenced(arc_buf_t *buf);
	#else
	#define arc_referenced(buf) ((void) sizeof (buf), 0)
	#endif

	int arc_read(zio_t pio, spa_t spa, const blkptr_t *bp,
	arc_read_done_func_t done, void priv, zio_priority_t priority,
	int flags, arc_flags_t arc_flags, const zbookmark_phys_t zb);
	zio_t arc_write(zio_t pio, spa_t *spa, uint64_t txg,
	blkptr_t bp, arc_buf_t buf, boolean_t l2arc, const zio_prop_t *zp,
	arc_write_done_func_t ready, arc_write_done_func_t child_ready,
	arc_write_done_func_t physdone, arc_write_done_func_t done,
	void *priv, zio_priority_t priority, int zio_flags,
	const zbookmark_phys_t *zb);

	arc_prune_t arc_add_prune_callback(arc_prune_func_t func, void *priv);
	void arc_remove_prune_callback(arc_prune_t *p);
	void arc_freed(spa_t spa, const blkptr_t bp);

	void arc_flush(spa_t *spa, boolean_t retry);
	void arc_tempreserve_clear(uint64_t reserve);
	int arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg);

	uint64_t arc_all_memory(void);
	uint64_t arc_default_max(uint64_t min, uint64_t allmem);
	uint64_t arc_target_bytes(void);
	void arc_set_limits(uint64_t);
	void arc_init(void);
	void arc_fini(void);

	/*
	* Level 2 ARC
	*/

	void l2arc_add_vdev(spa_t spa, vdev_t vd);
	void l2arc_remove_vdev(vdev_t *vd);
	boolean_t l2arc_vdev_present(vdev_t *vd);
	void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen);
	boolean_t l2arc_range_check_overlap(uint64_t bottom, uint64_t top,
	uint64_t check);
	void l2arc_init(void);
	void l2arc_fini(void);
	void l2arc_start(void);
	void l2arc_stop(void);
	void l2arc_spa_rebuild_start(spa_t *spa);

	#ifndef _KERNEL
	extern boolean_t arc_watch;
	#endif

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ARC_H */
	diff --git a/sys/contrib/openzfs/include/sys/arc_impl.h b/sys/contrib/openzfs/include/sys/arc_impl.h
	index 4e2cff7e2ed2..0f2cd956eb64 100644
	--- a/sys/contrib/openzfs/include/sys/arc_impl.h
	+++ b/sys/contrib/openzfs/include/sys/arc_impl.h
	@@ -1,1023 +1,1023 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, Delphix. All rights reserved.
	* Copyright (c) 2013, Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2020, George Amanakis. All rights reserved.
	*/

	#ifndef _SYS_ARC_IMPL_H
	#define _SYS_ARC_IMPL_H

	#include <sys/arc.h>
	#include <sys/zio_crypt.h>
	#include <sys/zthr.h>
	#include <sys/aggsum.h>
	#include <sys/wmsum.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Note that buffers can be in one of 6 states:
	* ARC_anon - anonymous (discussed below)
	* ARC_mru - recently used, currently cached
	* ARC_mru_ghost - recently used, no longer in cache
	* ARC_mfu - frequently used, currently cached
	* ARC_mfu_ghost - frequently used, no longer in cache
	* ARC_l2c_only - exists in L2ARC but not other states
	* When there are no active references to the buffer, they are
	* are linked onto a list in one of these arc states. These are
	* the only buffers that can be evicted or deleted. Within each
	* state there are multiple lists, one for meta-data and one for
	* non-meta-data. Meta-data (indirect blocks, blocks of dnodes,
	* etc.) is tracked separately so that it can be managed more
	* explicitly: favored over data, limited explicitly.
	*
	* Anonymous buffers are buffers that are not associated with
	* a DVA. These are buffers that hold dirty block copies
	* before they are written to stable storage. By definition,
	* they are "ref'd" and are considered part of arc_mru
	* that cannot be freed. Generally, they will acquire a DVA
	* as they are written and migrate onto the arc_mru list.
	*
	* The ARC_l2c_only state is for buffers that are in the second
	* level ARC but no longer in any of the ARC_m* lists. The second
	* level ARC itself may also contain buffers that are in any of
	* the ARC_m* states - meaning that a buffer can exist in two
	* places. The reason for the ARC_l2c_only state is to keep the
	* buffer header in the hash table, so that reads that hit the
	* second level ARC benefit from these fast lookups.
	*/

	typedef struct arc_state {
	/*
	* list of evictable buffers
	*/
	multilist_t arcs_list[ARC_BUFC_NUMTYPES];
	/*
	* supports the "dbufs" kstat
	*/
	arc_state_type_t arcs_state;
	/*
	* total amount of evictable data in this state
	*/
	zfs_refcount_t arcs_esize[ARC_BUFC_NUMTYPES] ____cacheline_aligned;
	/*
	* total amount of data in this state; this includes: evictable,
	* non-evictable, ARC_BUFC_DATA, and ARC_BUFC_METADATA.
	*/
	zfs_refcount_t arcs_size;
	} arc_state_t;

	typedef struct arc_callback arc_callback_t;

	struct arc_callback {
	void *acb_private;
	arc_read_done_func_t *acb_done;
	arc_buf_t *acb_buf;
	boolean_t acb_encrypted;
	boolean_t acb_compressed;
	boolean_t acb_noauth;
	boolean_t acb_nobuf;
	zbookmark_phys_t acb_zb;
	zio_t *acb_zio_dummy;
	zio_t *acb_zio_head;
	arc_callback_t *acb_next;
	};

	typedef struct arc_write_callback arc_write_callback_t;

	struct arc_write_callback {
	void *awcb_private;
	arc_write_done_func_t *awcb_ready;
	arc_write_done_func_t *awcb_children_ready;
	arc_write_done_func_t *awcb_physdone;
	arc_write_done_func_t *awcb_done;
	arc_buf_t *awcb_buf;
	};

	/*
	* ARC buffers are separated into multiple structs as a memory saving measure:
	* - Common fields struct, always defined, and embedded within it:
	* - L2-only fields, always allocated but undefined when not in L2ARC
	* - L1-only fields, only allocated when in L1ARC
	*
	* Buffer in L1 Buffer only in L2
	* +------------------------+ +------------------------+
	* \| arc_buf_hdr_t \| \| arc_buf_hdr_t \|
	* \| \| \| \|
	* \| \| \| \|
	* \| \| \| \|
	* +------------------------+ +------------------------+
	* \| l2arc_buf_hdr_t \| \| l2arc_buf_hdr_t \|
	* \| (undefined if L1-only) \| \| \|
	* +------------------------+ +------------------------+
	* \| l1arc_buf_hdr_t \|
	* \| \|
	* \| \|
	* \| \|
	* \| \|
	* +------------------------+
	*
	* Because it's possible for the L2ARC to become extremely large, we can wind
	* up eating a lot of memory in L2ARC buffer headers, so the size of a header
	* is minimized by only allocating the fields necessary for an L1-cached buffer
	* when a header is actually in the L1 cache. The sub-headers (l1arc_buf_hdr and
	* l2arc_buf_hdr) are embedded rather than allocated separately to save a couple
	* words in pointers. arc_hdr_realloc() is used to switch a header between
	* these two allocation states.
	*/
	typedef struct l1arc_buf_hdr {
	kmutex_t b_freeze_lock;
	zio_cksum_t *b_freeze_cksum;

	/* for waiting on reads to complete */
	kcondvar_t b_cv;
	uint8_t b_byteswap;

	/* protected by arc state mutex */
	arc_state_t *b_state;
	multilist_node_t b_arc_node;

	/* protected by hash lock */
	clock_t b_arc_access;
	uint32_t b_mru_hits;
	uint32_t b_mru_ghost_hits;
	uint32_t b_mfu_hits;
	uint32_t b_mfu_ghost_hits;
	uint32_t b_bufcnt;
	arc_buf_t *b_buf;

	/* self protecting */
	zfs_refcount_t b_refcnt;

	arc_callback_t *b_acb;
	abd_t *b_pabd;
	} l1arc_buf_hdr_t;

	typedef enum l2arc_dev_hdr_flags_t {
	L2ARC_DEV_HDR_EVICT_FIRST = (1 << 0) /* mirror of l2ad_first */
	} l2arc_dev_hdr_flags_t;

	/*
	* Pointer used in persistent L2ARC (for pointing to log blocks).
	*/
	typedef struct l2arc_log_blkptr {
	/*
	* Offset of log block within the device, in bytes
	*/
	uint64_t lbp_daddr;
	/*
	* Aligned payload size (in bytes) of the log block
	*/
	uint64_t lbp_payload_asize;
	/*
	* Offset in bytes of the first buffer in the payload
	*/
	uint64_t lbp_payload_start;
	/*
	* lbp_prop has the following format:
	* * logical size (in bytes)
	* * aligned (after compression) size (in bytes)
	* * compression algorithm (we always LZ4-compress l2arc logs)
	* * checksum algorithm (used for lbp_cksum)
	*/
	uint64_t lbp_prop;
	zio_cksum_t lbp_cksum; /* checksum of log */
	} l2arc_log_blkptr_t;

	/*
	* The persistent L2ARC device header.
	* Byte order of magic determines whether 64-bit bswap of fields is necessary.
	*/
	typedef struct l2arc_dev_hdr_phys {
	uint64_t dh_magic; /* L2ARC_DEV_HDR_MAGIC */
	uint64_t dh_version; /* Persistent L2ARC version */

	/*
	* Global L2ARC device state and metadata.
	*/
	uint64_t dh_spa_guid;
	uint64_t dh_vdev_guid;
	uint64_t dh_log_entries; /* mirror of l2ad_log_entries */
	uint64_t dh_evict; /* evicted offset in bytes */
	uint64_t dh_flags; /* l2arc_dev_hdr_flags_t */
	/*
	* Used in zdb.c for determining if a log block is valid, in the same
	* way that l2arc_rebuild() does.
	*/
	uint64_t dh_start; /* mirror of l2ad_start */
	uint64_t dh_end; /* mirror of l2ad_end */
	/*
	* Start of log block chain. [0] -> newest log, [1] -> one older (used
	* for initiating prefetch).
	*/
	l2arc_log_blkptr_t dh_start_lbps[2];
	/*
	* Aligned size of all log blocks as accounted by vdev_space_update().
	*/
	uint64_t dh_lb_asize; /* mirror of l2ad_lb_asize */
	uint64_t dh_lb_count; /* mirror of l2ad_lb_count */
	/*
	* Mirrors of vdev_trim_action_time and vdev_trim_state, used to
	* display when the cache device was fully trimmed for the last
	* time.
	*/
	uint64_t dh_trim_action_time;
	uint64_t dh_trim_state;
	const uint64_t dh_pad[30]; /* pad to 512 bytes */
	zio_eck_t dh_tail;
	} l2arc_dev_hdr_phys_t;
	_Static_assert(sizeof (l2arc_dev_hdr_phys_t) == SPA_MINBLOCKSIZE,
	"l2arc_dev_hdr_phys_t wrong size");

	/*
	* A single ARC buffer header entry in a l2arc_log_blk_phys_t.
	*/
	typedef struct l2arc_log_ent_phys {
	dva_t le_dva; /* dva of buffer */
	uint64_t le_birth; /* birth txg of buffer */
	/*
	* le_prop has the following format:
	* * logical size (in bytes)
	* * physical (compressed) size (in bytes)
	* * compression algorithm
	* * object type (used to restore arc_buf_contents_t)
	* * protected status (used for encryption)
	* * prefetch status (used in l2arc_read_done())
	*/
	uint64_t le_prop;
	uint64_t le_daddr; /* buf location on l2dev */
	uint64_t le_complevel;
	/*
	* We pad the size of each entry to a power of 2 so that the size of
	* l2arc_log_blk_phys_t is power-of-2 aligned with SPA_MINBLOCKSHIFT,
	* because of the L2ARC_SET_*SIZE macros.
	*/
	const uint64_t le_pad[2]; /* pad to 64 bytes */
	} l2arc_log_ent_phys_t;

	#define L2ARC_LOG_BLK_MAX_ENTRIES (1022)

	/*
	* A log block of up to 1022 ARC buffer log entries, chained into the
	* persistent L2ARC metadata linked list. Byte order of magic determines
	* whether 64-bit bswap of fields is necessary.
	*/
	typedef struct l2arc_log_blk_phys {
	uint64_t lb_magic; /* L2ARC_LOG_BLK_MAGIC */
	/*
	* There are 2 chains (headed by dh_start_lbps[2]), and this field
	* points back to the previous block in this chain. We alternate
	* which chain we append to, so they are time-wise and offset-wise
	* interleaved, but that is an optimization rather than for
	* correctness.
	*/
	l2arc_log_blkptr_t lb_prev_lbp; /* pointer to prev log block */
	/*
	* Pad header section to 128 bytes
	*/
	uint64_t lb_pad[7];
	/* Payload */
	l2arc_log_ent_phys_t lb_entries[L2ARC_LOG_BLK_MAX_ENTRIES];
	} l2arc_log_blk_phys_t; /* 64K total */

	/*
	* The size of l2arc_log_blk_phys_t has to be power-of-2 aligned with
	* SPA_MINBLOCKSHIFT because of L2BLK_SET_*SIZE macros.
	*/
	_Static_assert(IS_P2ALIGNED(sizeof (l2arc_log_blk_phys_t),
	1ULL << SPA_MINBLOCKSHIFT), "l2arc_log_blk_phys_t misaligned");
	_Static_assert(sizeof (l2arc_log_blk_phys_t) >= SPA_MINBLOCKSIZE,
	"l2arc_log_blk_phys_t too small");
	_Static_assert(sizeof (l2arc_log_blk_phys_t) <= SPA_MAXBLOCKSIZE,
	"l2arc_log_blk_phys_t too big");

	/*
	* These structures hold in-flight abd buffers for log blocks as they're being
	* written to the L2ARC device.
	*/
	typedef struct l2arc_lb_abd_buf {
	abd_t *abd;
	list_node_t node;
	} l2arc_lb_abd_buf_t;

	/*
	* These structures hold pointers to log blocks present on the L2ARC device.
	*/
	typedef struct l2arc_lb_ptr_buf {
	l2arc_log_blkptr_t *lb_ptr;
	list_node_t node;
	} l2arc_lb_ptr_buf_t;

	/* Macros for setting fields in le_prop and lbp_prop */
	#define L2BLK_GET_LSIZE(field) \
	BF64_GET_SB((field), 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)
	#define L2BLK_SET_LSIZE(field, x) \
	BF64_SET_SB((field), 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
	#define L2BLK_GET_PSIZE(field) \
	BF64_GET_SB((field), 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1)
	#define L2BLK_SET_PSIZE(field, x) \
	BF64_SET_SB((field), 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x)
	#define L2BLK_GET_COMPRESS(field) \
	BF64_GET((field), 32, SPA_COMPRESSBITS)
	#define L2BLK_SET_COMPRESS(field, x) \
	BF64_SET((field), 32, SPA_COMPRESSBITS, x)
	#define L2BLK_GET_PREFETCH(field) BF64_GET((field), 39, 1)
	#define L2BLK_SET_PREFETCH(field, x) BF64_SET((field), 39, 1, x)
	#define L2BLK_GET_CHECKSUM(field) BF64_GET((field), 40, 8)
	#define L2BLK_SET_CHECKSUM(field, x) BF64_SET((field), 40, 8, x)
	#define L2BLK_GET_TYPE(field) BF64_GET((field), 48, 8)
	#define L2BLK_SET_TYPE(field, x) BF64_SET((field), 48, 8, x)
	#define L2BLK_GET_PROTECTED(field) BF64_GET((field), 56, 1)
	#define L2BLK_SET_PROTECTED(field, x) BF64_SET((field), 56, 1, x)
	#define L2BLK_GET_STATE(field) BF64_GET((field), 57, 4)
	#define L2BLK_SET_STATE(field, x) BF64_SET((field), 57, 4, x)

	#define PTR_SWAP(x, y) \
	do { \
	void *tmp = (x);\
	x = y; \
	y = tmp; \
	} while (0)

	#define L2ARC_DEV_HDR_MAGIC 0x5a46534341434845LLU /* ASCII: "ZFSCACHE" */
	#define L2ARC_LOG_BLK_MAGIC 0x4c4f47424c4b4844LLU /* ASCII: "LOGBLKHD" */

	/*
	* L2ARC Internals
	*/
	typedef struct l2arc_dev {
	vdev_t l2ad_vdev; / vdev */
	spa_t l2ad_spa; / spa */
	uint64_t l2ad_hand; /* next write location */
	uint64_t l2ad_start; /* first addr on device */
	uint64_t l2ad_end; /* last addr on device */
	boolean_t l2ad_first; /* first sweep through */
	boolean_t l2ad_writing; /* currently writing */
	kmutex_t l2ad_mtx; /* lock for buffer list */
	list_t l2ad_buflist; /* buffer list */
	list_node_t l2ad_node; /* device list node */
	zfs_refcount_t l2ad_alloc; /* allocated bytes */
	/*
	* Persistence-related stuff
	*/
	l2arc_dev_hdr_phys_t l2ad_dev_hdr; / persistent device header */
	uint64_t l2ad_dev_hdr_asize; /* aligned hdr size */
	l2arc_log_blk_phys_t l2ad_log_blk; /* currently open log block */
	int l2ad_log_ent_idx; /* index into cur log blk */
	/* Number of bytes in current log block's payload */
	uint64_t l2ad_log_blk_payload_asize;
	/*
	* Offset (in bytes) of the first buffer in current log block's
	* payload.
	*/
	uint64_t l2ad_log_blk_payload_start;
	/* Flag indicating whether a rebuild is scheduled or is going on */
	boolean_t l2ad_rebuild;
	boolean_t l2ad_rebuild_cancel;
	boolean_t l2ad_rebuild_began;
	uint64_t l2ad_log_entries; /* entries per log blk */
	uint64_t l2ad_evict; /* evicted offset in bytes */
	/* List of pointers to log blocks present in the L2ARC device */
	list_t l2ad_lbptr_list;
	/*
	* Aligned size of all log blocks as accounted by vdev_space_update().
	*/
	zfs_refcount_t l2ad_lb_asize;
	/*
	* Number of log blocks present on the device.
	*/
	zfs_refcount_t l2ad_lb_count;
	boolean_t l2ad_trim_all; /* TRIM whole device */
	} l2arc_dev_t;

	/*
	* Encrypted blocks will need to be stored encrypted on the L2ARC
	* disk as they appear in the main pool. In order for this to work we
	* need to pass around the encryption parameters so they can be used
	* to write data to the L2ARC. This struct is only defined in the
	* arc_buf_hdr_t if the L1 header is defined and has the ARC_FLAG_ENCRYPTED
	* flag set.
	*/
	typedef struct arc_buf_hdr_crypt {
	abd_t b_rabd; / raw encrypted data */
	dmu_object_type_t b_ot; /* object type */
	uint32_t b_ebufcnt; /* count of encrypted buffers */

	/* dsobj for looking up encryption key for l2arc encryption */
	uint64_t b_dsobj;

	/* encryption parameters */
	uint8_t b_salt[ZIO_DATA_SALT_LEN];
	uint8_t b_iv[ZIO_DATA_IV_LEN];

	/*
	* Technically this could be removed since we will always be able to
	* get the mac from the bp when we need it. However, it is inconvenient
	* for callers of arc code to have to pass a bp in all the time. This
	* also allows us to assert that L2ARC data is properly encrypted to
	* match the data in the main storage pool.
	*/
	uint8_t b_mac[ZIO_DATA_MAC_LEN];
	} arc_buf_hdr_crypt_t;

	typedef struct l2arc_buf_hdr {
	/* protected by arc_buf_hdr mutex */
	l2arc_dev_t b_dev; / L2ARC device */
	uint64_t b_daddr; /* disk address, offset byte */
	uint32_t b_hits;
	arc_state_type_t b_arcs_state;
	list_node_t b_l2node;
	} l2arc_buf_hdr_t;

	typedef struct l2arc_write_callback {
	l2arc_dev_t l2wcb_dev; / device info */
	arc_buf_hdr_t l2wcb_head; / head of write buflist */
	/* in-flight list of log blocks */
	list_t l2wcb_abd_list;
	} l2arc_write_callback_t;

	struct arc_buf_hdr {
	/* protected by hash lock */
	dva_t b_dva;
	uint64_t b_birth;

	arc_buf_contents_t b_type;
	uint8_t b_complevel;
	uint8_t b_reserved1; /* used for 4 byte alignment */
	uint16_t b_reserved2; /* used for 4 byte alignment */
	arc_buf_hdr_t *b_hash_next;
	arc_flags_t b_flags;

	/*
	* This field stores the size of the data buffer after
	* compression, and is set in the arc's zio completion handlers.
	* It is in units of SPA_MINBLOCKSIZE (e.g. 1 == 512 bytes).
	*
	* While the block pointers can store up to 32MB in their psize
	* field, we can only store up to 32MB minus 512B. This is due
	* to the bp using a bias of 1, whereas we use a bias of 0 (i.e.
	* a field of zeros represents 512B in the bp). We can't use a
	* bias of 1 since we need to reserve a psize of zero, here, to
	* represent holes and embedded blocks.
	*
	* This isn't a problem in practice, since the maximum size of a
	* buffer is limited to 16MB, so we never need to store 32MB in
	* this field. Even in the upstream illumos code base, the
	* maximum size of a buffer is limited to 16MB.
	*/
	uint16_t b_psize;

	/*
	* This field stores the size of the data buffer before
	* compression, and cannot change once set. It is in units
	* of SPA_MINBLOCKSIZE (e.g. 2 == 1024 bytes)
	*/
	uint16_t b_lsize; /* immutable */
	uint64_t b_spa; /* immutable */

	/* L2ARC fields. Undefined when not in L2ARC. */
	l2arc_buf_hdr_t b_l2hdr;
	/* L1ARC fields. Undefined when in l2arc_only state */
	l1arc_buf_hdr_t b_l1hdr;
	/*
	* Encryption parameters. Defined only when ARC_FLAG_ENCRYPTED
	* is set and the L1 header exists.
	*/
	arc_buf_hdr_crypt_t b_crypt_hdr;
	};

	typedef struct arc_stats {
	kstat_named_t arcstat_hits;
	kstat_named_t arcstat_misses;
	kstat_named_t arcstat_demand_data_hits;
	kstat_named_t arcstat_demand_data_misses;
	kstat_named_t arcstat_demand_metadata_hits;
	kstat_named_t arcstat_demand_metadata_misses;
	kstat_named_t arcstat_prefetch_data_hits;
	kstat_named_t arcstat_prefetch_data_misses;
	kstat_named_t arcstat_prefetch_metadata_hits;
	kstat_named_t arcstat_prefetch_metadata_misses;
	kstat_named_t arcstat_mru_hits;
	kstat_named_t arcstat_mru_ghost_hits;
	kstat_named_t arcstat_mfu_hits;
	kstat_named_t arcstat_mfu_ghost_hits;
	kstat_named_t arcstat_deleted;
	/*
	* Number of buffers that could not be evicted because the hash lock
	* was held by another thread. The lock may not necessarily be held
	* by something using the same buffer, since hash locks are shared
	* by multiple buffers.
	*/
	kstat_named_t arcstat_mutex_miss;
	/*
	* Number of buffers skipped when updating the access state due to the
	* header having already been released after acquiring the hash lock.
	*/
	kstat_named_t arcstat_access_skip;
	/*
	* Number of buffers skipped because they have I/O in progress, are
	* indirect prefetch buffers that have not lived long enough, or are
	* not from the spa we're trying to evict from.
	*/
	kstat_named_t arcstat_evict_skip;
	/*
	* Number of times arc_evict_state() was unable to evict enough
	* buffers to reach its target amount.
	*/
	kstat_named_t arcstat_evict_not_enough;
	kstat_named_t arcstat_evict_l2_cached;
	kstat_named_t arcstat_evict_l2_eligible;
	kstat_named_t arcstat_evict_l2_eligible_mfu;
	kstat_named_t arcstat_evict_l2_eligible_mru;
	kstat_named_t arcstat_evict_l2_ineligible;
	kstat_named_t arcstat_evict_l2_skip;
	kstat_named_t arcstat_hash_elements;
	kstat_named_t arcstat_hash_elements_max;
	kstat_named_t arcstat_hash_collisions;
	kstat_named_t arcstat_hash_chains;
	kstat_named_t arcstat_hash_chain_max;
	kstat_named_t arcstat_p;
	kstat_named_t arcstat_c;
	kstat_named_t arcstat_c_min;
	kstat_named_t arcstat_c_max;
	kstat_named_t arcstat_size;
	/*
	* Number of compressed bytes stored in the arc_buf_hdr_t's b_pabd.
	* Note that the compressed bytes may match the uncompressed bytes
	* if the block is either not compressed or compressed arc is disabled.
	*/
	kstat_named_t arcstat_compressed_size;
	/*
	* Uncompressed size of the data stored in b_pabd. If compressed
	* arc is disabled then this value will be identical to the stat
	* above.
	*/
	kstat_named_t arcstat_uncompressed_size;
	/*
	* Number of bytes stored in all the arc_buf_t's. This is classified
	* as "overhead" since this data is typically short-lived and will
	* be evicted from the arc when it becomes unreferenced unless the
	* zfs_keep_uncompressed_metadata or zfs_keep_uncompressed_level
	* values have been set (see comment in dbuf.c for more information).
	*/
	kstat_named_t arcstat_overhead_size;
	/*
	* Number of bytes consumed by internal ARC structures necessary
	* for tracking purposes; these structures are not actually
	* backed by ARC buffers. This includes arc_buf_hdr_t structures
	* (allocated via arc_buf_hdr_t_full and arc_buf_hdr_t_l2only
	* caches), and arc_buf_t structures (allocated via arc_buf_t
	* cache).
	*/
	kstat_named_t arcstat_hdr_size;
	/*
	* Number of bytes consumed by ARC buffers of type equal to
	* ARC_BUFC_DATA. This is generally consumed by buffers backing
	* on disk user data (e.g. plain file contents).
	*/
	kstat_named_t arcstat_data_size;
	/*
	* Number of bytes consumed by ARC buffers of type equal to
	* ARC_BUFC_METADATA. This is generally consumed by buffers
	* backing on disk data that is used for internal ZFS
	* structures (e.g. ZAP, dnode, indirect blocks, etc).
	*/
	kstat_named_t arcstat_metadata_size;
	/*
	* Number of bytes consumed by dmu_buf_impl_t objects.
	*/
	kstat_named_t arcstat_dbuf_size;
	/*
	* Number of bytes consumed by dnode_t objects.
	*/
	kstat_named_t arcstat_dnode_size;
	/*
	* Number of bytes consumed by bonus buffers.
	*/
	kstat_named_t arcstat_bonus_size;
	#if defined(COMPAT_FREEBSD11)
	/*
	* Sum of the previous three counters, provided for compatibility.
	*/
	kstat_named_t arcstat_other_size;
	#endif

	/*
	* Total number of bytes consumed by ARC buffers residing in the
	* arc_anon state. This includes all buffers in the arc_anon
	* state; e.g. data, metadata, evictable, and unevictable buffers
	* are all included in this value.
	*/
	kstat_named_t arcstat_anon_size;
	/*
	* Number of bytes consumed by ARC buffers that meet the
	* following criteria: backing buffers of type ARC_BUFC_DATA,
	* residing in the arc_anon state, and are eligible for eviction
	* (e.g. have no outstanding holds on the buffer).
	*/
	kstat_named_t arcstat_anon_evictable_data;
	/*
	* Number of bytes consumed by ARC buffers that meet the
	* following criteria: backing buffers of type ARC_BUFC_METADATA,
	* residing in the arc_anon state, and are eligible for eviction
	* (e.g. have no outstanding holds on the buffer).
	*/
	kstat_named_t arcstat_anon_evictable_metadata;
	/*
	* Total number of bytes consumed by ARC buffers residing in the
	* arc_mru state. This includes all buffers in the arc_mru
	* state; e.g. data, metadata, evictable, and unevictable buffers
	* are all included in this value.
	*/
	kstat_named_t arcstat_mru_size;
	/*
	* Number of bytes consumed by ARC buffers that meet the
	* following criteria: backing buffers of type ARC_BUFC_DATA,
	* residing in the arc_mru state, and are eligible for eviction
	* (e.g. have no outstanding holds on the buffer).
	*/
	kstat_named_t arcstat_mru_evictable_data;
	/*
	* Number of bytes consumed by ARC buffers that meet the
	* following criteria: backing buffers of type ARC_BUFC_METADATA,
	* residing in the arc_mru state, and are eligible for eviction
	* (e.g. have no outstanding holds on the buffer).
	*/
	kstat_named_t arcstat_mru_evictable_metadata;
	/*
	* Total number of bytes that would have been consumed by ARC
	* buffers in the arc_mru_ghost state. The key thing to note
	* here, is the fact that this size doesn't actually indicate
	* RAM consumption. The ghost lists only consist of headers and
	* don't actually have ARC buffers linked off of these headers.
	* Thus, if the headers had associated ARC buffers, these
	* buffers would have consumed this number of bytes.
	*/
	kstat_named_t arcstat_mru_ghost_size;
	/*
	* Number of bytes that would have been consumed by ARC
	* buffers that are eligible for eviction, of type
	* ARC_BUFC_DATA, and linked off the arc_mru_ghost state.
	*/
	kstat_named_t arcstat_mru_ghost_evictable_data;
	/*
	* Number of bytes that would have been consumed by ARC
	* buffers that are eligible for eviction, of type
	* ARC_BUFC_METADATA, and linked off the arc_mru_ghost state.
	*/
	kstat_named_t arcstat_mru_ghost_evictable_metadata;
	/*
	* Total number of bytes consumed by ARC buffers residing in the
	* arc_mfu state. This includes all buffers in the arc_mfu
	* state; e.g. data, metadata, evictable, and unevictable buffers
	* are all included in this value.
	*/
	kstat_named_t arcstat_mfu_size;
	/*
	* Number of bytes consumed by ARC buffers that are eligible for
	* eviction, of type ARC_BUFC_DATA, and reside in the arc_mfu
	* state.
	*/
	kstat_named_t arcstat_mfu_evictable_data;
	/*
	* Number of bytes consumed by ARC buffers that are eligible for
	* eviction, of type ARC_BUFC_METADATA, and reside in the
	* arc_mfu state.
	*/
	kstat_named_t arcstat_mfu_evictable_metadata;
	/*
	* Total number of bytes that would have been consumed by ARC
	* buffers in the arc_mfu_ghost state. See the comment above
	* arcstat_mru_ghost_size for more details.
	*/
	kstat_named_t arcstat_mfu_ghost_size;
	/*
	* Number of bytes that would have been consumed by ARC
	* buffers that are eligible for eviction, of type
	* ARC_BUFC_DATA, and linked off the arc_mfu_ghost state.
	*/
	kstat_named_t arcstat_mfu_ghost_evictable_data;
	/*
	* Number of bytes that would have been consumed by ARC
	* buffers that are eligible for eviction, of type
	* ARC_BUFC_METADATA, and linked off the arc_mru_ghost state.
	*/
	kstat_named_t arcstat_mfu_ghost_evictable_metadata;
	kstat_named_t arcstat_l2_hits;
	kstat_named_t arcstat_l2_misses;
	/*
	* Allocated size (in bytes) of L2ARC cached buffers by ARC state.
	*/
	kstat_named_t arcstat_l2_prefetch_asize;
	kstat_named_t arcstat_l2_mru_asize;
	kstat_named_t arcstat_l2_mfu_asize;
	/*
	* Allocated size (in bytes) of L2ARC cached buffers by buffer content
	* type.
	*/
	kstat_named_t arcstat_l2_bufc_data_asize;
	kstat_named_t arcstat_l2_bufc_metadata_asize;
	kstat_named_t arcstat_l2_feeds;
	kstat_named_t arcstat_l2_rw_clash;
	kstat_named_t arcstat_l2_read_bytes;
	kstat_named_t arcstat_l2_write_bytes;
	kstat_named_t arcstat_l2_writes_sent;
	kstat_named_t arcstat_l2_writes_done;
	kstat_named_t arcstat_l2_writes_error;
	kstat_named_t arcstat_l2_writes_lock_retry;
	kstat_named_t arcstat_l2_evict_lock_retry;
	kstat_named_t arcstat_l2_evict_reading;
	kstat_named_t arcstat_l2_evict_l1cached;
	kstat_named_t arcstat_l2_free_on_write;
	kstat_named_t arcstat_l2_abort_lowmem;
	kstat_named_t arcstat_l2_cksum_bad;
	kstat_named_t arcstat_l2_io_error;
	kstat_named_t arcstat_l2_lsize;
	kstat_named_t arcstat_l2_psize;
	kstat_named_t arcstat_l2_hdr_size;
	/*
	* Number of L2ARC log blocks written. These are used for restoring the
	* L2ARC. Updated during writing of L2ARC log blocks.
	*/
	kstat_named_t arcstat_l2_log_blk_writes;
	/*
	* Moving average of the aligned size of the L2ARC log blocks, in
	* bytes. Updated during L2ARC rebuild and during writing of L2ARC
	* log blocks.
	*/
	kstat_named_t arcstat_l2_log_blk_avg_asize;
	/* Aligned size of L2ARC log blocks on L2ARC devices. */
	kstat_named_t arcstat_l2_log_blk_asize;
	/* Number of L2ARC log blocks present on L2ARC devices. */
	kstat_named_t arcstat_l2_log_blk_count;
	/*
	* Moving average of the aligned size of L2ARC restored data, in bytes,
	* to the aligned size of their metadata in L2ARC, in bytes.
	* Updated during L2ARC rebuild and during writing of L2ARC log blocks.
	*/
	kstat_named_t arcstat_l2_data_to_meta_ratio;
	/*
	* Number of times the L2ARC rebuild was successful for an L2ARC device.
	*/
	kstat_named_t arcstat_l2_rebuild_success;
	/*
	* Number of times the L2ARC rebuild failed because the device header
	* was in an unsupported format or corrupted.
	*/
	kstat_named_t arcstat_l2_rebuild_abort_unsupported;
	/*
	* Number of times the L2ARC rebuild failed because of IO errors
	* while reading a log block.
	*/
	kstat_named_t arcstat_l2_rebuild_abort_io_errors;
	/*
	* Number of times the L2ARC rebuild failed because of IO errors when
	* reading the device header.
	*/
	kstat_named_t arcstat_l2_rebuild_abort_dh_errors;
	/*
	* Number of L2ARC log blocks which failed to be restored due to
	* checksum errors.
	*/
	kstat_named_t arcstat_l2_rebuild_abort_cksum_lb_errors;
	/*
	* Number of times the L2ARC rebuild was aborted due to low system
	* memory.
	*/
	kstat_named_t arcstat_l2_rebuild_abort_lowmem;
	/* Logical size of L2ARC restored data, in bytes. */
	kstat_named_t arcstat_l2_rebuild_size;
	/* Aligned size of L2ARC restored data, in bytes. */
	kstat_named_t arcstat_l2_rebuild_asize;
	/*
	* Number of L2ARC log entries (buffers) that were successfully
	* restored in ARC.
	*/
	kstat_named_t arcstat_l2_rebuild_bufs;
	/*
	* Number of L2ARC log entries (buffers) already cached in ARC. These
	* were not restored again.
	*/
	kstat_named_t arcstat_l2_rebuild_bufs_precached;
	/*
	* Number of L2ARC log blocks that were restored successfully. Each
	* log block may hold up to L2ARC_LOG_BLK_MAX_ENTRIES buffers.
	*/
	kstat_named_t arcstat_l2_rebuild_log_blks;
	kstat_named_t arcstat_memory_throttle_count;
	kstat_named_t arcstat_memory_direct_count;
	kstat_named_t arcstat_memory_indirect_count;
	kstat_named_t arcstat_memory_all_bytes;
	kstat_named_t arcstat_memory_free_bytes;
	kstat_named_t arcstat_memory_available_bytes;
	kstat_named_t arcstat_no_grow;
	kstat_named_t arcstat_tempreserve;
	kstat_named_t arcstat_loaned_bytes;
	kstat_named_t arcstat_prune;
	kstat_named_t arcstat_meta_used;
	kstat_named_t arcstat_meta_limit;
	kstat_named_t arcstat_dnode_limit;
	kstat_named_t arcstat_meta_max;
	kstat_named_t arcstat_meta_min;
	kstat_named_t arcstat_async_upgrade_sync;
	kstat_named_t arcstat_demand_hit_predictive_prefetch;
	kstat_named_t arcstat_demand_hit_prescient_prefetch;
	kstat_named_t arcstat_need_free;
	kstat_named_t arcstat_sys_free;
	kstat_named_t arcstat_raw_size;
	kstat_named_t arcstat_cached_only_in_progress;
	kstat_named_t arcstat_abd_chunk_waste_size;
	} arc_stats_t;

	typedef struct arc_sums {
	wmsum_t arcstat_hits;
	wmsum_t arcstat_misses;
	wmsum_t arcstat_demand_data_hits;
	wmsum_t arcstat_demand_data_misses;
	wmsum_t arcstat_demand_metadata_hits;
	wmsum_t arcstat_demand_metadata_misses;
	wmsum_t arcstat_prefetch_data_hits;
	wmsum_t arcstat_prefetch_data_misses;
	wmsum_t arcstat_prefetch_metadata_hits;
	wmsum_t arcstat_prefetch_metadata_misses;
	wmsum_t arcstat_mru_hits;
	wmsum_t arcstat_mru_ghost_hits;
	wmsum_t arcstat_mfu_hits;
	wmsum_t arcstat_mfu_ghost_hits;
	wmsum_t arcstat_deleted;
	wmsum_t arcstat_mutex_miss;
	wmsum_t arcstat_access_skip;
	wmsum_t arcstat_evict_skip;
	wmsum_t arcstat_evict_not_enough;
	wmsum_t arcstat_evict_l2_cached;
	wmsum_t arcstat_evict_l2_eligible;
	wmsum_t arcstat_evict_l2_eligible_mfu;
	wmsum_t arcstat_evict_l2_eligible_mru;
	wmsum_t arcstat_evict_l2_ineligible;
	wmsum_t arcstat_evict_l2_skip;
	wmsum_t arcstat_hash_collisions;
	wmsum_t arcstat_hash_chains;
	aggsum_t arcstat_size;
	wmsum_t arcstat_compressed_size;
	wmsum_t arcstat_uncompressed_size;
	wmsum_t arcstat_overhead_size;
	wmsum_t arcstat_hdr_size;
	wmsum_t arcstat_data_size;
	wmsum_t arcstat_metadata_size;
	wmsum_t arcstat_dbuf_size;
	aggsum_t arcstat_dnode_size;
	wmsum_t arcstat_bonus_size;
	wmsum_t arcstat_l2_hits;
	wmsum_t arcstat_l2_misses;
	wmsum_t arcstat_l2_prefetch_asize;
	wmsum_t arcstat_l2_mru_asize;
	wmsum_t arcstat_l2_mfu_asize;
	wmsum_t arcstat_l2_bufc_data_asize;
	wmsum_t arcstat_l2_bufc_metadata_asize;
	wmsum_t arcstat_l2_feeds;
	wmsum_t arcstat_l2_rw_clash;
	wmsum_t arcstat_l2_read_bytes;
	wmsum_t arcstat_l2_write_bytes;
	wmsum_t arcstat_l2_writes_sent;
	wmsum_t arcstat_l2_writes_done;
	wmsum_t arcstat_l2_writes_error;
	wmsum_t arcstat_l2_writes_lock_retry;
	wmsum_t arcstat_l2_evict_lock_retry;
	wmsum_t arcstat_l2_evict_reading;
	wmsum_t arcstat_l2_evict_l1cached;
	wmsum_t arcstat_l2_free_on_write;
	wmsum_t arcstat_l2_abort_lowmem;
	wmsum_t arcstat_l2_cksum_bad;
	wmsum_t arcstat_l2_io_error;
	wmsum_t arcstat_l2_lsize;
	wmsum_t arcstat_l2_psize;
	aggsum_t arcstat_l2_hdr_size;
	wmsum_t arcstat_l2_log_blk_writes;
	wmsum_t arcstat_l2_log_blk_asize;
	wmsum_t arcstat_l2_log_blk_count;
	wmsum_t arcstat_l2_rebuild_success;
	wmsum_t arcstat_l2_rebuild_abort_unsupported;
	wmsum_t arcstat_l2_rebuild_abort_io_errors;
	wmsum_t arcstat_l2_rebuild_abort_dh_errors;
	wmsum_t arcstat_l2_rebuild_abort_cksum_lb_errors;
	wmsum_t arcstat_l2_rebuild_abort_lowmem;
	wmsum_t arcstat_l2_rebuild_size;
	wmsum_t arcstat_l2_rebuild_asize;
	wmsum_t arcstat_l2_rebuild_bufs;
	wmsum_t arcstat_l2_rebuild_bufs_precached;
	wmsum_t arcstat_l2_rebuild_log_blks;
	wmsum_t arcstat_memory_throttle_count;
	wmsum_t arcstat_memory_direct_count;
	wmsum_t arcstat_memory_indirect_count;
	wmsum_t arcstat_prune;
	aggsum_t arcstat_meta_used;
	wmsum_t arcstat_async_upgrade_sync;
	wmsum_t arcstat_demand_hit_predictive_prefetch;
	wmsum_t arcstat_demand_hit_prescient_prefetch;
	wmsum_t arcstat_raw_size;
	wmsum_t arcstat_cached_only_in_progress;
	wmsum_t arcstat_abd_chunk_waste_size;
	} arc_sums_t;

	typedef struct arc_evict_waiter {
	list_node_t aew_node;
	kcondvar_t aew_cv;
	uint64_t aew_count;
	} arc_evict_waiter_t;

	#define ARCSTAT(stat) (arc_stats.stat.value.ui64)

	#define ARCSTAT_INCR(stat, val) \
	wmsum_add(&arc_sums.stat, (val))

	#define ARCSTAT_BUMP(stat) ARCSTAT_INCR(stat, 1)
	#define ARCSTAT_BUMPDOWN(stat) ARCSTAT_INCR(stat, -1)

	#define arc_no_grow ARCSTAT(arcstat_no_grow) /* do not grow cache size */
	#define arc_p ARCSTAT(arcstat_p) /* target size of MRU */
	#define arc_c ARCSTAT(arcstat_c) /* target size of cache */
	#define arc_c_min ARCSTAT(arcstat_c_min) /* min target cache size */
	#define arc_c_max ARCSTAT(arcstat_c_max) /* max target cache size */
	#define arc_sys_free ARCSTAT(arcstat_sys_free) /* target system free bytes */

	#define arc_anon (&ARC_anon)
	#define arc_mru (&ARC_mru)
	#define arc_mru_ghost (&ARC_mru_ghost)
	#define arc_mfu (&ARC_mfu)
	#define arc_mfu_ghost (&ARC_mfu_ghost)
	#define arc_l2c_only (&ARC_l2c_only)

	extern taskq_t *arc_prune_taskq;
	extern arc_stats_t arc_stats;
	extern arc_sums_t arc_sums;
	extern hrtime_t arc_growtime;
	extern boolean_t arc_warm;
	-extern int arc_grow_retry;
	-extern int arc_no_grow_shift;
	-extern int arc_shrink_shift;
	+extern uint_t arc_grow_retry;
	+extern uint_t arc_no_grow_shift;
	+extern uint_t arc_shrink_shift;
	extern kmutex_t arc_prune_mtx;
	extern list_t arc_prune_list;
	extern arc_state_t ARC_mfu;
	extern arc_state_t ARC_mru;
	extern uint_t zfs_arc_pc_percent;
	-extern int arc_lotsfree_percent;
	+extern uint_t arc_lotsfree_percent;
	extern unsigned long zfs_arc_min;
	extern unsigned long zfs_arc_max;

	extern void arc_reduce_target_size(int64_t to_free);
	extern boolean_t arc_reclaim_needed(void);
	extern void arc_kmem_reap_soon(void);
	extern void arc_wait_for_eviction(uint64_t, boolean_t);

	extern void arc_lowmem_init(void);
	extern void arc_lowmem_fini(void);
	-extern void arc_prune_async(int64_t);
	+extern void arc_prune_async(uint64_t);
	extern int arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg);
	extern uint64_t arc_free_memory(void);
	extern int64_t arc_available_memory(void);
	extern void arc_tuning_update(boolean_t);
	extern void arc_register_hotplug(void);
	extern void arc_unregister_hotplug(void);

	extern int param_set_arc_long(ZFS_MODULE_PARAM_ARGS);
	extern int param_set_arc_int(ZFS_MODULE_PARAM_ARGS);
	extern int param_set_arc_min(ZFS_MODULE_PARAM_ARGS);
	extern int param_set_arc_max(ZFS_MODULE_PARAM_ARGS);

	/* used in zdb.c */
	boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t *lbp);

	/* used in vdev_trim.c */
	void l2arc_dev_hdr_update(l2arc_dev_t *dev);
	l2arc_dev_t l2arc_vdev_get(vdev_t vd);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ARC_IMPL_H */
	diff --git a/sys/contrib/openzfs/include/sys/dbuf.h b/sys/contrib/openzfs/include/sys/dbuf.h
	index a3aaac5e87ce..06489ea84bf8 100644
	--- a/sys/contrib/openzfs/include/sys/dbuf.h
	+++ b/sys/contrib/openzfs/include/sys/dbuf.h
	@@ -1,496 +1,498 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	*/

	#ifndef _SYS_DBUF_H
	#define _SYS_DBUF_H

	#include <sys/dmu.h>
	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/zio.h>
	#include <sys/arc.h>
	#include <sys/zfs_context.h>
	#include <sys/zfs_refcount.h>
	#include <sys/zrlock.h>
	#include <sys/multilist.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	#define IN_DMU_SYNC 2

	/*
	* define flags for dbuf_read
	*/

	#define DB_RF_MUST_SUCCEED (1 << 0)
	#define DB_RF_CANFAIL (1 << 1)
	#define DB_RF_HAVESTRUCT (1 << 2)
	#define DB_RF_NOPREFETCH (1 << 3)
	#define DB_RF_NEVERWAIT (1 << 4)
	#define DB_RF_CACHED (1 << 5)
	#define DB_RF_NO_DECRYPT (1 << 6)

	/*
	* The simplified state transition diagram for dbufs looks like:
	*
	* +----> READ ----+
	* \| \|
	* \| V
	* (alloc)-->UNCACHED CACHED-->EVICTING-->(free)
	* \| ^ ^
	* \| \| \|
	* +----> FILL ----+ \|
	* \| \|
	* \| \|
	* +--------> NOFILL -------+
	*
	* DB_SEARCH is an invalid state for a dbuf. It is used by dbuf_free_range
	* to find all dbufs in a range of a dnode and must be less than any other
	* dbuf_states_t (see comment on dn_dbufs in dnode.h).
	*/
	typedef enum dbuf_states {
	DB_SEARCH = -1,
	DB_UNCACHED,
	DB_FILL,
	DB_NOFILL,
	DB_READ,
	DB_CACHED,
	DB_EVICTING
	} dbuf_states_t;

	typedef enum dbuf_cached_state {
	DB_NO_CACHE = -1,
	DB_DBUF_CACHE,
	DB_DBUF_METADATA_CACHE,
	DB_CACHE_MAX
	} dbuf_cached_state_t;

	struct dnode;
	struct dmu_tx;

	/*
	* level = 0 means the user data
	* level = 1 means the single indirect block
	* etc.
	*/

	struct dmu_buf_impl;

	typedef enum override_states {
	DR_NOT_OVERRIDDEN,
	DR_IN_DMU_SYNC,
	DR_OVERRIDDEN
	} override_states_t;

	typedef enum db_lock_type {
	DLT_NONE,
	DLT_PARENT,
	DLT_OBJSET
	} db_lock_type_t;

	typedef struct dbuf_dirty_record {
	/* link on our parents dirty list */
	list_node_t dr_dirty_node;

	/* transaction group this data will sync in */
	uint64_t dr_txg;

	/* zio of outstanding write IO */
	zio_t *dr_zio;

	/* pointer back to our dbuf */
	struct dmu_buf_impl *dr_dbuf;

	/* list link for dbuf dirty records */
	list_node_t dr_dbuf_node;

	/*
	* The dnode we are part of. Note that the dnode can not be moved or
	* evicted due to the hold that's added by dnode_setdirty() or
	* dmu_objset_sync_dnodes(), and released by dnode_rele_task() or
	* userquota_updates_task(). This hold is necessary for
	* dirty_lightweight_leaf-type dirty records, which don't have a hold
	* on a dbuf.
	*/
	dnode_t *dr_dnode;

	/* pointer to parent dirty record */
	struct dbuf_dirty_record *dr_parent;

	/* How much space was changed to dsl_pool_dirty_space() for this? */
	unsigned int dr_accounted;

	/* A copy of the bp that points to us */
	blkptr_t dr_bp_copy;

	union dirty_types {
	struct dirty_indirect {

	/* protect access to list */
	kmutex_t dr_mtx;

	/* Our list of dirty children */
	list_t dr_children;
	} di;
	struct dirty_leaf {

	/*
	* dr_data is set when we dirty the buffer
	* so that we can retain the pointer even if it
	* gets COW'd in a subsequent transaction group.
	*/
	arc_buf_t *dr_data;
	blkptr_t dr_overridden_by;
	override_states_t dr_override_state;
	uint8_t dr_copies;
	boolean_t dr_nopwrite;
	boolean_t dr_has_raw_params;

	/*
	* If dr_has_raw_params is set, the following crypt
	* params will be set on the BP that's written.
	*/
	boolean_t dr_byteorder;
	uint8_t dr_salt[ZIO_DATA_SALT_LEN];
	uint8_t dr_iv[ZIO_DATA_IV_LEN];
	uint8_t dr_mac[ZIO_DATA_MAC_LEN];
	} dl;
	struct dirty_lightweight_leaf {
	/*
	* This dirty record refers to a leaf (level=0)
	* block, whose dbuf has not been instantiated for
	* performance reasons.
	*/
	uint64_t dr_blkid;
	abd_t *dr_abd;
	zio_prop_t dr_props;
	enum zio_flag dr_flags;
	} dll;
	} dt;
	} dbuf_dirty_record_t;

	typedef struct dmu_buf_impl {
	/*
	* The following members are immutable, with the exception of
	* db.db_data, which is protected by db_mtx.
	*/

	/* the publicly visible structure */
	dmu_buf_t db;

	/* the objset we belong to */
	struct objset *db_objset;

	/*
	* handle to safely access the dnode we belong to (NULL when evicted)
	*/
	struct dnode_handle *db_dnode_handle;

	/*
	* our parent buffer; if the dnode points to us directly,
	* db_parent == db_dnode_handle->dnh_dnode->dn_dbuf
	* only accessed by sync thread ???
	* (NULL when evicted)
	* May change from NULL to non-NULL under the protection of db_mtx
	* (see dbuf_check_blkptr())
	*/
	struct dmu_buf_impl *db_parent;

	/*
	* link for hash table of all dmu_buf_impl_t's
	*/
	struct dmu_buf_impl *db_hash_next;

	/*
	* Our link on the owner dnodes's dn_dbufs list.
	* Protected by its dn_dbufs_mtx. Should be on the same cache line
	* as db_level and db_blkid for the best avl_add() performance.
	*/
	avl_node_t db_link;

	/* our block number */
	uint64_t db_blkid;

	/*
	* Pointer to the blkptr_t which points to us. May be NULL if we
	* don't have one yet. (NULL when evicted)
	*/
	blkptr_t *db_blkptr;

	/*
	* Our indirection level. Data buffers have db_level==0.
	* Indirect buffers which point to data buffers have
	* db_level==1. etc. Buffers which contain dnodes have
	* db_level==0, since the dnodes are stored in a file.
	*/
	uint8_t db_level;

	/*
	* Protects db_buf's contents if they contain an indirect block or data
	* block of the meta-dnode. We use this lock to protect the structure of
	* the block tree. This means that when modifying this dbuf's data, we
	* grab its rwlock. When modifying its parent's data (including the
	* blkptr to this dbuf), we grab the parent's rwlock. The lock ordering
	* for this lock is:
	* 1) dn_struct_rwlock
	* 2) db_rwlock
	* We don't currently grab multiple dbufs' db_rwlocks at once.
	*/
	krwlock_t db_rwlock;

	/* buffer holding our data */
	arc_buf_t *db_buf;

	/* db_mtx protects the members below */
	kmutex_t db_mtx;

	/*
	* Current state of the buffer
	*/
	dbuf_states_t db_state;

	/*
	* Refcount accessed by dmu_buf_{hold,rele}.
	* If nonzero, the buffer can't be destroyed.
	* Protected by db_mtx.
	*/
	zfs_refcount_t db_holds;

	kcondvar_t db_changed;
	dbuf_dirty_record_t *db_data_pending;

	/* List of dirty records for the buffer sorted newest to oldest. */
	list_t db_dirty_records;

	/* Link in dbuf_cache or dbuf_metadata_cache */
	multilist_node_t db_cache_link;

	/* Tells us which dbuf cache this dbuf is in, if any */
	dbuf_cached_state_t db_caching_status;

	/* Data which is unique to data (leaf) blocks: */

	/* User callback information. */
	dmu_buf_user_t *db_user;

	/*
	* Evict user data as soon as the dirty and reference
	* counts are equal.
	*/
	uint8_t db_user_immediate_evict;

	/*
	* This block was freed while a read or write was
	* active.
	*/
	uint8_t db_freed_in_flight;

	/*
	* dnode_evict_dbufs() or dnode_evict_bonus() tried to
	* evict this dbuf, but couldn't due to outstanding
	* references. Evict once the refcount drops to 0.
	*/
	uint8_t db_pending_evict;

	uint8_t db_dirtycnt;
	} dmu_buf_impl_t;

	-#define DBUF_RWLOCKS 8192
	-#define DBUF_HASH_RWLOCK(h, idx) (&(h)->hash_rwlocks[(idx) & (DBUF_RWLOCKS-1)])
	+#define DBUF_HASH_MUTEX(h, idx) \
	+ (&(h)->hash_mutexes[(idx) & ((h)->hash_mutex_mask)])
	+
	typedef struct dbuf_hash_table {
	uint64_t hash_table_mask;
	+ uint64_t hash_mutex_mask;
	dmu_buf_impl_t **hash_table;
	- krwlock_t hash_rwlocks[DBUF_RWLOCKS] ____cacheline_aligned;
	+ kmutex_t *hash_mutexes;
	} dbuf_hash_table_t;

	typedef void (dbuf_prefetch_fn)(void , uint64_t, uint64_t, boolean_t);

	uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level,
	const uint64_t offset);

	void dbuf_create_bonus(struct dnode *dn);
	int dbuf_spill_set_blksz(dmu_buf_t db, uint64_t blksz, dmu_tx_t tx);

	void dbuf_rm_spill(struct dnode dn, dmu_tx_t tx);

	dmu_buf_impl_t dbuf_hold(struct dnode dn, uint64_t blkid, const void *tag);
	dmu_buf_impl_t dbuf_hold_level(struct dnode dn, int level, uint64_t blkid,
	const void *tag);
	int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid,
	boolean_t fail_sparse, boolean_t fail_uncached,
	const void tag, dmu_buf_impl_t *dbp);

	int dbuf_prefetch_impl(struct dnode *dn, int64_t level, uint64_t blkid,
	zio_priority_t prio, arc_flags_t aflags, dbuf_prefetch_fn cb,
	void *arg);
	int dbuf_prefetch(struct dnode *dn, int64_t level, uint64_t blkid,
	zio_priority_t prio, arc_flags_t aflags);

	void dbuf_add_ref(dmu_buf_impl_t db, const void tag);
	boolean_t dbuf_try_add_ref(dmu_buf_t db, objset_t os, uint64_t obj,
	uint64_t blkid, const void *tag);
	uint64_t dbuf_refcount(dmu_buf_impl_t *db);

	void dbuf_rele(dmu_buf_impl_t db, const void tag);
	void dbuf_rele_and_unlock(dmu_buf_impl_t db, const void tag,
	boolean_t evicting);

	dmu_buf_impl_t dbuf_find(struct objset os, uint64_t object, uint8_t level,
	uint64_t blkid);

	int dbuf_read(dmu_buf_impl_t db, zio_t zio, uint32_t flags);
	void dmu_buf_will_not_fill(dmu_buf_t db, dmu_tx_t tx);
	void dmu_buf_will_fill(dmu_buf_t db, dmu_tx_t tx);
	void dmu_buf_fill_done(dmu_buf_t db, dmu_tx_t tx);
	void dbuf_assign_arcbuf(dmu_buf_impl_t db, arc_buf_t buf, dmu_tx_t *tx);
	dbuf_dirty_record_t dbuf_dirty(dmu_buf_impl_t db, dmu_tx_t *tx);
	dbuf_dirty_record_t dbuf_dirty_lightweight(dnode_t dn, uint64_t blkid,
	dmu_tx_t *tx);
	arc_buf_t dbuf_loan_arcbuf(dmu_buf_impl_t db);
	void dmu_buf_write_embedded(dmu_buf_t dbuf, void data,
	bp_embedded_type_t etype, enum zio_compress comp,
	int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx);

	int dmu_lightweight_write_by_dnode(dnode_t dn, uint64_t offset, abd_t abd,
	const struct zio_prop zp, enum zio_flag flags, dmu_tx_t tx);

	void dmu_buf_redact(dmu_buf_t dbuf, dmu_tx_t tx);
	void dbuf_destroy(dmu_buf_impl_t *db);

	void dbuf_unoverride(dbuf_dirty_record_t *dr);
	void dbuf_sync_list(list_t list, int level, dmu_tx_t tx);
	void dbuf_release_bp(dmu_buf_impl_t *db);
	db_lock_type_t dmu_buf_lock_parent(dmu_buf_impl_t *db, krw_t rw,
	const void *tag);
	void dmu_buf_unlock_parent(dmu_buf_impl_t *db, db_lock_type_t type,
	const void *tag);

	void dbuf_free_range(struct dnode *dn, uint64_t start, uint64_t end,
	struct dmu_tx *);

	void dbuf_new_size(dmu_buf_impl_t db, int size, dmu_tx_t tx);

	void dbuf_stats_init(dbuf_hash_table_t *hash);
	void dbuf_stats_destroy(void);

	int dbuf_dnode_findbp(dnode_t *dn, uint64_t level, uint64_t blkid,
	blkptr_t bp, uint16_t datablkszsec, uint8_t *indblkshift);

	#define DB_DNODE(_db) ((_db)->db_dnode_handle->dnh_dnode)
	#define DB_DNODE_LOCK(_db) ((_db)->db_dnode_handle->dnh_zrlock)
	#define DB_DNODE_ENTER(_db) (zrl_add(&DB_DNODE_LOCK(_db)))
	#define DB_DNODE_EXIT(_db) (zrl_remove(&DB_DNODE_LOCK(_db)))
	#define DB_DNODE_HELD(_db) (!zrl_is_zero(&DB_DNODE_LOCK(_db)))

	void dbuf_init(void);
	void dbuf_fini(void);

	boolean_t dbuf_is_metadata(dmu_buf_impl_t *db);

	static inline dbuf_dirty_record_t *
	dbuf_find_dirty_lte(dmu_buf_impl_t *db, uint64_t txg)
	{
	dbuf_dirty_record_t *dr;

	for (dr = list_head(&db->db_dirty_records);
	dr != NULL && dr->dr_txg > txg;
	dr = list_next(&db->db_dirty_records, dr))
	continue;
	return (dr);
	}

	static inline dbuf_dirty_record_t *
	dbuf_find_dirty_eq(dmu_buf_impl_t *db, uint64_t txg)
	{
	dbuf_dirty_record_t *dr;

	dr = dbuf_find_dirty_lte(db, txg);
	if (dr && dr->dr_txg == txg)
	return (dr);
	return (NULL);
	}

	#define DBUF_GET_BUFC_TYPE(_db) \
	(dbuf_is_metadata(_db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)

	#define DBUF_IS_CACHEABLE(_db) \
	((_db)->db_objset->os_primary_cache == ZFS_CACHE_ALL \|\| \
	(dbuf_is_metadata(_db) && \
	((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))

	boolean_t dbuf_is_l2cacheable(dmu_buf_impl_t *db);

	#ifdef ZFS_DEBUG

	/*
	* There should be a ## between the string literal and fmt, to make it
	* clear that we're joining two strings together, but gcc does not
	* support that preprocessor token.
	*/
	#define dprintf_dbuf(dbuf, fmt, ...) do { \
	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
	char __db_buf[32]; \
	uint64_t __db_obj = (dbuf)->db.db_object; \
	if (__db_obj == DMU_META_DNODE_OBJECT) \
	(void) strlcpy(__db_buf, "mdn", sizeof (__db_buf)); \
	else \
	(void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
	(u_longlong_t)__db_obj); \
	dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \
	"obj=%s lvl=%u blkid=%lld " fmt, \
	__db_buf, (dbuf)->db_level, \
	(u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \
	} \
	} while (0)

	#define dprintf_dbuf_bp(db, bp, fmt, ...) do { \
	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \
	dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \
	kmem_free(__blkbuf, BP_SPRINTF_LEN); \
	} \
	} while (0)

	#define DBUF_VERIFY(db) dbuf_verify(db)

	#else

	#define dprintf_dbuf(db, fmt, ...)
	#define dprintf_dbuf_bp(db, bp, fmt, ...)
	#define DBUF_VERIFY(db)

	#endif


	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_DBUF_H */
	diff --git a/sys/contrib/openzfs/include/sys/dmu.h b/sys/contrib/openzfs/include/sys/dmu.h
	index 0a4827e5ec3e..d68700d371db 100644
	--- a/sys/contrib/openzfs/include/sys/dmu.h
	+++ b/sys/contrib/openzfs/include/sys/dmu.h
	@@ -1,1079 +1,1079 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	*/

	/* Portions Copyright 2010 Robert Milkowski */

	#ifndef _SYS_DMU_H
	#define _SYS_DMU_H

	/*
	* This file describes the interface that the DMU provides for its
	* consumers.
	*
	* The DMU also interacts with the SPA. That interface is described in
	* dmu_spa.h.
	*/

	#include <sys/zfs_context.h>
	#include <sys/inttypes.h>
	#include <sys/cred.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_priority.h>
	#include <sys/uio.h>
	#include <sys/zfs_file.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	struct page;
	struct vnode;
	struct spa;
	struct zilog;
	struct zio;
	struct blkptr;
	struct zap_cursor;
	struct dsl_dataset;
	struct dsl_pool;
	struct dnode;
	struct drr_begin;
	struct drr_end;
	struct zbookmark_phys;
	struct spa;
	struct nvlist;
	struct arc_buf;
	struct zio_prop;
	struct sa_handle;
	struct dsl_crypto_params;
	struct locked_range;

	typedef struct objset objset_t;
	typedef struct dmu_tx dmu_tx_t;
	typedef struct dsl_dir dsl_dir_t;
	typedef struct dnode dnode_t;

	typedef enum dmu_object_byteswap {
	DMU_BSWAP_UINT8,
	DMU_BSWAP_UINT16,
	DMU_BSWAP_UINT32,
	DMU_BSWAP_UINT64,
	DMU_BSWAP_ZAP,
	DMU_BSWAP_DNODE,
	DMU_BSWAP_OBJSET,
	DMU_BSWAP_ZNODE,
	DMU_BSWAP_OLDACL,
	DMU_BSWAP_ACL,
	/*
	* Allocating a new byteswap type number makes the on-disk format
	* incompatible with any other format that uses the same number.
	*
	* Data can usually be structured to work with one of the
	* DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types.
	*/
	DMU_BSWAP_NUMFUNCS
	} dmu_object_byteswap_t;

	#define DMU_OT_NEWTYPE 0x80
	#define DMU_OT_METADATA 0x40
	#define DMU_OT_ENCRYPTED 0x20
	#define DMU_OT_BYTESWAP_MASK 0x1f

	/*
	* Defines a uint8_t object type. Object types specify if the data
	* in the object is metadata (boolean) and how to byteswap the data
	* (dmu_object_byteswap_t). All of the types created by this method
	* are cached in the dbuf metadata cache.
	*/
	#define DMU_OT(byteswap, metadata, encrypted) \
	(DMU_OT_NEWTYPE \| \
	((metadata) ? DMU_OT_METADATA : 0) \| \
	((encrypted) ? DMU_OT_ENCRYPTED : 0) \| \
	((byteswap) & DMU_OT_BYTESWAP_MASK))

	#define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \
	((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \
	(ot) < DMU_OT_NUMTYPES)

	#define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
	B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache)

	/*
	* MDB doesn't have dmu_ot; it defines these macros itself.
	*/
	#ifndef ZFS_MDB
	#define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata)
	#define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt)
	#define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap)
	#endif

	#define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
	(((ot) & DMU_OT_METADATA) != 0) : \
	DMU_OT_IS_METADATA_IMPL(ot))

	#define DMU_OT_IS_DDT(ot) \
	((ot) == DMU_OT_DDT_ZAP)

	/* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */
	#define DMU_OT_IS_FILE(ot) \
	((ot) == DMU_OT_PLAIN_FILE_CONTENTS \|\| (ot) == DMU_OT_UINT64_OTHER)

	#define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
	(((ot) & DMU_OT_ENCRYPTED) != 0) : \
	DMU_OT_IS_ENCRYPTED_IMPL(ot))

	/*
	* These object types use bp_fill != 1 for their L0 bp's. Therefore they can't
	* have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill
	* is repurposed for embedded BPs.
	*/
	#define DMU_OT_HAS_FILL(ot) \
	((ot) == DMU_OT_DNODE \|\| (ot) == DMU_OT_OBJSET)

	#define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \
	((ot) & DMU_OT_BYTESWAP_MASK) : \
	DMU_OT_BYTESWAP_IMPL(ot))

	typedef enum dmu_object_type {
	DMU_OT_NONE,
	/* general: */
	DMU_OT_OBJECT_DIRECTORY, /* ZAP */
	DMU_OT_OBJECT_ARRAY, /* UINT64 */
	DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */
	DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */
	DMU_OT_BPOBJ, /* UINT64 */
	DMU_OT_BPOBJ_HDR, /* UINT64 */
	/* spa: */
	DMU_OT_SPACE_MAP_HEADER, /* UINT64 */
	DMU_OT_SPACE_MAP, /* UINT64 */
	/* zil: */
	DMU_OT_INTENT_LOG, /* UINT64 */
	/* dmu: */
	DMU_OT_DNODE, /* DNODE */
	DMU_OT_OBJSET, /* OBJSET */
	/* dsl: */
	DMU_OT_DSL_DIR, /* UINT64 */
	DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */
	DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */
	DMU_OT_DSL_PROPS, /* ZAP */
	DMU_OT_DSL_DATASET, /* UINT64 */
	/* zpl: */
	DMU_OT_ZNODE, /* ZNODE */
	DMU_OT_OLDACL, /* Old ACL */
	DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */
	DMU_OT_DIRECTORY_CONTENTS, /* ZAP */
	DMU_OT_MASTER_NODE, /* ZAP */
	DMU_OT_UNLINKED_SET, /* ZAP */
	/* zvol: */
	DMU_OT_ZVOL, /* UINT8 */
	DMU_OT_ZVOL_PROP, /* ZAP */
	/* other; for testing only! */
	DMU_OT_PLAIN_OTHER, /* UINT8 */
	DMU_OT_UINT64_OTHER, /* UINT64 */
	DMU_OT_ZAP_OTHER, /* ZAP */
	/* new object types: */
	DMU_OT_ERROR_LOG, /* ZAP */
	DMU_OT_SPA_HISTORY, /* UINT8 */
	DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */
	DMU_OT_POOL_PROPS, /* ZAP */
	DMU_OT_DSL_PERMS, /* ZAP */
	DMU_OT_ACL, /* ACL */
	DMU_OT_SYSACL, /* SYSACL */
	DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */
	DMU_OT_FUID_SIZE, /* FUID table size UINT64 */
	DMU_OT_NEXT_CLONES, /* ZAP */
	DMU_OT_SCAN_QUEUE, /* ZAP */
	DMU_OT_USERGROUP_USED, /* ZAP */
	DMU_OT_USERGROUP_QUOTA, /* ZAP */
	DMU_OT_USERREFS, /* ZAP */
	DMU_OT_DDT_ZAP, /* ZAP */
	DMU_OT_DDT_STATS, /* ZAP */
	DMU_OT_SA, /* System attr */
	DMU_OT_SA_MASTER_NODE, /* ZAP */
	DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */
	DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */
	DMU_OT_SCAN_XLATE, /* ZAP */
	DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */
	DMU_OT_DEADLIST, /* ZAP */
	DMU_OT_DEADLIST_HDR, /* UINT64 */
	DMU_OT_DSL_CLONES, /* ZAP */
	DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */
	/*
	* Do not allocate new object types here. Doing so makes the on-disk
	* format incompatible with any other format that uses the same object
	* type number.
	*
	* When creating an object which does not have one of the above types
	* use the DMU_OTN_* type with the correct byteswap and metadata
	* values.
	*
	* The DMU_OTN_* types do not have entries in the dmu_ot table,
	* use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead
	* of indexing into dmu_ot directly (this works for both DMU_OT_* types
	* and DMU_OTN_* types).
	*/
	DMU_OT_NUMTYPES,

	/*
	* Names for valid types declared with DMU_OT().
	*/
	DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE),
	DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE),
	DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE),
	DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE),
	DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE),
	DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE),
	DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE),
	DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE),
	DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE),
	DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE),

	DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE),
	DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE),
	DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE),
	DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE),
	DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE),
	DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE),
	DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE),
	DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE),
	DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE),
	DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE),
	} dmu_object_type_t;

	/*
	* These flags are intended to be used to specify the "txg_how"
	* parameter when calling the dmu_tx_assign() function. See the comment
	* above dmu_tx_assign() for more details on the meaning of these flags.
	*/
	#define TXG_NOWAIT (0ULL)
	#define TXG_WAIT (1ULL<<0)
	#define TXG_NOTHROTTLE (1ULL<<1)

	void byteswap_uint64_array(void *buf, size_t size);
	void byteswap_uint32_array(void *buf, size_t size);
	void byteswap_uint16_array(void *buf, size_t size);
	void byteswap_uint8_array(void *buf, size_t size);
	void zap_byteswap(void *buf, size_t size);
	void zfs_oldacl_byteswap(void *buf, size_t size);
	void zfs_acl_byteswap(void *buf, size_t size);
	void zfs_znode_byteswap(void *buf, size_t size);

	#define DS_FIND_SNAPSHOTS (1<<0)
	#define DS_FIND_CHILDREN (1<<1)
	#define DS_FIND_SERIALIZE (1<<2)

	/*
	* The maximum number of bytes that can be accessed as part of one
	* operation, including metadata.
	*/
	#define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */
	#define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */

	#define DMU_USERUSED_OBJECT (-1ULL)
	#define DMU_GROUPUSED_OBJECT (-2ULL)
	#define DMU_PROJECTUSED_OBJECT (-3ULL)

	/*
	* Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT.
	*/
	#define DMU_OBJACCT_PREFIX "obj-"
	#define DMU_OBJACCT_PREFIX_LEN 4

	/*
	* artificial blkids for bonus buffer and spill blocks
	*/
	#define DMU_BONUS_BLKID (-1ULL)
	#define DMU_SPILL_BLKID (-2ULL)

	/*
	* Public routines to create, destroy, open, and close objsets.
	*/
	typedef void dmu_objset_create_sync_func_t(objset_t os, void arg,
	cred_t cr, dmu_tx_t tx);

	int dmu_objset_hold(const char name, const void tag, objset_t **osp);
	int dmu_objset_own(const char *name, dmu_objset_type_t type,
	boolean_t readonly, boolean_t key_required, const void *tag,
	objset_t **osp);
	void dmu_objset_rele(objset_t os, const void tag);
	void dmu_objset_disown(objset_t os, boolean_t key_required, const void tag);
	int dmu_objset_open_ds(struct dsl_dataset ds, objset_t *osp);

	void dmu_objset_evict_dbufs(objset_t *os);
	int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
	struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func,
	void *arg);
	int dmu_objset_clone(const char name, const char origin);
	int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer,
	struct nvlist *errlist);
	int dmu_objset_snapshot_one(const char fsname, const char snapname);
	int dmu_objset_find(const char name, int func(const char , void ), void arg,
	int flags);
	void dmu_objset_byteswap(void *buf, size_t size);
	int dsl_dataset_rename_snapshot(const char *fsname,
	const char oldsnapname, const char newsnapname, boolean_t recursive);

	typedef struct dmu_buf {
	uint64_t db_object; /* object that this buffer is part of */
	uint64_t db_offset; /* byte offset in this object */
	uint64_t db_size; /* size of buffer in bytes */
	void db_data; / data in buffer */
	} dmu_buf_t;

	/*
	* The names of zap entries in the DIRECTORY_OBJECT of the MOS.
	*/
	#define DMU_POOL_DIRECTORY_OBJECT 1
	#define DMU_POOL_CONFIG "config"
	#define DMU_POOL_FEATURES_FOR_WRITE "features_for_write"
	#define DMU_POOL_FEATURES_FOR_READ "features_for_read"
	#define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions"
	#define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg"
	#define DMU_POOL_ROOT_DATASET "root_dataset"
	#define DMU_POOL_SYNC_BPOBJ "sync_bplist"
	#define DMU_POOL_ERRLOG_SCRUB "errlog_scrub"
	#define DMU_POOL_ERRLOG_LAST "errlog_last"
	#define DMU_POOL_SPARES "spares"
	#define DMU_POOL_DEFLATE "deflate"
	#define DMU_POOL_HISTORY "history"
	#define DMU_POOL_PROPS "pool_props"
	#define DMU_POOL_L2CACHE "l2cache"
	#define DMU_POOL_TMP_USERREFS "tmp_userrefs"
	#define DMU_POOL_DDT "DDT-%s-%s-%s"
	#define DMU_POOL_DDT_STATS "DDT-statistics"
	#define DMU_POOL_CREATION_VERSION "creation_version"
	#define DMU_POOL_SCAN "scan"
	#define DMU_POOL_FREE_BPOBJ "free_bpobj"
	#define DMU_POOL_BPTREE_OBJ "bptree_obj"
	#define DMU_POOL_EMPTY_BPOBJ "empty_bpobj"
	#define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt"
	#define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map"
	#define DMU_POOL_REMOVING "com.delphix:removing"
	#define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj"
	#define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect"
	#define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint"
	#define DMU_POOL_LOG_SPACEMAP_ZAP "com.delphix:log_spacemap_zap"
	#define DMU_POOL_DELETED_CLONES "com.delphix:deleted_clones"

	/*
	* Allocate an object from this objset. The range of object numbers
	* available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode.
	*
	* The transaction must be assigned to a txg. The newly allocated
	* object will be "held" in the transaction (ie. you can modify the
	* newly allocated object in this transaction).
	*
	* dmu_object_alloc() chooses an object and returns it in *objectp.
	*
	* dmu_object_claim() allocates a specific object number. If that
	* number is already allocated, it fails and returns EEXIST.
	*
	* Return 0 on success, or ENOSPC or EEXIST as specified above.
	*/
	uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
	uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift,
	dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
	uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonus_type, int bonus_len,
	int dnodesize, dmu_tx_t *tx);
	uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot,
	int blocksize, int indirect_blockshift, dmu_object_type_t bonustype,
	int bonuslen, int dnodesize, dnode_t *allocated_dnode, const void tag,
	dmu_tx_t *tx);
	int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx);
	int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonus_type, int bonus_len,
	int dnodesize, dmu_tx_t *tx);
	int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp);
	int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object,
	dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype,
	int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx);
	int dmu_object_rm_spill(objset_t os, uint64_t object, dmu_tx_t tx);

	/*
	* Free an object from this objset.
	*
	* The object's data will be freed as well (ie. you don't need to call
	* dmu_free(object, 0, -1, tx)).
	*
	* The object need not be held in the transaction.
	*
	* If there are any holds on this object's buffers (via dmu_buf_hold()),
	* or tx holds on the object (via dmu_tx_hold_object()), you can not
	* free it; it fails and returns EBUSY.
	*
	* If the object is not allocated, it fails and returns ENOENT.
	*
	* Return 0 on success, or EBUSY or ENOENT as specified above.
	*/
	int dmu_object_free(objset_t os, uint64_t object, dmu_tx_t tx);

	/*
	* Find the next allocated or free object.
	*
	* The objectp parameter is in-out. It will be updated to be the next
	* object which is allocated. Ignore objects which have not been
	* modified since txg.
	*
	* XXX Can only be called on a objset with no dirty data.
	*
	* Returns 0 on success, or ENOENT if there are no more objects.
	*/
	int dmu_object_next(objset_t os, uint64_t objectp,
	boolean_t hole, uint64_t txg);

	/*
	* Set the number of levels on a dnode. nlevels must be greater than the
	* current number of levels or an EINVAL will be returned.
	*/
	int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels,
	dmu_tx_t *tx);

	/*
	* Set the data blocksize for an object.
	*
	* The object cannot have any blocks allocated beyond the first. If
	* the first block is allocated already, the new size must be greater
	* than the current block size. If these conditions are not met,
	* ENOTSUP will be returned.
	*
	* Returns 0 on success, or EBUSY if there are any holds on the object
	* contents, or ENOTSUP as described above.
	*/
	int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size,
	int ibs, dmu_tx_t *tx);

	/*
	* Manually set the maxblkid on a dnode. This will adjust nlevels accordingly
	* to accommodate the change. When calling this function, the caller must
	* ensure that the object's nlevels can sufficiently support the new maxblkid.
	*/
	int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
	dmu_tx_t *tx);

	/*
	* Set the checksum property on a dnode. The new checksum algorithm will
	* apply to all newly written blocks; existing blocks will not be affected.
	*/
	void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
	dmu_tx_t *tx);

	/*
	* Set the compress property on a dnode. The new compression algorithm will
	* apply to all newly written blocks; existing blocks will not be affected.
	*/
	void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
	dmu_tx_t *tx);

	void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
	void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
	int compressed_size, int byteorder, dmu_tx_t *tx);
	void dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx);

	/*
	* Decide how to write a block: checksum, compression, number of copies, etc.
	*/
	#define WP_NOFILL 0x1
	#define WP_DMU_SYNC 0x2
	#define WP_SPILL 0x4

	void dmu_write_policy(objset_t os, dnode_t dn, int level, int wp,
	struct zio_prop *zp);

	/*
	* The bonus data is accessed more or less like a regular buffer.
	* You must dmu_bonus_hold() to get the buffer, which will give you a
	* dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus
	* data. As with any normal buffer, you must call dmu_buf_will_dirty()
	* before modifying it, and the
	* object must be held in an assigned transaction before calling
	* dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus
	* buffer as well. You must release what you hold with dmu_buf_rele().
	*
	* Returns ENOENT, EIO, or 0.
	*/
	int dmu_bonus_hold(objset_t os, uint64_t object, const void tag,
	dmu_buf_t **dbp);
	int dmu_bonus_hold_by_dnode(dnode_t dn, const void tag, dmu_buf_t **dbp,
	uint32_t flags);
	int dmu_bonus_max(void);
	int dmu_set_bonus(dmu_buf_t , int, dmu_tx_t );
	int dmu_set_bonustype(dmu_buf_t , dmu_object_type_t, dmu_tx_t );
	dmu_object_type_t dmu_get_bonustype(dmu_buf_t *);
	int dmu_rm_spill(objset_t , uint64_t, dmu_tx_t );

	/*
	* Special spill buffer support used by "SA" framework
	*/

	int dmu_spill_hold_by_bonus(dmu_buf_t bonus, uint32_t flags, const void tag,
	dmu_buf_t **dbp);
	int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags,
	const void tag, dmu_buf_t *dbp);
	int dmu_spill_hold_existing(dmu_buf_t bonus, const void tag, dmu_buf_t **dbp);

	/*
	* Obtain the DMU buffer from the specified object which contains the
	* specified offset. dmu_buf_hold() puts a "hold" on the buffer, so
	* that it will remain in memory. You must release the hold with
	* dmu_buf_rele(). You must not access the dmu_buf_t after releasing
	* what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU.
	*
	* You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill
	* on the returned buffer before reading or writing the buffer's
	* db_data. The comments for those routines describe what particular
	* operations are valid after calling them.
	*
	* The object number must be a valid, allocated object number.
	*/
	int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
	const void tag, dmu_buf_t *, int flags);
	int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t length, int read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp);
	int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
	const void tag, dmu_buf_t *dbp, int flags);
	int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
	uint64_t length, boolean_t read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp, uint32_t flags);
	/*
	* Add a reference to a dmu buffer that has already been held via
	* dmu_buf_hold() in the current context.
	*/
	void dmu_buf_add_ref(dmu_buf_t db, const void tag);

	/*
	* Attempt to add a reference to a dmu buffer that is in an unknown state,
	* using a pointer that may have been invalidated by eviction processing.
	* The request will succeed if the passed in dbuf still represents the
	* same os/object/blkid, is ineligible for eviction, and has at least
	* one hold by a user other than the syncer.
	*/
	boolean_t dmu_buf_try_add_ref(dmu_buf_t , objset_t os, uint64_t object,
	uint64_t blkid, const void *tag);

	void dmu_buf_rele(dmu_buf_t db, const void tag);
	uint64_t dmu_buf_refcount(dmu_buf_t *db);
	uint64_t dmu_buf_user_refcount(dmu_buf_t *db);

	/*
	* dmu_buf_hold_array holds the DMU buffers which contain all bytes in a
	* range of an object. A pointer to an array of dmu_buf_t*'s is
	* returned (in *dbpp).
	*
	* dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and
	* frees the array. The hold on the array of buffers MUST be released
	* with dmu_buf_rele_array. You can NOT release the hold on each buffer
	* individually with dmu_buf_rele.
	*/
	int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset,
	uint64_t length, boolean_t read, const void *tag,
	int numbufsp, dmu_buf_t **dbpp);
	void dmu_buf_rele_array(dmu_buf_t *, int numbufs, const void tag);

	typedef void dmu_buf_evict_func_t(void *user_ptr);

	/*
	* A DMU buffer user object may be associated with a dbuf for the
	* duration of its lifetime. This allows the user of a dbuf (client)
	* to attach private data to a dbuf (e.g. in-core only data such as a
	* dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified
	* when that dbuf has been evicted. Clients typically respond to the
	* eviction notification by freeing their private data, thus ensuring
	* the same lifetime for both dbuf and private data.
	*
	* The mapping from a dmu_buf_user_t to any client private data is the
	* client's responsibility. All current consumers of the API with private
	* data embed a dmu_buf_user_t as the first member of the structure for
	* their private data. This allows conversions between the two types
	* with a simple cast. Since the DMU buf user API never needs access
	* to the private data, other strategies can be employed if necessary
	* or convenient for the client (e.g. using container_of() to do the
	* conversion for private data that cannot have the dmu_buf_user_t as
	* its first member).
	*
	* Eviction callbacks are executed without the dbuf mutex held or any
	* other type of mechanism to guarantee that the dbuf is still available.
	* For this reason, users must assume the dbuf has already been freed
	* and not reference the dbuf from the callback context.
	*
	* Users requesting "immediate eviction" are notified as soon as the dbuf
	* is only referenced by dirty records (dirties == holds). Otherwise the
	* notification occurs after eviction processing for the dbuf begins.
	*/
	typedef struct dmu_buf_user {
	/*
	* Asynchronous user eviction callback state.
	*/
	taskq_ent_t dbu_tqent;

	/*
	* This instance's eviction function pointers.
	*
	* dbu_evict_func_sync is called synchronously and then
	* dbu_evict_func_async is executed asynchronously on a taskq.
	*/
	dmu_buf_evict_func_t *dbu_evict_func_sync;
	dmu_buf_evict_func_t *dbu_evict_func_async;
	#ifdef ZFS_DEBUG
	/*
	* Pointer to user's dbuf pointer. NULL for clients that do
	* not associate a dbuf with their user data.
	*
	* The dbuf pointer is cleared upon eviction so as to catch
	* use-after-evict bugs in clients.
	*/
	dmu_buf_t **dbu_clear_on_evict_dbufp;
	#endif
	} dmu_buf_user_t;

	/*
	* Initialize the given dmu_buf_user_t instance with the eviction function
	* evict_func, to be called when the user is evicted.
	*
	* NOTE: This function should only be called once on a given dmu_buf_user_t.
	* To allow enforcement of this, dbu must already be zeroed on entry.
	*/
	static inline void
	dmu_buf_init_user(dmu_buf_user_t dbu, dmu_buf_evict_func_t evict_func_sync,
	dmu_buf_evict_func_t *evict_func_async,
	dmu_buf_t **clear_on_evict_dbufp __maybe_unused)
	{
	ASSERT(dbu->dbu_evict_func_sync == NULL);
	ASSERT(dbu->dbu_evict_func_async == NULL);

	/* must have at least one evict func */
	IMPLY(evict_func_sync == NULL, evict_func_async != NULL);
	dbu->dbu_evict_func_sync = evict_func_sync;
	dbu->dbu_evict_func_async = evict_func_async;
	taskq_init_ent(&dbu->dbu_tqent);
	#ifdef ZFS_DEBUG
	dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp;
	#endif
	}

	/*
	* Attach user data to a dbuf and mark it for normal (when the dbuf's
	* data is cleared or its reference count goes to zero) eviction processing.
	*
	* Returns NULL on success, or the existing user if another user currently
	* owns the buffer.
	*/
	void dmu_buf_set_user(dmu_buf_t db, dmu_buf_user_t *user);

	/*
	* Attach user data to a dbuf and mark it for immediate (its dirty and
	* reference counts are equal) eviction processing.
	*
	* Returns NULL on success, or the existing user if another user currently
	* owns the buffer.
	*/
	void dmu_buf_set_user_ie(dmu_buf_t db, dmu_buf_user_t *user);

	/*
	* Replace the current user of a dbuf.
	*
	* If given the current user of a dbuf, replaces the dbuf's user with
	* "new_user" and returns the user data pointer that was replaced.
	* Otherwise returns the current, and unmodified, dbuf user pointer.
	*/
	void dmu_buf_replace_user(dmu_buf_t db,
	dmu_buf_user_t old_user, dmu_buf_user_t new_user);

	/*
	* Remove the specified user data for a DMU buffer.
	*
	* Returns the user that was removed on success, or the current user if
	* another user currently owns the buffer.
	*/
	void dmu_buf_remove_user(dmu_buf_t db, dmu_buf_user_t *user);

	/*
	* Returns the user data (dmu_buf_user_t *) associated with this dbuf.
	*/
	void dmu_buf_get_user(dmu_buf_t db);

	objset_t dmu_buf_get_objset(dmu_buf_t db);
	dnode_t dmu_buf_dnode_enter(dmu_buf_t db);
	void dmu_buf_dnode_exit(dmu_buf_t *db);

	/* Block until any in-progress dmu buf user evictions complete. */
	void dmu_buf_user_evict_wait(void);

	/*
	* Returns the blkptr associated with this dbuf, or NULL if not set.
	*/
	struct blkptr dmu_buf_get_blkptr(dmu_buf_t db);

	/*
	* Indicate that you are going to modify the buffer's data (db_data).
	*
	* The transaction (tx) must be assigned to a txg (ie. you've called
	* dmu_tx_assign()). The buffer's object must be held in the tx
	* (ie. you've called dmu_tx_hold_object(tx, db->db_object)).
	*/
	void dmu_buf_will_dirty(dmu_buf_t db, dmu_tx_t tx);
	boolean_t dmu_buf_is_dirty(dmu_buf_t db, dmu_tx_t tx);
	void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t mac, dmu_tx_t tx);

	/*
	* You must create a transaction, then hold the objects which you will
	* (or might) modify as part of this transaction. Then you must assign
	* the transaction to a transaction group. Once the transaction has
	* been assigned, you can modify buffers which belong to held objects as
	* part of this transaction. You can't modify buffers before the
	* transaction has been assigned; you can't modify buffers which don't
	* belong to objects which this transaction holds; you can't hold
	* objects once the transaction has been assigned. You may hold an
	* object which you are going to free (with dmu_object_free()), but you
	* don't have to.
	*
	* You can abort the transaction before it has been assigned.
	*
	* Note that you may hold buffers (with dmu_buf_hold) at any time,
	* regardless of transaction state.
	*/

	#define DMU_NEW_OBJECT (-1ULL)
	#define DMU_OBJECT_END (-1ULL)

	dmu_tx_t dmu_tx_create(objset_t os);
	void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len);
	void dmu_tx_hold_write_by_dnode(dmu_tx_t tx, dnode_t dn, uint64_t off,
	int len);
	void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off,
	uint64_t len);
	void dmu_tx_hold_free_by_dnode(dmu_tx_t tx, dnode_t dn, uint64_t off,
	uint64_t len);
	void dmu_tx_hold_zap(dmu_tx_t tx, uint64_t object, int add, const char name);
	void dmu_tx_hold_zap_by_dnode(dmu_tx_t tx, dnode_t dn, int add,
	const char *name);
	void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object);
	void dmu_tx_hold_bonus_by_dnode(dmu_tx_t tx, dnode_t dn);
	void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object);
	void dmu_tx_hold_sa(dmu_tx_t tx, struct sa_handle hdl, boolean_t may_grow);
	void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size);
	void dmu_tx_abort(dmu_tx_t *tx);
	int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how);
	void dmu_tx_wait(dmu_tx_t *tx);
	void dmu_tx_commit(dmu_tx_t *tx);
	void dmu_tx_mark_netfree(dmu_tx_t *tx);

	/*
	* To register a commit callback, dmu_tx_callback_register() must be called.
	*
	* dcb_data is a pointer to caller private data that is passed on as a
	* callback parameter. The caller is responsible for properly allocating and
	* freeing it.
	*
	* When registering a callback, the transaction must be already created, but
	* it cannot be committed or aborted. It can be assigned to a txg or not.
	*
	* The callback will be called after the transaction has been safely written
	* to stable storage and will also be called if the dmu_tx is aborted.
	* If there is any error which prevents the transaction from being committed to
	* disk, the callback will be called with a value of error != 0.
	*
	* When multiple callbacks are registered to the transaction, the callbacks
	* will be called in reverse order to let Lustre, the only user of commit
	* callback currently, take the fast path of its commit callback handling.
	*/
	typedef void dmu_tx_callback_func_t(void *dcb_data, int error);

	void dmu_tx_callback_register(dmu_tx_t tx, dmu_tx_callback_func_t dcb_func,
	void *dcb_data);
	void dmu_tx_do_callbacks(list_t *cb_list, int error);

	/*
	* Free up the data blocks for a defined range of a file. If size is
	* -1, the range from offset to end-of-file is freed.
	*/
	int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t size, dmu_tx_t *tx);
	int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t size);
	int dmu_free_long_object(objset_t *os, uint64_t object);

	/*
	* Convenience functions.
	*
	* Canfail routines will return 0 on success, or an errno if there is a
	* nonrecoverable I/O error.
	*/
	#define DMU_READ_PREFETCH 0 /* prefetch */
	#define DMU_READ_NO_PREFETCH 1 /* don't prefetch */
	#define DMU_READ_NO_DECRYPT 2 /* don't decrypt */
	int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	void *buf, uint32_t flags);
	int dmu_read_by_dnode(dnode_t dn, uint64_t offset, uint64_t size, void buf,
	uint32_t flags);
	void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx);
	void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx);
	void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx);
	#ifdef _KERNEL
	int dmu_read_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size);
	int dmu_read_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size);
	int dmu_read_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size);
	int dmu_write_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx);
	int dmu_write_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx);
	int dmu_write_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx);
	#endif
	struct arc_buf dmu_request_arcbuf(dmu_buf_t handle, int size);
	void dmu_return_arcbuf(struct arc_buf *buf);
	int dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset,
	struct arc_buf buf, dmu_tx_t tx);
	int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset,
	struct arc_buf buf, dmu_tx_t tx);
	#define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf
	-extern int zfs_max_recordsize;
	+extern uint_t zfs_max_recordsize;

	/*
	* Asynchronously try to read in the data.
	*/
	void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
	uint64_t len, enum zio_priority pri);

	typedef struct dmu_object_info {
	/* All sizes are in bytes unless otherwise indicated. */
	uint32_t doi_data_block_size;
	uint32_t doi_metadata_block_size;
	dmu_object_type_t doi_type;
	dmu_object_type_t doi_bonus_type;
	uint64_t doi_bonus_size;
	uint8_t doi_indirection; /* 2 = dnode->indirect->data */
	uint8_t doi_checksum;
	uint8_t doi_compress;
	uint8_t doi_nblkptr;
	uint8_t doi_pad[4];
	uint64_t doi_dnodesize;
	uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */
	uint64_t doi_max_offset;
	uint64_t doi_fill_count; /* number of non-empty blocks */
	} dmu_object_info_t;

	typedef void (const arc_byteswap_func_t)(void buf, size_t size);

	typedef struct dmu_object_type_info {
	dmu_object_byteswap_t ot_byteswap;
	boolean_t ot_metadata;
	boolean_t ot_dbuf_metadata_cache;
	boolean_t ot_encrypt;
	const char *ot_name;
	} dmu_object_type_info_t;

	typedef const struct dmu_object_byteswap_info {
	arc_byteswap_func_t ob_func;
	const char *ob_name;
	} dmu_object_byteswap_info_t;

	extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES];
	-extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];
	+extern dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS];

	/*
	* Get information on a DMU object.
	*
	* Return 0 on success or ENOENT if object is not allocated.
	*
	* If doi is NULL, just indicates whether the object exists.
	*/
	int dmu_object_info(objset_t os, uint64_t object, dmu_object_info_t doi);
	void __dmu_object_info_from_dnode(struct dnode dn, dmu_object_info_t doi);
	/* Like dmu_object_info, but faster if you have a held dnode in hand. */
	void dmu_object_info_from_dnode(dnode_t dn, dmu_object_info_t doi);
	/* Like dmu_object_info, but faster if you have a held dbuf in hand. */
	void dmu_object_info_from_db(dmu_buf_t db, dmu_object_info_t doi);
	/*
	* Like dmu_object_info_from_db, but faster still when you only care about
	* the size.
	*/
	void dmu_object_size_from_db(dmu_buf_t db, uint32_t blksize,
	u_longlong_t *nblk512);

	void dmu_object_dnsize_from_db(dmu_buf_t db, int dnsize);

	typedef struct dmu_objset_stats {
	uint64_t dds_num_clones; /* number of clones of this */
	uint64_t dds_creation_txg;
	uint64_t dds_guid;
	dmu_objset_type_t dds_type;
	uint8_t dds_is_snapshot;
	uint8_t dds_inconsistent;
	uint8_t dds_redacted;
	char dds_origin[ZFS_MAX_DATASET_NAME_LEN];
	} dmu_objset_stats_t;

	/*
	* Get stats on a dataset.
	*/
	void dmu_objset_fast_stat(objset_t os, dmu_objset_stats_t stat);

	/*
	* Add entries to the nvlist for all the objset's properties. See
	* zfs_prop_table[] and zfs(1m) for details on the properties.
	*/
	void dmu_objset_stats(objset_t os, struct nvlist nv);

	/*
	* Get the space usage statistics for statvfs().
	*
	* refdbytes is the amount of space "referenced" by this objset.
	* availbytes is the amount of space available to this objset, taking
	* into account quotas & reservations, assuming that no other objsets
	* use the space first. These values correspond to the 'referenced' and
	* 'available' properties, described in the zfs(1m) manpage.
	*
	* usedobjs and availobjs are the number of objects currently allocated,
	* and available.
	*/
	void dmu_objset_space(objset_t os, uint64_t refdbytesp, uint64_t *availbytesp,
	uint64_t usedobjsp, uint64_t availobjsp);

	/*
	* The fsid_guid is a 56-bit ID that can change to avoid collisions.
	* (Contrast with the ds_guid which is a 64-bit ID that will never
	* change, so there is a small probability that it will collide.)
	*/
	uint64_t dmu_objset_fsid_guid(objset_t *os);

	/*
	* Get the [cm]time for an objset's snapshot dir
	*/
	inode_timespec_t dmu_objset_snap_cmtime(objset_t *os);

	int dmu_objset_is_snapshot(objset_t *os);

	extern struct spa dmu_objset_spa(objset_t os);
	extern struct zilog dmu_objset_zil(objset_t os);
	extern struct dsl_pool dmu_objset_pool(objset_t os);
	extern struct dsl_dataset dmu_objset_ds(objset_t os);
	extern void dmu_objset_name(objset_t os, char buf);
	extern dmu_objset_type_t dmu_objset_type(objset_t *os);
	extern uint64_t dmu_objset_id(objset_t *os);
	extern uint64_t dmu_objset_dnodesize(objset_t *os);
	extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
	extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
	extern int dmu_objset_blksize(objset_t *os);
	extern int dmu_snapshot_list_next(objset_t os, int namelen, char name,
	uint64_t id, uint64_t offp, boolean_t *case_conflict);
	extern int dmu_snapshot_lookup(objset_t os, const char name, uint64_t *val);
	extern int dmu_snapshot_realname(objset_t os, const char name, char *real,
	int maxlen, boolean_t *conflict);
	extern int dmu_dir_list_next(objset_t os, int namelen, char name,
	uint64_t idp, uint64_t offp);

	typedef struct zfs_file_info {
	uint64_t zfi_user;
	uint64_t zfi_group;
	uint64_t zfi_project;
	uint64_t zfi_generation;
	} zfs_file_info_t;

	typedef int file_info_cb_t(dmu_object_type_t bonustype, const void *data,
	struct zfs_file_info *zoi);
	extern void dmu_objset_register_type(dmu_objset_type_t ost,
	file_info_cb_t *cb);
	extern void dmu_objset_set_user(objset_t os, void user_ptr);
	extern void dmu_objset_get_user(objset_t os);

	/*
	* Return the txg number for the given assigned transaction.
	*/
	uint64_t dmu_tx_get_txg(dmu_tx_t *tx);

	/*
	* Synchronous write.
	* If a parent zio is provided this function initiates a write on the
	* provided buffer as a child of the parent zio.
	* In the absence of a parent zio, the write is completed synchronously.
	* At write completion, blk is filled with the bp of the written block.
	* Note that while the data covered by this function will be on stable
	* storage when the write completes this new data does not become a
	* permanent part of the file until the associated transaction commits.
	*/

	/*
	* {zfs,zvol,ztest}_get_done() args
	*/
	typedef struct zgd {
	struct lwb *zgd_lwb;
	struct blkptr *zgd_bp;
	dmu_buf_t *zgd_db;
	struct zfs_locked_range *zgd_lr;
	void *zgd_private;
	} zgd_t;

	typedef void dmu_sync_cb_t(zgd_t *arg, int error);
	int dmu_sync(struct zio zio, uint64_t txg, dmu_sync_cb_t done, zgd_t *zgd);

	/*
	* Find the next hole or data block in file starting at *off
	* Return found offset in *off. Return ESRCH for end of file.
	*/
	int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole,
	uint64_t *off);

	/*
	* Initial setup and final teardown.
	*/
	extern void dmu_init(void);
	extern void dmu_fini(void);

	typedef void (dmu_traverse_cb_t)(objset_t os, void arg, struct blkptr bp,
	uint64_t object, uint64_t offset, int len);
	void dmu_traverse_objset(objset_t *os, uint64_t txg_start,
	dmu_traverse_cb_t cb, void *arg);

	int dmu_diff(const char tosnap_name, const char fromsnap_name,
	zfs_file_t fp, offset_t offp);

	/* CRC64 table */
	#define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */
	extern uint64_t zfs_crc64_table[256];

	-extern int dmu_prefetch_max;
	+extern uint_t dmu_prefetch_max;

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_DMU_H */
	diff --git a/sys/contrib/openzfs/include/sys/dsl_pool.h b/sys/contrib/openzfs/include/sys/dsl_pool.h
	index 226bb7f2c23a..9364106d94b7 100644
	--- a/sys/contrib/openzfs/include/sys/dsl_pool.h
	+++ b/sys/contrib/openzfs/include/sys/dsl_pool.h
	@@ -1,205 +1,205 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2018 by Delphix. All rights reserved.
	* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
	*/

	#ifndef _SYS_DSL_POOL_H
	#define _SYS_DSL_POOL_H

	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/txg_impl.h>
	#include <sys/zfs_context.h>
	#include <sys/zio.h>
	#include <sys/dnode.h>
	#include <sys/ddt.h>
	#include <sys/arc.h>
	#include <sys/bpobj.h>
	#include <sys/bptree.h>
	#include <sys/rrwlock.h>
	#include <sys/dsl_synctask.h>
	#include <sys/mmp.h>
	#include <sys/aggsum.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	extern int zfs_txg_synctime_ms;

	struct objset;
	struct dsl_dir;
	struct dsl_dataset;
	struct dsl_pool;
	struct dmu_tx;
	struct dsl_scan;
	struct dsl_crypto_params;
	struct dsl_deadlist;

	extern unsigned long zfs_dirty_data_max;
	extern unsigned long zfs_dirty_data_max_max;
	extern unsigned long zfs_wrlog_data_max;
	-extern int zfs_dirty_data_max_percent;
	-extern int zfs_dirty_data_max_max_percent;
	-extern int zfs_delay_min_dirty_percent;
	+extern uint_t zfs_dirty_data_max_percent;
	+extern uint_t zfs_dirty_data_max_max_percent;
	+extern uint_t zfs_delay_min_dirty_percent;
	extern unsigned long zfs_delay_scale;

	/* These macros are for indexing into the zfs_all_blkstats_t. */
	#define DMU_OT_DEFERRED DMU_OT_NONE
	#define DMU_OT_OTHER DMU_OT_NUMTYPES /* place holder for DMU_OT() types */
	#define DMU_OT_TOTAL (DMU_OT_NUMTYPES + 1)

	typedef struct zfs_blkstat {
	uint64_t zb_count;
	uint64_t zb_asize;
	uint64_t zb_lsize;
	uint64_t zb_psize;
	uint64_t zb_gangs;
	uint64_t zb_ditto_2_of_2_samevdev;
	uint64_t zb_ditto_2_of_3_samevdev;
	uint64_t zb_ditto_3_of_3_samevdev;
	} zfs_blkstat_t;

	typedef struct zfs_all_blkstats {
	zfs_blkstat_t zab_type[DN_MAX_LEVELS + 1][DMU_OT_TOTAL + 1];
	} zfs_all_blkstats_t;


	typedef struct dsl_pool {
	/* Immutable */
	spa_t *dp_spa;
	struct objset *dp_meta_objset;
	struct dsl_dir *dp_root_dir;
	struct dsl_dir *dp_mos_dir;
	struct dsl_dir *dp_free_dir;
	struct dsl_dir *dp_leak_dir;
	struct dsl_dataset *dp_origin_snap;
	uint64_t dp_root_dir_obj;
	struct taskq *dp_zrele_taskq;
	struct taskq *dp_unlinked_drain_taskq;

	/* No lock needed - sync context only */
	blkptr_t dp_meta_rootbp;
	uint64_t dp_tmp_userrefs_obj;
	bpobj_t dp_free_bpobj;
	uint64_t dp_bptree_obj;
	uint64_t dp_empty_bpobj;
	bpobj_t dp_obsolete_bpobj;

	struct dsl_scan *dp_scan;

	/* Uses dp_lock */
	kmutex_t dp_lock;
	kcondvar_t dp_spaceavail_cv;
	uint64_t dp_dirty_pertxg[TXG_SIZE];
	uint64_t dp_dirty_total;
	uint64_t dp_long_free_dirty_pertxg[TXG_SIZE];
	uint64_t dp_mos_used_delta;
	uint64_t dp_mos_compressed_delta;
	uint64_t dp_mos_uncompressed_delta;

	aggsum_t dp_wrlog_pertxg[TXG_SIZE];
	aggsum_t dp_wrlog_total;

	/*
	* Time of most recently scheduled (furthest in the future)
	* wakeup for delayed transactions.
	*/
	hrtime_t dp_last_wakeup;

	/* Has its own locking */
	tx_state_t dp_tx;
	txg_list_t dp_dirty_datasets;
	txg_list_t dp_dirty_zilogs;
	txg_list_t dp_dirty_dirs;
	txg_list_t dp_sync_tasks;
	txg_list_t dp_early_sync_tasks;
	taskq_t *dp_sync_taskq;
	taskq_t *dp_zil_clean_taskq;

	/*
	* Protects administrative changes (properties, namespace)
	*
	* It is only held for write in syncing context. Therefore
	* syncing context does not need to ever have it for read, since
	* nobody else could possibly have it for write.
	*/
	rrwlock_t dp_config_rwlock;

	zfs_all_blkstats_t *dp_blkstats;
	} dsl_pool_t;

	int dsl_pool_init(spa_t spa, uint64_t txg, dsl_pool_t *dpp);
	int dsl_pool_open(dsl_pool_t *dp);
	void dsl_pool_close(dsl_pool_t *dp);
	dsl_pool_t dsl_pool_create(spa_t spa, nvlist_t *zplprops,
	struct dsl_crypto_params *dcp, uint64_t txg);
	void dsl_pool_sync(dsl_pool_t *dp, uint64_t txg);
	void dsl_pool_sync_done(dsl_pool_t *dp, uint64_t txg);
	int dsl_pool_sync_context(dsl_pool_t *dp);
	uint64_t dsl_pool_adjustedsize(dsl_pool_t *dp, zfs_space_check_t slop_policy);
	uint64_t dsl_pool_unreserved_space(dsl_pool_t *dp,
	zfs_space_check_t slop_policy);
	uint64_t dsl_pool_deferred_space(dsl_pool_t *dp);
	void dsl_pool_wrlog_count(dsl_pool_t *dp, int64_t size, uint64_t txg);
	boolean_t dsl_pool_need_wrlog_delay(dsl_pool_t *dp);
	void dsl_pool_dirty_space(dsl_pool_t dp, int64_t space, dmu_tx_t tx);
	void dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg);
	void dsl_free(dsl_pool_t dp, uint64_t txg, const blkptr_t bpp);
	void dsl_free_sync(zio_t pio, dsl_pool_t dp, uint64_t txg,
	const blkptr_t *bpp);
	void dsl_pool_create_origin(dsl_pool_t dp, dmu_tx_t tx);
	void dsl_pool_upgrade_clones(dsl_pool_t dp, dmu_tx_t tx);
	void dsl_pool_upgrade_dir_clones(dsl_pool_t dp, dmu_tx_t tx);
	void dsl_pool_mos_diduse_space(dsl_pool_t *dp,
	int64_t used, int64_t comp, int64_t uncomp);
	void dsl_pool_ckpoint_diduse_space(dsl_pool_t *dp,
	int64_t used, int64_t comp, int64_t uncomp);
	boolean_t dsl_pool_need_dirty_delay(dsl_pool_t *dp);
	void dsl_pool_config_enter(dsl_pool_t dp, const void tag);
	void dsl_pool_config_enter_prio(dsl_pool_t dp, const void tag);
	void dsl_pool_config_exit(dsl_pool_t dp, const void tag);
	boolean_t dsl_pool_config_held(dsl_pool_t *dp);
	boolean_t dsl_pool_config_held_writer(dsl_pool_t *dp);

	taskq_t dsl_pool_zrele_taskq(dsl_pool_t dp);
	taskq_t dsl_pool_unlinked_drain_taskq(dsl_pool_t dp);

	int dsl_pool_user_hold(dsl_pool_t *dp, uint64_t dsobj,
	const char tag, uint64_t now, dmu_tx_t tx);
	int dsl_pool_user_release(dsl_pool_t *dp, uint64_t dsobj,
	const char tag, dmu_tx_t tx);
	void dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp);
	int dsl_pool_open_special_dir(dsl_pool_t dp, const char name, dsl_dir_t **);
	int dsl_pool_hold(const char name, const void tag, dsl_pool_t **dp);
	void dsl_pool_rele(dsl_pool_t dp, const void tag);

	void dsl_pool_create_obsolete_bpobj(dsl_pool_t dp, dmu_tx_t tx);
	void dsl_pool_destroy_obsolete_bpobj(dsl_pool_t dp, dmu_tx_t tx);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_DSL_POOL_H */
	diff --git a/sys/contrib/openzfs/include/sys/fm/util.h b/sys/contrib/openzfs/include/sys/fm/util.h
	index 7f91ca1da573..038162ab7524 100644
	--- a/sys/contrib/openzfs/include/sys/fm/util.h
	+++ b/sys/contrib/openzfs/include/sys/fm/util.h
	@@ -1,121 +1,121 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
	*/

	#ifndef _SYS_FM_UTIL_H
	#define _SYS_FM_UTIL_H

	#ifdef __cplusplus
	extern "C" {
	#endif

	#include <sys/nvpair.h>
	#include <sys/zfs_file.h>

	/*
	* Shared user/kernel definitions for class length, error channel name,
	* and kernel event publisher string.
	*/
	#define FM_MAX_CLASS 100
	#define FM_ERROR_CHAN "com.sun:fm:error"
	#define FM_PUB "fm"

	/*
	* ereport dump device transport support
	*
	* Ereports are written out to the dump device at a proscribed offset from the
	* end, similar to in-transit log messages. The ereports are represented as a
	* erpt_dump_t header followed by ed_size bytes of packed native nvlist data.
	*
	* NOTE: All of these constants and the header must be defined so they have the
	* same representation for both 32-bit and 64-bit producers and consumers.
	*/
	#define ERPT_MAGIC 0xf00d4eddU
	#define ERPT_MAX_ERRS 16
	#define ERPT_DATA_SZ (6 * 1024)
	#define ERPT_EVCH_MAX 256
	#define ERPT_HIWAT 64

	typedef struct erpt_dump {
	uint32_t ed_magic; /* ERPT_MAGIC or zero to indicate end */
	uint32_t ed_chksum; /* checksum32() of packed nvlist data */
	uint32_t ed_size; /* ereport (nvl) fixed buf size */
	uint32_t ed_pad; /* reserved for future use */
	hrtime_t ed_hrt_nsec; /* hrtime of this ereport */
	hrtime_t ed_hrt_base; /* hrtime sample corresponding to ed_tod_base */
	struct {
	uint64_t sec; /* seconds since gettimeofday() Epoch */
	uint64_t nsec; /* nanoseconds past ed_tod_base.sec */
	} ed_tod_base;
	} erpt_dump_t;

	#ifdef _KERNEL

	#define ZEVENT_SHUTDOWN 0x1

	typedef void zevent_cb_t(nvlist_t , nvlist_t );

	typedef struct zevent_s {
	nvlist_t ev_nvl; / protected by the zevent_lock */
	nvlist_t ev_detector; / " */
	list_t ev_ze_list; /* " */
	list_node_t ev_node; /* " */
	zevent_cb_t ev_cb; / " */
	uint64_t ev_eid;
	} zevent_t;

	typedef struct zfs_zevent {
	zevent_t ze_zevent; / protected by the zevent_lock */
	list_node_t ze_node; /* " */
	uint64_t ze_dropped; /* " */
	} zfs_zevent_t;

	extern void fm_init(void);
	extern void fm_fini(void);
	extern void zfs_zevent_post_cb(nvlist_t nvl, nvlist_t detector);
	extern int zfs_zevent_post(nvlist_t , nvlist_t , zevent_cb_t *);
	-extern void zfs_zevent_drain_all(int *);
	+extern void zfs_zevent_drain_all(uint_t *);
	extern zfs_file_t zfs_zevent_fd_hold(int, minor_t , zfs_zevent_t **);
	extern void zfs_zevent_fd_rele(zfs_file_t *);
	extern int zfs_zevent_next(zfs_zevent_t , nvlist_t , uint64_t , uint64_t *);
	extern int zfs_zevent_wait(zfs_zevent_t *);
	extern int zfs_zevent_seek(zfs_zevent_t *, uint64_t);
	extern void zfs_zevent_init(zfs_zevent_t **);
	extern void zfs_zevent_destroy(zfs_zevent_t *);

	extern void zfs_zevent_track_duplicate(void);
	extern void zfs_ereport_init(void);
	extern void zfs_ereport_fini(void);
	#else

	static inline void fm_init(void) { }
	static inline void fm_fini(void) { }

	#endif /* _KERNEL */

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_FM_UTIL_H */
	diff --git a/sys/contrib/openzfs/include/sys/fs/zfs.h b/sys/contrib/openzfs/include/sys/fs/zfs.h
	index dedee0e7bd5a..b3fecf489eb0 100644
	--- a/sys/contrib/openzfs/include/sys/fs/zfs.h
	+++ b/sys/contrib/openzfs/include/sys/fs/zfs.h
	@@ -1,1826 +1,1827 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2013, 2017 Joyent, Inc. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019 Datto Inc.
	* Portions Copyright 2010 Robert Milkowski
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	#ifndef _SYS_FS_ZFS_H
	#define _SYS_FS_ZFS_H extern __attribute__((visibility("default")))

	#include <sys/time.h>
	#include <sys/zio_priority.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Types and constants shared between userland and the kernel.
	*/

	/*
	* Each dataset can be one of the following types. These constants can be
	* combined into masks that can be passed to various functions.
	*/
	typedef enum {
	ZFS_TYPE_INVALID = 0,
	ZFS_TYPE_FILESYSTEM = (1 << 0),
	ZFS_TYPE_SNAPSHOT = (1 << 1),
	ZFS_TYPE_VOLUME = (1 << 2),
	ZFS_TYPE_POOL = (1 << 3),
	ZFS_TYPE_BOOKMARK = (1 << 4),
	ZFS_TYPE_VDEV = (1 << 5),
	} zfs_type_t;

	/*
	* NB: lzc_dataset_type should be updated whenever a new objset type is added,
	* if it represents a real type of a dataset that can be created from userland.
	*/
	typedef enum dmu_objset_type {
	DMU_OST_NONE,
	DMU_OST_META,
	DMU_OST_ZFS,
	DMU_OST_ZVOL,
	DMU_OST_OTHER, /* For testing only! */
	DMU_OST_ANY, /* Be careful! */
	DMU_OST_NUMTYPES
	} dmu_objset_type_t;

	#define ZFS_TYPE_DATASET \
	(ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME \| ZFS_TYPE_SNAPSHOT)

	/*
	* All of these include the terminating NUL byte.
	*/
	#define ZAP_MAXNAMELEN 256
	#define ZAP_MAXVALUELEN (1024 * 8)
	#define ZAP_OLDMAXVALUELEN 1024
	#define ZFS_MAX_DATASET_NAME_LEN 256

	/*
	* Dataset properties are identified by these constants and must be added to
	* the end of this list to ensure that external consumers are not affected
	* by the change. If you make any changes to this list, be sure to update
	* the property table in module/zcommon/zfs_prop.c.
	*/
	typedef enum {
	ZPROP_CONT = -2,
	ZPROP_INVAL = -1,
	ZPROP_USERPROP = ZPROP_INVAL,
	ZFS_PROP_TYPE = 0,
	ZFS_PROP_CREATION,
	ZFS_PROP_USED,
	ZFS_PROP_AVAILABLE,
	ZFS_PROP_REFERENCED,
	ZFS_PROP_COMPRESSRATIO,
	ZFS_PROP_MOUNTED,
	ZFS_PROP_ORIGIN,
	ZFS_PROP_QUOTA,
	ZFS_PROP_RESERVATION,
	ZFS_PROP_VOLSIZE,
	ZFS_PROP_VOLBLOCKSIZE,
	ZFS_PROP_RECORDSIZE,
	ZFS_PROP_MOUNTPOINT,
	ZFS_PROP_SHARENFS,
	ZFS_PROP_CHECKSUM,
	ZFS_PROP_COMPRESSION,
	ZFS_PROP_ATIME,
	ZFS_PROP_DEVICES,
	ZFS_PROP_EXEC,
	ZFS_PROP_SETUID,
	ZFS_PROP_READONLY,
	ZFS_PROP_ZONED,
	ZFS_PROP_SNAPDIR,
	ZFS_PROP_ACLMODE,
	ZFS_PROP_ACLINHERIT,
	ZFS_PROP_CREATETXG,
	ZFS_PROP_NAME, /* not exposed to the user */
	ZFS_PROP_CANMOUNT,
	ZFS_PROP_ISCSIOPTIONS, /* not exposed to the user */
	ZFS_PROP_XATTR,
	ZFS_PROP_NUMCLONES, /* not exposed to the user */
	ZFS_PROP_COPIES,
	ZFS_PROP_VERSION,
	ZFS_PROP_UTF8ONLY,
	ZFS_PROP_NORMALIZE,
	ZFS_PROP_CASE,
	ZFS_PROP_VSCAN,
	ZFS_PROP_NBMAND,
	ZFS_PROP_SHARESMB,
	ZFS_PROP_REFQUOTA,
	ZFS_PROP_REFRESERVATION,
	ZFS_PROP_GUID,
	ZFS_PROP_PRIMARYCACHE,
	ZFS_PROP_SECONDARYCACHE,
	ZFS_PROP_USEDSNAP,
	ZFS_PROP_USEDDS,
	ZFS_PROP_USEDCHILD,
	ZFS_PROP_USEDREFRESERV,
	ZFS_PROP_USERACCOUNTING, /* not exposed to the user */
	ZFS_PROP_STMF_SHAREINFO, /* not exposed to the user */
	ZFS_PROP_DEFER_DESTROY,
	ZFS_PROP_USERREFS,
	ZFS_PROP_LOGBIAS,
	ZFS_PROP_UNIQUE, /* not exposed to the user */
	ZFS_PROP_OBJSETID,
	ZFS_PROP_DEDUP,
	ZFS_PROP_MLSLABEL,
	ZFS_PROP_SYNC,
	ZFS_PROP_DNODESIZE,
	ZFS_PROP_REFRATIO,
	ZFS_PROP_WRITTEN,
	ZFS_PROP_CLONES,
	ZFS_PROP_LOGICALUSED,
	ZFS_PROP_LOGICALREFERENCED,
	ZFS_PROP_INCONSISTENT, /* not exposed to the user */
	ZFS_PROP_VOLMODE,
	ZFS_PROP_FILESYSTEM_LIMIT,
	ZFS_PROP_SNAPSHOT_LIMIT,
	ZFS_PROP_FILESYSTEM_COUNT,
	ZFS_PROP_SNAPSHOT_COUNT,
	ZFS_PROP_SNAPDEV,
	ZFS_PROP_ACLTYPE,
	ZFS_PROP_SELINUX_CONTEXT,
	ZFS_PROP_SELINUX_FSCONTEXT,
	ZFS_PROP_SELINUX_DEFCONTEXT,
	ZFS_PROP_SELINUX_ROOTCONTEXT,
	ZFS_PROP_RELATIME,
	ZFS_PROP_REDUNDANT_METADATA,
	ZFS_PROP_OVERLAY,
	ZFS_PROP_PREV_SNAP,
	ZFS_PROP_RECEIVE_RESUME_TOKEN,
	ZFS_PROP_ENCRYPTION,
	ZFS_PROP_KEYLOCATION,
	ZFS_PROP_KEYFORMAT,
	ZFS_PROP_PBKDF2_SALT,
	ZFS_PROP_PBKDF2_ITERS,
	ZFS_PROP_ENCRYPTION_ROOT,
	ZFS_PROP_KEY_GUID,
	ZFS_PROP_KEYSTATUS,
	ZFS_PROP_REMAPTXG, /* obsolete - no longer used */
	ZFS_PROP_SPECIAL_SMALL_BLOCKS,
	ZFS_PROP_IVSET_GUID, /* not exposed to the user */
	ZFS_PROP_REDACTED,
	ZFS_PROP_REDACT_SNAPS,
	ZFS_PROP_SNAPSHOTS_CHANGED,
	ZFS_NUM_PROPS
	} zfs_prop_t;

	typedef enum {
	ZFS_PROP_USERUSED,
	ZFS_PROP_USERQUOTA,
	ZFS_PROP_GROUPUSED,
	ZFS_PROP_GROUPQUOTA,
	ZFS_PROP_USEROBJUSED,
	ZFS_PROP_USEROBJQUOTA,
	ZFS_PROP_GROUPOBJUSED,
	ZFS_PROP_GROUPOBJQUOTA,
	ZFS_PROP_PROJECTUSED,
	ZFS_PROP_PROJECTQUOTA,
	ZFS_PROP_PROJECTOBJUSED,
	ZFS_PROP_PROJECTOBJQUOTA,
	ZFS_NUM_USERQUOTA_PROPS
	} zfs_userquota_prop_t;

	_SYS_FS_ZFS_H const char *const zfs_userquota_prop_prefixes[
	ZFS_NUM_USERQUOTA_PROPS];

	/*
	* Pool properties are identified by these constants and must be added to the
	* end of this list to ensure that external consumers are not affected
	* by the change. Properties must be registered in zfs_prop_init().
	*/
	typedef enum {
	ZPOOL_PROP_INVAL = -1,
	ZPOOL_PROP_NAME,
	ZPOOL_PROP_SIZE,
	ZPOOL_PROP_CAPACITY,
	ZPOOL_PROP_ALTROOT,
	ZPOOL_PROP_HEALTH,
	ZPOOL_PROP_GUID,
	ZPOOL_PROP_VERSION,
	ZPOOL_PROP_BOOTFS,
	ZPOOL_PROP_DELEGATION,
	ZPOOL_PROP_AUTOREPLACE,
	ZPOOL_PROP_CACHEFILE,
	ZPOOL_PROP_FAILUREMODE,
	ZPOOL_PROP_LISTSNAPS,
	ZPOOL_PROP_AUTOEXPAND,
	ZPOOL_PROP_DEDUPDITTO,
	ZPOOL_PROP_DEDUPRATIO,
	ZPOOL_PROP_FREE,
	ZPOOL_PROP_ALLOCATED,
	ZPOOL_PROP_READONLY,
	ZPOOL_PROP_ASHIFT,
	ZPOOL_PROP_COMMENT,
	ZPOOL_PROP_EXPANDSZ,
	ZPOOL_PROP_FREEING,
	ZPOOL_PROP_FRAGMENTATION,
	ZPOOL_PROP_LEAKED,
	ZPOOL_PROP_MAXBLOCKSIZE,
	ZPOOL_PROP_TNAME,
	ZPOOL_PROP_MAXDNODESIZE,
	ZPOOL_PROP_MULTIHOST,
	ZPOOL_PROP_CHECKPOINT,
	ZPOOL_PROP_LOAD_GUID,
	ZPOOL_PROP_AUTOTRIM,
	ZPOOL_PROP_COMPATIBILITY,
	ZPOOL_NUM_PROPS
	} zpool_prop_t;

	/* Small enough to not hog a whole line of printout in zpool(8). */
	#define ZPROP_MAX_COMMENT 32
	#define ZPROP_BOOLEAN_NA 2

	#define ZPROP_VALUE "value"
	#define ZPROP_SOURCE "source"

	typedef enum {
	ZPROP_SRC_NONE = 0x1,
	ZPROP_SRC_DEFAULT = 0x2,
	ZPROP_SRC_TEMPORARY = 0x4,
	ZPROP_SRC_LOCAL = 0x8,
	ZPROP_SRC_INHERITED = 0x10,
	ZPROP_SRC_RECEIVED = 0x20
	} zprop_source_t;

	#define ZPROP_SRC_ALL 0x3f

	#define ZPROP_SOURCE_VAL_RECVD "$recvd"
	#define ZPROP_N_MORE_ERRORS "N_MORE_ERRORS"

	/*
	* Dataset flag implemented as a special entry in the props zap object
	* indicating that the dataset has received properties on or after
	* SPA_VERSION_RECVD_PROPS. The first such receive blows away local properties
	* just as it did in earlier versions, and thereafter, local properties are
	* preserved.
	*/
	#define ZPROP_HAS_RECVD "$hasrecvd"

	typedef enum {
	ZPROP_ERR_NOCLEAR = 0x1, /* failure to clear existing props */
	ZPROP_ERR_NORESTORE = 0x2 /* failure to restore props on error */
	} zprop_errflags_t;

	typedef int (zprop_func)(int, void );

	/*
	* Properties to be set on the root file system of a new pool
	* are stuffed into their own nvlist, which is then included in
	* the properties nvlist with the pool properties.
	*/
	#define ZPOOL_ROOTFS_PROPS "root-props-nvl"

	/*
	* Length of 'written@' and 'written#'
	*/
	#define ZFS_WRITTEN_PROP_PREFIX_LEN 8

	/*
	* VDEV properties are identified by these constants and must be added to the
	* end of this list to ensure that external consumers are not affected
	* by the change. If you make any changes to this list, be sure to update
	* the property table in usr/src/common/zfs/zpool_prop.c.
	*/
	typedef enum {
	VDEV_PROP_INVAL = -1,
	VDEV_PROP_USERPROP = VDEV_PROP_INVAL,
	VDEV_PROP_NAME,
	VDEV_PROP_CAPACITY,
	VDEV_PROP_STATE,
	VDEV_PROP_GUID,
	VDEV_PROP_ASIZE,
	VDEV_PROP_PSIZE,
	VDEV_PROP_ASHIFT,
	VDEV_PROP_SIZE,
	VDEV_PROP_FREE,
	VDEV_PROP_ALLOCATED,
	VDEV_PROP_COMMENT,
	VDEV_PROP_EXPANDSZ,
	VDEV_PROP_FRAGMENTATION,
	VDEV_PROP_BOOTSIZE,
	VDEV_PROP_PARITY,
	VDEV_PROP_PATH,
	VDEV_PROP_DEVID,
	VDEV_PROP_PHYS_PATH,
	VDEV_PROP_ENC_PATH,
	VDEV_PROP_FRU,
	VDEV_PROP_PARENT,
	VDEV_PROP_CHILDREN,
	VDEV_PROP_NUMCHILDREN,
	VDEV_PROP_READ_ERRORS,
	VDEV_PROP_WRITE_ERRORS,
	VDEV_PROP_CHECKSUM_ERRORS,
	VDEV_PROP_INITIALIZE_ERRORS,
	VDEV_PROP_OPS_NULL,
	VDEV_PROP_OPS_READ,
	VDEV_PROP_OPS_WRITE,
	VDEV_PROP_OPS_FREE,
	VDEV_PROP_OPS_CLAIM,
	VDEV_PROP_OPS_TRIM,
	VDEV_PROP_BYTES_NULL,
	VDEV_PROP_BYTES_READ,
	VDEV_PROP_BYTES_WRITE,
	VDEV_PROP_BYTES_FREE,
	VDEV_PROP_BYTES_CLAIM,
	VDEV_PROP_BYTES_TRIM,
	VDEV_PROP_REMOVING,
	VDEV_PROP_ALLOCATING,
	VDEV_NUM_PROPS
	} vdev_prop_t;

	/*
	* Dataset property functions shared between libzfs and kernel.
	*/
	_SYS_FS_ZFS_H const char *zfs_prop_default_string(zfs_prop_t);
	_SYS_FS_ZFS_H uint64_t zfs_prop_default_numeric(zfs_prop_t);
	_SYS_FS_ZFS_H boolean_t zfs_prop_readonly(zfs_prop_t);
	_SYS_FS_ZFS_H boolean_t zfs_prop_visible(zfs_prop_t prop);
	_SYS_FS_ZFS_H boolean_t zfs_prop_inheritable(zfs_prop_t);
	_SYS_FS_ZFS_H boolean_t zfs_prop_setonce(zfs_prop_t);
	_SYS_FS_ZFS_H boolean_t zfs_prop_encryption_key_param(zfs_prop_t);
	_SYS_FS_ZFS_H boolean_t zfs_prop_valid_keylocation(const char *, boolean_t);
	_SYS_FS_ZFS_H const char *zfs_prop_to_name(zfs_prop_t);
	_SYS_FS_ZFS_H zfs_prop_t zfs_name_to_prop(const char *);
	_SYS_FS_ZFS_H boolean_t zfs_prop_user(const char *);
	_SYS_FS_ZFS_H boolean_t zfs_prop_userquota(const char *);
	_SYS_FS_ZFS_H boolean_t zfs_prop_written(const char *);
	_SYS_FS_ZFS_H int zfs_prop_index_to_string(zfs_prop_t, uint64_t, const char **);
	_SYS_FS_ZFS_H int zfs_prop_string_to_index(zfs_prop_t, const char *,
	uint64_t *);
	_SYS_FS_ZFS_H uint64_t zfs_prop_random_value(zfs_prop_t, uint64_t seed);
	_SYS_FS_ZFS_H boolean_t zfs_prop_valid_for_type(int, zfs_type_t, boolean_t);

	/*
	* Pool property functions shared between libzfs and kernel.
	*/
	_SYS_FS_ZFS_H zpool_prop_t zpool_name_to_prop(const char *);
	_SYS_FS_ZFS_H const char *zpool_prop_to_name(zpool_prop_t);
	_SYS_FS_ZFS_H const char *zpool_prop_default_string(zpool_prop_t);
	_SYS_FS_ZFS_H uint64_t zpool_prop_default_numeric(zpool_prop_t);
	_SYS_FS_ZFS_H boolean_t zpool_prop_readonly(zpool_prop_t);
	_SYS_FS_ZFS_H boolean_t zpool_prop_setonce(zpool_prop_t);
	_SYS_FS_ZFS_H boolean_t zpool_prop_feature(const char *);
	_SYS_FS_ZFS_H boolean_t zpool_prop_unsupported(const char *);
	_SYS_FS_ZFS_H int zpool_prop_index_to_string(zpool_prop_t, uint64_t,
	const char **);
	_SYS_FS_ZFS_H int zpool_prop_string_to_index(zpool_prop_t, const char *,
	uint64_t *);
	_SYS_FS_ZFS_H uint64_t zpool_prop_random_value(zpool_prop_t, uint64_t seed);

	/*
	* VDEV property functions shared between libzfs and kernel.
	*/
	_SYS_FS_ZFS_H vdev_prop_t vdev_name_to_prop(const char *);
	_SYS_FS_ZFS_H boolean_t vdev_prop_user(const char *name);
	_SYS_FS_ZFS_H const char *vdev_prop_to_name(vdev_prop_t);
	_SYS_FS_ZFS_H const char *vdev_prop_default_string(vdev_prop_t);
	_SYS_FS_ZFS_H uint64_t vdev_prop_default_numeric(vdev_prop_t);
	_SYS_FS_ZFS_H boolean_t vdev_prop_readonly(vdev_prop_t prop);
	_SYS_FS_ZFS_H int vdev_prop_index_to_string(vdev_prop_t, uint64_t,
	const char **);
	_SYS_FS_ZFS_H int vdev_prop_string_to_index(vdev_prop_t, const char *,
	uint64_t *);
	_SYS_FS_ZFS_H boolean_t zpool_prop_vdev(const char *name);
	_SYS_FS_ZFS_H uint64_t vdev_prop_random_value(vdev_prop_t prop, uint64_t seed);

	/*
	* Definitions for the Delegation.
	*/
	typedef enum {
	ZFS_DELEG_WHO_UNKNOWN = 0,
	ZFS_DELEG_USER = 'u',
	ZFS_DELEG_USER_SETS = 'U',
	ZFS_DELEG_GROUP = 'g',
	ZFS_DELEG_GROUP_SETS = 'G',
	ZFS_DELEG_EVERYONE = 'e',
	ZFS_DELEG_EVERYONE_SETS = 'E',
	ZFS_DELEG_CREATE = 'c',
	ZFS_DELEG_CREATE_SETS = 'C',
	ZFS_DELEG_NAMED_SET = 's',
	ZFS_DELEG_NAMED_SET_SETS = 'S'
	} zfs_deleg_who_type_t;

	typedef enum {
	ZFS_DELEG_NONE = 0,
	ZFS_DELEG_PERM_LOCAL = 1,
	ZFS_DELEG_PERM_DESCENDENT = 2,
	ZFS_DELEG_PERM_LOCALDESCENDENT = 3,
	ZFS_DELEG_PERM_CREATE = 4
	} zfs_deleg_inherit_t;

	#define ZFS_DELEG_PERM_UID "uid"
	#define ZFS_DELEG_PERM_GID "gid"
	#define ZFS_DELEG_PERM_GROUPS "groups"

	#define ZFS_MLSLABEL_DEFAULT "none"

	#define ZFS_SMB_ACL_SRC "src"
	#define ZFS_SMB_ACL_TARGET "target"

	typedef enum {
	ZFS_CANMOUNT_OFF = 0,
	ZFS_CANMOUNT_ON = 1,
	ZFS_CANMOUNT_NOAUTO = 2
	} zfs_canmount_type_t;

	typedef enum {
	ZFS_LOGBIAS_LATENCY = 0,
	ZFS_LOGBIAS_THROUGHPUT = 1
	} zfs_logbias_op_t;

	typedef enum zfs_share_op {
	ZFS_SHARE_NFS = 0,
	ZFS_UNSHARE_NFS = 1,
	ZFS_SHARE_SMB = 2,
	ZFS_UNSHARE_SMB = 3
	} zfs_share_op_t;

	typedef enum zfs_smb_acl_op {
	ZFS_SMB_ACL_ADD,
	ZFS_SMB_ACL_REMOVE,
	ZFS_SMB_ACL_RENAME,
	ZFS_SMB_ACL_PURGE
	} zfs_smb_acl_op_t;

	typedef enum zfs_cache_type {
	ZFS_CACHE_NONE = 0,
	ZFS_CACHE_METADATA = 1,
	ZFS_CACHE_ALL = 2
	} zfs_cache_type_t;

	typedef enum {
	ZFS_SYNC_STANDARD = 0,
	ZFS_SYNC_ALWAYS = 1,
	ZFS_SYNC_DISABLED = 2
	} zfs_sync_type_t;

	typedef enum {
	ZFS_XATTR_OFF = 0,
	ZFS_XATTR_DIR = 1,
	ZFS_XATTR_SA = 2
	} zfs_xattr_type_t;

	typedef enum {
	ZFS_DNSIZE_LEGACY = 0,
	ZFS_DNSIZE_AUTO = 1,
	ZFS_DNSIZE_1K = 1024,
	ZFS_DNSIZE_2K = 2048,
	ZFS_DNSIZE_4K = 4096,
	ZFS_DNSIZE_8K = 8192,
	ZFS_DNSIZE_16K = 16384
	} zfs_dnsize_type_t;

	typedef enum {
	ZFS_REDUNDANT_METADATA_ALL,
	ZFS_REDUNDANT_METADATA_MOST
	} zfs_redundant_metadata_type_t;

	typedef enum {
	ZFS_VOLMODE_DEFAULT = 0,
	ZFS_VOLMODE_GEOM = 1,
	ZFS_VOLMODE_DEV = 2,
	ZFS_VOLMODE_NONE = 3
	} zfs_volmode_t;

	typedef enum zfs_keystatus {
	ZFS_KEYSTATUS_NONE = 0,
	ZFS_KEYSTATUS_UNAVAILABLE,
	ZFS_KEYSTATUS_AVAILABLE,
	} zfs_keystatus_t;

	typedef enum zfs_keyformat {
	ZFS_KEYFORMAT_NONE = 0,
	ZFS_KEYFORMAT_RAW,
	ZFS_KEYFORMAT_HEX,
	ZFS_KEYFORMAT_PASSPHRASE,
	ZFS_KEYFORMAT_FORMATS
	} zfs_keyformat_t;

	typedef enum zfs_key_location {
	ZFS_KEYLOCATION_NONE = 0,
	ZFS_KEYLOCATION_PROMPT,
	ZFS_KEYLOCATION_URI,
	ZFS_KEYLOCATION_LOCATIONS
	} zfs_keylocation_t;

	#define DEFAULT_PBKDF2_ITERATIONS 350000
	#define MIN_PBKDF2_ITERATIONS 100000

	/*
	* On-disk version number.
	*/
	#define SPA_VERSION_1 1ULL
	#define SPA_VERSION_2 2ULL
	#define SPA_VERSION_3 3ULL
	#define SPA_VERSION_4 4ULL
	#define SPA_VERSION_5 5ULL
	#define SPA_VERSION_6 6ULL
	#define SPA_VERSION_7 7ULL
	#define SPA_VERSION_8 8ULL
	#define SPA_VERSION_9 9ULL
	#define SPA_VERSION_10 10ULL
	#define SPA_VERSION_11 11ULL
	#define SPA_VERSION_12 12ULL
	#define SPA_VERSION_13 13ULL
	#define SPA_VERSION_14 14ULL
	#define SPA_VERSION_15 15ULL
	#define SPA_VERSION_16 16ULL
	#define SPA_VERSION_17 17ULL
	#define SPA_VERSION_18 18ULL
	#define SPA_VERSION_19 19ULL
	#define SPA_VERSION_20 20ULL
	#define SPA_VERSION_21 21ULL
	#define SPA_VERSION_22 22ULL
	#define SPA_VERSION_23 23ULL
	#define SPA_VERSION_24 24ULL
	#define SPA_VERSION_25 25ULL
	#define SPA_VERSION_26 26ULL
	#define SPA_VERSION_27 27ULL
	#define SPA_VERSION_28 28ULL
	#define SPA_VERSION_5000 5000ULL

	/*
	* The incrementing pool version number has been replaced by pool feature
	* flags. For more details, see zfeature.c.
	*/
	#define SPA_VERSION SPA_VERSION_5000
	#define SPA_VERSION_STRING "5000"

	/*
	* Symbolic names for the changes that caused a SPA_VERSION switch.
	* Used in the code when checking for presence or absence of a feature.
	* Feel free to define multiple symbolic names for each version if there
	* were multiple changes to on-disk structures during that version.
	*
	* NOTE: When checking the current SPA_VERSION in your code, be sure
	* to use spa_version() since it reports the version of the
	* last synced uberblock. Checking the in-flight version can
	* be dangerous in some cases.
	*/
	#define SPA_VERSION_INITIAL SPA_VERSION_1
	#define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2
	#define SPA_VERSION_SPARES SPA_VERSION_3
	#define SPA_VERSION_RAIDZ2 SPA_VERSION_3
	#define SPA_VERSION_BPOBJ_ACCOUNT SPA_VERSION_3
	#define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3
	#define SPA_VERSION_DNODE_BYTES SPA_VERSION_3
	#define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4
	#define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5
	#define SPA_VERSION_BOOTFS SPA_VERSION_6
	#define SPA_VERSION_SLOGS SPA_VERSION_7
	#define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8
	#define SPA_VERSION_FUID SPA_VERSION_9
	#define SPA_VERSION_REFRESERVATION SPA_VERSION_9
	#define SPA_VERSION_REFQUOTA SPA_VERSION_9
	#define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9
	#define SPA_VERSION_L2CACHE SPA_VERSION_10
	#define SPA_VERSION_NEXT_CLONES SPA_VERSION_11
	#define SPA_VERSION_ORIGIN SPA_VERSION_11
	#define SPA_VERSION_DSL_SCRUB SPA_VERSION_11
	#define SPA_VERSION_SNAP_PROPS SPA_VERSION_12
	#define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13
	#define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14
	#define SPA_VERSION_USERSPACE SPA_VERSION_15
	#define SPA_VERSION_STMF_PROP SPA_VERSION_16
	#define SPA_VERSION_RAIDZ3 SPA_VERSION_17
	#define SPA_VERSION_USERREFS SPA_VERSION_18
	#define SPA_VERSION_HOLES SPA_VERSION_19
	#define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20
	#define SPA_VERSION_DEDUP SPA_VERSION_21
	#define SPA_VERSION_RECVD_PROPS SPA_VERSION_22
	#define SPA_VERSION_SLIM_ZIL SPA_VERSION_23
	#define SPA_VERSION_SA SPA_VERSION_24
	#define SPA_VERSION_SCAN SPA_VERSION_25
	#define SPA_VERSION_DIR_CLONES SPA_VERSION_26
	#define SPA_VERSION_DEADLISTS SPA_VERSION_26
	#define SPA_VERSION_FAST_SNAP SPA_VERSION_27
	#define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28
	#define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28
	#define SPA_VERSION_FEATURES SPA_VERSION_5000

	#define SPA_VERSION_IS_SUPPORTED(v) \
	(((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) \|\| \
	((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION))

	/*
	* ZPL version - rev'd whenever an incompatible on-disk format change
	* occurs. This is independent of SPA/DMU/ZAP versioning. You must
	* also update the version_table[] and help message in zfs_prop.c.
	*/
	#define ZPL_VERSION_1 1ULL
	#define ZPL_VERSION_2 2ULL
	#define ZPL_VERSION_3 3ULL
	#define ZPL_VERSION_4 4ULL
	#define ZPL_VERSION_5 5ULL
	#define ZPL_VERSION ZPL_VERSION_5
	#define ZPL_VERSION_STRING "5"

	#define ZPL_VERSION_INITIAL ZPL_VERSION_1
	#define ZPL_VERSION_DIRENT_TYPE ZPL_VERSION_2
	#define ZPL_VERSION_FUID ZPL_VERSION_3
	#define ZPL_VERSION_NORMALIZATION ZPL_VERSION_3
	#define ZPL_VERSION_SYSATTR ZPL_VERSION_3
	#define ZPL_VERSION_USERSPACE ZPL_VERSION_4
	#define ZPL_VERSION_SA ZPL_VERSION_5

	/* Persistent L2ARC version */
	#define L2ARC_PERSISTENT_VERSION_1 1ULL
	#define L2ARC_PERSISTENT_VERSION L2ARC_PERSISTENT_VERSION_1
	#define L2ARC_PERSISTENT_VERSION_STRING "1"

	/* Rewind policy information */
	#define ZPOOL_NO_REWIND 1 /* No policy - default behavior */
	#define ZPOOL_NEVER_REWIND 2 /* Do not search for best txg or rewind */
	#define ZPOOL_TRY_REWIND 4 /* Search for best txg, but do not rewind */
	#define ZPOOL_DO_REWIND 8 /* Rewind to best txg w/in deferred frees */
	#define ZPOOL_EXTREME_REWIND 16 /* Allow extreme measures to find best txg */
	#define ZPOOL_REWIND_MASK 28 /* All the possible rewind bits */
	#define ZPOOL_REWIND_POLICIES 31 /* All the possible policy bits */

	typedef struct zpool_load_policy {
	uint32_t zlp_rewind; /* rewind policy requested */
	uint64_t zlp_maxmeta; /* max acceptable meta-data errors */
	uint64_t zlp_maxdata; /* max acceptable data errors */
	uint64_t zlp_txg; /* specific txg to load */
	} zpool_load_policy_t;

	/*
	* The following are configuration names used in the nvlist describing a pool's
	* configuration. New on-disk names should be prefixed with "<reversed-DNS>:"
	* (e.g. "org.openzfs:") to avoid conflicting names being developed
	* independently.
	*/
	#define ZPOOL_CONFIG_VERSION "version"
	#define ZPOOL_CONFIG_POOL_NAME "name"
	#define ZPOOL_CONFIG_POOL_STATE "state"
	#define ZPOOL_CONFIG_POOL_TXG "txg"
	#define ZPOOL_CONFIG_POOL_GUID "pool_guid"
	#define ZPOOL_CONFIG_CREATE_TXG "create_txg"
	#define ZPOOL_CONFIG_TOP_GUID "top_guid"
	#define ZPOOL_CONFIG_VDEV_TREE "vdev_tree"
	#define ZPOOL_CONFIG_TYPE "type"
	#define ZPOOL_CONFIG_CHILDREN "children"
	#define ZPOOL_CONFIG_ID "id"
	#define ZPOOL_CONFIG_GUID "guid"
	#define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object"
	#define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births"
	#define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev"
	#define ZPOOL_CONFIG_PATH "path"
	#define ZPOOL_CONFIG_DEVID "devid"
	#define ZPOOL_CONFIG_SPARE_ID "spareid"
	#define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array"
	#define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift"
	#define ZPOOL_CONFIG_ASHIFT "ashift"
	#define ZPOOL_CONFIG_ASIZE "asize"
	#define ZPOOL_CONFIG_DTL "DTL"
	#define ZPOOL_CONFIG_SCAN_STATS "scan_stats" /* not stored on disk */
	#define ZPOOL_CONFIG_REMOVAL_STATS "removal_stats" /* not stored on disk */
	#define ZPOOL_CONFIG_CHECKPOINT_STATS "checkpoint_stats" /* not on disk */
	#define ZPOOL_CONFIG_VDEV_STATS "vdev_stats" /* not stored on disk */
	#define ZPOOL_CONFIG_INDIRECT_SIZE "indirect_size" /* not stored on disk */

	/* container nvlist of extended stats */
	#define ZPOOL_CONFIG_VDEV_STATS_EX "vdev_stats_ex"

	/* Active queue read/write stats */
	#define ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE "vdev_sync_r_active_queue"
	#define ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE "vdev_sync_w_active_queue"
	#define ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE "vdev_async_r_active_queue"
	#define ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE "vdev_async_w_active_queue"
	#define ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE "vdev_async_scrub_active_queue"
	#define ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE "vdev_async_trim_active_queue"
	#define ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE "vdev_rebuild_active_queue"

	/* Queue sizes */
	#define ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE "vdev_sync_r_pend_queue"
	#define ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE "vdev_sync_w_pend_queue"
	#define ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE "vdev_async_r_pend_queue"
	#define ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE "vdev_async_w_pend_queue"
	#define ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE "vdev_async_scrub_pend_queue"
	#define ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE "vdev_async_trim_pend_queue"
	#define ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE "vdev_rebuild_pend_queue"

	/* Latency read/write histogram stats */
	#define ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO "vdev_tot_r_lat_histo"
	#define ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO "vdev_tot_w_lat_histo"
	#define ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO "vdev_disk_r_lat_histo"
	#define ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO "vdev_disk_w_lat_histo"
	#define ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO "vdev_sync_r_lat_histo"
	#define ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO "vdev_sync_w_lat_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO "vdev_async_r_lat_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO "vdev_async_w_lat_histo"
	#define ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO "vdev_scrub_histo"
	#define ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO "vdev_trim_histo"
	#define ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO "vdev_rebuild_histo"

	/* Request size histograms */
	#define ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO "vdev_sync_ind_r_histo"
	#define ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO "vdev_sync_ind_w_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO "vdev_async_ind_r_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO "vdev_async_ind_w_histo"
	#define ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO "vdev_ind_scrub_histo"
	#define ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO "vdev_ind_trim_histo"
	#define ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO "vdev_ind_rebuild_histo"
	#define ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO "vdev_sync_agg_r_histo"
	#define ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO "vdev_sync_agg_w_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO "vdev_async_agg_r_histo"
	#define ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO "vdev_async_agg_w_histo"
	#define ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO "vdev_agg_scrub_histo"
	#define ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO "vdev_agg_trim_histo"
	#define ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO "vdev_agg_rebuild_histo"

	/* Number of slow IOs */
	#define ZPOOL_CONFIG_VDEV_SLOW_IOS "vdev_slow_ios"

	/* vdev enclosure sysfs path */
	#define ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH "vdev_enc_sysfs_path"

	#define ZPOOL_CONFIG_WHOLE_DISK "whole_disk"
	#define ZPOOL_CONFIG_ERRCOUNT "error_count"
	#define ZPOOL_CONFIG_NOT_PRESENT "not_present"
	#define ZPOOL_CONFIG_SPARES "spares"
	#define ZPOOL_CONFIG_IS_SPARE "is_spare"
	#define ZPOOL_CONFIG_NPARITY "nparity"
	#define ZPOOL_CONFIG_HOSTID "hostid"
	#define ZPOOL_CONFIG_HOSTNAME "hostname"
	#define ZPOOL_CONFIG_LOADED_TIME "initial_load_time"
	#define ZPOOL_CONFIG_UNSPARE "unspare"
	#define ZPOOL_CONFIG_PHYS_PATH "phys_path"
	#define ZPOOL_CONFIG_IS_LOG "is_log"
	#define ZPOOL_CONFIG_L2CACHE "l2cache"
	#define ZPOOL_CONFIG_HOLE_ARRAY "hole_array"
	#define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children"
	#define ZPOOL_CONFIG_IS_HOLE "is_hole"
	#define ZPOOL_CONFIG_DDT_HISTOGRAM "ddt_histogram"
	#define ZPOOL_CONFIG_DDT_OBJ_STATS "ddt_object_stats"
	#define ZPOOL_CONFIG_DDT_STATS "ddt_stats"
	#define ZPOOL_CONFIG_SPLIT "splitcfg"
	#define ZPOOL_CONFIG_ORIG_GUID "orig_guid"
	#define ZPOOL_CONFIG_SPLIT_GUID "split_guid"
	#define ZPOOL_CONFIG_SPLIT_LIST "guid_list"
	#define ZPOOL_CONFIG_NONALLOCATING "non_allocating"
	#define ZPOOL_CONFIG_REMOVING "removing"
	#define ZPOOL_CONFIG_RESILVER_TXG "resilver_txg"
	#define ZPOOL_CONFIG_REBUILD_TXG "rebuild_txg"
	#define ZPOOL_CONFIG_COMMENT "comment"
	#define ZPOOL_CONFIG_SUSPENDED "suspended" /* not stored on disk */
	#define ZPOOL_CONFIG_SUSPENDED_REASON "suspended_reason" /* not stored */
	#define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */
	#define ZPOOL_CONFIG_BOOTFS "bootfs" /* not stored on disk */
	#define ZPOOL_CONFIG_MISSING_DEVICES "missing_vdevs" /* not stored on disk */
	#define ZPOOL_CONFIG_LOAD_INFO "load_info" /* not stored on disk */
	#define ZPOOL_CONFIG_REWIND_INFO "rewind_info" /* not stored on disk */
	#define ZPOOL_CONFIG_UNSUP_FEAT "unsup_feat" /* not stored on disk */
	#define ZPOOL_CONFIG_ENABLED_FEAT "enabled_feat" /* not stored on disk */
	#define ZPOOL_CONFIG_CAN_RDONLY "can_rdonly" /* not stored on disk */
	#define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read"
	#define ZPOOL_CONFIG_FEATURE_STATS "feature_stats" /* not stored on disk */
	#define ZPOOL_CONFIG_ERRATA "errata" /* not stored on disk */
	#define ZPOOL_CONFIG_VDEV_TOP_ZAP "com.delphix:vdev_zap_top"
	#define ZPOOL_CONFIG_VDEV_LEAF_ZAP "com.delphix:vdev_zap_leaf"
	#define ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS "com.delphix:has_per_vdev_zaps"
	#define ZPOOL_CONFIG_RESILVER_DEFER "com.datto:resilver_defer"
	#define ZPOOL_CONFIG_CACHEFILE "cachefile" /* not stored on disk */
	#define ZPOOL_CONFIG_MMP_STATE "mmp_state" /* not stored on disk */
	#define ZPOOL_CONFIG_MMP_TXG "mmp_txg" /* not stored on disk */
	#define ZPOOL_CONFIG_MMP_SEQ "mmp_seq" /* not stored on disk */
	#define ZPOOL_CONFIG_MMP_HOSTNAME "mmp_hostname" /* not stored on disk */
	#define ZPOOL_CONFIG_MMP_HOSTID "mmp_hostid" /* not stored on disk */
	#define ZPOOL_CONFIG_ALLOCATION_BIAS "alloc_bias" /* not stored on disk */
	#define ZPOOL_CONFIG_EXPANSION_TIME "expansion_time" /* not stored */
	#define ZPOOL_CONFIG_REBUILD_STATS "org.openzfs:rebuild_stats"
	#define ZPOOL_CONFIG_COMPATIBILITY "compatibility"

	/*
	* The persistent vdev state is stored as separate values rather than a single
	* 'vdev_state' entry. This is because a device can be in multiple states, such
	* as offline and degraded.
	*/
	#define ZPOOL_CONFIG_OFFLINE "offline"
	#define ZPOOL_CONFIG_FAULTED "faulted"
	#define ZPOOL_CONFIG_DEGRADED "degraded"
	#define ZPOOL_CONFIG_REMOVED "removed"
	#define ZPOOL_CONFIG_FRU "fru"
	#define ZPOOL_CONFIG_AUX_STATE "aux_state"

	/* Pool load policy parameters */
	#define ZPOOL_LOAD_POLICY "load-policy"
	#define ZPOOL_LOAD_REWIND_POLICY "load-rewind-policy"
	#define ZPOOL_LOAD_REQUEST_TXG "load-request-txg"
	#define ZPOOL_LOAD_META_THRESH "load-meta-thresh"
	#define ZPOOL_LOAD_DATA_THRESH "load-data-thresh"

	/* Rewind data discovered */
	#define ZPOOL_CONFIG_LOAD_TIME "rewind_txg_ts"
	#define ZPOOL_CONFIG_LOAD_META_ERRORS "verify_meta_errors"
	#define ZPOOL_CONFIG_LOAD_DATA_ERRORS "verify_data_errors"
	#define ZPOOL_CONFIG_REWIND_TIME "seconds_of_rewind"

	/* dRAID configuration */
	#define ZPOOL_CONFIG_DRAID_NDATA "draid_ndata"
	#define ZPOOL_CONFIG_DRAID_NSPARES "draid_nspares"
	#define ZPOOL_CONFIG_DRAID_NGROUPS "draid_ngroups"

	#define VDEV_TYPE_ROOT "root"
	#define VDEV_TYPE_MIRROR "mirror"
	#define VDEV_TYPE_REPLACING "replacing"
	#define VDEV_TYPE_RAIDZ "raidz"
	#define VDEV_TYPE_DRAID "draid"
	#define VDEV_TYPE_DRAID_SPARE "dspare"
	#define VDEV_TYPE_DISK "disk"
	#define VDEV_TYPE_FILE "file"
	#define VDEV_TYPE_MISSING "missing"
	#define VDEV_TYPE_HOLE "hole"
	#define VDEV_TYPE_SPARE "spare"
	#define VDEV_TYPE_LOG "log"
	#define VDEV_TYPE_L2CACHE "l2cache"
	#define VDEV_TYPE_INDIRECT "indirect"

	#define VDEV_RAIDZ_MAXPARITY 3

	#define VDEV_DRAID_MAXPARITY 3
	#define VDEV_DRAID_MIN_CHILDREN 2
	#define VDEV_DRAID_MAX_CHILDREN UINT8_MAX

	/* VDEV_TOP_ZAP_* are used in top-level vdev ZAP objects. */
	#define VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM \
	"com.delphix:indirect_obsolete_sm"
	#define VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE \
	"com.delphix:obsolete_counts_are_precise"
	#define VDEV_TOP_ZAP_POOL_CHECKPOINT_SM \
	"com.delphix:pool_checkpoint_sm"
	#define VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS \
	"com.delphix:ms_unflushed_phys_txgs"

	#define VDEV_TOP_ZAP_VDEV_REBUILD_PHYS \
	"org.openzfs:vdev_rebuild"

	#define VDEV_TOP_ZAP_ALLOCATION_BIAS \
	"org.zfsonlinux:allocation_bias"

	/* vdev metaslab allocation bias */
	#define VDEV_ALLOC_BIAS_LOG "log"
	#define VDEV_ALLOC_BIAS_SPECIAL "special"
	#define VDEV_ALLOC_BIAS_DEDUP "dedup"

	/* vdev initialize state */
	#define VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET \
	"com.delphix:next_offset_to_initialize"
	#define VDEV_LEAF_ZAP_INITIALIZE_STATE \
	"com.delphix:vdev_initialize_state"
	#define VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME \
	"com.delphix:vdev_initialize_action_time"

	/* vdev TRIM state */
	#define VDEV_LEAF_ZAP_TRIM_LAST_OFFSET \
	"org.zfsonlinux:next_offset_to_trim"
	#define VDEV_LEAF_ZAP_TRIM_STATE \
	"org.zfsonlinux:vdev_trim_state"
	#define VDEV_LEAF_ZAP_TRIM_ACTION_TIME \
	"org.zfsonlinux:vdev_trim_action_time"
	#define VDEV_LEAF_ZAP_TRIM_RATE \
	"org.zfsonlinux:vdev_trim_rate"
	#define VDEV_LEAF_ZAP_TRIM_PARTIAL \
	"org.zfsonlinux:vdev_trim_partial"
	#define VDEV_LEAF_ZAP_TRIM_SECURE \
	"org.zfsonlinux:vdev_trim_secure"

	/*
	* This is needed in userland to report the minimum necessary device size.
	*/
	#define SPA_MINDEVSIZE (64ULL << 20)

	/*
	* Set if the fragmentation has not yet been calculated. This can happen
	* because the space maps have not been upgraded or the histogram feature
	* is not enabled.
	*/
	#define ZFS_FRAG_INVALID UINT64_MAX

	/*
	* The location of the pool configuration repository, shared between kernel and
	* userland.
	*/
	#define ZPOOL_CACHE_BOOT "/boot/zfs/zpool.cache"
	#define ZPOOL_CACHE "/etc/zfs/zpool.cache"
	/*
	* Settings for zpool compatibility features files
	*/
	#define ZPOOL_SYSCONF_COMPAT_D SYSCONFDIR "/zfs/compatibility.d"
	#define ZPOOL_DATA_COMPAT_D PKGDATADIR "/compatibility.d"
	#define ZPOOL_COMPAT_MAXSIZE 16384

	/*
	* Hard-wired compatibility settings
	*/
	#define ZPOOL_COMPAT_LEGACY "legacy"
	#define ZPOOL_COMPAT_OFF "off"

	/*
	* vdev states are ordered from least to most healthy.
	* A vdev that's CANT_OPEN or below is considered unusable.
	*/
	typedef enum vdev_state {
	VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */
	VDEV_STATE_CLOSED, /* Not currently open */
	VDEV_STATE_OFFLINE, /* Not allowed to open */
	VDEV_STATE_REMOVED, /* Explicitly removed from system */
	VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */
	VDEV_STATE_FAULTED, /* External request to fault device */
	VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */
	VDEV_STATE_HEALTHY /* Presumed good */
	} vdev_state_t;

	#define VDEV_STATE_ONLINE VDEV_STATE_HEALTHY

	/*
	* vdev aux states. When a vdev is in the CANT_OPEN state, the aux field
	* of the vdev stats structure uses these constants to distinguish why.
	*/
	typedef enum vdev_aux {
	VDEV_AUX_NONE, /* no error */
	VDEV_AUX_OPEN_FAILED, /* ldi_open_() or vn_open() failed /
	VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */
	VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */
	VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */
	VDEV_AUX_TOO_SMALL, /* vdev size is too small */
	VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */
	VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */
	VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */
	VDEV_AUX_UNSUP_FEAT, /* unsupported features */
	VDEV_AUX_SPARED, /* hot spare used in another pool */
	VDEV_AUX_ERR_EXCEEDED, /* too many errors */
	VDEV_AUX_IO_FAILURE, /* experienced I/O failure */
	VDEV_AUX_BAD_LOG, /* cannot read log chain(s) */
	VDEV_AUX_EXTERNAL, /* external diagnosis or forced fault */
	VDEV_AUX_SPLIT_POOL, /* vdev was split off into another pool */
	VDEV_AUX_BAD_ASHIFT, /* vdev ashift is invalid */
	VDEV_AUX_EXTERNAL_PERSIST, /* persistent forced fault */
	VDEV_AUX_ACTIVE, /* vdev active on a different host */
	VDEV_AUX_CHILDREN_OFFLINE, /* all children are offline */
	VDEV_AUX_ASHIFT_TOO_BIG, /* vdev's min block size is too large */
	} vdev_aux_t;

	/*
	* pool state. The following states are written to disk as part of the normal
	* SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE, L2CACHE. The remaining
	* states are software abstractions used at various levels to communicate
	* pool state.
	*/
	typedef enum pool_state {
	POOL_STATE_ACTIVE = 0, /* In active use */
	POOL_STATE_EXPORTED, /* Explicitly exported */
	POOL_STATE_DESTROYED, /* Explicitly destroyed */
	POOL_STATE_SPARE, /* Reserved for hot spare use */
	POOL_STATE_L2CACHE, /* Level 2 ARC device */
	POOL_STATE_UNINITIALIZED, /* Internal spa_t state */
	POOL_STATE_UNAVAIL, /* Internal libzfs state */
	POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */
	} pool_state_t;

	/*
	* mmp state. The following states provide additional detail describing
	* why a pool couldn't be safely imported.
	*/
	typedef enum mmp_state {
	MMP_STATE_ACTIVE = 0, /* In active use */
	MMP_STATE_INACTIVE, /* Inactive and safe to import */
	MMP_STATE_NO_HOSTID /* System hostid is not set */
	} mmp_state_t;

	/*
	* Scan Functions.
	*/
	typedef enum pool_scan_func {
	POOL_SCAN_NONE,
	POOL_SCAN_SCRUB,
	POOL_SCAN_RESILVER,
	POOL_SCAN_FUNCS
	} pool_scan_func_t;

	/*
	* Used to control scrub pause and resume.
	*/
	typedef enum pool_scrub_cmd {
	POOL_SCRUB_NORMAL = 0,
	POOL_SCRUB_PAUSE,
	POOL_SCRUB_FLAGS_END
	} pool_scrub_cmd_t;

	typedef enum {
	CS_NONE,
	CS_CHECKPOINT_EXISTS,
	CS_CHECKPOINT_DISCARDING,
	CS_NUM_STATES
	} checkpoint_state_t;

	typedef struct pool_checkpoint_stat {
	uint64_t pcs_state; /* checkpoint_state_t */
	uint64_t pcs_start_time; /* time checkpoint/discard started */
	uint64_t pcs_space; /* checkpointed space */
	} pool_checkpoint_stat_t;

	/*
	* ZIO types. Needed to interpret vdev statistics below.
	*/
	typedef enum zio_type {
	ZIO_TYPE_NULL = 0,
	ZIO_TYPE_READ,
	ZIO_TYPE_WRITE,
	ZIO_TYPE_FREE,
	ZIO_TYPE_CLAIM,
	ZIO_TYPE_IOCTL,
	ZIO_TYPE_TRIM,
	ZIO_TYPES
	} zio_type_t;

	/*
	* Pool statistics. Note: all fields should be 64-bit because this
	* is passed between kernel and userland as an nvlist uint64 array.
	*/
	typedef struct pool_scan_stat {
	/* values stored on disk */
	uint64_t pss_func; /* pool_scan_func_t */
	uint64_t pss_state; /* dsl_scan_state_t */
	uint64_t pss_start_time; /* scan start time */
	uint64_t pss_end_time; /* scan end time */
	uint64_t pss_to_examine; /* total bytes to scan */
	uint64_t pss_examined; /* total bytes located by scanner */
	uint64_t pss_to_process; /* total bytes to process */
	uint64_t pss_processed; /* total processed bytes */
	uint64_t pss_errors; /* scan errors */

	/* values not stored on disk */
	uint64_t pss_pass_exam; /* examined bytes per scan pass */
	uint64_t pss_pass_start; /* start time of a scan pass */
	uint64_t pss_pass_scrub_pause; /* pause time of a scrub pass */
	/* cumulative time scrub spent paused, needed for rate calculation */
	uint64_t pss_pass_scrub_spent_paused;
	uint64_t pss_pass_issued; /* issued bytes per scan pass */
	uint64_t pss_issued; /* total bytes checked by scanner */
	} pool_scan_stat_t;

	typedef struct pool_removal_stat {
	uint64_t prs_state; /* dsl_scan_state_t */
	uint64_t prs_removing_vdev;
	uint64_t prs_start_time;
	uint64_t prs_end_time;
	uint64_t prs_to_copy; /* bytes that need to be copied */
	uint64_t prs_copied; /* bytes copied so far */
	/*
	* bytes of memory used for indirect mappings.
	* This includes all removed vdevs.
	*/
	uint64_t prs_mapping_memory;
	} pool_removal_stat_t;

	typedef enum dsl_scan_state {
	DSS_NONE,
	DSS_SCANNING,
	DSS_FINISHED,
	DSS_CANCELED,
	DSS_NUM_STATES
	} dsl_scan_state_t;

	typedef struct vdev_rebuild_stat {
	uint64_t vrs_state; /* vdev_rebuild_state_t */
	uint64_t vrs_start_time; /* time_t */
	uint64_t vrs_end_time; /* time_t */
	uint64_t vrs_scan_time_ms; /* total run time (millisecs) */
	uint64_t vrs_bytes_scanned; /* allocated bytes scanned */
	uint64_t vrs_bytes_issued; /* read bytes issued */
	uint64_t vrs_bytes_rebuilt; /* rebuilt bytes */
	uint64_t vrs_bytes_est; /* total bytes to scan */
	uint64_t vrs_errors; /* scanning errors */
	uint64_t vrs_pass_time_ms; /* pass run time (millisecs) */
	uint64_t vrs_pass_bytes_scanned; /* bytes scanned since start/resume */
	uint64_t vrs_pass_bytes_issued; /* bytes rebuilt since start/resume */
	} vdev_rebuild_stat_t;

	/*
	* Errata described by https://openzfs.github.io/openzfs-docs/msg/ZFS-8000-ER.
	* The ordering of this enum must be maintained to ensure the errata identifiers
	* map to the correct documentation. New errata may only be appended to the
	* list and must contain corresponding documentation at the above link.
	*/
	typedef enum zpool_errata {
	ZPOOL_ERRATA_NONE,
	ZPOOL_ERRATA_ZOL_2094_SCRUB,
	ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY,
	ZPOOL_ERRATA_ZOL_6845_ENCRYPTION,
	ZPOOL_ERRATA_ZOL_8308_ENCRYPTION,
	} zpool_errata_t;

	/*
	* Vdev statistics. Note: all fields should be 64-bit because this
	* is passed between kernel and user land as an nvlist uint64 array.
	*
	* The vs_ops[] and vs_bytes[] arrays must always be an array size of 6 in
	* order to keep subsequent members at their known fixed offsets. When
	* adding a new field it must be added to the end the structure.
	*/
	#define VS_ZIO_TYPES 6

	typedef struct vdev_stat {
	hrtime_t vs_timestamp; /* time since vdev load */
	uint64_t vs_state; /* vdev state */
	uint64_t vs_aux; /* see vdev_aux_t */
	uint64_t vs_alloc; /* space allocated */
	uint64_t vs_space; /* total capacity */
	uint64_t vs_dspace; /* deflated capacity */
	uint64_t vs_rsize; /* replaceable dev size */
	uint64_t vs_esize; /* expandable dev size */
	uint64_t vs_ops[VS_ZIO_TYPES]; /* operation count */
	uint64_t vs_bytes[VS_ZIO_TYPES]; /* bytes read/written */
	uint64_t vs_read_errors; /* read errors */
	uint64_t vs_write_errors; /* write errors */
	uint64_t vs_checksum_errors; /* checksum errors */
	uint64_t vs_initialize_errors; /* initializing errors */
	uint64_t vs_self_healed; /* self-healed bytes */
	uint64_t vs_scan_removing; /* removing? */
	uint64_t vs_scan_processed; /* scan processed bytes */
	uint64_t vs_fragmentation; /* device fragmentation */
	uint64_t vs_initialize_bytes_done; /* bytes initialized */
	uint64_t vs_initialize_bytes_est; /* total bytes to initialize */
	uint64_t vs_initialize_state; /* vdev_initializing_state_t */
	uint64_t vs_initialize_action_time; /* time_t */
	uint64_t vs_checkpoint_space; /* checkpoint-consumed space */
	uint64_t vs_resilver_deferred; /* resilver deferred */
	uint64_t vs_slow_ios; /* slow IOs */
	uint64_t vs_trim_errors; /* trimming errors */
	uint64_t vs_trim_notsup; /* supported by device */
	uint64_t vs_trim_bytes_done; /* bytes trimmed */
	uint64_t vs_trim_bytes_est; /* total bytes to trim */
	uint64_t vs_trim_state; /* vdev_trim_state_t */
	uint64_t vs_trim_action_time; /* time_t */
	uint64_t vs_rebuild_processed; /* bytes rebuilt */
	uint64_t vs_configured_ashift; /* TLV vdev_ashift */
	uint64_t vs_logical_ashift; /* vdev_logical_ashift */
	uint64_t vs_physical_ashift; /* vdev_physical_ashift */
	uint64_t vs_noalloc; /* allocations halted? */
	uint64_t vs_pspace; /* physical capacity */
	} vdev_stat_t;

	#define VDEV_STAT_VALID(field, uint64_t_field_count) \
	((uint64_t_field_count * sizeof (uint64_t)) >= \
	(offsetof(vdev_stat_t, field) + sizeof (((vdev_stat_t *)NULL)->field)))

	/*
	* Extended stats
	*
	* These are stats which aren't included in the original iostat output. For
	* convenience, they are grouped together in vdev_stat_ex, although each stat
	* is individually exported as an nvlist.
	*/
	typedef struct vdev_stat_ex {
	/* Number of ZIOs issued to disk and waiting to finish */
	uint64_t vsx_active_queue[ZIO_PRIORITY_NUM_QUEUEABLE];

	/* Number of ZIOs pending to be issued to disk */
	uint64_t vsx_pend_queue[ZIO_PRIORITY_NUM_QUEUEABLE];

	/*
	* Below are the histograms for various latencies. Buckets are in
	* units of nanoseconds.
	*/

	/*
	* 2^37 nanoseconds = 134s. Timeouts will probably start kicking in
	* before this.
	*/
	#define VDEV_L_HISTO_BUCKETS 37 /* Latency histo buckets */
	#define VDEV_RQ_HISTO_BUCKETS 25 /* Request size histo buckets */

	/* Amount of time in ZIO queue (ns) */
	uint64_t vsx_queue_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
	[VDEV_L_HISTO_BUCKETS];

	/* Total ZIO latency (ns). Includes queuing and disk access time */
	uint64_t vsx_total_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];

	/* Amount of time to read/write the disk (ns) */
	uint64_t vsx_disk_histo[ZIO_TYPES][VDEV_L_HISTO_BUCKETS];

	/* "lookup the bucket for a value" histogram macros */
	#define HISTO(val, buckets) (val != 0 ? MIN(highbit64(val) - 1, \
	buckets - 1) : 0)
	#define L_HISTO(a) HISTO(a, VDEV_L_HISTO_BUCKETS)
	#define RQ_HISTO(a) HISTO(a, VDEV_RQ_HISTO_BUCKETS)

	/* Physical IO histogram */
	uint64_t vsx_ind_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
	[VDEV_RQ_HISTO_BUCKETS];

	/* Delegated (aggregated) physical IO histogram */
	uint64_t vsx_agg_histo[ZIO_PRIORITY_NUM_QUEUEABLE]
	[VDEV_RQ_HISTO_BUCKETS];

	} vdev_stat_ex_t;

	/*
	* Initialize functions.
	*/
	typedef enum pool_initialize_func {
	POOL_INITIALIZE_START,
	POOL_INITIALIZE_CANCEL,
	POOL_INITIALIZE_SUSPEND,
	POOL_INITIALIZE_FUNCS
	} pool_initialize_func_t;

	/*
	* TRIM functions.
	*/
	typedef enum pool_trim_func {
	POOL_TRIM_START,
	POOL_TRIM_CANCEL,
	POOL_TRIM_SUSPEND,
	POOL_TRIM_FUNCS
	} pool_trim_func_t;

	/*
	* DDT statistics. Note: all fields should be 64-bit because this
	* is passed between kernel and userland as an nvlist uint64 array.
	*/
	typedef struct ddt_object {
	uint64_t ddo_count; /* number of elements in ddt */
	uint64_t ddo_dspace; /* size of ddt on disk */
	uint64_t ddo_mspace; /* size of ddt in-core */
	} ddt_object_t;

	typedef struct ddt_stat {
	uint64_t dds_blocks; /* blocks */
	uint64_t dds_lsize; /* logical size */
	uint64_t dds_psize; /* physical size */
	uint64_t dds_dsize; /* deflated allocated size */
	uint64_t dds_ref_blocks; /* referenced blocks */
	uint64_t dds_ref_lsize; /* referenced lsize * refcnt */
	uint64_t dds_ref_psize; /* referenced psize * refcnt */
	uint64_t dds_ref_dsize; /* referenced dsize * refcnt */
	} ddt_stat_t;

	typedef struct ddt_histogram {
	ddt_stat_t ddh_stat[64]; /* power-of-two histogram buckets */
	} ddt_histogram_t;

	#define ZVOL_DRIVER "zvol"
	#define ZFS_DRIVER "zfs"
	#define ZFS_DEV "/dev/zfs"
	#define ZFS_DEVDIR "/dev"

	#define ZFS_SUPER_MAGIC 0x2fc12fc1

	/* general zvol path */
	#define ZVOL_DIR "/dev/zvol/"

	#define ZVOL_MAJOR 230
	#define ZVOL_MINOR_BITS 4
	#define ZVOL_MINOR_MASK ((1U << ZVOL_MINOR_BITS) - 1)
	#define ZVOL_MINORS (1 << 4)
	#define ZVOL_DEV_NAME "zd"

	#define ZVOL_PROP_NAME "name"
	#define ZVOL_DEFAULT_BLOCKSIZE 16384

	typedef enum {
	VDEV_INITIALIZE_NONE,
	VDEV_INITIALIZE_ACTIVE,
	VDEV_INITIALIZE_CANCELED,
	VDEV_INITIALIZE_SUSPENDED,
	VDEV_INITIALIZE_COMPLETE
	} vdev_initializing_state_t;

	typedef enum {
	VDEV_TRIM_NONE,
	VDEV_TRIM_ACTIVE,
	VDEV_TRIM_CANCELED,
	VDEV_TRIM_SUSPENDED,
	VDEV_TRIM_COMPLETE,
	} vdev_trim_state_t;

	typedef enum {
	VDEV_REBUILD_NONE,
	VDEV_REBUILD_ACTIVE,
	VDEV_REBUILD_CANCELED,
	VDEV_REBUILD_COMPLETE,
	} vdev_rebuild_state_t;

	/*
	* nvlist name constants. Facilitate restricting snapshot iteration range for
	* the "list next snapshot" ioctl
	*/
	#define SNAP_ITER_MIN_TXG "snap_iter_min_txg"
	#define SNAP_ITER_MAX_TXG "snap_iter_max_txg"

	/*
	* /dev/zfs ioctl numbers.
	*
	* These numbers cannot change over time. New ioctl numbers must be appended.
	*/
	typedef enum zfs_ioc {
	/*
	* Core features - 81/128 numbers reserved.
	*/
	#ifdef __FreeBSD__
	ZFS_IOC_FIRST = 0,
	#else
	ZFS_IOC_FIRST = ('Z' << 8),
	#endif
	ZFS_IOC = ZFS_IOC_FIRST,
	ZFS_IOC_POOL_CREATE = ZFS_IOC_FIRST, /* 0x5a00 */
	ZFS_IOC_POOL_DESTROY, /* 0x5a01 */
	ZFS_IOC_POOL_IMPORT, /* 0x5a02 */
	ZFS_IOC_POOL_EXPORT, /* 0x5a03 */
	ZFS_IOC_POOL_CONFIGS, /* 0x5a04 */
	ZFS_IOC_POOL_STATS, /* 0x5a05 */
	ZFS_IOC_POOL_TRYIMPORT, /* 0x5a06 */
	ZFS_IOC_POOL_SCAN, /* 0x5a07 */
	ZFS_IOC_POOL_FREEZE, /* 0x5a08 */
	ZFS_IOC_POOL_UPGRADE, /* 0x5a09 */
	ZFS_IOC_POOL_GET_HISTORY, /* 0x5a0a */
	ZFS_IOC_VDEV_ADD, /* 0x5a0b */
	ZFS_IOC_VDEV_REMOVE, /* 0x5a0c */
	ZFS_IOC_VDEV_SET_STATE, /* 0x5a0d */
	ZFS_IOC_VDEV_ATTACH, /* 0x5a0e */
	ZFS_IOC_VDEV_DETACH, /* 0x5a0f */
	ZFS_IOC_VDEV_SETPATH, /* 0x5a10 */
	ZFS_IOC_VDEV_SETFRU, /* 0x5a11 */
	ZFS_IOC_OBJSET_STATS, /* 0x5a12 */
	ZFS_IOC_OBJSET_ZPLPROPS, /* 0x5a13 */
	ZFS_IOC_DATASET_LIST_NEXT, /* 0x5a14 */
	ZFS_IOC_SNAPSHOT_LIST_NEXT, /* 0x5a15 */
	ZFS_IOC_SET_PROP, /* 0x5a16 */
	ZFS_IOC_CREATE, /* 0x5a17 */
	ZFS_IOC_DESTROY, /* 0x5a18 */
	ZFS_IOC_ROLLBACK, /* 0x5a19 */
	ZFS_IOC_RENAME, /* 0x5a1a */
	ZFS_IOC_RECV, /* 0x5a1b */
	ZFS_IOC_SEND, /* 0x5a1c */
	ZFS_IOC_INJECT_FAULT, /* 0x5a1d */
	ZFS_IOC_CLEAR_FAULT, /* 0x5a1e */
	ZFS_IOC_INJECT_LIST_NEXT, /* 0x5a1f */
	ZFS_IOC_ERROR_LOG, /* 0x5a20 */
	ZFS_IOC_CLEAR, /* 0x5a21 */
	ZFS_IOC_PROMOTE, /* 0x5a22 */
	ZFS_IOC_SNAPSHOT, /* 0x5a23 */
	ZFS_IOC_DSOBJ_TO_DSNAME, /* 0x5a24 */
	ZFS_IOC_OBJ_TO_PATH, /* 0x5a25 */
	ZFS_IOC_POOL_SET_PROPS, /* 0x5a26 */
	ZFS_IOC_POOL_GET_PROPS, /* 0x5a27 */
	ZFS_IOC_SET_FSACL, /* 0x5a28 */
	ZFS_IOC_GET_FSACL, /* 0x5a29 */
	ZFS_IOC_SHARE, /* 0x5a2a */
	ZFS_IOC_INHERIT_PROP, /* 0x5a2b */
	ZFS_IOC_SMB_ACL, /* 0x5a2c */
	ZFS_IOC_USERSPACE_ONE, /* 0x5a2d */
	ZFS_IOC_USERSPACE_MANY, /* 0x5a2e */
	ZFS_IOC_USERSPACE_UPGRADE, /* 0x5a2f */
	ZFS_IOC_HOLD, /* 0x5a30 */
	ZFS_IOC_RELEASE, /* 0x5a31 */
	ZFS_IOC_GET_HOLDS, /* 0x5a32 */
	ZFS_IOC_OBJSET_RECVD_PROPS, /* 0x5a33 */
	ZFS_IOC_VDEV_SPLIT, /* 0x5a34 */
	ZFS_IOC_NEXT_OBJ, /* 0x5a35 */
	ZFS_IOC_DIFF, /* 0x5a36 */
	ZFS_IOC_TMP_SNAPSHOT, /* 0x5a37 */
	ZFS_IOC_OBJ_TO_STATS, /* 0x5a38 */
	ZFS_IOC_SPACE_WRITTEN, /* 0x5a39 */
	ZFS_IOC_SPACE_SNAPS, /* 0x5a3a */
	ZFS_IOC_DESTROY_SNAPS, /* 0x5a3b */
	ZFS_IOC_POOL_REGUID, /* 0x5a3c */
	ZFS_IOC_POOL_REOPEN, /* 0x5a3d */
	ZFS_IOC_SEND_PROGRESS, /* 0x5a3e */
	ZFS_IOC_LOG_HISTORY, /* 0x5a3f */
	ZFS_IOC_SEND_NEW, /* 0x5a40 */
	ZFS_IOC_SEND_SPACE, /* 0x5a41 */
	ZFS_IOC_CLONE, /* 0x5a42 */
	ZFS_IOC_BOOKMARK, /* 0x5a43 */
	ZFS_IOC_GET_BOOKMARKS, /* 0x5a44 */
	ZFS_IOC_DESTROY_BOOKMARKS, /* 0x5a45 */
	ZFS_IOC_RECV_NEW, /* 0x5a46 */
	ZFS_IOC_POOL_SYNC, /* 0x5a47 */
	ZFS_IOC_CHANNEL_PROGRAM, /* 0x5a48 */
	ZFS_IOC_LOAD_KEY, /* 0x5a49 */
	ZFS_IOC_UNLOAD_KEY, /* 0x5a4a */
	ZFS_IOC_CHANGE_KEY, /* 0x5a4b */
	ZFS_IOC_REMAP, /* 0x5a4c */
	ZFS_IOC_POOL_CHECKPOINT, /* 0x5a4d */
	ZFS_IOC_POOL_DISCARD_CHECKPOINT, /* 0x5a4e */
	ZFS_IOC_POOL_INITIALIZE, /* 0x5a4f */
	ZFS_IOC_POOL_TRIM, /* 0x5a50 */
	ZFS_IOC_REDACT, /* 0x5a51 */
	ZFS_IOC_GET_BOOKMARK_PROPS, /* 0x5a52 */
	ZFS_IOC_WAIT, /* 0x5a53 */
	ZFS_IOC_WAIT_FS, /* 0x5a54 */
	ZFS_IOC_VDEV_GET_PROPS, /* 0x5a55 */
	ZFS_IOC_VDEV_SET_PROPS, /* 0x5a56 */

	/*
	* Per-platform (Optional) - 8/128 numbers reserved.
	*/
	ZFS_IOC_PLATFORM = ZFS_IOC_FIRST + 0x80,
	ZFS_IOC_EVENTS_NEXT, /* 0x81 (Linux) */
	ZFS_IOC_EVENTS_CLEAR, /* 0x82 (Linux) */
	ZFS_IOC_EVENTS_SEEK, /* 0x83 (Linux) */
	ZFS_IOC_NEXTBOOT, /* 0x84 (FreeBSD) */
	ZFS_IOC_JAIL, /* 0x85 (FreeBSD) */
	ZFS_IOC_USERNS_ATTACH = ZFS_IOC_JAIL, /* 0x85 (Linux) */
	ZFS_IOC_UNJAIL, /* 0x86 (FreeBSD) */
	ZFS_IOC_USERNS_DETACH = ZFS_IOC_UNJAIL, /* 0x86 (Linux) */
	ZFS_IOC_SET_BOOTENV, /* 0x87 */
	ZFS_IOC_GET_BOOTENV, /* 0x88 */
	ZFS_IOC_LAST
	} zfs_ioc_t;

	/*
	* zvol ioctl to get dataset name
	*/
	#define BLKZNAME _IOR(0x12, 125, char[ZFS_MAX_DATASET_NAME_LEN])

	#ifdef __linux__

	/*
	* IOCTLs to update and retrieve additional file level attributes on
	* Linux.
	*/
	#define ZFS_IOC_GETDOSFLAGS _IOR(0x83, 1, uint64_t)
	#define ZFS_IOC_SETDOSFLAGS _IOW(0x83, 2, uint64_t)

	/*
	* Additional file level attributes, that are stored
	* in the upper half of z_pflags
	*/
	#define ZFS_READONLY 0x0000000100000000ull
	#define ZFS_HIDDEN 0x0000000200000000ull
	#define ZFS_SYSTEM 0x0000000400000000ull
	#define ZFS_ARCHIVE 0x0000000800000000ull
	#define ZFS_IMMUTABLE 0x0000001000000000ull
	#define ZFS_NOUNLINK 0x0000002000000000ull
	#define ZFS_APPENDONLY 0x0000004000000000ull
	#define ZFS_NODUMP 0x0000008000000000ull
	#define ZFS_OPAQUE 0x0000010000000000ull
	#define ZFS_AV_QUARANTINED 0x0000020000000000ull
	#define ZFS_AV_MODIFIED 0x0000040000000000ull
	#define ZFS_REPARSE 0x0000080000000000ull
	#define ZFS_OFFLINE 0x0000100000000000ull
	#define ZFS_SPARSE 0x0000200000000000ull

	#define ZFS_DOS_FL_USER_VISIBLE (ZFS_IMMUTABLE \| ZFS_APPENDONLY \| \
	ZFS_NOUNLINK \| ZFS_ARCHIVE \| ZFS_NODUMP \| ZFS_SYSTEM \| \
	ZFS_HIDDEN \| ZFS_READONLY \| ZFS_REPARSE \| ZFS_OFFLINE \| \
	ZFS_SPARSE)

	#endif

	/*
	* ZFS-specific error codes used for returning descriptive errors
	* to the userland through zfs ioctls.
	*
	* The enum implicitly includes all the error codes from errno.h.
	* New code should use and extend this enum for errors that are
	* not described precisely by generic errno codes.
	*
	* These numbers should not change over time. New entries should be appended.
	*
	* (Keep in sync with contrib/pyzfs/libzfs_core/_constants.py)
	*/
	typedef enum {
	ZFS_ERR_CHECKPOINT_EXISTS = 1024,
	ZFS_ERR_DISCARDING_CHECKPOINT,
	ZFS_ERR_NO_CHECKPOINT,
	ZFS_ERR_DEVRM_IN_PROGRESS,
	ZFS_ERR_VDEV_TOO_BIG,
	ZFS_ERR_IOC_CMD_UNAVAIL,
	ZFS_ERR_IOC_ARG_UNAVAIL,
	ZFS_ERR_IOC_ARG_REQUIRED,
	ZFS_ERR_IOC_ARG_BADTYPE,
	ZFS_ERR_WRONG_PARENT,
	ZFS_ERR_FROM_IVSET_GUID_MISSING,
	ZFS_ERR_FROM_IVSET_GUID_MISMATCH,
	ZFS_ERR_SPILL_BLOCK_FLAG_MISSING,
	ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE,
	ZFS_ERR_EXPORT_IN_PROGRESS,
	ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR,
	ZFS_ERR_STREAM_TRUNCATED,
	ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH,
	ZFS_ERR_RESILVER_IN_PROGRESS,
	ZFS_ERR_REBUILD_IN_PROGRESS,
	ZFS_ERR_BADPROP,
	ZFS_ERR_VDEV_NOTSUP,
	ZFS_ERR_NOT_USER_NAMESPACE,
	+ ZFS_ERR_RESUME_EXISTS,
	} zfs_errno_t;

	/*
	* Internal SPA load state. Used by FMA diagnosis engine.
	*/
	typedef enum {
	SPA_LOAD_NONE, /* no load in progress */
	SPA_LOAD_OPEN, /* normal open */
	SPA_LOAD_IMPORT, /* import in progress */
	SPA_LOAD_TRYIMPORT, /* tryimport in progress */
	SPA_LOAD_RECOVER, /* recovery requested */
	SPA_LOAD_ERROR, /* load failed */
	SPA_LOAD_CREATE /* creation in progress */
	} spa_load_state_t;

	typedef enum {
	ZPOOL_WAIT_CKPT_DISCARD,
	ZPOOL_WAIT_FREE,
	ZPOOL_WAIT_INITIALIZE,
	ZPOOL_WAIT_REPLACE,
	ZPOOL_WAIT_REMOVE,
	ZPOOL_WAIT_RESILVER,
	ZPOOL_WAIT_SCRUB,
	ZPOOL_WAIT_TRIM,
	ZPOOL_WAIT_NUM_ACTIVITIES
	} zpool_wait_activity_t;

	typedef enum {
	ZFS_WAIT_DELETEQ,
	ZFS_WAIT_NUM_ACTIVITIES
	} zfs_wait_activity_t;

	/*
	* Bookmark name values.
	*/
	#define ZPOOL_ERR_LIST "error list"
	#define ZPOOL_ERR_DATASET "dataset"
	#define ZPOOL_ERR_OBJECT "object"

	#define HIS_MAX_RECORD_LEN (MAXPATHLEN + MAXPATHLEN + 1)

	/*
	* The following are names used in the nvlist describing
	* the pool's history log.
	*/
	#define ZPOOL_HIST_RECORD "history record"
	#define ZPOOL_HIST_TIME "history time"
	#define ZPOOL_HIST_CMD "history command"
	#define ZPOOL_HIST_WHO "history who"
	#define ZPOOL_HIST_ZONE "history zone"
	#define ZPOOL_HIST_HOST "history hostname"
	#define ZPOOL_HIST_TXG "history txg"
	#define ZPOOL_HIST_INT_EVENT "history internal event"
	#define ZPOOL_HIST_INT_STR "history internal str"
	#define ZPOOL_HIST_INT_NAME "internal_name"
	#define ZPOOL_HIST_IOCTL "ioctl"
	#define ZPOOL_HIST_INPUT_NVL "in_nvl"
	#define ZPOOL_HIST_OUTPUT_NVL "out_nvl"
	#define ZPOOL_HIST_OUTPUT_SIZE "out_size"
	#define ZPOOL_HIST_DSNAME "dsname"
	#define ZPOOL_HIST_DSID "dsid"
	#define ZPOOL_HIST_ERRNO "errno"
	#define ZPOOL_HIST_ELAPSED_NS "elapsed_ns"

	/*
	* Special nvlist name that will not have its args recorded in the pool's
	* history log.
	*/
	#define ZPOOL_HIDDEN_ARGS "hidden_args"

	/*
	* The following are names used when invoking ZFS_IOC_POOL_INITIALIZE.
	*/
	#define ZPOOL_INITIALIZE_COMMAND "initialize_command"
	#define ZPOOL_INITIALIZE_VDEVS "initialize_vdevs"

	/*
	* The following are names used when invoking ZFS_IOC_POOL_TRIM.
	*/
	#define ZPOOL_TRIM_COMMAND "trim_command"
	#define ZPOOL_TRIM_VDEVS "trim_vdevs"
	#define ZPOOL_TRIM_RATE "trim_rate"
	#define ZPOOL_TRIM_SECURE "trim_secure"

	/*
	* The following are names used when invoking ZFS_IOC_POOL_WAIT.
	*/
	#define ZPOOL_WAIT_ACTIVITY "wait_activity"
	#define ZPOOL_WAIT_TAG "wait_tag"
	#define ZPOOL_WAIT_WAITED "wait_waited"

	/*
	* The following are names used when invoking ZFS_IOC_VDEV_GET_PROP.
	*/
	#define ZPOOL_VDEV_PROPS_GET_VDEV "vdevprops_get_vdev"
	#define ZPOOL_VDEV_PROPS_GET_PROPS "vdevprops_get_props"

	/*
	* The following are names used when invoking ZFS_IOC_VDEV_SET_PROP.
	*/
	#define ZPOOL_VDEV_PROPS_SET_VDEV "vdevprops_set_vdev"
	#define ZPOOL_VDEV_PROPS_SET_PROPS "vdevprops_set_props"

	/*
	* The following are names used when invoking ZFS_IOC_WAIT_FS.
	*/
	#define ZFS_WAIT_ACTIVITY "wait_activity"
	#define ZFS_WAIT_WAITED "wait_waited"

	/*
	* Flags for ZFS_IOC_VDEV_SET_STATE
	*/
	#define ZFS_ONLINE_CHECKREMOVE 0x1
	#define ZFS_ONLINE_UNSPARE 0x2
	#define ZFS_ONLINE_FORCEFAULT 0x4
	#define ZFS_ONLINE_EXPAND 0x8
	#define ZFS_OFFLINE_TEMPORARY 0x1

	/*
	* Flags for ZFS_IOC_POOL_IMPORT
	*/
	#define ZFS_IMPORT_NORMAL 0x0
	#define ZFS_IMPORT_VERBATIM 0x1
	#define ZFS_IMPORT_ANY_HOST 0x2
	#define ZFS_IMPORT_MISSING_LOG 0x4
	#define ZFS_IMPORT_ONLY 0x8
	#define ZFS_IMPORT_TEMP_NAME 0x10
	#define ZFS_IMPORT_SKIP_MMP 0x20
	#define ZFS_IMPORT_LOAD_KEYS 0x40
	#define ZFS_IMPORT_CHECKPOINT 0x80

	/*
	* Channel program argument/return nvlist keys and defaults.
	*/
	#define ZCP_ARG_PROGRAM "program"
	#define ZCP_ARG_ARGLIST "arg"
	#define ZCP_ARG_SYNC "sync"
	#define ZCP_ARG_INSTRLIMIT "instrlimit"
	#define ZCP_ARG_MEMLIMIT "memlimit"

	#define ZCP_ARG_CLIARGV "argv"

	#define ZCP_RET_ERROR "error"
	#define ZCP_RET_RETURN "return"

	#define ZCP_DEFAULT_INSTRLIMIT (10 * 1000 * 1000)
	#define ZCP_MAX_INSTRLIMIT (10 * ZCP_DEFAULT_INSTRLIMIT)
	#define ZCP_DEFAULT_MEMLIMIT (10 * 1024 * 1024)
	#define ZCP_MAX_MEMLIMIT (10 * ZCP_DEFAULT_MEMLIMIT)

	/*
	* Sysevent payload members. ZFS will generate the following sysevents with the
	* given payloads:
	*
	* ESC_ZFS_RESILVER_START
	* ESC_ZFS_RESILVER_FINISH
	*
	* ZFS_EV_POOL_NAME DATA_TYPE_STRING
	* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
	* ZFS_EV_RESILVER_TYPE DATA_TYPE_STRING
	*
	* ESC_ZFS_POOL_DESTROY
	* ESC_ZFS_POOL_REGUID
	*
	* ZFS_EV_POOL_NAME DATA_TYPE_STRING
	* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
	*
	* ESC_ZFS_VDEV_REMOVE
	* ESC_ZFS_VDEV_CLEAR
	* ESC_ZFS_VDEV_CHECK
	*
	* ZFS_EV_POOL_NAME DATA_TYPE_STRING
	* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
	* ZFS_EV_VDEV_PATH DATA_TYPE_STRING (optional)
	* ZFS_EV_VDEV_GUID DATA_TYPE_UINT64
	*
	* ESC_ZFS_HISTORY_EVENT
	*
	* ZFS_EV_POOL_NAME DATA_TYPE_STRING
	* ZFS_EV_POOL_GUID DATA_TYPE_UINT64
	* ZFS_EV_HIST_TIME DATA_TYPE_UINT64 (optional)
	* ZFS_EV_HIST_CMD DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_WHO DATA_TYPE_UINT64 (optional)
	* ZFS_EV_HIST_ZONE DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_HOST DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_TXG DATA_TYPE_UINT64 (optional)
	* ZFS_EV_HIST_INT_EVENT DATA_TYPE_UINT64 (optional)
	* ZFS_EV_HIST_INT_STR DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_INT_NAME DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_IOCTL DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_DSNAME DATA_TYPE_STRING (optional)
	* ZFS_EV_HIST_DSID DATA_TYPE_UINT64 (optional)
	*
	* The ZFS_EV_HIST_* members will correspond to the ZPOOL_HIST_* members in the
	* history log nvlist. The keynames will be free of any spaces or other
	* characters that could be potentially unexpected to consumers of the
	* sysevents.
	*/
	#define ZFS_EV_POOL_NAME "pool_name"
	#define ZFS_EV_POOL_GUID "pool_guid"
	#define ZFS_EV_VDEV_PATH "vdev_path"
	#define ZFS_EV_VDEV_GUID "vdev_guid"
	#define ZFS_EV_HIST_TIME "history_time"
	#define ZFS_EV_HIST_CMD "history_command"
	#define ZFS_EV_HIST_WHO "history_who"
	#define ZFS_EV_HIST_ZONE "history_zone"
	#define ZFS_EV_HIST_HOST "history_hostname"
	#define ZFS_EV_HIST_TXG "history_txg"
	#define ZFS_EV_HIST_INT_EVENT "history_internal_event"
	#define ZFS_EV_HIST_INT_STR "history_internal_str"
	#define ZFS_EV_HIST_INT_NAME "history_internal_name"
	#define ZFS_EV_HIST_IOCTL "history_ioctl"
	#define ZFS_EV_HIST_DSNAME "history_dsname"
	#define ZFS_EV_HIST_DSID "history_dsid"
	#define ZFS_EV_RESILVER_TYPE "resilver_type"

	/*
	* We currently support block sizes from 512 bytes to 16MB.
	* The benefits of larger blocks, and thus larger IO, need to be weighed
	* against the cost of COWing a giant block to modify one byte, and the
	* large latency of reading or writing a large block.
	*
	* The recordsize property can not be set larger than zfs_max_recordsize
	* (default 16MB on 64-bit and 1MB on 32-bit). See the comment near
	* zfs_max_recordsize in dsl_dataset.c for details.
	*
	* Note that although the LSIZE field of the blkptr_t can store sizes up
	* to 32MB, the dnode's dn_datablkszsec can only store sizes up to
	* 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
	*/
	#define SPA_MINBLOCKSHIFT 9
	#define SPA_OLD_MAXBLOCKSHIFT 17
	#define SPA_MAXBLOCKSHIFT 24
	#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
	#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
	#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)

	/* supported encryption algorithms */
	enum zio_encrypt {
	ZIO_CRYPT_INHERIT = 0,
	ZIO_CRYPT_ON,
	ZIO_CRYPT_OFF,
	ZIO_CRYPT_AES_128_CCM,
	ZIO_CRYPT_AES_192_CCM,
	ZIO_CRYPT_AES_256_CCM,
	ZIO_CRYPT_AES_128_GCM,
	ZIO_CRYPT_AES_192_GCM,
	ZIO_CRYPT_AES_256_GCM,
	ZIO_CRYPT_FUNCTIONS
	};

	#define ZIO_CRYPT_ON_VALUE ZIO_CRYPT_AES_256_GCM
	#define ZIO_CRYPT_DEFAULT ZIO_CRYPT_OFF

	/*
	* xattr namespace prefixes. These are forbidden in xattr names.
	*
	* For cross-platform compatibility, xattrs in the user namespace should not be
	* prefixed with the namespace name, but for backwards compatibility with older
	* ZFS on Linux versions we do prefix the namespace.
	*/
	#define ZFS_XA_NS_FREEBSD_PREFIX "freebsd:"
	#define ZFS_XA_NS_FREEBSD_PREFIX_LEN strlen("freebsd:")
	#define ZFS_XA_NS_LINUX_SECURITY_PREFIX "security."
	#define ZFS_XA_NS_LINUX_SECURITY_PREFIX_LEN strlen("security.")
	#define ZFS_XA_NS_LINUX_SYSTEM_PREFIX "system."
	#define ZFS_XA_NS_LINUX_SYSTEM_PREFIX_LEN strlen("system.")
	#define ZFS_XA_NS_LINUX_TRUSTED_PREFIX "trusted."
	#define ZFS_XA_NS_LINUX_TRUSTED_PREFIX_LEN strlen("trusted.")
	#define ZFS_XA_NS_LINUX_USER_PREFIX "user."
	#define ZFS_XA_NS_LINUX_USER_PREFIX_LEN strlen("user.")

	#define ZFS_XA_NS_PREFIX_MATCH(ns, name) \
	(strncmp(name, ZFS_XA_NS_##ns##_PREFIX, \
	ZFS_XA_NS_##ns##_PREFIX_LEN) == 0)

	#define ZFS_XA_NS_PREFIX_FORBIDDEN(name) \
	(ZFS_XA_NS_PREFIX_MATCH(FREEBSD, name) \|\| \
	ZFS_XA_NS_PREFIX_MATCH(LINUX_SECURITY, name) \|\| \
	ZFS_XA_NS_PREFIX_MATCH(LINUX_SYSTEM, name) \|\| \
	ZFS_XA_NS_PREFIX_MATCH(LINUX_TRUSTED, name) \|\| \
	ZFS_XA_NS_PREFIX_MATCH(LINUX_USER, name))

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_FS_ZFS_H */
	diff --git a/sys/contrib/openzfs/include/sys/spa.h b/sys/contrib/openzfs/include/sys/spa.h
	index e185ce6b1d8e..3e68cb8c6511 100644
	--- a/sys/contrib/openzfs/include/sys/spa.h
	+++ b/sys/contrib/openzfs/include/sys/spa.h
	@@ -1,1232 +1,1232 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Datto Inc.
	*/

	#ifndef _SYS_SPA_H
	#define _SYS_SPA_H

	#include <sys/avl.h>
	#include <sys/zfs_context.h>
	#include <sys/kstat.h>
	#include <sys/nvpair.h>
	#include <sys/sysmacros.h>
	#include <sys/types.h>
	#include <sys/fs/zfs.h>
	#include <sys/spa_checksum.h>
	#include <sys/dmu.h>
	#include <sys/space_map.h>
	#include <sys/bitops.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Forward references that lots of things need.
	*/
	typedef struct spa spa_t;
	typedef struct vdev vdev_t;
	typedef struct metaslab metaslab_t;
	typedef struct metaslab_group metaslab_group_t;
	typedef struct metaslab_class metaslab_class_t;
	typedef struct zio zio_t;
	typedef struct zilog zilog_t;
	typedef struct spa_aux_vdev spa_aux_vdev_t;
	typedef struct ddt ddt_t;
	typedef struct ddt_entry ddt_entry_t;
	typedef struct zbookmark_phys zbookmark_phys_t;

	struct bpobj;
	struct bplist;
	struct dsl_pool;
	struct dsl_dataset;
	struct dsl_crypto_params;

	/*
	* Alignment Shift (ashift) is an immutable, internal top-level vdev property
	* which can only be set at vdev creation time. Physical writes are always done
	* according to it, which makes 2^ashift the smallest possible IO on a vdev.
	*
	* We currently allow values ranging from 512 bytes (2^9 = 512) to 64 KiB
	* (2^16 = 65,536).
	*/
	#define ASHIFT_MIN 9
	#define ASHIFT_MAX 16

	/*
	* Size of block to hold the configuration data (a packed nvlist)
	*/
	#define SPA_CONFIG_BLOCKSIZE (1ULL << 14)

	/*
	* The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
	* The ASIZE encoding should be at least 64 times larger (6 more bits)
	* to support up to 4-way RAID-Z mirror mode with worst-case gang block
	* overhead, three DVAs per bp, plus one more bit in case we do anything
	* else that expands the ASIZE.
	*/
	#define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */
	#define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */
	#define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */

	#define SPA_COMPRESSBITS 7
	#define SPA_VDEVBITS 24
	#define SPA_COMPRESSMASK ((1U << SPA_COMPRESSBITS) - 1)

	/*
	* All SPA data is represented by 128-bit data virtual addresses (DVAs).
	* The members of the dva_t should be considered opaque outside the SPA.
	*/
	typedef struct dva {
	uint64_t dva_word[2];
	} dva_t;


	/*
	* Some checksums/hashes need a 256-bit initialization salt. This salt is kept
	* secret and is suitable for use in MAC algorithms as the key.
	*/
	typedef struct zio_cksum_salt {
	uint8_t zcs_bytes[32];
	} zio_cksum_salt_t;

	/*
	* Each block is described by its DVAs, time of birth, checksum, etc.
	* The word-by-word, bit-by-bit layout of the blkptr is as follows:
	*
	* 64 56 48 40 32 24 16 8 0
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 0 \| pad \| vdev1 \| GRID \| ASIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 1 \|G\| offset1 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 2 \| pad \| vdev2 \| GRID \| ASIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 3 \|G\| offset2 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 4 \| pad \| vdev3 \| GRID \| ASIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 5 \|G\| offset3 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 6 \|BDX\|lvl\| type \| cksum \|E\| comp\| PSIZE \| LSIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 7 \| padding \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 8 \| padding \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 9 \| physical birth txg \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* a \| logical birth txg \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* b \| fill count \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* c \| checksum[0] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* d \| checksum[1] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* e \| checksum[2] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* f \| checksum[3] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	*
	* Legend:
	*
	* vdev virtual device ID
	* offset offset into virtual device
	* LSIZE logical size
	* PSIZE physical size (after compression)
	* ASIZE allocated size (including RAID-Z parity and gang block headers)
	* GRID RAID-Z layout information (reserved for future use)
	* cksum checksum function
	* comp compression function
	* G gang block indicator
	* B byteorder (endianness)
	* D dedup
	* X encryption
	* E blkptr_t contains embedded data (see below)
	* lvl level of indirection
	* type DMU object type
	* phys birth txg when dva[0] was written; zero if same as logical birth txg
	* note that typically all the dva's would be written in this
	* txg, but they could be different if they were moved by
	* device removal.
	* log. birth transaction group in which the block was logically born
	* fill count number of non-zero blocks under this bp
	* checksum[4] 256-bit checksum of the data this bp describes
	*/

	/*
	* The blkptr_t's of encrypted blocks also need to store the encryption
	* parameters so that the block can be decrypted. This layout is as follows:
	*
	* 64 56 48 40 32 24 16 8 0
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 0 \| vdev1 \| GRID \| ASIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 1 \|G\| offset1 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 2 \| vdev2 \| GRID \| ASIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 3 \|G\| offset2 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 4 \| salt \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 5 \| IV1 \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 6 \|BDX\|lvl\| type \| cksum \|E\| comp\| PSIZE \| LSIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 7 \| padding \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 8 \| padding \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 9 \| physical birth txg \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* a \| logical birth txg \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* b \| IV2 \| fill count \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* c \| checksum[0] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* d \| checksum[1] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* e \| MAC[0] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* f \| MAC[1] \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	*
	* Legend:
	*
	* salt Salt for generating encryption keys
	* IV1 First 64 bits of encryption IV
	* X Block requires encryption handling (set to 1)
	* E blkptr_t contains embedded data (set to 0, see below)
	* fill count number of non-zero blocks under this bp (truncated to 32 bits)
	* IV2 Last 32 bits of encryption IV
	* checksum[2] 128-bit checksum of the data this bp describes
	* MAC[2] 128-bit message authentication code for this data
	*
	* The X bit being set indicates that this block is one of 3 types. If this is
	* a level 0 block with an encrypted object type, the block is encrypted
	* (see BP_IS_ENCRYPTED()). If this is a level 0 block with an unencrypted
	* object type, this block is authenticated with an HMAC (see
	* BP_IS_AUTHENTICATED()). Otherwise (if level > 0), this bp will use the MAC
	* words to store a checksum-of-MACs from the level below (see
	* BP_HAS_INDIRECT_MAC_CKSUM()). For convenience in the code, BP_IS_PROTECTED()
	* refers to both encrypted and authenticated blocks and BP_USES_CRYPT()
	* refers to any of these 3 kinds of blocks.
	*
	* The additional encryption parameters are the salt, IV, and MAC which are
	* explained in greater detail in the block comment at the top of zio_crypt.c.
	* The MAC occupies half of the checksum space since it serves a very similar
	* purpose: to prevent data corruption on disk. The only functional difference
	* is that the checksum is used to detect on-disk corruption whether or not the
	* encryption key is loaded and the MAC provides additional protection against
	* malicious disk tampering. We use the 3rd DVA to store the salt and first
	* 64 bits of the IV. As a result encrypted blocks can only have 2 copies
	* maximum instead of the normal 3. The last 32 bits of the IV are stored in
	* the upper bits of what is usually the fill count. Note that only blocks at
	* level 0 or -2 are ever encrypted, which allows us to guarantee that these
	* 32 bits are not trampled over by other code (see zio_crypt.c for details).
	* The salt and IV are not used for authenticated bps or bps with an indirect
	* MAC checksum, so these blocks can utilize all 3 DVAs and the full 64 bits
	* for the fill count.
	*/

	/*
	* "Embedded" blkptr_t's don't actually point to a block, instead they
	* have a data payload embedded in the blkptr_t itself. See the comment
	* in blkptr.c for more details.
	*
	* The blkptr_t is laid out as follows:
	*
	* 64 56 48 40 32 24 16 8 0
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 0 \| payload \|
	* 1 \| payload \|
	* 2 \| payload \|
	* 3 \| payload \|
	* 4 \| payload \|
	* 5 \| payload \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 6 \|BDX\|lvl\| type \| etype \|E\| comp\| PSIZE\| LSIZE \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* 7 \| payload \|
	* 8 \| payload \|
	* 9 \| payload \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* a \| logical birth txg \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* b \| payload \|
	* c \| payload \|
	* d \| payload \|
	* e \| payload \|
	* f \| payload \|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	*
	* Legend:
	*
	* payload contains the embedded data
	* B (byteorder) byteorder (endianness)
	* D (dedup) padding (set to zero)
	* X encryption (set to zero)
	* E (embedded) set to one
	* lvl indirection level
	* type DMU object type
	* etype how to interpret embedded data (BP_EMBEDDED_TYPE_*)
	* comp compression function of payload
	* PSIZE size of payload after compression, in bytes
	* LSIZE logical size of payload, in bytes
	* note that 25 bits is enough to store the largest
	* "normal" BP's LSIZE (2^16 * 2^9) in bytes
	* log. birth transaction group in which the block was logically born
	*
	* Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded
	* bp's they are stored in units of SPA_MINBLOCKSHIFT.
	* Generally, the generic BP_GET_*() macros can be used on embedded BP's.
	* The B, D, X, lvl, type, and comp fields are stored the same as with normal
	* BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must
	* be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before
	* other macros, as they assert that they are only used on BP's of the correct
	* "embedded-ness". Encrypted blkptr_t's cannot be embedded because they use
	* the payload space for encryption parameters (see the comment above on
	* how encryption parameters are stored).
	*/

	#define BPE_GET_ETYPE(bp) \
	(ASSERT(BP_IS_EMBEDDED(bp)), \
	BF64_GET((bp)->blk_prop, 40, 8))
	#define BPE_SET_ETYPE(bp, t) do { \
	ASSERT(BP_IS_EMBEDDED(bp)); \
	BF64_SET((bp)->blk_prop, 40, 8, t); \
	} while (0)

	#define BPE_GET_LSIZE(bp) \
	(ASSERT(BP_IS_EMBEDDED(bp)), \
	BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1))
	#define BPE_SET_LSIZE(bp, x) do { \
	ASSERT(BP_IS_EMBEDDED(bp)); \
	BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \
	} while (0)

	#define BPE_GET_PSIZE(bp) \
	(ASSERT(BP_IS_EMBEDDED(bp)), \
	BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1))
	#define BPE_SET_PSIZE(bp, x) do { \
	ASSERT(BP_IS_EMBEDDED(bp)); \
	BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \
	} while (0)

	typedef enum bp_embedded_type {
	BP_EMBEDDED_TYPE_DATA,
	BP_EMBEDDED_TYPE_RESERVED, /* Reserved for Delphix byteswap feature. */
	BP_EMBEDDED_TYPE_REDACTED,
	NUM_BP_EMBEDDED_TYPES
	} bp_embedded_type_t;

	#define BPE_NUM_WORDS 14
	#define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t))
	#define BPE_IS_PAYLOADWORD(bp, wp) \
	((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth)

	#define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */
	#define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */
	#define SPA_SYNC_MIN_VDEVS 3 /* min vdevs to update during sync */

	/*
	* A block is a hole when it has either 1) never been written to, or
	* 2) is zero-filled. In both cases, ZFS can return all zeroes for all reads
	* without physically allocating disk space. Holes are represented in the
	* blkptr_t structure by zeroed blk_dva. Correct checking for holes is
	* done through the BP_IS_HOLE macro. For holes, the logical size, level,
	* DMU object type, and birth times are all also stored for holes that
	* were written to at some point (i.e. were punched after having been filled).
	*/
	typedef struct blkptr {
	dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
	uint64_t blk_prop; /* size, compression, type, etc */
	uint64_t blk_pad[2]; /* Extra space for the future */
	uint64_t blk_phys_birth; /* txg when block was allocated */
	uint64_t blk_birth; /* transaction group at birth */
	uint64_t blk_fill; /* fill count */
	zio_cksum_t blk_cksum; /* 256-bit checksum */
	} blkptr_t;

	/*
	* Macros to get and set fields in a bp or DVA.
	*/

	/*
	* Note, for gang blocks, DVA_GET_ASIZE() is the total space allocated for
	* this gang DVA including its children BP's. The space allocated at this
	* DVA's vdev/offset is vdev_gang_header_asize(vdev).
	*/
	#define DVA_GET_ASIZE(dva) \
	BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0)
	#define DVA_SET_ASIZE(dva, x) \
	BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \
	SPA_MINBLOCKSHIFT, 0, x)

	#define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8)
	#define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x)

	#define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, SPA_VDEVBITS)
	#define DVA_SET_VDEV(dva, x) \
	BF64_SET((dva)->dva_word[0], 32, SPA_VDEVBITS, x)

	#define DVA_GET_OFFSET(dva) \
	BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0)
	#define DVA_SET_OFFSET(dva, x) \
	BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x)

	#define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1)
	#define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x)

	#define BP_GET_LSIZE(bp) \
	(BP_IS_EMBEDDED(bp) ? \
	(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \
	BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1))
	#define BP_SET_LSIZE(bp, x) do { \
	ASSERT(!BP_IS_EMBEDDED(bp)); \
	BF64_SET_SB((bp)->blk_prop, \
	0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
	} while (0)

	#define BP_GET_PSIZE(bp) \
	(BP_IS_EMBEDDED(bp) ? 0 : \
	BF64_GET_SB((bp)->blk_prop, 16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1))
	#define BP_SET_PSIZE(bp, x) do { \
	ASSERT(!BP_IS_EMBEDDED(bp)); \
	BF64_SET_SB((bp)->blk_prop, \
	16, SPA_PSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \
	} while (0)

	#define BP_GET_COMPRESS(bp) \
	BF64_GET((bp)->blk_prop, 32, SPA_COMPRESSBITS)
	#define BP_SET_COMPRESS(bp, x) \
	BF64_SET((bp)->blk_prop, 32, SPA_COMPRESSBITS, x)

	#define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1)
	#define BP_SET_EMBEDDED(bp, x) BF64_SET((bp)->blk_prop, 39, 1, x)

	#define BP_GET_CHECKSUM(bp) \
	(BP_IS_EMBEDDED(bp) ? ZIO_CHECKSUM_OFF : \
	BF64_GET((bp)->blk_prop, 40, 8))
	#define BP_SET_CHECKSUM(bp, x) do { \
	ASSERT(!BP_IS_EMBEDDED(bp)); \
	BF64_SET((bp)->blk_prop, 40, 8, x); \
	} while (0)

	#define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8)
	#define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x)

	#define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5)
	#define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x)

	/* encrypted, authenticated, and MAC cksum bps use the same bit */
	#define BP_USES_CRYPT(bp) BF64_GET((bp)->blk_prop, 61, 1)
	#define BP_SET_CRYPT(bp, x) BF64_SET((bp)->blk_prop, 61, 1, x)

	#define BP_IS_ENCRYPTED(bp) \
	(BP_USES_CRYPT(bp) && \
	BP_GET_LEVEL(bp) <= 0 && \
	DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))

	#define BP_IS_AUTHENTICATED(bp) \
	(BP_USES_CRYPT(bp) && \
	BP_GET_LEVEL(bp) <= 0 && \
	!DMU_OT_IS_ENCRYPTED(BP_GET_TYPE(bp)))

	#define BP_HAS_INDIRECT_MAC_CKSUM(bp) \
	(BP_USES_CRYPT(bp) && BP_GET_LEVEL(bp) > 0)

	#define BP_IS_PROTECTED(bp) \
	(BP_IS_ENCRYPTED(bp) \|\| BP_IS_AUTHENTICATED(bp))

	#define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1)
	#define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x)

	#define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1)
	#define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x)

	#define BP_GET_FREE(bp) BF64_GET((bp)->blk_fill, 0, 1)
	#define BP_SET_FREE(bp, x) BF64_SET((bp)->blk_fill, 0, 1, x)

	#define BP_PHYSICAL_BIRTH(bp) \
	(BP_IS_EMBEDDED(bp) ? 0 : \
	(bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)

	#define BP_SET_BIRTH(bp, logical, physical) \
	{ \
	ASSERT(!BP_IS_EMBEDDED(bp)); \
	(bp)->blk_birth = (logical); \
	(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
	}

	#define BP_GET_FILL(bp) \
	((BP_IS_ENCRYPTED(bp)) ? BF64_GET((bp)->blk_fill, 0, 32) : \
	((BP_IS_EMBEDDED(bp)) ? 1 : (bp)->blk_fill))

	#define BP_SET_FILL(bp, fill) \
	{ \
	if (BP_IS_ENCRYPTED(bp)) \
	BF64_SET((bp)->blk_fill, 0, 32, fill); \
	else \
	(bp)->blk_fill = fill; \
	}

	#define BP_GET_IV2(bp) \
	(ASSERT(BP_IS_ENCRYPTED(bp)), \
	BF64_GET((bp)->blk_fill, 32, 32))
	#define BP_SET_IV2(bp, iv2) \
	{ \
	ASSERT(BP_IS_ENCRYPTED(bp)); \
	BF64_SET((bp)->blk_fill, 32, 32, iv2); \
	}

	#define BP_IS_METADATA(bp) \
	(BP_GET_LEVEL(bp) > 0 \|\| DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))

	#define BP_GET_ASIZE(bp) \
	(BP_IS_EMBEDDED(bp) ? 0 : \
	DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
	DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
	(DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))

	#define BP_GET_UCSIZE(bp) \
	(BP_IS_METADATA(bp) ? BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp))

	#define BP_GET_NDVAS(bp) \
	(BP_IS_EMBEDDED(bp) ? 0 : \
	!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \
	!!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
	(!!DVA_GET_ASIZE(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp)))

	#define BP_COUNT_GANG(bp) \
	(BP_IS_EMBEDDED(bp) ? 0 : \
	(DVA_GET_GANG(&(bp)->blk_dva[0]) + \
	DVA_GET_GANG(&(bp)->blk_dva[1]) + \
	(DVA_GET_GANG(&(bp)->blk_dva[2]) * !BP_IS_ENCRYPTED(bp))))

	#define DVA_EQUAL(dva1, dva2) \
	((dva1)->dva_word[1] == (dva2)->dva_word[1] && \
	(dva1)->dva_word[0] == (dva2)->dva_word[0])

	#define BP_EQUAL(bp1, bp2) \
	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) && \
	(bp1)->blk_birth == (bp2)->blk_birth && \
	DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) && \
	DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) && \
	DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))


	#define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0)

	#define BP_IDENTITY(bp) (ASSERT(!BP_IS_EMBEDDED(bp)), &(bp)->blk_dva[0])
	#define BP_IS_GANG(bp) \
	(BP_IS_EMBEDDED(bp) ? B_FALSE : DVA_GET_GANG(BP_IDENTITY(bp)))
	#define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \
	(dva)->dva_word[1] == 0ULL)
	#define BP_IS_HOLE(bp) \
	(!BP_IS_EMBEDDED(bp) && DVA_IS_EMPTY(BP_IDENTITY(bp)))

	#define BP_SET_REDACTED(bp) \
	{ \
	BP_SET_EMBEDDED(bp, B_TRUE); \
	BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_REDACTED); \
	}
	#define BP_IS_REDACTED(bp) \
	(BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_REDACTED)

	/* BP_IS_RAIDZ(bp) assumes no block compression */
	#define BP_IS_RAIDZ(bp) (DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
	BP_GET_PSIZE(bp))

	#define BP_ZERO(bp) \
	{ \
	(bp)->blk_dva[0].dva_word[0] = 0; \
	(bp)->blk_dva[0].dva_word[1] = 0; \
	(bp)->blk_dva[1].dva_word[0] = 0; \
	(bp)->blk_dva[1].dva_word[1] = 0; \
	(bp)->blk_dva[2].dva_word[0] = 0; \
	(bp)->blk_dva[2].dva_word[1] = 0; \
	(bp)->blk_prop = 0; \
	(bp)->blk_pad[0] = 0; \
	(bp)->blk_pad[1] = 0; \
	(bp)->blk_phys_birth = 0; \
	(bp)->blk_birth = 0; \
	(bp)->blk_fill = 0; \
	ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \
	}

	#ifdef _ZFS_BIG_ENDIAN
	#define ZFS_HOST_BYTEORDER (0ULL)
	#else
	#define ZFS_HOST_BYTEORDER (1ULL)
	#endif

	#define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER)

	#define BP_SPRINTF_LEN 400

	/*
	* This macro allows code sharing between zfs, libzpool, and mdb.
	* 'func' is either snprintf() or mdb_snprintf().
	* 'ws' (whitespace) can be ' ' for single-line format, '\n' for multi-line.
	*/

	#define SNPRINTF_BLKPTR(func, ws, buf, size, bp, type, checksum, compress) \
	{ \
	static const char *const copyname[] = \
	{ "zero", "single", "double", "triple" }; \
	int len = 0; \
	int copies = 0; \
	const char *crypt_type; \
	if (bp != NULL) { \
	if (BP_IS_ENCRYPTED(bp)) { \
	crypt_type = "encrypted"; \
	/* LINTED E_SUSPICIOUS_COMPARISON */ \
	} else if (BP_IS_AUTHENTICATED(bp)) { \
	crypt_type = "authenticated"; \
	} else if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { \
	crypt_type = "indirect-MAC"; \
	} else { \
	crypt_type = "unencrypted"; \
	} \
	} \
	if (bp == NULL) { \
	len += func(buf + len, size - len, "<NULL>"); \
	} else if (BP_IS_HOLE(bp)) { \
	len += func(buf + len, size - len, \
	"HOLE [L%llu %s] " \
	"size=%llxL birth=%lluL", \
	(u_longlong_t)BP_GET_LEVEL(bp), \
	type, \
	(u_longlong_t)BP_GET_LSIZE(bp), \
	(u_longlong_t)bp->blk_birth); \
	} else if (BP_IS_EMBEDDED(bp)) { \
	len = func(buf + len, size - len, \
	"EMBEDDED [L%llu %s] et=%u %s " \
	"size=%llxL/%llxP birth=%lluL", \
	(u_longlong_t)BP_GET_LEVEL(bp), \
	type, \
	(int)BPE_GET_ETYPE(bp), \
	compress, \
	(u_longlong_t)BPE_GET_LSIZE(bp), \
	(u_longlong_t)BPE_GET_PSIZE(bp), \
	(u_longlong_t)bp->blk_birth); \
	} else if (BP_IS_REDACTED(bp)) { \
	len += func(buf + len, size - len, \
	"REDACTED [L%llu %s] size=%llxL birth=%lluL", \
	(u_longlong_t)BP_GET_LEVEL(bp), \
	type, \
	(u_longlong_t)BP_GET_LSIZE(bp), \
	(u_longlong_t)bp->blk_birth); \
	} else { \
	for (int d = 0; d < BP_GET_NDVAS(bp); d++) { \
	const dva_t *dva = &bp->blk_dva[d]; \
	if (DVA_IS_VALID(dva)) \
	copies++; \
	len += func(buf + len, size - len, \
	"DVA[%d]=<%llu:%llx:%llx>%c", d, \
	(u_longlong_t)DVA_GET_VDEV(dva), \
	(u_longlong_t)DVA_GET_OFFSET(dva), \
	(u_longlong_t)DVA_GET_ASIZE(dva), \
	ws); \
	} \
	if (BP_IS_ENCRYPTED(bp)) { \
	len += func(buf + len, size - len, \
	"salt=%llx iv=%llx:%llx%c", \
	(u_longlong_t)bp->blk_dva[2].dva_word[0], \
	(u_longlong_t)bp->blk_dva[2].dva_word[1], \
	(u_longlong_t)BP_GET_IV2(bp), \
	ws); \
	} \
	if (BP_IS_GANG(bp) && \
	DVA_GET_ASIZE(&bp->blk_dva[2]) <= \
	DVA_GET_ASIZE(&bp->blk_dva[1]) / 2) \
	copies--; \
	len += func(buf + len, size - len, \
	"[L%llu %s] %s %s %s %s %s %s %s%c" \
	"size=%llxL/%llxP birth=%lluL/%lluP fill=%llu%c" \
	"cksum=%llx:%llx:%llx:%llx", \
	(u_longlong_t)BP_GET_LEVEL(bp), \
	type, \
	checksum, \
	compress, \
	crypt_type, \
	BP_GET_BYTEORDER(bp) == 0 ? "BE" : "LE", \
	BP_IS_GANG(bp) ? "gang" : "contiguous", \
	BP_GET_DEDUP(bp) ? "dedup" : "unique", \
	copyname[copies], \
	ws, \
	(u_longlong_t)BP_GET_LSIZE(bp), \
	(u_longlong_t)BP_GET_PSIZE(bp), \
	(u_longlong_t)bp->blk_birth, \
	(u_longlong_t)BP_PHYSICAL_BIRTH(bp), \
	(u_longlong_t)BP_GET_FILL(bp), \
	ws, \
	(u_longlong_t)bp->blk_cksum.zc_word[0], \
	(u_longlong_t)bp->blk_cksum.zc_word[1], \
	(u_longlong_t)bp->blk_cksum.zc_word[2], \
	(u_longlong_t)bp->blk_cksum.zc_word[3]); \
	} \
	ASSERT(len < size); \
	}

	#define BP_GET_BUFC_TYPE(bp) \
	(BP_IS_METADATA(bp) ? ARC_BUFC_METADATA : ARC_BUFC_DATA)

	typedef enum spa_import_type {
	SPA_IMPORT_EXISTING,
	SPA_IMPORT_ASSEMBLE
	} spa_import_type_t;

	typedef enum spa_mode {
	SPA_MODE_UNINIT = 0,
	SPA_MODE_READ = 1,
	SPA_MODE_WRITE = 2,
	} spa_mode_t;

	/*
	* Send TRIM commands in-line during normal pool operation while deleting.
	* OFF: no
	* ON: yes
	* NB: IN_FREEBSD_BASE is defined within the FreeBSD sources.
	*/
	typedef enum {
	SPA_AUTOTRIM_OFF = 0, /* default */
	SPA_AUTOTRIM_ON,
	#ifdef IN_FREEBSD_BASE
	SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_ON,
	#else
	SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_OFF,
	#endif
	} spa_autotrim_t;

	/*
	* Reason TRIM command was issued, used internally for accounting purposes.
	*/
	typedef enum trim_type {
	TRIM_TYPE_MANUAL = 0,
	TRIM_TYPE_AUTO = 1,
	TRIM_TYPE_SIMPLE = 2
	} trim_type_t;

	/* state manipulation functions */
	extern int spa_open(const char pool, spa_t , const void tag);
	extern int spa_open_rewind(const char pool, spa_t , const void tag,
	nvlist_t policy, nvlist_t *config);
	extern int spa_get_stats(const char pool, nvlist_t config, char altroot,
	size_t buflen);
	extern int spa_create(const char pool, nvlist_t nvroot, nvlist_t *props,
	nvlist_t zplprops, struct dsl_crypto_params dcp);
	extern int spa_import(char pool, nvlist_t config, nvlist_t *props,
	uint64_t flags);
	extern nvlist_t spa_tryimport(nvlist_t tryconfig);
	extern int spa_destroy(const char *pool);
	extern int spa_checkpoint(const char *pool);
	extern int spa_checkpoint_discard(const char *pool);
	extern int spa_export(const char pool, nvlist_t *oldconfig, boolean_t force,
	boolean_t hardforce);
	extern int spa_reset(const char *pool);
	extern void spa_async_request(spa_t *spa, int flag);
	extern void spa_async_unrequest(spa_t *spa, int flag);
	extern void spa_async_suspend(spa_t *spa);
	extern void spa_async_resume(spa_t *spa);
	extern int spa_async_tasks(spa_t *spa);
	extern spa_t spa_inject_addref(char pool);
	extern void spa_inject_delref(spa_t *spa);
	extern void spa_scan_stat_init(spa_t *spa);
	extern int spa_scan_get_stats(spa_t spa, pool_scan_stat_t ps);
	extern int bpobj_enqueue_alloc_cb(void arg, const blkptr_t bp, dmu_tx_t *tx);
	extern int bpobj_enqueue_free_cb(void arg, const blkptr_t bp, dmu_tx_t *tx);

	#define SPA_ASYNC_CONFIG_UPDATE 0x01
	#define SPA_ASYNC_REMOVE 0x02
	#define SPA_ASYNC_PROBE 0x04
	#define SPA_ASYNC_RESILVER_DONE 0x08
	#define SPA_ASYNC_RESILVER 0x10
	#define SPA_ASYNC_AUTOEXPAND 0x20
	#define SPA_ASYNC_REMOVE_DONE 0x40
	#define SPA_ASYNC_REMOVE_STOP 0x80
	#define SPA_ASYNC_INITIALIZE_RESTART 0x100
	#define SPA_ASYNC_TRIM_RESTART 0x200
	#define SPA_ASYNC_AUTOTRIM_RESTART 0x400
	#define SPA_ASYNC_L2CACHE_REBUILD 0x800
	#define SPA_ASYNC_L2CACHE_TRIM 0x1000
	#define SPA_ASYNC_REBUILD_DONE 0x2000

	/* device manipulation */
	extern int spa_vdev_add(spa_t spa, nvlist_t nvroot);
	extern int spa_vdev_attach(spa_t spa, uint64_t guid, nvlist_t nvroot,
	int replacing, int rebuild);
	extern int spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid,
	int replace_done);
	extern int spa_vdev_alloc(spa_t *spa, uint64_t guid);
	extern int spa_vdev_noalloc(spa_t *spa, uint64_t guid);
	extern boolean_t spa_vdev_remove_active(spa_t *spa);
	extern int spa_vdev_initialize(spa_t spa, nvlist_t nv, uint64_t cmd_type,
	nvlist_t *vdev_errlist);
	extern int spa_vdev_trim(spa_t spa, nvlist_t nv, uint64_t cmd_type,
	uint64_t rate, boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist);
	extern int spa_vdev_setpath(spa_t spa, uint64_t guid, const char newpath);
	extern int spa_vdev_setfru(spa_t spa, uint64_t guid, const char newfru);
	extern int spa_vdev_split_mirror(spa_t spa, const char newname,
	nvlist_t config, nvlist_t props, boolean_t exp);

	/* spare state (which is global across all pools) */
	extern void spa_spare_add(vdev_t *vd);
	extern void spa_spare_remove(vdev_t *vd);
	extern boolean_t spa_spare_exists(uint64_t guid, uint64_t pool, int refcnt);
	extern void spa_spare_activate(vdev_t *vd);

	/* L2ARC state (which is global across all pools) */
	extern void spa_l2cache_add(vdev_t *vd);
	extern void spa_l2cache_remove(vdev_t *vd);
	extern boolean_t spa_l2cache_exists(uint64_t guid, uint64_t *pool);
	extern void spa_l2cache_activate(vdev_t *vd);
	extern void spa_l2cache_drop(spa_t *spa);

	/* scanning */
	extern int spa_scan(spa_t *spa, pool_scan_func_t func);
	extern int spa_scan_stop(spa_t *spa);
	extern int spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t flag);

	/* spa syncing */
	extern void spa_sync(spa_t spa, uint64_t txg); / only for DMU use */
	extern void spa_sync_allpools(void);

	-extern int zfs_sync_pass_deferred_free;
	+extern uint_t zfs_sync_pass_deferred_free;

	/* spa namespace global mutex */
	extern kmutex_t spa_namespace_lock;

	/*
	* SPA configuration functions in spa_config.c
	*/

	#define SPA_CONFIG_UPDATE_POOL 0
	#define SPA_CONFIG_UPDATE_VDEVS 1

	-extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t);
	+extern void spa_write_cachefile(spa_t *, boolean_t, boolean_t, boolean_t);
	extern void spa_config_load(void);
	extern nvlist_t spa_all_configs(uint64_t );
	extern void spa_config_set(spa_t spa, nvlist_t config);
	extern nvlist_t spa_config_generate(spa_t spa, vdev_t *vd, uint64_t txg,
	int getstats);
	extern void spa_config_update(spa_t *spa, int what);
	extern int spa_config_parse(spa_t spa, vdev_t vdp, nvlist_t nv,
	vdev_t *parent, uint_t id, int atype);


	/*
	* Miscellaneous SPA routines in spa_misc.c
	*/

	/* Namespace manipulation */
	extern spa_t spa_lookup(const char name);
	extern spa_t spa_add(const char name, nvlist_t config, const char altroot);
	extern void spa_remove(spa_t *spa);
	extern spa_t spa_next(spa_t prev);

	/* Refcount functions */
	extern void spa_open_ref(spa_t spa, const void tag);
	extern void spa_close(spa_t spa, const void tag);
	extern void spa_async_close(spa_t spa, const void tag);
	extern boolean_t spa_refcount_zero(spa_t *spa);

	#define SCL_NONE 0x00
	#define SCL_CONFIG 0x01
	#define SCL_STATE 0x02
	#define SCL_L2ARC 0x04 /* hack until L2ARC 2.0 */
	#define SCL_ALLOC 0x08
	#define SCL_ZIO 0x10
	#define SCL_FREE 0x20
	#define SCL_VDEV 0x40
	#define SCL_LOCKS 7
	#define SCL_ALL ((1 << SCL_LOCKS) - 1)
	#define SCL_STATE_ALL (SCL_STATE \| SCL_L2ARC \| SCL_ZIO)

	/* Historical pool statistics */
	typedef struct spa_history_kstat {
	kmutex_t lock;
	uint64_t count;
	uint64_t size;
	kstat_t *kstat;
	void *priv;
	list_t list;
	} spa_history_kstat_t;

	typedef struct spa_history_list {
	uint64_t size;
	procfs_list_t procfs_list;
	} spa_history_list_t;

	typedef struct spa_stats {
	spa_history_list_t read_history;
	spa_history_list_t txg_history;
	spa_history_kstat_t tx_assign_histogram;
	spa_history_list_t mmp_history;
	spa_history_kstat_t state; /* pool state */
	spa_history_kstat_t guid; /* pool guid */
	spa_history_kstat_t iostats;
	} spa_stats_t;

	typedef enum txg_state {
	TXG_STATE_BIRTH = 0,
	TXG_STATE_OPEN = 1,
	TXG_STATE_QUIESCED = 2,
	TXG_STATE_WAIT_FOR_SYNC = 3,
	TXG_STATE_SYNCED = 4,
	TXG_STATE_COMMITTED = 5,
	} txg_state_t;

	typedef struct txg_stat {
	vdev_stat_t vs1;
	vdev_stat_t vs2;
	uint64_t txg;
	uint64_t ndirty;
	} txg_stat_t;

	/* Assorted pool IO kstats */
	typedef struct spa_iostats {
	kstat_named_t trim_extents_written;
	kstat_named_t trim_bytes_written;
	kstat_named_t trim_extents_skipped;
	kstat_named_t trim_bytes_skipped;
	kstat_named_t trim_extents_failed;
	kstat_named_t trim_bytes_failed;
	kstat_named_t autotrim_extents_written;
	kstat_named_t autotrim_bytes_written;
	kstat_named_t autotrim_extents_skipped;
	kstat_named_t autotrim_bytes_skipped;
	kstat_named_t autotrim_extents_failed;
	kstat_named_t autotrim_bytes_failed;
	kstat_named_t simple_trim_extents_written;
	kstat_named_t simple_trim_bytes_written;
	kstat_named_t simple_trim_extents_skipped;
	kstat_named_t simple_trim_bytes_skipped;
	kstat_named_t simple_trim_extents_failed;
	kstat_named_t simple_trim_bytes_failed;
	} spa_iostats_t;

	extern void spa_stats_init(spa_t *spa);
	extern void spa_stats_destroy(spa_t *spa);
	extern void spa_read_history_add(spa_t spa, const zbookmark_phys_t zb,
	uint32_t aflags);
	extern void spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time);
	extern int spa_txg_history_set(spa_t *spa, uint64_t txg,
	txg_state_t completed_state, hrtime_t completed_time);
	extern txg_stat_t spa_txg_history_init_io(spa_t , uint64_t,
	struct dsl_pool *);
	extern void spa_txg_history_fini_io(spa_t , txg_stat_t );
	extern void spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs);
	extern int spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_kstat_id);
	extern int spa_mmp_history_set(spa_t *spa, uint64_t mmp_kstat_id, int io_error,
	hrtime_t duration);
	extern void spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
	uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_kstat_id,
	int error);
	extern void spa_iostats_trim_add(spa_t *spa, trim_type_t type,
	uint64_t extents_written, uint64_t bytes_written,
	uint64_t extents_skipped, uint64_t bytes_skipped,
	uint64_t extents_failed, uint64_t bytes_failed);
	extern void spa_import_progress_add(spa_t *spa);
	extern void spa_import_progress_remove(uint64_t spa_guid);
	extern int spa_import_progress_set_mmp_check(uint64_t pool_guid,
	uint64_t mmp_sec_remaining);
	extern int spa_import_progress_set_max_txg(uint64_t pool_guid,
	uint64_t max_txg);
	extern int spa_import_progress_set_state(uint64_t pool_guid,
	spa_load_state_t spa_load_state);

	/* Pool configuration locks */
	extern int spa_config_tryenter(spa_t spa, int locks, const void tag,
	krw_t rw);
	extern void spa_config_enter(spa_t spa, int locks, const void tag, krw_t rw);
	extern void spa_config_exit(spa_t spa, int locks, const void tag);
	extern int spa_config_held(spa_t *spa, int locks, krw_t rw);

	/* Pool vdev add/remove lock */
	extern uint64_t spa_vdev_enter(spa_t *spa);
	extern uint64_t spa_vdev_detach_enter(spa_t *spa, uint64_t guid);
	extern uint64_t spa_vdev_config_enter(spa_t *spa);
	extern void spa_vdev_config_exit(spa_t spa, vdev_t vd, uint64_t txg,
	int error, const char *tag);
	extern int spa_vdev_exit(spa_t spa, vdev_t vd, uint64_t txg, int error);

	/* Pool vdev state change lock */
	extern void spa_vdev_state_enter(spa_t *spa, int oplock);
	extern int spa_vdev_state_exit(spa_t spa, vdev_t vd, int error);

	/* Log state */
	typedef enum spa_log_state {
	SPA_LOG_UNKNOWN = 0, /* unknown log state */
	SPA_LOG_MISSING, /* missing log(s) */
	SPA_LOG_CLEAR, /* clear the log(s) */
	SPA_LOG_GOOD, /* log(s) are good */
	} spa_log_state_t;

	extern spa_log_state_t spa_get_log_state(spa_t *spa);
	extern void spa_set_log_state(spa_t *spa, spa_log_state_t state);
	extern int spa_reset_logs(spa_t *spa);

	/* Log claim callback */
	extern void spa_claim_notify(zio_t *zio);
	extern void spa_deadman(void *);

	/* Accessor functions */
	extern boolean_t spa_shutting_down(spa_t *spa);
	extern struct dsl_pool spa_get_dsl(spa_t spa);
	extern boolean_t spa_is_initializing(spa_t *spa);
	extern boolean_t spa_indirect_vdevs_loaded(spa_t *spa);
	extern blkptr_t spa_get_rootblkptr(spa_t spa);
	extern void spa_set_rootblkptr(spa_t spa, const blkptr_t bp);
	extern void spa_altroot(spa_t , char , size_t);
	-extern int spa_sync_pass(spa_t *spa);
	+extern uint32_t spa_sync_pass(spa_t *spa);
	extern char spa_name(spa_t spa);
	extern uint64_t spa_guid(spa_t *spa);
	extern uint64_t spa_load_guid(spa_t *spa);
	extern uint64_t spa_last_synced_txg(spa_t *spa);
	extern uint64_t spa_first_txg(spa_t *spa);
	extern uint64_t spa_syncing_txg(spa_t *spa);
	extern uint64_t spa_final_dirty_txg(spa_t *spa);
	extern uint64_t spa_version(spa_t *spa);
	extern pool_state_t spa_state(spa_t *spa);
	extern spa_load_state_t spa_load_state(spa_t *spa);
	extern uint64_t spa_freeze_txg(spa_t *spa);
	extern uint64_t spa_get_worst_case_asize(spa_t *spa, uint64_t lsize);
	extern uint64_t spa_get_dspace(spa_t *spa);
	extern uint64_t spa_get_checkpoint_space(spa_t *spa);
	extern uint64_t spa_get_slop_space(spa_t *spa);
	extern void spa_update_dspace(spa_t *spa);
	extern uint64_t spa_version(spa_t *spa);
	extern boolean_t spa_deflate(spa_t *spa);
	extern metaslab_class_t spa_normal_class(spa_t spa);
	extern metaslab_class_t spa_log_class(spa_t spa);
	extern metaslab_class_t spa_embedded_log_class(spa_t spa);
	extern metaslab_class_t spa_special_class(spa_t spa);
	extern metaslab_class_t spa_dedup_class(spa_t spa);
	extern metaslab_class_t spa_preferred_class(spa_t spa, uint64_t size,
	dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);

	extern void spa_evicting_os_register(spa_t , objset_t os);
	extern void spa_evicting_os_deregister(spa_t , objset_t os);
	extern void spa_evicting_os_wait(spa_t *spa);
	extern int spa_max_replication(spa_t *spa);
	extern int spa_prev_software_version(spa_t *spa);
	extern uint64_t spa_get_failmode(spa_t *spa);
	extern uint64_t spa_get_deadman_failmode(spa_t *spa);
	extern void spa_set_deadman_failmode(spa_t spa, const char failmode);
	extern boolean_t spa_suspended(spa_t *spa);
	extern uint64_t spa_bootfs(spa_t *spa);
	extern uint64_t spa_delegation(spa_t *spa);
	extern objset_t spa_meta_objset(spa_t spa);
	extern space_map_t spa_syncing_log_sm(spa_t spa);
	extern uint64_t spa_deadman_synctime(spa_t *spa);
	extern uint64_t spa_deadman_ziotime(spa_t *spa);
	extern uint64_t spa_dirty_data(spa_t *spa);
	extern spa_autotrim_t spa_get_autotrim(spa_t *spa);

	/* Miscellaneous support routines */
	extern void spa_load_failed(spa_t spa, const char fmt, ...)
	__attribute__((format(printf, 2, 3)));
	extern void spa_load_note(spa_t spa, const char fmt, ...)
	__attribute__((format(printf, 2, 3)));
	extern void spa_activate_mos_feature(spa_t spa, const char feature,
	dmu_tx_t *tx);
	extern void spa_deactivate_mos_feature(spa_t spa, const char feature);
	extern spa_t *spa_by_guid(uint64_t pool_guid, uint64_t device_guid);
	extern boolean_t spa_guid_exists(uint64_t pool_guid, uint64_t device_guid);
	extern char spa_strdup(const char );
	extern void spa_strfree(char *);
	extern uint64_t spa_generate_guid(spa_t *spa);
	extern void snprintf_blkptr(char buf, size_t buflen, const blkptr_t bp);
	extern void spa_freeze(spa_t *spa);
	extern int spa_change_guid(spa_t *spa);
	extern void spa_upgrade(spa_t *spa, uint64_t version);
	extern void spa_evict_all(void);
	extern vdev_t spa_lookup_by_guid(spa_t spa, uint64_t guid,
	boolean_t l2cache);
	extern boolean_t spa_has_l2cache(spa_t *, uint64_t guid);
	extern boolean_t spa_has_spare(spa_t *, uint64_t guid);
	extern uint64_t dva_get_dsize_sync(spa_t spa, const dva_t dva);
	extern uint64_t bp_get_dsize_sync(spa_t spa, const blkptr_t bp);
	extern uint64_t bp_get_dsize(spa_t spa, const blkptr_t bp);
	extern boolean_t spa_has_slogs(spa_t *spa);
	extern boolean_t spa_is_root(spa_t *spa);
	extern boolean_t spa_writeable(spa_t *spa);
	extern boolean_t spa_has_pending_synctask(spa_t *spa);
	extern int spa_maxblocksize(spa_t *spa);
	extern int spa_maxdnodesize(spa_t *spa);
	extern boolean_t spa_has_checkpoint(spa_t *spa);
	extern boolean_t spa_importing_readonly_checkpoint(spa_t *spa);
	extern boolean_t spa_suspend_async_destroy(spa_t *spa);
	extern uint64_t spa_min_claim_txg(spa_t *spa);
	extern boolean_t zfs_dva_valid(spa_t spa, const dva_t dva,
	const blkptr_t *bp);
	typedef void (*spa_remap_cb_t)(uint64_t vdev, uint64_t offset, uint64_t size,
	void *arg);
	extern boolean_t spa_remap_blkptr(spa_t spa, blkptr_t bp,
	spa_remap_cb_t callback, void *arg);
	extern uint64_t spa_get_last_removal_txg(spa_t *spa);
	extern boolean_t spa_trust_config(spa_t *spa);
	extern uint64_t spa_missing_tvds_allowed(spa_t *spa);
	extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
	extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
	extern uint64_t spa_total_metaslabs(spa_t *spa);
	extern boolean_t spa_multihost(spa_t *spa);
	extern uint32_t spa_get_hostid(spa_t *spa);
	extern void spa_activate_allocation_classes(spa_t , dmu_tx_t );
	extern boolean_t spa_livelist_delete_check(spa_t *spa);

	extern spa_mode_t spa_mode(spa_t *spa);
	extern uint64_t zfs_strtonum(const char str, char *nptr);

	extern char *spa_his_ievent_table[];

	extern void spa_history_create_obj(spa_t spa, dmu_tx_t tx);
	extern int spa_history_get(spa_t spa, uint64_t offset, uint64_t *len_read,
	char *his_buf);
	extern int spa_history_log(spa_t spa, const char his_buf);
	extern int spa_history_log_nvl(spa_t spa, nvlist_t nvl);
	extern void spa_history_log_version(spa_t spa, const char operation,
	dmu_tx_t *tx);
	extern void spa_history_log_internal(spa_t spa, const char operation,
	dmu_tx_t tx, const char fmt, ...) __printflike(4, 5);
	extern void spa_history_log_internal_ds(struct dsl_dataset ds, const char op,
	dmu_tx_t tx, const char fmt, ...) __printflike(4, 5);
	extern void spa_history_log_internal_dd(dsl_dir_t dd, const char operation,
	dmu_tx_t tx, const char fmt, ...) __printflike(4, 5);

	extern const char spa_state_to_name(spa_t spa);

	/* error handling */
	struct zbookmark_phys;
	extern void spa_log_error(spa_t spa, const zbookmark_phys_t zb);
	extern void spa_remove_error(spa_t spa, zbookmark_phys_t zb);
	extern int zfs_ereport_post(const char clazz, spa_t spa, vdev_t *vd,
	const zbookmark_phys_t zb, zio_t zio, uint64_t state);
	extern boolean_t zfs_ereport_is_valid(const char clazz, spa_t spa, vdev_t *vd,
	zio_t *zio);
	extern void zfs_ereport_taskq_fini(void);
	extern void zfs_ereport_clear(spa_t spa, vdev_t vd);
	extern nvlist_t zfs_event_create(spa_t spa, vdev_t vd, const char type,
	const char name, nvlist_t aux);
	extern void zfs_post_remove(spa_t spa, vdev_t vd);
	extern void zfs_post_state_change(spa_t spa, vdev_t vd, uint64_t laststate);
	extern void zfs_post_autoreplace(spa_t spa, vdev_t vd);
	extern uint64_t spa_get_errlog_size(spa_t *spa);
	extern int spa_get_errlog(spa_t spa, void uaddr, uint64_t *count);
	extern void spa_errlog_rotate(spa_t *spa);
	extern void spa_errlog_drain(spa_t *spa);
	extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
	extern void spa_get_errlists(spa_t spa, avl_tree_t last, avl_tree_t *scrub);
	extern void spa_delete_dataset_errlog(spa_t spa, uint64_t ds, dmu_tx_t tx);
	extern void spa_swap_errlog(spa_t *spa, uint64_t new_head_ds,
	uint64_t old_head_ds, dmu_tx_t *tx);
	extern void sync_error_list(spa_t spa, avl_tree_t t, uint64_t *obj,
	dmu_tx_t *tx);
	extern void spa_upgrade_errlog(spa_t spa, dmu_tx_t tx);

	/* vdev cache */
	extern void vdev_cache_stat_init(void);
	extern void vdev_cache_stat_fini(void);

	/* vdev mirror */
	extern void vdev_mirror_stat_init(void);
	extern void vdev_mirror_stat_fini(void);

	/* Initialization and termination */
	extern void spa_init(spa_mode_t mode);
	extern void spa_fini(void);
	extern void spa_boot_init(void);

	/* properties */
	extern int spa_prop_set(spa_t spa, nvlist_t nvp);
	extern int spa_prop_get(spa_t spa, nvlist_t *nvp);
	extern void spa_prop_clear_bootfs(spa_t spa, uint64_t obj, dmu_tx_t tx);
	extern void spa_configfile_set(spa_t , nvlist_t , boolean_t);

	/* asynchronous event notification */
	extern void spa_event_notify(spa_t spa, vdev_t vdev, nvlist_t *hist_nvl,
	const char *name);
	extern void zfs_ereport_zvol_post(const char subclass, const char name,
	const char device_name, const char raw_name);

	/* waiting for pool activities to complete */
	extern int spa_wait(const char *pool, zpool_wait_activity_t activity,
	boolean_t *waited);
	extern int spa_wait_tag(const char *name, zpool_wait_activity_t activity,
	uint64_t tag, boolean_t *waited);
	extern void spa_notify_waiters(spa_t *spa);
	extern void spa_wake_waiters(spa_t *spa);

	extern void spa_import_os(spa_t *spa);
	extern void spa_export_os(spa_t *spa);
	extern void spa_activate_os(spa_t *spa);
	extern void spa_deactivate_os(spa_t *spa);

	/* module param call functions */
	int param_set_deadman_ziotime(ZFS_MODULE_PARAM_ARGS);
	int param_set_deadman_synctime(ZFS_MODULE_PARAM_ARGS);
	int param_set_slop_shift(ZFS_MODULE_PARAM_ARGS);
	int param_set_deadman_failmode(ZFS_MODULE_PARAM_ARGS);

	#ifdef ZFS_DEBUG
	#define dprintf_bp(bp, fmt, ...) do { \
	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
	char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \
	snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, (bp)); \
	dprintf(fmt " %s\n", __VA_ARGS__, __blkbuf); \
	kmem_free(__blkbuf, BP_SPRINTF_LEN); \
	} \
	} while (0)
	#else
	#define dprintf_bp(bp, fmt, ...)
	#endif

	extern spa_mode_t spa_mode_global;
	extern int zfs_deadman_enabled;
	extern unsigned long zfs_deadman_synctime_ms;
	extern unsigned long zfs_deadman_ziotime_ms;
	extern unsigned long zfs_deadman_checktime_ms;

	extern kmem_cache_t *zio_buf_cache[];
	extern kmem_cache_t *zio_data_buf_cache[];

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_SPA_H */
	diff --git a/sys/contrib/openzfs/include/sys/spa_impl.h b/sys/contrib/openzfs/include/sys/spa_impl.h
	index 469b1266e453..cde08ec9b0f4 100644
	--- a/sys/contrib/openzfs/include/sys/spa_impl.h
	+++ b/sys/contrib/openzfs/include/sys/spa_impl.h
	@@ -1,468 +1,468 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019 Datto Inc.
	*/

	#ifndef _SYS_SPA_IMPL_H
	#define _SYS_SPA_IMPL_H

	#include <sys/spa.h>
	#include <sys/spa_checkpoint.h>
	#include <sys/spa_log_spacemap.h>
	#include <sys/vdev.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_removal.h>
	#include <sys/metaslab.h>
	#include <sys/dmu.h>
	#include <sys/dsl_pool.h>
	#include <sys/uberblock_impl.h>
	#include <sys/zfs_context.h>
	#include <sys/avl.h>
	#include <sys/zfs_refcount.h>
	#include <sys/bplist.h>
	#include <sys/bpobj.h>
	#include <sys/dsl_crypt.h>
	#include <sys/zfeature.h>
	#include <sys/zthr.h>
	#include <sys/dsl_deadlist.h>
	#include <zfeature_common.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	typedef struct spa_alloc {
	kmutex_t spaa_lock;
	avl_tree_t spaa_tree;
	} ____cacheline_aligned spa_alloc_t;

	typedef struct spa_error_entry {
	zbookmark_phys_t se_bookmark;
	char *se_name;
	avl_node_t se_avl;
	} spa_error_entry_t;

	typedef struct spa_history_phys {
	uint64_t sh_pool_create_len; /* ending offset of zpool create */
	uint64_t sh_phys_max_off; /* physical EOF */
	uint64_t sh_bof; /* logical BOF */
	uint64_t sh_eof; /* logical EOF */
	uint64_t sh_records_lost; /* num of records overwritten */
	} spa_history_phys_t;

	/*
	* All members must be uint64_t, for byteswap purposes.
	*/
	typedef struct spa_removing_phys {
	uint64_t sr_state; /* dsl_scan_state_t */

	/*
	* The vdev ID that we most recently attempted to remove,
	* or -1 if no removal has been attempted.
	*/
	uint64_t sr_removing_vdev;

	/*
	* The vdev ID that we most recently successfully removed,
	* or -1 if no devices have been removed.
	*/
	uint64_t sr_prev_indirect_vdev;

	uint64_t sr_start_time;
	uint64_t sr_end_time;

	/*
	* Note that we can not use the space map's or indirect mapping's
	* accounting as a substitute for these values, because we need to
	* count frees of not-yet-copied data as though it did the copy.
	* Otherwise, we could get into a situation where copied > to_copy,
	* or we complete before copied == to_copy.
	*/
	uint64_t sr_to_copy; /* bytes that need to be copied */
	uint64_t sr_copied; /* bytes that have been copied or freed */
	} spa_removing_phys_t;

	/*
	* This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT
	* (with key DMU_POOL_CONDENSING_INDIRECT). It is present if a condense
	* of an indirect vdev's mapping object is in progress.
	*/
	typedef struct spa_condensing_indirect_phys {
	/*
	* The vdev ID of the indirect vdev whose indirect mapping is
	* being condensed.
	*/
	uint64_t scip_vdev;

	/*
	* The vdev's old obsolete spacemap. This spacemap's contents are
	* being integrated into the new mapping.
	*/
	uint64_t scip_prev_obsolete_sm_object;

	/*
	* The new mapping object that is being created.
	*/
	uint64_t scip_next_mapping_object;
	} spa_condensing_indirect_phys_t;

	struct spa_aux_vdev {
	uint64_t sav_object; /* MOS object for device list */
	nvlist_t sav_config; / cached device config */
	vdev_t *sav_vdevs; / devices */
	int sav_count; /* number devices */
	boolean_t sav_sync; /* sync the device list */
	nvlist_t *sav_pending; / pending device additions */
	uint_t sav_npending; /* # pending devices */
	};

	typedef struct spa_config_lock {
	kmutex_t scl_lock;
	kthread_t *scl_writer;
	int scl_write_wanted;
	int scl_count;
	kcondvar_t scl_cv;
	} ____cacheline_aligned spa_config_lock_t;

	typedef struct spa_config_dirent {
	list_node_t scd_link;
	char *scd_path;
	} spa_config_dirent_t;

	typedef enum zio_taskq_type {
	ZIO_TASKQ_ISSUE = 0,
	ZIO_TASKQ_ISSUE_HIGH,
	ZIO_TASKQ_INTERRUPT,
	ZIO_TASKQ_INTERRUPT_HIGH,
	ZIO_TASKQ_TYPES
	} zio_taskq_type_t;

	/*
	* State machine for the zpool-poolname process. The states transitions
	* are done as follows:
	*
	* From To Routine
	* PROC_NONE -> PROC_CREATED spa_activate()
	* PROC_CREATED -> PROC_ACTIVE spa_thread()
	* PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate()
	* PROC_DEACTIVATE -> PROC_GONE spa_thread()
	* PROC_GONE -> PROC_NONE spa_deactivate()
	*/
	typedef enum spa_proc_state {
	SPA_PROC_NONE, /* spa_proc = &p0, no process created */
	SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */
	SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */
	SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */
	SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */
	} spa_proc_state_t;

	typedef struct spa_taskqs {
	uint_t stqs_count;
	taskq_t **stqs_taskq;
	} spa_taskqs_t;

	typedef enum spa_all_vdev_zap_action {
	AVZ_ACTION_NONE = 0,
	AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */
	AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */
	AVZ_ACTION_INITIALIZE
	} spa_avz_action_t;

	typedef enum spa_config_source {
	SPA_CONFIG_SRC_NONE = 0,
	SPA_CONFIG_SRC_SCAN, /* scan of path (default: /dev/dsk) */
	SPA_CONFIG_SRC_CACHEFILE, /* any cachefile */
	SPA_CONFIG_SRC_TRYIMPORT, /* returned from call to tryimport */
	SPA_CONFIG_SRC_SPLIT, /* new pool in a pool split */
	SPA_CONFIG_SRC_MOS /* MOS, but not always from right txg */
	} spa_config_source_t;

	struct spa {
	/*
	* Fields protected by spa_namespace_lock.
	*/
	char spa_name[ZFS_MAX_DATASET_NAME_LEN]; /* pool name */
	char spa_comment; / comment */
	avl_node_t spa_avl; /* node in spa_namespace_avl */
	nvlist_t spa_config; / last synced config */
	nvlist_t spa_config_syncing; / currently syncing config */
	nvlist_t spa_config_splitting; / config for splitting */
	nvlist_t spa_load_info; / info and errors from load */
	uint64_t spa_config_txg; /* txg of last config change */
	- int spa_sync_pass; /* iterate-to-convergence */
	+ uint32_t spa_sync_pass; /* iterate-to-convergence */
	pool_state_t spa_state; /* pool state */
	int spa_inject_ref; /* injection references */
	uint8_t spa_sync_on; /* sync threads are running */
	spa_load_state_t spa_load_state; /* current load operation */
	boolean_t spa_indirect_vdevs_loaded; /* mappings loaded? */
	boolean_t spa_trust_config; /* do we trust vdev tree? */
	boolean_t spa_is_splitting; /* in the middle of a split? */
	spa_config_source_t spa_config_source; /* where config comes from? */
	uint64_t spa_import_flags; /* import specific flags */
	spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
	dsl_pool_t *spa_dsl_pool;
	boolean_t spa_is_initializing; /* true while opening pool */
	boolean_t spa_is_exporting; /* true while exporting pool */
	metaslab_class_t spa_normal_class; / normal data class */
	metaslab_class_t spa_log_class; / intent log data class */
	metaslab_class_t spa_embedded_log_class; / log on normal vdevs */
	metaslab_class_t spa_special_class; / special allocation class */
	metaslab_class_t spa_dedup_class; / dedup allocation class */
	uint64_t spa_first_txg; /* first txg after spa_open() */
	uint64_t spa_final_txg; /* txg of export/destroy */
	uint64_t spa_freeze_txg; /* freeze pool at this txg */
	uint64_t spa_load_max_txg; /* best initial ub_txg */
	uint64_t spa_claim_max_txg; /* highest claimed birth txg */
	inode_timespec_t spa_loaded_ts; /* 1st successful open time */
	objset_t spa_meta_objset; / copy of dp->dp_meta_objset */
	kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */
	list_t spa_evicting_os_list; /* Objsets being evicted. */
	kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */
	txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
	vdev_t spa_root_vdev; / top-level vdev container */
	uint64_t spa_min_ashift; /* of vdevs in normal class */
	uint64_t spa_max_ashift; /* of vdevs in normal class */
	uint64_t spa_min_alloc; /* of vdevs in normal class */
	uint64_t spa_config_guid; /* config pool guid */
	uint64_t spa_load_guid; /* spa_load initialized guid */
	uint64_t spa_last_synced_guid; /* last synced guid */
	list_t spa_config_dirty_list; /* vdevs with dirty config */
	list_t spa_state_dirty_list; /* vdevs with dirty state */
	/*
	* spa_allocs is an array, whose lengths is stored in spa_alloc_count.
	* There is one tree and one lock for each allocator, to help improve
	* allocation performance in write-heavy workloads.
	*/
	spa_alloc_t *spa_allocs;
	int spa_alloc_count;

	spa_aux_vdev_t spa_spares; /* hot spares */
	spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
	nvlist_t spa_label_features; / Features for reading MOS */
	uint64_t spa_config_object; /* MOS object for pool config */
	uint64_t spa_config_generation; /* config generation number */
	uint64_t spa_syncing_txg; /* txg currently syncing */
	bpobj_t spa_deferred_bpobj; /* deferred-free bplist */
	bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
	zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */
	/* checksum context templates */
	kmutex_t spa_cksum_tmpls_lock;
	void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
	uberblock_t spa_ubsync; /* last synced uberblock */
	uberblock_t spa_uberblock; /* current uberblock */
	boolean_t spa_extreme_rewind; /* rewind past deferred frees */
	kmutex_t spa_scrub_lock; /* resilver/scrub lock */
	uint64_t spa_scrub_inflight; /* in-flight scrub bytes */

	/* in-flight verification bytes */
	uint64_t spa_load_verify_bytes;
	kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
	uint8_t spa_scrub_active; /* active or suspended? */
	uint8_t spa_scrub_type; /* type of scrub we're doing */
	uint8_t spa_scrub_finished; /* indicator to rotate logs */
	uint8_t spa_scrub_started; /* started since last boot */
	uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
	uint64_t spa_scan_pass_start; /* start time per pass/reboot */
	uint64_t spa_scan_pass_scrub_pause; /* scrub pause time */
	uint64_t spa_scan_pass_scrub_spent_paused; /* total paused */
	uint64_t spa_scan_pass_exam; /* examined bytes per pass */
	uint64_t spa_scan_pass_issued; /* issued bytes per pass */

	/*
	* We are in the middle of a resilver, and another resilver
	* is needed once this one completes. This is set iff any
	* vdev_resilver_deferred is set.
	*/
	boolean_t spa_resilver_deferred;
	kmutex_t spa_async_lock; /* protect async state */
	kthread_t spa_async_thread; / thread doing async task */
	int spa_async_suspended; /* async tasks suspended */
	kcondvar_t spa_async_cv; /* wait for thread_exit() */
	uint16_t spa_async_tasks; /* async task mask */
	uint64_t spa_missing_tvds; /* unopenable tvds on load */
	uint64_t spa_missing_tvds_allowed; /* allow loading spa? */

	uint64_t spa_nonallocating_dspace;
	spa_removing_phys_t spa_removing_phys;
	spa_vdev_removal_t *spa_vdev_removal;

	spa_condensing_indirect_phys_t spa_condensing_indirect_phys;
	spa_condensing_indirect_t *spa_condensing_indirect;
	zthr_t spa_condense_zthr; / zthr doing condense. */

	uint64_t spa_checkpoint_txg; /* the txg of the checkpoint */
	spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */
	zthr_t *spa_checkpoint_discard_zthr;

	space_map_t spa_syncing_log_sm; / current log space map */
	avl_tree_t spa_sm_logs_by_txg;
	kmutex_t spa_flushed_ms_lock; /* for metaslabs_by_flushed */
	avl_tree_t spa_metaslabs_by_flushed;
	spa_unflushed_stats_t spa_unflushed_stats;
	list_t spa_log_summary;
	uint64_t spa_log_flushall_txg;

	zthr_t spa_livelist_delete_zthr; / deleting livelists */
	zthr_t spa_livelist_condense_zthr; / condensing livelists */
	uint64_t spa_livelists_to_delete; /* set of livelists to free */
	livelist_condense_entry_t spa_to_condense; /* next to condense */

	char spa_root; / alternate root directory */
	uint64_t spa_ena; /* spa-wide ereport ENA */
	int spa_last_open_failed; /* error if last open failed */
	uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */
	uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */
	uint64_t spa_load_txg; /* ub txg that loaded */
	uint64_t spa_load_txg_ts; /* timestamp from that ub */
	uint64_t spa_load_meta_errors; /* verify metadata err count */
	uint64_t spa_load_data_errors; /* verify data err count */
	uint64_t spa_verify_min_txg; /* start txg of verify scrub */
	kmutex_t spa_errlog_lock; /* error log lock */
	uint64_t spa_errlog_last; /* last error log object */
	uint64_t spa_errlog_scrub; /* scrub error log object */
	kmutex_t spa_errlist_lock; /* error list/ereport lock */
	avl_tree_t spa_errlist_last; /* last error list */
	avl_tree_t spa_errlist_scrub; /* scrub error list */
	avl_tree_t spa_errlist_healed; /* list of healed blocks */
	uint64_t spa_deflate; /* should we deflate? */
	uint64_t spa_history; /* history object */
	kmutex_t spa_history_lock; /* history lock */
	vdev_t spa_pending_vdev; / pending vdev additions */
	kmutex_t spa_props_lock; /* property lock */
	uint64_t spa_pool_props_object; /* object for properties */
	uint64_t spa_bootfs; /* default boot filesystem */
	uint64_t spa_failmode; /* failure mode for the pool */
	uint64_t spa_deadman_failmode; /* failure mode for deadman */
	uint64_t spa_delegation; /* delegation on/off */
	list_t spa_config_list; /* previous cache file(s) */
	/* per-CPU array of root of async I/O: */
	zio_t **spa_async_zio_root;
	zio_t spa_suspend_zio_root; / root of all suspended I/O */
	zio_t spa_txg_zio[TXG_SIZE]; / spa_sync() waits for this */
	kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
	kcondvar_t spa_suspend_cv; /* notification of resume */
	zio_suspend_reason_t spa_suspended; /* pool is suspended */
	uint8_t spa_claiming; /* pool is doing zil_claim() */
	boolean_t spa_is_root; /* pool is root */
	int spa_minref; /* num refs when first opened */
	spa_mode_t spa_mode; /* SPA_MODE_{READ\|WRITE} */
	boolean_t spa_read_spacemaps; /* spacemaps available if ro */
	spa_log_state_t spa_log_state; /* log state */
	uint64_t spa_autoexpand; /* lun expansion on/off */
	ddt_t spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; / in-core DDTs */
	uint64_t spa_ddt_stat_object; /* DDT statistics */
	uint64_t spa_dedup_dspace; /* Cache get_dedup_dspace() */
	uint64_t spa_dedup_checksum; /* default dedup checksum */
	uint64_t spa_dspace; /* dspace in normal class */
	kmutex_t spa_vdev_top_lock; /* dueling offline/remove */
	kmutex_t spa_proc_lock; /* protects spa_proc* */
	kcondvar_t spa_proc_cv; /* spa_proc_state transitions */
	spa_proc_state_t spa_proc_state; /* see definition */
	proc_t spa_proc; / "zpool-poolname" process */
	uintptr_t spa_did; /* if procp != p0, did of t1 */
	boolean_t spa_autoreplace; /* autoreplace set in open */
	int spa_vdev_locks; /* locks grabbed */
	uint64_t spa_creation_version; /* version at pool creation */
	uint64_t spa_prev_software_version; /* See ub_software_version */
	uint64_t spa_feat_for_write_obj; /* required to write to pool */
	uint64_t spa_feat_for_read_obj; /* required to read from pool */
	uint64_t spa_feat_desc_obj; /* Feature descriptions */
	uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */
	kmutex_t spa_feat_stats_lock; /* protects spa_feat_stats */
	nvlist_t spa_feat_stats; / Cache of enabled features */
	/* cache feature refcounts */
	uint64_t spa_feat_refcount_cache[SPA_FEATURES];
	taskqid_t spa_deadman_tqid; /* Task id */
	uint64_t spa_deadman_calls; /* number of deadman calls */
	hrtime_t spa_sync_starttime; /* starting time of spa_sync */
	uint64_t spa_deadman_synctime; /* deadman sync expiration */
	uint64_t spa_deadman_ziotime; /* deadman zio expiration */
	uint64_t spa_all_vdev_zaps; /* ZAP of per-vd ZAP obj #s */
	spa_avz_action_t spa_avz_action; /* destroy/rebuild AVZ? */
	uint64_t spa_autotrim; /* automatic background trim? */
	uint64_t spa_errata; /* errata issues detected */
	spa_stats_t spa_stats; /* assorted spa statistics */
	spa_keystore_t spa_keystore; /* loaded crypto keys */

	/* arc_memory_throttle() parameters during low memory condition */
	uint64_t spa_lowmem_page_load; /* memory load during txg */
	uint64_t spa_lowmem_last_txg; /* txg window start */

	hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */
	taskq_t spa_zvol_taskq; / Taskq for minor management */
	taskq_t spa_prefetch_taskq; / Taskq for prefetch threads */
	uint64_t spa_multihost; /* multihost aware (mmp) */
	mmp_thread_t spa_mmp; /* multihost mmp thread */
	list_t spa_leaf_list; /* list of leaf vdevs */
	uint64_t spa_leaf_list_gen; /* track leaf_list changes */
	uint32_t spa_hostid; /* cached system hostid */

	/* synchronization for threads in spa_wait */
	kmutex_t spa_activities_lock;
	kcondvar_t spa_activities_cv;
	kcondvar_t spa_waiters_cv;
	int spa_waiters; /* number of waiting threads */
	boolean_t spa_waiters_cancel; /* waiters should return */

	char spa_compatibility; / compatibility file(s) */

	/*
	* spa_refcount & spa_config_lock must be the last elements
	* because zfs_refcount_t changes size based on compilation options.
	* In order for the MDB module to function correctly, the other
	* fields must remain in the same location.
	*/
	spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
	zfs_refcount_t spa_refcount; /* number of opens */

	taskq_t spa_upgrade_taskq; / taskq for upgrade jobs */
	};

	extern char *spa_config_path;
	extern const char *zfs_deadman_failmode;
	-extern int spa_slop_shift;
	+extern uint_t spa_slop_shift;
	extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	task_func_t func, void arg, uint_t flags, taskq_ent_t *ent);
	extern void spa_taskq_dispatch_sync(spa_t *, zio_type_t t, zio_taskq_type_t q,
	task_func_t func, void arg, uint_t flags);
	extern void spa_load_spares(spa_t *spa);
	extern void spa_load_l2cache(spa_t *spa);
	extern sysevent_t spa_event_create(spa_t spa, vdev_t vd, nvlist_t hist_nvl,
	const char *name);
	extern void spa_event_post(sysevent_t *ev);
	extern int param_set_deadman_failmode_common(const char *val);
	extern void spa_set_deadman_synctime(hrtime_t ns);
	extern void spa_set_deadman_ziotime(hrtime_t ns);
	extern const char *spa_history_zone(void);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_SPA_IMPL_H */
	diff --git a/sys/contrib/openzfs/include/sys/txg.h b/sys/contrib/openzfs/include/sys/txg.h
	index c0ee348a61bc..46945210cdb5 100644
	--- a/sys/contrib/openzfs/include/sys/txg.h
	+++ b/sys/contrib/openzfs/include/sys/txg.h
	@@ -1,154 +1,154 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/
	/*
	* Copyright (c) 2012, 2017 by Delphix. All rights reserved.
	*/

	#ifndef _SYS_TXG_H
	#define _SYS_TXG_H

	#include <sys/spa.h>
	#include <sys/zfs_context.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	#define TXG_CONCURRENT_STATES 3 /* open, quiescing, syncing */
	#define TXG_SIZE 4 /* next power of 2 */
	#define TXG_MASK (TXG_SIZE - 1) /* mask for size */
	#define TXG_INITIAL TXG_SIZE /* initial txg */
	#define TXG_IDX (txg & TXG_MASK)
	#define TXG_UNKNOWN 0

	/* Number of txgs worth of frees we defer adding to in-core spacemaps */
	#define TXG_DEFER_SIZE 2

	typedef struct tx_cpu tx_cpu_t;

	typedef struct txg_handle {
	tx_cpu_t *th_cpu;
	uint64_t th_txg;
	} txg_handle_t;

	typedef struct txg_node {
	struct txg_node *tn_next[TXG_SIZE];
	uint8_t tn_member[TXG_SIZE];
	} txg_node_t;

	typedef struct txg_list {
	kmutex_t tl_lock;
	size_t tl_offset;
	spa_t *tl_spa;
	txg_node_t *tl_head[TXG_SIZE];
	} txg_list_t;

	struct dsl_pool;

	extern void txg_init(struct dsl_pool *dp, uint64_t txg);
	extern void txg_fini(struct dsl_pool *dp);
	extern void txg_sync_start(struct dsl_pool *dp);
	extern void txg_sync_stop(struct dsl_pool *dp);
	extern uint64_t txg_hold_open(struct dsl_pool dp, txg_handle_t txghp);
	extern void txg_rele_to_quiesce(txg_handle_t *txghp);
	extern void txg_rele_to_sync(txg_handle_t *txghp);
	extern void txg_register_callbacks(txg_handle_t txghp, list_t tx_callbacks);

	extern void txg_delay(struct dsl_pool *dp, uint64_t txg, hrtime_t delta,
	hrtime_t resolution);
	extern void txg_kick(struct dsl_pool *dp, uint64_t txg);

	/*
	* Wait until the given transaction group has finished syncing.
	* Try to make this happen as soon as possible (eg. kick off any
	* necessary syncs immediately). If txg==0, wait for the currently open
	* txg to finish syncing.
	*/
	extern void txg_wait_synced(struct dsl_pool *dp, uint64_t txg);

	/*
	* Wait as above. Returns true if the thread was signaled while waiting.
	*/
	extern boolean_t txg_wait_synced_sig(struct dsl_pool *dp, uint64_t txg);

	/*
	* Wait until the given transaction group, or one after it, is
	* the open transaction group. Try to make this happen as soon
	* as possible (eg. kick off any necessary syncs immediately) when
	* should_quiesce is set. If txg == 0, wait for the next open txg.
	*/
	extern void txg_wait_open(struct dsl_pool *dp, uint64_t txg,
	boolean_t should_quiesce);

	/*
	* Returns TRUE if we are "backed up" waiting for the syncing
	* transaction to complete; otherwise returns FALSE.
	*/
	extern boolean_t txg_stalled(struct dsl_pool *dp);

	/* returns TRUE if someone is waiting for the next txg to sync */
	extern boolean_t txg_sync_waiting(struct dsl_pool *dp);

	extern void txg_verify(spa_t *spa, uint64_t txg);

	/*
	* Wait for pending commit callbacks of already-synced transactions to finish
	* processing.
	*/
	extern void txg_wait_callbacks(struct dsl_pool *dp);

	/*
	* Per-txg object lists.
	*/

	#define TXG_CLEAN(txg) ((txg) - 1)

	extern void txg_list_create(txg_list_t tl, spa_t spa, size_t offset);
	extern void txg_list_destroy(txg_list_t *tl);
	extern boolean_t txg_list_empty(txg_list_t *tl, uint64_t txg);
	extern boolean_t txg_all_lists_empty(txg_list_t *tl);
	extern boolean_t txg_list_add(txg_list_t tl, void p, uint64_t txg);
	extern boolean_t txg_list_add_tail(txg_list_t tl, void p, uint64_t txg);
	extern void txg_list_remove(txg_list_t tl, uint64_t txg);
	extern void txg_list_remove_this(txg_list_t tl, void *p, uint64_t txg);
	extern boolean_t txg_list_member(txg_list_t tl, void p, uint64_t txg);
	extern void txg_list_head(txg_list_t tl, uint64_t txg);
	extern void txg_list_next(txg_list_t tl, void *p, uint64_t txg);

	/* Global tuning */
	-extern int zfs_txg_timeout;
	+extern uint_t zfs_txg_timeout;


	#ifdef ZFS_DEBUG
	#define TXG_VERIFY(spa, txg) txg_verify(spa, txg)
	#else
	#define TXG_VERIFY(spa, txg)
	#endif

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_TXG_H */
	diff --git a/sys/contrib/openzfs/include/sys/vdev.h b/sys/contrib/openzfs/include/sys/vdev.h
	index 5fec1d51a5f2..7a7c70dc1598 100644
	--- a/sys/contrib/openzfs/include/sys/vdev.h
	+++ b/sys/contrib/openzfs/include/sys/vdev.h
	@@ -1,229 +1,232 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, Datto Inc. All rights reserved.
	*/

	#ifndef _SYS_VDEV_H
	#define _SYS_VDEV_H

	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/dmu.h>
	#include <sys/space_map.h>
	#include <sys/metaslab.h>
	#include <sys/fs/zfs.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	typedef enum vdev_dtl_type {
	DTL_MISSING, /* 0% replication: no copies of the data */
	DTL_PARTIAL, /* less than 100% replication: some copies missing */
	DTL_SCRUB, /* unable to fully repair during scrub/resilver */
	DTL_OUTAGE, /* temporarily missing (used to attempt detach) */
	DTL_TYPES
	} vdev_dtl_type_t;

	extern int zfs_nocacheflush;

	typedef boolean_t vdev_open_children_func_t(vdev_t *vd);

	extern void vdev_dbgmsg(vdev_t vd, const char fmt, ...)
	__attribute__((format(printf, 2, 3)));
	extern void vdev_dbgmsg_print_tree(vdev_t *, int);
	extern int vdev_open(vdev_t *);
	extern void vdev_open_children(vdev_t *);
	extern void vdev_open_children_subset(vdev_t , vdev_open_children_func_t );
	extern int vdev_validate(vdev_t *);
	extern int vdev_copy_path_strict(vdev_t , vdev_t );
	extern void vdev_copy_path_relaxed(vdev_t , vdev_t );
	extern void vdev_close(vdev_t *);
	extern int vdev_create(vdev_t *, uint64_t txg, boolean_t isreplace);
	extern void vdev_reopen(vdev_t *);
	extern int vdev_validate_aux(vdev_t *vd);
	extern zio_t vdev_probe(vdev_t vd, zio_t *pio);
	extern boolean_t vdev_is_concrete(vdev_t *vd);
	extern boolean_t vdev_is_bootable(vdev_t *vd);
	extern vdev_t vdev_lookup_top(spa_t spa, uint64_t vdev);
	extern vdev_t vdev_lookup_by_guid(vdev_t vd, uint64_t guid);
	extern int vdev_count_leaves(spa_t *spa);
	extern void vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t d,
	uint64_t txg, uint64_t size);
	extern boolean_t vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t d,
	uint64_t txg, uint64_t size);
	extern boolean_t vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t d);
	extern boolean_t vdev_default_need_resilver(vdev_t vd, const dva_t dva,
	size_t psize, uint64_t phys_birth);
	extern boolean_t vdev_dtl_need_resilver(vdev_t vd, const dva_t dva,
	size_t psize, uint64_t phys_birth);
	extern void vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
	boolean_t scrub_done, boolean_t rebuild_done);
	extern boolean_t vdev_dtl_required(vdev_t *vd);
	extern boolean_t vdev_resilver_needed(vdev_t *vd,
	uint64_t minp, uint64_t maxp);
	extern void vdev_destroy_unlink_zap(vdev_t *vd, uint64_t zapobj,
	dmu_tx_t *tx);
	extern uint64_t vdev_create_link_zap(vdev_t vd, dmu_tx_t tx);
	extern void vdev_construct_zaps(vdev_t vd, dmu_tx_t tx);
	extern void vdev_destroy_spacemaps(vdev_t vd, dmu_tx_t tx);
	extern void vdev_indirect_mark_obsolete(vdev_t *vd, uint64_t offset,
	uint64_t size);
	extern void spa_vdev_indirect_mark_obsolete(spa_t *spa, uint64_t vdev,
	uint64_t offset, uint64_t size, dmu_tx_t *tx);
	extern boolean_t vdev_replace_in_progress(vdev_t *vdev);

	extern void vdev_hold(vdev_t *);
	extern void vdev_rele(vdev_t *);

	extern int vdev_metaslab_init(vdev_t *vd, uint64_t txg);
	extern void vdev_metaslab_fini(vdev_t *vd);
	extern void vdev_metaslab_set_size(vdev_t *);
	extern void vdev_expand(vdev_t *vd, uint64_t txg);
	extern void vdev_split(vdev_t *vd);
	extern void vdev_deadman(vdev_t vd, const char tag);

	typedef void vdev_xlate_func_t(void arg, range_seg64_t physical_rs);

	extern boolean_t vdev_xlate_is_empty(range_seg64_t *rs);
	extern void vdev_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs);
	extern void vdev_xlate_walk(vdev_t vd, const range_seg64_t logical_rs,
	vdev_xlate_func_t func, void arg);

	extern void vdev_get_stats_ex(vdev_t vd, vdev_stat_t vs, vdev_stat_ex_t *vsx);

	extern metaslab_group_t vdev_get_mg(vdev_t vd, metaslab_class_t *mc);

	extern void vdev_get_stats(vdev_t vd, vdev_stat_t vs);
	extern void vdev_clear_stats(vdev_t *vd);
	extern void vdev_stat_update(zio_t *zio, uint64_t psize);
	extern void vdev_scan_stat_init(vdev_t *vd);
	extern void vdev_propagate_state(vdev_t *vd);
	extern void vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state,
	vdev_aux_t aux);
	extern boolean_t vdev_children_are_offline(vdev_t *vd);

	extern void vdev_space_update(vdev_t *vd,
	int64_t alloc_delta, int64_t defer_delta, int64_t space_delta);

	extern int64_t vdev_deflated_space(vdev_t *vd, int64_t space);

	extern uint64_t vdev_psize_to_asize(vdev_t *vd, uint64_t psize);

	/*
	* Return the amount of space allocated for a gang block header.
	*/
	static inline uint64_t
	vdev_gang_header_asize(vdev_t *vd)
	{
	return (vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE));
	}

	extern int vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux);
	extern int vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux);
	extern int vdev_online(spa_t *spa, uint64_t guid, uint64_t flags,
	vdev_state_t *);
	extern int vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags);
	+extern int vdev_remove_wanted(spa_t *spa, uint64_t guid);
	extern void vdev_clear(spa_t spa, vdev_t vd);

	extern boolean_t vdev_is_dead(vdev_t *vd);
	extern boolean_t vdev_readable(vdev_t *vd);
	extern boolean_t vdev_writeable(vdev_t *vd);
	extern boolean_t vdev_allocatable(vdev_t *vd);
	extern boolean_t vdev_accessible(vdev_t vd, zio_t zio);
	extern boolean_t vdev_is_spacemap_addressable(vdev_t *vd);

	extern void vdev_cache_init(vdev_t *vd);
	extern void vdev_cache_fini(vdev_t *vd);
	extern boolean_t vdev_cache_read(zio_t *zio);
	extern void vdev_cache_write(zio_t *zio);
	extern void vdev_cache_purge(vdev_t *vd);

	extern void vdev_queue_init(vdev_t *vd);
	extern void vdev_queue_fini(vdev_t *vd);
	extern zio_t vdev_queue_io(zio_t zio);
	extern void vdev_queue_io_done(zio_t *zio);
	extern void vdev_queue_change_io_priority(zio_t *zio, zio_priority_t priority);

	extern int vdev_queue_length(vdev_t *vd);
	extern uint64_t vdev_queue_last_offset(vdev_t *vd);

	extern void vdev_config_dirty(vdev_t *vd);
	extern void vdev_config_clean(vdev_t *vd);
	extern int vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg);

	extern void vdev_state_dirty(vdev_t *vd);
	extern void vdev_state_clean(vdev_t *vd);

	extern void vdev_defer_resilver(vdev_t *vd);
	extern boolean_t vdev_clear_resilver_deferred(vdev_t vd, dmu_tx_t tx);

	typedef enum vdev_config_flag {
	VDEV_CONFIG_SPARE = 1 << 0,
	VDEV_CONFIG_L2CACHE = 1 << 1,
	VDEV_CONFIG_REMOVING = 1 << 2,
	VDEV_CONFIG_MOS = 1 << 3,
	VDEV_CONFIG_MISSING = 1 << 4
	} vdev_config_flag_t;

	+extern void vdev_post_kobj_evt(vdev_t *vd);
	+extern void vdev_clear_kobj_evt(vdev_t *vd);
	extern void vdev_top_config_generate(spa_t spa, nvlist_t config);
	extern nvlist_t vdev_config_generate(spa_t spa, vdev_t *vd,
	boolean_t getstats, vdev_config_flag_t flags);

	/*
	* Label routines
	*/
	struct uberblock;
	extern uint64_t vdev_label_offset(uint64_t psize, int l, uint64_t offset);
	extern int vdev_label_number(uint64_t psise, uint64_t offset);
	extern nvlist_t vdev_label_read_config(vdev_t vd, uint64_t txg);
	extern void vdev_uberblock_load(vdev_t , struct uberblock , nvlist_t **);
	extern void vdev_config_generate_stats(vdev_t vd, nvlist_t nv);
	extern void vdev_label_write(zio_t zio, vdev_t vd, int l, abd_t *buf, uint64_t
	offset, uint64_t size, zio_done_func_t done, void priv, int flags);
	extern int vdev_label_read_bootenv(vdev_t , nvlist_t );
	extern int vdev_label_write_bootenv(vdev_t , nvlist_t );

	typedef enum {
	VDEV_LABEL_CREATE, /* create/add a new device */
	VDEV_LABEL_REPLACE, /* replace an existing device */
	VDEV_LABEL_SPARE, /* add a new hot spare */
	VDEV_LABEL_REMOVE, /* remove an existing device */
	VDEV_LABEL_L2CACHE, /* add an L2ARC cache device */
	VDEV_LABEL_SPLIT /* generating new label for split-off dev */
	} vdev_labeltype_t;

	extern int vdev_label_init(vdev_t *vd, uint64_t txg, vdev_labeltype_t reason);

	extern int vdev_prop_set(vdev_t vd, nvlist_t innvl, nvlist_t *outnvl);
	extern int vdev_prop_get(vdev_t vd, nvlist_t nvprops, nvlist_t *outnvl);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_VDEV_H */
	diff --git a/sys/contrib/openzfs/include/sys/vdev_impl.h b/sys/contrib/openzfs/include/sys/vdev_impl.h
	index 470eaa763d5a..b789d2c05d59 100644
	--- a/sys/contrib/openzfs/include/sys/vdev_impl.h
	+++ b/sys/contrib/openzfs/include/sys/vdev_impl.h
	@@ -1,658 +1,661 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	*/

	#ifndef _SYS_VDEV_IMPL_H
	#define _SYS_VDEV_IMPL_H

	#include <sys/avl.h>
	#include <sys/bpobj.h>
	#include <sys/dmu.h>
	#include <sys/metaslab.h>
	#include <sys/nvpair.h>
	#include <sys/space_map.h>
	#include <sys/vdev.h>
	#include <sys/dkio.h>
	#include <sys/uberblock_impl.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/vdev_indirect_births.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_removal.h>
	#include <sys/zfs_ratelimit.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Virtual device descriptors.
	*
	* All storage pool operations go through the virtual device framework,
	* which provides data replication and I/O scheduling.
	*/

	/*
	* Forward declarations that lots of things need.
	*/
	typedef struct vdev_queue vdev_queue_t;
	typedef struct vdev_cache vdev_cache_t;
	typedef struct vdev_cache_entry vdev_cache_entry_t;
	struct abd;

	-extern int zfs_vdev_queue_depth_pct;
	-extern int zfs_vdev_def_queue_depth;
	-extern uint32_t zfs_vdev_async_write_max_active;
	+extern uint_t zfs_vdev_queue_depth_pct;
	+extern uint_t zfs_vdev_def_queue_depth;
	+extern uint_t zfs_vdev_async_write_max_active;

	/*
	* Virtual device operations
	*/
	typedef int vdev_init_func_t(spa_t spa, nvlist_t nv, void **tsd);
	+typedef void vdev_kobj_post_evt_func_t(vdev_t *vd);
	typedef void vdev_fini_func_t(vdev_t *vd);
	typedef int vdev_open_func_t(vdev_t vd, uint64_t size, uint64_t *max_size,
	uint64_t ashift, uint64_t pshift);
	typedef void vdev_close_func_t(vdev_t *vd);
	typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
	typedef uint64_t vdev_min_asize_func_t(vdev_t *vd);
	typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd);
	typedef void vdev_io_start_func_t(zio_t *zio);
	typedef void vdev_io_done_func_t(zio_t *zio);
	typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
	typedef boolean_t vdev_need_resilver_func_t(vdev_t vd, const dva_t dva,
	size_t psize, uint64_t phys_birth);
	typedef void vdev_hold_func_t(vdev_t *vd);
	typedef void vdev_rele_func_t(vdev_t *vd);

	typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
	uint64_t offset, uint64_t size, void *arg);
	typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
	vdev_remap_cb_t callback, void *arg);
	/*
	* Given a target vdev, translates the logical range "in" to the physical
	* range "res"
	*/
	typedef void vdev_xlation_func_t(vdev_t cvd, const range_seg64_t logical,
	range_seg64_t physical, range_seg64_t remain);
	typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start,
	uint64_t size, uint64_t max_segment);
	typedef void vdev_metaslab_init_func_t(vdev_t vd, uint64_t startp,
	uint64_t *sizep);
	typedef void vdev_config_generate_func_t(vdev_t vd, nvlist_t nv);
	typedef uint64_t vdev_nparity_func_t(vdev_t *vd);
	typedef uint64_t vdev_ndisks_func_t(vdev_t *vd);

	typedef const struct vdev_ops {
	vdev_init_func_t *vdev_op_init;
	vdev_fini_func_t *vdev_op_fini;
	vdev_open_func_t *vdev_op_open;
	vdev_close_func_t *vdev_op_close;
	vdev_asize_func_t *vdev_op_asize;
	vdev_min_asize_func_t *vdev_op_min_asize;
	vdev_min_alloc_func_t *vdev_op_min_alloc;
	vdev_io_start_func_t *vdev_op_io_start;
	vdev_io_done_func_t *vdev_op_io_done;
	vdev_state_change_func_t *vdev_op_state_change;
	vdev_need_resilver_func_t *vdev_op_need_resilver;
	vdev_hold_func_t *vdev_op_hold;
	vdev_rele_func_t *vdev_op_rele;
	vdev_remap_func_t *vdev_op_remap;
	vdev_xlation_func_t *vdev_op_xlate;
	vdev_rebuild_asize_func_t *vdev_op_rebuild_asize;
	vdev_metaslab_init_func_t *vdev_op_metaslab_init;
	vdev_config_generate_func_t *vdev_op_config_generate;
	vdev_nparity_func_t *vdev_op_nparity;
	vdev_ndisks_func_t *vdev_op_ndisks;
	+ vdev_kobj_post_evt_func_t *vdev_op_kobj_evt_post;
	char vdev_op_type[16];
	boolean_t vdev_op_leaf;
	} vdev_ops_t;

	/*
	* Virtual device properties
	*/
	struct vdev_cache_entry {
	struct abd *ve_abd;
	uint64_t ve_offset;
	clock_t ve_lastused;
	avl_node_t ve_offset_node;
	avl_node_t ve_lastused_node;
	uint32_t ve_hits;
	uint16_t ve_missed_update;
	zio_t *ve_fill_io;
	};

	struct vdev_cache {
	avl_tree_t vc_offset_tree;
	avl_tree_t vc_lastused_tree;
	kmutex_t vc_lock;
	};

	typedef struct vdev_queue_class {
	uint32_t vqc_active;

	/*
	* Sorted by offset or timestamp, depending on if the queue is
	* LBA-ordered vs FIFO.
	*/
	avl_tree_t vqc_queued_tree;
	} vdev_queue_class_t;

	struct vdev_queue {
	vdev_t *vq_vdev;
	vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
	avl_tree_t vq_active_tree;
	avl_tree_t vq_read_offset_tree;
	avl_tree_t vq_write_offset_tree;
	avl_tree_t vq_trim_offset_tree;
	uint64_t vq_last_offset;
	zio_priority_t vq_last_prio; /* Last sent I/O priority. */
	uint32_t vq_ia_active; /* Active interactive I/Os. */
	uint32_t vq_nia_credit; /* Non-interactive I/Os credit. */
	hrtime_t vq_io_complete_ts; /* time last i/o completed */
	hrtime_t vq_io_delta_ts;
	zio_t vq_io_search; /* used as local for stack reduction */
	kmutex_t vq_lock;
	};

	typedef enum vdev_alloc_bias {
	VDEV_BIAS_NONE,
	VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */
	VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */
	VDEV_BIAS_DEDUP /* dedicated to dedup metadata */
	} vdev_alloc_bias_t;


	/*
	* On-disk indirect vdev state.
	*
	* An indirect vdev is described exclusively in the MOS config of a pool.
	* The config for an indirect vdev includes several fields, which are
	* accessed in memory by a vdev_indirect_config_t.
	*/
	typedef struct vdev_indirect_config {
	/*
	* Object (in MOS) which contains the indirect mapping. This object
	* contains an array of vdev_indirect_mapping_entry_phys_t ordered by
	* vimep_src. The bonus buffer for this object is a
	* vdev_indirect_mapping_phys_t. This object is allocated when a vdev
	* removal is initiated.
	*
	* Note that this object can be empty if none of the data on the vdev
	* has been copied yet.
	*/
	uint64_t vic_mapping_object;

	/*
	* Object (in MOS) which contains the birth times for the mapping
	* entries. This object contains an array of
	* vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
	* buffer for this object is a vdev_indirect_birth_phys_t. This object
	* is allocated when a vdev removal is initiated.
	*
	* Note that this object can be empty if none of the vdev has yet been
	* copied.
	*/
	uint64_t vic_births_object;

	/*
	* This is the vdev ID which was removed previous to this vdev, or
	* UINT64_MAX if there are no previously removed vdevs.
	*/
	uint64_t vic_prev_indirect_vdev;
	} vdev_indirect_config_t;

	/*
	* Virtual device descriptor
	*/
	struct vdev {
	/*
	* Common to all vdev types.
	*/
	uint64_t vdev_id; /* child number in vdev parent */
	uint64_t vdev_guid; /* unique ID for this vdev */
	uint64_t vdev_guid_sum; /* self guid + all child guids */
	uint64_t vdev_orig_guid; /* orig. guid prior to remove */
	uint64_t vdev_asize; /* allocatable device capacity */
	uint64_t vdev_min_asize; /* min acceptable asize */
	uint64_t vdev_max_asize; /* max acceptable asize */
	uint64_t vdev_ashift; /* block alignment shift */

	/*
	* Logical block alignment shift
	*
	* The smallest sized/aligned I/O supported by the device.
	*/
	uint64_t vdev_logical_ashift;
	/*
	* Physical block alignment shift
	*
	* The device supports logical I/Os with vdev_logical_ashift
	* size/alignment, but optimum performance will be achieved by
	* aligning/sizing requests to vdev_physical_ashift. Smaller
	* requests may be inflated or incur device level read-modify-write
	* operations.
	*
	* May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
	*/
	uint64_t vdev_physical_ashift;
	uint64_t vdev_state; /* see VDEV_STATE_* #defines */
	uint64_t vdev_prevstate; /* used when reopening a vdev */
	vdev_ops_t vdev_ops; / vdev operations */
	spa_t vdev_spa; / spa for this vdev */
	void vdev_tsd; / type-specific data */
	vdev_t vdev_top; / top-level vdev */
	vdev_t vdev_parent; / parent vdev */
	vdev_t *vdev_child; / array of children */
	uint64_t vdev_children; /* number of children */
	vdev_stat_t vdev_stat; /* virtual device statistics */
	vdev_stat_ex_t vdev_stat_ex; /* extended statistics */
	boolean_t vdev_expanding; /* expand the vdev? */
	boolean_t vdev_reopening; /* reopen in progress? */
	boolean_t vdev_nonrot; /* true if solid state */
	int vdev_load_error; /* error on last load */
	int vdev_open_error; /* error on last open */
	int vdev_validate_error; /* error on last validate */
	kthread_t vdev_open_thread; / thread opening children */
	kthread_t vdev_validate_thread; / thread validating children */
	uint64_t vdev_crtxg; /* txg when top-level was added */

	/*
	* Top-level vdev state.
	*/
	uint64_t vdev_ms_array; /* metaslab array object */
	uint64_t vdev_ms_shift; /* metaslab size shift */
	uint64_t vdev_ms_count; /* number of metaslabs */
	metaslab_group_t vdev_mg; / metaslab group */
	metaslab_group_t vdev_log_mg; / embedded slog metaslab group */
	metaslab_t *vdev_ms; / metaslab array */
	uint64_t vdev_pending_fastwrite; /* allocated fastwrites */
	txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
	txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
	txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
	boolean_t vdev_remove_wanted; /* async remove wanted? */
	boolean_t vdev_probe_wanted; /* async probe wanted? */
	list_node_t vdev_config_dirty_node; /* config dirty list */
	list_node_t vdev_state_dirty_node; /* state dirty list */
	uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
	uint64_t vdev_islog; /* is an intent log device */
	uint64_t vdev_noalloc; /* device is passivated? */
	uint64_t vdev_removing; /* device is being removed? */
	boolean_t vdev_ishole; /* is a hole in the namespace */
	uint64_t vdev_top_zap;
	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */

	/* pool checkpoint related */
	space_map_t vdev_checkpoint_sm; / contains reserved blocks */

	/* Initialize related */
	boolean_t vdev_initialize_exit_wanted;
	vdev_initializing_state_t vdev_initialize_state;
	list_node_t vdev_initialize_node;
	kthread_t *vdev_initialize_thread;
	/* Protects vdev_initialize_thread and vdev_initialize_state. */
	kmutex_t vdev_initialize_lock;
	kcondvar_t vdev_initialize_cv;
	uint64_t vdev_initialize_offset[TXG_SIZE];
	uint64_t vdev_initialize_last_offset;
	range_tree_t vdev_initialize_tree; / valid while initializing */
	uint64_t vdev_initialize_bytes_est;
	uint64_t vdev_initialize_bytes_done;
	uint64_t vdev_initialize_action_time; /* start and end time */

	/* TRIM related */
	boolean_t vdev_trim_exit_wanted;
	boolean_t vdev_autotrim_exit_wanted;
	vdev_trim_state_t vdev_trim_state;
	list_node_t vdev_trim_node;
	kmutex_t vdev_autotrim_lock;
	kcondvar_t vdev_autotrim_cv;
	kthread_t *vdev_autotrim_thread;
	/* Protects vdev_trim_thread and vdev_trim_state. */
	kmutex_t vdev_trim_lock;
	kcondvar_t vdev_trim_cv;
	kthread_t *vdev_trim_thread;
	uint64_t vdev_trim_offset[TXG_SIZE];
	uint64_t vdev_trim_last_offset;
	uint64_t vdev_trim_bytes_est;
	uint64_t vdev_trim_bytes_done;
	uint64_t vdev_trim_rate; /* requested rate (bytes/sec) */
	uint64_t vdev_trim_partial; /* requested partial TRIM */
	uint64_t vdev_trim_secure; /* requested secure TRIM */
	uint64_t vdev_trim_action_time; /* start and end time */

	/* Rebuild related */
	boolean_t vdev_rebuilding;
	boolean_t vdev_rebuild_exit_wanted;
	boolean_t vdev_rebuild_cancel_wanted;
	boolean_t vdev_rebuild_reset_wanted;
	kmutex_t vdev_rebuild_lock;
	kcondvar_t vdev_rebuild_cv;
	kthread_t *vdev_rebuild_thread;
	vdev_rebuild_t vdev_rebuild_config;

	/* For limiting outstanding I/Os (initialize, TRIM) */
	kmutex_t vdev_initialize_io_lock;
	kcondvar_t vdev_initialize_io_cv;
	uint64_t vdev_initialize_inflight;
	kmutex_t vdev_trim_io_lock;
	kcondvar_t vdev_trim_io_cv;
	uint64_t vdev_trim_inflight[3];

	/*
	* Values stored in the config for an indirect or removing vdev.
	*/
	vdev_indirect_config_t vdev_indirect_config;

	/*
	* The vdev_indirect_rwlock protects the vdev_indirect_mapping
	* pointer from changing on indirect vdevs (when it is condensed).
	* Note that removing (not yet indirect) vdevs have different
	* access patterns (the mapping is not accessed from open context,
	* e.g. from zio_read) and locking strategy (e.g. svr_lock).
	*/
	krwlock_t vdev_indirect_rwlock;
	vdev_indirect_mapping_t *vdev_indirect_mapping;
	vdev_indirect_births_t *vdev_indirect_births;

	/*
	* In memory data structures used to manage the obsolete sm, for
	* indirect or removing vdevs.
	*
	* The vdev_obsolete_segments is the in-core record of the segments
	* that are no longer referenced anywhere in the pool (due to
	* being freed or remapped and not referenced by any snapshots).
	* During a sync, segments are added to vdev_obsolete_segments
	* via vdev_indirect_mark_obsolete(); at the end of each sync
	* pass, this is appended to vdev_obsolete_sm via
	* vdev_indirect_sync_obsolete(). The vdev_obsolete_lock
	* protects against concurrent modifications of vdev_obsolete_segments
	* from multiple zio threads.
	*/
	kmutex_t vdev_obsolete_lock;
	range_tree_t *vdev_obsolete_segments;
	space_map_t *vdev_obsolete_sm;

	/*
	* Protects the vdev_scan_io_queue field itself as well as the
	* structure's contents (when present).
	*/
	kmutex_t vdev_scan_io_queue_lock;
	struct dsl_scan_io_queue *vdev_scan_io_queue;

	/*
	* Leaf vdev state.
	*/
	range_tree_t vdev_dtl[DTL_TYPES]; / dirty time logs */
	space_map_t vdev_dtl_sm; / dirty time log space map */
	txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
	uint64_t vdev_dtl_object; /* DTL object */
	uint64_t vdev_psize; /* physical device capacity */
	uint64_t vdev_wholedisk; /* true if this is a whole disk */
	uint64_t vdev_offline; /* persistent offline state */
	uint64_t vdev_faulted; /* persistent faulted state */
	uint64_t vdev_degraded; /* persistent degraded state */
	uint64_t vdev_removed; /* persistent removed state */
	uint64_t vdev_resilver_txg; /* persistent resilvering state */
	uint64_t vdev_rebuild_txg; /* persistent rebuilding state */
	char vdev_path; / vdev path (if any) */
	char vdev_devid; / vdev devid (if any) */
	char vdev_physpath; / vdev device path (if any) */
	char vdev_enc_sysfs_path; / enclosure sysfs path */
	char vdev_fru; / physical FRU location */
	uint64_t vdev_not_present; /* not present during import */
	uint64_t vdev_unspare; /* unspare when resilvering done */
	boolean_t vdev_nowritecache; /* true if flushwritecache failed */
	boolean_t vdev_has_trim; /* TRIM is supported */
	boolean_t vdev_has_securetrim; /* secure TRIM is supported */
	boolean_t vdev_checkremove; /* temporary online test */
	boolean_t vdev_forcefault; /* force online fault */
	boolean_t vdev_splitting; /* split or repair in progress */
	boolean_t vdev_delayed_close; /* delayed device close? */
	boolean_t vdev_tmpoffline; /* device taken offline temporarily? */
	boolean_t vdev_detached; /* device detached? */
	boolean_t vdev_cant_read; /* vdev is failing all reads */
	boolean_t vdev_cant_write; /* vdev is failing all writes */
	boolean_t vdev_isspare; /* was a hot spare */
	boolean_t vdev_isl2cache; /* was a l2cache device */
	boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */
	boolean_t vdev_resilver_deferred; /* resilver deferred */
	+ boolean_t vdev_kobj_flag; /* kobj event record */
	vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
	vdev_cache_t vdev_cache; /* physical block cache */
	spa_aux_vdev_t vdev_aux; / for l2cache and spares vdevs */
	zio_t vdev_probe_zio; / root of current probe */
	vdev_aux_t vdev_label_aux; /* on-disk aux state */
	uint64_t vdev_leaf_zap;
	hrtime_t vdev_mmp_pending; /* 0 if write finished */
	uint64_t vdev_mmp_kstat_id; /* to find kstat entry */
	uint64_t vdev_expansion_time; /* vdev's last expansion time */
	list_node_t vdev_leaf_node; /* leaf vdev list */

	/*
	* For DTrace to work in userland (libzpool) context, these fields must
	* remain at the end of the structure. DTrace will use the kernel's
	* CTF definition for 'struct vdev', and since the size of a kmutex_t is
	* larger in userland, the offsets for the rest of the fields would be
	* incorrect.
	*/
	kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
	kmutex_t vdev_stat_lock; /* vdev_stat */
	kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */

	/*
	* We rate limit ZIO delay, deadman, and checksum events, since they
	* can flood ZED with tons of events when a drive is acting up.
	*/
	zfs_ratelimit_t vdev_delay_rl;
	zfs_ratelimit_t vdev_deadman_rl;
	zfs_ratelimit_t vdev_checksum_rl;
	};

	#define VDEV_PAD_SIZE (8 << 10)
	/* 2 padding areas (vl_pad1 and vl_be) to skip */
	#define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
	#define VDEV_PHYS_SIZE (112 << 10)
	#define VDEV_UBERBLOCK_RING (128 << 10)

	/*
	* MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
	* ring when MMP is enabled.
	*/
	#define MMP_BLOCKS_PER_LABEL 1

	/* The largest uberblock we support is 8k. */
	#define MAX_UBERBLOCK_SHIFT (13)
	#define VDEV_UBERBLOCK_SHIFT(vd) \
	MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
	MAX_UBERBLOCK_SHIFT)
	#define VDEV_UBERBLOCK_COUNT(vd) \
	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
	#define VDEV_UBERBLOCK_OFFSET(vd, n) \
	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
	#define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))

	typedef struct vdev_phys {
	char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
	zio_eck_t vp_zbt;
	} vdev_phys_t;

	typedef enum vbe_vers {
	/*
	* The bootenv file is stored as ascii text in the envblock.
	* It is used by the GRUB bootloader used on Linux to store the
	* contents of the grubenv file. The file is stored as raw ASCII,
	* and is protected by an embedded checksum. By default, GRUB will
	* check if the boot filesystem supports storing the environment data
	* in a special location, and if so, will invoke filesystem specific
	* logic to retrieve it. This can be overridden by a variable, should
	* the user so desire.
	*/
	VB_RAW = 0,

	/*
	* The bootenv file is converted to an nvlist and then packed into the
	* envblock.
	*/
	VB_NVLIST = 1
	} vbe_vers_t;

	typedef struct vdev_boot_envblock {
	uint64_t vbe_version;
	char vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
	sizeof (zio_eck_t)];
	zio_eck_t vbe_zbt;
	} vdev_boot_envblock_t;
	_Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE,
	"vdev_boot_envblock_t wrong size");

	typedef struct vdev_label {
	char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
	vdev_boot_envblock_t vl_be; /* 8K */
	vdev_phys_t vl_vdev_phys; /* 112K */
	char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
	} vdev_label_t; /* 256K total */

	/*
	* vdev_dirty() flags
	*/
	#define VDD_METASLAB 0x01
	#define VDD_DTL 0x02

	/* Offset of embedded boot loader region on each label */
	#define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
	/*
	* Size of embedded boot loader region on each label.
	* The total size of the first two labels plus the boot area is 4MB.
	*/
	#define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */

	/*
	* Size of label regions at the start and end of each leaf device.
	*/
	#define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
	#define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
	#define VDEV_LABELS 4
	#define VDEV_BEST_LABEL VDEV_LABELS
	#define VDEV_OFFSET_IS_LABEL(vd, off) \
	(((off) < VDEV_LABEL_START_SIZE) \|\| \
	((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE)))

	#define VDEV_ALLOC_LOAD 0
	#define VDEV_ALLOC_ADD 1
	#define VDEV_ALLOC_SPARE 2
	#define VDEV_ALLOC_L2CACHE 3
	#define VDEV_ALLOC_ROOTPOOL 4
	#define VDEV_ALLOC_SPLIT 5
	#define VDEV_ALLOC_ATTACH 6

	/*
	* Allocate or free a vdev
	*/
	extern vdev_t vdev_alloc_common(spa_t spa, uint_t id, uint64_t guid,
	vdev_ops_t *ops);
	extern int vdev_alloc(spa_t spa, vdev_t vdp, nvlist_t config,
	vdev_t *parent, uint_t id, int alloctype);
	extern void vdev_free(vdev_t *vd);

	/*
	* Add or remove children and parents
	*/
	extern void vdev_add_child(vdev_t pvd, vdev_t cvd);
	extern void vdev_remove_child(vdev_t pvd, vdev_t cvd);
	extern void vdev_compact_children(vdev_t *pvd);
	extern vdev_t vdev_add_parent(vdev_t cvd, vdev_ops_t *ops);
	extern void vdev_remove_parent(vdev_t *cvd);

	/*
	* vdev sync load and sync
	*/
	extern boolean_t vdev_log_state_valid(vdev_t *vd);
	extern int vdev_load(vdev_t *vd);
	extern int vdev_dtl_load(vdev_t *vd);
	extern void vdev_sync(vdev_t *vd, uint64_t txg);
	extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
	extern void vdev_dirty(vdev_t vd, int flags, void arg, uint64_t txg);
	extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);

	/*
	* Available vdev types.
	*/
	extern vdev_ops_t vdev_root_ops;
	extern vdev_ops_t vdev_mirror_ops;
	extern vdev_ops_t vdev_replacing_ops;
	extern vdev_ops_t vdev_raidz_ops;
	extern vdev_ops_t vdev_draid_ops;
	extern vdev_ops_t vdev_draid_spare_ops;
	extern vdev_ops_t vdev_disk_ops;
	extern vdev_ops_t vdev_file_ops;
	extern vdev_ops_t vdev_missing_ops;
	extern vdev_ops_t vdev_hole_ops;
	extern vdev_ops_t vdev_spare_ops;
	extern vdev_ops_t vdev_indirect_ops;

	/*
	* Common size functions
	*/
	extern void vdev_default_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs);
	extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
	extern uint64_t vdev_default_min_asize(vdev_t *vd);
	extern uint64_t vdev_get_min_asize(vdev_t *vd);
	extern void vdev_set_min_asize(vdev_t *vd);
	extern uint64_t vdev_get_min_alloc(vdev_t *vd);
	extern uint64_t vdev_get_nparity(vdev_t *vd);
	extern uint64_t vdev_get_ndisks(vdev_t *vd);

	/*
	* Global variables
	*/
	extern int zfs_vdev_standard_sm_blksz;

	/*
	* Functions from vdev_indirect.c
	*/
	extern void vdev_indirect_sync_obsolete(vdev_t vd, dmu_tx_t tx);
	extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
	extern void spa_condense_indirect_start_sync(vdev_t vd, dmu_tx_t tx);
	extern int vdev_obsolete_sm_object(vdev_t vd, uint64_t sm_obj);
	extern int vdev_obsolete_counts_are_precise(vdev_t vd, boolean_t are_precise);

	/*
	* Other miscellaneous functions
	*/
	int vdev_checkpoint_sm_object(vdev_t vd, uint64_t sm_obj);
	void vdev_metaslab_group_create(vdev_t *vd);
	uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b);

	/*
	* Vdev ashift optimization tunables
	*/
	extern uint64_t zfs_vdev_min_auto_ashift;
	extern uint64_t zfs_vdev_max_auto_ashift;
	int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS);
	int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_VDEV_IMPL_H */
	diff --git a/sys/contrib/openzfs/include/sys/vdev_removal.h b/sys/contrib/openzfs/include/sys/vdev_removal.h
	index e3bab0658d62..70b743f4ec6b 100644
	--- a/sys/contrib/openzfs/include/sys/vdev_removal.h
	+++ b/sys/contrib/openzfs/include/sys/vdev_removal.h
	@@ -1,96 +1,96 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2014, 2019 by Delphix. All rights reserved.
	*/

	#ifndef _SYS_VDEV_REMOVAL_H
	#define _SYS_VDEV_REMOVAL_H

	#include <sys/spa.h>
	#include <sys/bpobj.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/vdev_indirect_births.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	typedef struct spa_vdev_removal {
	uint64_t svr_vdev_id;
	uint64_t svr_max_offset_to_sync[TXG_SIZE];
	/* Thread performing a vdev removal. */
	kthread_t *svr_thread;
	/* Segments left to copy from the current metaslab. */
	range_tree_t *svr_allocd_segs;
	kmutex_t svr_lock;
	kcondvar_t svr_cv;
	boolean_t svr_thread_exit;

	/*
	* New mappings to write out each txg.
	*/
	list_t svr_new_segments[TXG_SIZE];

	/*
	* Ranges that were freed while a mapping was in flight. This is
	* a subset of the ranges covered by vdev_im_new_segments.
	*/
	range_tree_t *svr_frees[TXG_SIZE];

	/*
	* Number of bytes which we have finished our work for
	* in each txg. This could be data copied (which will be part of
	* the mappings in vdev_im_new_segments), or data freed before
	* we got around to copying it.
	*/
	uint64_t svr_bytes_done[TXG_SIZE];

	/* List of leaf zap objects to be unlinked */
	nvlist_t *svr_zaplist;
	} spa_vdev_removal_t;

	typedef struct spa_condensing_indirect {
	/*
	* New mappings to write out each txg.
	*/
	list_t sci_new_mapping_entries[TXG_SIZE];

	vdev_indirect_mapping_t *sci_new_mapping;
	} spa_condensing_indirect_t;

	extern int spa_remove_init(spa_t *);
	extern void spa_restart_removal(spa_t *);
	extern int spa_condense_init(spa_t *);
	extern void spa_condense_fini(spa_t *);
	extern void spa_start_indirect_condensing_thread(spa_t *);
	extern void spa_vdev_condense_suspend(spa_t *);
	extern int spa_vdev_remove(spa_t *, uint64_t, boolean_t);
	extern void free_from_removing_vdev(vdev_t *, uint64_t, uint64_t);
	extern int spa_removal_get_stats(spa_t , pool_removal_stat_t );
	extern void svr_sync(spa_t , dmu_tx_t );
	extern void spa_vdev_remove_suspend(spa_t *);
	extern int spa_vdev_remove_cancel(spa_t *);
	extern void spa_vdev_removal_destroy(spa_vdev_removal_t *);
	extern uint64_t spa_remove_max_segment(spa_t *);

	-extern int vdev_removal_max_span;
	+extern uint_t vdev_removal_max_span;

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_VDEV_REMOVAL_H */
	diff --git a/sys/contrib/openzfs/include/sys/zfs_context.h b/sys/contrib/openzfs/include/sys/zfs_context.h
	index 83ed97fbec7f..d29d7118ff00 100644
	--- a/sys/contrib/openzfs/include/sys/zfs_context.h
	+++ b/sys/contrib/openzfs/include/sys/zfs_context.h
	@@ -1,778 +1,783 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	*/

	#ifndef _SYS_ZFS_CONTEXT_H
	#define _SYS_ZFS_CONTEXT_H

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* This code compiles in three different contexts. When __KERNEL__ is defined,
	* the code uses "unix-like" kernel interfaces. When _STANDALONE is defined, the
	* code is running in a reduced capacity environment of the boot loader which is
	* generally a subset of both POSIX and kernel interfaces (with a few unique
	* interfaces too). When neither are defined, it's in a userland POSIX or
	* similar environment.
	*/
	#if defined(__KERNEL__) \|\| defined(_STANDALONE)
	#include <sys/types.h>
	#include <sys/atomic.h>
	#include <sys/sysmacros.h>
	#include <sys/vmsystm.h>
	#include <sys/condvar.h>
	#include <sys/cmn_err.h>
	#include <sys/kmem.h>
	#include <sys/kmem_cache.h>
	#include <sys/vmem.h>
	+#include <sys/misc.h>
	#include <sys/taskq.h>
	#include <sys/param.h>
	#include <sys/disp.h>
	#include <sys/debug.h>
	#include <sys/random.h>
	#include <sys/string.h>
	#include <sys/byteorder.h>
	#include <sys/list.h>
	#include <sys/time.h>
	#include <sys/zone.h>
	#include <sys/kstat.h>
	#include <sys/zfs_debug.h>
	#include <sys/sysevent.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/zfs_delay.h>
	#include <sys/sunddi.h>
	#include <sys/ctype.h>
	#include <sys/disp.h>
	#include <sys/trace.h>
	#include <sys/procfs_list.h>
	#include <sys/mod.h>
	#include <sys/uio_impl.h>
	#include <sys/zfs_context_os.h>
	#else /* _KERNEL \|\| _STANDALONE */

	#define _SYS_MUTEX_H
	#define _SYS_RWLOCK_H
	#define _SYS_CONDVAR_H
	#define _SYS_VNODE_H
	#define _SYS_VFS_H
	#define _SYS_SUNDDI_H
	#define _SYS_CALLB_H

	#include <stdio.h>
	#include <stdlib.h>
	#include <stddef.h>
	#include <stdarg.h>
	#include <fcntl.h>
	#include <unistd.h>
	#include <errno.h>
	#include <string.h>
	#include <pthread.h>
	#include <setjmp.h>
	#include <assert.h>
	#include <umem.h>
	#include <limits.h>
	#include <atomic.h>
	#include <dirent.h>
	#include <time.h>
	#include <ctype.h>
	#include <signal.h>
	#include <sys/mman.h>
	#include <sys/types.h>
	#include <sys/cred.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <sys/byteorder.h>
	#include <sys/list.h>
	#include <sys/mod.h>
	#include <sys/uio.h>
	#include <sys/zfs_debug.h>
	#include <sys/kstat.h>
	#include <sys/u8_textprep.h>
	#include <sys/sysevent.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/sunddi.h>
	#include <sys/debug.h>
	#include <sys/utsname.h>
	#include <sys/trace_zfs.h>

	#include <sys/zfs_context_os.h>

	/*
	* Stack
	*/

	#define noinline __attribute__((noinline))
	#define likely(x) __builtin_expect((x), 1)
	#define unlikely(x) __builtin_expect((x), 0)

	/*
	* Debugging
	*/

	/*
	* Note that we are not using the debugging levels.
	*/

	#define CE_CONT 0 /* continuation */
	#define CE_NOTE 1 /* notice */
	#define CE_WARN 2 /* warning */
	#define CE_PANIC 3 /* panic */
	#define CE_IGNORE 4 /* print nothing */

	/*
	* ZFS debugging
	*/

	extern void dprintf_setup(int argc, char *argv);

	-extern void cmn_err(int, const char *, ...);
	-extern void vcmn_err(int, const char *, va_list);
	-extern __attribute__((noreturn)) void panic(const char *, ...);
	-extern __attribute__((noreturn)) void vpanic(const char *, va_list);
	+extern void cmn_err(int, const char *, ...)
	+ __attribute__((format(printf, 2, 3)));
	+extern void vcmn_err(int, const char *, va_list)
	+ __attribute__((format(printf, 2, 0)));
	+extern void panic(const char *, ...)
	+ __attribute__((format(printf, 1, 2), noreturn));
	+extern void vpanic(const char *, va_list)
	+ __attribute__((format(printf, 1, 0), noreturn));

	#define fm_panic panic

	/*
	* DTrace SDT probes have different signatures in userland than they do in
	* the kernel. If they're being used in kernel code, re-define them out of
	* existence for their counterparts in libzpool.
	*
	* Here's an example of how to use the set-error probes in userland:
	* zfs$target:::set-error /arg0 == EBUSY/ {stack();}
	*
	* Here's an example of how to use DTRACE_PROBE probes in userland:
	* If there is a probe declared as follows:
	* DTRACE_PROBE2(zfs__probe_name, uint64_t, blkid, dnode_t *, dn);
	* Then you can use it as follows:
	* zfs$target:::probe2 /copyinstr(arg0) == "zfs__probe_name"/
	* {printf("%u %p\n", arg1, arg2);}
	*/

	#ifdef DTRACE_PROBE
	#undef DTRACE_PROBE
	#endif /* DTRACE_PROBE */
	#define DTRACE_PROBE(a)

	#ifdef DTRACE_PROBE1
	#undef DTRACE_PROBE1
	#endif /* DTRACE_PROBE1 */
	#define DTRACE_PROBE1(a, b, c)

	#ifdef DTRACE_PROBE2
	#undef DTRACE_PROBE2
	#endif /* DTRACE_PROBE2 */
	#define DTRACE_PROBE2(a, b, c, d, e)

	#ifdef DTRACE_PROBE3
	#undef DTRACE_PROBE3
	#endif /* DTRACE_PROBE3 */
	#define DTRACE_PROBE3(a, b, c, d, e, f, g)

	#ifdef DTRACE_PROBE4
	#undef DTRACE_PROBE4
	#endif /* DTRACE_PROBE4 */
	#define DTRACE_PROBE4(a, b, c, d, e, f, g, h, i)

	/*
	* Tunables.
	*/
	typedef struct zfs_kernel_param {
	const char name; / unused stub */
	} zfs_kernel_param_t;

	#define ZFS_MODULE_PARAM(scope_prefix, name_prefix, name, type, perm, desc)
	#define ZFS_MODULE_PARAM_ARGS void
	#define ZFS_MODULE_PARAM_CALL(scope_prefix, name_prefix, name, setfunc, \
	getfunc, perm, desc)

	/*
	* Threads.
	*/
	typedef pthread_t kthread_t;

	#define TS_RUN 0x00000002
	#define TS_JOINABLE 0x00000004

	#define curthread ((void *)(uintptr_t)pthread_self())
	#define getcomm() "unknown"

	#define thread_create_named(name, stk, stksize, func, arg, len, \
	pp, state, pri) \
	zk_thread_create(func, arg, stksize, state)
	#define thread_create(stk, stksize, func, arg, len, pp, state, pri) \
	zk_thread_create(func, arg, stksize, state)
	#define thread_exit() pthread_exit(NULL)
	#define thread_join(t) pthread_join((pthread_t)(t), NULL)

	#define newproc(f, a, cid, pri, ctp, pid) (ENOSYS)

	/* in libzpool, p0 exists only to have its address taken */
	typedef struct proc {
	uintptr_t this_is_never_used_dont_dereference_it;
	} proc_t;

	extern struct proc p0;
	#define curproc (&p0)

	#define PS_NONE -1

	extern kthread_t zk_thread_create(void (func)(void ), void arg,
	size_t stksize, int state);

	#define issig(why) (FALSE)
	#define ISSIG(thr, why) (FALSE)

	#define KPREEMPT_SYNC (-1)

	#define kpreempt(x) sched_yield()
	#define kpreempt_disable() ((void)0)
	#define kpreempt_enable() ((void)0)

	/*
	* Mutexes
	*/
	typedef struct kmutex {
	pthread_mutex_t m_lock;
	pthread_t m_owner;
	} kmutex_t;

	#define MUTEX_DEFAULT 0
	#define MUTEX_NOLOCKDEP MUTEX_DEFAULT
	#define MUTEX_HELD(mp) pthread_equal((mp)->m_owner, pthread_self())
	#define MUTEX_NOT_HELD(mp) !MUTEX_HELD(mp)

	extern void mutex_init(kmutex_t mp, char name, int type, void *cookie);
	extern void mutex_destroy(kmutex_t *mp);
	extern void mutex_enter(kmutex_t *mp);
	extern void mutex_exit(kmutex_t *mp);
	extern int mutex_tryenter(kmutex_t *mp);

	#define NESTED_SINGLE 1
	#define mutex_enter_nested(mp, class) mutex_enter(mp)
	/*
	* RW locks
	*/
	typedef struct krwlock {
	pthread_rwlock_t rw_lock;
	pthread_t rw_owner;
	uint_t rw_readers;
	} krwlock_t;

	typedef int krw_t;

	#define RW_READER 0
	#define RW_WRITER 1
	#define RW_DEFAULT RW_READER
	#define RW_NOLOCKDEP RW_READER

	#define RW_READ_HELD(rw) ((rw)->rw_readers > 0)
	#define RW_WRITE_HELD(rw) pthread_equal((rw)->rw_owner, pthread_self())
	#define RW_LOCK_HELD(rw) (RW_READ_HELD(rw) \|\| RW_WRITE_HELD(rw))

	extern void rw_init(krwlock_t rwlp, char name, int type, void *arg);
	extern void rw_destroy(krwlock_t *rwlp);
	extern void rw_enter(krwlock_t *rwlp, krw_t rw);
	extern int rw_tryenter(krwlock_t *rwlp, krw_t rw);
	extern int rw_tryupgrade(krwlock_t *rwlp);
	extern void rw_exit(krwlock_t *rwlp);
	#define rw_downgrade(rwlp) do { } while (0)

	/*
	* Credentials
	*/
	extern uid_t crgetuid(cred_t *cr);
	extern uid_t crgetruid(cred_t *cr);
	extern gid_t crgetgid(cred_t *cr);
	extern int crgetngroups(cred_t *cr);
	extern gid_t crgetgroups(cred_t cr);

	/*
	* Condition variables
	*/
	typedef pthread_cond_t kcondvar_t;

	#define CV_DEFAULT 0
	#define CALLOUT_FLAG_ABSOLUTE 0x2

	extern void cv_init(kcondvar_t cv, char name, int type, void *arg);
	extern void cv_destroy(kcondvar_t *cv);
	extern void cv_wait(kcondvar_t cv, kmutex_t mp);
	extern int cv_wait_sig(kcondvar_t cv, kmutex_t mp);
	extern int cv_timedwait(kcondvar_t cv, kmutex_t mp, clock_t abstime);
	extern int cv_timedwait_hires(kcondvar_t cvp, kmutex_t mp, hrtime_t tim,
	hrtime_t res, int flag);
	extern void cv_signal(kcondvar_t *cv);
	extern void cv_broadcast(kcondvar_t *cv);

	#define cv_timedwait_io(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_timedwait_idle(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_timedwait_sig(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_wait_io(cv, mp) cv_wait(cv, mp)
	#define cv_wait_idle(cv, mp) cv_wait(cv, mp)
	#define cv_wait_io_sig(cv, mp) cv_wait_sig(cv, mp)
	#define cv_timedwait_sig_hires(cv, mp, t, r, f) \
	cv_timedwait_hires(cv, mp, t, r, f)
	#define cv_timedwait_idle_hires(cv, mp, t, r, f) \
	cv_timedwait_hires(cv, mp, t, r, f)

	/*
	* Thread-specific data
	*/
	#define tsd_get(k) pthread_getspecific(k)
	#define tsd_set(k, v) pthread_setspecific(k, v)
	#define tsd_create(kp, d) pthread_key_create((pthread_key_t *)kp, d)
	#define tsd_destroy(kp) /* nothing */
	#ifdef __FreeBSD__
	typedef off_t loff_t;
	#endif

	/*
	* kstat creation, installation and deletion
	*/
	extern kstat_t kstat_create(const char , int,
	const char , const char , uchar_t, ulong_t, uchar_t);
	extern void kstat_install(kstat_t *);
	extern void kstat_delete(kstat_t *);
	extern void kstat_set_raw_ops(kstat_t *ksp,
	int (headers)(char buf, size_t size),
	int (data)(char buf, size_t size, void *data),
	void (addr)(kstat_t *ksp, loff_t index));

	/*
	* procfs list manipulation
	*/

	typedef struct procfs_list {
	void *pl_private;
	kmutex_t pl_lock;
	list_t pl_list;
	uint64_t pl_next_id;
	size_t pl_node_offset;
	} procfs_list_t;

	#ifndef __cplusplus
	struct seq_file { };
	void seq_printf(struct seq_file m, const char fmt, ...);

	typedef struct procfs_list_node {
	list_node_t pln_link;
	uint64_t pln_id;
	} procfs_list_node_t;

	void procfs_list_install(const char *module,
	const char *submodule,
	const char *name,
	mode_t mode,
	procfs_list_t *procfs_list,
	int (show)(struct seq_file f, void *p),
	int (show_header)(struct seq_file f),
	int (clear)(procfs_list_t procfs_list),
	size_t procfs_list_node_off);
	void procfs_list_uninstall(procfs_list_t *procfs_list);
	void procfs_list_destroy(procfs_list_t *procfs_list);
	void procfs_list_add(procfs_list_t procfs_list, void p);
	#endif

	/*
	* Kernel memory
	*/
	#define KM_SLEEP UMEM_NOFAIL
	#define KM_PUSHPAGE KM_SLEEP
	#define KM_NOSLEEP UMEM_DEFAULT
	#define KM_NORMALPRI 0 /* not needed with UMEM_DEFAULT */
	#define KMC_NODEBUG UMC_NODEBUG
	#define KMC_KVMEM 0x0
	#define kmem_alloc(_s, _f) umem_alloc(_s, _f)
	#define kmem_zalloc(_s, _f) umem_zalloc(_s, _f)
	#define kmem_free(_b, _s) umem_free(_b, _s)
	#define vmem_alloc(_s, _f) kmem_alloc(_s, _f)
	#define vmem_zalloc(_s, _f) kmem_zalloc(_s, _f)
	#define vmem_free(_b, _s) kmem_free(_b, _s)
	#define kmem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i) \
	umem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i)
	#define kmem_cache_destroy(_c) umem_cache_destroy(_c)
	#define kmem_cache_alloc(_c, _f) umem_cache_alloc(_c, _f)
	#define kmem_cache_free(_c, _b) umem_cache_free(_c, _b)
	#define kmem_debugging() 0
	#define kmem_cache_reap_now(_c) umem_cache_reap_now(_c);
	#define kmem_cache_set_move(_c, _cb) /* nothing */
	#define POINTER_INVALIDATE(_pp) /* nothing */
	#define POINTER_IS_VALID(_p) 0

	typedef umem_cache_t kmem_cache_t;

	typedef enum kmem_cbrc {
	KMEM_CBRC_YES,
	KMEM_CBRC_NO,
	KMEM_CBRC_LATER,
	KMEM_CBRC_DONT_NEED,
	KMEM_CBRC_DONT_KNOW
	} kmem_cbrc_t;

	/*
	* Task queues
	*/

	#define TASKQ_NAMELEN 31

	typedef uintptr_t taskqid_t;
	typedef void (task_func_t)(void *);

	typedef struct taskq_ent {
	struct taskq_ent *tqent_next;
	struct taskq_ent *tqent_prev;
	task_func_t *tqent_func;
	void *tqent_arg;
	uintptr_t tqent_flags;
	} taskq_ent_t;

	typedef struct taskq {
	char tq_name[TASKQ_NAMELEN + 1];
	kmutex_t tq_lock;
	krwlock_t tq_threadlock;
	kcondvar_t tq_dispatch_cv;
	kcondvar_t tq_wait_cv;
	kthread_t **tq_threadlist;
	int tq_flags;
	int tq_active;
	int tq_nthreads;
	int tq_nalloc;
	int tq_minalloc;
	int tq_maxalloc;
	kcondvar_t tq_maxalloc_cv;
	int tq_maxalloc_wait;
	taskq_ent_t *tq_freelist;
	taskq_ent_t tq_task;
	} taskq_t;

	#define TQENT_FLAG_PREALLOC 0x1 /* taskq_dispatch_ent used */

	#define TASKQ_PREPOPULATE 0x0001
	#define TASKQ_CPR_SAFE 0x0002 /* Use CPR safe protocol */
	#define TASKQ_DYNAMIC 0x0004 /* Use dynamic thread scheduling */
	#define TASKQ_THREADS_CPU_PCT 0x0008 /* Scale # threads by # cpus */
	#define TASKQ_DC_BATCH 0x0010 /* Mark threads as batch */

	#define TQ_SLEEP KM_SLEEP /* Can block for memory */
	#define TQ_NOSLEEP KM_NOSLEEP /* cannot block for memory; may fail */
	#define TQ_NOQUEUE 0x02 /* Do not enqueue if can't dispatch */
	#define TQ_FRONT 0x08 /* Queue in front */

	#define TASKQID_INVALID ((taskqid_t)0)

	extern taskq_t *system_taskq;
	extern taskq_t *system_delay_taskq;

	extern taskq_t taskq_create(const char , int, pri_t, int, int, uint_t);
	#define taskq_create_proc(a, b, c, d, e, p, f) \
	(taskq_create(a, b, c, d, e, f))
	#define taskq_create_sysdc(a, b, d, e, p, dc, f) \
	((void) sizeof (dc), taskq_create(a, b, maxclsyspri, d, e, f))
	extern taskqid_t taskq_dispatch(taskq_t , task_func_t, void , uint_t);
	extern taskqid_t taskq_dispatch_delay(taskq_t , task_func_t, void , uint_t,
	clock_t);
	extern void taskq_dispatch_ent(taskq_t , task_func_t, void , uint_t,
	taskq_ent_t *);
	extern int taskq_empty_ent(taskq_ent_t *);
	extern void taskq_init_ent(taskq_ent_t *);
	extern void taskq_destroy(taskq_t *);
	extern void taskq_wait(taskq_t *);
	extern void taskq_wait_id(taskq_t *, taskqid_t);
	extern void taskq_wait_outstanding(taskq_t *, taskqid_t);
	extern int taskq_member(taskq_t , kthread_t );
	extern taskq_t *taskq_of_curthread(void);
	extern int taskq_cancel_id(taskq_t *, taskqid_t);
	extern void system_taskq_init(void);
	extern void system_taskq_fini(void);

	#define XVA_MAPSIZE 3
	#define XVA_MAGIC 0x78766174

	extern char *vn_dumpdir;
	#define AV_SCANSTAMP_SZ 32 /* length of anti-virus scanstamp */

	typedef struct xoptattr {
	inode_timespec_t xoa_createtime; /* Create time of file */
	uint8_t xoa_archive;
	uint8_t xoa_system;
	uint8_t xoa_readonly;
	uint8_t xoa_hidden;
	uint8_t xoa_nounlink;
	uint8_t xoa_immutable;
	uint8_t xoa_appendonly;
	uint8_t xoa_nodump;
	uint8_t xoa_settable;
	uint8_t xoa_opaque;
	uint8_t xoa_av_quarantined;
	uint8_t xoa_av_modified;
	uint8_t xoa_av_scanstamp[AV_SCANSTAMP_SZ];
	uint8_t xoa_reparse;
	uint8_t xoa_offline;
	uint8_t xoa_sparse;
	} xoptattr_t;

	typedef struct vattr {
	uint_t va_mask; /* bit-mask of attributes */
	u_offset_t va_size; /* file size in bytes */
	} vattr_t;


	typedef struct xvattr {
	vattr_t xva_vattr; /* Embedded vattr structure */
	uint32_t xva_magic; /* Magic Number */
	uint32_t xva_mapsize; /* Size of attr bitmap (32-bit words) */
	uint32_t xva_rtnattrmapp; / Ptr to xva_rtnattrmap[] */
	uint32_t xva_reqattrmap[XVA_MAPSIZE]; /* Requested attrs */
	uint32_t xva_rtnattrmap[XVA_MAPSIZE]; /* Returned attrs */
	xoptattr_t xva_xoptattrs; /* Optional attributes */
	} xvattr_t;

	typedef struct vsecattr {
	uint_t vsa_mask; /* See below */
	int vsa_aclcnt; /* ACL entry count */
	void vsa_aclentp; / pointer to ACL entries */
	int vsa_dfaclcnt; /* default ACL entry count */
	void vsa_dfaclentp; / pointer to default ACL entries */
	size_t vsa_aclentsz; /* ACE size in bytes of vsa_aclentp */
	} vsecattr_t;

	#define AT_MODE 0x00002
	#define AT_UID 0x00004
	#define AT_GID 0x00008
	#define AT_FSID 0x00010
	#define AT_NODEID 0x00020
	#define AT_NLINK 0x00040
	#define AT_SIZE 0x00080
	#define AT_ATIME 0x00100
	#define AT_MTIME 0x00200
	#define AT_CTIME 0x00400
	#define AT_RDEV 0x00800
	#define AT_BLKSIZE 0x01000
	#define AT_NBLOCKS 0x02000
	#define AT_SEQ 0x08000
	#define AT_XVATTR 0x10000

	#define CRCREAT 0

	#define F_FREESP 11
	#define FIGNORECASE 0x80000 /* request case-insensitive lookups */

	/*
	* Random stuff
	*/
	#define ddi_get_lbolt() (gethrtime() >> 23)
	#define ddi_get_lbolt64() (gethrtime() >> 23)
	#define hz 119 /* frequency when using gethrtime() >> 23 for lbolt */

	#define ddi_time_before(a, b) (a < b)
	#define ddi_time_after(a, b) ddi_time_before(b, a)
	#define ddi_time_before_eq(a, b) (!ddi_time_after(a, b))
	#define ddi_time_after_eq(a, b) ddi_time_before_eq(b, a)

	#define ddi_time_before64(a, b) (a < b)
	#define ddi_time_after64(a, b) ddi_time_before64(b, a)
	#define ddi_time_before_eq64(a, b) (!ddi_time_after64(a, b))
	#define ddi_time_after_eq64(a, b) ddi_time_before_eq64(b, a)

	extern void delay(clock_t ticks);

	#define SEC_TO_TICK(sec) ((sec) * hz)
	#define MSEC_TO_TICK(msec) (howmany((hrtime_t)(msec) * hz, MILLISEC))
	#define USEC_TO_TICK(usec) (howmany((hrtime_t)(usec) * hz, MICROSEC))
	#define NSEC_TO_TICK(nsec) (howmany((hrtime_t)(nsec) * hz, NANOSEC))

	#define max_ncpus 64
	#define boot_ncpus (sysconf(_SC_NPROCESSORS_ONLN))

	/*
	* Process priorities as defined by setpriority(2) and getpriority(2).
	*/
	#define minclsyspri 19
	#define maxclsyspri -20
	#define defclsyspri 0

	#define CPU_SEQID ((uintptr_t)pthread_self() & (max_ncpus - 1))
	#define CPU_SEQID_UNSTABLE CPU_SEQID

	#define kcred NULL
	#define CRED() NULL

	#define ptob(x) ((x) * PAGESIZE)

	#define NN_DIVISOR_1000 (1U << 0)
	#define NN_NUMBUF_SZ (6)

	extern uint64_t physmem;
	extern const char *random_path;
	extern const char *urandom_path;

	extern int highbit64(uint64_t i);
	extern int lowbit64(uint64_t i);
	extern int random_get_bytes(uint8_t *ptr, size_t len);
	extern int random_get_pseudo_bytes(uint8_t *ptr, size_t len);

	static __inline__ uint32_t
	random_in_range(uint32_t range)
	{
	uint32_t r;

	ASSERT(range != 0);

	if (range == 1)
	return (0);

	(void) random_get_pseudo_bytes((uint8_t *)&r, sizeof (r));

	return (r % range);
	}

	extern void kernel_init(int mode);
	extern void kernel_fini(void);
	extern void random_init(void);
	extern void random_fini(void);

	struct spa;
	extern void show_pool_stats(struct spa *);
	extern int set_global_var(char const *arg);

	typedef struct callb_cpr {
	kmutex_t *cc_lockp;
	} callb_cpr_t;

	#define CALLB_CPR_INIT(cp, lockp, func, name) { \
	(cp)->cc_lockp = lockp; \
	}

	#define CALLB_CPR_SAFE_BEGIN(cp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	}

	#define CALLB_CPR_SAFE_END(cp, lockp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	}

	#define CALLB_CPR_EXIT(cp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	mutex_exit((cp)->cc_lockp); \
	}

	#define zone_dataset_visible(x, y) (1)
	#define INGLOBALZONE(z) (1)
	extern uint32_t zone_get_hostid(void *zonep);

	extern char kmem_vasprintf(const char fmt, va_list adx);
	extern char kmem_asprintf(const char fmt, ...);
	#define kmem_strfree(str) kmem_free((str), strlen(str) + 1)
	#define kmem_strdup(s) strdup(s)

	/*
	* Hostname information
	*/
	extern int ddi_strtoull(const char str, char *nptr, int base,
	u_longlong_t *result);

	typedef struct utsname utsname_t;
	extern utsname_t *utsname(void);

	/* ZFS Boot Related stuff. */

	struct _buf {
	intptr_t _fd;
	};

	struct bootstat {
	uint64_t st_size;
	};

	typedef struct ace_object {
	uid_t a_who;
	uint32_t a_access_mask;
	uint16_t a_flags;
	uint16_t a_type;
	uint8_t a_obj_type[16];
	uint8_t a_inherit_obj_type[16];
	} ace_object_t;


	#define ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE 0x05
	#define ACE_ACCESS_DENIED_OBJECT_ACE_TYPE 0x06
	#define ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE 0x07
	#define ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE 0x08

	extern int zfs_secpolicy_snapshot_perms(const char name, cred_t cr);
	extern int zfs_secpolicy_rename_perms(const char from, const char to,
	cred_t *cr);
	extern int zfs_secpolicy_destroy_perms(const char name, cred_t cr);
	extern int secpolicy_zfs(const cred_t *cr);
	extern int secpolicy_zfs_proc(const cred_t cr, proc_t proc);
	extern zoneid_t getzoneid(void);

	/* SID stuff */
	typedef struct ksiddomain {
	uint_t kd_ref;
	uint_t kd_len;
	char *kd_name;
	} ksiddomain_t;

	ksiddomain_t ksid_lookupdomain(const char );
	void ksiddomain_rele(ksiddomain_t *);

	#define DDI_SLEEP KM_SLEEP
	#define ddi_log_sysevent(_a, _b, _c, _d, _e, _f, _g) \
	sysevent_post_event(_c, _d, _b, "libzpool", _e, _f)

	#define zfs_sleep_until(wakeup) \
	do { \
	hrtime_t delta = wakeup - gethrtime(); \
	struct timespec ts; \
	ts.tv_sec = delta / NANOSEC; \
	ts.tv_nsec = delta % NANOSEC; \
	(void) nanosleep(&ts, NULL); \
	} while (0)

	typedef int fstrans_cookie_t;

	extern fstrans_cookie_t spl_fstrans_mark(void);
	extern void spl_fstrans_unmark(fstrans_cookie_t);
	extern int __spl_pf_fstrans_check(void);
	extern int kmem_cache_reap_active(void);


	/*
	* Kernel modules
	*/
	#define __init
	#define __exit

	#endif /* _KERNEL \|\| _STANDALONE */

	#ifdef __cplusplus
	};
	#endif

	#endif /* _SYS_ZFS_CONTEXT_H */
	diff --git a/sys/contrib/openzfs/include/sys/zfs_ioctl.h b/sys/contrib/openzfs/include/sys/zfs_ioctl.h
	index a44c96364418..91439d4b7861 100644
	--- a/sys/contrib/openzfs/include/sys/zfs_ioctl.h
	+++ b/sys/contrib/openzfs/include/sys/zfs_ioctl.h
	@@ -1,581 +1,587 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright 2016 RackTop Systems.
	* Copyright (c) 2017, Intel Corporation.
	*/

	#ifndef _SYS_ZFS_IOCTL_H
	#define _SYS_ZFS_IOCTL_H

	#include <sys/cred.h>
	#include <sys/dmu.h>
	#include <sys/zio.h>
	#include <sys/dsl_deleg.h>
	#include <sys/spa.h>
	#include <sys/zfs_stat.h>

	#ifdef _KERNEL
	#include <sys/nvpair.h>
	#endif /* _KERNEL */

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* The structures in this file are passed between userland and the
	* kernel. Userland may be running a 32-bit process, while the kernel
	* is 64-bit. Therefore, these structures need to compile the same in
	* 32-bit and 64-bit. This means not using type "long", and adding
	* explicit padding so that the 32-bit structure will not be packed more
	* tightly than the 64-bit structure (which requires 64-bit alignment).
	*/

	/*
	* Property values for snapdir
	*/
	#define ZFS_SNAPDIR_HIDDEN 0
	#define ZFS_SNAPDIR_VISIBLE 1

	/*
	* Property values for snapdev
	*/
	#define ZFS_SNAPDEV_HIDDEN 0
	#define ZFS_SNAPDEV_VISIBLE 1
	/*
	* Property values for acltype
	*/
	#define ZFS_ACLTYPE_OFF 0
	#define ZFS_ACLTYPE_POSIX 1
	#define ZFS_ACLTYPE_NFSV4 2

	/*
	* Field manipulation macros for the drr_versioninfo field of the
	* send stream header.
	*/

	/*
	* Header types for zfs send streams.
	*/
	typedef enum drr_headertype {
	DMU_SUBSTREAM = 0x1,
	DMU_COMPOUNDSTREAM = 0x2
	} drr_headertype_t;

	#define DMU_GET_STREAM_HDRTYPE(vi) BF64_GET((vi), 0, 2)
	#define DMU_SET_STREAM_HDRTYPE(vi, x) BF64_SET((vi), 0, 2, x)

	#define DMU_GET_FEATUREFLAGS(vi) BF64_GET((vi), 2, 30)
	#define DMU_SET_FEATUREFLAGS(vi, x) BF64_SET((vi), 2, 30, x)

	/*
	* Feature flags for zfs send streams (flags in drr_versioninfo)
	*/

	#define DMU_BACKUP_FEATURE_DEDUP (1 << 0)
	#define DMU_BACKUP_FEATURE_DEDUPPROPS (1 << 1)
	#define DMU_BACKUP_FEATURE_SA_SPILL (1 << 2)
	/* flags #3 - #15 are reserved for incompatible closed-source implementations */
	#define DMU_BACKUP_FEATURE_EMBED_DATA (1 << 16)
	#define DMU_BACKUP_FEATURE_LZ4 (1 << 17)
	/* flag #18 is reserved for a Delphix feature */
	#define DMU_BACKUP_FEATURE_LARGE_BLOCKS (1 << 19)
	#define DMU_BACKUP_FEATURE_RESUMING (1 << 20)
	#define DMU_BACKUP_FEATURE_REDACTED (1 << 21)
	#define DMU_BACKUP_FEATURE_COMPRESSED (1 << 22)
	#define DMU_BACKUP_FEATURE_LARGE_DNODE (1 << 23)
	#define DMU_BACKUP_FEATURE_RAW (1 << 24)
	#define DMU_BACKUP_FEATURE_ZSTD (1 << 25)
	#define DMU_BACKUP_FEATURE_HOLDS (1 << 26)
	/*
	* The SWITCH_TO_LARGE_BLOCKS feature indicates that we can receive
	* incremental LARGE_BLOCKS streams (those with WRITE records of >128KB) even
	* if the previous send did not use LARGE_BLOCKS, and thus its large blocks
	* were split into multiple 128KB WRITE records. (See
	* flush_write_batch_impl() and receive_object()). Older software that does
	* not support this flag may encounter a bug when switching to large blocks,
	* which causes files to incorrectly be zeroed.
	*
	* This flag is currently not set on any send streams. In the future, we
	* intend for incremental send streams of snapshots that have large blocks to
	* use LARGE_BLOCKS by default, and these streams will also have the
	* SWITCH_TO_LARGE_BLOCKS feature set. This ensures that streams from the
	* default use of "zfs send" won't encounter the bug mentioned above.
	*/
	#define DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS (1 << 27)
	-#define DMU_BACKUP_FEATURE_BLAKE3 (1 << 28)
	+/* flag #28 is reserved for a Nutanix feature */
	+/*
	+ * flag #29 is the last unused bit. It is reserved to indicate a to-be-designed
	+ * extension to the stream format which will accomodate more feature flags.
	+ * If you need to add another feature flag, please reach out to the OpenZFS
	+ * community, e.g., on GitHub or Slack.
	+ */

	/*
	* Mask of all supported backup features
	*/
	#define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_SA_SPILL \| \
	DMU_BACKUP_FEATURE_EMBED_DATA \| DMU_BACKUP_FEATURE_LZ4 \| \
	DMU_BACKUP_FEATURE_RESUMING \| DMU_BACKUP_FEATURE_LARGE_BLOCKS \| \
	DMU_BACKUP_FEATURE_COMPRESSED \| DMU_BACKUP_FEATURE_LARGE_DNODE \| \
	DMU_BACKUP_FEATURE_RAW \| DMU_BACKUP_FEATURE_HOLDS \| \
	DMU_BACKUP_FEATURE_REDACTED \| DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS \| \
	- DMU_BACKUP_FEATURE_ZSTD \| DMU_BACKUP_FEATURE_BLAKE3)
	+ DMU_BACKUP_FEATURE_ZSTD)

	/* Are all features in the given flag word currently supported? */
	#define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK))

	typedef enum dmu_send_resume_token_version {
	ZFS_SEND_RESUME_TOKEN_VERSION = 1
	} dmu_send_resume_token_version_t;

	/*
	* The drr_versioninfo field of the dmu_replay_record has the
	* following layout:
	*
	* 64 56 48 40 32 24 16 8 0
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	* \| reserved \| feature-flags \|C\|S\|
	* +-------+-------+-------+-------+-------+-------+-------+-------+
	*
	* The low order two bits indicate the header type: SUBSTREAM (0x1)
	* or COMPOUNDSTREAM (0x2). Using two bits for this is historical:
	* this field used to be a version number, where the two version types
	* were 1 and 2. Using two bits for this allows earlier versions of
	* the code to be able to recognize send streams that don't use any
	* of the features indicated by feature flags.
	*/

	#define DMU_BACKUP_MAGIC 0x2F5bacbacULL

	/*
	* Send stream flags. Bits 24-31 are reserved for vendor-specific
	* implementations and should not be used.
	*/
	#define DRR_FLAG_CLONE (1<<0)
	#define DRR_FLAG_CI_DATA (1<<1)
	/*
	* This send stream, if it is a full send, includes the FREE and FREEOBJECT
	* records that are created by the sending process. This means that the send
	* stream can be received as a clone, even though it is not an incremental.
	* This is not implemented as a feature flag, because the receiving side does
	* not need to have implemented it to receive this stream; it is fully backwards
	* compatible. We need a flag, though, because full send streams without it
	* cannot necessarily be received as a clone correctly.
	*/
	#define DRR_FLAG_FREERECORDS (1<<2)
	/*
	* When DRR_FLAG_SPILL_BLOCK is set it indicates the DRR_OBJECT_SPILL
	* and DRR_SPILL_UNMODIFIED flags are meaningful in the send stream.
	*
	* When DRR_FLAG_SPILL_BLOCK is set, DRR_OBJECT records will have
	* DRR_OBJECT_SPILL set if and only if they should have a spill block
	* (either an existing one, or a new one in the send stream). When clear
	* the object does not have a spill block and any existing spill block
	* should be freed.
	*
	* Similarly, when DRR_FLAG_SPILL_BLOCK is set, DRR_SPILL records will
	* have DRR_SPILL_UNMODIFIED set if and only if they were included for
	* backward compatibility purposes, and can be safely ignored by new versions
	* of zfs receive. Previous versions of ZFS which do not understand the
	* DRR_FLAG_SPILL_BLOCK will process this record and recreate any missing
	* spill blocks.
	*/
	#define DRR_FLAG_SPILL_BLOCK (1<<3)

	/*
	* flags in the drr_flags field in the DRR_WRITE, DRR_SPILL, DRR_OBJECT,
	* DRR_WRITE_BYREF, and DRR_OBJECT_RANGE blocks
	*/
	#define DRR_CHECKSUM_DEDUP (1<<0) /* not used for SPILL records */
	#define DRR_RAW_BYTESWAP (1<<1)
	#define DRR_OBJECT_SPILL (1<<2) /* OBJECT record has a spill block */
	#define DRR_SPILL_UNMODIFIED (1<<2) /* SPILL record for unmodified block */

	#define DRR_IS_DEDUP_CAPABLE(flags) ((flags) & DRR_CHECKSUM_DEDUP)
	#define DRR_IS_RAW_BYTESWAPPED(flags) ((flags) & DRR_RAW_BYTESWAP)
	#define DRR_OBJECT_HAS_SPILL(flags) ((flags) & DRR_OBJECT_SPILL)
	#define DRR_SPILL_IS_UNMODIFIED(flags) ((flags) & DRR_SPILL_UNMODIFIED)

	/* deal with compressed drr_write replay records */
	#define DRR_WRITE_COMPRESSED(drrw) ((drrw)->drr_compressiontype != 0)
	#define DRR_WRITE_PAYLOAD_SIZE(drrw) \
	(DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
	(drrw)->drr_logical_size)
	#define DRR_SPILL_PAYLOAD_SIZE(drrs) \
	((drrs)->drr_compressed_size ? \
	(drrs)->drr_compressed_size : (drrs)->drr_length)
	#define DRR_OBJECT_PAYLOAD_SIZE(drro) \
	((drro)->drr_raw_bonuslen != 0 ? \
	(drro)->drr_raw_bonuslen : P2ROUNDUP((drro)->drr_bonuslen, 8))

	/*
	* zfs ioctl command structure
	*/

	/* Header is used in C++ so can't forward declare untagged struct */
	struct drr_begin {
	uint64_t drr_magic;
	uint64_t drr_versioninfo; /* was drr_version */
	uint64_t drr_creation_time;
	dmu_objset_type_t drr_type;
	uint32_t drr_flags;
	uint64_t drr_toguid;
	uint64_t drr_fromguid;
	char drr_toname[MAXNAMELEN];
	};

	typedef struct dmu_replay_record {
	enum {
	DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
	DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
	DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_OBJECT_RANGE, DRR_REDACT,
	DRR_NUMTYPES
	} drr_type;
	uint32_t drr_payloadlen;
	union {
	struct drr_begin drr_begin;
	struct drr_end {
	zio_cksum_t drr_checksum;
	uint64_t drr_toguid;
	} drr_end;
	struct drr_object {
	uint64_t drr_object;
	dmu_object_type_t drr_type;
	dmu_object_type_t drr_bonustype;
	uint32_t drr_blksz;
	uint32_t drr_bonuslen;
	uint8_t drr_checksumtype;
	uint8_t drr_compress;
	uint8_t drr_dn_slots;
	uint8_t drr_flags;
	uint32_t drr_raw_bonuslen;
	uint64_t drr_toguid;
	/* only (possibly) nonzero for raw streams */
	uint8_t drr_indblkshift;
	uint8_t drr_nlevels;
	uint8_t drr_nblkptr;
	uint8_t drr_pad[5];
	uint64_t drr_maxblkid;
	/* bonus content follows */
	} drr_object;
	struct drr_freeobjects {
	uint64_t drr_firstobj;
	uint64_t drr_numobjs;
	uint64_t drr_toguid;
	} drr_freeobjects;
	struct drr_write {
	uint64_t drr_object;
	dmu_object_type_t drr_type;
	uint32_t drr_pad;
	uint64_t drr_offset;
	uint64_t drr_logical_size;
	uint64_t drr_toguid;
	uint8_t drr_checksumtype;
	uint8_t drr_flags;
	uint8_t drr_compressiontype;
	uint8_t drr_pad2[5];
	/* deduplication key */
	ddt_key_t drr_key;
	/* only nonzero if drr_compressiontype is not 0 */
	uint64_t drr_compressed_size;
	/* only nonzero for raw streams */
	uint8_t drr_salt[ZIO_DATA_SALT_LEN];
	uint8_t drr_iv[ZIO_DATA_IV_LEN];
	uint8_t drr_mac[ZIO_DATA_MAC_LEN];
	/* content follows */
	} drr_write;
	struct drr_free {
	uint64_t drr_object;
	uint64_t drr_offset;
	uint64_t drr_length;
	uint64_t drr_toguid;
	} drr_free;
	struct drr_write_byref {
	/* where to put the data */
	uint64_t drr_object;
	uint64_t drr_offset;
	uint64_t drr_length;
	uint64_t drr_toguid;
	/* where to find the prior copy of the data */
	uint64_t drr_refguid;
	uint64_t drr_refobject;
	uint64_t drr_refoffset;
	/* properties of the data */
	uint8_t drr_checksumtype;
	uint8_t drr_flags;
	uint8_t drr_pad2[6];
	ddt_key_t drr_key; /* deduplication key */
	} drr_write_byref;
	struct drr_spill {
	uint64_t drr_object;
	uint64_t drr_length;
	uint64_t drr_toguid;
	uint8_t drr_flags;
	uint8_t drr_compressiontype;
	uint8_t drr_pad[6];
	/* only nonzero for raw streams */
	uint64_t drr_compressed_size;
	uint8_t drr_salt[ZIO_DATA_SALT_LEN];
	uint8_t drr_iv[ZIO_DATA_IV_LEN];
	uint8_t drr_mac[ZIO_DATA_MAC_LEN];
	dmu_object_type_t drr_type;
	/* spill data follows */
	} drr_spill;
	struct drr_write_embedded {
	uint64_t drr_object;
	uint64_t drr_offset;
	/* logical length, should equal blocksize */
	uint64_t drr_length;
	uint64_t drr_toguid;
	uint8_t drr_compression;
	uint8_t drr_etype;
	uint8_t drr_pad[6];
	uint32_t drr_lsize; /* uncompressed size of payload */
	uint32_t drr_psize; /* compr. (real) size of payload */
	/* (possibly compressed) content follows */
	} drr_write_embedded;
	struct drr_object_range {
	uint64_t drr_firstobj;
	uint64_t drr_numslots;
	uint64_t drr_toguid;
	uint8_t drr_salt[ZIO_DATA_SALT_LEN];
	uint8_t drr_iv[ZIO_DATA_IV_LEN];
	uint8_t drr_mac[ZIO_DATA_MAC_LEN];
	uint8_t drr_flags;
	uint8_t drr_pad[3];
	} drr_object_range;
	struct drr_redact {
	uint64_t drr_object;
	uint64_t drr_offset;
	uint64_t drr_length;
	uint64_t drr_toguid;
	} drr_redact;

	/*
	* Note: drr_checksum is overlaid with all record types
	* except DRR_BEGIN. Therefore its (non-pad) members
	* must not overlap with members from the other structs.
	* We accomplish this by putting its members at the very
	* end of the struct.
	*/
	struct drr_checksum {
	uint64_t drr_pad[34];
	/*
	* fletcher-4 checksum of everything preceding the
	* checksum.
	*/
	zio_cksum_t drr_checksum;
	} drr_checksum;
	} drr_u;
	} dmu_replay_record_t;

	/* diff record range types */
	typedef enum diff_type {
	DDR_NONE = 0x1,
	DDR_INUSE = 0x2,
	DDR_FREE = 0x4
	} diff_type_t;

	/*
	* The diff reports back ranges of free or in-use objects.
	*/
	typedef struct dmu_diff_record {
	uint64_t ddr_type;
	uint64_t ddr_first;
	uint64_t ddr_last;
	} dmu_diff_record_t;

	typedef struct zinject_record {
	uint64_t zi_objset;
	uint64_t zi_object;
	uint64_t zi_start;
	uint64_t zi_end;
	uint64_t zi_guid;
	uint32_t zi_level;
	uint32_t zi_error;
	uint64_t zi_type;
	uint32_t zi_freq;
	uint32_t zi_failfast;
	char zi_func[MAXNAMELEN];
	uint32_t zi_iotype;
	int32_t zi_duration;
	uint64_t zi_timer;
	uint64_t zi_nlanes;
	uint32_t zi_cmd;
	uint32_t zi_dvas;
	} zinject_record_t;

	#define ZINJECT_NULL 0x1
	#define ZINJECT_FLUSH_ARC 0x2
	#define ZINJECT_UNLOAD_SPA 0x4
	#define ZINJECT_CALC_RANGE 0x8

	#define ZEVENT_NONE 0x0
	#define ZEVENT_NONBLOCK 0x1
	#define ZEVENT_SIZE 1024

	#define ZEVENT_SEEK_START 0
	#define ZEVENT_SEEK_END UINT64_MAX

	/* scaled frequency ranges */
	#define ZI_PERCENTAGE_MIN 4294UL
	#define ZI_PERCENTAGE_MAX UINT32_MAX

	#define ZI_NO_DVA (-1)

	typedef enum zinject_type {
	ZINJECT_UNINITIALIZED,
	ZINJECT_DATA_FAULT,
	ZINJECT_DEVICE_FAULT,
	ZINJECT_LABEL_FAULT,
	ZINJECT_IGNORED_WRITES,
	ZINJECT_PANIC,
	ZINJECT_DELAY_IO,
	ZINJECT_DECRYPT_FAULT,
	} zinject_type_t;

	typedef struct zfs_share {
	uint64_t z_exportdata;
	uint64_t z_sharedata;
	uint64_t z_sharetype; /* 0 = share, 1 = unshare */
	uint64_t z_sharemax; /* max length of share string */
	} zfs_share_t;

	/*
	* ZFS file systems may behave the usual, POSIX-compliant way, where
	* name lookups are case-sensitive. They may also be set up so that
	* all the name lookups are case-insensitive, or so that only some
	* lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
	*/
	typedef enum zfs_case {
	ZFS_CASE_SENSITIVE,
	ZFS_CASE_INSENSITIVE,
	ZFS_CASE_MIXED
	} zfs_case_t;

	/*
	* Note: this struct must have the same layout in 32-bit and 64-bit, so
	* that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
	* kernel. Therefore, we add padding to it so that no "hidden" padding
	* is automatically added on 64-bit (but not on 32-bit).
	*/
	typedef struct zfs_cmd {
	char zc_name[MAXPATHLEN]; /* name of pool or dataset */
	uint64_t zc_nvlist_src; /* really (char ) /
	uint64_t zc_nvlist_src_size;
	uint64_t zc_nvlist_dst; /* really (char ) /
	uint64_t zc_nvlist_dst_size;
	boolean_t zc_nvlist_dst_filled; /* put an nvlist in dst? */
	int zc_pad2;

	/*
	* The following members are for legacy ioctls which haven't been
	* converted to the new method.
	*/
	uint64_t zc_history; /* really (char ) /
	char zc_value[MAXPATHLEN * 2];
	char zc_string[MAXNAMELEN];
	uint64_t zc_guid;
	uint64_t zc_nvlist_conf; /* really (char ) /
	uint64_t zc_nvlist_conf_size;
	uint64_t zc_cookie;
	uint64_t zc_objset_type;
	uint64_t zc_perm_action;
	uint64_t zc_history_len;
	uint64_t zc_history_offset;
	uint64_t zc_obj;
	uint64_t zc_iflags; /* internal to zfs(7fs) */
	zfs_share_t zc_share;
	dmu_objset_stats_t zc_objset_stats;
	struct drr_begin zc_begin_record;
	zinject_record_t zc_inject_record;
	uint32_t zc_defer_destroy;
	uint32_t zc_flags;
	uint64_t zc_action_handle;
	int zc_cleanup_fd;
	uint8_t zc_simple;
	uint8_t zc_pad[3]; /* alignment */
	uint64_t zc_sendobj;
	uint64_t zc_fromobj;
	uint64_t zc_createtxg;
	zfs_stat_t zc_stat;
	uint64_t zc_zoneid;
	} zfs_cmd_t;

	typedef struct zfs_useracct {
	char zu_domain[256];
	uid_t zu_rid;
	uint32_t zu_pad;
	uint64_t zu_space;
	} zfs_useracct_t;

	#define ZFSDEV_MAX_MINOR (1 << 16)

	#define ZPOOL_EXPORT_AFTER_SPLIT 0x1

	#ifdef _KERNEL
	struct objset;
	struct zfsvfs;

	typedef struct zfs_creat {
	nvlist_t *zct_zplprops;
	nvlist_t *zct_props;
	} zfs_creat_t;

	extern int zfs_secpolicy_snapshot_perms(const char , cred_t );
	extern int zfs_secpolicy_rename_perms(const char , const char , cred_t *);
	extern int zfs_secpolicy_destroy_perms(const char , cred_t );
	extern void zfs_unmount_snap(const char *);
	extern void zfs_destroy_unmount_origin(const char *);
	extern int getzfsvfs_impl(struct objset , struct zfsvfs *);
	extern int getzfsvfs(const char , struct zfsvfs *);

	enum zfsdev_state_type {
	ZST_ONEXIT,
	ZST_ZEVENT,
	ZST_ALL,
	};

	/*
	* The zfsdev_state_t structure is managed as a singly-linked list
	* from which items are never deleted. This allows for lock-free
	* reading of the list so long as assignments to the zs_next and
	* reads from zs_minor are performed atomically. Empty items are
	* indicated by storing -1 into zs_minor.
	*/
	typedef struct zfsdev_state {
	struct zfsdev_state zs_next; / next zfsdev_state_t link */
	minor_t zs_minor; /* made up minor number */
	void zs_onexit; / onexit data */
	void zs_zevent; / zevent data */
	} zfsdev_state_t;

	extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which);
	extern int zfsdev_getminor(zfs_file_t fp, minor_t minorp);

	extern uint_t zfs_fsyncer_key;
	extern uint_t zfs_allow_log_key;

	#endif /* _KERNEL */

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ZFS_IOCTL_H */
	diff --git a/sys/contrib/openzfs/include/sys/zio_checksum.h b/sys/contrib/openzfs/include/sys/zio_checksum.h
	index 84e371d6c3ef..5903678dfb41 100644
	--- a/sys/contrib/openzfs/include/sys/zio_checksum.h
	+++ b/sys/contrib/openzfs/include/sys/zio_checksum.h
	@@ -1,155 +1,155 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2014, 2016 by Delphix. All rights reserved.
	* Copyright (c) 2013 Saso Kiselkov, All rights reserved.
	* Copyright (c) 2021 Tino Reichardt <milky-zfs@mcmilk.de>
	*/

	#ifndef _SYS_ZIO_CHECKSUM_H
	#define _SYS_ZIO_CHECKSUM_H extern __attribute__((visibility("default")))

	#include <sys/zio.h>
	#include <zfeature_common.h>
	#include <zfs_fletcher.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	struct abd;

	/*
	* Signature for checksum functions.
	*/
	typedef void zio_checksum_t(struct abd *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp);
	typedef void zio_checksum_tmpl_init_t(const zio_cksum_salt_t salt);
	typedef void zio_checksum_tmpl_free_t(void *ctx_template);

	typedef enum zio_checksum_flags {
	/* Strong enough for metadata? */
	ZCHECKSUM_FLAG_METADATA = (1 << 1),
	/* ZIO embedded checksum */
	ZCHECKSUM_FLAG_EMBEDDED = (1 << 2),
	/* Strong enough for dedup (without verification)? */
	ZCHECKSUM_FLAG_DEDUP = (1 << 3),
	/* Uses salt value */
	ZCHECKSUM_FLAG_SALTED = (1 << 4),
	/* Strong enough for nopwrite? */
	ZCHECKSUM_FLAG_NOPWRITE = (1 << 5)
	} zio_checksum_flags_t;

	typedef enum {
	ZIO_CHECKSUM_NATIVE,
	ZIO_CHECKSUM_BYTESWAP
	} zio_byteorder_t;

	typedef struct zio_abd_checksum_data {
	zio_byteorder_t acd_byteorder;
	fletcher_4_ctx_t *acd_ctx;
	zio_cksum_t *acd_zcp;
	void *acd_private;
	} zio_abd_checksum_data_t;

	typedef void zio_abd_checksum_init_t(zio_abd_checksum_data_t *);
	typedef void zio_abd_checksum_fini_t(zio_abd_checksum_data_t *);
	typedef int zio_abd_checksum_iter_t(void , size_t, void );

	typedef const struct zio_abd_checksum_func {
	zio_abd_checksum_init_t *acf_init;
	zio_abd_checksum_fini_t *acf_fini;
	zio_abd_checksum_iter_t *acf_iter;
	} zio_abd_checksum_func_t;

	/*
	* Information about each checksum function.
	*/
	typedef const struct zio_checksum_info {
	/* checksum function for each byteorder */
	zio_checksum_t *ci_func[2];
	zio_checksum_tmpl_init_t *ci_tmpl_init;
	zio_checksum_tmpl_free_t *ci_tmpl_free;
	zio_checksum_flags_t ci_flags;
	const char ci_name; / descriptive name */
	} zio_checksum_info_t;

	typedef struct zio_bad_cksum {
	zio_cksum_t zbc_expected;
	zio_cksum_t zbc_actual;
	const char *zbc_checksum_name;
	uint8_t zbc_byteswapped;
	uint8_t zbc_injected;
	uint8_t zbc_has_cksum; /* expected/actual valid */
	} zio_bad_cksum_t;

	-_SYS_ZIO_CHECKSUM_H const zio_checksum_info_t
	+_SYS_ZIO_CHECKSUM_H zio_checksum_info_t
	zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS];

	/*
	* Checksum routines.
	*/

	/* SHA2 */
	extern zio_checksum_t abd_checksum_SHA256;
	extern zio_checksum_t abd_checksum_SHA512_native;
	extern zio_checksum_t abd_checksum_SHA512_byteswap;

	/* Skein */
	extern zio_checksum_t abd_checksum_skein_native;
	extern zio_checksum_t abd_checksum_skein_byteswap;
	extern zio_checksum_tmpl_init_t abd_checksum_skein_tmpl_init;
	extern zio_checksum_tmpl_free_t abd_checksum_skein_tmpl_free;

	/* Edon-R */
	extern zio_checksum_t abd_checksum_edonr_native;
	extern zio_checksum_t abd_checksum_edonr_byteswap;
	extern zio_checksum_tmpl_init_t abd_checksum_edonr_tmpl_init;
	extern zio_checksum_tmpl_free_t abd_checksum_edonr_tmpl_free;

	/* BLAKE3 */
	extern zio_checksum_t abd_checksum_blake3_native;
	extern zio_checksum_t abd_checksum_blake3_byteswap;
	extern zio_checksum_tmpl_init_t abd_checksum_blake3_tmpl_init;
	extern zio_checksum_tmpl_free_t abd_checksum_blake3_tmpl_free;

	/* Fletcher 4 */
	_SYS_ZIO_CHECKSUM_H zio_abd_checksum_func_t fletcher_4_abd_ops;
	extern zio_checksum_t abd_fletcher_4_native;
	extern zio_checksum_t abd_fletcher_4_byteswap;

	extern int zio_checksum_equal(spa_t , blkptr_t , enum zio_checksum,
	void , uint64_t, uint64_t, zio_bad_cksum_t );
	extern void zio_checksum_compute(zio_t *, enum zio_checksum,
	struct abd *, uint64_t);
	extern int zio_checksum_error_impl(spa_t , const blkptr_t , enum zio_checksum,
	struct abd , uint64_t, uint64_t, zio_bad_cksum_t );
	extern int zio_checksum_error(zio_t zio, zio_bad_cksum_t out);
	extern enum zio_checksum spa_dedup_checksum(spa_t *spa);
	extern void zio_checksum_templates_free(spa_t *spa);
	extern spa_feature_t zio_checksum_to_feature(enum zio_checksum cksum);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ZIO_CHECKSUM_H */
	diff --git a/sys/contrib/openzfs/include/sys/zio_compress.h b/sys/contrib/openzfs/include/sys/zio_compress.h
	index a736d8091986..19fc44952db6 100644
	--- a/sys/contrib/openzfs/include/sys/zio_compress.h
	+++ b/sys/contrib/openzfs/include/sys/zio_compress.h
	@@ -1,198 +1,198 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2019, Klara Inc.
	* Use is subject to license terms.
	* Copyright (c) 2015, 2016 by Delphix. All rights reserved.
	*/

	#ifndef _SYS_ZIO_COMPRESS_H
	#define _SYS_ZIO_COMPRESS_H

	#include <sys/abd.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	enum zio_compress {
	ZIO_COMPRESS_INHERIT = 0,
	ZIO_COMPRESS_ON,
	ZIO_COMPRESS_OFF,
	ZIO_COMPRESS_LZJB,
	ZIO_COMPRESS_EMPTY,
	ZIO_COMPRESS_GZIP_1,
	ZIO_COMPRESS_GZIP_2,
	ZIO_COMPRESS_GZIP_3,
	ZIO_COMPRESS_GZIP_4,
	ZIO_COMPRESS_GZIP_5,
	ZIO_COMPRESS_GZIP_6,
	ZIO_COMPRESS_GZIP_7,
	ZIO_COMPRESS_GZIP_8,
	ZIO_COMPRESS_GZIP_9,
	ZIO_COMPRESS_ZLE,
	ZIO_COMPRESS_LZ4,
	ZIO_COMPRESS_ZSTD,
	ZIO_COMPRESS_FUNCTIONS
	};

	/* Compression algorithms that have levels */
	#define ZIO_COMPRESS_HASLEVEL(compress) ((compress == ZIO_COMPRESS_ZSTD \|\| \
	(compress >= ZIO_COMPRESS_GZIP_1 && \
	compress <= ZIO_COMPRESS_GZIP_9)))

	#define ZIO_COMPLEVEL_INHERIT 0
	#define ZIO_COMPLEVEL_DEFAULT 255

	enum zio_zstd_levels {
	ZIO_ZSTD_LEVEL_INHERIT = 0,
	ZIO_ZSTD_LEVEL_1,
	#define ZIO_ZSTD_LEVEL_MIN ZIO_ZSTD_LEVEL_1
	ZIO_ZSTD_LEVEL_2,
	ZIO_ZSTD_LEVEL_3,
	#define ZIO_ZSTD_LEVEL_DEFAULT ZIO_ZSTD_LEVEL_3
	ZIO_ZSTD_LEVEL_4,
	ZIO_ZSTD_LEVEL_5,
	ZIO_ZSTD_LEVEL_6,
	ZIO_ZSTD_LEVEL_7,
	ZIO_ZSTD_LEVEL_8,
	ZIO_ZSTD_LEVEL_9,
	ZIO_ZSTD_LEVEL_10,
	ZIO_ZSTD_LEVEL_11,
	ZIO_ZSTD_LEVEL_12,
	ZIO_ZSTD_LEVEL_13,
	ZIO_ZSTD_LEVEL_14,
	ZIO_ZSTD_LEVEL_15,
	ZIO_ZSTD_LEVEL_16,
	ZIO_ZSTD_LEVEL_17,
	ZIO_ZSTD_LEVEL_18,
	ZIO_ZSTD_LEVEL_19,
	#define ZIO_ZSTD_LEVEL_MAX ZIO_ZSTD_LEVEL_19
	ZIO_ZSTD_LEVEL_RESERVE = 101, /* Leave room for new positive levels */
	ZIO_ZSTD_LEVEL_FAST, /* Fast levels are negative */
	ZIO_ZSTD_LEVEL_FAST_1,
	#define ZIO_ZSTD_LEVEL_FAST_DEFAULT ZIO_ZSTD_LEVEL_FAST_1
	ZIO_ZSTD_LEVEL_FAST_2,
	ZIO_ZSTD_LEVEL_FAST_3,
	ZIO_ZSTD_LEVEL_FAST_4,
	ZIO_ZSTD_LEVEL_FAST_5,
	ZIO_ZSTD_LEVEL_FAST_6,
	ZIO_ZSTD_LEVEL_FAST_7,
	ZIO_ZSTD_LEVEL_FAST_8,
	ZIO_ZSTD_LEVEL_FAST_9,
	ZIO_ZSTD_LEVEL_FAST_10,
	ZIO_ZSTD_LEVEL_FAST_20,
	ZIO_ZSTD_LEVEL_FAST_30,
	ZIO_ZSTD_LEVEL_FAST_40,
	ZIO_ZSTD_LEVEL_FAST_50,
	ZIO_ZSTD_LEVEL_FAST_60,
	ZIO_ZSTD_LEVEL_FAST_70,
	ZIO_ZSTD_LEVEL_FAST_80,
	ZIO_ZSTD_LEVEL_FAST_90,
	ZIO_ZSTD_LEVEL_FAST_100,
	ZIO_ZSTD_LEVEL_FAST_500,
	ZIO_ZSTD_LEVEL_FAST_1000,
	#define ZIO_ZSTD_LEVEL_FAST_MAX ZIO_ZSTD_LEVEL_FAST_1000
	ZIO_ZSTD_LEVEL_AUTO = 251, /* Reserved for future use */
	ZIO_ZSTD_LEVEL_LEVELS
	};

	/* Forward Declaration to avoid visibility problems */
	struct zio_prop;

	/* Common signature for all zio compress functions. */
	typedef size_t zio_compress_func_t(void src, void dst,
	size_t s_len, size_t d_len, int);
	/* Common signature for all zio decompress functions. */
	typedef int zio_decompress_func_t(void src, void dst,
	size_t s_len, size_t d_len, int);
	/* Common signature for all zio decompress and get level functions. */
	typedef int zio_decompresslevel_func_t(void src, void dst,
	size_t s_len, size_t d_len, uint8_t *level);
	/* Common signature for all zio get-compression-level functions. */
	typedef int zio_getlevel_func_t(void src, size_t s_len, uint8_t level);


	/*
	* Common signature for all zio decompress functions using an ABD as input.
	* This is helpful if you have both compressed ARC and scatter ABDs enabled,
	* but is not a requirement for all compression algorithms.
	*/
	typedef int zio_decompress_abd_func_t(abd_t src, void dst,
	size_t s_len, size_t d_len, int);
	/*
	* Information about each compression function.
	*/
	typedef const struct zio_compress_info {
	const char *ci_name;
	int ci_level;
	zio_compress_func_t *ci_compress;
	zio_decompress_func_t *ci_decompress;
	zio_decompresslevel_func_t *ci_decompress_level;
	} zio_compress_info_t;

	-extern const zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS];
	+extern zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS];

	/*
	* lz4 compression init & free
	*/
	extern void lz4_init(void);
	extern void lz4_fini(void);

	/*
	* Compression routines.
	*/
	extern size_t lzjb_compress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern int lzjb_decompress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern size_t gzip_compress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern int gzip_decompress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern size_t zle_compress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern int zle_decompress(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern size_t lz4_compress_zfs(void src, void dst, size_t s_len, size_t d_len,
	int level);
	extern int lz4_decompress_zfs(void src, void dst, size_t s_len, size_t d_len,
	int level);

	/*
	* Compress and decompress data if necessary.
	*/
	extern size_t zio_compress_data(enum zio_compress c, abd_t src, void dst,
	size_t s_len, uint8_t level);
	extern int zio_decompress_data(enum zio_compress c, abd_t src, void dst,
	size_t s_len, size_t d_len, uint8_t *level);
	extern int zio_decompress_data_buf(enum zio_compress c, void src, void dst,
	size_t s_len, size_t d_len, uint8_t *level);
	extern int zio_compress_to_feature(enum zio_compress comp);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ZIO_COMPRESS_H */
	diff --git a/sys/contrib/openzfs/lib/libshare/nfs.c b/sys/contrib/openzfs/lib/libshare/nfs.c
	index bbaea93fca5c..3251913cba63 100644
	--- a/sys/contrib/openzfs/lib/libshare/nfs.c
	+++ b/sys/contrib/openzfs/lib/libshare/nfs.c
	@@ -1,313 +1,313 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/


	#include <sys/types.h>
	#include <sys/stat.h>
	#include <sys/file.h>
	#include <fcntl.h>
	#include <ctype.h>
	#include <stdio.h>
	#include <errno.h>
	#include <libshare.h>
	#include <unistd.h>
	#include "nfs.h"


	/*
	* nfs_exports_[lock\|unlock] are used to guard against conconcurrent
	* updates to the exports file. Each protocol is responsible for
	* providing the necessary locking to ensure consistency.
	*/
	static int
	nfs_exports_lock(const char name, int nfs_lock_fd)
	{
	int err;

	*nfs_lock_fd = open(name, O_RDWR \| O_CREAT \| O_CLOEXEC, 0600);
	if (*nfs_lock_fd == -1) {
	err = errno;
	fprintf(stderr, "failed to lock %s: %s\n", name, strerror(err));
	return (err);
	}

	while ((err = flock(*nfs_lock_fd, LOCK_EX)) != 0 && errno == EINTR)
	;
	if (err != 0) {
	err = errno;
	fprintf(stderr, "failed to lock %s: %s\n", name, strerror(err));
	(void) close(*nfs_lock_fd);
	*nfs_lock_fd = -1;
	return (err);
	}

	return (0);
	}

	static void
	nfs_exports_unlock(const char name, int nfs_lock_fd)
	{
	verify(*nfs_lock_fd > 0);

	if (flock(*nfs_lock_fd, LOCK_UN) != 0)
	fprintf(stderr, "failed to unlock %s: %s\n",
	name, strerror(errno));

	(void) close(*nfs_lock_fd);
	*nfs_lock_fd = -1;
	}

	struct tmpfile {
	/*
	* This only needs to be as wide as ZFS_EXPORTS_FILE and mktemp suffix,
	* 64 is more than enough.
	*/
	char name[64];
	FILE *fp;
	};

	static boolean_t
	nfs_init_tmpfile(const char prefix, const char mdir, struct tmpfile *tmpf)
	{
	if (mdir != NULL &&
	mkdir(mdir, 0755) < 0 &&
	errno != EEXIST) {
	fprintf(stderr, "failed to create %s: %s\n",
	mdir, strerror(errno));
	return (B_FALSE);
	}

	- strcpy(tmpf->name, prefix);
	- strcat(tmpf->name, ".XXXXXXXX");
	+ strlcpy(tmpf->name, prefix, sizeof (tmpf->name));
	+ strlcat(tmpf->name, ".XXXXXXXX", sizeof (tmpf->name) - strlen(prefix));

	int fd = mkostemp(tmpf->name, O_CLOEXEC);
	if (fd == -1) {
	fprintf(stderr, "Unable to create temporary file: %s",
	strerror(errno));
	return (B_FALSE);
	}

	tmpf->fp = fdopen(fd, "w+");
	if (tmpf->fp == NULL) {
	fprintf(stderr, "Unable to reopen temporary file: %s",
	strerror(errno));
	close(fd);
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	static void
	nfs_abort_tmpfile(struct tmpfile *tmpf)
	{
	unlink(tmpf->name);
	fclose(tmpf->fp);
	}

	static int
	nfs_fini_tmpfile(const char exports, struct tmpfile tmpf)
	{
	if (fflush(tmpf->fp) != 0) {
	fprintf(stderr, "Failed to write to temporary file: %s\n",
	strerror(errno));
	nfs_abort_tmpfile(tmpf);
	return (SA_SYSTEM_ERR);
	}

	if (rename(tmpf->name, exports) == -1) {
	fprintf(stderr, "Unable to rename %s -> %s: %s\n",
	tmpf->name, exports, strerror(errno));
	nfs_abort_tmpfile(tmpf);
	return (SA_SYSTEM_ERR);
	}

	(void) fchmod(fileno(tmpf->fp), 0644);
	fclose(tmpf->fp);
	return (SA_OK);
	}

	int
	nfs_escape_mountpoint(const char mp, char out, boolean_t need_free)
	{
	if (strpbrk(mp, "\t\n\v\f\r \\") == NULL) {
	out = (char )mp;
	*need_free = B_FALSE;
	return (SA_OK);
	} else {
	size_t len = strlen(mp);
	out = malloc(len 4 + 1);
	if (!*out)
	return (SA_NO_MEMORY);
	*need_free = B_TRUE;

	char oc = out;
	for (const char *c = mp; c < mp + len; ++c)
	if (memchr("\t\n\v\f\r \\", *c,
	strlen("\t\n\v\f\r \\"))) {
	sprintf(oc, "\\%03hho", *c);
	oc += 4;
	} else
	oc++ = c;
	*oc = '\0';
	}

	return (SA_OK);
	}

	static int
	nfs_process_exports(const char exports, const char mountpoint,
	boolean_t (cbk)(void userdata, char *line, boolean_t found_mountpoint),
	void *userdata)
	{
	int error = SA_OK;
	boolean_t cont = B_TRUE;

	FILE *oldfp = fopen(exports, "re");
	if (oldfp != NULL) {
	boolean_t need_mp_free;
	char *mp;
	if ((error = nfs_escape_mountpoint(mountpoint,
	&mp, &need_mp_free)) != SA_OK) {
	(void) fclose(oldfp);
	return (error);
	}

	char buf = NULL, sep;
	size_t buflen = 0, mplen = strlen(mp);

	while (cont && getline(&buf, &buflen, oldfp) != -1) {
	if (buf[0] == '\n' \|\| buf[0] == '#')
	continue;

	cont = cbk(userdata, buf,
	(sep = strpbrk(buf, "\t \n")) != NULL &&
	sep - buf == mplen &&
	strncmp(buf, mp, mplen) == 0);
	}
	free(buf);
	if (need_mp_free)
	free(mp);

	if (ferror(oldfp) != 0)
	error = ferror(oldfp);

	if (fclose(oldfp) != 0) {
	fprintf(stderr, "Unable to close file %s: %s\n",
	exports, strerror(errno));
	error = error != SA_OK ? error : SA_SYSTEM_ERR;
	}
	}

	return (error);
	}

	static boolean_t
	nfs_copy_entries_cb(void userdata, char line, boolean_t found_mountpoint)
	{
	FILE *newfp = userdata;
	if (!found_mountpoint)
	fputs(line, newfp);
	return (B_TRUE);
	}

	/*
	* Copy all entries from the exports file (if it exists) to newfp,
	* omitting any entries for the specified mountpoint.
	*/
	static int
	nfs_copy_entries(FILE newfp, const char exports, const char *mountpoint)
	{
	fputs(FILE_HEADER, newfp);

	int error = nfs_process_exports(
	exports, mountpoint, nfs_copy_entries_cb, newfp);

	if (error == SA_OK && ferror(newfp) != 0)
	error = ferror(newfp);

	return (error);
	}

	int
	nfs_toggle_share(const char lockfile, const char exports,
	const char *expdir, sa_share_impl_t impl_share,
	int(cbk)(sa_share_impl_t impl_share, FILE tmpfile))
	{
	int error, nfs_lock_fd = -1;
	struct tmpfile tmpf;

	if (!nfs_init_tmpfile(exports, expdir, &tmpf))
	return (SA_SYSTEM_ERR);

	error = nfs_exports_lock(lockfile, &nfs_lock_fd);
	if (error != 0) {
	nfs_abort_tmpfile(&tmpf);
	return (error);
	}

	error = nfs_copy_entries(tmpf.fp, exports, impl_share->sa_mountpoint);
	if (error != SA_OK)
	goto fullerr;

	error = cbk(impl_share, tmpf.fp);
	if (error != SA_OK)
	goto fullerr;

	error = nfs_fini_tmpfile(exports, &tmpf);
	nfs_exports_unlock(lockfile, &nfs_lock_fd);
	return (error);

	fullerr:
	nfs_abort_tmpfile(&tmpf);
	nfs_exports_unlock(lockfile, &nfs_lock_fd);
	return (error);
	}

	void
	nfs_reset_shares(const char lockfile, const char exports)
	{
	int nfs_lock_fd = -1;

	if (nfs_exports_lock(lockfile, &nfs_lock_fd) == 0) {
	(void) ! truncate(exports, 0);
	nfs_exports_unlock(lockfile, &nfs_lock_fd);
	}
	}

	static boolean_t
	nfs_is_shared_cb(void userdata, char line, boolean_t found_mountpoint)
	{
	(void) line;

	boolean_t *found = userdata;
	*found = found_mountpoint;
	return (!found_mountpoint);
	}

	boolean_t
	nfs_is_shared_impl(const char *exports, sa_share_impl_t impl_share)
	{
	boolean_t found = B_FALSE;
	nfs_process_exports(exports, impl_share->sa_mountpoint,
	nfs_is_shared_cb, &found);
	return (found);
	}
	diff --git a/sys/contrib/openzfs/lib/libspl/assert.c b/sys/contrib/openzfs/lib/libspl/assert.c
	index 207da3a1d860..9d44740d4e3c 100644
	--- a/sys/contrib/openzfs/lib/libspl/assert.c
	+++ b/sys/contrib/openzfs/lib/libspl/assert.c
	@@ -1,52 +1,55 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include <assert.h>

	static boolean_t libspl_assert_ok = B_FALSE;

	void
	libspl_set_assert_ok(boolean_t val)
	{
	libspl_assert_ok = val;
	}

	/* printf version of libspl_assert */
	void
	libspl_assertf(const char file, const char func, int line,
	const char *format, ...)
	{
	va_list args;

	va_start(args, format);
	vfprintf(stderr, format, args);
	fprintf(stderr, "\n");
	fprintf(stderr, "ASSERT at %s:%d:%s()", file, line, func);
	va_end(args);
	+
	+#if !__has_feature(attribute_analyzer_noreturn) && !defined(__COVERITY__)
	if (libspl_assert_ok) {
	return;
	}
	+#endif
	abort();
	}
	diff --git a/sys/contrib/openzfs/lib/libspl/include/assert.h b/sys/contrib/openzfs/lib/libspl/include/assert.h
	index aaaa0af096ef..c5bf0f0cc8f1 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/assert.h
	+++ b/sys/contrib/openzfs/lib/libspl/include/assert.h
	@@ -1,150 +1,162 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include_next <assert.h>

	#ifndef _LIBSPL_ASSERT_H
	#define _LIBSPL_ASSERT_H

	#include <stdio.h>
	#include <stdlib.h>
	#include <stdarg.h>
	#include <sys/types.h>

	+/* Workaround for non-Clang compilers */
	+#ifndef __has_feature
	+#define __has_feature(x) 0
	+#endif
	+
	+/* We need to workaround libspl_set_assert_ok() that we have for zdb */
	+#if __has_feature(attribute_analyzer_noreturn) \|\| defined(__COVERITY__)
	+#define NORETURN __attribute__((__noreturn__))
	+#else
	+#define NORETURN
	+#endif
	+
	/* Set to non-zero to avoid abort()ing on an assertion failure */
	extern void libspl_set_assert_ok(boolean_t val);

	/* printf version of libspl_assert */
	extern void libspl_assertf(const char file, const char func, int line,
	- const char *format, ...);
	+ const char *format, ...) NORETURN __attribute__((format(printf, 4, 5)));

	static inline int
	libspl_assert(const char buf, const char file, const char *func, int line)
	{
	libspl_assertf(file, func, line, "%s", buf);
	return (0);
	}

	#ifdef verify
	#undef verify
	#endif

	#define VERIFY(cond) \
	(void) ((!(cond)) && \
	libspl_assert(#cond, __FILE__, __FUNCTION__, __LINE__))
	#define verify(cond) \
	(void) ((!(cond)) && \
	libspl_assert(#cond, __FILE__, __FUNCTION__, __LINE__))

	#define VERIFY3B(LEFT, OP, RIGHT) \
	do { \
	const boolean_t __left = (boolean_t)(LEFT); \
	const boolean_t __right = (boolean_t)(RIGHT); \
	if (!(__left OP __right)) \
	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, \
	"%s %s %s (0x%llx %s 0x%llx)", #LEFT, #OP, #RIGHT, \
	(u_longlong_t)__left, #OP, (u_longlong_t)__right); \
	} while (0)

	#define VERIFY3S(LEFT, OP, RIGHT) \
	do { \
	const int64_t __left = (int64_t)(LEFT); \
	const int64_t __right = (int64_t)(RIGHT); \
	if (!(__left OP __right)) \
	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, \
	"%s %s %s (0x%llx %s 0x%llx)", #LEFT, #OP, #RIGHT, \
	(u_longlong_t)__left, #OP, (u_longlong_t)__right); \
	} while (0)

	#define VERIFY3U(LEFT, OP, RIGHT) \
	do { \
	const uint64_t __left = (uint64_t)(LEFT); \
	const uint64_t __right = (uint64_t)(RIGHT); \
	if (!(__left OP __right)) \
	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, \
	"%s %s %s (0x%llx %s 0x%llx)", #LEFT, #OP, #RIGHT, \
	(u_longlong_t)__left, #OP, (u_longlong_t)__right); \
	} while (0)

	#define VERIFY3P(LEFT, OP, RIGHT) \
	do { \
	const uintptr_t __left = (uintptr_t)(LEFT); \
	const uintptr_t __right = (uintptr_t)(RIGHT); \
	if (!(__left OP __right)) \
	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, \
	"%s %s %s (0x%llx %s 0x%llx)", #LEFT, #OP, #RIGHT, \
	(u_longlong_t)__left, #OP, (u_longlong_t)__right); \
	} while (0)

	#define VERIFY0(LEFT) \
	do { \
	const uint64_t __left = (uint64_t)(LEFT); \
	if (!(__left == 0)) \
	libspl_assertf(__FILE__, __FUNCTION__, __LINE__, \
	"%s == 0 (0x%llx == 0)", #LEFT, \
	(u_longlong_t)__left); \
	} while (0)

	#ifdef assert
	#undef assert
	#endif

	#ifdef NDEBUG
	#define ASSERT3B(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3S(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3U(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT3P(x, y, z) \
	((void) sizeof ((uintptr_t)(x)), (void) sizeof ((uintptr_t)(z)))
	#define ASSERT0(x) ((void) sizeof ((uintptr_t)(x)))
	#define ASSERT(x) ((void) sizeof ((uintptr_t)(x)))
	#define assert(x) ((void) sizeof ((uintptr_t)(x)))
	#define IMPLY(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))
	#define EQUIV(A, B) \
	((void) sizeof ((uintptr_t)(A)), (void) sizeof ((uintptr_t)(B)))
	#else
	#define ASSERT3B VERIFY3B
	#define ASSERT3S VERIFY3S
	#define ASSERT3U VERIFY3U
	#define ASSERT3P VERIFY3P
	#define ASSERT0 VERIFY0
	#define ASSERT VERIFY
	#define assert VERIFY
	#define IMPLY(A, B) \
	((void)(((!(A)) \|\| (B)) \|\| \
	libspl_assert("(" #A ") implies (" #B ")", \
	__FILE__, __FUNCTION__, __LINE__)))
	#define EQUIV(A, B) \
	((void)((!!(A) == !!(B)) \|\| \
	libspl_assert("(" #A ") is equivalent to (" #B ")", \
	__FILE__, __FUNCTION__, __LINE__)))

	#endif /* NDEBUG */

	#endif /* _LIBSPL_ASSERT_H */
	diff --git a/sys/contrib/openzfs/lib/libspl/include/os/freebsd/sys/byteorder.h b/sys/contrib/openzfs/lib/libspl/include/os/freebsd/sys/byteorder.h
	index 8de1104dc8a2..bd6728820ed4 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/os/freebsd/sys/byteorder.h
	+++ b/sys/contrib/openzfs/lib/libspl/include/os/freebsd/sys/byteorder.h
	@@ -1,192 +1,206 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */
	/* All Rights Reserved */

	/*
	* University Copyright- Copyright (c) 1982, 1986, 1988
	* The Regents of the University of California
	* All Rights Reserved
	*
	* University Acknowledgment- Portions of this document are derived from
	* software developed by the University of California, Berkeley, and its
	* contributors.
	*/

	#ifndef _SYS_BYTEORDER_H
	#define _SYS_BYTEORDER_H

	#include <sys/endian.h>
	#include <netinet/in.h>
	#include <sys/isa_defs.h>
	#include <inttypes.h>

	#if defined(__GNUC__) && defined(_ASM_INLINES) && \
	(defined(__i386) \|\| defined(__amd64))
	#include <asm/byteorder.h>
	#endif

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* macros for conversion between host and (internet) network byte order
	*/
	#if !defined(_XPG4_2) \|\| defined(__EXTENSIONS__)

	+#ifdef __COVERITY__
	+/*
	+ * Coverity's taint warnings from byteswapping are false positives for us.
	+ * Suppress them by hiding byteswapping from Coverity.
	+ */
	+#define BSWAP_8(x) ((x) & 0xff)
	+#define BSWAP_16(x) ((x) & 0xffff)
	+#define BSWAP_32(x) ((x) & 0xffffffff)
	+#define BSWAP_64(x) (x)
	+
	+#else /* __COVERITY__ */
	+
	/*
	* Macros to reverse byte order
	*/
	#define BSWAP_8(x) ((x) & 0xff)
	#define BSWAP_16(x) ((BSWAP_8(x) << 8) \| BSWAP_8((x) >> 8))
	#define BSWAP_32(x) ((BSWAP_16(x) << 16) \| BSWAP_16((x) >> 16))
	#define BSWAP_64(x) ((BSWAP_32(x) << 32) \| BSWAP_32((x) >> 32))

	+#endif /* __COVERITY__ */
	+
	#define BMASK_8(x) ((x) & 0xff)
	#define BMASK_16(x) ((x) & 0xffff)
	#define BMASK_32(x) ((x) & 0xffffffff)
	#define BMASK_64(x) (x)

	/*
	* Macros to convert from a specific byte order to/from native byte order
	*/
	#ifdef _ZFS_BIG_ENDIAN
	#define BE_8(x) BMASK_8(x)
	#define BE_16(x) BMASK_16(x)
	#define BE_32(x) BMASK_32(x)
	#define BE_64(x) BMASK_64(x)
	#define LE_8(x) BSWAP_8(x)
	#define LE_16(x) BSWAP_16(x)
	#define LE_32(x) BSWAP_32(x)
	#define LE_64(x) BSWAP_64(x)
	#else
	#define LE_8(x) BMASK_8(x)
	#define LE_16(x) BMASK_16(x)
	#define LE_32(x) BMASK_32(x)
	#define LE_64(x) BMASK_64(x)
	#define BE_8(x) BSWAP_8(x)
	#define BE_16(x) BSWAP_16(x)
	#define BE_32(x) BSWAP_32(x)
	#define BE_64(x) BSWAP_64(x)
	#endif

	#ifdef _ZFS_BIG_ENDIAN
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return (n);
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return (n);
	}
	#else
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return ((((uint64_t)htonl(n)) << 32) + htonl(n >> 32));
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return ((((uint64_t)ntohl(n)) << 32) + ntohl(n >> 32));
	}
	#endif

	/*
	* Macros to read unaligned values from a specific byte order to
	* native byte order
	*/

	#define BE_IN8(xa) \
	((uint8_t )(xa))

	#define BE_IN16(xa) \
	(((uint16_t)BE_IN8(xa) << 8) \| BE_IN8((uint8_t *)(xa)+1))

	#define BE_IN32(xa) \
	(((uint32_t)BE_IN16(xa) << 16) \| BE_IN16((uint8_t *)(xa)+2))

	#define BE_IN64(xa) \
	(((uint64_t)BE_IN32(xa) << 32) \| BE_IN32((uint8_t *)(xa)+4))

	#define LE_IN8(xa) \
	((uint8_t )(xa))

	#define LE_IN16(xa) \
	(((uint16_t)LE_IN8((uint8_t *)(xa) + 1) << 8) \| LE_IN8(xa))

	#define LE_IN32(xa) \
	(((uint32_t)LE_IN16((uint8_t *)(xa) + 2) << 16) \| LE_IN16(xa))

	#define LE_IN64(xa) \
	(((uint64_t)LE_IN32((uint8_t *)(xa) + 4) << 32) \| LE_IN32(xa))

	/*
	* Macros to write unaligned values from native byte order to a specific byte
	* order.
	*/

	#define BE_OUT8(xa, yv) ((uint8_t )(xa)) = (uint8_t)(yv);

	#define BE_OUT16(xa, yv) \
	BE_OUT8((uint8_t *)(xa) + 1, yv); \
	BE_OUT8((uint8_t *)(xa), (yv) >> 8);

	#define BE_OUT32(xa, yv) \
	BE_OUT16((uint8_t *)(xa) + 2, yv); \
	BE_OUT16((uint8_t *)(xa), (yv) >> 16);

	#define BE_OUT64(xa, yv) \
	BE_OUT32((uint8_t *)(xa) + 4, yv); \
	BE_OUT32((uint8_t *)(xa), (yv) >> 32);

	#define LE_OUT8(xa, yv) ((uint8_t )(xa)) = (uint8_t)(yv);

	#define LE_OUT16(xa, yv) \
	LE_OUT8((uint8_t *)(xa), yv); \
	LE_OUT8((uint8_t *)(xa) + 1, (yv) >> 8);

	#define LE_OUT32(xa, yv) \
	LE_OUT16((uint8_t *)(xa), yv); \
	LE_OUT16((uint8_t *)(xa) + 2, (yv) >> 16);

	#define LE_OUT64(xa, yv) \
	LE_OUT32((uint8_t *)(xa), yv); \
	LE_OUT32((uint8_t *)(xa) + 4, (yv) >> 32);

	#endif /* !defined(_XPG4_2) \|\| defined(__EXTENSIONS__) */

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_BYTEORDER_H */
	diff --git a/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/byteorder.h b/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/byteorder.h
	index c8413ea76352..50f0f1e77936 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/byteorder.h
	+++ b/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/byteorder.h
	@@ -1,223 +1,237 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */
	/* All Rights Reserved */

	/*
	* University Copyright- Copyright (c) 1982, 1986, 1988
	* The Regents of the University of California
	* All Rights Reserved
	*
	* University Acknowledgment- Portions of this document are derived from
	* software developed by the University of California, Berkeley, and its
	* contributors.
	*/

	#ifndef _SYS_BYTEORDER_H
	#define _SYS_BYTEORDER_H

	#if defined(__GNUC__) && defined(_ASM_INLINES) && \
	(defined(__i386) \|\| defined(__amd64))
	#include <asm/byteorder.h>
	#endif

	#include <sys/isa_defs.h>
	#include <inttypes.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* macros for conversion between host and (internet) network byte order
	*/

	#if defined(_ZFS_BIG_ENDIAN) && !defined(ntohl) && !defined(__lint)
	/* big-endian */
	#define ntohl(x) (x)
	#define ntohs(x) (x)
	#define htonl(x) (x)
	#define htons(x) (x)

	#elif !defined(ntohl) /* little-endian */

	#ifndef _IN_PORT_T
	#define _IN_PORT_T
	typedef uint16_t in_port_t;
	#endif

	#ifndef _IN_ADDR_T
	#define _IN_ADDR_T
	typedef uint32_t in_addr_t;
	#endif

	#if !defined(_XPG4_2) \|\| defined(__EXTENSIONS__) \|\| defined(_XPG5)
	extern uint32_t htonl(uint32_t);
	extern uint16_t htons(uint16_t);
	extern uint32_t ntohl(uint32_t);
	extern uint16_t ntohs(uint16_t);
	#else
	extern in_addr_t htonl(in_addr_t);
	extern in_port_t htons(in_port_t);
	extern in_addr_t ntohl(in_addr_t);
	extern in_port_t ntohs(in_port_t);
	#endif /* !defined(_XPG4_2) \|\| defined(__EXTENSIONS__) \|\| defined(_XPG5) */
	#endif

	#if !defined(_XPG4_2) \|\| defined(__EXTENSIONS__)

	+#ifdef __COVERITY__
	+/*
	+ * Coverity's taint warnings from byteswapping are false positives for us.
	+ * Suppress them by hiding byteswapping from Coverity.
	+ */
	+#define BSWAP_8(x) ((x) & 0xff)
	+#define BSWAP_16(x) ((x) & 0xffff)
	+#define BSWAP_32(x) ((x) & 0xffffffff)
	+#define BSWAP_64(x) (x)
	+
	+#else /* __COVERITY__ */
	+
	/*
	* Macros to reverse byte order
	*/
	#define BSWAP_8(x) ((x) & 0xff)
	#define BSWAP_16(x) ((BSWAP_8(x) << 8) \| BSWAP_8((x) >> 8))
	#define BSWAP_32(x) ((BSWAP_16(x) << 16) \| BSWAP_16((x) >> 16))
	#define BSWAP_64(x) ((BSWAP_32(x) << 32) \| BSWAP_32((x) >> 32))

	+#endif /* __COVERITY__ */
	+
	#define BMASK_8(x) ((x) & 0xff)
	#define BMASK_16(x) ((x) & 0xffff)
	#define BMASK_32(x) ((x) & 0xffffffff)
	#define BMASK_64(x) (x)

	/*
	* Macros to convert from a specific byte order to/from native byte order
	*/
	#ifdef _ZFS_BIG_ENDIAN
	#define BE_8(x) BMASK_8(x)
	#define BE_16(x) BMASK_16(x)
	#define BE_32(x) BMASK_32(x)
	#define BE_64(x) BMASK_64(x)
	#define LE_8(x) BSWAP_8(x)
	#define LE_16(x) BSWAP_16(x)
	#define LE_32(x) BSWAP_32(x)
	#define LE_64(x) BSWAP_64(x)
	#else
	#define LE_8(x) BMASK_8(x)
	#define LE_16(x) BMASK_16(x)
	#define LE_32(x) BMASK_32(x)
	#define LE_64(x) BMASK_64(x)
	#define BE_8(x) BSWAP_8(x)
	#define BE_16(x) BSWAP_16(x)
	#define BE_32(x) BSWAP_32(x)
	#define BE_64(x) BSWAP_64(x)
	#endif

	#ifdef _ZFS_BIG_ENDIAN
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return (n);
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return (n);
	}
	#else
	static __inline__ uint64_t
	htonll(uint64_t n)
	{
	return ((((uint64_t)htonl(n)) << 32) + htonl(n >> 32));
	}

	static __inline__ uint64_t
	ntohll(uint64_t n)
	{
	return ((((uint64_t)ntohl(n)) << 32) + ntohl(n >> 32));
	}
	#endif

	/*
	* Macros to read unaligned values from a specific byte order to
	* native byte order
	*/

	#define BE_IN8(xa) \
	((uint8_t )(xa))

	#define BE_IN16(xa) \
	(((uint16_t)BE_IN8(xa) << 8) \| BE_IN8((uint8_t *)(xa)+1))

	#define BE_IN32(xa) \
	(((uint32_t)BE_IN16(xa) << 16) \| BE_IN16((uint8_t *)(xa)+2))

	#define BE_IN64(xa) \
	(((uint64_t)BE_IN32(xa) << 32) \| BE_IN32((uint8_t *)(xa)+4))

	#define LE_IN8(xa) \
	((uint8_t )(xa))

	#define LE_IN16(xa) \
	(((uint16_t)LE_IN8((uint8_t *)(xa) + 1) << 8) \| LE_IN8(xa))

	#define LE_IN32(xa) \
	(((uint32_t)LE_IN16((uint8_t *)(xa) + 2) << 16) \| LE_IN16(xa))

	#define LE_IN64(xa) \
	(((uint64_t)LE_IN32((uint8_t *)(xa) + 4) << 32) \| LE_IN32(xa))

	/*
	* Macros to write unaligned values from native byte order to a specific byte
	* order.
	*/

	#define BE_OUT8(xa, yv) ((uint8_t )(xa)) = (uint8_t)(yv);

	#define BE_OUT16(xa, yv) \
	BE_OUT8((uint8_t *)(xa) + 1, yv); \
	BE_OUT8((uint8_t *)(xa), (yv) >> 8);

	#define BE_OUT32(xa, yv) \
	BE_OUT16((uint8_t *)(xa) + 2, yv); \
	BE_OUT16((uint8_t *)(xa), (yv) >> 16);

	#define BE_OUT64(xa, yv) \
	BE_OUT32((uint8_t *)(xa) + 4, yv); \
	BE_OUT32((uint8_t *)(xa), (yv) >> 32);

	#define LE_OUT8(xa, yv) ((uint8_t )(xa)) = (uint8_t)(yv);

	#define LE_OUT16(xa, yv) \
	LE_OUT8((uint8_t *)(xa), yv); \
	LE_OUT8((uint8_t *)(xa) + 1, (yv) >> 8);

	#define LE_OUT32(xa, yv) \
	LE_OUT16((uint8_t *)(xa), yv); \
	LE_OUT16((uint8_t *)(xa) + 2, (yv) >> 16);

	#define LE_OUT64(xa, yv) \
	LE_OUT32((uint8_t *)(xa), yv); \
	LE_OUT32((uint8_t *)(xa) + 4, (yv) >> 32);

	#endif /* !defined(_XPG4_2) \|\| defined(__EXTENSIONS__) */

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_BYTEORDER_H */
	diff --git a/sys/contrib/openzfs/lib/libspl/include/umem.h b/sys/contrib/openzfs/lib/libspl/include/umem.h
	index 2a366e29472d..82976f756b73 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/umem.h
	+++ b/sys/contrib/openzfs/lib/libspl/include/umem.h
	@@ -1,208 +1,211 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _LIBSPL_UMEM_H
	#define _LIBSPL_UMEM_H

	/*
	* XXX: We should use the real portable umem library if it is detected
	* at configure time. However, if the library is not available, we can
	* use a trivial malloc based implementation. This obviously impacts
	* performance, but unless you are using a full userspace build of zpool for
	* something other than ztest, you are likely not going to notice or care.
	*
	* https://labs.omniti.com/trac/portableumem
	*/
	#include <sys/debug.h>

	#include <stdlib.h>
	#include <stdio.h>
	#include <string.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	typedef void vmem_t;

	/*
	* Flags for umem_alloc/umem_free
	*/
	#define UMEM_DEFAULT 0x0000 /* normal -- may fail */
	#define UMEM_NOFAIL 0x0100 /* Never fails */

	/*
	* Flags for umem_cache_create()
	*/
	#define UMC_NODEBUG 0x00020000

	#define UMEM_CACHE_NAMELEN 31

	typedef int umem_nofail_callback_t(void);
	typedef int umem_constructor_t(void , void , int);
	typedef void umem_destructor_t(void , void );
	typedef void umem_reclaim_t(void *);

	typedef struct umem_cache {
	char cache_name[UMEM_CACHE_NAMELEN + 1];
	size_t cache_bufsize;
	size_t cache_align;
	umem_constructor_t *cache_constructor;
	umem_destructor_t *cache_destructor;
	umem_reclaim_t *cache_reclaim;
	void *cache_private;
	void *cache_arena;
	int cache_cflags;
	} umem_cache_t;

	/* Prototypes for functions to provide defaults for umem envvars */
	const char *_umem_debug_init(void);
	const char *_umem_options_init(void);
	const char *_umem_logging_init(void);

	+__attribute__((alloc_size(1)))
	static inline void *
	umem_alloc(size_t size, int flags)
	{
	void *ptr = NULL;

	do {
	ptr = malloc(size);
	} while (ptr == NULL && (flags & UMEM_NOFAIL));

	return (ptr);
	}

	+__attribute__((alloc_size(1)))
	static inline void *
	umem_alloc_aligned(size_t size, size_t align, int flags)
	{
	void *ptr = NULL;
	int rc = EINVAL;

	do {
	rc = posix_memalign(&ptr, align, size);
	} while (rc == ENOMEM && (flags & UMEM_NOFAIL));

	if (rc == EINVAL) {
	fprintf(stderr, "%s: invalid memory alignment (%zd)\n",
	__func__, align);
	if (flags & UMEM_NOFAIL)
	abort();
	return (NULL);
	}

	return (ptr);
	}

	+__attribute__((alloc_size(1)))
	static inline void *
	umem_zalloc(size_t size, int flags)
	{
	void *ptr = NULL;

	ptr = umem_alloc(size, flags);
	if (ptr)
	memset(ptr, 0, size);

	return (ptr);
	}

	static inline void
	umem_free(const void *ptr, size_t size __maybe_unused)
	{
	free((void *)ptr);
	}

	static inline void
	umem_nofail_callback(umem_nofail_callback_t *cb __maybe_unused)
	{}

	static inline umem_cache_t *
	umem_cache_create(
	const char *name, size_t bufsize, size_t align,
	umem_constructor_t *constructor,
	umem_destructor_t *destructor,
	umem_reclaim_t *reclaim,
	void priv, void vmp, int cflags)
	{
	umem_cache_t *cp;

	cp = (umem_cache_t *)umem_alloc(sizeof (umem_cache_t), UMEM_DEFAULT);
	if (cp) {
	strlcpy(cp->cache_name, name, UMEM_CACHE_NAMELEN);
	cp->cache_bufsize = bufsize;
	cp->cache_align = align;
	cp->cache_constructor = constructor;
	cp->cache_destructor = destructor;
	cp->cache_reclaim = reclaim;
	cp->cache_private = priv;
	cp->cache_arena = vmp;
	cp->cache_cflags = cflags;
	}

	return (cp);
	}

	static inline void
	umem_cache_destroy(umem_cache_t *cp)
	{
	umem_free(cp, sizeof (umem_cache_t));
	}

	static inline void *
	umem_cache_alloc(umem_cache_t *cp, int flags)
	{
	void *ptr = NULL;

	if (cp->cache_align != 0)
	ptr = umem_alloc_aligned(
	cp->cache_bufsize, cp->cache_align, flags);
	else
	ptr = umem_alloc(cp->cache_bufsize, flags);

	if (ptr && cp->cache_constructor)
	cp->cache_constructor(ptr, cp->cache_private, UMEM_DEFAULT);

	return (ptr);
	}

	static inline void
	umem_cache_free(umem_cache_t cp, void ptr)
	{
	if (cp->cache_destructor)
	cp->cache_destructor(ptr, cp->cache_private);

	umem_free(ptr, cp->cache_bufsize);
	}

	static inline void
	umem_cache_reap_now(umem_cache_t *cp __maybe_unused)
	{
	}

	#ifdef __cplusplus
	}
	#endif

	#endif
	diff --git a/sys/contrib/openzfs/lib/libspl/os/freebsd/mnttab.c b/sys/contrib/openzfs/lib/libspl/os/freebsd/mnttab.c
	index f0cc04d89ded..a4673084ad5f 100644
	--- a/sys/contrib/openzfs/lib/libspl/os/freebsd/mnttab.c
	+++ b/sys/contrib/openzfs/lib/libspl/os/freebsd/mnttab.c
	@@ -1,237 +1,237 @@
	/*
	* Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*/

	/*
	* This file implements Solaris compatible getmntany() and hasmntopt()
	* functions.
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");

	#include <sys/param.h>
	#include <sys/mount.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>

	#include <ctype.h>
	#include <errno.h>
	#include <pthread.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>

	static char *
	mntopt(char **p)
	{
	char cp = p;
	char *retstr;

	while (cp && isspace(cp))
	cp++;

	retstr = cp;
	while (cp && cp != ',')
	cp++;

	if (*cp) {
	*cp = '\0';
	cp++;
	}

	*p = cp;
	return (retstr);
	}

	char *
	hasmntopt(struct mnttab mnt, const char opt)
	{
	char tmpopts[MNT_LINE_MAX];
	char f, opts = tmpopts;

	if (mnt->mnt_mntopts == NULL)
	return (NULL);
	- (void) strcpy(opts, mnt->mnt_mntopts);
	+ (void) strlcpy(opts, mnt->mnt_mntopts, MNT_LINE_MAX);
	f = mntopt(&opts);
	for (; *f; f = mntopt(&opts)) {
	if (strncmp(opt, f, strlen(opt)) == 0)
	return (f - tmpopts + mnt->mnt_mntopts);
	}
	return (NULL);
	}

	static void
	optadd(char mntopts, size_t size, const char opt)
	{

	if (mntopts[0] != '\0')
	strlcat(mntopts, ",", size);
	strlcat(mntopts, opt, size);
	}

	static __thread char gfstypename[MFSNAMELEN];
	static __thread char gmntfromname[MNAMELEN];
	static __thread char gmntonname[MNAMELEN];
	static __thread char gmntopts[MNTMAXSTR];

	void
	statfs2mnttab(struct statfs sfs, struct mnttab mp)
	{
	long flags;

	strlcpy(gfstypename, sfs->f_fstypename, sizeof (gfstypename));
	mp->mnt_fstype = gfstypename;

	strlcpy(gmntfromname, sfs->f_mntfromname, sizeof (gmntfromname));
	mp->mnt_special = gmntfromname;

	strlcpy(gmntonname, sfs->f_mntonname, sizeof (gmntonname));
	mp->mnt_mountp = gmntonname;

	flags = sfs->f_flags;
	gmntopts[0] = '\0';
	#define OPTADD(opt) optadd(gmntopts, sizeof (gmntopts), (opt))
	if (flags & MNT_RDONLY)
	OPTADD(MNTOPT_RO);
	else
	OPTADD(MNTOPT_RW);
	if (flags & MNT_NOSUID)
	OPTADD(MNTOPT_NOSETUID);
	else
	OPTADD(MNTOPT_SETUID);
	if (flags & MNT_UPDATE)
	OPTADD(MNTOPT_REMOUNT);
	if (flags & MNT_NOATIME)
	OPTADD(MNTOPT_NOATIME);
	else
	OPTADD(MNTOPT_ATIME);
	OPTADD(MNTOPT_NOXATTR);
	if (flags & MNT_NOEXEC)
	OPTADD(MNTOPT_NOEXEC);
	else
	OPTADD(MNTOPT_EXEC);
	#undef OPTADD
	mp->mnt_mntopts = gmntopts;
	}

	static pthread_rwlock_t gsfs_lock = PTHREAD_RWLOCK_INITIALIZER;
	static struct statfs *gsfs = NULL;
	static int allfs = 0;

	static int
	statfs_init(void)
	{
	struct statfs *sfs;
	int error;

	(void) pthread_rwlock_wrlock(&gsfs_lock);

	if (gsfs != NULL) {
	free(gsfs);
	gsfs = NULL;
	}
	allfs = getfsstat(NULL, 0, MNT_NOWAIT);
	if (allfs == -1)
	goto fail;
	gsfs = malloc(sizeof (gsfs[0]) * allfs * 2);
	if (gsfs == NULL)
	goto fail;
	allfs = getfsstat(gsfs, (long)(sizeof (gsfs[0]) * allfs * 2),
	MNT_NOWAIT);
	if (allfs == -1)
	goto fail;
	sfs = realloc(gsfs, allfs * sizeof (gsfs[0]));
	if (sfs != NULL)
	gsfs = sfs;
	(void) pthread_rwlock_unlock(&gsfs_lock);
	return (0);
	fail:
	error = errno;
	if (gsfs != NULL)
	free(gsfs);
	gsfs = NULL;
	allfs = 0;
	(void) pthread_rwlock_unlock(&gsfs_lock);
	return (error);
	}

	int
	getmntany(FILE fd __unused, struct mnttab mgetp, struct mnttab *mrefp)
	{
	int i, error;

	error = statfs_init();
	if (error != 0)
	return (error);

	(void) pthread_rwlock_rdlock(&gsfs_lock);

	for (i = 0; i < allfs; i++) {
	if (mrefp->mnt_special != NULL &&
	strcmp(mrefp->mnt_special, gsfs[i].f_mntfromname) != 0) {
	continue;
	}
	if (mrefp->mnt_mountp != NULL &&
	strcmp(mrefp->mnt_mountp, gsfs[i].f_mntonname) != 0) {
	continue;
	}
	if (mrefp->mnt_fstype != NULL &&
	strcmp(mrefp->mnt_fstype, gsfs[i].f_fstypename) != 0) {
	continue;
	}
	statfs2mnttab(&gsfs[i], mgetp);
	(void) pthread_rwlock_unlock(&gsfs_lock);
	return (0);
	}
	(void) pthread_rwlock_unlock(&gsfs_lock);
	return (-1);
	}

	int
	getmntent(FILE fp, struct mnttab mp)
	{
	int error, nfs;

	nfs = (int)lseek(fileno(fp), 0, SEEK_CUR);
	if (nfs == -1)
	return (errno);
	/* If nfs is 0, we want to refresh out cache. */
	if (nfs == 0 \|\| gsfs == NULL) {
	error = statfs_init();
	if (error != 0)
	return (error);
	}
	(void) pthread_rwlock_rdlock(&gsfs_lock);
	if (nfs >= allfs) {
	(void) pthread_rwlock_unlock(&gsfs_lock);
	return (-1);
	}
	statfs2mnttab(&gsfs[nfs], mp);
	(void) pthread_rwlock_unlock(&gsfs_lock);
	if (lseek(fileno(fp), 1, SEEK_CUR) == -1)
	return (errno);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs.abi b/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	index 7dd12df81718..3471bcac9412 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	@@ -1,5852 +1,5858 @@
	<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs.so.4'>
	<elf-needed>
	<dependency name='libzfs_core.so.3'/>
	<dependency name='libnvpair.so.3'/>
	<dependency name='libuuid.so.1'/>
	<dependency name='librt.so.1'/>
	<dependency name='libblkid.so.1'/>
	<dependency name='libudev.so.1'/>
	<dependency name='libuutil.so.3'/>
	<dependency name='libm.so.6'/>
	<dependency name='libcrypto.so.1.1'/>
	<dependency name='libz.so.1'/>
	<dependency name='libdl.so.2'/>
	<dependency name='libpthread.so.0'/>
	<dependency name='libc.so.6'/>
	<dependency name='ld-linux-x86-64.so.2'/>
	</elf-needed>
	<elf-function-symbols>
	<elf-symbol name='_sol_getmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_clear_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_set_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy_nodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert_here' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_nearest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_swap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_gt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_lt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='bookmark_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='cityhash4' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='color_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='color_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='dataset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='dataset_nestcheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_alloc_and_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_alloc_and_read' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_err_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_rescan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_use_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_write' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='entity_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_impl_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_native_varsize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_dataset_depth' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_system_hostid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getexecname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getextmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getmntany' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getprop_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getzoneid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='is_mpath_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_set_assert_ok' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_add_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_envvar_is_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_errno' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_action' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_description' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_free_str_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_cache' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_print_on_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process_get_stdout' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process_get_stdout_nopath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_active' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_replace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_move_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_consumer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_enter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_exit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_producer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mkdirp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mountpoint_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='permset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='pool_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='print_timestamp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='printf_color' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_disable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_enable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_errorstr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_truncate_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_validate_shareopts' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='snapshot_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spl_pagesize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcpy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_abandon' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_dispatch' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_member' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_suspend' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_suspended' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='update_vdev_config_dev_strs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_depends_on' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_is_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_is_valid_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_lookup_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_lookup_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_adjust_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_allocatable_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_append_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_basename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_bookmark_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_clone' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_component_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_create_ancestors' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_attempt_load_keys' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_clone_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_get_encryption_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_load_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_rewrap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_unload_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dataset_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dataset_name_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_canonicalize_perm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_verify_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_whokey' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps_nvl_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_flush' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_is_dm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_is_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_device_get_devid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_device_get_physical' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dirnamelen' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_foreach_mountpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_all_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_clones_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_enclosure_sysfs_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_holds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_pool_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_pool_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_recvd_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_underlying_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_underlying_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_user_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_handle_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_hold' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_hold_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_ioctl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_isnumber' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_bookmarks' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_children' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_dependents' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_filesystems' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots_sorted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapspec' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mod_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount_at' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount_delegation_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_name_valid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicebytes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicenum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicenum_format' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_niceraw' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicestrtonum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicetime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_parent_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_parse_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_path_to_zhandle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_promote' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_delegatable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_encryption_key_param' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_recvd' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_userquota_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_written_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_inherit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_inheritable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_is_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_set_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_valid_keylocation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_visible' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prune_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_receive' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_refresh_properties' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_resolve_shortname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_rollback' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_save_arguments' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_progress' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_resume_token_to_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_saved' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_set_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_show_diffs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_purge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_snapshot' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_snapshot_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_spa_version' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_spa_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_special_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_standard_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strcmp_pathname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strip_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strip_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_truncate_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_type_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unmount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unmountall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unshare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unshareall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userns' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userspace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_valid_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_kernel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_print' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_userland' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_zpl_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_clear_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_default_search_paths' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_datasets_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_volume_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_discard_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_dump_ddt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_enable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_seek' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_explain_recover' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_export' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_export_force' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_feature_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_vdev_by_physpath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_free_handles' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_all_vdev_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_errlog' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_features' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_load_policy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_prop_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_state' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_state_str' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_vdev_prop_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_history_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_in_use' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_initialize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_initialize_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_is_draid_spare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_label_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_label_disk_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_load_compat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_log_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_obj_to_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_obj_to_path_ds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_open_canfail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_pool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_print_unsup_feat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_unsupported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_props_refresh' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_read_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_refresh_stats' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_reguid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_reopen_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_scan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_search_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_skip_pool' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_sync_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_trim' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_upgrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_attach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_degrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_detach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_fault' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_indirect_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_offline' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_online' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_path_to_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_remove_cancel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	+ <elf-symbol name='zpool_vdev_remove_wanted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_split' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_free_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_get_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_iter_common' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_print_one_property' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_impl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_number' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_valid_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_width' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zvol_volsize_to_reservation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-function-symbols>
	<elf-variable-symbols>
	<elf-symbol name='efi_debug' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_abd_ops' size='24' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx2_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx512bw_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx512f_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_sse2_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_ssse3_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_superscalar4_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_superscalar_ops' size='64' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_config_ops' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_protocol_names' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spa_feature_table' size='2072' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_checks_disable' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_perm_tab' size='512' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_history_event_names' size='328' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_max_dataset_nesting' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userquota_prop_prefixes' size='96' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-variable-symbols>
	<abi-instr address-size='64' path='lib/libefi/rdwr_efi.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='288' id='16e6f2c6'>
	<subrange length='36' type-id='7359adad' id='ae666bde'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a65ae39c' size-in-bits='960' id='fa198beb'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='48' id='0f562bd0'>
	<subrange length='6' type-id='7359adad' id='52fa524b'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='384' id='dba89ba3'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='256' id='01d84ed4'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<class-decl name='dk_part' size-in-bits='960' is-struct='yes' visibility='default' id='a65ae39c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='p_start' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='p_size' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='p_guid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='p_tag' type-id='d908a348' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='272'>
	<var-decl name='p_flag' type-id='d908a348' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='p_name' type-id='16e6f2c6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='p_uguid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='p_resv' type-id='01d84ed4' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='dk_gpt' size-in-bits='1920' is-struct='yes' visibility='default' id='dd4a2e5a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='efi_version' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='efi_nparts' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='efi_part_size' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='efi_lbasize' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='efi_last_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='efi_first_u_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='efi_last_u_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='efi_disk_uguid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='efi_flags' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='efi_reserved1' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='efi_altern_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='efi_reserved' type-id='dba89ba3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='efi_parts' type-id='fa198beb' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uuid' size-in-bits='128' is-struct='yes' visibility='default' id='214f32ea'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='time_low' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='time_mid' type-id='149c6638' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='time_hi_and_version' type-id='149c6638' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='clock_seq_hi_and_reserved' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='72'>
	<var-decl name='clock_seq_low' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='80'>
	<var-decl name='node_addr' type-id='0f562bd0' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ushort_t' type-id='8efea9e5' id='d908a348'/>
	<pointer-type-def type-id='dd4a2e5a' size-in-bits='64' id='0d8119a8'/>
	<pointer-type-def type-id='0d8119a8' size-in-bits='64' id='c43b27a6'/>
	<var-decl name='efi_debug' type-id='95e97e5e' mangled-name='efi_debug' visibility='default' elf-symbol-id='efi_debug'/>
	<function-decl name='efi_alloc_and_init' mangled-name='efi_alloc_and_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_init'>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='8f92235e' name='nparts'/>
	<parameter type-id='c43b27a6' name='vtoc'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_alloc_and_read' mangled-name='efi_alloc_and_read' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_read'>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='c43b27a6' name='vtoc'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_rescan' mangled-name='efi_rescan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_rescan'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_use_whole_disk' mangled-name='efi_use_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_use_whole_disk'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_write' mangled-name='efi_write' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_write'>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='0d8119a8' name='vtoc'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_free' mangled-name='efi_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_free'>
	<parameter type-id='0d8119a8' name='ptr'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='efi_err_check' mangled-name='efi_err_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_err_check'>
	<parameter type-id='0d8119a8' name='vtoc'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/libshare.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='128' id='2d6895a3'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<var-decl name='sa_protocol_names' type-id='2d6895a3' mangled-name='sa_protocol_names' visibility='default' elf-symbol-id='sa_protocol_names'/>
	<function-decl name='sa_enable_share' mangled-name='sa_enable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_enable_share'>
	<parameter type-id='80f4b756' name='zfsname'/>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<parameter type-id='80f4b756' name='shareopts'/>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_disable_share' mangled-name='sa_disable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_disable_share'>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_is_shared' mangled-name='sa_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_is_shared'>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='sa_commit_shares' mangled-name='sa_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_commit_shares'>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='sa_truncate_shares' mangled-name='sa_truncate_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_truncate_shares'>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='sa_validate_shareopts' mangled-name='sa_validate_shareopts' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_validate_shareopts'>
	<parameter type-id='80f4b756' name='options'/>
	<parameter type-id='9155d4b5' name='protocol'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_errorstr' mangled-name='sa_errorstr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_errorstr'>
	<parameter type-id='95e97e5e' name='err'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<enum-decl name='sa_protocol' id='9155d4b5'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='SA_PROTOCOL_NFS' value='0'/>
	<enumerator name='SA_PROTOCOL_SMB' value='1'/>
	<enumerator name='SA_PROTOCOL_COUNT' value='2'/>
	</enum-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/os/linux/nfs.c' language='LANG_C99'>
	<class-decl name='sa_share_impl' size-in-bits='192' is-struct='yes' visibility='default' id='72b09bf8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='sa_zfsname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='sa_mountpoint' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='sa_shareopts' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sa_share_impl_t' type-id='946a2c6b' id='a48b47d0'/>
	<class-decl name='sa_fstype_t' size-in-bits='384' is-struct='yes' naming-typedef-id='639af739' visibility='default' id='944afa86'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='enable_share' type-id='2f78a9c1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='disable_share' type-id='2f78a9c1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='is_shared' type-id='81020bc2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='validate_shareopts' type-id='f194a8fb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='commit_shares' type-id='797ee7da' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='truncate_shares' type-id='5d51038b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sa_fstype_t' type-id='944afa86' id='639af739'/>
	<qualified-type-def type-id='639af739' const='yes' id='d19dbca9'/>
	<qualified-type-def type-id='72b09bf8' const='yes' id='484950e3'/>
	<pointer-type-def type-id='484950e3' size-in-bits='64' id='946a2c6b'/>
	<pointer-type-def type-id='276427e1' size-in-bits='64' id='1db260e5'/>
	<qualified-type-def type-id='1db260e5' const='yes' id='797ee7da'/>
	<pointer-type-def type-id='5113b296' size-in-bits='64' id='70487b28'/>
	<qualified-type-def type-id='70487b28' const='yes' id='f194a8fb'/>
	<pointer-type-def type-id='c13578bc' size-in-bits='64' id='fa1f29ce'/>
	<qualified-type-def type-id='fa1f29ce' const='yes' id='2f78a9c1'/>
	<pointer-type-def type-id='86373eb1' size-in-bits='64' id='f337456d'/>
	<qualified-type-def type-id='f337456d' const='yes' id='81020bc2'/>
	<pointer-type-def type-id='ee076206' size-in-bits='64' id='953b12f8'/>
	<qualified-type-def type-id='953b12f8' const='yes' id='5d51038b'/>
	<var-decl name='libshare_nfs_type' type-id='d19dbca9' visibility='default'/>
	<function-type size-in-bits='64' id='276427e1'>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='5113b296'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='c13578bc'>
	<parameter type-id='a48b47d0'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='86373eb1'>
	<parameter type-id='a48b47d0'/>
	<return type-id='c19b74c3'/>
	</function-type>
	<function-type size-in-bits='64' id='ee076206'>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/os/linux/smb.c' language='LANG_C99'>
	<var-decl name='libshare_smb_type' type-id='d19dbca9' visibility='default'/>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/assert.c' language='LANG_C99'>
	<function-decl name='libspl_set_assert_ok' mangled-name='libspl_set_assert_ok' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_set_assert_ok'>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
	<parameter type-id='80f4b756' name='file'/>
	<parameter type-id='80f4b756' name='func'/>
	<parameter type-id='95e97e5e' name='line'/>
	<parameter type-id='80f4b756' name='format'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/atomic.c' language='LANG_C99'>
	<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
	<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
	<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
	<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
	<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
	<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
	<qualified-type-def type-id='149c6638' volatile='yes' id='5120c5f7'/>
	<pointer-type-def type-id='5120c5f7' size-in-bits='64' id='93977ae7'/>
	<qualified-type-def type-id='8f92235e' volatile='yes' id='430e0681'/>
	<pointer-type-def type-id='430e0681' size-in-bits='64' id='3a147f31'/>
	<qualified-type-def type-id='b96825af' volatile='yes' id='84ff7d66'/>
	<pointer-type-def type-id='84ff7d66' size-in-bits='64' id='aa323ea4'/>
	<qualified-type-def type-id='ee1f298e' volatile='yes' id='6f7e09cb'/>
	<pointer-type-def type-id='6f7e09cb' size-in-bits='64' id='64698d33'/>
	<qualified-type-def type-id='48b5725f' volatile='yes' id='b0b3cbf9'/>
	<pointer-type-def type-id='b0b3cbf9' size-in-bits='64' id='fe09dd29'/>
	<function-decl name='atomic_inc_8' mangled-name='atomic_inc_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_16' mangled-name='atomic_inc_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_32' mangled-name='atomic_inc_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong' mangled-name='atomic_inc_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_8' mangled-name='atomic_dec_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_16' mangled-name='atomic_dec_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_32' mangled-name='atomic_dec_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong' mangled-name='atomic_dec_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_ptr' mangled-name='atomic_add_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_8' mangled-name='atomic_add_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_16' mangled-name='atomic_add_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_32' mangled-name='atomic_add_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_long' mangled-name='atomic_add_long' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr' mangled-name='atomic_sub_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_8' mangled-name='atomic_sub_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_16' mangled-name='atomic_sub_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_32' mangled-name='atomic_sub_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_long' mangled-name='atomic_sub_long' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_8' mangled-name='atomic_or_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_16' mangled-name='atomic_or_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_32' mangled-name='atomic_or_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_ulong' mangled-name='atomic_or_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_8' mangled-name='atomic_and_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_16' mangled-name='atomic_and_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_32' mangled-name='atomic_and_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_ulong' mangled-name='atomic_and_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_8_nv' mangled-name='atomic_inc_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_inc_16_nv' mangled-name='atomic_inc_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_inc_32_nv' mangled-name='atomic_inc_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong_nv' mangled-name='atomic_inc_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_dec_8_nv' mangled-name='atomic_dec_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_dec_16_nv' mangled-name='atomic_dec_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_dec_32_nv' mangled-name='atomic_dec_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong_nv' mangled-name='atomic_dec_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_add_ptr_nv' mangled-name='atomic_add_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_add_8_nv' mangled-name='atomic_add_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_add_16_nv' mangled-name='atomic_add_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_add_32_nv' mangled-name='atomic_add_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_add_long_nv' mangled-name='atomic_add_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr_nv' mangled-name='atomic_sub_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_sub_8_nv' mangled-name='atomic_sub_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_sub_16_nv' mangled-name='atomic_sub_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_sub_32_nv' mangled-name='atomic_sub_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_sub_long_nv' mangled-name='atomic_sub_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_or_8_nv' mangled-name='atomic_or_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_or_16_nv' mangled-name='atomic_or_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_or_32_nv' mangled-name='atomic_or_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_or_ulong_nv' mangled-name='atomic_or_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_and_8_nv' mangled-name='atomic_and_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_and_16_nv' mangled-name='atomic_and_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_and_32_nv' mangled-name='atomic_and_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_and_ulong_nv' mangled-name='atomic_and_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_cas_ptr' mangled-name='atomic_cas_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='exp'/>
	<parameter type-id='eaa32e2f' name='des'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_cas_8' mangled-name='atomic_cas_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='exp'/>
	<parameter type-id='b96825af' name='des'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_cas_16' mangled-name='atomic_cas_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='exp'/>
	<parameter type-id='149c6638' name='des'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_cas_32' mangled-name='atomic_cas_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='exp'/>
	<parameter type-id='8f92235e' name='des'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_cas_ulong' mangled-name='atomic_cas_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='exp'/>
	<parameter type-id='ee1f298e' name='des'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_8' mangled-name='atomic_swap_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_swap_16' mangled-name='atomic_swap_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_swap_32' mangled-name='atomic_swap_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_swap_ulong' mangled-name='atomic_swap_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_ptr' mangled-name='atomic_swap_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_set_long_excl' mangled-name='atomic_set_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_set_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='atomic_clear_long_excl' mangled-name='atomic_clear_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_clear_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='membar_enter' mangled-name='membar_enter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_enter'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_producer' mangled-name='membar_producer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_producer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_consumer' mangled-name='membar_consumer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_consumer'>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/getexecname.c' language='LANG_C99'>
	<function-decl name='getexecname' mangled-name='getexecname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getexecname'>
	<return type-id='80f4b756'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/list.c' language='LANG_C99'>
	<typedef-decl name='list_node_t' type-id='b0b5e45e' id='b21843b2'/>
	<typedef-decl name='list_t' type-id='e824dae9' id='0899125f'/>
	<class-decl name='list_node' size-in-bits='128' is-struct='yes' visibility='default' id='b0b5e45e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='next' type-id='b03eadb4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='prev' type-id='b03eadb4' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='list' size-in-bits='256' is-struct='yes' visibility='default' id='e824dae9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='list_size' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='list_head' type-id='b0b5e45e' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='b0b5e45e' size-in-bits='64' id='b03eadb4'/>
	<pointer-type-def type-id='b21843b2' size-in-bits='64' id='ccc38265'/>
	<pointer-type-def type-id='0899125f' size-in-bits='64' id='352ec160'/>
	<function-decl name='list_create' mangled-name='list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_create'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='b59d7dce' name='offset'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_destroy' mangled-name='list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_destroy'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_after' mangled-name='list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_after'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_before' mangled-name='list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_before'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_head' mangled-name='list_insert_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_head'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_tail' mangled-name='list_insert_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_tail'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove' mangled-name='list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove_head' mangled-name='list_remove_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_remove_tail' mangled-name='list_remove_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_head' mangled-name='list_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_tail' mangled-name='list_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_next' mangled-name='list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_next'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_prev' mangled-name='list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_prev'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_move_tail' mangled-name='list_move_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_move_tail'>
	<parameter type-id='352ec160' name='dst'/>
	<parameter type-id='352ec160' name='src'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_replace' mangled-name='list_link_replace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_replace'>
	<parameter type-id='ccc38265' name='lold'/>
	<parameter type-id='ccc38265' name='lnew'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_init' mangled-name='list_link_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_init'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_active' mangled-name='list_link_active' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_active'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='list_is_empty' mangled-name='list_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_is_empty'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/mkdirp.c' language='LANG_C99'>
	<typedef-decl name='mode_t' type-id='e1c52942' id='d50d396c'/>
	<function-decl name='mkdirp' mangled-name='mkdirp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mkdirp'>
	<parameter type-id='80f4b756' name='d'/>
	<parameter type-id='d50d396c' name='mode'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/gethostid.c' language='LANG_C99'>
	<function-decl name='get_system_hostid' mangled-name='get_system_hostid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_system_hostid'>
	<return type-id='7359adad'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getmntany.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='03085adc' size-in-bits='192' id='083f8d58'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
	<subrange length='20' type-id='7359adad' id='fdca39cf'/>
	</array-type-def>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='extmnttab' size-in-bits='320' is-struct='yes' visibility='default' id='0c544dc0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_major' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_minor' type-id='3502e3ff' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
	<typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
	<typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
	<typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
	<typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
	<typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
	<typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
	<typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
	<typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
	<typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
	<typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
	<typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
	<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
	<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1040'>
	<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1048'>
	<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1344'>
	<var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1472'>
	<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1568'>
	<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='timespec' size-in-bits='128' is-struct='yes' visibility='default' id='a9c79a1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tv_sec' type-id='65eda9c0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tv_nsec' type-id='03085adc' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
	<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
	<pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
	<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
	<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
	<pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
	<pointer-type-def type-id='0c544dc0' size-in-bits='64' id='394fc496'/>
	<pointer-type-def type-id='0bbec9cd' size-in-bits='64' id='62f7a03d'/>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<function-decl name='getmntany' mangled-name='getmntany' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getmntany'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<parameter type-id='9d424d31' name='mrefp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='_sol_getmntent' mangled-name='_sol_getmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='_sol_getmntent'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getextmntent' mangled-name='getextmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getextmntent'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='394fc496' name='entry'/>
	<parameter type-id='62f7a03d' name='statbuf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/zone.c' language='LANG_C99'>
	<typedef-decl name='zoneid_t' type-id='3502e3ff' id='4da03624'/>
	<function-decl name='getzoneid' mangled-name='getzoneid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getzoneid'>
	<return type-id='4da03624'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/page.c' language='LANG_C99'>
	<function-decl name='spl_pagesize' mangled-name='spl_pagesize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='spl_pagesize'>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/strlcat.c' language='LANG_C99'>
	<function-decl name='strlcat' mangled-name='strlcat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcat'>
	<parameter type-id='26a90f95' name='dst'/>
	<parameter type-id='80f4b756' name='src'/>
	<parameter type-id='b59d7dce' name='dstsize'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/strlcpy.c' language='LANG_C99'>
	<function-decl name='strlcpy' mangled-name='strlcpy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcpy'>
	<parameter type-id='26a90f95' name='dst'/>
	<parameter type-id='80f4b756' name='src'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/timestamp.c' language='LANG_C99'>
	<function-decl name='print_timestamp' mangled-name='print_timestamp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='print_timestamp'>
	<parameter type-id='3502e3ff' name='timestamp_fmt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libtpool/thread_pool.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='384' id='36d7f119'>
	<subrange length='48' type-id='7359adad' id='8f6d2a81'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='448' id='6093ff7c'>
	<subrange length='56' type-id='7359adad' id='f8137894'/>
	</array-type-def>
	<type-decl name='long long unsigned int' size-in-bits='64' id='3a47d82b'/>
	<array-type-def dimensions='1' type-id='f0981eeb' size-in-bits='64' id='0d532ec1'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<typedef-decl name='tpool_t' type-id='88d1b7f9' id='b1bbf10d'/>
	<typedef-decl name='pthread_t' type-id='7359adad' id='4051f5e7'/>
	<union-decl name='pthread_attr_t' size-in-bits='448' visibility='default' id='b63afacd'>
	<data-member access='public'>
	<var-decl name='__size' type-id='6093ff7c' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_attr_t' type-id='b63afacd' id='7d8569fd'/>
	<union-decl name='pthread_cond_t' size-in-bits='384' naming-typedef-id='62fab762' visibility='default' id='cbb12c12'>
	<data-member access='public'>
	<var-decl name='__data' type-id='c987b47c' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36d7f119' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='1eb56b1e' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_cond_t' type-id='cbb12c12' id='62fab762'/>
	<class-decl name='__pthread_cond_s' size-in-bits='384' is-struct='yes' visibility='default' id='c987b47c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='' type-id='ac5ab595' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='' type-id='ac5ab596' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__g_refs' type-id='0d532ec1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__g_size' type-id='0d532ec1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='__g1_orig_size' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__wrefs' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='__g_signals' type-id='0d532ec1' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__1' size-in-bits='64' is-anonymous='yes' visibility='default' id='ac5ab595'>
	<data-member access='public'>
	<var-decl name='__wseq' type-id='3a47d82b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__wseq32' type-id='e7f43f72' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='__anonymous_struct__' size-in-bits='64' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f72'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__low' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__high' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__2' size-in-bits='64' is-anonymous='yes' visibility='default' id='ac5ab596'>
	<data-member access='public'>
	<var-decl name='__g1_start' type-id='3a47d82b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__g1_start32' type-id='e7f43f72' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='tpool_job_t' type-id='3b8579e5' id='66a0afc9'/>
	<class-decl name='tpool_job' size-in-bits='192' is-struct='yes' visibility='default' id='3b8579e5'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tpj_next' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tpj_func' type-id='b7f9d8e6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tpj_arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='tpool_active_t' type-id='c8d086f4' id='6fcda10e'/>
	<class-decl name='tpool_active' size-in-bits='128' is-struct='yes' visibility='default' id='c8d086f4'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tpa_next' type-id='ad33e5e7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tpa_tid' type-id='4051f5e7' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='tpool' size-in-bits='2496' is-struct='yes' visibility='default' id='88d1b7f9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tp_forw' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tp_back' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tp_mutex' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='tp_busycv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='tp_workcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='tp_waitcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1600'>
	<var-decl name='tp_active' type-id='ad33e5e7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1664'>
	<var-decl name='tp_head' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1728'>
	<var-decl name='tp_tail' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1792'>
	<var-decl name='tp_attr' type-id='7d8569fd' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='tp_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2272'>
	<var-decl name='tp_linger' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='tp_njobs' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2336'>
	<var-decl name='tp_minimum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2368'>
	<var-decl name='tp_maximum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2400'>
	<var-decl name='tp_current' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='tp_idle' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='7d8569fd' size-in-bits='64' id='7347a39e'/>
	<pointer-type-def type-id='6fcda10e' size-in-bits='64' id='ad33e5e7'/>
	<pointer-type-def type-id='66a0afc9' size-in-bits='64' id='f32b30e4'/>
	<pointer-type-def type-id='b1bbf10d' size-in-bits='64' id='9cf59a50'/>
	<pointer-type-def type-id='c5c76c9c' size-in-bits='64' id='b7f9d8e6'/>
	<function-decl name='tpool_create' mangled-name='tpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_create'>
	<parameter type-id='3502e3ff' name='min_threads'/>
	<parameter type-id='3502e3ff' name='max_threads'/>
	<parameter type-id='3502e3ff' name='linger'/>
	<parameter type-id='7347a39e' name='attr'/>
	<return type-id='9cf59a50'/>
	</function-decl>
	<function-decl name='tpool_dispatch' mangled-name='tpool_dispatch' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_dispatch'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<parameter type-id='b7f9d8e6' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='tpool_destroy' mangled-name='tpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_destroy'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_abandon' mangled-name='tpool_abandon' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_abandon'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_wait' mangled-name='tpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_wait'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_suspend' mangled-name='tpool_suspend' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspend'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_suspended' mangled-name='tpool_suspended' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspended'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='tpool_resume' mangled-name='tpool_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_resume'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_member' mangled-name='tpool_member' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_member'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='c5c76c9c'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_changelist.c' language='LANG_C99'>
	<type-decl name='void' id='48b5725f'/>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_config.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='bf311473' size-in-bits='128' id='f0f65199'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<type-decl name='char' size-in-bits='8' id='a84c031d'/>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8192' id='b54ce520'>
	<subrange length='1024' type-id='7359adad' id='c60446f8'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='2048' id='d1617432'>
	<subrange length='256' type-id='7359adad' id='36e5b9fa'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
	<subrange length='40' type-id='7359adad' id='8f80b239'/>
	</array-type-def>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
	<class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
	<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
	<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
	<type-decl name='long long int' size-in-bits='64' id='1eb56b1e'/>
	<type-decl name='short int' size-in-bits='16' id='a2185560'/>
	<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
	<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
	<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<typedef-decl name='uu_avl_pool_t' type-id='12a530a8' id='7f84e390'/>
	<typedef-decl name='uu_avl_t' type-id='4af029d1' id='bb7f0973'/>
	<typedef-decl name='zfs_handle_t' type-id='f6ee4445' id='775509eb'/>
	<typedef-decl name='zpool_handle_t' type-id='67002a8a' id='b1efc708'/>
	<typedef-decl name='libzfs_handle_t' type-id='c8a9d9d8' id='95942d0c'/>
	<typedef-decl name='zpool_iter_f' type-id='3aebb66f' id='fa476e62'/>
	<typedef-decl name='zfs_iter_f' type-id='5571cde4' id='d8e49ab9'/>
	<typedef-decl name='avl_tree_t' type-id='b351119f' id='f20fbd51'/>
	<class-decl name='avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='428b67b3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_child' type-id='f0f65199' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_pcb' type-id='e475ab95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='avl_tree' size-in-bits='320' is-struct='yes' visibility='default' id='b351119f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_root' type-id='bf311473' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='avl_compar' type-id='585e1de9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='avl_numnodes' type-id='ee1f298e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='avl_pad' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='dmu_objset_stats' size-in-bits='2304' is-struct='yes' visibility='default' id='098f0221'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dds_num_clones' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='dds_creation_txg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='dds_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='dds_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='dds_is_snapshot' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='232'>
	<var-decl name='dds_inconsistent' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='240'>
	<var-decl name='dds_redacted' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='248'>
	<var-decl name='dds_origin' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='dmu_objset_stats_t' type-id='098f0221' id='b2c14f17'/>
	<enum-decl name='zfs_type_t' naming-typedef-id='2e45de5d' id='5d8f7321'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_TYPE_INVALID' value='0'/>
	<enumerator name='ZFS_TYPE_FILESYSTEM' value='1'/>
	<enumerator name='ZFS_TYPE_SNAPSHOT' value='2'/>
	<enumerator name='ZFS_TYPE_VOLUME' value='4'/>
	<enumerator name='ZFS_TYPE_POOL' value='8'/>
	<enumerator name='ZFS_TYPE_BOOKMARK' value='16'/>
	<enumerator name='ZFS_TYPE_VDEV' value='32'/>
	</enum-decl>
	<typedef-decl name='zfs_type_t' type-id='5d8f7321' id='2e45de5d'/>
	<enum-decl name='dmu_objset_type' id='6b1b19f9'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DMU_OST_NONE' value='0'/>
	<enumerator name='DMU_OST_META' value='1'/>
	<enumerator name='DMU_OST_ZFS' value='2'/>
	<enumerator name='DMU_OST_ZVOL' value='3'/>
	<enumerator name='DMU_OST_OTHER' value='4'/>
	<enumerator name='DMU_OST_ANY' value='5'/>
	<enumerator name='DMU_OST_NUMTYPES' value='6'/>
	</enum-decl>
	<typedef-decl name='dmu_objset_type_t' type-id='6b1b19f9' id='230f1e16'/>
	<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
	<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='B_FALSE' value='0'/>
	<enumerator name='B_TRUE' value='1'/>
	</enum-decl>
	<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
	<typedef-decl name='ulong_t' type-id='7359adad' id='ee1f298e'/>
	<typedef-decl name='longlong_t' type-id='1eb56b1e' id='9b3ff54f'/>
	<typedef-decl name='diskaddr_t' type-id='9b3ff54f' id='804dc465'/>
	<typedef-decl name='__re_long_size_t' type-id='7359adad' id='ba516949'/>
	<typedef-decl name='reg_syntax_t' type-id='7359adad' id='1b72c3b3'/>
	<class-decl name='re_pattern_buffer' size-in-bits='512' is-struct='yes' visibility='default' id='19fc9a8c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='buffer' type-id='33976309' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='allocated' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='used' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='syntax' type-id='1b72c3b3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fastmap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='translate' type-id='cf536864' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='re_nsub' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='can_be_null' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='449'>
	<var-decl name='regs_allocated' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='451'>
	<var-decl name='fastmap_accurate' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='452'>
	<var-decl name='no_sub' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='453'>
	<var-decl name='not_bol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='454'>
	<var-decl name='not_eol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='455'>
	<var-decl name='newline_anchor' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='regex_t' type-id='19fc9a8c' id='aca3bac8'/>
	<typedef-decl name='uintptr_t' type-id='7359adad' id='e475ab95'/>
	<union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
	<data-member access='public'>
	<var-decl name='__data' type-id='4c734837' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36c46961' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
	<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
	<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
	<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
	<typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
	<class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='__spins' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='176'>
	<var-decl name='__elision' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__list' type-id='518fb49c' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
	<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
	<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
	<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
	<typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
	<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
	<class-decl name='libzfs_handle' size-in-bits='18240' is-struct='yes' visibility='default' id='c8a9d9d8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='libzfs_error' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='libzfs_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='libzfs_pool_handles' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='libzfs_ns_avlpool' type-id='de82c773' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='libzfs_ns_avl' type-id='a5c21a38' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='libzfs_ns_gen' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='libzfs_desc_active' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='libzfs_action' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8544'>
	<var-decl name='libzfs_desc' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16736'>
	<var-decl name='libzfs_printerr' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16768'>
	<var-decl name='libzfs_mnttab_enable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16832'>
	<var-decl name='libzfs_mnttab_cache_lock' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17152'>
	<var-decl name='libzfs_mnttab_cache' type-id='f20fbd51' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17472'>
	<var-decl name='libzfs_pool_iter' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17504'>
	<var-decl name='libzfs_prop_debug' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17536'>
	<var-decl name='libzfs_urire' type-id='aca3bac8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18048'>
	<var-decl name='libzfs_max_nvlist' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18112'>
	<var-decl name='libfetch' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18176'>
	<var-decl name='libfetch_load_error' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zfs_handle' size-in-bits='4928' is-struct='yes' visibility='default' id='f6ee4445'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zfs_hdl' type-id='b0382bb3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zpool_hdl' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zfs_name' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='zfs_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2208'>
	<var-decl name='zfs_head_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='zfs_dmustats' type-id='b2c14f17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4544'>
	<var-decl name='zfs_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4608'>
	<var-decl name='zfs_user_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4672'>
	<var-decl name='zfs_recvd_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4736'>
	<var-decl name='zfs_mntcheck' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4800'>
	<var-decl name='zfs_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4864'>
	<var-decl name='zfs_props_table' type-id='ae3e8ca6' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zpool_handle' size-in-bits='2560' is-struct='yes' visibility='default' id='67002a8a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zpool_hdl' type-id='b0382bb3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zpool_next' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zpool_name' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='zpool_state' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='zpool_config_size' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='zpool_config' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2368'>
	<var-decl name='zpool_old_config' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='zpool_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2496'>
	<var-decl name='zpool_start_block' type-id='804dc465' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
	<pointer-type-def type-id='428b67b3' size-in-bits='64' id='bf311473'/>
	<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
	<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
	<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
	<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
	<pointer-type-def type-id='96ee24a5' size-in-bits='64' id='585e1de9'/>
	<pointer-type-def type-id='cb9628fa' size-in-bits='64' id='5571cde4'/>
	<pointer-type-def type-id='2bce87e3' size-in-bits='64' id='3aebb66f'/>
	<pointer-type-def type-id='95942d0c' size-in-bits='64' id='b0382bb3'/>
	<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
	<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
	<pointer-type-def type-id='b48d2441' size-in-bits='64' id='33976309'/>
	<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
	<pointer-type-def type-id='002ac4a6' size-in-bits='64' id='cf536864'/>
	<pointer-type-def type-id='7f84e390' size-in-bits='64' id='de82c773'/>
	<pointer-type-def type-id='bb7f0973' size-in-bits='64' id='a5c21a38'/>
	<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
	<pointer-type-def type-id='775509eb' size-in-bits='64' id='9200a744'/>
	<pointer-type-def type-id='b1efc708' size-in-bits='64' id='4c81de99'/>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
	<class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
	<function-decl name='zpool_get_config' mangled-name='zpool_get_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_config'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='oldconfig'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_get_features' mangled-name='zpool_get_features' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_features'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_refresh_stats' mangled-name='zpool_refresh_stats' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_refresh_stats'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_skip_pool' mangled-name='zpool_skip_pool' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_skip_pool'>
	<parameter type-id='80f4b756' name='poolname'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_iter' mangled-name='zpool_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_iter'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='fa476e62' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_root' mangled-name='zfs_iter_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_root'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='96ee24a5'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cb9628fa'>
	<parameter type-id='9200a744'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='2bce87e3'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_crypto.c' language='LANG_C99'>
	<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
	<pointer-type-def type-id='ae3e8ca6' size-in-bits='64' id='d8774064'/>
	<pointer-type-def type-id='3502e3ff' size-in-bits='64' id='4dd26a40'/>
	<function-decl name='zfs_crypto_get_encryption_root' mangled-name='zfs_crypto_get_encryption_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_get_encryption_root'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='37e3bd22' name='is_encroot'/>
	<parameter type-id='26a90f95' name='buf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_create' mangled-name='zfs_crypto_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='parent_name'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='pool_props'/>
	<parameter type-id='c19b74c3' name='stdin_available'/>
	<parameter type-id='d8774064' name='wkeydata_out'/>
	<parameter type-id='4dd26a40' name='wkeylen_out'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_clone_check' mangled-name='zfs_crypto_clone_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_clone_check'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='9200a744' name='origin_zhp'/>
	<parameter type-id='26a90f95' name='parent_name'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_attempt_load_keys' mangled-name='zfs_crypto_attempt_load_keys' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_attempt_load_keys'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_load_key' mangled-name='zfs_crypto_load_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_load_key'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='noop'/>
	<parameter type-id='80f4b756' name='alt_keylocation'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_unload_key' mangled-name='zfs_crypto_unload_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_unload_key'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_rewrap' mangled-name='zfs_crypto_rewrap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_rewrap'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='5ce45b60' name='raw_props'/>
	<parameter type-id='c19b74c3' name='inheritkey'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_dataset.c' language='LANG_C99'>
	<class-decl name='zprop_list' size-in-bits='448' is-struct='yes' visibility='default' id='bd9b4291'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pl_prop' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pl_user_prop' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='pl_next' type-id='9f1a1109' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='pl_all' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='pl_width' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='pl_recvd_width' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='pl_fixed' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_list_t' type-id='bd9b4291' id='bdb8ac4f'/>
	<class-decl name='renameflags' size-in-bits='32' is-struct='yes' visibility='default' id='7aee5792'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='recursive' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1'>
	<var-decl name='nounmount' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2'>
	<var-decl name='forceunmount' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='renameflags_t' type-id='7aee5792' id='067170c2'/>
	<typedef-decl name='zfs_userspace_cb_t' type-id='ca64ff60' id='16c5f410'/>
	<enum-decl name='zfs_prop_t' naming-typedef-id='58603c44' id='4b000d60'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPROP_CONT' value='-2'/>
	<enumerator name='ZPROP_INVAL' value='-1'/>
	<enumerator name='ZPROP_USERPROP' value='-1'/>
	<enumerator name='ZFS_PROP_TYPE' value='0'/>
	<enumerator name='ZFS_PROP_CREATION' value='1'/>
	<enumerator name='ZFS_PROP_USED' value='2'/>
	<enumerator name='ZFS_PROP_AVAILABLE' value='3'/>
	<enumerator name='ZFS_PROP_REFERENCED' value='4'/>
	<enumerator name='ZFS_PROP_COMPRESSRATIO' value='5'/>
	<enumerator name='ZFS_PROP_MOUNTED' value='6'/>
	<enumerator name='ZFS_PROP_ORIGIN' value='7'/>
	<enumerator name='ZFS_PROP_QUOTA' value='8'/>
	<enumerator name='ZFS_PROP_RESERVATION' value='9'/>
	<enumerator name='ZFS_PROP_VOLSIZE' value='10'/>
	<enumerator name='ZFS_PROP_VOLBLOCKSIZE' value='11'/>
	<enumerator name='ZFS_PROP_RECORDSIZE' value='12'/>
	<enumerator name='ZFS_PROP_MOUNTPOINT' value='13'/>
	<enumerator name='ZFS_PROP_SHARENFS' value='14'/>
	<enumerator name='ZFS_PROP_CHECKSUM' value='15'/>
	<enumerator name='ZFS_PROP_COMPRESSION' value='16'/>
	<enumerator name='ZFS_PROP_ATIME' value='17'/>
	<enumerator name='ZFS_PROP_DEVICES' value='18'/>
	<enumerator name='ZFS_PROP_EXEC' value='19'/>
	<enumerator name='ZFS_PROP_SETUID' value='20'/>
	<enumerator name='ZFS_PROP_READONLY' value='21'/>
	<enumerator name='ZFS_PROP_ZONED' value='22'/>
	<enumerator name='ZFS_PROP_SNAPDIR' value='23'/>
	<enumerator name='ZFS_PROP_ACLMODE' value='24'/>
	<enumerator name='ZFS_PROP_ACLINHERIT' value='25'/>
	<enumerator name='ZFS_PROP_CREATETXG' value='26'/>
	<enumerator name='ZFS_PROP_NAME' value='27'/>
	<enumerator name='ZFS_PROP_CANMOUNT' value='28'/>
	<enumerator name='ZFS_PROP_ISCSIOPTIONS' value='29'/>
	<enumerator name='ZFS_PROP_XATTR' value='30'/>
	<enumerator name='ZFS_PROP_NUMCLONES' value='31'/>
	<enumerator name='ZFS_PROP_COPIES' value='32'/>
	<enumerator name='ZFS_PROP_VERSION' value='33'/>
	<enumerator name='ZFS_PROP_UTF8ONLY' value='34'/>
	<enumerator name='ZFS_PROP_NORMALIZE' value='35'/>
	<enumerator name='ZFS_PROP_CASE' value='36'/>
	<enumerator name='ZFS_PROP_VSCAN' value='37'/>
	<enumerator name='ZFS_PROP_NBMAND' value='38'/>
	<enumerator name='ZFS_PROP_SHARESMB' value='39'/>
	<enumerator name='ZFS_PROP_REFQUOTA' value='40'/>
	<enumerator name='ZFS_PROP_REFRESERVATION' value='41'/>
	<enumerator name='ZFS_PROP_GUID' value='42'/>
	<enumerator name='ZFS_PROP_PRIMARYCACHE' value='43'/>
	<enumerator name='ZFS_PROP_SECONDARYCACHE' value='44'/>
	<enumerator name='ZFS_PROP_USEDSNAP' value='45'/>
	<enumerator name='ZFS_PROP_USEDDS' value='46'/>
	<enumerator name='ZFS_PROP_USEDCHILD' value='47'/>
	<enumerator name='ZFS_PROP_USEDREFRESERV' value='48'/>
	<enumerator name='ZFS_PROP_USERACCOUNTING' value='49'/>
	<enumerator name='ZFS_PROP_STMF_SHAREINFO' value='50'/>
	<enumerator name='ZFS_PROP_DEFER_DESTROY' value='51'/>
	<enumerator name='ZFS_PROP_USERREFS' value='52'/>
	<enumerator name='ZFS_PROP_LOGBIAS' value='53'/>
	<enumerator name='ZFS_PROP_UNIQUE' value='54'/>
	<enumerator name='ZFS_PROP_OBJSETID' value='55'/>
	<enumerator name='ZFS_PROP_DEDUP' value='56'/>
	<enumerator name='ZFS_PROP_MLSLABEL' value='57'/>
	<enumerator name='ZFS_PROP_SYNC' value='58'/>
	<enumerator name='ZFS_PROP_DNODESIZE' value='59'/>
	<enumerator name='ZFS_PROP_REFRATIO' value='60'/>
	<enumerator name='ZFS_PROP_WRITTEN' value='61'/>
	<enumerator name='ZFS_PROP_CLONES' value='62'/>
	<enumerator name='ZFS_PROP_LOGICALUSED' value='63'/>
	<enumerator name='ZFS_PROP_LOGICALREFERENCED' value='64'/>
	<enumerator name='ZFS_PROP_INCONSISTENT' value='65'/>
	<enumerator name='ZFS_PROP_VOLMODE' value='66'/>
	<enumerator name='ZFS_PROP_FILESYSTEM_LIMIT' value='67'/>
	<enumerator name='ZFS_PROP_SNAPSHOT_LIMIT' value='68'/>
	<enumerator name='ZFS_PROP_FILESYSTEM_COUNT' value='69'/>
	<enumerator name='ZFS_PROP_SNAPSHOT_COUNT' value='70'/>
	<enumerator name='ZFS_PROP_SNAPDEV' value='71'/>
	<enumerator name='ZFS_PROP_ACLTYPE' value='72'/>
	<enumerator name='ZFS_PROP_SELINUX_CONTEXT' value='73'/>
	<enumerator name='ZFS_PROP_SELINUX_FSCONTEXT' value='74'/>
	<enumerator name='ZFS_PROP_SELINUX_DEFCONTEXT' value='75'/>
	<enumerator name='ZFS_PROP_SELINUX_ROOTCONTEXT' value='76'/>
	<enumerator name='ZFS_PROP_RELATIME' value='77'/>
	<enumerator name='ZFS_PROP_REDUNDANT_METADATA' value='78'/>
	<enumerator name='ZFS_PROP_OVERLAY' value='79'/>
	<enumerator name='ZFS_PROP_PREV_SNAP' value='80'/>
	<enumerator name='ZFS_PROP_RECEIVE_RESUME_TOKEN' value='81'/>
	<enumerator name='ZFS_PROP_ENCRYPTION' value='82'/>
	<enumerator name='ZFS_PROP_KEYLOCATION' value='83'/>
	<enumerator name='ZFS_PROP_KEYFORMAT' value='84'/>
	<enumerator name='ZFS_PROP_PBKDF2_SALT' value='85'/>
	<enumerator name='ZFS_PROP_PBKDF2_ITERS' value='86'/>
	<enumerator name='ZFS_PROP_ENCRYPTION_ROOT' value='87'/>
	<enumerator name='ZFS_PROP_KEY_GUID' value='88'/>
	<enumerator name='ZFS_PROP_KEYSTATUS' value='89'/>
	<enumerator name='ZFS_PROP_REMAPTXG' value='90'/>
	<enumerator name='ZFS_PROP_SPECIAL_SMALL_BLOCKS' value='91'/>
	<enumerator name='ZFS_PROP_IVSET_GUID' value='92'/>
	<enumerator name='ZFS_PROP_REDACTED' value='93'/>
	<enumerator name='ZFS_PROP_REDACT_SNAPS' value='94'/>
	<enumerator name='ZFS_PROP_SNAPSHOTS_CHANGED' value='95'/>
	<enumerator name='ZFS_NUM_PROPS' value='96'/>
	</enum-decl>
	<typedef-decl name='zfs_prop_t' type-id='4b000d60' id='58603c44'/>
	<enum-decl name='zfs_userquota_prop_t' naming-typedef-id='279fde6a' id='5258d2f6'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_PROP_USERUSED' value='0'/>
	<enumerator name='ZFS_PROP_USERQUOTA' value='1'/>
	<enumerator name='ZFS_PROP_GROUPUSED' value='2'/>
	<enumerator name='ZFS_PROP_GROUPQUOTA' value='3'/>
	<enumerator name='ZFS_PROP_USEROBJUSED' value='4'/>
	<enumerator name='ZFS_PROP_USEROBJQUOTA' value='5'/>
	<enumerator name='ZFS_PROP_GROUPOBJUSED' value='6'/>
	<enumerator name='ZFS_PROP_GROUPOBJQUOTA' value='7'/>
	<enumerator name='ZFS_PROP_PROJECTUSED' value='8'/>
	<enumerator name='ZFS_PROP_PROJECTQUOTA' value='9'/>
	<enumerator name='ZFS_PROP_PROJECTOBJUSED' value='10'/>
	<enumerator name='ZFS_PROP_PROJECTOBJQUOTA' value='11'/>
	<enumerator name='ZFS_NUM_USERQUOTA_PROPS' value='12'/>
	</enum-decl>
	<typedef-decl name='zfs_userquota_prop_t' type-id='5258d2f6' id='279fde6a'/>
	<enum-decl name='zprop_source_t' naming-typedef-id='a2256d42' id='5903f80e'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPROP_SRC_NONE' value='1'/>
	<enumerator name='ZPROP_SRC_DEFAULT' value='2'/>
	<enumerator name='ZPROP_SRC_TEMPORARY' value='4'/>
	<enumerator name='ZPROP_SRC_LOCAL' value='8'/>
	<enumerator name='ZPROP_SRC_INHERITED' value='16'/>
	<enumerator name='ZPROP_SRC_RECEIVED' value='32'/>
	</enum-decl>
	<typedef-decl name='zprop_source_t' type-id='5903f80e' id='a2256d42'/>
	<enum-decl name='zfs_wait_activity_t' naming-typedef-id='3024501a' id='527d5dc6'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_WAIT_DELETEQ' value='0'/>
	<enumerator name='ZFS_WAIT_NUM_ACTIVITIES' value='1'/>
	</enum-decl>
	<typedef-decl name='zfs_wait_activity_t' type-id='527d5dc6' id='3024501a'/>
	<class-decl name='mnttab' size-in-bits='256' is-struct='yes' visibility='default' id='1b055409'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__uid_t' type-id='f0981eeb' id='cc5fcceb'/>
	<typedef-decl name='uid_t' type-id='cc5fcceb' id='354978ed'/>
	<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
	<qualified-type-def type-id='775509eb' const='yes' id='5eadf2db'/>
	<pointer-type-def type-id='5eadf2db' size-in-bits='64' id='fcd57163'/>
	<pointer-type-def type-id='7e291ce6' size-in-bits='64' id='ca64ff60'/>
	<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
	<pointer-type-def type-id='1b055409' size-in-bits='64' id='9d424d31'/>
	<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
	<pointer-type-def type-id='bd9b4291' size-in-bits='64' id='9f1a1109'/>
	<pointer-type-def type-id='bdb8ac4f' size-in-bits='64' id='3a9b2288'/>
	<pointer-type-def type-id='3a9b2288' size-in-bits='64' id='e4378506'/>
	<pointer-type-def type-id='a2256d42' size-in-bits='64' id='debc6aa3'/>
	<function-decl name='zfs_type_to_name' mangled-name='zfs_type_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_type_to_name'>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_name_valid' mangled-name='zfs_name_valid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_valid'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_free_handles' mangled-name='zpool_free_handles' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_free_handles'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_refresh_properties' mangled-name='zfs_refresh_properties' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_refresh_properties'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_handle_dup' mangled-name='zfs_handle_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_handle_dup'>
	<parameter type-id='9200a744' name='zhp_orig'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zfs_bookmark_exists' mangled-name='zfs_bookmark_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_bookmark_exists'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_open' mangled-name='zfs_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_open'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='95e97e5e' name='types'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zfs_close' mangled-name='zfs_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_close'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_init' mangled-name='libzfs_mnttab_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_init'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_fini' mangled-name='libzfs_mnttab_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_fini'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_cache' mangled-name='libzfs_mnttab_cache' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_cache'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='c19b74c3' name='enable'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_find' mangled-name='libzfs_mnttab_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_find'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='9d424d31' name='entry'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_add' mangled-name='libzfs_mnttab_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_add'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='special'/>
	<parameter type-id='80f4b756' name='mountp'/>
	<parameter type-id='80f4b756' name='mntopts'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_remove' mangled-name='libzfs_mnttab_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_remove'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_spa_version' mangled-name='zfs_spa_version' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='7292109c' name='spa_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_valid_proplist' mangled-name='zfs_valid_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_valid_proplist'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='9c313c2d' name='zoned'/>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='4c81de99' name='zpool_hdl'/>
	<parameter type-id='c19b74c3' name='key_params_ok'/>
	<parameter type-id='80f4b756' name='errbuf'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_prop_set' mangled-name='zfs_prop_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_set_list' mangled-name='zfs_prop_set_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set_list'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_inherit' mangled-name='zfs_prop_inherit' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inherit'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='c19b74c3' name='received'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getprop_uint64' mangled-name='getprop_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getprop_uint64'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='9b23c9ad' name='source'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_get_recvd' mangled-name='zfs_prop_get_recvd' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_recvd'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='b59d7dce' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_clones_nvl' mangled-name='zfs_get_clones_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_clones_nvl'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_prop_get' mangled-name='zfs_prop_get' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='b59d7dce' name='proplen'/>
	<parameter type-id='debc6aa3' name='src'/>
	<parameter type-id='26a90f95' name='statbuf'/>
	<parameter type-id='b59d7dce' name='statlen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_int' mangled-name='zfs_prop_get_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_int'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_get_numeric' mangled-name='zfs_prop_get_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_numeric'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='5d6479ae' name='value'/>
	<parameter type-id='debc6aa3' name='src'/>
	<parameter type-id='26a90f95' name='statbuf'/>
	<parameter type-id='b59d7dce' name='statlen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_userquota_int' mangled-name='zfs_prop_get_userquota_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota_int'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='5d6479ae' name='propvalue'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_userquota' mangled-name='zfs_prop_get_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='95e97e5e' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_written_int' mangled-name='zfs_prop_get_written_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written_int'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='5d6479ae' name='propvalue'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_written' mangled-name='zfs_prop_get_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='95e97e5e' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_name' mangled-name='zfs_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_name'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_get_pool_name' mangled-name='zfs_get_pool_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_name'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_get_type' mangled-name='zfs_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_type'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='2e45de5d'/>
	</function-decl>
	<function-decl name='zfs_get_underlying_type' mangled-name='zfs_get_underlying_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_type'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='2e45de5d'/>
	</function-decl>
	<function-decl name='zfs_parent_name' mangled-name='zfs_parent_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parent_name'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_dataset_exists' mangled-name='zfs_dataset_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_exists'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='types'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_create_ancestors' mangled-name='zfs_create_ancestors' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create_ancestors'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_create' mangled-name='zfs_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy' mangled-name='zfs_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps' mangled-name='zfs_destroy_snaps' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='26a90f95' name='snapname'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps_nvl' mangled-name='zfs_destroy_snaps_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_clone' mangled-name='zfs_clone' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_clone'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_promote' mangled-name='zfs_promote' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_promote'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_snapshot_nvl' mangled-name='zfs_snapshot_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot_nvl'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_snapshot' mangled-name='zfs_snapshot' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_rollback' mangled-name='zfs_rollback' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rollback'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='9200a744' name='snap'/>
	<parameter type-id='c19b74c3' name='force'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_rename' mangled-name='zfs_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rename'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='067170c2' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_all_props' mangled-name='zfs_get_all_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_all_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_get_recvd_props' mangled-name='zfs_get_recvd_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_recvd_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_get_user_props' mangled-name='zfs_get_user_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_user_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_expand_proplist' mangled-name='zfs_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_expand_proplist'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='e4378506' name='plp'/>
	<parameter type-id='c19b74c3' name='received'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prune_proplist' mangled-name='zfs_prune_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prune_proplist'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='ae3e8ca6' name='props'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_add' mangled-name='zfs_smb_acl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_add'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='resource'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_remove' mangled-name='zfs_smb_acl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_remove'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='resource'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_purge' mangled-name='zfs_smb_acl_purge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_purge'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_rename' mangled-name='zfs_smb_acl_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_rename'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='oldname'/>
	<parameter type-id='26a90f95' name='newname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_userspace' mangled-name='zfs_userspace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_userspace'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='279fde6a' name='type'/>
	<parameter type-id='16c5f410' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_hold' mangled-name='zfs_hold' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='tag'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_hold_nvl' mangled-name='zfs_hold_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold_nvl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='5ce45b60' name='holds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_release' mangled-name='zfs_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_release'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='tag'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_fsacl' mangled-name='zfs_get_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_fsacl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='857bb57e' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_set_fsacl' mangled-name='zfs_set_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_set_fsacl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='un'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_holds' mangled-name='zfs_get_holds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_holds'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='857bb57e' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zvol_volsize_to_reservation' mangled-name='zvol_volsize_to_reservation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zvol_volsize_to_reservation'>
	<parameter type-id='4c81de99' name='zph'/>
	<parameter type-id='9c313c2d' name='volsize'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_wait_status' mangled-name='zfs_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_wait_status'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='3024501a' name='activity'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='7e291ce6'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='354978ed'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_diff.c' language='LANG_C99'>
	<function-decl name='zfs_show_diffs' mangled-name='zfs_show_diffs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_show_diffs'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='fromsnap'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_import.c' language='LANG_C99'>
	<typedef-decl name='refresh_config_func_t' type-id='29f040d2' id='b7c58eaa'/>
	<typedef-decl name='pool_active_func_t' type-id='baa42fef' id='de5d1d8f'/>
	<class-decl name='pool_config_ops' size-in-bits='128' is-struct='yes' visibility='default' id='8b092c69'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pco_refresh_config' type-id='e7c00489' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pco_pool_active' type-id='9eadf5e0' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='pool_config_ops_t' type-id='1a21babe' id='b1e62775'/>
	<enum-decl name='pool_state' id='4871ac24'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_STATE_ACTIVE' value='0'/>
	<enumerator name='POOL_STATE_EXPORTED' value='1'/>
	<enumerator name='POOL_STATE_DESTROYED' value='2'/>
	<enumerator name='POOL_STATE_SPARE' value='3'/>
	<enumerator name='POOL_STATE_L2CACHE' value='4'/>
	<enumerator name='POOL_STATE_UNINITIALIZED' value='5'/>
	<enumerator name='POOL_STATE_UNAVAIL' value='6'/>
	<enumerator name='POOL_STATE_POTENTIALLY_ACTIVE' value='7'/>
	</enum-decl>
	<typedef-decl name='pool_state_t' type-id='4871ac24' id='084a08a3'/>
	<qualified-type-def type-id='8b092c69' const='yes' id='1a21babe'/>
	<pointer-type-def type-id='de5d1d8f' size-in-bits='64' id='9eadf5e0'/>
	<pointer-type-def type-id='084a08a3' size-in-bits='64' id='b9ea57b8'/>
	<pointer-type-def type-id='b7c58eaa' size-in-bits='64' id='e7c00489'/>
	<var-decl name='libzfs_config_ops' type-id='b1e62775' mangled-name='libzfs_config_ops' visibility='default' elf-symbol-id='libzfs_config_ops'/>
	<function-decl name='zpool_clear_label' mangled-name='zpool_clear_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear_label'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_in_use' mangled-name='zpool_in_use' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_in_use'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='b9ea57b8' name='state'/>
	<parameter type-id='9b23c9ad' name='namestr'/>
	<parameter type-id='37e3bd22' name='inuse'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='baa42fef'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='29f040d2'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='5ce45b60'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_iter.c' language='LANG_C99'>
	<function-decl name='zfs_iter_filesystems' mangled-name='zfs_iter_filesystems' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_filesystems'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots' mangled-name='zfs_iter_snapshots' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='simple'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='9c313c2d' name='min_txg'/>
	<parameter type-id='9c313c2d' name='max_txg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_bookmarks' mangled-name='zfs_iter_bookmarks' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_bookmarks'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots_sorted' mangled-name='zfs_iter_snapshots_sorted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots_sorted'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='callback'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='9c313c2d' name='min_txg'/>
	<parameter type-id='9c313c2d' name='max_txg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapspec' mangled-name='zfs_iter_snapspec' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapspec'>
	<parameter type-id='9200a744' name='fs_zhp'/>
	<parameter type-id='80f4b756' name='spec_orig'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_children' mangled-name='zfs_iter_children' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_children'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_dependents' mangled-name='zfs_iter_dependents' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_dependents'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='allowrecursion'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_mounted' mangled-name='zfs_iter_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_mounted'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_mount.c' language='LANG_C99'>
	<class-decl name='get_all_cb' size-in-bits='192' is-struct='yes' visibility='default' id='803dac95'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_handles' type-id='4507922a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='cb_alloc' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='cb_used' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='get_all_cb_t' type-id='803dac95' id='9b293607'/>
	<enum-decl name='sa_protocol' id='9155d4b5'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='SA_PROTOCOL_NFS' value='0'/>
	<enumerator name='SA_PROTOCOL_SMB' value='1'/>
	<enumerator name='SA_PROTOCOL_COUNT' value='2'/>
	</enum-decl>
	<qualified-type-def type-id='9155d4b5' const='yes' id='9f2c1699'/>
	<pointer-type-def type-id='9f2c1699' size-in-bits='64' id='4567bbc9'/>
	<pointer-type-def type-id='9b293607' size-in-bits='64' id='77bf1784'/>
	<pointer-type-def type-id='9200a744' size-in-bits='64' id='4507922a'/>
	<function-decl name='is_mounted' mangled-name='is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mounted'>
	<parameter type-id='b0382bb3' name='zfs_hdl'/>
	<parameter type-id='80f4b756' name='special'/>
	<parameter type-id='9b23c9ad' name='where'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_is_mounted' mangled-name='zfs_is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_mounted'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='9b23c9ad' name='where'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_mount' mangled-name='zfs_mount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='options'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_mount_at' mangled-name='zfs_mount_at' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_at'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='options'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unmount' mangled-name='zfs_unmount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmount'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unmountall' mangled-name='zfs_unmountall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmountall'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_share' mangled-name='zfs_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_share'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_is_shared' mangled-name='zfs_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_shared'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='9b23c9ad' name='where'/>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_commit_shares' mangled-name='zfs_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_shares'>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_truncate_shares' mangled-name='zfs_truncate_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_truncate_shares'>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_unshare' mangled-name='zfs_unshare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshare'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unshareall' mangled-name='zfs_unshareall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_add_handle' mangled-name='libzfs_add_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_add_handle'>
	<parameter type-id='77bf1784' name='cbp'/>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_foreach_mountpoint' mangled-name='zfs_foreach_mountpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_foreach_mountpoint'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4507922a' name='handles'/>
	<parameter type-id='b59d7dce' name='num_handles'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='c19b74c3' name='parallel'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_enable_datasets' mangled-name='zpool_enable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_enable_datasets'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='mntopts'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_disable_datasets' mangled-name='zpool_disable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='c19b74c3' name='force'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_pool.c' language='LANG_C99'>
	<class-decl name='splitflags' size-in-bits='64' is-struct='yes' visibility='default' id='dc01bf52'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dryrun' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1'>
	<var-decl name='import' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='name_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='splitflags_t' type-id='dc01bf52' id='325c1e34'/>
	<class-decl name='trimflags' size-in-bits='192' is-struct='yes' visibility='default' id='8ef58008'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='fullpool' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='secure' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='wait' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='rate' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='trimflags_t' type-id='8ef58008' id='a093cbb8'/>
	<enum-decl name='zpool_compat_status_t' naming-typedef-id='901b78d1' id='20676925'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_COMPATIBILITY_OK' value='0'/>
	<enumerator name='ZPOOL_COMPATIBILITY_WARNTOKEN' value='1'/>
	<enumerator name='ZPOOL_COMPATIBILITY_BADTOKEN' value='2'/>
	<enumerator name='ZPOOL_COMPATIBILITY_BADFILE' value='3'/>
	<enumerator name='ZPOOL_COMPATIBILITY_NOFILES' value='4'/>
	</enum-decl>
	<typedef-decl name='zpool_compat_status_t' type-id='20676925' id='901b78d1'/>
	<enum-decl name='zpool_prop_t' naming-typedef-id='5d0c23fb' id='af1ba157'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_PROP_INVAL' value='-1'/>
	<enumerator name='ZPOOL_PROP_NAME' value='0'/>
	<enumerator name='ZPOOL_PROP_SIZE' value='1'/>
	<enumerator name='ZPOOL_PROP_CAPACITY' value='2'/>
	<enumerator name='ZPOOL_PROP_ALTROOT' value='3'/>
	<enumerator name='ZPOOL_PROP_HEALTH' value='4'/>
	<enumerator name='ZPOOL_PROP_GUID' value='5'/>
	<enumerator name='ZPOOL_PROP_VERSION' value='6'/>
	<enumerator name='ZPOOL_PROP_BOOTFS' value='7'/>
	<enumerator name='ZPOOL_PROP_DELEGATION' value='8'/>
	<enumerator name='ZPOOL_PROP_AUTOREPLACE' value='9'/>
	<enumerator name='ZPOOL_PROP_CACHEFILE' value='10'/>
	<enumerator name='ZPOOL_PROP_FAILUREMODE' value='11'/>
	<enumerator name='ZPOOL_PROP_LISTSNAPS' value='12'/>
	<enumerator name='ZPOOL_PROP_AUTOEXPAND' value='13'/>
	<enumerator name='ZPOOL_PROP_DEDUPDITTO' value='14'/>
	<enumerator name='ZPOOL_PROP_DEDUPRATIO' value='15'/>
	<enumerator name='ZPOOL_PROP_FREE' value='16'/>
	<enumerator name='ZPOOL_PROP_ALLOCATED' value='17'/>
	<enumerator name='ZPOOL_PROP_READONLY' value='18'/>
	<enumerator name='ZPOOL_PROP_ASHIFT' value='19'/>
	<enumerator name='ZPOOL_PROP_COMMENT' value='20'/>
	<enumerator name='ZPOOL_PROP_EXPANDSZ' value='21'/>
	<enumerator name='ZPOOL_PROP_FREEING' value='22'/>
	<enumerator name='ZPOOL_PROP_FRAGMENTATION' value='23'/>
	<enumerator name='ZPOOL_PROP_LEAKED' value='24'/>
	<enumerator name='ZPOOL_PROP_MAXBLOCKSIZE' value='25'/>
	<enumerator name='ZPOOL_PROP_TNAME' value='26'/>
	<enumerator name='ZPOOL_PROP_MAXDNODESIZE' value='27'/>
	<enumerator name='ZPOOL_PROP_MULTIHOST' value='28'/>
	<enumerator name='ZPOOL_PROP_CHECKPOINT' value='29'/>
	<enumerator name='ZPOOL_PROP_LOAD_GUID' value='30'/>
	<enumerator name='ZPOOL_PROP_AUTOTRIM' value='31'/>
	<enumerator name='ZPOOL_PROP_COMPATIBILITY' value='32'/>
	<enumerator name='ZPOOL_NUM_PROPS' value='33'/>
	</enum-decl>
	<typedef-decl name='zpool_prop_t' type-id='af1ba157' id='5d0c23fb'/>
	<enum-decl name='vdev_prop_t' naming-typedef-id='5aa5c90c' id='1573bec8'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_PROP_INVAL' value='-1'/>
	<enumerator name='VDEV_PROP_USERPROP' value='-1'/>
	<enumerator name='VDEV_PROP_NAME' value='0'/>
	<enumerator name='VDEV_PROP_CAPACITY' value='1'/>
	<enumerator name='VDEV_PROP_STATE' value='2'/>
	<enumerator name='VDEV_PROP_GUID' value='3'/>
	<enumerator name='VDEV_PROP_ASIZE' value='4'/>
	<enumerator name='VDEV_PROP_PSIZE' value='5'/>
	<enumerator name='VDEV_PROP_ASHIFT' value='6'/>
	<enumerator name='VDEV_PROP_SIZE' value='7'/>
	<enumerator name='VDEV_PROP_FREE' value='8'/>
	<enumerator name='VDEV_PROP_ALLOCATED' value='9'/>
	<enumerator name='VDEV_PROP_COMMENT' value='10'/>
	<enumerator name='VDEV_PROP_EXPANDSZ' value='11'/>
	<enumerator name='VDEV_PROP_FRAGMENTATION' value='12'/>
	<enumerator name='VDEV_PROP_BOOTSIZE' value='13'/>
	<enumerator name='VDEV_PROP_PARITY' value='14'/>
	<enumerator name='VDEV_PROP_PATH' value='15'/>
	<enumerator name='VDEV_PROP_DEVID' value='16'/>
	<enumerator name='VDEV_PROP_PHYS_PATH' value='17'/>
	<enumerator name='VDEV_PROP_ENC_PATH' value='18'/>
	<enumerator name='VDEV_PROP_FRU' value='19'/>
	<enumerator name='VDEV_PROP_PARENT' value='20'/>
	<enumerator name='VDEV_PROP_CHILDREN' value='21'/>
	<enumerator name='VDEV_PROP_NUMCHILDREN' value='22'/>
	<enumerator name='VDEV_PROP_READ_ERRORS' value='23'/>
	<enumerator name='VDEV_PROP_WRITE_ERRORS' value='24'/>
	<enumerator name='VDEV_PROP_CHECKSUM_ERRORS' value='25'/>
	<enumerator name='VDEV_PROP_INITIALIZE_ERRORS' value='26'/>
	<enumerator name='VDEV_PROP_OPS_NULL' value='27'/>
	<enumerator name='VDEV_PROP_OPS_READ' value='28'/>
	<enumerator name='VDEV_PROP_OPS_WRITE' value='29'/>
	<enumerator name='VDEV_PROP_OPS_FREE' value='30'/>
	<enumerator name='VDEV_PROP_OPS_CLAIM' value='31'/>
	<enumerator name='VDEV_PROP_OPS_TRIM' value='32'/>
	<enumerator name='VDEV_PROP_BYTES_NULL' value='33'/>
	<enumerator name='VDEV_PROP_BYTES_READ' value='34'/>
	<enumerator name='VDEV_PROP_BYTES_WRITE' value='35'/>
	<enumerator name='VDEV_PROP_BYTES_FREE' value='36'/>
	<enumerator name='VDEV_PROP_BYTES_CLAIM' value='37'/>
	<enumerator name='VDEV_PROP_BYTES_TRIM' value='38'/>
	<enumerator name='VDEV_PROP_REMOVING' value='39'/>
	<enumerator name='VDEV_PROP_ALLOCATING' value='40'/>
	<enumerator name='VDEV_NUM_PROPS' value='41'/>
	</enum-decl>
	<typedef-decl name='vdev_prop_t' type-id='1573bec8' id='5aa5c90c'/>
	<enum-decl name='vdev_state' id='21566197'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_STATE_UNKNOWN' value='0'/>
	<enumerator name='VDEV_STATE_CLOSED' value='1'/>
	<enumerator name='VDEV_STATE_OFFLINE' value='2'/>
	<enumerator name='VDEV_STATE_REMOVED' value='3'/>
	<enumerator name='VDEV_STATE_CANT_OPEN' value='4'/>
	<enumerator name='VDEV_STATE_FAULTED' value='5'/>
	<enumerator name='VDEV_STATE_DEGRADED' value='6'/>
	<enumerator name='VDEV_STATE_HEALTHY' value='7'/>
	</enum-decl>
	<typedef-decl name='vdev_state_t' type-id='21566197' id='35acf840'/>
	<enum-decl name='vdev_aux' id='7f5bcca4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_AUX_NONE' value='0'/>
	<enumerator name='VDEV_AUX_OPEN_FAILED' value='1'/>
	<enumerator name='VDEV_AUX_CORRUPT_DATA' value='2'/>
	<enumerator name='VDEV_AUX_NO_REPLICAS' value='3'/>
	<enumerator name='VDEV_AUX_BAD_GUID_SUM' value='4'/>
	<enumerator name='VDEV_AUX_TOO_SMALL' value='5'/>
	<enumerator name='VDEV_AUX_BAD_LABEL' value='6'/>
	<enumerator name='VDEV_AUX_VERSION_NEWER' value='7'/>
	<enumerator name='VDEV_AUX_VERSION_OLDER' value='8'/>
	<enumerator name='VDEV_AUX_UNSUP_FEAT' value='9'/>
	<enumerator name='VDEV_AUX_SPARED' value='10'/>
	<enumerator name='VDEV_AUX_ERR_EXCEEDED' value='11'/>
	<enumerator name='VDEV_AUX_IO_FAILURE' value='12'/>
	<enumerator name='VDEV_AUX_BAD_LOG' value='13'/>
	<enumerator name='VDEV_AUX_EXTERNAL' value='14'/>
	<enumerator name='VDEV_AUX_SPLIT_POOL' value='15'/>
	<enumerator name='VDEV_AUX_BAD_ASHIFT' value='16'/>
	<enumerator name='VDEV_AUX_EXTERNAL_PERSIST' value='17'/>
	<enumerator name='VDEV_AUX_ACTIVE' value='18'/>
	<enumerator name='VDEV_AUX_CHILDREN_OFFLINE' value='19'/>
	<enumerator name='VDEV_AUX_ASHIFT_TOO_BIG' value='20'/>
	</enum-decl>
	<typedef-decl name='vdev_aux_t' type-id='7f5bcca4' id='9d774e0b'/>
	<enum-decl name='pool_scan_func' id='1b092565'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_SCAN_NONE' value='0'/>
	<enumerator name='POOL_SCAN_SCRUB' value='1'/>
	<enumerator name='POOL_SCAN_RESILVER' value='2'/>
	<enumerator name='POOL_SCAN_FUNCS' value='3'/>
	</enum-decl>
	<typedef-decl name='pool_scan_func_t' type-id='1b092565' id='7313fbe2'/>
	<enum-decl name='pool_scrub_cmd' id='a1474cbd'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_SCRUB_NORMAL' value='0'/>
	<enumerator name='POOL_SCRUB_PAUSE' value='1'/>
	<enumerator name='POOL_SCRUB_FLAGS_END' value='2'/>
	</enum-decl>
	<typedef-decl name='pool_scrub_cmd_t' type-id='a1474cbd' id='b51cf3c2'/>
	<enum-decl name='pool_initialize_func' id='5c246ad4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_INITIALIZE_START' value='0'/>
	<enumerator name='POOL_INITIALIZE_CANCEL' value='1'/>
	<enumerator name='POOL_INITIALIZE_SUSPEND' value='2'/>
	<enumerator name='POOL_INITIALIZE_FUNCS' value='3'/>
	</enum-decl>
	<typedef-decl name='pool_initialize_func_t' type-id='5c246ad4' id='7063e1ab'/>
	<enum-decl name='pool_trim_func' id='54ed608a'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_TRIM_START' value='0'/>
	<enumerator name='POOL_TRIM_CANCEL' value='1'/>
	<enumerator name='POOL_TRIM_SUSPEND' value='2'/>
	<enumerator name='POOL_TRIM_FUNCS' value='3'/>
	</enum-decl>
	<typedef-decl name='pool_trim_func_t' type-id='54ed608a' id='b1146b8d'/>
	<enum-decl name='zpool_wait_activity_t' naming-typedef-id='73446457' id='849338e3'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_WAIT_CKPT_DISCARD' value='0'/>
	<enumerator name='ZPOOL_WAIT_FREE' value='1'/>
	<enumerator name='ZPOOL_WAIT_INITIALIZE' value='2'/>
	<enumerator name='ZPOOL_WAIT_REPLACE' value='3'/>
	<enumerator name='ZPOOL_WAIT_REMOVE' value='4'/>
	<enumerator name='ZPOOL_WAIT_RESILVER' value='5'/>
	<enumerator name='ZPOOL_WAIT_SCRUB' value='6'/>
	<enumerator name='ZPOOL_WAIT_TRIM' value='7'/>
	<enumerator name='ZPOOL_WAIT_NUM_ACTIVITIES' value='8'/>
	</enum-decl>
	<typedef-decl name='zpool_wait_activity_t' type-id='849338e3' id='73446457'/>
	<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
	<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
	<pointer-type-def type-id='a093cbb8' size-in-bits='64' id='b13f38c3'/>
	<pointer-type-def type-id='35acf840' size-in-bits='64' id='17f3480d'/>
	<function-decl name='zpool_props_refresh' mangled-name='zpool_props_refresh' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_props_refresh'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_prop_int' mangled-name='zpool_get_prop_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop_int'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='debc6aa3' name='src'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_state_to_name' mangled-name='zpool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_state_to_name'>
	<parameter type-id='35acf840' name='state'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_pool_state_to_name' mangled-name='zpool_pool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_pool_state_to_name'>
	<parameter type-id='084a08a3' name='state'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_state_str' mangled-name='zpool_get_state_str' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state_str'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_prop' mangled-name='zpool_get_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_prop' mangled-name='zpool_set_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_expand_proplist' mangled-name='zpool_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_expand_proplist'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='e4378506' name='plp'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='vdev_expand_proplist' mangled-name='vdev_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_expand_proplist'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='e4378506' name='plp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_get_feature' mangled-name='zpool_prop_get_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_feature'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_open_canfail' mangled-name='zpool_open_canfail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open_canfail'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='pool'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='zpool_open' mangled-name='zpool_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='pool'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='zpool_close' mangled-name='zpool_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_close'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_get_name' mangled-name='zpool_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_name'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_state' mangled-name='zpool_get_state' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_is_draid_spare' mangled-name='zpool_is_draid_spare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_is_draid_spare'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_create' mangled-name='zpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='fsprops'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_destroy' mangled-name='zpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_destroy'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_checkpoint' mangled-name='zpool_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_checkpoint'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_discard_checkpoint' mangled-name='zpool_discard_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_discard_checkpoint'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_add' mangled-name='zpool_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_add'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_export' mangled-name='zpool_export' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_export_force' mangled-name='zpool_export_force' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export_force'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_explain_recover' mangled-name='zpool_explain_recover' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_explain_recover'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='95e97e5e' name='reason'/>
	<parameter type-id='5ce45b60' name='config'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_import' mangled-name='zpool_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='80f4b756' name='newname'/>
	<parameter type-id='26a90f95' name='altroot'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_print_unsup_feat' mangled-name='zpool_print_unsup_feat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_print_unsup_feat'>
	<parameter type-id='5ce45b60' name='config'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_import_props' mangled-name='zpool_import_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_props'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='80f4b756' name='newname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_initialize' mangled-name='zpool_initialize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7063e1ab' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_initialize_wait' mangled-name='zpool_initialize_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize_wait'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7063e1ab' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_trim' mangled-name='zpool_trim' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_trim'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='b1146b8d' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<parameter type-id='b13f38c3' name='trim_flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_scan' mangled-name='zpool_scan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_scan'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7313fbe2' name='func'/>
	<parameter type-id='b51cf3c2' name='cmd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_find_vdev_by_physpath' mangled-name='zpool_find_vdev_by_physpath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev_by_physpath'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='ppath'/>
	<parameter type-id='37e3bd22' name='avail_spare'/>
	<parameter type-id='37e3bd22' name='l2cache'/>
	<parameter type-id='37e3bd22' name='log'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_find_vdev' mangled-name='zpool_find_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='37e3bd22' name='avail_spare'/>
	<parameter type-id='37e3bd22' name='l2cache'/>
	<parameter type-id='37e3bd22' name='log'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_vdev_path_to_guid' mangled-name='zpool_vdev_path_to_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_path_to_guid'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_vdev_online' mangled-name='zpool_vdev_online' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_online'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='17f3480d' name='newstate'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_offline' mangled-name='zpool_vdev_offline' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_offline'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='c19b74c3' name='istmp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	+ <function-decl name='zpool_vdev_remove_wanted' mangled-name='zpool_vdev_remove_wanted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove_wanted'>
	+ <parameter type-id='4c81de99' name='zhp'/>
	+ <parameter type-id='80f4b756' name='path'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='zpool_vdev_fault' mangled-name='zpool_vdev_fault' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_fault'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_degrade' mangled-name='zpool_vdev_degrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_degrade'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_attach' mangled-name='zpool_vdev_attach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_attach'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='old_disk'/>
	<parameter type-id='80f4b756' name='new_disk'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	<parameter type-id='95e97e5e' name='replacing'/>
	<parameter type-id='c19b74c3' name='rebuild'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_detach' mangled-name='zpool_vdev_detach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_detach'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_split' mangled-name='zpool_vdev_split' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_split'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='26a90f95' name='newname'/>
	<parameter type-id='857bb57e' name='newroot'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='325c1e34' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_remove' mangled-name='zpool_vdev_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_remove_cancel' mangled-name='zpool_vdev_remove_cancel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove_cancel'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_indirect_size' mangled-name='zpool_vdev_indirect_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_indirect_size'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='5d6479ae' name='sizep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_clear' mangled-name='zpool_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='5ce45b60' name='rewindnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_clear' mangled-name='zpool_vdev_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_clear'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_reguid' mangled-name='zpool_reguid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reguid'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_reopen_one' mangled-name='zpool_reopen_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reopen_one'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_sync_one' mangled-name='zpool_sync_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_sync_one'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_name' mangled-name='zpool_vdev_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_name'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5ce45b60' name='nv'/>
	<parameter type-id='95e97e5e' name='name_flags'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zpool_get_errlog' mangled-name='zpool_get_errlog' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_errlog'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nverrlistp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_upgrade' mangled-name='zpool_upgrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_upgrade'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='new_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_save_arguments' mangled-name='zfs_save_arguments' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_save_arguments'>
	<parameter type-id='95e97e5e' name='argc'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='26a90f95' name='string'/>
	<parameter type-id='95e97e5e' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_log_history' mangled-name='zpool_log_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_log_history'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='message'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_history' mangled-name='zpool_get_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_history'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nvhisp'/>
	<parameter type-id='5d6479ae' name='off'/>
	<parameter type-id='37e3bd22' name='eof'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_next' mangled-name='zpool_events_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_next'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='857bb57e' name='nvp'/>
	<parameter type-id='7292109c' name='dropped'/>
	<parameter type-id='f0981eeb' name='flags'/>
	<parameter type-id='95e97e5e' name='zevent_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_clear' mangled-name='zpool_events_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_clear'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='7292109c' name='count'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_seek' mangled-name='zpool_events_seek' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_seek'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='9c313c2d' name='eid'/>
	<parameter type-id='95e97e5e' name='zevent_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_obj_to_path' mangled-name='zpool_obj_to_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='dsobj'/>
	<parameter type-id='9c313c2d' name='obj'/>
	<parameter type-id='26a90f95' name='pathname'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_obj_to_path_ds' mangled-name='zpool_obj_to_path_ds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path_ds'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='dsobj'/>
	<parameter type-id='9c313c2d' name='obj'/>
	<parameter type-id='26a90f95' name='pathname'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_wait' mangled-name='zpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='73446457' name='activity'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_wait_status' mangled-name='zpool_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait_status'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='73446457' name='activity'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_bootenv' mangled-name='zpool_set_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_bootenv'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='22cce67b' name='envmap'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_bootenv' mangled-name='zpool_get_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_bootenv'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_load_compat' mangled-name='zpool_load_compat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_load_compat'>
	<parameter type-id='80f4b756' name='compat'/>
	<parameter type-id='37e3bd22' name='features'/>
	<parameter type-id='26a90f95' name='report'/>
	<parameter type-id='b59d7dce' name='rlen'/>
	<return type-id='901b78d1'/>
	</function-decl>
	<function-decl name='zpool_get_vdev_prop_value' mangled-name='zpool_get_vdev_prop_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop_value'>
	<parameter type-id='5ce45b60' name='nvprop'/>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='26a90f95' name='prop_name'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_vdev_prop' mangled-name='zpool_get_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='26a90f95' name='prop_name'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_all_vdev_props' mangled-name='zpool_get_all_vdev_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_all_vdev_props'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_vdev_prop' mangled-name='zpool_set_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_vdev_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_sendrecv.c' language='LANG_C99'>
	<class-decl name='sendflags' size-in-bits='544' is-struct='yes' visibility='default' id='f6aa15be'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='verbosity' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='replicate' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='skipmissing' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='doall' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='fromorigin' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='pad' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='props' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='parsable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='progress' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='largeblock' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='embed_data' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='compress' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='raw' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='backup' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='saved' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sendflags_t' type-id='f6aa15be' id='945467e6'/>
	<typedef-decl name='snapfilter_cb_t' type-id='d2a5e211' id='3d3ffb69'/>
	<class-decl name='recvflags' size-in-bits='448' is-struct='yes' visibility='default' id='34a384dc'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='verbose' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='isprefix' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='istail' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='force' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='canmountoff' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='resumable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='byteswap' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='nomount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='skipholds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='domount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='forceunmount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='heal' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='recvflags_t' type-id='34a384dc' id='9e59d1d4'/>
	<pointer-type-def type-id='f20fbd51' size-in-bits='64' id='a3681dea'/>
	<pointer-type-def type-id='9e59d1d4' size-in-bits='64' id='4ea84b4f'/>
	<pointer-type-def type-id='945467e6' size-in-bits='64' id='8def7735'/>
	<pointer-type-def type-id='3d3ffb69' size-in-bits='64' id='72a26210'/>
	<function-decl name='zfs_send_progress' mangled-name='zfs_send_progress' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_progress'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='5d6479ae' name='bytes_written'/>
	<parameter type-id='5d6479ae' name='blocks_visited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_resume_token_to_nvlist' mangled-name='zfs_send_resume_token_to_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume_token_to_nvlist'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='token'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_send_resume' mangled-name='zfs_send_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='resume_token'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_saved' mangled-name='zfs_send_saved' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_saved'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='resume_token'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send' mangled-name='zfs_send' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='fromsnap'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='72a26210' name='filter_func'/>
	<parameter type-id='eaa32e2f' name='cb_arg'/>
	<parameter type-id='857bb57e' name='debugnvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_one' mangled-name='zfs_send_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_one'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='80f4b756' name='redactbook'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_receive' mangled-name='zfs_receive' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_receive'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='4ea84b4f' name='flags'/>
	<parameter type-id='95e97e5e' name='infd'/>
	<parameter type-id='a3681dea' name='stream_avl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='d2a5e211'>
	<parameter type-id='9200a744'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='c19b74c3'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_status.c' language='LANG_C99'>
	<enum-decl name='zpool_status_t' naming-typedef-id='d3dd6294' id='5e770b40'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_CACHE' value='0'/>
	<enumerator name='ZPOOL_STATUS_MISSING_DEV_R' value='1'/>
	<enumerator name='ZPOOL_STATUS_MISSING_DEV_NR' value='2'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_R' value='3'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_NR' value='4'/>
	<enumerator name='ZPOOL_STATUS_BAD_GUID_SUM' value='5'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_POOL' value='6'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_DATA' value='7'/>
	<enumerator name='ZPOOL_STATUS_FAILING_DEV' value='8'/>
	<enumerator name='ZPOOL_STATUS_VERSION_NEWER' value='9'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_MISMATCH' value='10'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_ACTIVE' value='11'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_REQUIRED' value='12'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_WAIT' value='13'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_CONTINUE' value='14'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_MMP' value='15'/>
	<enumerator name='ZPOOL_STATUS_BAD_LOG' value='16'/>
	<enumerator name='ZPOOL_STATUS_ERRATA' value='17'/>
	<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_READ' value='18'/>
	<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_WRITE' value='19'/>
	<enumerator name='ZPOOL_STATUS_FAULTED_DEV_R' value='20'/>
	<enumerator name='ZPOOL_STATUS_FAULTED_DEV_NR' value='21'/>
	<enumerator name='ZPOOL_STATUS_VERSION_OLDER' value='22'/>
	<enumerator name='ZPOOL_STATUS_FEAT_DISABLED' value='23'/>
	<enumerator name='ZPOOL_STATUS_RESILVERING' value='24'/>
	<enumerator name='ZPOOL_STATUS_OFFLINE_DEV' value='25'/>
	<enumerator name='ZPOOL_STATUS_REMOVED_DEV' value='26'/>
	<enumerator name='ZPOOL_STATUS_REBUILDING' value='27'/>
	<enumerator name='ZPOOL_STATUS_REBUILD_SCRUB' value='28'/>
	<enumerator name='ZPOOL_STATUS_NON_NATIVE_ASHIFT' value='29'/>
	<enumerator name='ZPOOL_STATUS_COMPATIBILITY_ERR' value='30'/>
	<enumerator name='ZPOOL_STATUS_INCOMPATIBLE_FEAT' value='31'/>
	<enumerator name='ZPOOL_STATUS_OK' value='32'/>
	</enum-decl>
	<typedef-decl name='zpool_status_t' type-id='5e770b40' id='d3dd6294'/>
	<enum-decl name='zpool_errata' id='d9abbf54'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_ERRATA_NONE' value='0'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_2094_SCRUB' value='1'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY' value='2'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_6845_ENCRYPTION' value='3'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_8308_ENCRYPTION' value='4'/>
	</enum-decl>
	<typedef-decl name='zpool_errata_t' type-id='d9abbf54' id='688c495b'/>
	<pointer-type-def type-id='80f4b756' size-in-bits='64' id='7d3cd834'/>
	<pointer-type-def type-id='688c495b' size-in-bits='64' id='cec6f2e4'/>
	<function-decl name='zpool_get_status' mangled-name='zpool_get_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_status'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7d3cd834' name='msgid'/>
	<parameter type-id='cec6f2e4' name='errata'/>
	<return type-id='d3dd6294'/>
	</function-decl>
	<function-decl name='zpool_import_status' mangled-name='zpool_import_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_status'>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='7d3cd834' name='msgid'/>
	<parameter type-id='cec6f2e4' name='errata'/>
	<return type-id='d3dd6294'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_util.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='192' id='e41bdf22'>
	<subrange length='6' type-id='7359adad' id='52fa524b'/>
	</array-type-def>
	<type-decl name='variadic parameter type' id='2c1145c5'/>
	<array-type-def dimensions='1' type-id='19cefcee' size-in-bits='160' alignment-in-bits='32' id='3fcf57d2'>
	<subrange length='5' type-id='7359adad' id='53010e10'/>
	</array-type-def>
	<enum-decl name='zfs_get_column_t' naming-typedef-id='19cefcee' id='223bdcaa'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='GET_COL_NONE' value='0'/>
	<enumerator name='GET_COL_NAME' value='1'/>
	<enumerator name='GET_COL_PROPERTY' value='2'/>
	<enumerator name='GET_COL_VALUE' value='3'/>
	<enumerator name='GET_COL_RECVD' value='4'/>
	<enumerator name='GET_COL_SOURCE' value='5'/>
	</enum-decl>
	<typedef-decl name='zfs_get_column_t' type-id='223bdcaa' id='19cefcee'/>
	<class-decl name='vdev_cbdata' size-in-bits='192' is-struct='yes' visibility='default' id='b8006be8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_name_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='cb_names' type-id='9b23c9ad' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='cb_names_count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='vdev_cbdata_t' type-id='b8006be8' id='a9679c94'/>
	<class-decl name='zprop_get_cbdata' size-in-bits='832' is-struct='yes' visibility='default' id='f3d3c319'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_sources' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='cb_columns' type-id='3fcf57d2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='cb_colwidths' type-id='e41bdf22' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='cb_scripted' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='cb_literal' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='cb_first' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='cb_proplist' type-id='3a9b2288' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='cb_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='cb_vdevs' type-id='a9679c94' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_get_cbdata_t' type-id='f3d3c319' id='f3d87113'/>
	<typedef-decl name='zprop_func' type-id='2e711a2a' id='1ec3747a'/>
	<pointer-type-def type-id='9b23c9ad' size-in-bits='64' id='c0563f85'/>
	<pointer-type-def type-id='c70fa2e8' size-in-bits='64' id='2e711a2a'/>
	<pointer-type-def type-id='f3d87113' size-in-bits='64' id='0d2a0670'/>
	<function-decl name='libzfs_errno' mangled-name='libzfs_errno' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_errno'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_error_action' mangled-name='libzfs_error_action' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_action'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='libzfs_error_description' mangled-name='libzfs_error_description' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_description'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_standard_error' mangled-name='zfs_standard_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_standard_error'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='95e97e5e' name='error'/>
	<parameter type-id='80f4b756' name='msg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_print_on_error' mangled-name='libzfs_print_on_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_print_on_error'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='c19b74c3' name='printerr'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_run_process' mangled-name='libzfs_run_process' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_run_process_get_stdout' mangled-name='libzfs_run_process_get_stdout' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='9b23c9ad' name='env'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_run_process_get_stdout_nopath' mangled-name='libzfs_run_process_get_stdout_nopath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout_nopath'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='9b23c9ad' name='env'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_free_str_array' mangled-name='libzfs_free_str_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_free_str_array'>
	<parameter type-id='9b23c9ad' name='strs'/>
	<parameter type-id='95e97e5e' name='count'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_envvar_is_set' mangled-name='libzfs_envvar_is_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_envvar_is_set'>
	<parameter type-id='80f4b756' name='envvar'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='libzfs_init' mangled-name='libzfs_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_init'>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='libzfs_fini' mangled-name='libzfs_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_fini'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_get_handle' mangled-name='zpool_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_handle'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='zfs_get_handle' mangled-name='zfs_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_handle'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='zfs_get_pool_handle' mangled-name='zfs_get_pool_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_handle'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='zfs_path_to_zhandle' mangled-name='zfs_path_to_zhandle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_path_to_zhandle'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='argtype'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zprop_print_one_property' mangled-name='zprop_print_one_property' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_print_one_property'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='0d2a0670' name='cbp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='value'/>
	<parameter type-id='a2256d42' name='sourcetype'/>
	<parameter type-id='80f4b756' name='source'/>
	<parameter type-id='80f4b756' name='recvd_value'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicestrtonum' mangled-name='zfs_nicestrtonum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicestrtonum'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='value'/>
	<parameter type-id='5d6479ae' name='num'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_get_list' mangled-name='zprop_get_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_get_list'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='props'/>
	<parameter type-id='e4378506' name='listp'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_free_list' mangled-name='zprop_free_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_free_list'>
	<parameter type-id='3a9b2288' name='pl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_iter' mangled-name='zprop_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter'>
	<parameter type-id='1ec3747a' name='func'/>
	<parameter type-id='eaa32e2f' name='cb'/>
	<parameter type-id='c19b74c3' name='show_all'/>
	<parameter type-id='c19b74c3' name='ordered'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_version_userland' mangled-name='zfs_version_userland' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_userland'>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_version_print' mangled-name='zfs_version_print' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_print'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='color_start' mangled-name='color_start' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_start'>
	<parameter type-id='80f4b756' name='color'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='color_end' mangled-name='color_end' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_end'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='printf_color' mangled-name='printf_color' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='printf_color'>
	<parameter type-id='80f4b756' name='color'/>
	<parameter type-id='80f4b756' name='format'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='c70fa2e8'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_mount_os.c' language='LANG_C99'>
	<pointer-type-def type-id='7359adad' size-in-bits='64' id='1d2c2b85'/>
	<function-decl name='zfs_parse_mount_options' mangled-name='zfs_parse_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parse_mount_options'>
	<parameter type-id='80f4b756' name='mntopts'/>
	<parameter type-id='1d2c2b85' name='mntflags'/>
	<parameter type-id='1d2c2b85' name='zfsflags'/>
	<parameter type-id='95e97e5e' name='sloppy'/>
	<parameter type-id='26a90f95' name='badopt'/>
	<parameter type-id='26a90f95' name='mtabopt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_adjust_mount_options' mangled-name='zfs_adjust_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_adjust_mount_options'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='mntpoint'/>
	<parameter type-id='26a90f95' name='mntopts'/>
	<parameter type-id='26a90f95' name='mtabopt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_mount_delegation_check' mangled-name='zfs_mount_delegation_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_delegation_check'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_disable_datasets_os' mangled-name='zpool_disable_datasets_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets_os'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='c19b74c3' name='force'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_disable_volume_os' mangled-name='zpool_disable_volume_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_volume_os'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_pool_os.c' language='LANG_C99'>
	<function-decl name='zpool_label_disk' mangled-name='zpool_label_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_util_os.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32768' id='d16c6df4'>
	<subrange length='4096' type-id='7359adad' id='bc1b5ddc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='65536' id='163f6aa5'>
	<subrange length='8192' type-id='7359adad' id='c88f397d'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='128' id='c1c22e6c'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='24' id='d3490169'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<class-decl name='drr_begin' size-in-bits='2432' is-struct='yes' visibility='default' id='09fcdc01'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_magic' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_versioninfo' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_creation_time' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='drr_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_fromguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_toname' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zinject_record' size-in-bits='2816' is-struct='yes' visibility='default' id='3216f820'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zi_objset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zi_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zi_start' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zi_end' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='zi_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='zi_level' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='zi_error' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='zi_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='zi_freq' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='zi_failfast' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='zi_func' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2560'>
	<var-decl name='zi_iotype' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2592'>
	<var-decl name='zi_duration' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2624'>
	<var-decl name='zi_timer' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2688'>
	<var-decl name='zi_nlanes' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2752'>
	<var-decl name='zi_cmd' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2784'>
	<var-decl name='zi_dvas' type-id='8f92235e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zinject_record_t' type-id='3216f820' id='a4301ca6'/>
	<class-decl name='zfs_share' size-in-bits='256' is-struct='yes' visibility='default' id='feb6f2da'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='z_exportdata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='z_sharedata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='z_sharetype' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='z_sharemax' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_share_t' type-id='feb6f2da' id='ee5cec36'/>
	<class-decl name='zfs_cmd' size-in-bits='109952' is-struct='yes' visibility='default' id='3522cd69'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_name' type-id='d16c6df4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32768'>
	<var-decl name='zc_nvlist_src' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32832'>
	<var-decl name='zc_nvlist_src_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32896'>
	<var-decl name='zc_nvlist_dst' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32960'>
	<var-decl name='zc_nvlist_dst_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33024'>
	<var-decl name='zc_nvlist_dst_filled' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33056'>
	<var-decl name='zc_pad2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33088'>
	<var-decl name='zc_history' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33152'>
	<var-decl name='zc_value' type-id='163f6aa5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='98688'>
	<var-decl name='zc_string' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100736'>
	<var-decl name='zc_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100800'>
	<var-decl name='zc_nvlist_conf' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100864'>
	<var-decl name='zc_nvlist_conf_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100928'>
	<var-decl name='zc_cookie' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100992'>
	<var-decl name='zc_objset_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101056'>
	<var-decl name='zc_perm_action' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101120'>
	<var-decl name='zc_history_len' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101184'>
	<var-decl name='zc_history_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101248'>
	<var-decl name='zc_obj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101312'>
	<var-decl name='zc_iflags' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101376'>
	<var-decl name='zc_share' type-id='ee5cec36' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101632'>
	<var-decl name='zc_objset_stats' type-id='b2c14f17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='103936'>
	<var-decl name='zc_begin_record' type-id='09fcdc01' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='106368'>
	<var-decl name='zc_inject_record' type-id='a4301ca6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109184'>
	<var-decl name='zc_defer_destroy' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109216'>
	<var-decl name='zc_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109248'>
	<var-decl name='zc_action_handle' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109312'>
	<var-decl name='zc_cleanup_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109344'>
	<var-decl name='zc_simple' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109352'>
	<var-decl name='zc_pad' type-id='d3490169' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109376'>
	<var-decl name='zc_sendobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109440'>
	<var-decl name='zc_fromobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109504'>
	<var-decl name='zc_createtxg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109568'>
	<var-decl name='zc_stat' type-id='0371a9c7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109888'>
	<var-decl name='zc_zoneid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_cmd_t' type-id='3522cd69' id='a5559cdd'/>
	<class-decl name='zfs_stat' size-in-bits='320' is-struct='yes' visibility='default' id='6417f0b9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zs_gen' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zs_mode' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zs_links' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zs_ctime' type-id='c1c22e6c' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_stat_t' type-id='6417f0b9' id='0371a9c7'/>
	<pointer-type-def type-id='a5559cdd' size-in-bits='64' id='e4ec4540'/>
	<function-decl name='zfs_ioctl' mangled-name='zfs_ioctl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_ioctl'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='95e97e5e' name='request'/>
	<parameter type-id='e4ec4540' name='zc'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_error_init' mangled-name='libzfs_error_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_init'>
	<parameter type-id='95e97e5e' name='error'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps_nvl_os' mangled-name='zfs_destroy_snaps_nvl_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl_os'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='snaps'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_version_kernel' mangled-name='zfs_version_kernel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_kernel'>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_userns' mangled-name='zfs_userns' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_userns'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='nspath'/>
	<parameter type-id='95e97e5e' name='attach'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/os/linux/zutil_device_path_os.c' language='LANG_C99'>
	<function-decl name='zfs_append_partition' mangled-name='zfs_append_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_append_partition'>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='b59d7dce' name='max_len'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_strip_partition' mangled-name='zfs_strip_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_partition'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_strip_path' mangled-name='zfs_strip_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_path'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_get_enclosure_sysfs_path' mangled-name='zfs_get_enclosure_sysfs_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_enclosure_sysfs_path'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_dev_is_dm' mangled-name='zfs_dev_is_dm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_dm'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_dev_is_whole_disk' mangled-name='zfs_dev_is_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_whole_disk'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_get_underlying_path' mangled-name='zfs_get_underlying_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_path'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='is_mpath_whole_disk' mangled-name='is_mpath_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mpath_whole_disk'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/os/linux/zutil_import_os.c' language='LANG_C99'>
	<class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
	<pointer-type-def type-id='b99c00c9' size-in-bits='64' id='13956559'/>
	<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
	<pointer-type-def type-id='640b33ca' size-in-bits='64' id='b32bae08'/>
	<class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
	<function-decl name='zfs_dev_flush' mangled-name='zfs_dev_flush' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_flush'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_default_search_paths' mangled-name='zpool_default_search_paths' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_default_search_paths'>
	<parameter type-id='78c01427' name='count'/>
	<return type-id='13956559'/>
	</function-decl>
	<function-decl name='zfs_device_get_devid' mangled-name='zfs_device_get_devid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_devid'>
	<parameter type-id='b32bae08' name='dev'/>
	<parameter type-id='26a90f95' name='bufptr'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_device_get_physical' mangled-name='zfs_device_get_physical' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_physical'>
	<parameter type-id='b32bae08' name='dev'/>
	<parameter type-id='26a90f95' name='bufptr'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_label_disk_wait' mangled-name='zpool_label_disk_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk_wait'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='95e97e5e' name='timeout_ms'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='update_vdev_config_dev_strs' mangled-name='update_vdev_config_dev_strs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='update_vdev_config_dev_strs'>
	<parameter type-id='5ce45b60' name='nv'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_device_path.c' language='LANG_C99'>
	<typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
	<typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
	<function-decl name='zfs_basename' mangled-name='zfs_basename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_basename'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_dirnamelen' mangled-name='zfs_dirnamelen' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dirnamelen'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='zfs_resolve_shortname' mangled-name='zfs_resolve_shortname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_resolve_shortname'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_strcmp_pathname' mangled-name='zfs_strcmp_pathname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strcmp_pathname'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='80f4b756' name='cmp'/>
	<parameter type-id='95e97e5e' name='wholedisk'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_import.c' language='LANG_C99'>
	<class-decl name='importargs' size-in-bits='448' is-struct='yes' visibility='default' id='7ac83801'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='path' type-id='9b23c9ad' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='paths' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='poolname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='cachefile' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='can_be_active' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='scan' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='policy' type-id='5ce45b60' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='importargs_t' type-id='7ac83801' id='7a842a6b'/>
	<pointer-type-def type-id='7a842a6b' size-in-bits='64' id='07ee4a58'/>
	<pointer-type-def type-id='b1e62775' size-in-bits='64' id='f095e320'/>
	<function-decl name='zpool_read_label' mangled-name='zpool_read_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_read_label'>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='857bb57e' name='config'/>
	<parameter type-id='7292109c' name='num_labels'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_search_import' mangled-name='zpool_search_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_search_import'>
	<parameter type-id='eaa32e2f' name='hdl'/>
	<parameter type-id='07ee4a58' name='import'/>
	<parameter type-id='f095e320' name='pco'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_find_config' mangled-name='zpool_find_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_config'>
	<parameter type-id='eaa32e2f' name='hdl'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='857bb57e' name='configp'/>
	<parameter type-id='07ee4a58' name='args'/>
	<parameter type-id='f095e320' name='pco'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_nicenum.c' language='LANG_C99'>
	<enum-decl name='zfs_nicenum_format' id='29cf1969'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_NICENUM_1024' value='0'/>
	<enumerator name='ZFS_NICENUM_BYTES' value='1'/>
	<enumerator name='ZFS_NICENUM_TIME' value='2'/>
	<enumerator name='ZFS_NICENUM_RAW' value='3'/>
	<enumerator name='ZFS_NICENUM_RAWTIME' value='4'/>
	</enum-decl>
	<function-decl name='zfs_isnumber' mangled-name='zfs_isnumber' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_isnumber'>
	<parameter type-id='80f4b756' name='str'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_nicenum_format' mangled-name='zfs_nicenum_format' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum_format'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<parameter type-id='29cf1969' name='format'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicenum' mangled-name='zfs_nicenum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicetime' mangled-name='zfs_nicetime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicetime'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_niceraw' mangled-name='zfs_niceraw' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_niceraw'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicebytes' mangled-name='zfs_nicebytes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicebytes'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_pool.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='853fd5dc' size-in-bits='32768' id='b505fc2f'>
	<subrange length='64' type-id='7359adad' id='b10be967'/>
	</array-type-def>
	<class-decl name='ddt_stat' size-in-bits='512' is-struct='yes' visibility='default' id='65242dfe'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dds_blocks' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='dds_lsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='dds_psize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='dds_dsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='dds_ref_blocks' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='dds_ref_lsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='dds_ref_psize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='dds_ref_dsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ddt_stat_t' type-id='65242dfe' id='853fd5dc'/>
	<class-decl name='ddt_histogram' size-in-bits='32768' is-struct='yes' visibility='default' id='bc2b3086'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ddh_stat' type-id='b505fc2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ddt_histogram_t' type-id='bc2b3086' id='2d7fe832'/>
	<qualified-type-def type-id='2d7fe832' const='yes' id='ec92d602'/>
	<pointer-type-def type-id='ec92d602' size-in-bits='64' id='932720f8'/>
	<qualified-type-def type-id='853fd5dc' const='yes' id='764c298c'/>
	<pointer-type-def type-id='764c298c' size-in-bits='64' id='dfe59052'/>
	<pointer-type-def type-id='857bb57e' size-in-bits='64' id='75be733c'/>
	<function-decl name='zpool_dump_ddt' mangled-name='zpool_dump_ddt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_dump_ddt'>
	<parameter type-id='dfe59052' name='dds_total'/>
	<parameter type-id='932720f8' name='ddh'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_history_unpack' mangled-name='zpool_history_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_history_unpack'>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='9c313c2d' name='bytes_read'/>
	<parameter type-id='5d6479ae' name='leftover'/>
	<parameter type-id='75be733c' name='records'/>
	<parameter type-id='4dd26a40' name='numrecords'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/avl/avl.c' language='LANG_C99'>
	<typedef-decl name='avl_index_t' type-id='e475ab95' id='fba6cb51'/>
	<pointer-type-def type-id='fba6cb51' size-in-bits='64' id='32adbf30'/>
	<pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
	<function-decl name='avl_walk' mangled-name='avl_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_walk'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='oldnode'/>
	<parameter type-id='95e97e5e' name='left'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_first' mangled-name='avl_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_first'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_last' mangled-name='avl_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_last'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_nearest' mangled-name='avl_nearest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_nearest'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='fba6cb51' name='where'/>
	<parameter type-id='95e97e5e' name='direction'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_find' mangled-name='avl_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_find'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='value'/>
	<parameter type-id='32adbf30' name='where'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_insert' mangled-name='avl_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_data'/>
	<parameter type-id='fba6cb51' name='where'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_insert_here' mangled-name='avl_insert_here' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert_here'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_data'/>
	<parameter type-id='eaa32e2f' name='here'/>
	<parameter type-id='95e97e5e' name='direction'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_add' mangled-name='avl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_add'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_node'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_remove' mangled-name='avl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_remove'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_update_lt' mangled-name='avl_update_lt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_lt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update_gt' mangled-name='avl_update_gt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_gt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update' mangled-name='avl_update' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_swap' mangled-name='avl_swap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_swap'>
	<parameter type-id='a3681dea' name='tree1'/>
	<parameter type-id='a3681dea' name='tree2'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_create' mangled-name='avl_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_create'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='585e1de9' name='compar'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='b59d7dce' name='offset'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_destroy' mangled-name='avl_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_numnodes' mangled-name='avl_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_numnodes'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='avl_is_empty' mangled-name='avl_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_is_empty'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_destroy_nodes' mangled-name='avl_destroy_nodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy_nodes'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='63e171df' name='cookie'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/cityhash.c' language='LANG_C99'>
	<function-decl name='cityhash4' mangled-name='cityhash4' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash4'>
	<parameter type-id='9c313c2d' name='w1'/>
	<parameter type-id='9c313c2d' name='w2'/>
	<parameter type-id='9c313c2d' name='w3'/>
	<parameter type-id='9c313c2d' name='w4'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfeature_common.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='83f29ca2' size-in-bits='16576' id='d95b2b0b'>
	<subrange length='37' type-id='7359adad' id='aa6426fb'/>
	</array-type-def>
	<enum-decl name='spa_feature' id='33ecb627'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='SPA_FEATURE_NONE' value='-1'/>
	<enumerator name='SPA_FEATURE_ASYNC_DESTROY' value='0'/>
	<enumerator name='SPA_FEATURE_EMPTY_BPOBJ' value='1'/>
	<enumerator name='SPA_FEATURE_LZ4_COMPRESS' value='2'/>
	<enumerator name='SPA_FEATURE_MULTI_VDEV_CRASH_DUMP' value='3'/>
	<enumerator name='SPA_FEATURE_SPACEMAP_HISTOGRAM' value='4'/>
	<enumerator name='SPA_FEATURE_ENABLED_TXG' value='5'/>
	<enumerator name='SPA_FEATURE_HOLE_BIRTH' value='6'/>
	<enumerator name='SPA_FEATURE_EXTENSIBLE_DATASET' value='7'/>
	<enumerator name='SPA_FEATURE_EMBEDDED_DATA' value='8'/>
	<enumerator name='SPA_FEATURE_BOOKMARKS' value='9'/>
	<enumerator name='SPA_FEATURE_FS_SS_LIMIT' value='10'/>
	<enumerator name='SPA_FEATURE_LARGE_BLOCKS' value='11'/>
	<enumerator name='SPA_FEATURE_LARGE_DNODE' value='12'/>
	<enumerator name='SPA_FEATURE_SHA512' value='13'/>
	<enumerator name='SPA_FEATURE_SKEIN' value='14'/>
	<enumerator name='SPA_FEATURE_EDONR' value='15'/>
	<enumerator name='SPA_FEATURE_USEROBJ_ACCOUNTING' value='16'/>
	<enumerator name='SPA_FEATURE_ENCRYPTION' value='17'/>
	<enumerator name='SPA_FEATURE_PROJECT_QUOTA' value='18'/>
	<enumerator name='SPA_FEATURE_DEVICE_REMOVAL' value='19'/>
	<enumerator name='SPA_FEATURE_OBSOLETE_COUNTS' value='20'/>
	<enumerator name='SPA_FEATURE_POOL_CHECKPOINT' value='21'/>
	<enumerator name='SPA_FEATURE_SPACEMAP_V2' value='22'/>
	<enumerator name='SPA_FEATURE_ALLOCATION_CLASSES' value='23'/>
	<enumerator name='SPA_FEATURE_RESILVER_DEFER' value='24'/>
	<enumerator name='SPA_FEATURE_BOOKMARK_V2' value='25'/>
	<enumerator name='SPA_FEATURE_REDACTION_BOOKMARKS' value='26'/>
	<enumerator name='SPA_FEATURE_REDACTED_DATASETS' value='27'/>
	<enumerator name='SPA_FEATURE_BOOKMARK_WRITTEN' value='28'/>
	<enumerator name='SPA_FEATURE_LOG_SPACEMAP' value='29'/>
	<enumerator name='SPA_FEATURE_LIVELIST' value='30'/>
	<enumerator name='SPA_FEATURE_DEVICE_REBUILD' value='31'/>
	<enumerator name='SPA_FEATURE_ZSTD_COMPRESS' value='32'/>
	<enumerator name='SPA_FEATURE_DRAID' value='33'/>
	<enumerator name='SPA_FEATURE_ZILSAXATTR' value='34'/>
	<enumerator name='SPA_FEATURE_HEAD_ERRLOG' value='35'/>
	<enumerator name='SPA_FEATURE_BLAKE3' value='36'/>
	<enumerator name='SPA_FEATURES' value='37'/>
	</enum-decl>
	<typedef-decl name='spa_feature_t' type-id='33ecb627' id='d6618c78'/>
	<enum-decl name='zfeature_flags' id='6db816a4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFEATURE_FLAG_READONLY_COMPAT' value='1'/>
	<enumerator name='ZFEATURE_FLAG_MOS' value='2'/>
	<enumerator name='ZFEATURE_FLAG_ACTIVATE_ON_ENABLE' value='4'/>
	<enumerator name='ZFEATURE_FLAG_PER_DATASET' value='8'/>
	</enum-decl>
	<typedef-decl name='zfeature_flags_t' type-id='6db816a4' id='fc329033'/>
	<enum-decl name='zfeature_type' id='c4fa2355'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFEATURE_TYPE_BOOLEAN' value='0'/>
	<enumerator name='ZFEATURE_TYPE_UINT64_ARRAY' value='1'/>
	<enumerator name='ZFEATURE_NUM_TYPES' value='2'/>
	</enum-decl>
	<typedef-decl name='zfeature_type_t' type-id='c4fa2355' id='732d2bb2'/>
	<class-decl name='zfeature_info' size-in-bits='448' is-struct='yes' visibility='default' id='1178d146'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='fi_feature' type-id='d6618c78' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='fi_uname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='fi_guid' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='fi_desc' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fi_flags' type-id='fc329033' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='fi_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='fi_type' type-id='732d2bb2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='fi_depends' type-id='1acff326' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfeature_info_t' type-id='1178d146' id='83f29ca2'/>
	<class-decl name='zfs_mod_supported_features' size-in-bits='128' is-struct='yes' visibility='default' id='3eee3342'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tree' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='all_features' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<qualified-type-def type-id='d6618c78' const='yes' id='81a65028'/>
	<pointer-type-def type-id='81a65028' size-in-bits='64' id='1acff326'/>
	<qualified-type-def type-id='3eee3342' const='yes' id='0c1d5bbb'/>
	<pointer-type-def type-id='0c1d5bbb' size-in-bits='64' id='a3372543'/>
	<pointer-type-def type-id='d6618c78' size-in-bits='64' id='a8425263'/>
	<var-decl name='spa_feature_table' type-id='d95b2b0b' mangled-name='spa_feature_table' visibility='default' elf-symbol-id='spa_feature_table'/>
	<var-decl name='zfeature_checks_disable' type-id='c19b74c3' mangled-name='zfeature_checks_disable' visibility='default' elf-symbol-id='zfeature_checks_disable'/>
	<function-decl name='zfeature_is_valid_guid' mangled-name='zfeature_is_valid_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_valid_guid'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfeature_is_supported' mangled-name='zfeature_is_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_supported'>
	<parameter type-id='80f4b756' name='guid'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfeature_lookup_guid' mangled-name='zfeature_lookup_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_guid'>
	<parameter type-id='80f4b756' name='guid'/>
	<parameter type-id='a8425263' name='res'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfeature_lookup_name' mangled-name='zfeature_lookup_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_name'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a8425263' name='res'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfeature_depends_on' mangled-name='zfeature_depends_on' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_depends_on'>
	<parameter type-id='d6618c78' name='fid'/>
	<parameter type-id='d6618c78' name='check'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_mod_supported' mangled-name='zfs_mod_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mod_supported'>
	<parameter type-id='80f4b756' name='scope'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_feature_init' mangled-name='zpool_feature_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_feature_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_comutil.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='2624' id='5ce15418'>
	<subrange length='41' type-id='7359adad' id='cb834f44'/>
	</array-type-def>
	<class-decl name='zpool_load_policy' size-in-bits='256' is-struct='yes' visibility='default' id='2f65b36f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zlp_rewind' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zlp_maxmeta' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zlp_maxdata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zlp_txg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zpool_load_policy_t' type-id='2f65b36f' id='d11b7617'/>
	<qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
	<pointer-type-def type-id='d11b7617' size-in-bits='64' id='23432aaa'/>
	<var-decl name='zfs_history_event_names' type-id='5ce15418' mangled-name='zfs_history_event_names' visibility='default' elf-symbol-id='zfs_history_event_names'/>
	<function-decl name='zfs_allocatable_devs' mangled-name='zfs_allocatable_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_allocatable_devs'>
	<parameter type-id='5ce45b60' name='nv'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_special_devs' mangled-name='zfs_special_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_special_devs'>
	<parameter type-id='5ce45b60' name='nv'/>
	<parameter type-id='80f4b756' name='type'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_get_load_policy' mangled-name='zpool_get_load_policy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_load_policy'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='23432aaa' name='zlpp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_zpl_version_map' mangled-name='zfs_zpl_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_zpl_version_map'>
	<parameter type-id='95e97e5e' name='spa_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_spa_version_map' mangled-name='zfs_spa_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version_map'>
	<parameter type-id='95e97e5e' name='zpl_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_dataset_name_hidden' mangled-name='zfs_dataset_name_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_name_hidden'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_deleg.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='fa1870fd' size-in-bits='infinite' id='7c00e69d'>
	<subrange length='infinite' id='031f2035'/>
	</array-type-def>
	<enum-decl name='zfs_deleg_who_type_t' naming-typedef-id='36d4bd5a' id='b5fa5816'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_DELEG_WHO_UNKNOWN' value='0'/>
	<enumerator name='ZFS_DELEG_USER' value='117'/>
	<enumerator name='ZFS_DELEG_USER_SETS' value='85'/>
	<enumerator name='ZFS_DELEG_GROUP' value='103'/>
	<enumerator name='ZFS_DELEG_GROUP_SETS' value='71'/>
	<enumerator name='ZFS_DELEG_EVERYONE' value='101'/>
	<enumerator name='ZFS_DELEG_EVERYONE_SETS' value='69'/>
	<enumerator name='ZFS_DELEG_CREATE' value='99'/>
	<enumerator name='ZFS_DELEG_CREATE_SETS' value='67'/>
	<enumerator name='ZFS_DELEG_NAMED_SET' value='115'/>
	<enumerator name='ZFS_DELEG_NAMED_SET_SETS' value='83'/>
	</enum-decl>
	<typedef-decl name='zfs_deleg_who_type_t' type-id='b5fa5816' id='36d4bd5a'/>
	<enum-decl name='zfs_deleg_note_t' naming-typedef-id='4613c173' id='729d4547'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_DELEG_NOTE_CREATE' value='0'/>
	<enumerator name='ZFS_DELEG_NOTE_DESTROY' value='1'/>
	<enumerator name='ZFS_DELEG_NOTE_SNAPSHOT' value='2'/>
	<enumerator name='ZFS_DELEG_NOTE_ROLLBACK' value='3'/>
	<enumerator name='ZFS_DELEG_NOTE_CLONE' value='4'/>
	<enumerator name='ZFS_DELEG_NOTE_PROMOTE' value='5'/>
	<enumerator name='ZFS_DELEG_NOTE_RENAME' value='6'/>
	<enumerator name='ZFS_DELEG_NOTE_SEND' value='7'/>
	<enumerator name='ZFS_DELEG_NOTE_RECEIVE' value='8'/>
	<enumerator name='ZFS_DELEG_NOTE_ALLOW' value='9'/>
	<enumerator name='ZFS_DELEG_NOTE_USERPROP' value='10'/>
	<enumerator name='ZFS_DELEG_NOTE_MOUNT' value='11'/>
	<enumerator name='ZFS_DELEG_NOTE_SHARE' value='12'/>
	<enumerator name='ZFS_DELEG_NOTE_USERQUOTA' value='13'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPQUOTA' value='14'/>
	<enumerator name='ZFS_DELEG_NOTE_USERUSED' value='15'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPUSED' value='16'/>
	<enumerator name='ZFS_DELEG_NOTE_USEROBJQUOTA' value='17'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPOBJQUOTA' value='18'/>
	<enumerator name='ZFS_DELEG_NOTE_USEROBJUSED' value='19'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPOBJUSED' value='20'/>
	<enumerator name='ZFS_DELEG_NOTE_HOLD' value='21'/>
	<enumerator name='ZFS_DELEG_NOTE_RELEASE' value='22'/>
	<enumerator name='ZFS_DELEG_NOTE_DIFF' value='23'/>
	<enumerator name='ZFS_DELEG_NOTE_BOOKMARK' value='24'/>
	<enumerator name='ZFS_DELEG_NOTE_LOAD_KEY' value='25'/>
	<enumerator name='ZFS_DELEG_NOTE_CHANGE_KEY' value='26'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTUSED' value='27'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTQUOTA' value='28'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJUSED' value='29'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJQUOTA' value='30'/>
	<enumerator name='ZFS_DELEG_NOTE_NONE' value='31'/>
	</enum-decl>
	<typedef-decl name='zfs_deleg_note_t' type-id='729d4547' id='4613c173'/>
	<class-decl name='zfs_deleg_perm_tab' size-in-bits='128' is-struct='yes' visibility='default' id='5aa05c1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='z_perm' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='z_note' type-id='4613c173' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_deleg_perm_tab_t' type-id='5aa05c1f' id='f3f851ad'/>
	<qualified-type-def type-id='f3f851ad' const='yes' id='fa1870fd'/>
	<var-decl name='zfs_deleg_perm_tab' type-id='7c00e69d' mangled-name='zfs_deleg_perm_tab' visibility='default' elf-symbol-id='zfs_deleg_perm_tab'/>
	<function-decl name='zfs_deleg_canonicalize_perm' mangled-name='zfs_deleg_canonicalize_perm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_canonicalize_perm'>
	<parameter type-id='80f4b756' name='perm'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_deleg_verify_nvlist' mangled-name='zfs_deleg_verify_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_verify_nvlist'>
	<parameter type-id='5ce45b60' name='nvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_deleg_whokey' mangled-name='zfs_deleg_whokey' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_whokey'>
	<parameter type-id='26a90f95' name='attr'/>
	<parameter type-id='36d4bd5a' name='type'/>
	<parameter type-id='a84c031d' name='inheritchr'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='256' id='85c64d26'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='512' id='c5d13f42'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='90dbb6d6' size-in-bits='2048' id='16582e69'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='8240361c' size-in-bits='1024' id='481f90b1'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='7c1ab40c' size-in-bits='512' id='cbd91ec1'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='6d059eaa' size-in-bits='1024' id='729b6ebb'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<class-decl name='zio_cksum' size-in-bits='256' is-struct='yes' visibility='default' id='1d53e28b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_word' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_cksum_t' type-id='1d53e28b' id='39730d0b'/>
	<enum-decl name='zio_byteorder_t' naming-typedef-id='595a65ec' id='fc861be0'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZIO_CHECKSUM_NATIVE' value='0'/>
	<enumerator name='ZIO_CHECKSUM_BYTESWAP' value='1'/>
	</enum-decl>
	<typedef-decl name='zio_byteorder_t' type-id='fc861be0' id='595a65ec'/>
	<class-decl name='zio_abd_checksum_data' size-in-bits='256' is-struct='yes' visibility='default' id='4bf4b004'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='acd_byteorder' type-id='595a65ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='acd_ctx' type-id='0f7df99e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='acd_zcp' type-id='c24fc2ee' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='acd_private' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_abd_checksum_data_t' type-id='4bf4b004' id='74e39470'/>
	<typedef-decl name='zio_abd_checksum_init_t' type-id='a5444274' id='029a8ebe'/>
	<typedef-decl name='zio_abd_checksum_fini_t' type-id='a5444274' id='d6fd5c6c'/>
	<typedef-decl name='zio_abd_checksum_iter_t' type-id='f4a1892e' id='cefa0f4a'/>
	<class-decl name='zio_abd_checksum_func' size-in-bits='192' is-struct='yes' visibility='default' id='aa14691a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='acf_init' type-id='0bcca125' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='acf_fini' type-id='bfe36153' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='acf_iter' type-id='1e276399' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_abd_checksum_func_t' type-id='3f8e8d11' id='c2eb138a'/>
	<class-decl name='zfs_fletcher_superscalar' size-in-bits='256' is-struct='yes' visibility='default' id='28efb250'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_superscalar_t' type-id='28efb250' id='6d059eaa'/>
	<class-decl name='zfs_fletcher_sse' size-in-bits='128' is-struct='yes' visibility='default' id='acd4019a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='c1c22e6c' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_sse_t' type-id='acd4019a' id='7c1ab40c'/>
	<class-decl name='zfs_fletcher_avx' size-in-bits='256' is-struct='yes' visibility='default' id='8c208dfa'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_avx_t' type-id='8c208dfa' id='8240361c'/>
	<class-decl name='zfs_fletcher_avx512' size-in-bits='512' is-struct='yes' visibility='default' id='c6d0c382'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='c5d13f42' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_avx512_t' type-id='c6d0c382' id='90dbb6d6'/>
	<union-decl name='fletcher_4_ctx' size-in-bits='2048' visibility='default' id='1f951ade'>
	<data-member access='public'>
	<var-decl name='scalar' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='superscalar' type-id='729b6ebb' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='sse' type-id='cbd91ec1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='avx' type-id='481f90b1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='avx512' type-id='16582e69' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='fletcher_4_ctx_t' type-id='1f951ade' id='4b675395'/>
	<qualified-type-def type-id='aa14691a' const='yes' id='3f8e8d11'/>
	<pointer-type-def type-id='4b675395' size-in-bits='64' id='0f7df99e'/>
	<pointer-type-def type-id='74e39470' size-in-bits='64' id='eefe7427'/>
	<pointer-type-def type-id='d6fd5c6c' size-in-bits='64' id='bfe36153'/>
	<pointer-type-def type-id='029a8ebe' size-in-bits='64' id='0bcca125'/>
	<pointer-type-def type-id='cefa0f4a' size-in-bits='64' id='1e276399'/>
	<pointer-type-def type-id='39730d0b' size-in-bits='64' id='c24fc2ee'/>
	<var-decl name='fletcher_4_abd_ops' type-id='c2eb138a' mangled-name='fletcher_4_abd_ops' visibility='default' elf-symbol-id='fletcher_4_abd_ops'/>
	<function-decl name='fletcher_init' mangled-name='fletcher_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_init'>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_2_incremental_native' mangled-name='fletcher_2_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_2_native' mangled-name='fletcher_2_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_2_incremental_byteswap' mangled-name='fletcher_2_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_2_byteswap' mangled-name='fletcher_2_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_impl_set' mangled-name='fletcher_4_impl_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_impl_set'>
	<parameter type-id='80f4b756' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_4_native' mangled-name='fletcher_4_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_native_varsize' mangled-name='fletcher_4_native_varsize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native_varsize'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_byteswap' mangled-name='fletcher_4_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_incremental_native' mangled-name='fletcher_4_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_4_incremental_byteswap' mangled-name='fletcher_4_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_4_init' mangled-name='fletcher_4_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_fini' mangled-name='fletcher_4_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_fini'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='f4a1892e'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='a5444274'>
	<parameter type-id='eefe7427'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_avx512.c' language='LANG_C99'>
	<typedef-decl name='fletcher_4_init_f' type-id='173aa527' id='b9ae1656'/>
	<typedef-decl name='fletcher_4_fini_f' type-id='0ad5b8a8' id='c4c1f4fc'/>
	<typedef-decl name='fletcher_4_compute_f' type-id='38147eff' id='ad1dc4cb'/>
	<class-decl name='fletcher_4_func' size-in-bits='512' is-struct='yes' visibility='default' id='57f479a0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='init_native' type-id='b9ae1656' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='fini_native' type-id='c4c1f4fc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='compute_native' type-id='ad1dc4cb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='init_byteswap' type-id='b9ae1656' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fini_byteswap' type-id='c4c1f4fc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='compute_byteswap' type-id='ad1dc4cb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='valid' type-id='297d38bc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='name' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='fletcher_4_ops_t' type-id='57f479a0' id='eba91718'/>
	<qualified-type-def type-id='eba91718' const='yes' id='9eeabdc8'/>
	<pointer-type-def type-id='e9e61702' size-in-bits='64' id='297d38bc'/>
	<pointer-type-def type-id='fe40251b' size-in-bits='64' id='173aa527'/>
	<pointer-type-def type-id='17fb1f83' size-in-bits='64' id='38147eff'/>
	<pointer-type-def type-id='fb39e25e' size-in-bits='64' id='0ad5b8a8'/>
	<var-decl name='fletcher_4_avx512f_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512f_ops' visibility='default' elf-symbol-id='fletcher_4_avx512f_ops'/>
	<var-decl name='fletcher_4_avx512bw_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512bw_ops' visibility='default' elf-symbol-id='fletcher_4_avx512bw_ops'/>
	<function-type size-in-bits='64' id='e9e61702'>
	<return type-id='c19b74c3'/>
	</function-type>
	<function-type size-in-bits='64' id='fe40251b'>
	<parameter type-id='0f7df99e'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='17fb1f83'>
	<parameter type-id='0f7df99e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='fb39e25e'>
	<parameter type-id='0f7df99e'/>
	<parameter type-id='c24fc2ee'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_intel.c' language='LANG_C99'>
	<var-decl name='fletcher_4_avx2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx2_ops' visibility='default' elf-symbol-id='fletcher_4_avx2_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_sse.c' language='LANG_C99'>
	<var-decl name='fletcher_4_sse2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_sse2_ops' visibility='default' elf-symbol-id='fletcher_4_sse2_ops'/>
	<var-decl name='fletcher_4_ssse3_ops' type-id='9eeabdc8' mangled-name='fletcher_4_ssse3_ops' visibility='default' elf-symbol-id='fletcher_4_ssse3_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_superscalar.c' language='LANG_C99'>
	<var-decl name='fletcher_4_superscalar_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_superscalar4.c' language='LANG_C99'>
	<var-decl name='fletcher_4_superscalar4_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar4_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar4_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_namecheck.c' language='LANG_C99'>
	<enum-decl name='namecheck_err_t' naming-typedef-id='8e0af06e' id='f43bbcda'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='NAME_ERR_LEADING_SLASH' value='0'/>
	<enumerator name='NAME_ERR_EMPTY_COMPONENT' value='1'/>
	<enumerator name='NAME_ERR_TRAILING_SLASH' value='2'/>
	<enumerator name='NAME_ERR_INVALCHAR' value='3'/>
	<enumerator name='NAME_ERR_MULTIPLE_DELIMITERS' value='4'/>
	<enumerator name='NAME_ERR_NOLETTER' value='5'/>
	<enumerator name='NAME_ERR_RESERVED' value='6'/>
	<enumerator name='NAME_ERR_DISKLIKE' value='7'/>
	<enumerator name='NAME_ERR_TOOLONG' value='8'/>
	<enumerator name='NAME_ERR_SELF_REF' value='9'/>
	<enumerator name='NAME_ERR_PARENT_REF' value='10'/>
	<enumerator name='NAME_ERR_NO_AT' value='11'/>
	<enumerator name='NAME_ERR_NO_POUND' value='12'/>
	</enum-decl>
	<typedef-decl name='namecheck_err_t' type-id='f43bbcda' id='8e0af06e'/>
	<pointer-type-def type-id='8e0af06e' size-in-bits='64' id='053457bd'/>
	<var-decl name='zfs_max_dataset_nesting' type-id='95e97e5e' mangled-name='zfs_max_dataset_nesting' visibility='default' elf-symbol-id='zfs_max_dataset_nesting'/>
	<function-decl name='get_dataset_depth' mangled-name='get_dataset_depth' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_dataset_depth'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_component_namecheck' mangled-name='zfs_component_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_component_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='permset_namecheck' mangled-name='permset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='permset_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dataset_nestcheck' mangled-name='dataset_nestcheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_nestcheck'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='entity_namecheck' mangled-name='entity_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='entity_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dataset_namecheck' mangled-name='dataset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='bookmark_namecheck' mangled-name='bookmark_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='bookmark_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='snapshot_namecheck' mangled-name='snapshot_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='snapshot_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='mountpoint_namecheck' mangled-name='mountpoint_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mountpoint_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pool_namecheck' mangled-name='pool_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='pool_namecheck'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_prop.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='768' id='bcc77e38'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<enum-decl name='zprop_type_t' naming-typedef-id='31429eff' id='87676253'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='PROP_TYPE_NUMBER' value='0'/>
	<enumerator name='PROP_TYPE_STRING' value='1'/>
	<enumerator name='PROP_TYPE_INDEX' value='2'/>
	</enum-decl>
	<typedef-decl name='zprop_type_t' type-id='87676253' id='31429eff'/>
	<enum-decl name='zprop_attr_t' naming-typedef-id='999701cc' id='77d05200'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='PROP_DEFAULT' value='0'/>
	<enumerator name='PROP_READONLY' value='1'/>
	<enumerator name='PROP_INHERIT' value='2'/>
	<enumerator name='PROP_ONETIME' value='3'/>
	<enumerator name='PROP_ONETIME_DEFAULT' value='4'/>
	</enum-decl>
	<typedef-decl name='zprop_attr_t' type-id='77d05200' id='999701cc'/>
	<class-decl name='zfs_index' size-in-bits='128' is-struct='yes' visibility='default' id='87957af9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pi_name' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pi_value' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_index_t' type-id='87957af9' id='64636ce3'/>
	<class-decl name='zprop_desc_t' size-in-bits='640' is-struct='yes' naming-typedef-id='ffa52b96' visibility='default' id='bbff5e4b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pd_name' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pd_propnum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='pd_proptype' type-id='31429eff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='pd_strdefault' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='pd_numdefault' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='pd_attr' type-id='999701cc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='pd_types' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='pd_values' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='pd_colname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='pd_rightalign' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='449'>
	<var-decl name='pd_visible' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='450'>
	<var-decl name='pd_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='451'>
	<var-decl name='pd_always_flex' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='pd_table' type-id='c8bc397b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='pd_table_size' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_desc_t' type-id='bbff5e4b' id='ffa52b96'/>
	<qualified-type-def type-id='64636ce3' const='yes' id='072f7953'/>
	<pointer-type-def type-id='072f7953' size-in-bits='64' id='c8bc397b'/>
	<pointer-type-def type-id='ffa52b96' size-in-bits='64' id='76c8174b'/>
	<var-decl name='zfs_userquota_prop_prefixes' type-id='bcc77e38' mangled-name='zfs_userquota_prop_prefixes' visibility='default' elf-symbol-id='zfs_userquota_prop_prefixes'/>
	<function-decl name='zfs_prop_get_table' mangled-name='zfs_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='zfs_prop_init' mangled-name='zfs_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_prop_delegatable' mangled-name='zfs_prop_delegatable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_delegatable'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_name_to_prop' mangled-name='zfs_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_to_prop'>
	<parameter type-id='80f4b756' name='propname'/>
	<return type-id='58603c44'/>
	</function-decl>
	<function-decl name='zfs_prop_user' mangled-name='zfs_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_user'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_userquota' mangled-name='zfs_prop_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_userquota'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_written' mangled-name='zfs_prop_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_written'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_string_to_index' mangled-name='zfs_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_string_to_index'>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_index_to_string' mangled-name='zfs_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_index_to_string'>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='9c313c2d' name='index'/>
	<parameter type-id='7d3cd834' name='string'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_random_value' mangled-name='zfs_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_random_value'>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_valid_for_type' mangled-name='zfs_prop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_for_type'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='2e45de5d' name='types'/>
	<parameter type-id='c19b74c3' name='headcheck'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_get_type' mangled-name='zfs_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_type'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='zfs_prop_readonly' mangled-name='zfs_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_readonly'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_visible' mangled-name='zfs_prop_visible' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_visible'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_setonce' mangled-name='zfs_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_setonce'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_default_string' mangled-name='zfs_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_string'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_default_numeric' mangled-name='zfs_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_numeric'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_to_name' mangled-name='zfs_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_to_name'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_inheritable' mangled-name='zfs_prop_inheritable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inheritable'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_encryption_key_param' mangled-name='zfs_prop_encryption_key_param' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_encryption_key_param'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_valid_keylocation' mangled-name='zfs_prop_valid_keylocation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_keylocation'>
	<parameter type-id='80f4b756' name='str'/>
	<parameter type-id='c19b74c3' name='encrypted'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_values' mangled-name='zfs_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_values'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_is_string' mangled-name='zfs_prop_is_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_is_string'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_column_name' mangled-name='zfs_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_column_name'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_align_right' mangled-name='zfs_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_align_right'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zpool_prop.c' language='LANG_C99'>
	<function-decl name='zpool_prop_get_table' mangled-name='zpool_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='zpool_prop_init' mangled-name='zpool_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_name_to_prop' mangled-name='zpool_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_name_to_prop'>
	<parameter type-id='80f4b756' name='propname'/>
	<return type-id='5d0c23fb'/>
	</function-decl>
	<function-decl name='zpool_prop_to_name' mangled-name='zpool_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_to_name'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_get_type' mangled-name='zpool_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_type'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='zpool_prop_readonly' mangled-name='zpool_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_readonly'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_setonce' mangled-name='zpool_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_setonce'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_default_string' mangled-name='zpool_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_string'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_default_numeric' mangled-name='zpool_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_numeric'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_prop_feature' mangled-name='zpool_prop_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_feature'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_unsupported' mangled-name='zpool_prop_unsupported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_unsupported'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_string_to_index' mangled-name='zpool_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_string_to_index'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_index_to_string' mangled-name='zpool_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_index_to_string'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='9c313c2d' name='index'/>
	<parameter type-id='7d3cd834' name='string'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_random_value' mangled-name='zpool_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_random_value'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_prop_values' mangled-name='zpool_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_values'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_column_name' mangled-name='zpool_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_column_name'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_align_right' mangled-name='zpool_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_align_right'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_get_table' mangled-name='vdev_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='vdev_prop_init' mangled-name='vdev_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='vdev_name_to_prop' mangled-name='vdev_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_name_to_prop'>
	<parameter type-id='80f4b756' name='propname'/>
	<return type-id='5aa5c90c'/>
	</function-decl>
	<function-decl name='vdev_prop_user' mangled-name='vdev_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_user'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_to_name' mangled-name='vdev_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_to_name'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_get_type' mangled-name='vdev_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_type'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='vdev_prop_readonly' mangled-name='vdev_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_readonly'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_default_string' mangled-name='vdev_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_string'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_default_numeric' mangled-name='vdev_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_numeric'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='vdev_prop_string_to_index' mangled-name='vdev_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_string_to_index'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='vdev_prop_index_to_string' mangled-name='vdev_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_index_to_string'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='9c313c2d' name='index'/>
	<parameter type-id='7d3cd834' name='string'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_vdev' mangled-name='zpool_prop_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_vdev'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_random_value' mangled-name='vdev_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_random_value'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='vdev_prop_values' mangled-name='vdev_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_values'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_column_name' mangled-name='vdev_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_column_name'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_align_right' mangled-name='vdev_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_align_right'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zprop_common.c' language='LANG_C99'>
	<function-decl name='zprop_register_impl' mangled-name='zprop_register_impl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_impl'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='31429eff' name='type'/>
	<parameter type-id='9c313c2d' name='numdefault'/>
	<parameter type-id='80f4b756' name='strdefault'/>
	<parameter type-id='999701cc' name='attr'/>
	<parameter type-id='95e97e5e' name='objset_types'/>
	<parameter type-id='80f4b756' name='values'/>
	<parameter type-id='80f4b756' name='colname'/>
	<parameter type-id='c19b74c3' name='rightalign'/>
	<parameter type-id='c19b74c3' name='visible'/>
	<parameter type-id='c19b74c3' name='flex'/>
	<parameter type-id='c8bc397b' name='idx_tbl'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_string' mangled-name='zprop_register_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_string'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='80f4b756' name='def'/>
	<parameter type-id='999701cc' name='attr'/>
	<parameter type-id='95e97e5e' name='objset_types'/>
	<parameter type-id='80f4b756' name='values'/>
	<parameter type-id='80f4b756' name='colname'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_number' mangled-name='zprop_register_number' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_number'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9c313c2d' name='def'/>
	<parameter type-id='999701cc' name='attr'/>
	<parameter type-id='95e97e5e' name='objset_types'/>
	<parameter type-id='80f4b756' name='values'/>
	<parameter type-id='80f4b756' name='colname'/>
	<parameter type-id='c19b74c3' name='flex'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_index' mangled-name='zprop_register_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_index'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9c313c2d' name='def'/>
	<parameter type-id='999701cc' name='attr'/>
	<parameter type-id='95e97e5e' name='objset_types'/>
	<parameter type-id='80f4b756' name='values'/>
	<parameter type-id='80f4b756' name='colname'/>
	<parameter type-id='c8bc397b' name='idx_tbl'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_hidden' mangled-name='zprop_register_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_hidden'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='31429eff' name='type'/>
	<parameter type-id='999701cc' name='attr'/>
	<parameter type-id='95e97e5e' name='objset_types'/>
	<parameter type-id='80f4b756' name='colname'/>
	<parameter type-id='c19b74c3' name='flex'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_iter_common' mangled-name='zprop_iter_common' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter_common'>
	<parameter type-id='1ec3747a' name='func'/>
	<parameter type-id='eaa32e2f' name='cb'/>
	<parameter type-id='c19b74c3' name='show_all'/>
	<parameter type-id='c19b74c3' name='ordered'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_name_to_prop' mangled-name='zprop_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_name_to_prop'>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_string_to_index' mangled-name='zprop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_string_to_index'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_index_to_string' mangled-name='zprop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_index_to_string'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='9c313c2d' name='index'/>
	<parameter type-id='7d3cd834' name='string'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_random_value' mangled-name='zprop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_random_value'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zprop_values' mangled-name='zprop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_values'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zprop_valid_for_type' mangled-name='zprop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_for_type'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='c19b74c3' name='headcheck'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zprop_valid_char' mangled-name='zprop_valid_char' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_char'>
	<parameter type-id='a84c031d' name='c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_width' mangled-name='zprop_width' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_width'>
	<parameter type-id='95e97e5e' name='prop'/>
	<parameter type-id='37e3bd22' name='fixed'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	</abi-corpus>
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c b/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c
	index 047a254882f7..133b3b358831 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c
	@@ -1,5589 +1,5592 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2019 Joyent, Inc.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2012 DEY Storage Systems, Inc. All rights reserved.
	* Copyright (c) 2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
	* Copyright (c) 2013 Martin Matuska. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright 2017-2018 RackTop Systems.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
	* Copyright (c) 2021 Matt Fiddaman
	*/

	#include <ctype.h>
	#include <errno.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <strings.h>
	#include <unistd.h>
	#include <stddef.h>
	#include <zone.h>
	#include <fcntl.h>
	#include <sys/mntent.h>
	#include <sys/mount.h>
	#include <pwd.h>
	#include <grp.h>
	#ifdef HAVE_IDMAP
	#include <idmap.h>
	#include <aclutils.h>
	#include <directory.h>
	#endif /* HAVE_IDMAP */

	#include <sys/dnode.h>
	#include <sys/spa.h>
	#include <sys/zap.h>
	#include <sys/dsl_crypt.h>
	#include <libzfs.h>
	#include <libzutil.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "libzfs_impl.h"
	#include "zfs_deleg.h"

	static int userquota_propname_decode(const char *propname, boolean_t zoned,
	zfs_userquota_prop_t typep, char domain, int domainlen, uint64_t *ridp);

	/*
	* Given a single type (not a mask of types), return the type in a human
	* readable form.
	*/
	const char *
	zfs_type_to_name(zfs_type_t type)
	{
	switch (type) {
	case ZFS_TYPE_FILESYSTEM:
	return (dgettext(TEXT_DOMAIN, "filesystem"));
	case ZFS_TYPE_SNAPSHOT:
	return (dgettext(TEXT_DOMAIN, "snapshot"));
	case ZFS_TYPE_VOLUME:
	return (dgettext(TEXT_DOMAIN, "volume"));
	case ZFS_TYPE_POOL:
	return (dgettext(TEXT_DOMAIN, "pool"));
	case ZFS_TYPE_BOOKMARK:
	return (dgettext(TEXT_DOMAIN, "bookmark"));
	default:
	assert(!"unhandled zfs_type_t");
	}

	return (NULL);
	}

	/*
	* Validate a ZFS path. This is used even before trying to open the dataset, to
	* provide a more meaningful error message. We call zfs_error_aux() to
	* explain exactly why the name was not valid.
	*/
	int
	zfs_validate_name(libzfs_handle_t hdl, const char path, int type,
	boolean_t modifying)
	{
	namecheck_err_t why;
	char what;

	if (!(type & ZFS_TYPE_SNAPSHOT) && strchr(path, '@') != NULL) {
	if (hdl != NULL)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"snapshot delimiter '@' is not expected here"));
	return (0);
	}

	if (type == ZFS_TYPE_SNAPSHOT && strchr(path, '@') == NULL) {
	if (hdl != NULL)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"missing '@' delimiter in snapshot name"));
	return (0);
	}

	if (!(type & ZFS_TYPE_BOOKMARK) && strchr(path, '#') != NULL) {
	if (hdl != NULL)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"bookmark delimiter '#' is not expected here"));
	return (0);
	}

	if (type == ZFS_TYPE_BOOKMARK && strchr(path, '#') == NULL) {
	if (hdl != NULL)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"missing '#' delimiter in bookmark name"));
	return (0);
	}

	if (modifying && strchr(path, '%') != NULL) {
	if (hdl != NULL)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid character %c in name"), '%');
	return (0);
	}

	if (entity_namecheck(path, &why, &what) != 0) {
	if (hdl != NULL) {
	switch (why) {
	case NAME_ERR_TOOLONG:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"name is too long"));
	break;

	case NAME_ERR_LEADING_SLASH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"leading slash in name"));
	break;

	case NAME_ERR_EMPTY_COMPONENT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"empty component or misplaced '@'"
	" or '#' delimiter in name"));
	break;

	case NAME_ERR_TRAILING_SLASH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"trailing slash in name"));
	break;

	case NAME_ERR_INVALCHAR:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "invalid character "
	"'%c' in name"), what);
	break;

	case NAME_ERR_MULTIPLE_DELIMITERS:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"multiple '@' and/or '#' delimiters in "
	"name"));
	break;

	case NAME_ERR_NOLETTER:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool doesn't begin with a letter"));
	break;

	case NAME_ERR_RESERVED:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"name is reserved"));
	break;

	case NAME_ERR_DISKLIKE:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"reserved disk name"));
	break;

	case NAME_ERR_SELF_REF:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"self reference, '.' is found in name"));
	break;

	case NAME_ERR_PARENT_REF:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"parent reference, '..' is found in name"));
	break;

	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"(%d) not defined"), why);
	break;
	}
	}

	return (0);
	}

	return (-1);
	}

	int
	zfs_name_valid(const char *name, zfs_type_t type)
	{
	if (type == ZFS_TYPE_POOL)
	return (zpool_name_valid(NULL, B_FALSE, name));
	return (zfs_validate_name(NULL, name, type, B_FALSE));
	}

	/*
	* This function takes the raw DSL properties, and filters out the user-defined
	* properties into a separate nvlist.
	*/
	static nvlist_t *
	process_user_props(zfs_handle_t zhp, nvlist_t props)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvpair_t *elem;
	nvlist_t *nvl;

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) {
	(void) no_memory(hdl);
	return (NULL);
	}

	elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	if (!zfs_prop_user(nvpair_name(elem)))
	continue;

	nvlist_t *propval = fnvpair_value_nvlist(elem);
	if (nvlist_add_nvlist(nvl, nvpair_name(elem), propval) != 0) {
	nvlist_free(nvl);
	(void) no_memory(hdl);
	return (NULL);
	}
	}

	return (nvl);
	}

	static zpool_handle_t *
	zpool_add_handle(zfs_handle_t zhp, const char pool_name)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	zpool_handle_t *zph;

	if ((zph = zpool_open_canfail(hdl, pool_name)) != NULL) {
	if (hdl->libzfs_pool_handles != NULL)
	zph->zpool_next = hdl->libzfs_pool_handles;
	hdl->libzfs_pool_handles = zph;
	}
	return (zph);
	}

	static zpool_handle_t *
	zpool_find_handle(zfs_handle_t zhp, const char pool_name, int len)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	zpool_handle_t *zph = hdl->libzfs_pool_handles;

	while ((zph != NULL) &&
	(strncmp(pool_name, zpool_get_name(zph), len) != 0))
	zph = zph->zpool_next;
	return (zph);
	}

	/*
	* Returns a handle to the pool that contains the provided dataset.
	* If a handle to that pool already exists then that handle is returned.
	* Otherwise, a new handle is created and added to the list of handles.
	*/
	static zpool_handle_t *
	zpool_handle(zfs_handle_t *zhp)
	{
	char *pool_name;
	int len;
	zpool_handle_t *zph;

	len = strcspn(zhp->zfs_name, "/@#") + 1;
	pool_name = zfs_alloc(zhp->zfs_hdl, len);
	(void) strlcpy(pool_name, zhp->zfs_name, len);

	zph = zpool_find_handle(zhp, pool_name, len);
	if (zph == NULL)
	zph = zpool_add_handle(zhp, pool_name);

	free(pool_name);
	return (zph);
	}

	void
	zpool_free_handles(libzfs_handle_t *hdl)
	{
	zpool_handle_t next, zph = hdl->libzfs_pool_handles;

	while (zph != NULL) {
	next = zph->zpool_next;
	zpool_close(zph);
	zph = next;
	}
	hdl->libzfs_pool_handles = NULL;
	}

	/*
	* Utility function to gather stats (objset and zpl) for the given object.
	*/
	static int
	get_stats_ioctl(zfs_handle_t zhp, zfs_cmd_t zc)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;

	(void) strlcpy(zc->zc_name, zhp->zfs_name, sizeof (zc->zc_name));

	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, zc) != 0) {
	if (errno == ENOMEM)
	zcmd_expand_dst_nvlist(hdl, zc);
	else
	return (-1);
	}
	return (0);
	}

	/*
	* Utility function to get the received properties of the given object.
	*/
	static int
	get_recvd_props_ioctl(zfs_handle_t *zhp)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t *recvdprops;
	zfs_cmd_t zc = {"\0"};
	int err;

	zcmd_alloc_dst_nvlist(hdl, &zc, 0);

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_RECVD_PROPS, &zc) != 0) {
	if (errno == ENOMEM)
	zcmd_expand_dst_nvlist(hdl, &zc);
	else {
	zcmd_free_nvlists(&zc);
	return (-1);
	}
	}

	err = zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &recvdprops);
	zcmd_free_nvlists(&zc);
	if (err != 0)
	return (-1);

	nvlist_free(zhp->zfs_recvd_props);
	zhp->zfs_recvd_props = recvdprops;

	return (0);
	}

	static int
	put_stats_zhdl(zfs_handle_t zhp, zfs_cmd_t zc)
	{
	nvlist_t allprops, userprops;

	zhp->zfs_dmustats = zc->zc_objset_stats; /* structure assignment */

	if (zcmd_read_dst_nvlist(zhp->zfs_hdl, zc, &allprops) != 0) {
	return (-1);
	}

	/*
	* XXX Why do we store the user props separately, in addition to
	* storing them in zfs_props?
	*/
	if ((userprops = process_user_props(zhp, allprops)) == NULL) {
	nvlist_free(allprops);
	return (-1);
	}

	nvlist_free(zhp->zfs_props);
	nvlist_free(zhp->zfs_user_props);

	zhp->zfs_props = allprops;
	zhp->zfs_user_props = userprops;

	return (0);
	}

	static int
	get_stats(zfs_handle_t *zhp)
	{
	int rc = 0;
	zfs_cmd_t zc = {"\0"};

	zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);

	if (get_stats_ioctl(zhp, &zc) != 0)
	rc = -1;
	else if (put_stats_zhdl(zhp, &zc) != 0)
	rc = -1;
	zcmd_free_nvlists(&zc);
	return (rc);
	}

	/*
	* Refresh the properties currently stored in the handle.
	*/
	void
	zfs_refresh_properties(zfs_handle_t *zhp)
	{
	(void) get_stats(zhp);
	}

	/*
	* Makes a handle from the given dataset name. Used by zfs_open() and
	* zfs_iter_* to create child handles on the fly.
	*/
	static int
	make_dataset_handle_common(zfs_handle_t zhp, zfs_cmd_t zc)
	{
	if (put_stats_zhdl(zhp, zc) != 0)
	return (-1);

	/*
	* We've managed to open the dataset and gather statistics. Determine
	* the high-level type.
	*/
	if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL) {
	zhp->zfs_head_type = ZFS_TYPE_VOLUME;
	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS) {
	zhp->zfs_head_type = ZFS_TYPE_FILESYSTEM;
	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_OTHER) {
	errno = EINVAL;
	return (-1);
	} else if (zhp->zfs_dmustats.dds_inconsistent) {
	errno = EBUSY;
	return (-1);
	} else {
	abort();
	}

	if (zhp->zfs_dmustats.dds_is_snapshot)
	zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL)
	zhp->zfs_type = ZFS_TYPE_VOLUME;
	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS)
	zhp->zfs_type = ZFS_TYPE_FILESYSTEM;
	else
	abort(); /* we should never see any other types */

	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL)
	return (-1);

	return (0);
	}

	zfs_handle_t *
	make_dataset_handle(libzfs_handle_t hdl, const char path)
	{
	zfs_cmd_t zc = {"\0"};

	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));

	if (zhp == NULL)
	return (NULL);

	zhp->zfs_hdl = hdl;
	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));
	zcmd_alloc_dst_nvlist(hdl, &zc, 0);

	if (get_stats_ioctl(zhp, &zc) == -1) {
	zcmd_free_nvlists(&zc);
	free(zhp);
	return (NULL);
	}
	if (make_dataset_handle_common(zhp, &zc) == -1) {
	free(zhp);
	zhp = NULL;
	}
	zcmd_free_nvlists(&zc);
	return (zhp);
	}

	zfs_handle_t *
	make_dataset_handle_zc(libzfs_handle_t hdl, zfs_cmd_t zc)
	{
	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));

	if (zhp == NULL)
	return (NULL);

	zhp->zfs_hdl = hdl;
	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
	if (make_dataset_handle_common(zhp, zc) == -1) {
	free(zhp);
	return (NULL);
	}
	return (zhp);
	}

	zfs_handle_t *
	make_dataset_simple_handle_zc(zfs_handle_t pzhp, zfs_cmd_t zc)
	{
	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));

	if (zhp == NULL)
	return (NULL);

	zhp->zfs_hdl = pzhp->zfs_hdl;
	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
	zhp->zfs_head_type = pzhp->zfs_type;
	zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
	zhp->zpool_hdl = zpool_handle(zhp);
	zhp->zfs_dmustats = zc->zc_objset_stats;

	return (zhp);
	}

	zfs_handle_t *
	zfs_handle_dup(zfs_handle_t *zhp_orig)
	{
	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));

	if (zhp == NULL)
	return (NULL);

	zhp->zfs_hdl = zhp_orig->zfs_hdl;
	zhp->zpool_hdl = zhp_orig->zpool_hdl;
	(void) strlcpy(zhp->zfs_name, zhp_orig->zfs_name,
	sizeof (zhp->zfs_name));
	zhp->zfs_type = zhp_orig->zfs_type;
	zhp->zfs_head_type = zhp_orig->zfs_head_type;
	zhp->zfs_dmustats = zhp_orig->zfs_dmustats;
	if (zhp_orig->zfs_props != NULL) {
	if (nvlist_dup(zhp_orig->zfs_props, &zhp->zfs_props, 0) != 0) {
	(void) no_memory(zhp->zfs_hdl);
	zfs_close(zhp);
	return (NULL);
	}
	}
	if (zhp_orig->zfs_user_props != NULL) {
	if (nvlist_dup(zhp_orig->zfs_user_props,
	&zhp->zfs_user_props, 0) != 0) {
	(void) no_memory(zhp->zfs_hdl);
	zfs_close(zhp);
	return (NULL);
	}
	}
	if (zhp_orig->zfs_recvd_props != NULL) {
	if (nvlist_dup(zhp_orig->zfs_recvd_props,
	&zhp->zfs_recvd_props, 0)) {
	(void) no_memory(zhp->zfs_hdl);
	zfs_close(zhp);
	return (NULL);
	}
	}
	zhp->zfs_mntcheck = zhp_orig->zfs_mntcheck;
	if (zhp_orig->zfs_mntopts != NULL) {
	zhp->zfs_mntopts = zfs_strdup(zhp_orig->zfs_hdl,
	zhp_orig->zfs_mntopts);
	}
	zhp->zfs_props_table = zhp_orig->zfs_props_table;
	return (zhp);
	}

	boolean_t
	zfs_bookmark_exists(const char *path)
	{
	nvlist_t *bmarks;
	nvlist_t *props;
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	char *bmark_name;
	char *pound;
	int err;
	boolean_t rv;

	(void) strlcpy(fsname, path, sizeof (fsname));
	pound = strchr(fsname, '#');
	if (pound == NULL)
	return (B_FALSE);

	*pound = '\0';
	bmark_name = pound + 1;
	props = fnvlist_alloc();
	err = lzc_get_bookmarks(fsname, props, &bmarks);
	nvlist_free(props);
	if (err != 0) {
	nvlist_free(bmarks);
	return (B_FALSE);
	}

	rv = nvlist_exists(bmarks, bmark_name);
	nvlist_free(bmarks);
	return (rv);
	}

	zfs_handle_t *
	make_bookmark_handle(zfs_handle_t parent, const char path,
	nvlist_t *bmark_props)
	{
	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));

	if (zhp == NULL)
	return (NULL);

	/* Fill in the name. */
	zhp->zfs_hdl = parent->zfs_hdl;
	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));

	/* Set the property lists. */
	if (nvlist_dup(bmark_props, &zhp->zfs_props, 0) != 0) {
	free(zhp);
	return (NULL);
	}

	/* Set the types. */
	zhp->zfs_head_type = parent->zfs_head_type;
	zhp->zfs_type = ZFS_TYPE_BOOKMARK;

	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL) {
	nvlist_free(zhp->zfs_props);
	free(zhp);
	return (NULL);
	}

	return (zhp);
	}

	struct zfs_open_bookmarks_cb_data {
	const char *path;
	zfs_handle_t *zhp;
	};

	static int
	zfs_open_bookmarks_cb(zfs_handle_t zhp, void data)
	{
	struct zfs_open_bookmarks_cb_data *dp = data;

	/*
	* Is it the one we are looking for?
	*/
	if (strcmp(dp->path, zfs_get_name(zhp)) == 0) {
	/*
	* We found it. Save it and let the caller know we are done.
	*/
	dp->zhp = zhp;
	return (EEXIST);
	}

	/*
	* Not found. Close the handle and ask for another one.
	*/
	zfs_close(zhp);
	return (0);
	}

	/*
	* Opens the given snapshot, bookmark, filesystem, or volume. The 'types'
	* argument is a mask of acceptable types. The function will print an
	* appropriate error message and return NULL if it can't be opened.
	*/
	zfs_handle_t *
	zfs_open(libzfs_handle_t hdl, const char path, int types)
	{
	zfs_handle_t *zhp;
	char errbuf[ERRBUFLEN];
	char *bookp;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot open '%s'"), path);

	/*
	* Validate the name before we even try to open it.
	*/
	if (!zfs_validate_name(hdl, path, types, B_FALSE)) {
	(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	return (NULL);
	}

	/*
	* Bookmarks needs to be handled separately.
	*/
	bookp = strchr(path, '#');
	if (bookp == NULL) {
	/*
	* Try to get stats for the dataset, which will tell us if it
	* exists.
	*/
	errno = 0;
	if ((zhp = make_dataset_handle(hdl, path)) == NULL) {
	(void) zfs_standard_error(hdl, errno, errbuf);
	return (NULL);
	}
	} else {
	char dsname[ZFS_MAX_DATASET_NAME_LEN];
	zfs_handle_t *pzhp;
	struct zfs_open_bookmarks_cb_data cb_data = {path, NULL};

	/*
	* We need to cut out '#' and everything after '#'
	* to get the parent dataset name only.
	*/
	assert(bookp - path < sizeof (dsname));
	- (void) strncpy(dsname, path, bookp - path);
	- dsname[bookp - path] = '\0';
	+ (void) strlcpy(dsname, path,
	+ MIN(sizeof (dsname), bookp - path + 1));

	/*
	* Create handle for the parent dataset.
	*/
	errno = 0;
	if ((pzhp = make_dataset_handle(hdl, dsname)) == NULL) {
	(void) zfs_standard_error(hdl, errno, errbuf);
	return (NULL);
	}

	/*
	* Iterate bookmarks to find the right one.
	*/
	errno = 0;
	if ((zfs_iter_bookmarks(pzhp, zfs_open_bookmarks_cb,
	&cb_data) == 0) && (cb_data.zhp == NULL)) {
	(void) zfs_error(hdl, EZFS_NOENT, errbuf);
	zfs_close(pzhp);
	return (NULL);
	}
	if (cb_data.zhp == NULL) {
	(void) zfs_standard_error(hdl, errno, errbuf);
	zfs_close(pzhp);
	return (NULL);
	}
	zhp = cb_data.zhp;

	/*
	* Cleanup.
	*/
	zfs_close(pzhp);
	}

	if (!(types & zhp->zfs_type)) {
	(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
	zfs_close(zhp);
	return (NULL);
	}

	return (zhp);
	}

	/*
	* Release a ZFS handle. Nothing to do but free the associated memory.
	*/
	void
	zfs_close(zfs_handle_t *zhp)
	{
	if (zhp->zfs_mntopts)
	free(zhp->zfs_mntopts);
	nvlist_free(zhp->zfs_props);
	nvlist_free(zhp->zfs_user_props);
	nvlist_free(zhp->zfs_recvd_props);
	free(zhp);
	}

	typedef struct mnttab_node {
	struct mnttab mtn_mt;
	avl_node_t mtn_node;
	} mnttab_node_t;

	static int
	libzfs_mnttab_cache_compare(const void arg1, const void arg2)
	{
	const mnttab_node_t mtn1 = (const mnttab_node_t )arg1;
	const mnttab_node_t mtn2 = (const mnttab_node_t )arg2;
	int rv;

	rv = strcmp(mtn1->mtn_mt.mnt_special, mtn2->mtn_mt.mnt_special);

	return (TREE_ISIGN(rv));
	}

	void
	libzfs_mnttab_init(libzfs_handle_t *hdl)
	{
	pthread_mutex_init(&hdl->libzfs_mnttab_cache_lock, NULL);
	assert(avl_numnodes(&hdl->libzfs_mnttab_cache) == 0);
	avl_create(&hdl->libzfs_mnttab_cache, libzfs_mnttab_cache_compare,
	sizeof (mnttab_node_t), offsetof(mnttab_node_t, mtn_node));
	}

	static int
	libzfs_mnttab_update(libzfs_handle_t *hdl)
	{
	FILE *mnttab;
	struct mnttab entry;

	if ((mnttab = fopen(MNTTAB, "re")) == NULL)
	return (ENOENT);

	while (getmntent(mnttab, &entry) == 0) {
	mnttab_node_t *mtn;
	avl_index_t where;

	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
	continue;

	mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
	mtn->mtn_mt.mnt_special = zfs_strdup(hdl, entry.mnt_special);
	mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, entry.mnt_mountp);
	mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, entry.mnt_fstype);
	mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, entry.mnt_mntopts);

	/* Exclude duplicate mounts */
	if (avl_find(&hdl->libzfs_mnttab_cache, mtn, &where) != NULL) {
	free(mtn->mtn_mt.mnt_special);
	free(mtn->mtn_mt.mnt_mountp);
	free(mtn->mtn_mt.mnt_fstype);
	free(mtn->mtn_mt.mnt_mntopts);
	free(mtn);
	continue;
	}

	avl_add(&hdl->libzfs_mnttab_cache, mtn);
	}

	(void) fclose(mnttab);
	return (0);
	}

	void
	libzfs_mnttab_fini(libzfs_handle_t *hdl)
	{
	void *cookie = NULL;
	mnttab_node_t *mtn;

	while ((mtn = avl_destroy_nodes(&hdl->libzfs_mnttab_cache, &cookie))
	!= NULL) {
	free(mtn->mtn_mt.mnt_special);
	free(mtn->mtn_mt.mnt_mountp);
	free(mtn->mtn_mt.mnt_fstype);
	free(mtn->mtn_mt.mnt_mntopts);
	free(mtn);
	}
	avl_destroy(&hdl->libzfs_mnttab_cache);
	(void) pthread_mutex_destroy(&hdl->libzfs_mnttab_cache_lock);
	}

	void
	libzfs_mnttab_cache(libzfs_handle_t *hdl, boolean_t enable)
	{
	hdl->libzfs_mnttab_enable = enable;
	}

	int
	libzfs_mnttab_find(libzfs_handle_t hdl, const char fsname,
	struct mnttab *entry)
	{
	FILE *mnttab;
	mnttab_node_t find;
	mnttab_node_t *mtn;
	int ret = ENOENT;

	if (!hdl->libzfs_mnttab_enable) {
	struct mnttab srch = { 0 };

	if (avl_numnodes(&hdl->libzfs_mnttab_cache))
	libzfs_mnttab_fini(hdl);

	if ((mnttab = fopen(MNTTAB, "re")) == NULL)
	return (ENOENT);

	srch.mnt_special = (char *)fsname;
	srch.mnt_fstype = (char *)MNTTYPE_ZFS;
	ret = getmntany(mnttab, entry, &srch) ? ENOENT : 0;
	(void) fclose(mnttab);
	return (ret);
	}

	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
	if (avl_numnodes(&hdl->libzfs_mnttab_cache) == 0) {
	int error;

	if ((error = libzfs_mnttab_update(hdl)) != 0) {
	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
	return (error);
	}
	}

	find.mtn_mt.mnt_special = (char *)fsname;
	mtn = avl_find(&hdl->libzfs_mnttab_cache, &find, NULL);
	if (mtn) {
	*entry = mtn->mtn_mt;
	ret = 0;
	}
	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
	return (ret);
	}

	void
	libzfs_mnttab_add(libzfs_handle_t hdl, const char special,
	const char mountp, const char mntopts)
	{
	mnttab_node_t *mtn;

	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
	if (avl_numnodes(&hdl->libzfs_mnttab_cache) != 0) {
	mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
	mtn->mtn_mt.mnt_special = zfs_strdup(hdl, special);
	mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, mountp);
	mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, MNTTYPE_ZFS);
	mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, mntopts);
	/*
	* Another thread may have already added this entry
	* via libzfs_mnttab_update. If so we should skip it.
	*/
	if (avl_find(&hdl->libzfs_mnttab_cache, mtn, NULL) != NULL) {
	free(mtn->mtn_mt.mnt_special);
	free(mtn->mtn_mt.mnt_mountp);
	free(mtn->mtn_mt.mnt_fstype);
	free(mtn->mtn_mt.mnt_mntopts);
	free(mtn);
	} else {
	avl_add(&hdl->libzfs_mnttab_cache, mtn);
	}
	}
	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
	}

	void
	libzfs_mnttab_remove(libzfs_handle_t hdl, const char fsname)
	{
	mnttab_node_t find;
	mnttab_node_t *ret;

	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
	find.mtn_mt.mnt_special = (char *)fsname;
	if ((ret = avl_find(&hdl->libzfs_mnttab_cache, (void *)&find, NULL))
	!= NULL) {
	avl_remove(&hdl->libzfs_mnttab_cache, ret);
	free(ret->mtn_mt.mnt_special);
	free(ret->mtn_mt.mnt_mountp);
	free(ret->mtn_mt.mnt_fstype);
	free(ret->mtn_mt.mnt_mntopts);
	free(ret);
	}
	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
	}

	int
	zfs_spa_version(zfs_handle_t zhp, int spa_version)
	{
	zpool_handle_t *zpool_handle = zhp->zpool_hdl;

	if (zpool_handle == NULL)
	return (-1);

	*spa_version = zpool_get_prop_int(zpool_handle,
	ZPOOL_PROP_VERSION, NULL);
	return (0);
	}

	/*
	* The choice of reservation property depends on the SPA version.
	*/
	static int
	zfs_which_resv_prop(zfs_handle_t zhp, zfs_prop_t resv_prop)
	{
	int spa_version;

	if (zfs_spa_version(zhp, &spa_version) < 0)
	return (-1);

	if (spa_version >= SPA_VERSION_REFRESERVATION)
	*resv_prop = ZFS_PROP_REFRESERVATION;
	else
	*resv_prop = ZFS_PROP_RESERVATION;

	return (0);
	}

	/*
	* Given an nvlist of properties to set, validates that they are correct, and
	* parses any numeric properties (index, boolean, etc) if they are specified as
	* strings.
	*/
	nvlist_t *
	zfs_valid_proplist(libzfs_handle_t hdl, zfs_type_t type, nvlist_t nvl,
	uint64_t zoned, zfs_handle_t zhp, zpool_handle_t zpool_hdl,
	boolean_t key_params_ok, const char *errbuf)
	{
	nvpair_t *elem;
	uint64_t intval;
	char *strval;
	zfs_prop_t prop;
	nvlist_t *ret;
	int chosen_normal = -1;
	int chosen_utf = -1;

	if (nvlist_alloc(&ret, NV_UNIQUE_NAME, 0) != 0) {
	(void) no_memory(hdl);
	return (NULL);
	}

	/*
	* Make sure this property is valid and applies to this type.
	*/

	elem = NULL;
	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
	const char *propname = nvpair_name(elem);

	prop = zfs_name_to_prop(propname);
	if (prop == ZPROP_USERPROP && zfs_prop_user(propname)) {
	/*
	* This is a user property: make sure it's a
	* string, and that it's less than ZAP_MAXNAMELEN.
	*/
	if (nvpair_type(elem) != DATA_TYPE_STRING) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a string"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property name '%s' is too long"),
	propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	(void) nvpair_value_string(elem, &strval);
	if (nvlist_add_string(ret, propname, strval) != 0) {
	(void) no_memory(hdl);
	goto error;
	}
	continue;
	}

	/*
	* Currently, only user properties can be modified on
	* snapshots.
	*/
	if (type == ZFS_TYPE_SNAPSHOT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"this property can not be modified for snapshots"));
	(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
	goto error;
	}

	if (prop == ZPROP_USERPROP && zfs_prop_userquota(propname)) {
	zfs_userquota_prop_t uqtype;
	char *newpropname = NULL;
	char domain[128];
	uint64_t rid;
	uint64_t valary[3];
	int rc;

	if (userquota_propname_decode(propname, zoned,
	&uqtype, domain, sizeof (domain), &rid) != 0) {
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN,
	"'%s' has an invalid user/group name"),
	propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (uqtype != ZFS_PROP_USERQUOTA &&
	uqtype != ZFS_PROP_GROUPQUOTA &&
	uqtype != ZFS_PROP_USEROBJQUOTA &&
	uqtype != ZFS_PROP_GROUPOBJQUOTA &&
	uqtype != ZFS_PROP_PROJECTQUOTA &&
	uqtype != ZFS_PROP_PROJECTOBJQUOTA) {
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "'%s' is readonly"),
	propname);
	(void) zfs_error(hdl, EZFS_PROPREADONLY,
	errbuf);
	goto error;
	}

	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	(void) nvpair_value_string(elem, &strval);
	if (strcmp(strval, "none") == 0) {
	intval = 0;
	} else if (zfs_nicestrtonum(hdl,
	strval, &intval) != 0) {
	(void) zfs_error(hdl,
	EZFS_BADPROP, errbuf);
	goto error;
	}
	} else if (nvpair_type(elem) ==
	DATA_TYPE_UINT64) {
	(void) nvpair_value_uint64(elem, &intval);
	if (intval == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"use 'none' to disable "
	"{user\|group\|project}quota"));
	goto error;
	}
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a number"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	/*
	* Encode the prop name as
	* userquota@<hex-rid>-domain, to make it easy
	* for the kernel to decode.
	*/
	rc = asprintf(&newpropname, "%s%llx-%s",
	zfs_userquota_prop_prefixes[uqtype],
	(longlong_t)rid, domain);
	if (rc == -1 \|\| newpropname == NULL) {
	(void) no_memory(hdl);
	goto error;
	}

	valary[0] = uqtype;
	valary[1] = rid;
	valary[2] = intval;
	if (nvlist_add_uint64_array(ret, newpropname,
	valary, 3) != 0) {
	free(newpropname);
	(void) no_memory(hdl);
	goto error;
	}
	free(newpropname);
	continue;
	} else if (prop == ZPROP_USERPROP &&
	zfs_prop_written(propname)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is readonly"),
	propname);
	(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
	goto error;
	}

	if (prop == ZPROP_INVAL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property '%s'"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (!zfs_prop_valid_for_type(prop, type, B_FALSE)) {
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "'%s' does not "
	"apply to datasets of this type"), propname);
	(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
	goto error;
	}

	if (zfs_prop_readonly(prop) &&
	!(zfs_prop_setonce(prop) && zhp == NULL) &&
	!(zfs_prop_encryption_key_param(prop) && key_params_ok)) {
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "'%s' is readonly"),
	propname);
	(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
	goto error;
	}

	if (zprop_parse_value(hdl, elem, prop, type, ret,
	&strval, &intval, errbuf) != 0)
	goto error;

	/*
	* Perform some additional checks for specific properties.
	*/
	switch (prop) {
	case ZFS_PROP_VERSION:
	{
	int version;

	if (zhp == NULL)
	break;
	version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
	if (intval < version) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Can not downgrade; already at version %u"),
	version);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	}

	case ZFS_PROP_VOLBLOCKSIZE:
	case ZFS_PROP_RECORDSIZE:
	{
	int maxbs = SPA_MAXBLOCKSIZE;
	char buf[64];

	if (zpool_hdl != NULL) {
	maxbs = zpool_get_prop_int(zpool_hdl,
	ZPOOL_PROP_MAXBLOCKSIZE, NULL);
	}
	/*
	* The value must be a power of two between
	* SPA_MINBLOCKSIZE and maxbs.
	*/
	if (intval < SPA_MINBLOCKSIZE \|\|
	intval > maxbs \|\| !ISP2(intval)) {
	zfs_nicebytes(maxbs, buf, sizeof (buf));
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be power of 2 from 512B "
	"to %s"), propname, buf);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	}

	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
	{
	int maxbs = SPA_OLD_MAXBLOCKSIZE;
	char buf[64];

	if (zpool_hdl != NULL) {
	char state[64] = "";

	maxbs = zpool_get_prop_int(zpool_hdl,
	ZPOOL_PROP_MAXBLOCKSIZE, NULL);

	/*
	* Issue a warning but do not fail so that
	* tests for settable properties succeed.
	*/
	if (zpool_prop_get_feature(zpool_hdl,
	"feature@allocation_classes", state,
	sizeof (state)) != 0 \|\|
	strcmp(state, ZFS_FEATURE_ACTIVE) != 0) {
	(void) fprintf(stderr, gettext(
	"%s: property requires a special "
	"device in the pool\n"), propname);
	}
	}
	if (intval != 0 &&
	(intval < SPA_MINBLOCKSIZE \|\|
	intval > maxbs \|\| !ISP2(intval))) {
	zfs_nicebytes(maxbs, buf, sizeof (buf));
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid '%s=%llu' property: must be zero "
	"or a power of 2 from 512B to %s"),
	propname, (unsigned long long)intval, buf);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	}

	case ZFS_PROP_MLSLABEL:
	{
	#ifdef HAVE_MLSLABEL
	/*
	* Verify the mlslabel string and convert to
	* internal hex label string.
	*/

	m_label_t *new_sl;
	char hex = NULL; / internal label string */

	/* Default value is already OK. */
	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
	break;

	/* Verify the label can be converted to binary form */
	if (((new_sl = m_label_alloc(MAC_LABEL)) == NULL) \|\|
	(str_to_label(strval, &new_sl, MAC_LABEL,
	L_NO_CORRECTION, NULL) == -1)) {
	goto badlabel;
	}

	/* Now translate to hex internal label string */
	if (label_to_str(new_sl, &hex, M_INTERNAL,
	DEF_NAMES) != 0) {
	if (hex)
	free(hex);
	goto badlabel;
	}
	m_label_free(new_sl);

	/* If string is already in internal form, we're done. */
	if (strcmp(strval, hex) == 0) {
	free(hex);
	break;
	}

	/* Replace the label string with the internal form. */
	(void) nvlist_remove(ret, zfs_prop_to_name(prop),
	DATA_TYPE_STRING);
	fnvlist_add_string(ret, zfs_prop_to_name(prop), hex);
	free(hex);

	break;

	badlabel:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid mlslabel '%s'"), strval);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	m_label_free(new_sl); /* OK if null */
	goto error;
	#else
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"mlslabels are unsupported"));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	#endif /* HAVE_MLSLABEL */
	}

	case ZFS_PROP_MOUNTPOINT:
	{
	namecheck_err_t why;

	if (strcmp(strval, ZFS_MOUNTPOINT_NONE) == 0 \|\|
	strcmp(strval, ZFS_MOUNTPOINT_LEGACY) == 0)
	break;

	if (mountpoint_namecheck(strval, &why)) {
	switch (why) {
	case NAME_ERR_LEADING_SLASH:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN,
	"'%s' must be an absolute path, "
	"'none', or 'legacy'"), propname);
	break;
	case NAME_ERR_TOOLONG:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN,
	"component of '%s' is too long"),
	propname);
	break;

	default:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN,
	"(%d) not defined"),
	why);
	break;
	}
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	zfs_fallthrough;
	}

	case ZFS_PROP_SHARESMB:
	case ZFS_PROP_SHARENFS:
	/*
	* For the mountpoint and sharenfs or sharesmb
	* properties, check if it can be set in a
	* global/non-global zone based on
	* the zoned property value:
	*
	* global zone non-global zone
	* --------------------------------------------------
	* zoned=on mountpoint (no) mountpoint (yes)
	* sharenfs (no) sharenfs (no)
	* sharesmb (no) sharesmb (no)
	*
	* zoned=off mountpoint (yes) N/A
	* sharenfs (yes)
	* sharesmb (yes)
	*/
	if (zoned) {
	if (getzoneid() == GLOBAL_ZONEID) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' cannot be set on "
	"dataset in a non-global zone"),
	propname);
	(void) zfs_error(hdl, EZFS_ZONED,
	errbuf);
	goto error;
	} else if (prop == ZFS_PROP_SHARENFS \|\|
	prop == ZFS_PROP_SHARESMB) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' cannot be set in "
	"a non-global zone"), propname);
	(void) zfs_error(hdl, EZFS_ZONED,
	errbuf);
	goto error;
	}
	} else if (getzoneid() != GLOBAL_ZONEID) {
	/*
	* If zoned property is 'off', this must be in
	* a global zone. If not, something is wrong.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' cannot be set while dataset "
	"'zoned' property is set"), propname);
	(void) zfs_error(hdl, EZFS_ZONED, errbuf);
	goto error;
	}

	/*
	* At this point, it is legitimate to set the
	* property. Now we want to make sure that the
	* property value is valid if it is sharenfs.
	*/
	if ((prop == ZFS_PROP_SHARENFS \|\|
	prop == ZFS_PROP_SHARESMB) &&
	strcmp(strval, "on") != 0 &&
	strcmp(strval, "off") != 0) {
	enum sa_protocol proto;

	if (prop == ZFS_PROP_SHARESMB)
	proto = SA_PROTOCOL_SMB;
	else
	proto = SA_PROTOCOL_NFS;

	if (sa_validate_shareopts(strval, proto) !=
	SA_OK) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' cannot be set to invalid "
	"options"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}
	}

	break;

	case ZFS_PROP_KEYLOCATION:
	if (!zfs_prop_valid_keylocation(strval, B_FALSE)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid keylocation"));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (zhp != NULL) {
	uint64_t crypt =
	zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION);

	if (crypt == ZIO_CRYPT_OFF &&
	strcmp(strval, "none") != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"keylocation must be 'none' "
	"for unencrypted datasets"));
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	} else if (crypt != ZIO_CRYPT_OFF &&
	strcmp(strval, "none") == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"keylocation must not be 'none' "
	"for encrypted datasets"));
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}
	}
	break;

	case ZFS_PROP_PBKDF2_ITERS:
	if (intval < MIN_PBKDF2_ITERATIONS) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"minimum pbkdf2 iterations is %u"),
	MIN_PBKDF2_ITERATIONS);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;

	case ZFS_PROP_UTF8ONLY:
	chosen_utf = (int)intval;
	break;

	case ZFS_PROP_NORMALIZE:
	chosen_normal = (int)intval;
	break;

	default:
	break;
	}

	/*
	* For changes to existing volumes, we have some additional
	* checks to enforce.
	*/
	if (type == ZFS_TYPE_VOLUME && zhp != NULL) {
	uint64_t blocksize = zfs_prop_get_int(zhp,
	ZFS_PROP_VOLBLOCKSIZE);
	char buf[64];

	switch (prop) {
	case ZFS_PROP_VOLSIZE:
	if (intval % blocksize != 0) {
	zfs_nicebytes(blocksize, buf,
	sizeof (buf));
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a multiple of "
	"volume block size (%s)"),
	propname, buf);
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}

	if (intval == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' cannot be zero"),
	propname);
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}
	break;

	default:
	break;
	}
	}

	/* check encryption properties */
	if (zhp != NULL) {
	int64_t crypt = zfs_prop_get_int(zhp,
	ZFS_PROP_ENCRYPTION);

	switch (prop) {
	case ZFS_PROP_COPIES:
	if (crypt != ZIO_CRYPT_OFF && intval > 2) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encrypted datasets cannot have "
	"3 copies"));
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}
	break;
	default:
	break;
	}
	}
	}

	/*
	* If normalization was chosen, but no UTF8 choice was made,
	* enforce rejection of non-UTF8 names.
	*
	* If normalization was chosen, but rejecting non-UTF8 names
	* was explicitly not chosen, it is an error.
	*
	* If utf8only was turned off, but the parent has normalization,
	* turn off normalization.
	*/
	if (chosen_normal > 0 && chosen_utf < 0) {
	if (nvlist_add_uint64(ret,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY), 1) != 0) {
	(void) no_memory(hdl);
	goto error;
	}
	} else if (chosen_normal > 0 && chosen_utf == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be set 'on' if normalization chosen"),
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	} else if (chosen_normal < 0 && chosen_utf == 0) {
	if (nvlist_add_uint64(ret,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE), 0) != 0) {
	(void) no_memory(hdl);
	goto error;
	}
	}
	return (ret);

	error:
	nvlist_free(ret);
	return (NULL);
	}

	static int
	zfs_add_synthetic_resv(zfs_handle_t zhp, nvlist_t nvl)
	{
	uint64_t old_volsize;
	uint64_t new_volsize;
	uint64_t old_reservation;
	uint64_t new_reservation;
	zfs_prop_t resv_prop;
	nvlist_t *props;
	zpool_handle_t *zph = zpool_handle(zhp);

	/*
	* If this is an existing volume, and someone is setting the volsize,
	* make sure that it matches the reservation, or add it if necessary.
	*/
	old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
	if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
	return (-1);
	old_reservation = zfs_prop_get_int(zhp, resv_prop);

	props = fnvlist_alloc();
	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));

	if ((zvol_volsize_to_reservation(zph, old_volsize, props) !=
	old_reservation) \|\| nvlist_exists(nvl,
	zfs_prop_to_name(resv_prop))) {
	fnvlist_free(props);
	return (0);
	}
	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
	&new_volsize) != 0) {
	fnvlist_free(props);
	return (-1);
	}
	new_reservation = zvol_volsize_to_reservation(zph, new_volsize, props);
	fnvlist_free(props);

	if (nvlist_add_uint64(nvl, zfs_prop_to_name(resv_prop),
	new_reservation) != 0) {
	(void) no_memory(zhp->zfs_hdl);
	return (-1);
	}
	return (1);
	}

	/*
	* Helper for 'zfs {set\|clone} refreservation=auto'. Must be called after
	* zfs_valid_proplist(), as it is what sets the UINT64_MAX sentinel value.
	* Return codes must match zfs_add_synthetic_resv().
	*/
	static int
	zfs_fix_auto_resv(zfs_handle_t zhp, nvlist_t nvl)
	{
	uint64_t volsize;
	uint64_t resvsize;
	zfs_prop_t prop;
	nvlist_t *props;

	if (!ZFS_IS_VOLUME(zhp)) {
	return (0);
	}

	if (zfs_which_resv_prop(zhp, &prop) != 0) {
	return (-1);
	}

	if (prop != ZFS_PROP_REFRESERVATION) {
	return (0);
	}

	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(prop), &resvsize) != 0) {
	/* No value being set, so it can't be "auto" */
	return (0);
	}
	if (resvsize != UINT64_MAX) {
	/* Being set to a value other than "auto" */
	return (0);
	}

	props = fnvlist_alloc();

	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));

	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
	&volsize) != 0) {
	volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
	}

	resvsize = zvol_volsize_to_reservation(zpool_handle(zhp), volsize,
	props);
	fnvlist_free(props);

	(void) nvlist_remove_all(nvl, zfs_prop_to_name(prop));
	if (nvlist_add_uint64(nvl, zfs_prop_to_name(prop), resvsize) != 0) {
	(void) no_memory(zhp->zfs_hdl);
	return (-1);
	}
	return (1);
	}

	static boolean_t
	zfs_is_namespace_prop(zfs_prop_t prop)
	{
	switch (prop) {

	case ZFS_PROP_ATIME:
	case ZFS_PROP_RELATIME:
	case ZFS_PROP_DEVICES:
	case ZFS_PROP_EXEC:
	case ZFS_PROP_SETUID:
	case ZFS_PROP_READONLY:
	case ZFS_PROP_XATTR:
	case ZFS_PROP_NBMAND:
	return (B_TRUE);

	default:
	return (B_FALSE);
	}
	}

	/*
	* Given a property name and value, set the property for the given dataset.
	*/
	int
	zfs_prop_set(zfs_handle_t zhp, const char propname, const char *propval)
	{
	int ret = -1;
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t *nvl = NULL;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
	zhp->zfs_name);

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0 \|\|
	nvlist_add_string(nvl, propname, propval) != 0) {
	(void) no_memory(hdl);
	goto error;
	}

	ret = zfs_prop_set_list(zhp, nvl);

	error:
	nvlist_free(nvl);
	return (ret);
	}



	/*
	* Given an nvlist of property names and values, set the properties for the
	* given dataset.
	*/
	int
	zfs_prop_set_list(zfs_handle_t zhp, nvlist_t props)
	{
	zfs_cmd_t zc = {"\0"};
	int ret = -1;
	prop_changelist_t **cls = NULL;
	int cl_idx;
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t *nvl;
	int nvl_len = 0;
	int added_resv = 0;
	zfs_prop_t prop = 0;
	nvpair_t *elem;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
	zhp->zfs_name);

	if ((nvl = zfs_valid_proplist(hdl, zhp->zfs_type, props,
	zfs_prop_get_int(zhp, ZFS_PROP_ZONED), zhp, zhp->zpool_hdl,
	B_FALSE, errbuf)) == NULL)
	goto error;

	/*
	* We have to check for any extra properties which need to be added
	* before computing the length of the nvlist.
	*/
	for (elem = nvlist_next_nvpair(nvl, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(nvl, elem)) {
	if (zfs_name_to_prop(nvpair_name(elem)) == ZFS_PROP_VOLSIZE &&
	(added_resv = zfs_add_synthetic_resv(zhp, nvl)) == -1) {
	goto error;
	}
	}

	if (added_resv != 1 &&
	(added_resv = zfs_fix_auto_resv(zhp, nvl)) == -1) {
	goto error;
	}

	/*
	* Check how many properties we're setting and allocate an array to
	* store changelist pointers for postfix().
	*/
	for (elem = nvlist_next_nvpair(nvl, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(nvl, elem))
	nvl_len++;
	if ((cls = calloc(nvl_len, sizeof (prop_changelist_t *))) == NULL)
	goto error;

	cl_idx = 0;
	for (elem = nvlist_next_nvpair(nvl, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(nvl, elem)) {

	prop = zfs_name_to_prop(nvpair_name(elem));

	assert(cl_idx < nvl_len);
	/*
	* We don't want to unmount & remount the dataset when changing
	* its canmount property to 'on' or 'noauto'. We only use
	* the changelist logic to unmount when setting canmount=off.
	*/
	if (prop != ZFS_PROP_CANMOUNT \|\|
	(fnvpair_value_uint64(elem) == ZFS_CANMOUNT_OFF &&
	zfs_is_mounted(zhp, NULL))) {
	cls[cl_idx] = changelist_gather(zhp, prop, 0, 0);
	if (cls[cl_idx] == NULL)
	goto error;
	}

	if (prop == ZFS_PROP_MOUNTPOINT &&
	changelist_haszonedchild(cls[cl_idx])) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"child dataset with inherited mountpoint is used "
	"in a non-global zone"));
	ret = zfs_error(hdl, EZFS_ZONED, errbuf);
	goto error;
	}

	if (cls[cl_idx] != NULL &&
	(ret = changelist_prefix(cls[cl_idx])) != 0)
	goto error;

	cl_idx++;
	}
	assert(cl_idx == nvl_len);

	/*
	* Execute the corresponding ioctl() to set this list of properties.
	*/
	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	zcmd_write_src_nvlist(hdl, &zc, nvl);
	zcmd_alloc_dst_nvlist(hdl, &zc, 0);

	ret = zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);

	if (ret != 0) {
	if (zc.zc_nvlist_dst_filled == B_FALSE) {
	(void) zfs_standard_error(hdl, errno, errbuf);
	goto error;
	}

	/* Get the list of unset properties back and report them. */
	nvlist_t *errorprops = NULL;
	if (zcmd_read_dst_nvlist(hdl, &zc, &errorprops) != 0)
	goto error;
	for (nvpair_t *elem = nvlist_next_nvpair(errorprops, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(errorprops, elem)) {
	prop = zfs_name_to_prop(nvpair_name(elem));
	zfs_setprop_error(hdl, prop, errno, errbuf);
	}
	nvlist_free(errorprops);

	if (added_resv && errno == ENOSPC) {
	/* clean up the volsize property we tried to set */
	uint64_t old_volsize = zfs_prop_get_int(zhp,
	ZFS_PROP_VOLSIZE);
	nvlist_free(nvl);
	nvl = NULL;
	zcmd_free_nvlists(&zc);

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
	goto error;
	if (nvlist_add_uint64(nvl,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE),
	old_volsize) != 0)
	goto error;
	zcmd_write_src_nvlist(hdl, &zc, nvl);
	(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
	}
	} else {
	for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
	if (cls[cl_idx] != NULL) {
	int clp_err = changelist_postfix(cls[cl_idx]);
	if (clp_err != 0)
	ret = clp_err;
	}
	}

	if (ret == 0) {
	/*
	* Refresh the statistics so the new property
	* value is reflected.
	*/
	(void) get_stats(zhp);

	/*
	* Remount the filesystem to propagate the change
	* if one of the options handled by the generic
	* Linux namespace layer has been modified.
	*/
	if (zfs_is_namespace_prop(prop) &&
	zfs_is_mounted(zhp, NULL))
	ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
	}
	}

	error:
	nvlist_free(nvl);
	zcmd_free_nvlists(&zc);
	if (cls != NULL) {
	for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
	if (cls[cl_idx] != NULL)
	changelist_free(cls[cl_idx]);
	}
	free(cls);
	}
	return (ret);
	}

	/*
	* Given a property, inherit the value from the parent dataset, or if received
	* is TRUE, revert to the received value, if any.
	*/
	int
	zfs_prop_inherit(zfs_handle_t zhp, const char propname, boolean_t received)
	{
	zfs_cmd_t zc = {"\0"};
	int ret;
	prop_changelist_t *cl;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char errbuf[ERRBUFLEN];
	zfs_prop_t prop;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot inherit %s for '%s'"), propname, zhp->zfs_name);

	zc.zc_cookie = received;
	if ((prop = zfs_name_to_prop(propname)) == ZPROP_USERPROP) {
	/*
	* For user properties, the amount of work we have to do is very
	* small, so just do it here.
	*/
	if (!zfs_prop_user(propname)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));

	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0)
	return (zfs_standard_error(hdl, errno, errbuf));

	(void) get_stats(zhp);
	return (0);
	}

	/*
	* Verify that this property is inheritable.
	*/
	if (zfs_prop_readonly(prop))
	return (zfs_error(hdl, EZFS_PROPREADONLY, errbuf));

	if (!zfs_prop_inheritable(prop) && !received)
	return (zfs_error(hdl, EZFS_PROPNONINHERIT, errbuf));

	/*
	* Check to see if the value applies to this type
	*/
	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
	return (zfs_error(hdl, EZFS_PROPTYPE, errbuf));

	/*
	* Normalize the name, to get rid of shorthand abbreviations.
	*/
	propname = zfs_prop_to_name(prop);
	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));

	if (prop == ZFS_PROP_MOUNTPOINT && getzoneid() == GLOBAL_ZONEID &&
	zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset is used in a non-global zone"));
	return (zfs_error(hdl, EZFS_ZONED, errbuf));
	}

	/*
	* Determine datasets which will be affected by this change, if any.
	*/
	if ((cl = changelist_gather(zhp, prop, 0, 0)) == NULL)
	return (-1);

	if (prop == ZFS_PROP_MOUNTPOINT && changelist_haszonedchild(cl)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"child dataset with inherited mountpoint is used "
	"in a non-global zone"));
	ret = zfs_error(hdl, EZFS_ZONED, errbuf);
	goto error;
	}

	if ((ret = changelist_prefix(cl)) != 0)
	goto error;

	if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc)) != 0) {
	+ changelist_free(cl);
	return (zfs_standard_error(hdl, errno, errbuf));
	} else {

	if ((ret = changelist_postfix(cl)) != 0)
	goto error;

	/*
	* Refresh the statistics so the new property is reflected.
	*/
	(void) get_stats(zhp);

	/*
	* Remount the filesystem to propagate the change
	* if one of the options handled by the generic
	* Linux namespace layer has been modified.
	*/
	if (zfs_is_namespace_prop(prop) &&
	zfs_is_mounted(zhp, NULL))
	ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
	}

	error:
	changelist_free(cl);
	return (ret);
	}

	/*
	* True DSL properties are stored in an nvlist. The following two functions
	* extract them appropriately.
	*/
	uint64_t
	getprop_uint64(zfs_handle_t zhp, zfs_prop_t prop, char *source)
	{
	nvlist_t *nv;
	uint64_t value;

	*source = NULL;
	if (nvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(prop), &nv) == 0) {
	value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
	(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
	} else {
	verify(!zhp->zfs_props_table \|\|
	zhp->zfs_props_table[prop] == B_TRUE);
	value = zfs_prop_default_numeric(prop);
	source = (char )"";
	}

	return (value);
	}

	static const char *
	getprop_string(zfs_handle_t zhp, zfs_prop_t prop, char *source)
	{
	nvlist_t *nv;
	const char *value;

	*source = NULL;
	if (nvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(prop), &nv) == 0) {
	value = fnvlist_lookup_string(nv, ZPROP_VALUE);
	(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
	} else {
	verify(!zhp->zfs_props_table \|\|
	zhp->zfs_props_table[prop] == B_TRUE);
	value = zfs_prop_default_string(prop);
	source = (char )"";
	}

	return (value);
	}

	static boolean_t
	zfs_is_recvd_props_mode(zfs_handle_t *zhp)
	{
	return (zhp->zfs_props == zhp->zfs_recvd_props);
	}

	static void
	zfs_set_recvd_props_mode(zfs_handle_t zhp, uint64_t cookie)
	{
	*cookie = (uint64_t)(uintptr_t)zhp->zfs_props;
	zhp->zfs_props = zhp->zfs_recvd_props;
	}

	static void
	zfs_unset_recvd_props_mode(zfs_handle_t zhp, uint64_t cookie)
	{
	zhp->zfs_props = (nvlist_t )(uintptr_t)cookie;
	*cookie = 0;
	}

	/*
	* Internal function for getting a numeric property. Both zfs_prop_get() and
	* zfs_prop_get_int() are built using this interface.
	*
	* Certain properties can be overridden using 'mount -o'. In this case, scan
	* the contents of the /proc/self/mounts entry, searching for the
	* appropriate options. If they differ from the on-disk values, report the
	* current values and mark the source "temporary".
	*/
	static int
	get_numeric_property(zfs_handle_t zhp, zfs_prop_t prop, zprop_source_t src,
	char *source, uint64_t val)
	{
	zfs_cmd_t zc = {"\0"};
	nvlist_t *zplprops = NULL;
	struct mnttab mnt;
	const char *mntopt_on = NULL;
	const char *mntopt_off = NULL;
	boolean_t received = zfs_is_recvd_props_mode(zhp);

	*source = NULL;

	/*
	* If the property is being fetched for a snapshot, check whether
	* the property is valid for the snapshot's head dataset type.
	*/
	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT &&
	!zfs_prop_valid_for_type(prop, zhp->zfs_head_type, B_TRUE)) {
	*val = zfs_prop_default_numeric(prop);
	return (-1);
	}

	switch (prop) {
	case ZFS_PROP_ATIME:
	mntopt_on = MNTOPT_ATIME;
	mntopt_off = MNTOPT_NOATIME;
	break;

	case ZFS_PROP_RELATIME:
	mntopt_on = MNTOPT_RELATIME;
	mntopt_off = MNTOPT_NORELATIME;
	break;

	case ZFS_PROP_DEVICES:
	mntopt_on = MNTOPT_DEVICES;
	mntopt_off = MNTOPT_NODEVICES;
	break;

	case ZFS_PROP_EXEC:
	mntopt_on = MNTOPT_EXEC;
	mntopt_off = MNTOPT_NOEXEC;
	break;

	case ZFS_PROP_READONLY:
	mntopt_on = MNTOPT_RO;
	mntopt_off = MNTOPT_RW;
	break;

	case ZFS_PROP_SETUID:
	mntopt_on = MNTOPT_SETUID;
	mntopt_off = MNTOPT_NOSETUID;
	break;

	case ZFS_PROP_XATTR:
	mntopt_on = MNTOPT_XATTR;
	mntopt_off = MNTOPT_NOXATTR;
	break;

	case ZFS_PROP_NBMAND:
	mntopt_on = MNTOPT_NBMAND;
	mntopt_off = MNTOPT_NONBMAND;
	break;

	default:
	break;
	}

	/*
	* Because looking up the mount options is potentially expensive
	* (iterating over all of /proc/self/mounts), we defer its
	* calculation until we're looking up a property which requires
	* its presence.
	*/
	if (!zhp->zfs_mntcheck &&
	(mntopt_on != NULL \|\| prop == ZFS_PROP_MOUNTED)) {
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	struct mnttab entry;

	if (libzfs_mnttab_find(hdl, zhp->zfs_name, &entry) == 0)
	zhp->zfs_mntopts = zfs_strdup(hdl,
	entry.mnt_mntopts);

	zhp->zfs_mntcheck = B_TRUE;
	}

	if (zhp->zfs_mntopts == NULL)
	mnt.mnt_mntopts = (char *)"";
	else
	mnt.mnt_mntopts = zhp->zfs_mntopts;

	switch (prop) {
	case ZFS_PROP_ATIME:
	case ZFS_PROP_RELATIME:
	case ZFS_PROP_DEVICES:
	case ZFS_PROP_EXEC:
	case ZFS_PROP_READONLY:
	case ZFS_PROP_SETUID:
	#ifndef __FreeBSD__
	case ZFS_PROP_XATTR:
	#endif
	case ZFS_PROP_NBMAND:
	*val = getprop_uint64(zhp, prop, source);

	if (received)
	break;

	if (hasmntopt(&mnt, mntopt_on) && !*val) {
	*val = B_TRUE;
	if (src)
	*src = ZPROP_SRC_TEMPORARY;
	} else if (hasmntopt(&mnt, mntopt_off) && *val) {
	*val = B_FALSE;
	if (src)
	*src = ZPROP_SRC_TEMPORARY;
	}
	break;

	case ZFS_PROP_CANMOUNT:
	case ZFS_PROP_VOLSIZE:
	case ZFS_PROP_QUOTA:
	case ZFS_PROP_REFQUOTA:
	case ZFS_PROP_RESERVATION:
	case ZFS_PROP_REFRESERVATION:
	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	case ZFS_PROP_FILESYSTEM_COUNT:
	case ZFS_PROP_SNAPSHOT_COUNT:
	*val = getprop_uint64(zhp, prop, source);

	if (*source == NULL) {
	/* not default, must be local */
	*source = zhp->zfs_name;
	}
	break;

	case ZFS_PROP_MOUNTED:
	*val = (zhp->zfs_mntopts != NULL);
	break;

	case ZFS_PROP_NUMCLONES:
	*val = zhp->zfs_dmustats.dds_num_clones;
	break;

	case ZFS_PROP_VERSION:
	case ZFS_PROP_NORMALIZE:
	case ZFS_PROP_UTF8ONLY:
	case ZFS_PROP_CASE:
	zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_ZPLPROPS, &zc)) {
	zcmd_free_nvlists(&zc);
	if (prop == ZFS_PROP_VERSION &&
	zhp->zfs_type == ZFS_TYPE_VOLUME)
	*val = zfs_prop_default_numeric(prop);
	return (-1);
	}
	if (zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &zplprops) != 0 \|\|
	nvlist_lookup_uint64(zplprops, zfs_prop_to_name(prop),
	val) != 0) {
	zcmd_free_nvlists(&zc);
	return (-1);
	}
	nvlist_free(zplprops);
	zcmd_free_nvlists(&zc);
	break;

	case ZFS_PROP_INCONSISTENT:
	*val = zhp->zfs_dmustats.dds_inconsistent;
	break;

	case ZFS_PROP_REDACTED:
	*val = zhp->zfs_dmustats.dds_redacted;
	break;

	case ZFS_PROP_CREATETXG:
	/*
	* We can directly read createtxg property from zfs
	* handle for Filesystem, Snapshot and ZVOL types.
	*/
	if ((zhp->zfs_type == ZFS_TYPE_FILESYSTEM) \|\|
	(zhp->zfs_type == ZFS_TYPE_SNAPSHOT) \|\|
	(zhp->zfs_type == ZFS_TYPE_VOLUME)) {
	*val = zhp->zfs_dmustats.dds_creation_txg;
	break;
	}
	zfs_fallthrough;

	default:
	switch (zfs_prop_get_type(prop)) {
	case PROP_TYPE_NUMBER:
	case PROP_TYPE_INDEX:
	*val = getprop_uint64(zhp, prop, source);
	/*
	* If we tried to use a default value for a
	* readonly property, it means that it was not
	* present. Note this only applies to "truly"
	* readonly properties, not set-once properties
	* like volblocksize.
	*/
	if (zfs_prop_readonly(prop) &&
	!zfs_prop_setonce(prop) &&
	source != NULL && (source)[0] == '\0') {
	*source = NULL;
	return (-1);
	}
	break;

	case PROP_TYPE_STRING:
	default:
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"cannot get non-numeric property"));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADPROP,
	dgettext(TEXT_DOMAIN, "internal error")));
	}
	}

	return (0);
	}

	/*
	* Calculate the source type, given the raw source string.
	*/
	static void
	get_source(zfs_handle_t zhp, zprop_source_t srctype, char *source,
	char *statbuf, size_t statlen)
	{
	if (statbuf == NULL \|\|
	srctype == NULL \|\| *srctype == ZPROP_SRC_TEMPORARY) {
	return;
	}

	if (source == NULL) {
	*srctype = ZPROP_SRC_NONE;
	} else if (source[0] == '\0') {
	*srctype = ZPROP_SRC_DEFAULT;
	} else if (strstr(source, ZPROP_SOURCE_VAL_RECVD) != NULL) {
	*srctype = ZPROP_SRC_RECEIVED;
	} else {
	if (strcmp(source, zhp->zfs_name) == 0) {
	*srctype = ZPROP_SRC_LOCAL;
	} else {
	(void) strlcpy(statbuf, source, statlen);
	*srctype = ZPROP_SRC_INHERITED;
	}
	}

	}

	int
	zfs_prop_get_recvd(zfs_handle_t zhp, const char propname, char *propbuf,
	size_t proplen, boolean_t literal)
	{
	zfs_prop_t prop;
	int err = 0;

	if (zhp->zfs_recvd_props == NULL)
	if (get_recvd_props_ioctl(zhp) != 0)
	return (-1);

	prop = zfs_name_to_prop(propname);

	if (prop != ZPROP_USERPROP) {
	uint64_t cookie;
	if (!nvlist_exists(zhp->zfs_recvd_props, propname))
	return (-1);
	zfs_set_recvd_props_mode(zhp, &cookie);
	err = zfs_prop_get(zhp, prop, propbuf, proplen,
	NULL, NULL, 0, literal);
	zfs_unset_recvd_props_mode(zhp, &cookie);
	} else {
	nvlist_t *propval;
	char *recvdval;
	if (nvlist_lookup_nvlist(zhp->zfs_recvd_props,
	propname, &propval) != 0)
	return (-1);
	recvdval = fnvlist_lookup_string(propval, ZPROP_VALUE);
	(void) strlcpy(propbuf, recvdval, proplen);
	}

	return (err == 0 ? 0 : -1);
	}

	static int
	get_clones_string(zfs_handle_t zhp, char propbuf, size_t proplen)
	{
	nvlist_t *value;
	nvpair_t *pair;

	value = zfs_get_clones_nvl(zhp);
	if (value == NULL \|\| nvlist_empty(value))
	return (-1);

	propbuf[0] = '\0';
	for (pair = nvlist_next_nvpair(value, NULL); pair != NULL;
	pair = nvlist_next_nvpair(value, pair)) {
	if (propbuf[0] != '\0')
	(void) strlcat(propbuf, ",", proplen);
	(void) strlcat(propbuf, nvpair_name(pair), proplen);
	}

	return (0);
	}

	struct get_clones_arg {
	uint64_t numclones;
	nvlist_t *value;
	const char *origin;
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	};

	static int
	get_clones_cb(zfs_handle_t zhp, void arg)
	{
	struct get_clones_arg *gca = arg;

	if (gca->numclones == 0) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get(zhp, ZFS_PROP_ORIGIN, gca->buf, sizeof (gca->buf),
	NULL, NULL, 0, B_TRUE) != 0)
	goto out;
	if (strcmp(gca->buf, gca->origin) == 0) {
	fnvlist_add_boolean(gca->value, zfs_get_name(zhp));
	gca->numclones--;
	}

	out:
	(void) zfs_iter_children(zhp, get_clones_cb, gca);
	zfs_close(zhp);
	return (0);
	}

	nvlist_t *
	zfs_get_clones_nvl(zfs_handle_t *zhp)
	{
	nvlist_t nv, value;

	if (nvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_CLONES), &nv) != 0) {
	struct get_clones_arg gca;

	/*
	* if this is a snapshot, then the kernel wasn't able
	* to get the clones. Do it by slowly iterating.
	*/
	if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT)
	return (NULL);
	if (nvlist_alloc(&nv, NV_UNIQUE_NAME, 0) != 0)
	return (NULL);
	if (nvlist_alloc(&value, NV_UNIQUE_NAME, 0) != 0) {
	nvlist_free(nv);
	return (NULL);
	}

	gca.numclones = zfs_prop_get_int(zhp, ZFS_PROP_NUMCLONES);
	gca.value = value;
	gca.origin = zhp->zfs_name;

	if (gca.numclones != 0) {
	zfs_handle_t *root;
	char pool[ZFS_MAX_DATASET_NAME_LEN];
	char *cp = pool;

	/* get the pool name */
	(void) strlcpy(pool, zhp->zfs_name, sizeof (pool));
	(void) strsep(&cp, "/@");
	root = zfs_open(zhp->zfs_hdl, pool,
	ZFS_TYPE_FILESYSTEM);
	if (root == NULL) {
	nvlist_free(nv);
	nvlist_free(value);
	return (NULL);
	}

	(void) get_clones_cb(root, &gca);
	}

	if (gca.numclones != 0 \|\|
	nvlist_add_nvlist(nv, ZPROP_VALUE, value) != 0 \|\|
	nvlist_add_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_CLONES), nv) != 0) {
	nvlist_free(nv);
	nvlist_free(value);
	return (NULL);
	}
	nvlist_free(nv);
	nvlist_free(value);
	nv = fnvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_CLONES));
	}

	return (fnvlist_lookup_nvlist(nv, ZPROP_VALUE));
	}

	static int
	get_rsnaps_string(zfs_handle_t zhp, char propbuf, size_t proplen)
	{
	nvlist_t *value;
	uint64_t *snaps;
	uint_t nsnaps;

	if (nvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS), &value) != 0)
	return (-1);
	if (nvlist_lookup_uint64_array(value, ZPROP_VALUE, &snaps,
	&nsnaps) != 0)
	return (-1);
	if (nsnaps == 0) {
	/* There's no redaction snapshots; pass a special value back */
	(void) snprintf(propbuf, proplen, "none");
	return (0);
	}
	propbuf[0] = '\0';
	for (int i = 0; i < nsnaps; i++) {
	char buf[128];
	if (propbuf[0] != '\0')
	(void) strlcat(propbuf, ",", proplen);
	(void) snprintf(buf, sizeof (buf), "%llu",
	(u_longlong_t)snaps[i]);
	(void) strlcat(propbuf, buf, proplen);
	}

	return (0);
	}

	/*
	* Accepts a property and value and checks that the value
	* matches the one found by the channel program. If they are
	* not equal, print both of them.
	*/
	static void
	zcp_check(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t intval,
	const char *strval)
	{
	if (!zhp->zfs_hdl->libzfs_prop_debug)
	return;
	int error;
	char *poolname = zhp->zpool_hdl->zpool_name;
	const char *prop_name = zfs_prop_to_name(prop);
	const char *program =
	"args = ...\n"
	"ds = args['dataset']\n"
	"prop = args['property']\n"
	"value, setpoint = zfs.get_prop(ds, prop)\n"
	"return {value=value, setpoint=setpoint}\n";
	nvlist_t *outnvl;
	nvlist_t *retnvl;
	nvlist_t *argnvl = fnvlist_alloc();

	fnvlist_add_string(argnvl, "dataset", zhp->zfs_name);
	fnvlist_add_string(argnvl, "property", zfs_prop_to_name(prop));

	error = lzc_channel_program_nosync(poolname, program,
	10 * 1000 * 1000, 10 * 1024 * 1024, argnvl, &outnvl);

	if (error == 0) {
	retnvl = fnvlist_lookup_nvlist(outnvl, "return");
	if (zfs_prop_get_type(prop) == PROP_TYPE_NUMBER) {
	int64_t ans;
	error = nvlist_lookup_int64(retnvl, "value", &ans);
	if (error != 0) {
	(void) fprintf(stderr, "%s: zcp check error: "
	"%u\n", prop_name, error);
	return;
	}
	if (ans != intval) {
	(void) fprintf(stderr, "%s: zfs found %llu, "
	"but zcp found %llu\n", prop_name,
	(u_longlong_t)intval, (u_longlong_t)ans);
	}
	} else {
	char *str_ans;
	error = nvlist_lookup_string(retnvl, "value", &str_ans);
	if (error != 0) {
	(void) fprintf(stderr, "%s: zcp check error: "
	"%u\n", prop_name, error);
	return;
	}
	if (strcmp(strval, str_ans) != 0) {
	(void) fprintf(stderr,
	"%s: zfs found '%s', but zcp found '%s'\n",
	prop_name, strval, str_ans);
	}
	}
	} else {
	(void) fprintf(stderr, "%s: zcp check failed, channel program "
	"error: %u\n", prop_name, error);
	}
	nvlist_free(argnvl);
	nvlist_free(outnvl);
	}

	/*
	* Retrieve a property from the given object. If 'literal' is specified, then
	* numbers are left as exact values. Otherwise, numbers are converted to a
	* human-readable form.
	*
	* Returns 0 on success, or -1 on error.
	*/
	int
	zfs_prop_get(zfs_handle_t zhp, zfs_prop_t prop, char propbuf, size_t proplen,
	zprop_source_t src, char statbuf, size_t statlen, boolean_t literal)
	{
	char *source = NULL;
	uint64_t val;
	const char *str;
	const char *strval;
	boolean_t received = zfs_is_recvd_props_mode(zhp);

	/*
	* Check to see if this property applies to our object
	*/
	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
	return (-1);

	if (received && zfs_prop_readonly(prop))
	return (-1);

	if (src)
	*src = ZPROP_SRC_NONE;

	switch (prop) {
	case ZFS_PROP_CREATION:
	/*
	* 'creation' is a time_t stored in the statistics. We convert
	* this into a string unless 'literal' is specified.
	*/
	{
	val = getprop_uint64(zhp, prop, &source);
	time_t time = (time_t)val;
	struct tm t;

	if (literal \|\|
	localtime_r(&time, &t) == NULL \|\|
	strftime(propbuf, proplen, "%a %b %e %k:%M %Y",
	&t) == 0)
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	}
	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_MOUNTPOINT:
	/*
	* Getting the precise mountpoint can be tricky.
	*
	* - for 'none' or 'legacy', return those values.
	* - for inherited mountpoints, we want to take everything
	* after our ancestor and append it to the inherited value.
	*
	* If the pool has an alternate root, we want to prepend that
	* root to any values we return.
	*/

	str = getprop_string(zhp, prop, &source);

	if (str[0] == '/') {
	char buf[MAXPATHLEN];
	char *root = buf;
	const char *relpath;

	/*
	* If we inherit the mountpoint, even from a dataset
	* with a received value, the source will be the path of
	* the dataset we inherit from. If source is
	* ZPROP_SOURCE_VAL_RECVD, the received value is not
	* inherited.
	*/
	if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) {
	relpath = "";
	} else {
	relpath = zhp->zfs_name + strlen(source);
	if (relpath[0] == '/')
	relpath++;
	}

	if ((zpool_get_prop(zhp->zpool_hdl,
	ZPOOL_PROP_ALTROOT, buf, MAXPATHLEN, NULL,
	B_FALSE)) \|\| (strcmp(root, "-") == 0))
	root[0] = '\0';
	/*
	* Special case an alternate root of '/'. This will
	* avoid having multiple leading slashes in the
	* mountpoint path.
	*/
	if (strcmp(root, "/") == 0)
	root++;

	/*
	* If the mountpoint is '/' then skip over this
	* if we are obtaining either an alternate root or
	* an inherited mountpoint.
	*/
	if (str[1] == '\0' && (root[0] != '\0' \|\|
	relpath[0] != '\0'))
	str++;

	if (relpath[0] == '\0')
	(void) snprintf(propbuf, proplen, "%s%s",
	root, str);
	else
	(void) snprintf(propbuf, proplen, "%s%s%s%s",
	root, str, relpath[0] == '@' ? "" : "/",
	relpath);
	} else {
	/* 'legacy' or 'none' */
	(void) strlcpy(propbuf, str, proplen);
	}
	zcp_check(zhp, prop, 0, propbuf);
	break;

	case ZFS_PROP_ORIGIN:
	str = getprop_string(zhp, prop, &source);
	if (str == NULL)
	return (-1);
	(void) strlcpy(propbuf, str, proplen);
	zcp_check(zhp, prop, 0, str);
	break;

	case ZFS_PROP_REDACT_SNAPS:
	if (get_rsnaps_string(zhp, propbuf, proplen) != 0)
	return (-1);
	break;

	case ZFS_PROP_CLONES:
	if (get_clones_string(zhp, propbuf, proplen) != 0)
	return (-1);
	break;

	case ZFS_PROP_QUOTA:
	case ZFS_PROP_REFQUOTA:
	case ZFS_PROP_RESERVATION:
	case ZFS_PROP_REFRESERVATION:

	if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
	return (-1);
	/*
	* If quota or reservation is 0, we translate this into 'none'
	* (unless literal is set), and indicate that it's the default
	* value. Otherwise, we print the number nicely and indicate
	* that its set locally.
	*/
	if (val == 0) {
	if (literal)
	(void) strlcpy(propbuf, "0", proplen);
	else
	(void) strlcpy(propbuf, "none", proplen);
	} else {
	if (literal)
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	else
	zfs_nicebytes(val, propbuf, proplen);
	}
	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	case ZFS_PROP_FILESYSTEM_COUNT:
	case ZFS_PROP_SNAPSHOT_COUNT:

	if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
	return (-1);

	/*
	* If limit is UINT64_MAX, we translate this into 'none', and
	* indicate that it's the default value. Otherwise, we print
	* the number nicely and indicate that it's set locally.
	*/
	if (val == UINT64_MAX) {
	(void) strlcpy(propbuf, "none", proplen);
	} else if (literal) {
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	} else {
	zfs_nicenum(val, propbuf, proplen);
	}

	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_REFRATIO:
	case ZFS_PROP_COMPRESSRATIO:
	if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
	return (-1);
	if (literal)
	(void) snprintf(propbuf, proplen, "%llu.%02llu",
	(u_longlong_t)(val / 100),
	(u_longlong_t)(val % 100));
	else
	(void) snprintf(propbuf, proplen, "%llu.%02llux",
	(u_longlong_t)(val / 100),
	(u_longlong_t)(val % 100));
	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_TYPE:
	switch (zhp->zfs_type) {
	case ZFS_TYPE_FILESYSTEM:
	str = "filesystem";
	break;
	case ZFS_TYPE_VOLUME:
	str = "volume";
	break;
	case ZFS_TYPE_SNAPSHOT:
	str = "snapshot";
	break;
	case ZFS_TYPE_BOOKMARK:
	str = "bookmark";
	break;
	default:
	abort();
	}
	(void) snprintf(propbuf, proplen, "%s", str);
	zcp_check(zhp, prop, 0, propbuf);
	break;

	case ZFS_PROP_MOUNTED:
	/*
	* The 'mounted' property is a pseudo-property that described
	* whether the filesystem is currently mounted. Even though
	* it's a boolean value, the typical values of "on" and "off"
	* don't make sense, so we translate to "yes" and "no".
	*/
	if (get_numeric_property(zhp, ZFS_PROP_MOUNTED,
	src, &source, &val) != 0)
	return (-1);
	if (val)
	(void) strlcpy(propbuf, "yes", proplen);
	else
	(void) strlcpy(propbuf, "no", proplen);
	break;

	case ZFS_PROP_NAME:
	/*
	* The 'name' property is a pseudo-property derived from the
	* dataset name. It is presented as a real property to simplify
	* consumers.
	*/
	(void) strlcpy(propbuf, zhp->zfs_name, proplen);
	zcp_check(zhp, prop, 0, propbuf);
	break;

	case ZFS_PROP_MLSLABEL:
	{
	#ifdef HAVE_MLSLABEL
	m_label_t *new_sl = NULL;
	char ascii = NULL; / human readable label */

	(void) strlcpy(propbuf,
	getprop_string(zhp, prop, &source), proplen);

	if (literal \|\| (strcasecmp(propbuf,
	ZFS_MLSLABEL_DEFAULT) == 0))
	break;

	/*
	* Try to translate the internal hex string to
	* human-readable output. If there are any
	* problems just use the hex string.
	*/

	if (str_to_label(propbuf, &new_sl, MAC_LABEL,
	L_NO_CORRECTION, NULL) == -1) {
	m_label_free(new_sl);
	break;
	}

	if (label_to_str(new_sl, &ascii, M_LABEL,
	DEF_NAMES) != 0) {
	if (ascii)
	free(ascii);
	m_label_free(new_sl);
	break;
	}
	m_label_free(new_sl);

	(void) strlcpy(propbuf, ascii, proplen);
	free(ascii);
	#else
	(void) strlcpy(propbuf,
	getprop_string(zhp, prop, &source), proplen);
	#endif /* HAVE_MLSLABEL */
	}
	break;

	case ZFS_PROP_GUID:
	case ZFS_PROP_KEY_GUID:
	case ZFS_PROP_IVSET_GUID:
	case ZFS_PROP_CREATETXG:
	case ZFS_PROP_OBJSETID:
	case ZFS_PROP_PBKDF2_ITERS:
	/*
	* These properties are stored as numbers, but they are
	* identifiers or counters.
	* We don't want them to be pretty printed, because pretty
	* printing truncates their values making them useless.
	*/
	if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
	return (-1);
	(void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val);
	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_REFERENCED:
	case ZFS_PROP_AVAILABLE:
	case ZFS_PROP_USED:
	case ZFS_PROP_USEDSNAP:
	case ZFS_PROP_USEDDS:
	case ZFS_PROP_USEDREFRESERV:
	case ZFS_PROP_USEDCHILD:
	if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
	return (-1);
	if (literal) {
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	} else {
	zfs_nicebytes(val, propbuf, proplen);
	}
	zcp_check(zhp, prop, val, NULL);
	break;

	case ZFS_PROP_SNAPSHOTS_CHANGED:
	{
	if ((get_numeric_property(zhp, prop, src, &source,
	&val) != 0) \|\| val == 0) {
	return (-1);
	}

	time_t time = (time_t)val;
	struct tm t;

	if (literal \|\|
	localtime_r(&time, &t) == NULL \|\|
	strftime(propbuf, proplen, "%a %b %e %k:%M:%S %Y",
	&t) == 0)
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	}
	zcp_check(zhp, prop, val, NULL);
	break;

	default:
	switch (zfs_prop_get_type(prop)) {
	case PROP_TYPE_NUMBER:
	if (get_numeric_property(zhp, prop, src,
	&source, &val) != 0) {
	return (-1);
	}

	if (literal) {
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)val);
	} else {
	zfs_nicenum(val, propbuf, proplen);
	}
	zcp_check(zhp, prop, val, NULL);
	break;

	case PROP_TYPE_STRING:
	str = getprop_string(zhp, prop, &source);
	if (str == NULL)
	return (-1);

	(void) strlcpy(propbuf, str, proplen);
	zcp_check(zhp, prop, 0, str);
	break;

	case PROP_TYPE_INDEX:
	if (get_numeric_property(zhp, prop, src,
	&source, &val) != 0)
	return (-1);
	if (zfs_prop_index_to_string(prop, val, &strval) != 0)
	return (-1);

	(void) strlcpy(propbuf, strval, proplen);
	zcp_check(zhp, prop, 0, strval);
	break;

	default:
	abort();
	}
	}

	get_source(zhp, src, source, statbuf, statlen);

	return (0);
	}

	/*
	* Utility function to get the given numeric property. Does no validation that
	* the given property is the appropriate type; should only be used with
	* hard-coded property types.
	*/
	uint64_t
	zfs_prop_get_int(zfs_handle_t *zhp, zfs_prop_t prop)
	{
	char *source;
	uint64_t val = 0;

	(void) get_numeric_property(zhp, prop, NULL, &source, &val);

	return (val);
	}

	static int
	zfs_prop_set_int(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t val)
	{
	char buf[64];

	(void) snprintf(buf, sizeof (buf), "%llu", (longlong_t)val);
	return (zfs_prop_set(zhp, zfs_prop_to_name(prop), buf));
	}

	/*
	* Similar to zfs_prop_get(), but returns the value as an integer.
	*/
	int
	zfs_prop_get_numeric(zfs_handle_t zhp, zfs_prop_t prop, uint64_t value,
	zprop_source_t src, char statbuf, size_t statlen)
	{
	char *source;

	/*
	* Check to see if this property applies to our object
	*/
	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE)) {
	return (zfs_error_fmt(zhp->zfs_hdl, EZFS_PROPTYPE,
	dgettext(TEXT_DOMAIN, "cannot get property '%s'"),
	zfs_prop_to_name(prop)));
	}

	if (src)
	*src = ZPROP_SRC_NONE;

	if (get_numeric_property(zhp, prop, src, &source, value) != 0)
	return (-1);

	get_source(zhp, src, source, statbuf, statlen);

	return (0);
	}

	#ifdef HAVE_IDMAP
	static int
	idmap_id_to_numeric_domain_rid(uid_t id, boolean_t isuser,
	char *domainp, idmap_rid_t ridp)
	{
	idmap_get_handle_t *get_hdl = NULL;
	idmap_stat status;
	int err = EINVAL;

	if (idmap_get_create(&get_hdl) != IDMAP_SUCCESS)
	goto out;

	if (isuser) {
	err = idmap_get_sidbyuid(get_hdl, id,
	IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
	} else {
	err = idmap_get_sidbygid(get_hdl, id,
	IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
	}
	if (err == IDMAP_SUCCESS &&
	idmap_get_mappings(get_hdl) == IDMAP_SUCCESS &&
	status == IDMAP_SUCCESS)
	err = 0;
	else
	err = EINVAL;
	out:
	if (get_hdl)
	idmap_get_destroy(get_hdl);
	return (err);
	}
	#endif /* HAVE_IDMAP */

	/*
	* convert the propname into parameters needed by kernel
	* Eg: userquota@ahrens -> ZFS_PROP_USERQUOTA, "", 126829
	* Eg: userused@matt@domain -> ZFS_PROP_USERUSED, "S-1-123-456", 789
	* Eg: groupquota@staff -> ZFS_PROP_GROUPQUOTA, "", 1234
	* Eg: groupused@staff -> ZFS_PROP_GROUPUSED, "", 1234
	* Eg: projectquota@123 -> ZFS_PROP_PROJECTQUOTA, "", 123
	* Eg: projectused@789 -> ZFS_PROP_PROJECTUSED, "", 789
	*/
	static int
	userquota_propname_decode(const char *propname, boolean_t zoned,
	zfs_userquota_prop_t typep, char domain, int domainlen, uint64_t *ridp)
	{
	zfs_userquota_prop_t type;
	char *cp;
	boolean_t isuser;
	boolean_t isgroup;
	boolean_t isproject;
	struct passwd *pw;
	struct group *gr;

	domain[0] = '\0';

	/* Figure out the property type ({user\|group\|project}{quota\|space}) */
	for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) {
	if (strncmp(propname, zfs_userquota_prop_prefixes[type],
	strlen(zfs_userquota_prop_prefixes[type])) == 0)
	break;
	}
	if (type == ZFS_NUM_USERQUOTA_PROPS)
	return (EINVAL);
	*typep = type;

	isuser = (type == ZFS_PROP_USERQUOTA \|\| type == ZFS_PROP_USERUSED \|\|
	type == ZFS_PROP_USEROBJQUOTA \|\|
	type == ZFS_PROP_USEROBJUSED);
	isgroup = (type == ZFS_PROP_GROUPQUOTA \|\| type == ZFS_PROP_GROUPUSED \|\|
	type == ZFS_PROP_GROUPOBJQUOTA \|\|
	type == ZFS_PROP_GROUPOBJUSED);
	isproject = (type == ZFS_PROP_PROJECTQUOTA \|\|
	type == ZFS_PROP_PROJECTUSED \|\| type == ZFS_PROP_PROJECTOBJQUOTA \|\|
	type == ZFS_PROP_PROJECTOBJUSED);

	cp = strchr(propname, '@') + 1;

	if (isuser && (pw = getpwnam(cp)) != NULL) {
	if (zoned && getzoneid() == GLOBAL_ZONEID)
	return (ENOENT);
	*ridp = pw->pw_uid;
	} else if (isgroup && (gr = getgrnam(cp)) != NULL) {
	if (zoned && getzoneid() == GLOBAL_ZONEID)
	return (ENOENT);
	*ridp = gr->gr_gid;
	} else if (!isproject && strchr(cp, '@')) {
	#ifdef HAVE_IDMAP
	/*
	* It's a SID name (eg "user@domain") that needs to be
	* turned into S-1-domainID-RID.
	*/
	directory_error_t e;
	char *numericsid = NULL;
	char *end;

	if (zoned && getzoneid() == GLOBAL_ZONEID)
	return (ENOENT);
	if (isuser) {
	e = directory_sid_from_user_name(NULL,
	cp, &numericsid);
	} else {
	e = directory_sid_from_group_name(NULL,
	cp, &numericsid);
	}
	if (e != NULL) {
	directory_error_free(e);
	return (ENOENT);
	}
	if (numericsid == NULL)
	return (ENOENT);
	cp = numericsid;
	(void) strlcpy(domain, cp, domainlen);
	cp = strrchr(domain, '-');
	*cp = '\0';
	cp++;

	errno = 0;
	*ridp = strtoull(cp, &end, 10);
	free(numericsid);

	if (errno != 0 \|\| *end != '\0')
	return (EINVAL);
	#else
	(void) domainlen;
	return (ENOSYS);
	#endif /* HAVE_IDMAP */
	} else {
	/* It's a user/group/project ID (eg "12345"). */
	uid_t id;
	char *end;
	id = strtoul(cp, &end, 10);
	if (*end != '\0')
	return (EINVAL);
	if (id > MAXUID && !isproject) {
	#ifdef HAVE_IDMAP
	/* It's an ephemeral ID. */
	idmap_rid_t rid;
	char *mapdomain;

	if (idmap_id_to_numeric_domain_rid(id, isuser,
	&mapdomain, &rid) != 0)
	return (ENOENT);
	(void) strlcpy(domain, mapdomain, domainlen);
	*ridp = rid;
	#else
	return (ENOSYS);
	#endif /* HAVE_IDMAP */
	} else {
	*ridp = id;
	}
	}

	return (0);
	}

	static int
	zfs_prop_get_userquota_common(zfs_handle_t zhp, const char propname,
	uint64_t propvalue, zfs_userquota_prop_t typep)
	{
	int err;
	zfs_cmd_t zc = {"\0"};

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	err = userquota_propname_decode(propname,
	zfs_prop_get_int(zhp, ZFS_PROP_ZONED),
	typep, zc.zc_value, sizeof (zc.zc_value), &zc.zc_guid);
	zc.zc_objset_type = *typep;
	if (err)
	return (err);

	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_USERSPACE_ONE, &zc);
	if (err)
	return (err);

	*propvalue = zc.zc_cookie;
	return (0);
	}

	int
	zfs_prop_get_userquota_int(zfs_handle_t zhp, const char propname,
	uint64_t *propvalue)
	{
	zfs_userquota_prop_t type;

	return (zfs_prop_get_userquota_common(zhp, propname, propvalue,
	&type));
	}

	int
	zfs_prop_get_userquota(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal)
	{
	int err;
	uint64_t propvalue;
	zfs_userquota_prop_t type;

	err = zfs_prop_get_userquota_common(zhp, propname, &propvalue,
	&type);

	if (err)
	return (err);

	if (literal) {
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)propvalue);
	} else if (propvalue == 0 &&
	(type == ZFS_PROP_USERQUOTA \|\| type == ZFS_PROP_GROUPQUOTA \|\|
	type == ZFS_PROP_USEROBJQUOTA \|\| type == ZFS_PROP_GROUPOBJQUOTA \|\|
	type == ZFS_PROP_PROJECTQUOTA \|\|
	type == ZFS_PROP_PROJECTOBJQUOTA)) {
	(void) strlcpy(propbuf, "none", proplen);
	} else if (type == ZFS_PROP_USERQUOTA \|\| type == ZFS_PROP_GROUPQUOTA \|\|
	type == ZFS_PROP_USERUSED \|\| type == ZFS_PROP_GROUPUSED \|\|
	type == ZFS_PROP_PROJECTUSED \|\| type == ZFS_PROP_PROJECTQUOTA) {
	zfs_nicebytes(propvalue, propbuf, proplen);
	} else {
	zfs_nicenum(propvalue, propbuf, proplen);
	}
	return (0);
	}

	/*
	* propname must start with "written@" or "written#".
	*/
	int
	zfs_prop_get_written_int(zfs_handle_t zhp, const char propname,
	uint64_t *propvalue)
	{
	int err;
	zfs_cmd_t zc = {"\0"};
	const char *snapname;

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	assert(zfs_prop_written(propname));
	snapname = propname + strlen("written@");
	if (strchr(snapname, '@') != NULL \|\| strchr(snapname, '#') != NULL) {
	/* full snapshot or bookmark name specified */
	(void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value));
	} else {
	/* snapname is the short name, append it to zhp's fsname */
	char *cp;

	(void) strlcpy(zc.zc_value, zhp->zfs_name,
	sizeof (zc.zc_value));
	cp = strchr(zc.zc_value, '@');
	if (cp != NULL)
	*cp = '\0';
	(void) strlcat(zc.zc_value, snapname - 1, sizeof (zc.zc_value));
	}

	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SPACE_WRITTEN, &zc);
	if (err)
	return (err);

	*propvalue = zc.zc_cookie;
	return (0);
	}

	int
	zfs_prop_get_written(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal)
	{
	int err;
	uint64_t propvalue;

	err = zfs_prop_get_written_int(zhp, propname, &propvalue);

	if (err)
	return (err);

	if (literal) {
	(void) snprintf(propbuf, proplen, "%llu",
	(u_longlong_t)propvalue);
	} else {
	zfs_nicebytes(propvalue, propbuf, proplen);
	}

	return (0);
	}

	/*
	* Returns the name of the given zfs handle.
	*/
	const char *
	zfs_get_name(const zfs_handle_t *zhp)
	{
	return (zhp->zfs_name);
	}

	/*
	* Returns the name of the parent pool for the given zfs handle.
	*/
	const char *
	zfs_get_pool_name(const zfs_handle_t *zhp)
	{
	return (zhp->zpool_hdl->zpool_name);
	}

	/*
	* Returns the type of the given zfs handle.
	*/
	zfs_type_t
	zfs_get_type(const zfs_handle_t *zhp)
	{
	return (zhp->zfs_type);
	}

	/*
	* Returns the type of the given zfs handle,
	* or, if a snapshot, the type of the snapshotted dataset.
	*/
	zfs_type_t
	zfs_get_underlying_type(const zfs_handle_t *zhp)
	{
	return (zhp->zfs_head_type);
	}

	/*
	* Is one dataset name a child dataset of another?
	*
	* Needs to handle these cases:
	* Dataset 1 "a/foo" "a/foo" "a/foo" "a/foo"
	* Dataset 2 "a/fo" "a/foobar" "a/bar/baz" "a/foo/bar"
	* Descendant? No. No. No. Yes.
	*/
	static boolean_t
	is_descendant(const char ds1, const char ds2)
	{
	size_t d1len = strlen(ds1);

	/* ds2 can't be a descendant if it's smaller */
	if (strlen(ds2) < d1len)
	return (B_FALSE);

	/* otherwise, compare strings and verify that there's a '/' char */
	return (ds2[d1len] == '/' && (strncmp(ds1, ds2, d1len) == 0));
	}

	/*
	* Given a complete name, return just the portion that refers to the parent.
	* Will return -1 if there is no parent (path is just the name of the
	* pool).
	*/
	static int
	parent_name(const char path, char buf, size_t buflen)
	{
	char *slashp;

	(void) strlcpy(buf, path, buflen);

	if ((slashp = strrchr(buf, '/')) == NULL)
	return (-1);
	*slashp = '\0';

	return (0);
	}

	int
	zfs_parent_name(zfs_handle_t zhp, char buf, size_t buflen)
	{
	return (parent_name(zfs_get_name(zhp), buf, buflen));
	}

	/*
	* If accept_ancestor is false, then check to make sure that the given path has
	* a parent, and that it exists. If accept_ancestor is true, then find the
	* closest existing ancestor for the given path. In prefixlen return the
	* length of already existing prefix of the given path. We also fetch the
	* 'zoned' property, which is used to validate property settings when creating
	* new datasets.
	*/
	static int
	check_parents(libzfs_handle_t hdl, const char path, uint64_t *zoned,
	boolean_t accept_ancestor, int *prefixlen)
	{
	zfs_cmd_t zc = {"\0"};
	char parent[ZFS_MAX_DATASET_NAME_LEN];
	char *slash;
	zfs_handle_t *zhp;
	char errbuf[ERRBUFLEN];
	uint64_t is_zoned;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot create '%s'"), path);

	/* get parent, and check to see if this is just a pool */
	if (parent_name(path, parent, sizeof (parent)) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"missing dataset name"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}

	/* check to see if the pool exists */
	if ((slash = strchr(parent, '/')) == NULL)
	slash = parent + strlen(parent);
	- (void) strncpy(zc.zc_name, parent, slash - parent);
	- zc.zc_name[slash - parent] = '\0';
	+ (void) strlcpy(zc.zc_name, parent,
	+ MIN(sizeof (zc.zc_name), slash - parent + 1));
	if (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0 &&
	errno == ENOENT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no such pool '%s'"), zc.zc_name);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}

	/* check to see if the parent dataset exists */
	while ((zhp = make_dataset_handle(hdl, parent)) == NULL) {
	if (errno == ENOENT && accept_ancestor) {
	/*
	* Go deeper to find an ancestor, give up on top level.
	*/
	if (parent_name(parent, parent, sizeof (parent)) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no such pool '%s'"), zc.zc_name);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}
	} else if (errno == ENOENT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"parent does not exist"));
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	} else
	return (zfs_standard_error(hdl, errno, errbuf));
	}

	is_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
	if (zoned != NULL)
	*zoned = is_zoned;

	/* we are in a non-global zone, but parent is in the global zone */
	if (getzoneid() != GLOBAL_ZONEID && !is_zoned) {
	(void) zfs_standard_error(hdl, EPERM, errbuf);
	zfs_close(zhp);
	return (-1);
	}

	/* make sure parent is a filesystem */
	if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"parent is not a filesystem"));
	(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
	zfs_close(zhp);
	return (-1);
	}

	zfs_close(zhp);
	if (prefixlen != NULL)
	*prefixlen = strlen(parent);
	return (0);
	}

	/*
	* Finds whether the dataset of the given type(s) exists.
	*/
	boolean_t
	zfs_dataset_exists(libzfs_handle_t hdl, const char path, zfs_type_t types)
	{
	zfs_handle_t *zhp;

	if (!zfs_validate_name(hdl, path, types, B_FALSE))
	return (B_FALSE);

	/*
	* Try to get stats for the dataset, which will tell us if it exists.
	*/
	if ((zhp = make_dataset_handle(hdl, path)) != NULL) {
	int ds_type = zhp->zfs_type;

	zfs_close(zhp);
	if (types & ds_type)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Given a path to 'target', create all the ancestors between
	* the prefixlen portion of the path, and the target itself.
	* Fail if the initial prefixlen-ancestor does not already exist.
	*/
	int
	create_parents(libzfs_handle_t hdl, char target, int prefixlen)
	{
	zfs_handle_t *h;
	char *cp;
	const char *opname;

	/* make sure prefix exists */
	cp = target + prefixlen;
	if (*cp != '/') {
	assert(strchr(cp, '/') == NULL);
	h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
	} else {
	*cp = '\0';
	h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
	*cp = '/';
	}
	if (h == NULL)
	return (-1);
	zfs_close(h);

	/*
	* Attempt to create, mount, and share any ancestor filesystems,
	* up to the prefixlen-long one.
	*/
	for (cp = target + prefixlen + 1;
	(cp = strchr(cp, '/')) != NULL; *cp = '/', cp++) {

	*cp = '\0';

	h = make_dataset_handle(hdl, target);
	if (h) {
	/* it already exists, nothing to do here */
	zfs_close(h);
	continue;
	}

	if (zfs_create(hdl, target, ZFS_TYPE_FILESYSTEM,
	NULL) != 0) {
	opname = dgettext(TEXT_DOMAIN, "create");
	goto ancestorerr;
	}

	h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
	if (h == NULL) {
	opname = dgettext(TEXT_DOMAIN, "open");
	goto ancestorerr;
	}

	if (zfs_mount(h, NULL, 0) != 0) {
	opname = dgettext(TEXT_DOMAIN, "mount");
	goto ancestorerr;
	}

	if (zfs_share(h, NULL) != 0) {
	opname = dgettext(TEXT_DOMAIN, "share");
	goto ancestorerr;
	}

	zfs_close(h);
	}
	zfs_commit_shares(NULL);

	return (0);

	ancestorerr:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to %s ancestor '%s'"), opname, target);
	return (-1);
	}

	/*
	* Creates non-existing ancestors of the given path.
	*/
	int
	zfs_create_ancestors(libzfs_handle_t hdl, const char path)
	{
	int prefix;
	char *path_copy;
	char errbuf[ERRBUFLEN];
	int rc = 0;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot create '%s'"), path);

	/*
	* Check that we are not passing the nesting limit
	* before we start creating any ancestors.
	*/
	if (dataset_nestcheck(path) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"maximum name nesting depth exceeded"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}

	if (check_parents(hdl, path, NULL, B_TRUE, &prefix) != 0)
	return (-1);

	if ((path_copy = strdup(path)) != NULL) {
	rc = create_parents(hdl, path_copy, prefix);
	free(path_copy);
	}
	if (path_copy == NULL \|\| rc != 0)
	return (-1);

	return (0);
	}

	/*
	* Create a new filesystem or volume.
	*/
	int
	zfs_create(libzfs_handle_t hdl, const char path, zfs_type_t type,
	nvlist_t *props)
	{
	int ret;
	uint64_t size = 0;
	uint64_t blocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
	uint64_t zoned;
	enum lzc_dataset_type ost;
	zpool_handle_t *zpool_handle;
	uint8_t *wkeydata = NULL;
	uint_t wkeylen = 0;
	char errbuf[ERRBUFLEN];
	char parent[ZFS_MAX_DATASET_NAME_LEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot create '%s'"), path);

	/* validate the path, taking care to note the extended error message */
	if (!zfs_validate_name(hdl, path, type, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	if (dataset_nestcheck(path) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"maximum name nesting depth exceeded"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}

	/* validate parents exist */
	if (check_parents(hdl, path, &zoned, B_FALSE, NULL) != 0)
	return (-1);

	/*
	* The failure modes when creating a dataset of a different type over
	* one that already exists is a little strange. In particular, if you
	* try to create a dataset on top of an existing dataset, the ioctl()
	* will return ENOENT, not EEXIST. To prevent this from happening, we
	* first try to see if the dataset exists.
	*/
	if (zfs_dataset_exists(hdl, path, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset already exists"));
	return (zfs_error(hdl, EZFS_EXISTS, errbuf));
	}

	if (type == ZFS_TYPE_VOLUME)
	ost = LZC_DATSET_TYPE_ZVOL;
	else
	ost = LZC_DATSET_TYPE_ZFS;

	/* open zpool handle for prop validation */
	char pool_path[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(pool_path, path, sizeof (pool_path));

	/* truncate pool_path at first slash */
	char *p = strchr(pool_path, '/');
	if (p != NULL)
	*p = '\0';

	if ((zpool_handle = zpool_open(hdl, pool_path)) == NULL)
	return (-1);

	if (props && (props = zfs_valid_proplist(hdl, type, props,
	zoned, NULL, zpool_handle, B_TRUE, errbuf)) == 0) {
	zpool_close(zpool_handle);
	return (-1);
	}
	zpool_close(zpool_handle);

	if (type == ZFS_TYPE_VOLUME) {
	/*
	* If we are creating a volume, the size and block size must
	* satisfy a few restraints. First, the blocksize must be a
	* valid block size between SPA_{MIN,MAX}BLOCKSIZE. Second, the
	* volsize must be a multiple of the block size, and cannot be
	* zero.
	*/
	if (props == NULL \|\| nvlist_lookup_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), &size) != 0) {
	nvlist_free(props);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"missing volume size"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}

	if ((ret = nvlist_lookup_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	&blocksize)) != 0) {
	if (ret == ENOENT) {
	blocksize = zfs_prop_default_numeric(
	ZFS_PROP_VOLBLOCKSIZE);
	} else {
	nvlist_free(props);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"missing volume block size"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	}

	if (size == 0) {
	nvlist_free(props);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"volume size cannot be zero"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}

	if (size % blocksize != 0) {
	nvlist_free(props);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"volume size must be a multiple of volume block "
	"size"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	}

	(void) parent_name(path, parent, sizeof (parent));
	if (zfs_crypto_create(hdl, parent, props, NULL, B_TRUE,
	&wkeydata, &wkeylen) != 0) {
	nvlist_free(props);
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
	}

	/* create the dataset */
	ret = lzc_create(path, ost, props, wkeydata, wkeylen);
	nvlist_free(props);
	if (wkeydata != NULL)
	free(wkeydata);

	/* check for failure */
	if (ret != 0) {
	switch (errno) {
	case ENOENT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no such parent '%s'"), parent);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to set this "
	"property or value"));
	return (zfs_error(hdl, EZFS_BADVERSION, errbuf));

	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption root's key is not loaded "
	"or provided"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));

	case ERANGE:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property value(s) specified"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	#ifdef _ILP32
	case EOVERFLOW:
	/*
	* This platform can't address a volume this big.
	*/
	if (type == ZFS_TYPE_VOLUME)
	return (zfs_error(hdl, EZFS_VOLTOOBIG,
	errbuf));
	zfs_fallthrough;
	#endif
	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	}

	return (0);
	}

	/*
	* Destroys the given dataset. The caller must make sure that the filesystem
	* isn't mounted, and that there are no active dependents. If the file system
	* does not exist this function does nothing.
	*/
	int
	zfs_destroy(zfs_handle_t *zhp, boolean_t defer)
	{
	int error;

	if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT && defer)
	return (EINVAL);

	if (zhp->zfs_type == ZFS_TYPE_BOOKMARK) {
	nvlist_t *nv = fnvlist_alloc();
	fnvlist_add_boolean(nv, zhp->zfs_name);
	error = lzc_destroy_bookmarks(nv, NULL);
	fnvlist_free(nv);
	if (error != 0) {
	return (zfs_standard_error_fmt(zhp->zfs_hdl, error,
	dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
	zhp->zfs_name));
	}
	return (0);
	}

	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
	nvlist_t *nv = fnvlist_alloc();
	fnvlist_add_boolean(nv, zhp->zfs_name);
	error = lzc_destroy_snaps(nv, defer, NULL);
	fnvlist_free(nv);
	} else {
	error = lzc_destroy(zhp->zfs_name);
	}

	if (error != 0 && error != ENOENT) {
	return (zfs_standard_error_fmt(zhp->zfs_hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
	zhp->zfs_name));
	}

	remove_mountpoint(zhp);

	return (0);
	}

	struct destroydata {
	nvlist_t *nvl;
	const char *snapname;
	};

	static int
	zfs_check_snap_cb(zfs_handle_t zhp, void arg)
	{
	struct destroydata *dd = arg;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int rv = 0;

	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
	dd->snapname) >= sizeof (name))
	return (EINVAL);

	if (lzc_exists(name))
	fnvlist_add_boolean(dd->nvl, name);

	rv = zfs_iter_filesystems(zhp, zfs_check_snap_cb, dd);
	zfs_close(zhp);
	return (rv);
	}

	/*
	* Destroys all snapshots with the given name in zhp & descendants.
	*/
	int
	zfs_destroy_snaps(zfs_handle_t zhp, char snapname, boolean_t defer)
	{
	int ret;
	struct destroydata dd = { 0 };

	dd.snapname = snapname;
	dd.nvl = fnvlist_alloc();
	(void) zfs_check_snap_cb(zfs_handle_dup(zhp), &dd);

	if (nvlist_empty(dd.nvl)) {
	ret = zfs_standard_error_fmt(zhp->zfs_hdl, ENOENT,
	dgettext(TEXT_DOMAIN, "cannot destroy '%s@%s'"),
	zhp->zfs_name, snapname);
	} else {
	ret = zfs_destroy_snaps_nvl(zhp->zfs_hdl, dd.nvl, defer);
	}
	fnvlist_free(dd.nvl);
	return (ret);
	}

	/*
	* Destroys all the snapshots named in the nvlist.
	*/
	int
	zfs_destroy_snaps_nvl(libzfs_handle_t hdl, nvlist_t snaps, boolean_t defer)
	{
	nvlist_t *errlist = NULL;
	nvpair_t *pair;

	int ret = zfs_destroy_snaps_nvl_os(hdl, snaps);
	if (ret != 0)
	return (ret);

	ret = lzc_destroy_snaps(snaps, defer, &errlist);

	if (ret == 0) {
	nvlist_free(errlist);
	return (0);
	}

	if (nvlist_empty(errlist)) {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot destroy snapshots"));

	ret = zfs_standard_error(hdl, ret, errbuf);
	}
	for (pair = nvlist_next_nvpair(errlist, NULL);
	pair != NULL; pair = nvlist_next_nvpair(errlist, pair)) {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot destroy snapshot %s"),
	nvpair_name(pair));

	switch (fnvpair_value_int32(pair)) {
	case EEXIST:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "snapshot is cloned"));
	ret = zfs_error(hdl, EZFS_EXISTS, errbuf);
	break;
	default:
	ret = zfs_standard_error(hdl, errno, errbuf);
	break;
	}
	}

	nvlist_free(errlist);
	return (ret);
	}

	/*
	* Clones the given dataset. The target must be of the same type as the source.
	*/
	int
	zfs_clone(zfs_handle_t zhp, const char target, nvlist_t *props)
	{
	char parent[ZFS_MAX_DATASET_NAME_LEN];
	int ret;
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	uint64_t zoned;

	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot create '%s'"), target);

	/* validate the target/clone name */
	if (!zfs_validate_name(hdl, target, ZFS_TYPE_FILESYSTEM, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	/* validate parents exist */
	if (check_parents(hdl, target, &zoned, B_FALSE, NULL) != 0)
	return (-1);

	(void) parent_name(target, parent, sizeof (parent));

	/* do the clone */

	if (props) {
	zfs_type_t type = ZFS_TYPE_FILESYSTEM;

	if (ZFS_IS_VOLUME(zhp))
	type = ZFS_TYPE_VOLUME;
	if ((props = zfs_valid_proplist(hdl, type, props, zoned,
	zhp, zhp->zpool_hdl, B_TRUE, errbuf)) == NULL)
	return (-1);
	if (zfs_fix_auto_resv(zhp, props) == -1) {
	nvlist_free(props);
	return (-1);
	}
	}

	if (zfs_crypto_clone_check(hdl, zhp, parent, props) != 0) {
	nvlist_free(props);
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
	}

	ret = lzc_clone(target, zhp->zfs_name, props);
	nvlist_free(props);

	if (ret != 0) {
	switch (errno) {

	case ENOENT:
	/*
	* The parent doesn't exist. We should have caught this
	* above, but there may a race condition that has since
	* destroyed the parent.
	*
	* At this point, we don't know whether it's the source
	* that doesn't exist anymore, or whether the target
	* dataset doesn't exist.
	*/
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"no such parent '%s'"), parent);
	return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));

	case EXDEV:
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"source and target pools differ"));
	return (zfs_error(zhp->zfs_hdl, EZFS_CROSSTARGET,
	errbuf));

	default:
	return (zfs_standard_error(zhp->zfs_hdl, errno,
	errbuf));
	}
	}

	return (ret);
	}

	/*
	* Promotes the given clone fs to be the clone parent.
	*/
	int
	zfs_promote(zfs_handle_t *zhp)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char snapname[ZFS_MAX_DATASET_NAME_LEN];
	int ret;
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot promote '%s'"), zhp->zfs_name);

	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"snapshots can not be promoted"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}

	if (zhp->zfs_dmustats.dds_origin[0] == '\0') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not a cloned filesystem"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}

	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	ret = lzc_promote(zhp->zfs_name, snapname, sizeof (snapname));

	if (ret != 0) {
	switch (ret) {
	case EACCES:
	/*
	* Promoting encrypted dataset outside its
	* encryption root.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot promote dataset outside its "
	"encryption root"));
	return (zfs_error(hdl, EZFS_EXISTS, errbuf));

	case EEXIST:
	/* There is a conflicting snapshot name. */
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"conflicting snapshot '%s' from parent '%s'"),
	snapname, zhp->zfs_dmustats.dds_origin);
	return (zfs_error(hdl, EZFS_EXISTS, errbuf));

	default:
	return (zfs_standard_error(hdl, ret, errbuf));
	}
	}
	return (ret);
	}

	typedef struct snapdata {
	nvlist_t *sd_nvl;
	const char *sd_snapname;
	} snapdata_t;

	static int
	zfs_snapshot_cb(zfs_handle_t zhp, void arg)
	{
	snapdata_t *sd = arg;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int rv = 0;

	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) == 0) {
	if (snprintf(name, sizeof (name), "%s@%s", zfs_get_name(zhp),
	sd->sd_snapname) >= sizeof (name))
	return (EINVAL);

	fnvlist_add_boolean(sd->sd_nvl, name);

	rv = zfs_iter_filesystems(zhp, zfs_snapshot_cb, sd);
	}
	zfs_close(zhp);

	return (rv);
	}

	/*
	* Creates snapshots. The keys in the snaps nvlist are the snapshots to be
	* created.
	*/
	int
	zfs_snapshot_nvl(libzfs_handle_t hdl, nvlist_t snaps, nvlist_t *props)
	{
	int ret;
	char errbuf[ERRBUFLEN];
	nvpair_t *elem;
	nvlist_t *errors;
	zpool_handle_t *zpool_hdl;
	char pool[ZFS_MAX_DATASET_NAME_LEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot create snapshots "));

	elem = NULL;
	while ((elem = nvlist_next_nvpair(snaps, elem)) != NULL) {
	const char *snapname = nvpair_name(elem);

	/* validate the target name */
	if (!zfs_validate_name(hdl, snapname, ZFS_TYPE_SNAPSHOT,
	B_TRUE)) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot create snapshot '%s'"), snapname);
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}
	}

	/*
	* get pool handle for prop validation. assumes all snaps are in the
	* same pool, as does lzc_snapshot (below).
	*/
	elem = nvlist_next_nvpair(snaps, NULL);
	+ if (elem == NULL)
	+ return (-1);
	(void) strlcpy(pool, nvpair_name(elem), sizeof (pool));
	pool[strcspn(pool, "/@")] = '\0';
	zpool_hdl = zpool_open(hdl, pool);
	if (zpool_hdl == NULL)
	return (-1);

	if (props != NULL &&
	(props = zfs_valid_proplist(hdl, ZFS_TYPE_SNAPSHOT,
	props, B_FALSE, NULL, zpool_hdl, B_FALSE, errbuf)) == NULL) {
	zpool_close(zpool_hdl);
	return (-1);
	}
	zpool_close(zpool_hdl);

	ret = lzc_snapshot(snaps, props, &errors);

	if (ret != 0) {
	boolean_t printed = B_FALSE;
	for (elem = nvlist_next_nvpair(errors, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(errors, elem)) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot create snapshot '%s'"), nvpair_name(elem));
	(void) zfs_standard_error(hdl,
	fnvpair_value_int32(elem), errbuf);
	printed = B_TRUE;
	}
	if (!printed) {
	switch (ret) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"multiple snapshots of same "
	"fs not allowed"));
	(void) zfs_error(hdl, EZFS_EXISTS, errbuf);

	break;
	default:
	(void) zfs_standard_error(hdl, ret, errbuf);
	}
	}
	}

	nvlist_free(props);
	nvlist_free(errors);
	return (ret);
	}

	int
	zfs_snapshot(libzfs_handle_t hdl, const char path, boolean_t recursive,
	nvlist_t *props)
	{
	int ret;
	snapdata_t sd = { 0 };
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	char *cp;
	zfs_handle_t *zhp;
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot snapshot %s"), path);

	if (!zfs_validate_name(hdl, path, ZFS_TYPE_SNAPSHOT, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	(void) strlcpy(fsname, path, sizeof (fsname));
	cp = strchr(fsname, '@');
	*cp = '\0';
	sd.sd_snapname = cp + 1;

	if ((zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) == NULL) {
	return (-1);
	}

	sd.sd_nvl = fnvlist_alloc();
	if (recursive) {
	(void) zfs_snapshot_cb(zfs_handle_dup(zhp), &sd);
	} else {
	fnvlist_add_boolean(sd.sd_nvl, path);
	}

	ret = zfs_snapshot_nvl(hdl, sd.sd_nvl, props);
	fnvlist_free(sd.sd_nvl);
	zfs_close(zhp);
	return (ret);
	}

	/*
	* Destroy any more recent snapshots. We invoke this callback on any dependents
	* of the snapshot first. If the 'cb_dependent' member is non-zero, then this
	* is a dependent and we should just destroy it without checking the transaction
	* group.
	*/
	typedef struct rollback_data {
	const char cb_target; / the snapshot */
	uint64_t cb_create; /* creation time reference */
	boolean_t cb_error;
	boolean_t cb_force;
	} rollback_data_t;

	static int
	rollback_destroy_dependent(zfs_handle_t zhp, void data)
	{
	rollback_data_t *cbp = data;
	prop_changelist_t *clp;

	/* We must destroy this clone; first unmount it */
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	cbp->cb_force ? MS_FORCE: 0);
	if (clp == NULL \|\| changelist_prefix(clp) != 0) {
	cbp->cb_error = B_TRUE;
	zfs_close(zhp);
	return (0);
	}
	if (zfs_destroy(zhp, B_FALSE) != 0)
	cbp->cb_error = B_TRUE;
	else
	changelist_remove(clp, zhp->zfs_name);
	(void) changelist_postfix(clp);
	changelist_free(clp);

	zfs_close(zhp);
	return (0);
	}

	static int
	rollback_destroy(zfs_handle_t zhp, void data)
	{
	rollback_data_t *cbp = data;

	if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) {
	cbp->cb_error \|= zfs_iter_dependents(zhp, B_FALSE,
	rollback_destroy_dependent, cbp);

	cbp->cb_error \|= zfs_destroy(zhp, B_FALSE);
	}

	zfs_close(zhp);
	return (0);
	}

	/*
	* Given a dataset, rollback to a specific snapshot, discarding any
	* data changes since then and making it the active dataset.
	*
	* Any snapshots and bookmarks more recent than the target are
	* destroyed, along with their dependents (i.e. clones).
	*/
	int
	zfs_rollback(zfs_handle_t zhp, zfs_handle_t snap, boolean_t force)
	{
	rollback_data_t cb = { 0 };
	int err;
	boolean_t restore_resv = 0;
	uint64_t old_volsize = 0, new_volsize;
	zfs_prop_t resv_prop = { 0 };
	uint64_t min_txg = 0;

	assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM \|\|
	zhp->zfs_type == ZFS_TYPE_VOLUME);

	/*
	* Destroy all recent snapshots and their dependents.
	*/
	cb.cb_force = force;
	cb.cb_target = snap->zfs_name;
	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);

	if (cb.cb_create > 0)
	min_txg = cb.cb_create;

	(void) zfs_iter_snapshots(zhp, B_FALSE, rollback_destroy, &cb,
	min_txg, 0);

	(void) zfs_iter_bookmarks(zhp, rollback_destroy, &cb);

	if (cb.cb_error)
	return (-1);

	/*
	* Now that we have verified that the snapshot is the latest,
	* rollback to the given snapshot.
	*/

	if (zhp->zfs_type == ZFS_TYPE_VOLUME) {
	if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
	return (-1);
	old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
	restore_resv =
	(old_volsize == zfs_prop_get_int(zhp, resv_prop));
	}

	/*
	* Pass both the filesystem and the wanted snapshot names,
	* we would get an error back if the snapshot is destroyed or
	* a new snapshot is created before this request is processed.
	*/
	err = lzc_rollback_to(zhp->zfs_name, snap->zfs_name);
	if (err != 0) {
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot rollback '%s'"),
	zhp->zfs_name);
	switch (err) {
	case EEXIST:
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"there is a snapshot or bookmark more recent "
	"than '%s'"), snap->zfs_name);
	(void) zfs_error(zhp->zfs_hdl, EZFS_EXISTS, errbuf);
	break;
	case ESRCH:
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"'%s' is not found among snapshots of '%s'"),
	snap->zfs_name, zhp->zfs_name);
	(void) zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf);
	break;
	case EINVAL:
	(void) zfs_error(zhp->zfs_hdl, EZFS_BADTYPE, errbuf);
	break;
	default:
	(void) zfs_standard_error(zhp->zfs_hdl, err, errbuf);
	}
	return (err);
	}

	/*
	* For volumes, if the pre-rollback volsize matched the pre-
	* rollback reservation and the volsize has changed then set
	* the reservation property to the post-rollback volsize.
	* Make a new handle since the rollback closed the dataset.
	*/
	if ((zhp->zfs_type == ZFS_TYPE_VOLUME) &&
	(zhp = make_dataset_handle(zhp->zfs_hdl, zhp->zfs_name))) {
	if (restore_resv) {
	new_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
	if (old_volsize != new_volsize)
	err = zfs_prop_set_int(zhp, resv_prop,
	new_volsize);
	}
	zfs_close(zhp);
	}
	return (err);
	}

	/*
	* Renames the given dataset.
	*/
	int
	zfs_rename(zfs_handle_t zhp, const char target, renameflags_t flags)
	{
	int ret = 0;
	zfs_cmd_t zc = {"\0"};
	char *delim;
	prop_changelist_t *cl = NULL;
	char parent[ZFS_MAX_DATASET_NAME_LEN];
	char property[ZFS_MAXPROPLEN];
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char errbuf[ERRBUFLEN];

	/* if we have the same exact name, just return success */
	if (strcmp(zhp->zfs_name, target) == 0)
	return (0);

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot rename to '%s'"), target);

	/* make sure source name is valid */
	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	/*
	* Make sure the target name is valid
	*/
	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
	if ((strchr(target, '@') == NULL) \|\|
	*target == '@') {
	/*
	* Snapshot target name is abbreviated,
	* reconstruct full dataset name
	*/
	(void) strlcpy(parent, zhp->zfs_name,
	sizeof (parent));
	delim = strchr(parent, '@');
	if (strchr(target, '@') == NULL)
	*(++delim) = '\0';
	else
	*delim = '\0';
	(void) strlcat(parent, target, sizeof (parent));
	target = parent;
	} else {
	/*
	* Make sure we're renaming within the same dataset.
	*/
	delim = strchr(target, '@');
	if (strncmp(zhp->zfs_name, target, delim - target)
	!= 0 \|\| zhp->zfs_name[delim - target] != '@') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"snapshots must be part of same "
	"dataset"));
	return (zfs_error(hdl, EZFS_CROSSTARGET,
	errbuf));
	}
	}

	if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	} else {
	if (flags.recursive) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"recursive rename must be a snapshot"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}

	if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	/* validate parents */
	if (check_parents(hdl, target, NULL, B_FALSE, NULL) != 0)
	return (-1);

	/* make sure we're in the same pool */
	verify((delim = strchr(target, '/')) != NULL);
	if (strncmp(zhp->zfs_name, target, delim - target) != 0 \|\|
	zhp->zfs_name[delim - target] != '/') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"datasets must be within same pool"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
	}

	/* new name cannot be a child of the current dataset name */
	if (is_descendant(zhp->zfs_name, target)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"New dataset name cannot be a descendant of "
	"current dataset name"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}
	}

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot rename '%s'"), zhp->zfs_name);

	if (getzoneid() == GLOBAL_ZONEID &&
	zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset is used in a non-global zone"));
	return (zfs_error(hdl, EZFS_ZONED, errbuf));
	}

	/*
	* Avoid unmounting file systems with mountpoint property set to
	* 'legacy' or 'none' even if -u option is not given.
	*/
	if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
	!flags.recursive && !flags.nounmount &&
	zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, property,
	sizeof (property), NULL, NULL, 0, B_FALSE) == 0 &&
	(strcmp(property, "legacy") == 0 \|\|
	strcmp(property, "none") == 0)) {
	flags.nounmount = B_TRUE;
	}
	if (flags.recursive) {
	char *parentname = zfs_strdup(zhp->zfs_hdl, zhp->zfs_name);
	delim = strchr(parentname, '@');
	*delim = '\0';
	zfs_handle_t *zhrp = zfs_open(zhp->zfs_hdl, parentname,
	ZFS_TYPE_DATASET);
	free(parentname);
	if (zhrp == NULL) {
	ret = -1;
	goto error;
	}
	zfs_close(zhrp);
	} else if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) {
	if ((cl = changelist_gather(zhp, ZFS_PROP_NAME,
	flags.nounmount ? CL_GATHER_DONT_UNMOUNT :
	CL_GATHER_ITER_MOUNTED,
	flags.forceunmount ? MS_FORCE : 0)) == NULL)
	return (-1);

	if (changelist_haszonedchild(cl)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"child dataset with inherited mountpoint is used "
	"in a non-global zone"));
	(void) zfs_error(hdl, EZFS_ZONED, errbuf);
	ret = -1;
	goto error;
	}

	if ((ret = changelist_prefix(cl)) != 0)
	goto error;
	}

	if (ZFS_IS_VOLUME(zhp))
	zc.zc_objset_type = DMU_OST_ZVOL;
	else
	zc.zc_objset_type = DMU_OST_ZFS;

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	(void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value));

	zc.zc_cookie = !!flags.recursive;
	zc.zc_cookie \|= (!!flags.nounmount) << 1;

	if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_RENAME, &zc)) != 0) {
	/*
	* if it was recursive, the one that actually failed will
	* be in zc.zc_name
	*/
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot rename '%s'"), zc.zc_name);

	if (flags.recursive && errno == EEXIST) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"a child dataset already has a snapshot "
	"with the new name"));
	(void) zfs_error(hdl, EZFS_EXISTS, errbuf);
	} else if (errno == EACCES) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot move encrypted child outside of "
	"its encryption root"));
	(void) zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	} else {
	(void) zfs_standard_error(zhp->zfs_hdl, errno, errbuf);
	}

	/*
	* On failure, we still want to remount any filesystems that
	* were previously mounted, so we don't alter the system state.
	*/
	if (cl != NULL)
	(void) changelist_postfix(cl);
	} else {
	if (cl != NULL) {
	changelist_rename(cl, zfs_get_name(zhp), target);
	ret = changelist_postfix(cl);
	}
	}

	error:
	if (cl != NULL) {
	changelist_free(cl);
	}
	return (ret);
	}

	nvlist_t *
	zfs_get_all_props(zfs_handle_t *zhp)
	{
	return (zhp->zfs_props);
	}

	nvlist_t *
	zfs_get_recvd_props(zfs_handle_t *zhp)
	{
	if (zhp->zfs_recvd_props == NULL)
	if (get_recvd_props_ioctl(zhp) != 0)
	return (NULL);
	return (zhp->zfs_recvd_props);
	}

	nvlist_t *
	zfs_get_user_props(zfs_handle_t *zhp)
	{
	return (zhp->zfs_user_props);
	}

	/*
	* This function is used by 'zfs list' to determine the exact set of columns to
	* display, and their maximum widths. This does two main things:
	*
	* - If this is a list of all properties, then expand the list to include
	* all native properties, and set a flag so that for each dataset we look
	* for new unique user properties and add them to the list.
	*
	* - For non fixed-width properties, keep track of the maximum width seen
	* so that we can size the column appropriately. If the user has
	* requested received property values, we also need to compute the width
	* of the RECEIVED column.
	*/
	int
	zfs_expand_proplist(zfs_handle_t zhp, zprop_list_t *plp, boolean_t received,
	boolean_t literal)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	zprop_list_t *entry;
	zprop_list_t last, start;
	nvlist_t userprops, propval;
	nvpair_t *elem;
	char *strval;
	char buf[ZFS_MAXPROPLEN];

	if (zprop_expand_list(hdl, plp, ZFS_TYPE_DATASET) != 0)
	return (-1);

	userprops = zfs_get_user_props(zhp);

	entry = *plp;
	if (entry->pl_all && nvlist_next_nvpair(userprops, NULL) != NULL) {
	/*
	* Go through and add any user properties as necessary. We
	* start by incrementing our list pointer to the first
	* non-native property.
	*/
	start = plp;
	while (*start != NULL) {
	if ((*start)->pl_prop == ZPROP_USERPROP)
	break;
	start = &(*start)->pl_next;
	}

	elem = NULL;
	while ((elem = nvlist_next_nvpair(userprops, elem)) != NULL) {
	/*
	* See if we've already found this property in our list.
	*/
	for (last = start; *last != NULL;
	last = &(*last)->pl_next) {
	if (strcmp((*last)->pl_user_prop,
	nvpair_name(elem)) == 0)
	break;
	}

	if (*last == NULL) {
	entry = zfs_alloc(hdl, sizeof (zprop_list_t));
	entry->pl_user_prop =
	zfs_strdup(hdl, nvpair_name(elem));
	entry->pl_prop = ZPROP_USERPROP;
	entry->pl_width = strlen(nvpair_name(elem));
	entry->pl_all = B_TRUE;
	*last = entry;
	}
	}
	}

	/*
	* Now go through and check the width of any non-fixed columns
	*/
	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
	if (entry->pl_fixed && !literal)
	continue;

	if (entry->pl_prop != ZPROP_USERPROP) {
	if (zfs_prop_get(zhp, entry->pl_prop,
	buf, sizeof (buf), NULL, NULL, 0, literal) == 0) {
	if (strlen(buf) > entry->pl_width)
	entry->pl_width = strlen(buf);
	}
	if (received && zfs_prop_get_recvd(zhp,
	zfs_prop_to_name(entry->pl_prop),
	buf, sizeof (buf), literal) == 0)
	if (strlen(buf) > entry->pl_recvd_width)
	entry->pl_recvd_width = strlen(buf);
	} else {
	if (nvlist_lookup_nvlist(userprops, entry->pl_user_prop,
	&propval) == 0) {
	strval = fnvlist_lookup_string(propval,
	ZPROP_VALUE);
	if (strlen(strval) > entry->pl_width)
	entry->pl_width = strlen(strval);
	}
	if (received && zfs_prop_get_recvd(zhp,
	entry->pl_user_prop,
	buf, sizeof (buf), literal) == 0)
	if (strlen(buf) > entry->pl_recvd_width)
	entry->pl_recvd_width = strlen(buf);
	}
	}

	return (0);
	}

	void
	zfs_prune_proplist(zfs_handle_t zhp, uint8_t props)
	{
	nvpair_t *curr;
	nvpair_t *next;

	/*
	* Keep a reference to the props-table against which we prune the
	* properties.
	*/
	zhp->zfs_props_table = props;

	curr = nvlist_next_nvpair(zhp->zfs_props, NULL);

	while (curr) {
	zfs_prop_t zfs_prop = zfs_name_to_prop(nvpair_name(curr));
	next = nvlist_next_nvpair(zhp->zfs_props, curr);

	/*
	* User properties will result in ZPROP_USERPROP (an alias
	* for ZPROP_INVAL), and since we
	* only know how to prune standard ZFS properties, we always
	* leave these in the list. This can also happen if we
	* encounter an unknown DSL property (when running older
	* software, for example).
	*/
	if (zfs_prop != ZPROP_USERPROP && props[zfs_prop] == B_FALSE)
	(void) nvlist_remove(zhp->zfs_props,
	nvpair_name(curr), nvpair_type(curr));
	curr = next;
	}
	}

	static int
	zfs_smb_acl_mgmt(libzfs_handle_t hdl, char dataset, char *path,
	zfs_smb_acl_op_t cmd, char resource1, char resource2)
	{
	zfs_cmd_t zc = {"\0"};
	nvlist_t *nvlist = NULL;
	int error;

	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
	(void) strlcpy(zc.zc_value, path, sizeof (zc.zc_value));
	zc.zc_cookie = (uint64_t)cmd;

	if (cmd == ZFS_SMB_ACL_RENAME) {
	if (nvlist_alloc(&nvlist, NV_UNIQUE_NAME, 0) != 0) {
	(void) no_memory(hdl);
	return (0);
	}
	}

	switch (cmd) {
	case ZFS_SMB_ACL_ADD:
	case ZFS_SMB_ACL_REMOVE:
	(void) strlcpy(zc.zc_string, resource1, sizeof (zc.zc_string));
	break;
	case ZFS_SMB_ACL_RENAME:
	if (nvlist_add_string(nvlist, ZFS_SMB_ACL_SRC,
	resource1) != 0) {
	(void) no_memory(hdl);
	return (-1);
	}
	if (nvlist_add_string(nvlist, ZFS_SMB_ACL_TARGET,
	resource2) != 0) {
	(void) no_memory(hdl);
	return (-1);
	}
	zcmd_write_src_nvlist(hdl, &zc, nvlist);
	break;
	case ZFS_SMB_ACL_PURGE:
	break;
	default:
	return (-1);
	}
	error = ioctl(hdl->libzfs_fd, ZFS_IOC_SMB_ACL, &zc);
	nvlist_free(nvlist);
	return (error);
	}

	int
	zfs_smb_acl_add(libzfs_handle_t hdl, char dataset,
	char path, char resource)
	{
	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_ADD,
	resource, NULL));
	}

	int
	zfs_smb_acl_remove(libzfs_handle_t hdl, char dataset,
	char path, char resource)
	{
	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_REMOVE,
	resource, NULL));
	}

	int
	zfs_smb_acl_purge(libzfs_handle_t hdl, char dataset, char *path)
	{
	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_PURGE,
	NULL, NULL));
	}

	int
	zfs_smb_acl_rename(libzfs_handle_t hdl, char dataset, char *path,
	char oldname, char newname)
	{
	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_RENAME,
	oldname, newname));
	}

	int
	zfs_userspace(zfs_handle_t *zhp, zfs_userquota_prop_t type,
	zfs_userspace_cb_t func, void *arg)
	{
	zfs_cmd_t zc = {"\0"};
	zfs_useracct_t buf[100];
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	int ret;

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	zc.zc_objset_type = type;
	zc.zc_nvlist_dst = (uintptr_t)buf;

	for (;;) {
	zfs_useracct_t *zua = buf;

	zc.zc_nvlist_dst_size = sizeof (buf);
	if (zfs_ioctl(hdl, ZFS_IOC_USERSPACE_MANY, &zc) != 0) {
	if ((errno == ENOTSUP &&
	(type == ZFS_PROP_USEROBJUSED \|\|
	type == ZFS_PROP_GROUPOBJUSED \|\|
	type == ZFS_PROP_USEROBJQUOTA \|\|
	type == ZFS_PROP_GROUPOBJQUOTA \|\|
	type == ZFS_PROP_PROJECTOBJUSED \|\|
	type == ZFS_PROP_PROJECTOBJQUOTA \|\|
	type == ZFS_PROP_PROJECTUSED \|\|
	type == ZFS_PROP_PROJECTQUOTA)))
	break;

	return (zfs_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN,
	"cannot get used/quota for %s"), zc.zc_name));
	}
	if (zc.zc_nvlist_dst_size == 0)
	break;

	while (zc.zc_nvlist_dst_size > 0) {
	if ((ret = func(arg, zua->zu_domain, zua->zu_rid,
	zua->zu_space)) != 0)
	return (ret);
	zua++;
	zc.zc_nvlist_dst_size -= sizeof (zfs_useracct_t);
	}
	}

	return (0);
	}

	struct holdarg {
	nvlist_t *nvl;
	const char *snapname;
	const char *tag;
	boolean_t recursive;
	int error;
	};

	static int
	zfs_hold_one(zfs_handle_t zhp, void arg)
	{
	struct holdarg *ha = arg;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int rv = 0;

	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
	ha->snapname) >= sizeof (name))
	return (EINVAL);

	if (lzc_exists(name))
	fnvlist_add_string(ha->nvl, name, ha->tag);

	if (ha->recursive)
	rv = zfs_iter_filesystems(zhp, zfs_hold_one, ha);
	zfs_close(zhp);
	return (rv);
	}

	int
	zfs_hold(zfs_handle_t zhp, const char snapname, const char *tag,
	boolean_t recursive, int cleanup_fd)
	{
	int ret;
	struct holdarg ha;

	ha.nvl = fnvlist_alloc();
	ha.snapname = snapname;
	ha.tag = tag;
	ha.recursive = recursive;
	(void) zfs_hold_one(zfs_handle_dup(zhp), &ha);

	if (nvlist_empty(ha.nvl)) {
	char errbuf[ERRBUFLEN];

	fnvlist_free(ha.nvl);
	ret = ENOENT;
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot hold snapshot '%s@%s'"),
	zhp->zfs_name, snapname);
	(void) zfs_standard_error(zhp->zfs_hdl, ret, errbuf);
	return (ret);
	}

	ret = zfs_hold_nvl(zhp, cleanup_fd, ha.nvl);
	fnvlist_free(ha.nvl);

	return (ret);
	}

	int
	zfs_hold_nvl(zfs_handle_t zhp, int cleanup_fd, nvlist_t holds)
	{
	int ret;
	nvlist_t *errors;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char errbuf[ERRBUFLEN];
	nvpair_t *elem;

	errors = NULL;
	ret = lzc_hold(holds, cleanup_fd, &errors);

	if (ret == 0) {
	/* There may be errors even in the success case. */
	fnvlist_free(errors);
	return (0);
	}

	if (nvlist_empty(errors)) {
	/* no hold-specific errors */
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot hold"));
	switch (ret) {
	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded"));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case EINVAL:
	(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	default:
	(void) zfs_standard_error(hdl, ret, errbuf);
	}
	}

	for (elem = nvlist_next_nvpair(errors, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(errors, elem)) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot hold snapshot '%s'"), nvpair_name(elem));
	switch (fnvpair_value_int32(elem)) {
	case E2BIG:
	/*
	* Temporary tags wind up having the ds object id
	* prepended. So even if we passed the length check
	* above, it's still possible for the tag to wind
	* up being slightly too long.
	*/
	(void) zfs_error(hdl, EZFS_TAGTOOLONG, errbuf);
	break;
	case EINVAL:
	(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	case EEXIST:
	(void) zfs_error(hdl, EZFS_REFTAG_HOLD, errbuf);
	break;
	default:
	(void) zfs_standard_error(hdl,
	fnvpair_value_int32(elem), errbuf);
	}
	}

	fnvlist_free(errors);
	return (ret);
	}

	static int
	zfs_release_one(zfs_handle_t zhp, void arg)
	{
	struct holdarg *ha = arg;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int rv = 0;
	nvlist_t *existing_holds;

	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
	ha->snapname) >= sizeof (name)) {
	ha->error = EINVAL;
	rv = EINVAL;
	}

	if (lzc_get_holds(name, &existing_holds) != 0) {
	ha->error = ENOENT;
	} else if (!nvlist_exists(existing_holds, ha->tag)) {
	ha->error = ESRCH;
	} else {
	nvlist_t *torelease = fnvlist_alloc();
	fnvlist_add_boolean(torelease, ha->tag);
	fnvlist_add_nvlist(ha->nvl, name, torelease);
	fnvlist_free(torelease);
	}

	if (ha->recursive)
	rv = zfs_iter_filesystems(zhp, zfs_release_one, ha);
	zfs_close(zhp);
	return (rv);
	}

	int
	zfs_release(zfs_handle_t zhp, const char snapname, const char *tag,
	boolean_t recursive)
	{
	int ret;
	struct holdarg ha;
	nvlist_t *errors = NULL;
	nvpair_t *elem;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char errbuf[ERRBUFLEN];

	ha.nvl = fnvlist_alloc();
	ha.snapname = snapname;
	ha.tag = tag;
	ha.recursive = recursive;
	ha.error = 0;
	(void) zfs_release_one(zfs_handle_dup(zhp), &ha);

	if (nvlist_empty(ha.nvl)) {
	fnvlist_free(ha.nvl);
	ret = ha.error;
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot release hold from snapshot '%s@%s'"),
	zhp->zfs_name, snapname);
	if (ret == ESRCH) {
	(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
	} else {
	(void) zfs_standard_error(hdl, ret, errbuf);
	}
	return (ret);
	}

	ret = lzc_release(ha.nvl, &errors);
	fnvlist_free(ha.nvl);

	if (ret == 0) {
	/* There may be errors even in the success case. */
	fnvlist_free(errors);
	return (0);
	}

	if (nvlist_empty(errors)) {
	/* no hold-specific errors */
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot release"));
	switch (errno) {
	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded"));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	default:
	(void) zfs_standard_error(hdl, errno, errbuf);
	}
	}

	for (elem = nvlist_next_nvpair(errors, NULL);
	elem != NULL;
	elem = nvlist_next_nvpair(errors, elem)) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot release hold from snapshot '%s'"),
	nvpair_name(elem));
	switch (fnvpair_value_int32(elem)) {
	case ESRCH:
	(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
	break;
	case EINVAL:
	(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	default:
	(void) zfs_standard_error(hdl,
	fnvpair_value_int32(elem), errbuf);
	}
	}

	fnvlist_free(errors);
	return (ret);
	}

	int
	zfs_get_fsacl(zfs_handle_t zhp, nvlist_t *nvl)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	int nvsz = 2048;
	void *nvbuf;
	int err = 0;
	char errbuf[ERRBUFLEN];

	assert(zhp->zfs_type == ZFS_TYPE_VOLUME \|\|
	zhp->zfs_type == ZFS_TYPE_FILESYSTEM);

	tryagain:

	nvbuf = malloc(nvsz);
	if (nvbuf == NULL) {
	err = (zfs_error(hdl, EZFS_NOMEM, strerror(errno)));
	goto out;
	}

	zc.zc_nvlist_dst_size = nvsz;
	zc.zc_nvlist_dst = (uintptr_t)nvbuf;

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	if (zfs_ioctl(hdl, ZFS_IOC_GET_FSACL, &zc) != 0) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot get permissions on '%s'"),
	zc.zc_name);
	switch (errno) {
	case ENOMEM:
	free(nvbuf);
	nvsz = zc.zc_nvlist_dst_size;
	goto tryagain;

	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded"));
	err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case EINVAL:
	err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	case ENOENT:
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	break;
	default:
	err = zfs_standard_error(hdl, errno, errbuf);
	break;
	}
	} else {
	/* success */
	int rc = nvlist_unpack(nvbuf, zc.zc_nvlist_dst_size, nvl, 0);
	if (rc) {
	err = zfs_standard_error_fmt(hdl, rc, dgettext(
	TEXT_DOMAIN, "cannot get permissions on '%s'"),
	zc.zc_name);
	}
	}

	free(nvbuf);
	out:
	return (err);
	}

	int
	zfs_set_fsacl(zfs_handle_t zhp, boolean_t un, nvlist_t nvl)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char *nvbuf;
	char errbuf[ERRBUFLEN];
	size_t nvsz;
	int err;

	assert(zhp->zfs_type == ZFS_TYPE_VOLUME \|\|
	zhp->zfs_type == ZFS_TYPE_FILESYSTEM);

	err = nvlist_size(nvl, &nvsz, NV_ENCODE_NATIVE);
	assert(err == 0);

	nvbuf = malloc(nvsz);

	err = nvlist_pack(nvl, &nvbuf, &nvsz, NV_ENCODE_NATIVE, 0);
	assert(err == 0);

	zc.zc_nvlist_src_size = nvsz;
	zc.zc_nvlist_src = (uintptr_t)nvbuf;
	zc.zc_perm_action = un;

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));

	if (zfs_ioctl(hdl, ZFS_IOC_SET_FSACL, &zc) != 0) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot set permissions on '%s'"),
	zc.zc_name);
	switch (errno) {
	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded"));
	err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case EINVAL:
	err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	case ENOENT:
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	break;
	default:
	err = zfs_standard_error(hdl, errno, errbuf);
	break;
	}
	}

	free(nvbuf);

	return (err);
	}

	int
	zfs_get_holds(zfs_handle_t zhp, nvlist_t *nvl)
	{
	int err;
	char errbuf[ERRBUFLEN];

	err = lzc_get_holds(zhp->zfs_name, nvl);

	if (err != 0) {
	libzfs_handle_t *hdl = zhp->zfs_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot get holds for '%s'"),
	zhp->zfs_name);
	switch (err) {
	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded"));
	err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case EINVAL:
	err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
	break;
	case ENOENT:
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	break;
	default:
	err = zfs_standard_error(hdl, errno, errbuf);
	break;
	}
	}

	return (err);
	}

	/*
	* The theory of raidz space accounting
	*
	* The "referenced" property of RAIDZ vdevs is scaled such that a 128KB block
	* will "reference" 128KB, even though it allocates more than that, to store the
	* parity information (and perhaps skip sectors). This concept of the
	* "referenced" (and other DMU space accounting) being lower than the allocated
	* space by a constant factor is called "raidz deflation."
	*
	* As mentioned above, the constant factor for raidz deflation assumes a 128KB
	* block size. However, zvols typically have a much smaller block size (default
	* 8KB). These smaller blocks may require proportionally much more parity
	* information (and perhaps skip sectors). In this case, the change to the
	* "referenced" property may be much more than the logical block size.
	*
	* Suppose a raidz vdev has 5 disks with ashift=12. A 128k block may be written
	* as follows.
	*
	* +-------+-------+-------+-------+-------+
	* \| disk1 \| disk2 \| disk3 \| disk4 \| disk5 \|
	* +-------+-------+-------+-------+-------+
	* \| P0 \| D0 \| D8 \| D16 \| D24 \|
	* \| P1 \| D1 \| D9 \| D17 \| D25 \|
	* \| P2 \| D2 \| D10 \| D18 \| D26 \|
	* \| P3 \| D3 \| D11 \| D19 \| D27 \|
	* \| P4 \| D4 \| D12 \| D20 \| D28 \|
	* \| P5 \| D5 \| D13 \| D21 \| D29 \|
	* \| P6 \| D6 \| D14 \| D22 \| D30 \|
	* \| P7 \| D7 \| D15 \| D23 \| D31 \|
	* +-------+-------+-------+-------+-------+
	*
	* Above, notice that 160k was allocated: 8 x 4k parity sectors + 32 x 4k data
	* sectors. The dataset's referenced will increase by 128k and the pool's
	* allocated and free properties will be adjusted by 160k.
	*
	* A 4k block written to the same raidz vdev will require two 4k sectors. The
	* blank cells represent unallocated space.
	*
	* +-------+-------+-------+-------+-------+
	* \| disk1 \| disk2 \| disk3 \| disk4 \| disk5 \|
	* +-------+-------+-------+-------+-------+
	* \| P0 \| D0 \| \| \| \|
	* +-------+-------+-------+-------+-------+
	*
	* Above, notice that the 4k block required one sector for parity and another
	* for data. vdev_raidz_asize() will return 8k and as such the pool's allocated
	* and free properties will be adjusted by 8k. The dataset will not be charged
	* 8k. Rather, it will be charged a value that is scaled according to the
	* overhead of the 128k block on the same vdev. This 8k allocation will be
	* charged 8k * 128k / 160k. 128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as
	* calculated in the 128k block example above.
	*
	* Every raidz allocation is sized to be a multiple of nparity+1 sectors. That
	* is, every raidz1 allocation will be a multiple of 2 sectors, raidz2
	* allocations are a multiple of 3 sectors, and raidz3 allocations are a
	* multiple of of 4 sectors. When a block does not fill the required number of
	* sectors, skip blocks (sectors) are used.
	*
	* An 8k block being written to a raidz vdev may be written as follows:
	*
	* +-------+-------+-------+-------+-------+
	* \| disk1 \| disk2 \| disk3 \| disk4 \| disk5 \|
	* +-------+-------+-------+-------+-------+
	* \| P0 \| D0 \| D1 \| S0 \| \|
	* +-------+-------+-------+-------+-------+
	*
	* In order to maintain the nparity+1 allocation size, a skip block (S0) was
	* added. For this 8k block, the pool's allocated and free properties are
	* adjusted by 16k and the dataset's referenced is increased by 16k * 128k /
	* 160k. Again, 128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as calculated in
	* the 128k block example above.
	*
	* The situation is slightly different for dRAID since the minimum allocation
	* size is the full group width. The same 8K block above would be written as
	* follows in a dRAID group:
	*
	* +-------+-------+-------+-------+-------+
	* \| disk1 \| disk2 \| disk3 \| disk4 \| disk5 \|
	* +-------+-------+-------+-------+-------+
	* \| P0 \| D0 \| D1 \| S0 \| S1 \|
	* +-------+-------+-------+-------+-------+
	*
	* Compression may lead to a variety of block sizes being written for the same
	* volume or file. There is no clear way to reserve just the amount of space
	* that will be required, so the worst case (no compression) is assumed.
	* Note that metadata blocks will typically be compressed, so the reservation
	* size returned by zvol_volsize_to_reservation() will generally be slightly
	* larger than the maximum that the volume can reference.
	*/

	/*
	* Derived from function of same name in module/zfs/vdev_raidz.c. Returns the
	* amount of space (in bytes) that will be allocated for the specified block
	* size. Note that the "referenced" space accounted will be less than this, but
	* not necessarily equal to "blksize", due to RAIDZ deflation.
	*/
	static uint64_t
	vdev_raidz_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
	uint64_t blksize)
	{
	uint64_t asize, ndata;

	ASSERT3U(ndisks, >, nparity);
	ndata = ndisks - nparity;
	asize = ((blksize - 1) >> ashift) + 1;
	asize += nparity * ((asize + ndata - 1) / ndata);
	asize = roundup(asize, nparity + 1) << ashift;

	return (asize);
	}

	/*
	* Derived from function of same name in module/zfs/vdev_draid.c. Returns the
	* amount of space (in bytes) that will be allocated for the specified block
	* size.
	*/
	static uint64_t
	vdev_draid_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
	uint64_t blksize)
	{
	ASSERT3U(ndisks, >, nparity);
	uint64_t ndata = ndisks - nparity;
	uint64_t rows = ((blksize - 1) / (ndata << ashift)) + 1;
	uint64_t asize = (rows * ndisks) << ashift;

	return (asize);
	}

	/*
	* Determine how much space will be allocated if it lands on the most space-
	* inefficient top-level vdev. Returns the size in bytes required to store one
	* copy of the volume data. See theory comment above.
	*/
	static uint64_t
	volsize_from_vdevs(zpool_handle_t *zhp, uint64_t nblocks, uint64_t blksize)
	{
	nvlist_t config, tree, **vdevs;
	uint_t nvdevs;
	uint64_t ret = 0;

	config = zpool_get_config(zhp, NULL);
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 \|\|
	nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
	&vdevs, &nvdevs) != 0) {
	return (nblocks * blksize);
	}

	for (int v = 0; v < nvdevs; v++) {
	char *type;
	uint64_t nparity, ashift, asize, tsize;
	uint64_t volsize;

	if (nvlist_lookup_string(vdevs[v], ZPOOL_CONFIG_TYPE,
	&type) != 0)
	continue;

	if (strcmp(type, VDEV_TYPE_RAIDZ) != 0 &&
	strcmp(type, VDEV_TYPE_DRAID) != 0)
	continue;

	if (nvlist_lookup_uint64(vdevs[v],
	ZPOOL_CONFIG_NPARITY, &nparity) != 0)
	continue;

	if (nvlist_lookup_uint64(vdevs[v],
	ZPOOL_CONFIG_ASHIFT, &ashift) != 0)
	continue;

	if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
	nvlist_t **disks;
	uint_t ndisks;

	if (nvlist_lookup_nvlist_array(vdevs[v],
	ZPOOL_CONFIG_CHILDREN, &disks, &ndisks) != 0)
	continue;

	/* allocation size for the "typical" 128k block */
	tsize = vdev_raidz_asize(ndisks, nparity, ashift,
	SPA_OLD_MAXBLOCKSIZE);

	/* allocation size for the blksize block */
	asize = vdev_raidz_asize(ndisks, nparity, ashift,
	blksize);
	} else {
	uint64_t ndata;

	if (nvlist_lookup_uint64(vdevs[v],
	ZPOOL_CONFIG_DRAID_NDATA, &ndata) != 0)
	continue;

	/* allocation size for the "typical" 128k block */
	tsize = vdev_draid_asize(ndata + nparity, nparity,
	ashift, SPA_OLD_MAXBLOCKSIZE);

	/* allocation size for the blksize block */
	asize = vdev_draid_asize(ndata + nparity, nparity,
	ashift, blksize);
	}

	/*
	* Scale this size down as a ratio of 128k / tsize.
	* See theory statement above.
	*/
	volsize = nblocks * asize * SPA_OLD_MAXBLOCKSIZE / tsize;
	if (volsize > ret) {
	ret = volsize;
	}
	}

	if (ret == 0) {
	ret = nblocks * blksize;
	}

	return (ret);
	}

	/*
	* Convert the zvol's volume size to an appropriate reservation. See theory
	* comment above.
	*
	* Note: If this routine is updated, it is necessary to update the ZFS test
	* suite's shell version in reservation.shlib.
	*/
	uint64_t
	zvol_volsize_to_reservation(zpool_handle_t *zph, uint64_t volsize,
	nvlist_t *props)
	{
	uint64_t numdb;
	uint64_t nblocks, volblocksize;
	int ncopies;
	char *strval;

	if (nvlist_lookup_string(props,
	zfs_prop_to_name(ZFS_PROP_COPIES), &strval) == 0)
	ncopies = atoi(strval);
	else
	ncopies = 1;
	if (nvlist_lookup_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	&volblocksize) != 0)
	volblocksize = ZVOL_DEFAULT_BLOCKSIZE;

	nblocks = volsize / volblocksize;
	/*
	* Metadata defaults to using 128k blocks, not volblocksize blocks. For
	* this reason, only the data blocks are scaled based on vdev config.
	*/
	volsize = volsize_from_vdevs(zph, nblocks, volblocksize);

	/* start with metadnode L0-L6 */
	numdb = 7;
	/* calculate number of indirects */
	while (nblocks > 1) {
	nblocks += DNODES_PER_LEVEL - 1;
	nblocks /= DNODES_PER_LEVEL;
	numdb += nblocks;
	}
	numdb *= MIN(SPA_DVAS_PER_BP, ncopies + 1);
	volsize *= ncopies;
	/*
	* this is exactly DN_MAX_INDBLKSHIFT when metadata isn't
	* compressed, but in practice they compress down to about
	* 1100 bytes
	*/
	numdb *= 1ULL << DN_MAX_INDBLKSHIFT;
	volsize += numdb;
	return (volsize);
	}

	/*
	* Wait for the given activity and return the status of the wait (whether or not
	* any waiting was done) in the 'waited' parameter. Non-existent fses are
	* reported via the 'missing' parameter, rather than by printing an error
	* message. This is convenient when this function is called in a loop over a
	* long period of time (as it is, for example, by zfs's wait cmd). In that
	* scenario, a fs being exported or destroyed should be considered a normal
	* event, so we don't want to print an error when we find that the fs doesn't
	* exist.
	*/
	int
	zfs_wait_status(zfs_handle_t *zhp, zfs_wait_activity_t activity,
	boolean_t missing, boolean_t waited)
	{
	int error = lzc_wait_fs(zhp->zfs_name, activity, waited);
	*missing = (error == ENOENT);
	if (*missing)
	return (0);

	if (error != 0) {
	(void) zfs_standard_error_fmt(zhp->zfs_hdl, error,
	dgettext(TEXT_DOMAIN, "error waiting in fs '%s'"),
	zhp->zfs_name);
	}

	return (error);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_diff.c b/sys/contrib/openzfs/lib/libzfs/libzfs_diff.c
	index 93f6e19e9127..80588a860c18 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_diff.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_diff.c
	@@ -1,790 +1,788 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2015, 2018 by Delphix. All rights reserved.
	* Copyright 2016 Joyent, Inc.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	*/

	/*
	* zfs diff support
	*/
	#include <ctype.h>
	#include <errno.h>
	#include <libintl.h>
	#include <string.h>
	#include <sys/types.h>
	#include <sys/stat.h>
	#include <fcntl.h>
	#include <stddef.h>
	#include <unistd.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <pthread.h>
	#include <sys/zfs_ioctl.h>
	#include <libzfs.h>
	#include "libzfs_impl.h"

	#define ZDIFF_SNAPDIR "/.zfs/snapshot/"
	#define ZDIFF_PREFIX "zfs-diff-%d"

	#define ZDIFF_ADDED '+'
	#define ZDIFF_MODIFIED "M"
	#define ZDIFF_REMOVED '-'
	#define ZDIFF_RENAMED "R"


	/*
	* Given a {dsname, object id}, get the object path
	*/
	static int
	get_stats_for_obj(differ_info_t di, const char dsname, uint64_t obj,
	char pn, int maxlen, zfs_stat_t sb)
	{
	zfs_cmd_t zc = {"\0"};
	int error;

	(void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
	zc.zc_obj = obj;

	errno = 0;
	error = zfs_ioctl(di->zhp->zfs_hdl, ZFS_IOC_OBJ_TO_STATS, &zc);
	di->zerr = errno;

	/* we can get stats even if we failed to get a path */
	(void) memcpy(sb, &zc.zc_stat, sizeof (zfs_stat_t));
	if (error == 0) {
	ASSERT(di->zerr == 0);
	(void) strlcpy(pn, zc.zc_value, maxlen);
	return (0);
	}

	if (di->zerr == ESTALE) {
	(void) snprintf(pn, maxlen, "(on_delete_queue)");
	return (0);
	} else if (di->zerr == EPERM) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"The sys_config privilege or diff delegated permission "
	"is needed\nto discover path names"));
	return (-1);
	} else if (di->zerr == EACCES) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Key must be loaded to discover path names"));
	return (-1);
	} else {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Unable to determine path or stats for "
	"object %lld in %s"), (longlong_t)obj, dsname);
	return (-1);
	}
	}

	/*
	* stream_bytes
	*
	* Prints a file name out a character at a time. If the character is
	* not in the range of what we consider "printable" ASCII, display it
	* as an escaped 4-digit octal value. ASCII values less than a space
	* are all control characters and we declare the upper end as the
	* DELete character. This also is the last 7-bit ASCII character.
	* We choose to treat all 8-bit ASCII as not printable for this
	* application.
	*/
	static void
	stream_bytes(FILE fp, const char string)
	{
	char c;

	while ((c = *string++) != '\0') {
	if (c > ' ' && c != '\\' && c < '\177') {
	(void) fputc(c, fp);
	} else {
	(void) fprintf(fp, "\\%04hho", (uint8_t)c);
	}
	}
	}

	static char
	get_what(mode_t what)
	{
	switch (what & S_IFMT) {
	case S_IFBLK:
	return ('B');
	case S_IFCHR:
	return ('C');
	case S_IFDIR:
	return ('/');
	#ifdef S_IFDOOR
	case S_IFDOOR:
	return ('>');
	#endif
	case S_IFIFO:
	return ('\|');
	case S_IFLNK:
	return ('@');
	#ifdef S_IFPORT
	case S_IFPORT:
	return ('P');
	#endif
	case S_IFSOCK:
	return ('=');
	case S_IFREG:
	return ('F');
	default:
	return ('?');
	}
	}

	static void
	print_cmn(FILE fp, differ_info_t di, const char *file)
	{
	if (!di->no_mangle) {
	stream_bytes(fp, di->dsmnt);
	stream_bytes(fp, file);
	} else {
	(void) fputs(di->dsmnt, fp);
	(void) fputs(file, fp);
	}
	}

	static void
	print_rename(FILE fp, differ_info_t di, const char old, const char new,
	zfs_stat_t *isb)
	{
	if (di->timestamped)
	(void) fprintf(fp, "%10lld.%09lld\t",
	(longlong_t)isb->zs_ctime[0],
	(longlong_t)isb->zs_ctime[1]);
	(void) fputs(ZDIFF_RENAMED "\t", fp);
	if (di->classify)
	(void) fprintf(fp, "%c\t", get_what(isb->zs_mode));
	print_cmn(fp, di, old);
	(void) fputs(di->scripted ? "\t" : " -> ", fp);
	print_cmn(fp, di, new);
	(void) fputc('\n', fp);
	}

	static void
	print_link_change(FILE fp, differ_info_t di, int delta, const char *file,
	zfs_stat_t *isb)
	{
	if (di->timestamped)
	(void) fprintf(fp, "%10lld.%09lld\t",
	(longlong_t)isb->zs_ctime[0],
	(longlong_t)isb->zs_ctime[1]);
	(void) fputs(ZDIFF_MODIFIED "\t", fp);
	if (di->classify)
	(void) fprintf(fp, "%c\t", get_what(isb->zs_mode));
	print_cmn(fp, di, file);
	(void) fprintf(fp, "\t(%+d)\n", delta);
	}

	static void
	print_file(FILE fp, differ_info_t di, char type, const char *file,
	zfs_stat_t *isb)
	{
	if (di->timestamped)
	(void) fprintf(fp, "%10lld.%09lld\t",
	(longlong_t)isb->zs_ctime[0],
	(longlong_t)isb->zs_ctime[1]);
	(void) fprintf(fp, "%c\t", type);
	if (di->classify)
	(void) fprintf(fp, "%c\t", get_what(isb->zs_mode));
	print_cmn(fp, di, file);
	(void) fputc('\n', fp);
	}

	static int
	write_inuse_diffs_one(FILE fp, differ_info_t di, uint64_t dobj)
	{
	struct zfs_stat fsb, tsb;
	mode_t fmode, tmode;
	char fobjname[MAXPATHLEN], tobjname[MAXPATHLEN];
	boolean_t already_logged = B_FALSE;
	int fobjerr, tobjerr;
	int change;

	if (dobj == di->shares)
	return (0);

	/*
	* Check the from and to snapshots for info on the object. If
	* we get ENOENT, then the object just didn't exist in that
	* snapshot. If we get ENOTSUP, then we tried to get
	* info on a non-ZPL object, which we don't care about anyway.
	* For any other error we print a warning which includes the
	* errno and continue.
	*/

	fobjerr = get_stats_for_obj(di, di->fromsnap, dobj, fobjname,
	MAXPATHLEN, &fsb);
	if (fobjerr && di->zerr != ENOTSUP && di->zerr != ENOENT) {
	zfs_error_aux(di->zhp->zfs_hdl, "%s", strerror(di->zerr));
	zfs_error(di->zhp->zfs_hdl, di->zerr, di->errbuf);
	/*
	* Let's not print an error for the same object more than
	* once if it happens in both snapshots
	*/
	already_logged = B_TRUE;
	}

	tobjerr = get_stats_for_obj(di, di->tosnap, dobj, tobjname,
	MAXPATHLEN, &tsb);

	if (tobjerr && di->zerr != ENOTSUP && di->zerr != ENOENT) {
	if (!already_logged) {
	zfs_error_aux(di->zhp->zfs_hdl,
	"%s", strerror(di->zerr));
	zfs_error(di->zhp->zfs_hdl, di->zerr, di->errbuf);
	}
	}
	/*
	* Unallocated object sharing the same meta dnode block
	*/
	if (fobjerr && tobjerr) {
	di->zerr = 0;
	return (0);
	}

	di->zerr = 0; /* negate get_stats_for_obj() from side that failed */
	fmode = fsb.zs_mode & S_IFMT;
	tmode = tsb.zs_mode & S_IFMT;
	if (fmode == S_IFDIR \|\| tmode == S_IFDIR \|\| fsb.zs_links == 0 \|\|
	tsb.zs_links == 0)
	change = 0;
	else
	change = tsb.zs_links - fsb.zs_links;

	if (fobjerr) {
	if (change) {
	print_link_change(fp, di, change, tobjname, &tsb);
	return (0);
	}
	print_file(fp, di, ZDIFF_ADDED, tobjname, &tsb);
	return (0);
	} else if (tobjerr) {
	if (change) {
	print_link_change(fp, di, change, fobjname, &fsb);
	return (0);
	}
	print_file(fp, di, ZDIFF_REMOVED, fobjname, &fsb);
	return (0);
	}

	if (fmode != tmode && fsb.zs_gen == tsb.zs_gen)
	tsb.zs_gen++; /* Force a generational difference */

	/* Simple modification or no change */
	if (fsb.zs_gen == tsb.zs_gen) {
	/* No apparent changes. Could we assert !this? */
	if (fsb.zs_ctime[0] == tsb.zs_ctime[0] &&
	fsb.zs_ctime[1] == tsb.zs_ctime[1])
	return (0);
	if (change) {
	print_link_change(fp, di, change,
	change > 0 ? fobjname : tobjname, &tsb);
	} else if (strcmp(fobjname, tobjname) == 0) {
	print_file(fp, di, *ZDIFF_MODIFIED, fobjname, &tsb);
	} else {
	print_rename(fp, di, fobjname, tobjname, &tsb);
	}
	return (0);
	} else {
	/* file re-created or object re-used */
	print_file(fp, di, ZDIFF_REMOVED, fobjname, &fsb);
	print_file(fp, di, ZDIFF_ADDED, tobjname, &tsb);
	return (0);
	}
	}

	static int
	write_inuse_diffs(FILE fp, differ_info_t di, dmu_diff_record_t *dr)
	{
	uint64_t o;
	int err;

	for (o = dr->ddr_first; o <= dr->ddr_last; o++) {
	if ((err = write_inuse_diffs_one(fp, di, o)) != 0)
	return (err);
	}
	return (0);
	}

	static int
	describe_free(FILE fp, differ_info_t di, uint64_t object, char *namebuf,
	int maxlen)
	{
	struct zfs_stat sb;

	(void) get_stats_for_obj(di, di->fromsnap, object, namebuf,
	maxlen, &sb);

	/* Don't print if in the delete queue on from side */
	if (di->zerr == ESTALE \|\| di->zerr == ENOENT) {
	di->zerr = 0;
	return (0);
	}

	print_file(fp, di, ZDIFF_REMOVED, namebuf, &sb);
	return (0);
	}

	static int
	write_free_diffs(FILE fp, differ_info_t di, dmu_diff_record_t *dr)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *lhdl = di->zhp->zfs_hdl;
	char fobjname[MAXPATHLEN];

	(void) strlcpy(zc.zc_name, di->fromsnap, sizeof (zc.zc_name));
	zc.zc_obj = dr->ddr_first - 1;

	ASSERT(di->zerr == 0);

	while (zc.zc_obj < dr->ddr_last) {
	int err;

	err = zfs_ioctl(lhdl, ZFS_IOC_NEXT_OBJ, &zc);
	if (err == 0) {
	if (zc.zc_obj == di->shares) {
	zc.zc_obj++;
	continue;
	}
	if (zc.zc_obj > dr->ddr_last) {
	break;
	}
	err = describe_free(fp, di, zc.zc_obj, fobjname,
	MAXPATHLEN);
	} else if (errno == ESRCH) {
	break;
	} else {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"next allocated object (> %lld) find failure"),
	(longlong_t)zc.zc_obj);
	di->zerr = errno;
	break;
	}
	}
	if (di->zerr)
	return (-1);
	return (0);
	}

	static void *
	differ(void *arg)
	{
	differ_info_t *di = arg;
	dmu_diff_record_t dr;
	FILE *ofp;
	int err = 0;

	if ((ofp = fdopen(di->outputfd, "w")) == NULL) {
	di->zerr = errno;
	strlcpy(di->errbuf, strerror(errno), sizeof (di->errbuf));
	(void) close(di->datafd);
	return ((void *)-1);
	}

	for (;;) {
	char cp = (char )&dr;
	int len = sizeof (dr);
	int rv;

	do {
	rv = read(di->datafd, cp, len);
	cp += rv;
	len -= rv;
	} while (len > 0 && rv > 0);

	if (rv < 0 \|\| (rv == 0 && len != sizeof (dr))) {
	di->zerr = EPIPE;
	break;
	} else if (rv == 0) {
	/* end of file at a natural breaking point */
	break;
	}

	switch (dr.ddr_type) {
	case DDR_FREE:
	err = write_free_diffs(ofp, di, &dr);
	break;
	case DDR_INUSE:
	err = write_inuse_diffs(ofp, di, &dr);
	break;
	default:
	di->zerr = EPIPE;
	break;
	}

	if (err \|\| di->zerr)
	break;
	}

	(void) fclose(ofp);
	(void) close(di->datafd);
	if (err)
	return ((void *)-1);
	if (di->zerr) {
	ASSERT(di->zerr == EPIPE);
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Internal error: bad data from diff IOCTL"));
	return ((void *)-1);
	}
	return ((void *)0);
	}

	static int
	make_temp_snapshot(differ_info_t *di)
	{
	libzfs_handle_t *hdl = di->zhp->zfs_hdl;
	zfs_cmd_t zc = {"\0"};

	(void) snprintf(zc.zc_value, sizeof (zc.zc_value),
	ZDIFF_PREFIX, getpid());
	(void) strlcpy(zc.zc_name, di->ds, sizeof (zc.zc_name));
	zc.zc_cleanup_fd = di->cleanupfd;

	if (zfs_ioctl(hdl, ZFS_IOC_TMP_SNAPSHOT, &zc) != 0) {
	int err = errno;
	if (err == EPERM) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN, "The diff delegated "
	"permission is needed in order\nto create a "
	"just-in-time snapshot for diffing\n"));
	return (zfs_error(hdl, EZFS_DIFF, di->errbuf));
	} else {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN, "Cannot create just-in-time "
	"snapshot of '%s'"), zc.zc_name);
	return (zfs_standard_error(hdl, err, di->errbuf));
	}
	}

	di->tmpsnap = zfs_strdup(hdl, zc.zc_value);
	di->tosnap = zfs_asprintf(hdl, "%s@%s", di->ds, di->tmpsnap);
	return (0);
	}

	static void
	teardown_differ_info(differ_info_t *di)
	{
	free(di->ds);
	free(di->dsmnt);
	free(di->fromsnap);
	free(di->frommnt);
	free(di->tosnap);
	free(di->tmpsnap);
	free(di->tomnt);
	(void) close(di->cleanupfd);
	}

	static int
	get_snapshot_names(differ_info_t di, const char fromsnap,
	const char *tosnap)
	{
	libzfs_handle_t *hdl = di->zhp->zfs_hdl;
	char *atptrf = NULL;
	char *atptrt = NULL;
	int fdslen, fsnlen;
	int tdslen, tsnlen;

	/*
	* Can accept
	* fdslen fsnlen tdslen tsnlen
	* dataset@snap1
	* 0. dataset@snap1 dataset@snap2 >0 >1 >0 >1
	* 1. dataset@snap1 @snap2 >0 >1 ==0 >1
	* 2. dataset@snap1 dataset >0 >1 >0 ==0
	* 3. @snap1 dataset@snap2 ==0 >1 >0 >1
	* 4. @snap1 dataset ==0 >1 >0 ==0
	*/
	if (tosnap == NULL) {
	/* only a from snapshot given, must be valid */
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Badly formed snapshot name %s"), fromsnap);

	if (!zfs_validate_name(hdl, fromsnap, ZFS_TYPE_SNAPSHOT,
	B_FALSE)) {
	return (zfs_error(hdl, EZFS_INVALIDNAME,
	di->errbuf));
	}

	atptrf = strchr(fromsnap, '@');
	ASSERT(atptrf != NULL);
	fdslen = atptrf - fromsnap;

	di->fromsnap = zfs_strdup(hdl, fromsnap);
	di->ds = zfs_strdup(hdl, fromsnap);
	di->ds[fdslen] = '\0';

	/* the to snap will be a just-in-time snap of the head */
	return (make_temp_snapshot(di));
	}

	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Unable to determine which snapshots to compare"));

	atptrf = strchr(fromsnap, '@');
	atptrt = strchr(tosnap, '@');
	fdslen = atptrf ? atptrf - fromsnap : strlen(fromsnap);
	tdslen = atptrt ? atptrt - tosnap : strlen(tosnap);
	fsnlen = strlen(fromsnap) - fdslen; /* includes @ sign */
	tsnlen = strlen(tosnap) - tdslen; /* includes @ sign */

	if (fsnlen <= 1 \|\| tsnlen == 1 \|\| (fdslen == 0 && tdslen == 0)) {
	return (zfs_error(hdl, EZFS_INVALIDNAME, di->errbuf));
	} else if ((fdslen > 0 && tdslen > 0) &&
	((tdslen != fdslen \|\| strncmp(fromsnap, tosnap, fdslen) != 0))) {
	/*
	* not the same dataset name, might be okay if
	* tosnap is a clone of a fromsnap descendant.
	*/
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	zprop_source_t src;
	zfs_handle_t *zhp;

	di->ds = zfs_alloc(di->zhp->zfs_hdl, tdslen + 1);
	- (void) strncpy(di->ds, tosnap, tdslen);
	- di->ds[tdslen] = '\0';
	+ (void) strlcpy(di->ds, tosnap, tdslen + 1);

	zhp = zfs_open(hdl, di->ds, ZFS_TYPE_FILESYSTEM);
	while (zhp != NULL) {
	if (zfs_prop_get(zhp, ZFS_PROP_ORIGIN, origin,
	sizeof (origin), &src, NULL, 0, B_FALSE) != 0) {
	(void) zfs_close(zhp);
	zhp = NULL;
	break;
	}
	if (strncmp(origin, fromsnap, fsnlen) == 0)
	break;

	(void) zfs_close(zhp);
	zhp = zfs_open(hdl, origin, ZFS_TYPE_FILESYSTEM);
	}

	if (zhp == NULL) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, di->errbuf));
	} else {
	(void) zfs_close(zhp);
	}

	di->isclone = B_TRUE;
	di->fromsnap = zfs_strdup(hdl, fromsnap);
	if (tsnlen)
	di->tosnap = zfs_strdup(hdl, tosnap);
	else
	return (make_temp_snapshot(di));
	} else {
	int dslen = fdslen ? fdslen : tdslen;

	di->ds = zfs_alloc(hdl, dslen + 1);
	- (void) strncpy(di->ds, fdslen ? fromsnap : tosnap, dslen);
	- di->ds[dslen] = '\0';
	+ (void) strlcpy(di->ds, fdslen ? fromsnap : tosnap, dslen + 1);

	di->fromsnap = zfs_asprintf(hdl, "%s%s", di->ds, atptrf);
	if (tsnlen) {
	di->tosnap = zfs_asprintf(hdl, "%s%s", di->ds, atptrt);
	} else {
	return (make_temp_snapshot(di));
	}
	}
	return (0);
	}

	static int
	get_mountpoint(differ_info_t di, char dsnm, char **mntpt)
	{
	boolean_t mounted;

	mounted = is_mounted(di->zhp->zfs_hdl, dsnm, mntpt);
	if (mounted == B_FALSE) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN,
	"Cannot diff an unmounted snapshot"));
	return (zfs_error(di->zhp->zfs_hdl, EZFS_BADTYPE, di->errbuf));
	}

	/* Avoid a double slash at the beginning of root-mounted datasets */
	if (*mntpt == '/' && (*mntpt + 1) == '\0')
	**mntpt = '\0';
	return (0);
	}

	static int
	get_mountpoints(differ_info_t *di)
	{
	char *strptr;
	char *frommntpt;

	/*
	* first get the mountpoint for the parent dataset
	*/
	if (get_mountpoint(di, di->ds, &di->dsmnt) != 0)
	return (-1);

	strptr = strchr(di->tosnap, '@');
	ASSERT3P(strptr, !=, NULL);
	di->tomnt = zfs_asprintf(di->zhp->zfs_hdl, "%s%s%s", di->dsmnt,
	ZDIFF_SNAPDIR, ++strptr);

	strptr = strchr(di->fromsnap, '@');
	ASSERT3P(strptr, !=, NULL);

	frommntpt = di->dsmnt;
	if (di->isclone) {
	char *mntpt;
	int err;

	*strptr = '\0';
	err = get_mountpoint(di, di->fromsnap, &mntpt);
	*strptr = '@';
	if (err != 0)
	return (-1);
	frommntpt = mntpt;
	}

	di->frommnt = zfs_asprintf(di->zhp->zfs_hdl, "%s%s%s", frommntpt,
	ZDIFF_SNAPDIR, ++strptr);

	if (di->isclone)
	free(frommntpt);

	return (0);
	}

	static int
	setup_differ_info(zfs_handle_t zhp, const char fromsnap,
	const char tosnap, differ_info_t di)
	{
	di->zhp = zhp;

	di->cleanupfd = open(ZFS_DEV, O_RDWR \| O_CLOEXEC);
	VERIFY(di->cleanupfd >= 0);

	if (get_snapshot_names(di, fromsnap, tosnap) != 0)
	return (-1);

	if (get_mountpoints(di) != 0)
	return (-1);

	if (find_shares_object(di) != 0)
	return (-1);

	return (0);
	}

	int
	zfs_show_diffs(zfs_handle_t zhp, int outfd, const char fromsnap,
	const char *tosnap, int flags)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	differ_info_t di = { 0 };
	pthread_t tid;
	int pipefd[2];
	int iocerr;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "zfs diff failed"));

	if (setup_differ_info(zhp, fromsnap, tosnap, &di)) {
	teardown_differ_info(&di);
	return (-1);
	}

	if (pipe2(pipefd, O_CLOEXEC)) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(errno));
	teardown_differ_info(&di);
	return (zfs_error(zhp->zfs_hdl, EZFS_PIPEFAILED, errbuf));
	}

	di.scripted = (flags & ZFS_DIFF_PARSEABLE);
	di.classify = (flags & ZFS_DIFF_CLASSIFY);
	di.timestamped = (flags & ZFS_DIFF_TIMESTAMP);
	di.no_mangle = (flags & ZFS_DIFF_NO_MANGLE);

	di.outputfd = outfd;
	di.datafd = pipefd[0];

	if (pthread_create(&tid, NULL, differ, &di)) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(errno));
	(void) close(pipefd[0]);
	(void) close(pipefd[1]);
	teardown_differ_info(&di);
	return (zfs_error(zhp->zfs_hdl,
	EZFS_THREADCREATEFAILED, errbuf));
	}

	/* do the ioctl() */
	(void) strlcpy(zc.zc_value, di.fromsnap, strlen(di.fromsnap) + 1);
	(void) strlcpy(zc.zc_name, di.tosnap, strlen(di.tosnap) + 1);
	zc.zc_cookie = pipefd[1];

	iocerr = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_DIFF, &zc);
	if (iocerr != 0) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "Unable to obtain diffs"));
	if (errno == EPERM) {
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"\n The sys_mount privilege or diff delegated "
	"permission is needed\n to execute the "
	"diff ioctl"));
	} else if (errno == EXDEV) {
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"\n Not an earlier snapshot from the same fs"));
	} else if (errno != EPIPE \|\| di.zerr == 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(errno));
	}
	(void) close(pipefd[1]);
	(void) pthread_cancel(tid);
	(void) pthread_join(tid, NULL);
	teardown_differ_info(&di);
	if (di.zerr != 0 && di.zerr != EPIPE) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(di.zerr));
	return (zfs_error(zhp->zfs_hdl, EZFS_DIFF, di.errbuf));
	} else {
	return (zfs_error(zhp->zfs_hdl, EZFS_DIFFDATA, errbuf));
	}
	}

	(void) close(pipefd[1]);
	(void) pthread_join(tid, NULL);

	if (di.zerr != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(di.zerr));
	return (zfs_error(zhp->zfs_hdl, EZFS_DIFF, di.errbuf));
	}
	teardown_differ_info(&di);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c b/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c
	index eea388cf348f..b9806dc30dac 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c
	@@ -1,5181 +1,5218 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright (c) 2018 Datto Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	* Copyright (c) 2021, Klara Inc.
	*/

	#include <errno.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <strings.h>
	#include <unistd.h>
	#include <libgen.h>
	#include <zone.h>
	#include <sys/stat.h>
	#include <sys/efi_partition.h>
	#include <sys/systeminfo.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_sysfs.h>
	#include <sys/vdev_disk.h>
	#include <sys/types.h>
	#include <dlfcn.h>
	#include <libzutil.h>
	#include <fcntl.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "libzfs_impl.h"
	#include "zfs_comutil.h"
	#include "zfeature_common.h"

	static boolean_t zpool_vdev_is_interior(const char *name);

	typedef struct prop_flags {
	int create:1; /* Validate property on creation */
	int import:1; /* Validate property on import */
	int vdevprop:1; /* Validate property as a VDEV property */
	} prop_flags_t;

	/*
	* ====================================================================
	* zpool property functions
	* ====================================================================
	*/

	static int
	zpool_get_all_props(zpool_handle_t *zhp)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));

	zcmd_alloc_dst_nvlist(hdl, &zc, 0);

	while (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
	if (errno == ENOMEM)
	zcmd_expand_dst_nvlist(hdl, &zc);
	else {
	zcmd_free_nvlists(&zc);
	return (-1);
	}
	}

	if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
	zcmd_free_nvlists(&zc);
	return (-1);
	}

	zcmd_free_nvlists(&zc);

	return (0);
	}

	int
	zpool_props_refresh(zpool_handle_t *zhp)
	{
	nvlist_t *old_props;

	old_props = zhp->zpool_props;

	if (zpool_get_all_props(zhp) != 0)
	return (-1);

	nvlist_free(old_props);
	return (0);
	}

	static const char *
	zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
	zprop_source_t *src)
	{
	nvlist_t nv, nvl;
	const char *value;
	zprop_source_t source;

	nvl = zhp->zpool_props;
	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
	source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	value = fnvlist_lookup_string(nv, ZPROP_VALUE);
	} else {
	source = ZPROP_SRC_DEFAULT;
	if ((value = zpool_prop_default_string(prop)) == NULL)
	value = "-";
	}

	if (src)
	*src = source;

	return (value);
	}

	uint64_t
	zpool_get_prop_int(zpool_handle_t zhp, zpool_prop_t prop, zprop_source_t src)
	{
	nvlist_t nv, nvl;
	uint64_t value;
	zprop_source_t source;

	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
	/*
	* zpool_get_all_props() has most likely failed because
	* the pool is faulted, but if all we need is the top level
	* vdev's guid then get it from the zhp config nvlist.
	*/
	if ((prop == ZPOOL_PROP_GUID) &&
	(nvlist_lookup_nvlist(zhp->zpool_config,
	ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
	(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
	== 0)) {
	return (value);
	}
	return (zpool_prop_default_numeric(prop));
	}

	nvl = zhp->zpool_props;
	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
	source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
	} else {
	source = ZPROP_SRC_DEFAULT;
	value = zpool_prop_default_numeric(prop);
	}

	if (src)
	*src = source;

	return (value);
	}

	/*
	* Map VDEV STATE to printed strings.
	*/
	const char *
	zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
	{
	switch (state) {
	case VDEV_STATE_CLOSED:
	case VDEV_STATE_OFFLINE:
	return (gettext("OFFLINE"));
	case VDEV_STATE_REMOVED:
	return (gettext("REMOVED"));
	case VDEV_STATE_CANT_OPEN:
	if (aux == VDEV_AUX_CORRUPT_DATA \|\| aux == VDEV_AUX_BAD_LOG)
	return (gettext("FAULTED"));
	else if (aux == VDEV_AUX_SPLIT_POOL)
	return (gettext("SPLIT"));
	else
	return (gettext("UNAVAIL"));
	case VDEV_STATE_FAULTED:
	return (gettext("FAULTED"));
	case VDEV_STATE_DEGRADED:
	return (gettext("DEGRADED"));
	case VDEV_STATE_HEALTHY:
	return (gettext("ONLINE"));

	default:
	break;
	}

	return (gettext("UNKNOWN"));
	}

	/*
	* Map POOL STATE to printed strings.
	*/
	const char *
	zpool_pool_state_to_name(pool_state_t state)
	{
	switch (state) {
	default:
	break;
	case POOL_STATE_ACTIVE:
	return (gettext("ACTIVE"));
	case POOL_STATE_EXPORTED:
	return (gettext("EXPORTED"));
	case POOL_STATE_DESTROYED:
	return (gettext("DESTROYED"));
	case POOL_STATE_SPARE:
	return (gettext("SPARE"));
	case POOL_STATE_L2CACHE:
	return (gettext("L2CACHE"));
	case POOL_STATE_UNINITIALIZED:
	return (gettext("UNINITIALIZED"));
	case POOL_STATE_UNAVAIL:
	return (gettext("UNAVAIL"));
	case POOL_STATE_POTENTIALLY_ACTIVE:
	return (gettext("POTENTIALLY_ACTIVE"));
	}

	return (gettext("UNKNOWN"));
	}

	/*
	* Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
	* "SUSPENDED", etc).
	*/
	const char *
	zpool_get_state_str(zpool_handle_t *zhp)
	{
	zpool_errata_t errata;
	zpool_status_t status;
	const char *str;

	status = zpool_get_status(zhp, NULL, &errata);

	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
	str = gettext("FAULTED");
	} else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT \|\|
	status == ZPOOL_STATUS_IO_FAILURE_MMP) {
	str = gettext("SUSPENDED");
	} else {
	nvlist_t *nvroot = fnvlist_lookup_nvlist(
	zpool_get_config(zhp, NULL), ZPOOL_CONFIG_VDEV_TREE);
	uint_t vsc;
	vdev_stat_t vs = (vdev_stat_t )fnvlist_lookup_uint64_array(
	nvroot, ZPOOL_CONFIG_VDEV_STATS, &vsc);
	str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
	}
	return (str);
	}

	/*
	* Get a zpool property value for 'prop' and return the value in
	* a pre-allocated buffer.
	*/
	int
	zpool_get_prop(zpool_handle_t zhp, zpool_prop_t prop, char buf,
	size_t len, zprop_source_t *srctype, boolean_t literal)
	{
	uint64_t intval;
	const char *strval;
	zprop_source_t src = ZPROP_SRC_NONE;

	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
	switch (prop) {
	case ZPOOL_PROP_NAME:
	(void) strlcpy(buf, zpool_get_name(zhp), len);
	break;

	case ZPOOL_PROP_HEALTH:
	(void) strlcpy(buf, zpool_get_state_str(zhp), len);
	break;

	case ZPOOL_PROP_GUID:
	intval = zpool_get_prop_int(zhp, prop, &src);
	(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
	break;

	case ZPOOL_PROP_ALTROOT:
	case ZPOOL_PROP_CACHEFILE:
	case ZPOOL_PROP_COMMENT:
	case ZPOOL_PROP_COMPATIBILITY:
	if (zhp->zpool_props != NULL \|\|
	zpool_get_all_props(zhp) == 0) {
	(void) strlcpy(buf,
	zpool_get_prop_string(zhp, prop, &src),
	len);
	break;
	}
	zfs_fallthrough;
	default:
	(void) strlcpy(buf, "-", len);
	break;
	}

	if (srctype != NULL)
	*srctype = src;
	return (0);
	}

	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
	prop != ZPOOL_PROP_NAME)
	return (-1);

	switch (zpool_prop_get_type(prop)) {
	case PROP_TYPE_STRING:
	(void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
	len);
	break;

	case PROP_TYPE_NUMBER:
	intval = zpool_get_prop_int(zhp, prop, &src);

	switch (prop) {
	case ZPOOL_PROP_SIZE:
	case ZPOOL_PROP_ALLOCATED:
	case ZPOOL_PROP_FREE:
	case ZPOOL_PROP_FREEING:
	case ZPOOL_PROP_LEAKED:
	case ZPOOL_PROP_ASHIFT:
	case ZPOOL_PROP_MAXBLOCKSIZE:
	case ZPOOL_PROP_MAXDNODESIZE:
	if (literal)
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	else
	(void) zfs_nicenum(intval, buf, len);
	break;

	case ZPOOL_PROP_EXPANDSZ:
	case ZPOOL_PROP_CHECKPOINT:
	if (intval == 0) {
	(void) strlcpy(buf, "-", len);
	} else if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) zfs_nicebytes(intval, buf, len);
	}
	break;

	case ZPOOL_PROP_CAPACITY:
	if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) snprintf(buf, len, "%llu%%",
	(u_longlong_t)intval);
	}
	break;

	case ZPOOL_PROP_FRAGMENTATION:
	if (intval == UINT64_MAX) {
	(void) strlcpy(buf, "-", len);
	} else if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) snprintf(buf, len, "%llu%%",
	(u_longlong_t)intval);
	}
	break;

	case ZPOOL_PROP_DEDUPRATIO:
	if (literal)
	(void) snprintf(buf, len, "%llu.%02llu",
	(u_longlong_t)(intval / 100),
	(u_longlong_t)(intval % 100));
	else
	(void) snprintf(buf, len, "%llu.%02llux",
	(u_longlong_t)(intval / 100),
	(u_longlong_t)(intval % 100));
	break;

	case ZPOOL_PROP_HEALTH:
	(void) strlcpy(buf, zpool_get_state_str(zhp), len);
	break;
	case ZPOOL_PROP_VERSION:
	if (intval >= SPA_VERSION_FEATURES) {
	(void) snprintf(buf, len, "-");
	break;
	}
	zfs_fallthrough;
	default:
	(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
	}
	break;

	case PROP_TYPE_INDEX:
	intval = zpool_get_prop_int(zhp, prop, &src);
	if (zpool_prop_index_to_string(prop, intval, &strval)
	!= 0)
	return (-1);
	(void) strlcpy(buf, strval, len);
	break;

	default:
	abort();
	}

	if (srctype)
	*srctype = src;

	return (0);
	}

	/*
	* Check if the bootfs name has the same pool name as it is set to.
	* Assuming bootfs is a valid dataset name.
	*/
	static boolean_t
	bootfs_name_valid(const char pool, const char bootfs)
	{
	int len = strlen(pool);
	if (bootfs[0] == '\0')
	return (B_TRUE);

	if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM\|ZFS_TYPE_SNAPSHOT))
	return (B_FALSE);

	if (strncmp(pool, bootfs, len) == 0 &&
	(bootfs[len] == '/' \|\| bootfs[len] == '\0'))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Given an nvlist of zpool properties to be set, validate that they are
	* correct, and parse any numeric properties (index, boolean, etc) if they are
	* specified as strings.
	*/
	static nvlist_t *
	zpool_valid_proplist(libzfs_handle_t hdl, const char poolname,
	nvlist_t props, uint64_t version, prop_flags_t flags, char errbuf)
	{
	nvpair_t *elem;
	nvlist_t *retprops;
	zpool_prop_t prop;
	char *strval;
	uint64_t intval;
	char slash, check;
	struct stat64 statbuf;
	zpool_handle_t *zhp;
	char report[1024];

	if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
	(void) no_memory(hdl);
	return (NULL);
	}

	elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	const char *propname = nvpair_name(elem);

	if (flags.vdevprop && zpool_prop_vdev(propname)) {
	vdev_prop_t vprop = vdev_name_to_prop(propname);

	if (vdev_prop_readonly(vprop)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
	"is readonly"), propname);
	(void) zfs_error(hdl, EZFS_PROPREADONLY,
	errbuf);
	goto error;
	}

	if (zprop_parse_value(hdl, elem, vprop, ZFS_TYPE_VDEV,
	retprops, &strval, &intval, errbuf) != 0)
	goto error;

	continue;
	} else if (flags.vdevprop && vdev_prop_user(propname)) {
	if (nvlist_add_nvpair(retprops, elem) != 0) {
	(void) no_memory(hdl);
	goto error;
	}
	continue;
	} else if (flags.vdevprop) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property: '%s'"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	prop = zpool_name_to_prop(propname);
	if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
	int err;
	char *fname = strchr(propname, '@') + 1;

	err = zfeature_lookup_name(fname, NULL);
	if (err != 0) {
	ASSERT3U(err, ==, ENOENT);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"feature '%s' unsupported by kernel"),
	fname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (nvpair_type(elem) != DATA_TYPE_STRING) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a string"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	(void) nvpair_value_string(elem, &strval);
	if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
	strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' can only be set to "
	"'enabled' or 'disabled'"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (!flags.create &&
	strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' can only be set to "
	"'disabled' at creation time"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (nvlist_add_uint64(retprops, propname, 0) != 0) {
	(void) no_memory(hdl);
	goto error;
	}
	continue;
	}

	/*
	* Make sure this property is valid and applies to this type.
	*/
	if (prop == ZPOOL_PROP_INVAL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property '%s'"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (zpool_prop_readonly(prop)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
	"is readonly"), propname);
	(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
	goto error;
	}

	if (!flags.create && zpool_prop_setonce(prop)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' can only be set at "
	"creation time"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
	&strval, &intval, errbuf) != 0)
	goto error;

	/*
	* Perform additional checking for specific properties.
	*/
	switch (prop) {
	case ZPOOL_PROP_VERSION:
	if (intval < version \|\|
	!SPA_VERSION_IS_SUPPORTED(intval)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' number %llu is invalid."),
	propname, (unsigned long long)intval);
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	goto error;
	}
	break;

	case ZPOOL_PROP_ASHIFT:
	if (intval != 0 &&
	(intval < ASHIFT_MIN \|\| intval > ASHIFT_MAX)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' number %llu is invalid, "
	"only values between %" PRId32 " and %"
	PRId32 " are allowed."),
	propname, (unsigned long long)intval,
	ASHIFT_MIN, ASHIFT_MAX);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;

	case ZPOOL_PROP_BOOTFS:
	if (flags.create \|\| flags.import) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' cannot be set at creation "
	"or import time"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (version < SPA_VERSION_BOOTFS) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to support "
	"'%s' property"), propname);
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	goto error;
	}

	/*
	* bootfs property value has to be a dataset name and
	* the dataset has to be in the same pool as it sets to.
	*/
	if (!bootfs_name_valid(poolname, strval)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
	"is an invalid name"), strval);
	(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto error;
	}

	if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"could not open pool '%s'"), poolname);
	(void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
	goto error;
	}
	zpool_close(zhp);
	break;

	case ZPOOL_PROP_ALTROOT:
	if (!flags.create && !flags.import) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' can only be set during pool "
	"creation or import"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	if (strval[0] != '/') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"bad alternate root '%s'"), strval);
	(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
	goto error;
	}
	break;

	case ZPOOL_PROP_CACHEFILE:
	if (strval[0] == '\0')
	break;

	if (strcmp(strval, "none") == 0)
	break;

	if (strval[0] != '/') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' must be empty, an "
	"absolute path, or 'none'"), propname);
	(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
	goto error;
	}

	slash = strrchr(strval, '/');

	if (slash[1] == '\0' \|\| strcmp(slash, "/.") == 0 \|\|
	strcmp(slash, "/..") == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is not a valid file"), strval);
	(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
	goto error;
	}

	*slash = '\0';

	if (strval[0] != '\0' &&
	(stat64(strval, &statbuf) != 0 \|\|
	!S_ISDIR(statbuf.st_mode))) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is not a valid directory"),
	strval);
	(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
	goto error;
	}

	*slash = '/';
	break;

	case ZPOOL_PROP_COMPATIBILITY:
	switch (zpool_load_compat(strval, NULL, report, 1024)) {
	case ZPOOL_COMPATIBILITY_OK:
	case ZPOOL_COMPATIBILITY_WARNTOKEN:
	break;
	case ZPOOL_COMPATIBILITY_BADFILE:
	case ZPOOL_COMPATIBILITY_BADTOKEN:
	case ZPOOL_COMPATIBILITY_NOFILES:
	zfs_error_aux(hdl, "%s", report);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;

	case ZPOOL_PROP_COMMENT:
	for (check = strval; *check != '\0'; check++) {
	if (!isprint(*check)) {
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN,
	"comment may only have printable "
	"characters"));
	(void) zfs_error(hdl, EZFS_BADPROP,
	errbuf);
	goto error;
	}
	}
	if (strlen(strval) > ZPROP_MAX_COMMENT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"comment must not exceed %d characters"),
	ZPROP_MAX_COMMENT);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	case ZPOOL_PROP_READONLY:
	if (!flags.import) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' can only be set at "
	"import time"), propname);
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	case ZPOOL_PROP_MULTIHOST:
	if (get_system_hostid() == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"requires a non-zero system hostid"));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	break;
	case ZPOOL_PROP_DEDUPDITTO:
	printf("Note: property '%s' no longer has "
	"any effect\n", propname);
	break;

	default:
	break;
	}
	}

	return (retprops);
	error:
	nvlist_free(retprops);
	return (NULL);
	}

	/*
	* Set zpool property : propname=propval.
	*/
	int
	zpool_set_prop(zpool_handle_t zhp, const char propname, const char *propval)
	{
	zfs_cmd_t zc = {"\0"};
	int ret = -1;
	char errbuf[ERRBUFLEN];
	nvlist_t *nvl = NULL;
	nvlist_t *realprops;
	uint64_t version;
	prop_flags_t flags = { 0 };

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
	zhp->zpool_name);

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));

	if (nvlist_add_string(nvl, propname, propval) != 0) {
	nvlist_free(nvl);
	return (no_memory(zhp->zpool_hdl));
	}

	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
	zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
	nvlist_free(nvl);
	return (-1);
	}

	nvlist_free(nvl);
	nvl = realprops;

	/*
	* Execute the corresponding ioctl() to set this property.
	*/
	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));

	zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl);

	ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);

	zcmd_free_nvlists(&zc);
	nvlist_free(nvl);

	if (ret)
	(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
	else
	(void) zpool_props_refresh(zhp);

	return (ret);
	}

	int
	zpool_expand_proplist(zpool_handle_t zhp, zprop_list_t *plp,
	zfs_type_t type, boolean_t literal)
	{
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	zprop_list_t *entry;
	char buf[ZFS_MAXPROPLEN];
	nvlist_t *features = NULL;
	nvpair_t *nvp;
	zprop_list_t **last;
	boolean_t firstexpand = (NULL == *plp);
	int i;

	if (zprop_expand_list(hdl, plp, type) != 0)
	return (-1);

	if (type == ZFS_TYPE_VDEV)
	return (0);

	last = plp;
	while (*last != NULL)
	last = &(*last)->pl_next;

	if ((*plp)->pl_all)
	features = zpool_get_features(zhp);

	if ((*plp)->pl_all && firstexpand) {
	for (i = 0; i < SPA_FEATURES; i++) {
	zprop_list_t *entry = zfs_alloc(hdl,
	sizeof (zprop_list_t));
	entry->pl_prop = ZPROP_USERPROP;
	entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
	spa_feature_table[i].fi_uname);
	entry->pl_width = strlen(entry->pl_user_prop);
	entry->pl_all = B_TRUE;

	*last = entry;
	last = &entry->pl_next;
	}
	}

	/* add any unsupported features */
	for (nvp = nvlist_next_nvpair(features, NULL);
	nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
	char *propname;
	boolean_t found;
	zprop_list_t *entry;

	if (zfeature_is_supported(nvpair_name(nvp)))
	continue;

	propname = zfs_asprintf(hdl, "unsupported@%s",
	nvpair_name(nvp));

	/*
	* Before adding the property to the list make sure that no
	* other pool already added the same property.
	*/
	found = B_FALSE;
	entry = *plp;
	while (entry != NULL) {
	if (entry->pl_user_prop != NULL &&
	strcmp(propname, entry->pl_user_prop) == 0) {
	found = B_TRUE;
	break;
	}
	entry = entry->pl_next;
	}
	if (found) {
	free(propname);
	continue;
	}

	entry = zfs_alloc(hdl, sizeof (zprop_list_t));
	entry->pl_prop = ZPROP_USERPROP;
	entry->pl_user_prop = propname;
	entry->pl_width = strlen(entry->pl_user_prop);
	entry->pl_all = B_TRUE;

	*last = entry;
	last = &entry->pl_next;
	}

	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
	if (entry->pl_fixed && !literal)
	continue;

	if (entry->pl_prop != ZPROP_USERPROP &&
	zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
	NULL, literal) == 0) {
	if (strlen(buf) > entry->pl_width)
	entry->pl_width = strlen(buf);
	}
	}

	return (0);
	}

	int
	vdev_expand_proplist(zpool_handle_t zhp, const char vdevname,
	zprop_list_t **plp)
	{
	zprop_list_t *entry;
	char buf[ZFS_MAXPROPLEN];
	char *strval = NULL;
	int err = 0;
	nvpair_t *elem = NULL;
	nvlist_t *vprops = NULL;
	nvlist_t *propval = NULL;
	const char *propname;
	vdev_prop_t prop;
	zprop_list_t **last;

	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
	if (entry->pl_fixed)
	continue;

	if (zpool_get_vdev_prop(zhp, vdevname, entry->pl_prop,
	entry->pl_user_prop, buf, sizeof (buf), NULL,
	B_FALSE) == 0) {
	if (strlen(buf) > entry->pl_width)
	entry->pl_width = strlen(buf);
	}
	if (entry->pl_prop == VDEV_PROP_NAME &&
	strlen(vdevname) > entry->pl_width)
	entry->pl_width = strlen(vdevname);
	}

	/* Handle the all properties case */
	last = plp;
	if (last != NULL && (last)->pl_all == B_TRUE) {
	while (*last != NULL)
	last = &(*last)->pl_next;

	err = zpool_get_all_vdev_props(zhp, vdevname, &vprops);
	if (err != 0)
	return (err);

	while ((elem = nvlist_next_nvpair(vprops, elem)) != NULL) {
	propname = nvpair_name(elem);

	/* Skip properties that are not user defined */
	if ((prop = vdev_name_to_prop(propname)) !=
	VDEV_PROP_USERPROP)
	continue;

	if (nvpair_value_nvlist(elem, &propval) != 0)
	continue;

	strval = fnvlist_lookup_string(propval, ZPROP_VALUE);

	entry = zfs_alloc(zhp->zpool_hdl,
	sizeof (zprop_list_t));
	entry->pl_prop = prop;
	entry->pl_user_prop = zfs_strdup(zhp->zpool_hdl,
	propname);
	entry->pl_width = strlen(strval);
	entry->pl_all = B_TRUE;
	*last = entry;
	last = &entry->pl_next;
	}
	}

	return (0);
	}

	/*
	* Get the state for the given feature on the given ZFS pool.
	*/
	int
	zpool_prop_get_feature(zpool_handle_t zhp, const char propname, char *buf,
	size_t len)
	{
	uint64_t refcount;
	boolean_t found = B_FALSE;
	nvlist_t *features = zpool_get_features(zhp);
	boolean_t supported;
	const char *feature = strchr(propname, '@') + 1;

	supported = zpool_prop_feature(propname);
	ASSERT(supported \|\| zpool_prop_unsupported(propname));

	/*
	* Convert from feature name to feature guid. This conversion is
	* unnecessary for unsupported@... properties because they already
	* use guids.
	*/
	if (supported) {
	int ret;
	spa_feature_t fid;

	ret = zfeature_lookup_name(feature, &fid);
	if (ret != 0) {
	(void) strlcpy(buf, "-", len);
	return (ENOTSUP);
	}
	feature = spa_feature_table[fid].fi_guid;
	}

	if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
	found = B_TRUE;

	if (supported) {
	if (!found) {
	(void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
	} else {
	if (refcount == 0)
	(void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
	else
	(void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
	}
	} else {
	if (found) {
	if (refcount == 0) {
	(void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
	} else {
	(void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
	}
	} else {
	(void) strlcpy(buf, "-", len);
	return (ENOTSUP);
	}
	}

	return (0);
	}

	/*
	* Validate the given pool name, optionally putting an extended error message in
	* 'buf'.
	*/
	boolean_t
	zpool_name_valid(libzfs_handle_t hdl, boolean_t isopen, const char pool)
	{
	namecheck_err_t why;
	char what;
	int ret;

	ret = pool_namecheck(pool, &why, &what);

	/*
	* The rules for reserved pool names were extended at a later point.
	* But we need to support users with existing pools that may now be
	* invalid. So we only check for this expanded set of names during a
	* create (or import), and only in userland.
	*/
	if (ret == 0 && !isopen &&
	(strncmp(pool, "mirror", 6) == 0 \|\|
	strncmp(pool, "raidz", 5) == 0 \|\|
	strncmp(pool, "draid", 5) == 0 \|\|
	strncmp(pool, "spare", 5) == 0 \|\|
	strcmp(pool, "log") == 0)) {
	if (hdl != NULL)
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "name is reserved"));
	return (B_FALSE);
	}


	if (ret != 0) {
	if (hdl != NULL) {
	switch (why) {
	case NAME_ERR_TOOLONG:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "name is too long"));
	break;

	case NAME_ERR_INVALCHAR:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "invalid character "
	"'%c' in pool name"), what);
	break;

	case NAME_ERR_NOLETTER:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"name must begin with a letter"));
	break;

	case NAME_ERR_RESERVED:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"name is reserved"));
	break;

	case NAME_ERR_DISKLIKE:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool name is reserved"));
	break;

	case NAME_ERR_LEADING_SLASH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"leading slash in name"));
	break;

	case NAME_ERR_EMPTY_COMPONENT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"empty component in name"));
	break;

	case NAME_ERR_TRAILING_SLASH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"trailing slash in name"));
	break;

	case NAME_ERR_MULTIPLE_DELIMITERS:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"multiple '@' and/or '#' delimiters in "
	"name"));
	break;

	case NAME_ERR_NO_AT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"permission set is missing '@'"));
	break;

	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"(%d) not defined"), why);
	break;
	}
	}
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	/*
	* Open a handle to the given pool, even if the pool is currently in the FAULTED
	* state.
	*/
	zpool_handle_t *
	zpool_open_canfail(libzfs_handle_t hdl, const char pool)
	{
	zpool_handle_t *zhp;
	boolean_t missing;

	/*
	* Make sure the pool name is valid.
	*/
	if (!zpool_name_valid(hdl, B_TRUE, pool)) {
	(void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
	dgettext(TEXT_DOMAIN, "cannot open '%s'"),
	pool);
	return (NULL);
	}

	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));

	zhp->zpool_hdl = hdl;
	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));

	if (zpool_refresh_stats(zhp, &missing) != 0) {
	zpool_close(zhp);
	return (NULL);
	}

	if (missing) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
	(void) zfs_error_fmt(hdl, EZFS_NOENT,
	dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
	zpool_close(zhp);
	return (NULL);
	}

	return (zhp);
	}

	/*
	* Like the above, but silent on error. Used when iterating over pools (because
	* the configuration cache may be out of date).
	*/
	int
	zpool_open_silent(libzfs_handle_t hdl, const char pool, zpool_handle_t **ret)
	{
	zpool_handle_t *zhp;
	boolean_t missing;

	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));

	zhp->zpool_hdl = hdl;
	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));

	if (zpool_refresh_stats(zhp, &missing) != 0) {
	zpool_close(zhp);
	return (-1);
	}

	if (missing) {
	zpool_close(zhp);
	*ret = NULL;
	return (0);
	}

	*ret = zhp;
	return (0);
	}

	/*
	* Similar to zpool_open_canfail(), but refuses to open pools in the faulted
	* state.
	*/
	zpool_handle_t *
	zpool_open(libzfs_handle_t hdl, const char pool)
	{
	zpool_handle_t *zhp;

	if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
	return (NULL);

	if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
	(void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
	dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
	zpool_close(zhp);
	return (NULL);
	}

	return (zhp);
	}

	/*
	* Close the handle. Simply frees the memory associated with the handle.
	*/
	void
	zpool_close(zpool_handle_t *zhp)
	{
	nvlist_free(zhp->zpool_config);
	nvlist_free(zhp->zpool_old_config);
	nvlist_free(zhp->zpool_props);
	free(zhp);
	}

	/*
	* Return the name of the pool.
	*/
	const char *
	zpool_get_name(zpool_handle_t *zhp)
	{
	return (zhp->zpool_name);
	}


	/*
	* Return the state of the pool (ACTIVE or UNAVAILABLE)
	*/
	int
	zpool_get_state(zpool_handle_t *zhp)
	{
	return (zhp->zpool_state);
	}

	/*
	* Check if vdev list contains a special vdev
	*/
	static boolean_t
	zpool_has_special_vdev(nvlist_t *nvroot)
	{
	nvlist_t **child;
	uint_t children;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
	&children) == 0) {
	for (uint_t c = 0; c < children; c++) {
	char *bias;

	if (nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
	strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
	return (B_TRUE);
	}
	}
	}
	return (B_FALSE);
	}

	/*
	* Check if vdev list contains a dRAID vdev
	*/
	static boolean_t
	zpool_has_draid_vdev(nvlist_t *nvroot)
	{
	nvlist_t **child;
	uint_t children;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (uint_t c = 0; c < children; c++) {
	char *type;

	if (nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_TYPE, &type) == 0 &&
	strcmp(type, VDEV_TYPE_DRAID) == 0) {
	return (B_TRUE);
	}
	}
	}
	return (B_FALSE);
	}

	/*
	* Output a dRAID top-level vdev name in to the provided buffer.
	*/
	static char *
	zpool_draid_name(char *name, int len, uint64_t data, uint64_t parity,
	uint64_t spares, uint64_t children)
	{
	snprintf(name, len, "%s%llu:%llud:%lluc:%llus",
	VDEV_TYPE_DRAID, (u_longlong_t)parity, (u_longlong_t)data,
	(u_longlong_t)children, (u_longlong_t)spares);

	return (name);
	}

	/*
	* Return B_TRUE if the provided name is a dRAID spare name.
	*/
	boolean_t
	zpool_is_draid_spare(const char *name)
	{
	uint64_t spare_id, parity, vdev_id;

	if (sscanf(name, VDEV_TYPE_DRAID "%llu-%llu-%llu",
	(u_longlong_t )&parity, (u_longlong_t )&vdev_id,
	(u_longlong_t *)&spare_id) == 3) {
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Create the named pool, using the provided vdev list. It is assumed
	* that the consumer has already validated the contents of the nvlist, so we
	* don't have to worry about error semantics.
	*/
	int
	zpool_create(libzfs_handle_t hdl, const char pool, nvlist_t *nvroot,
	nvlist_t props, nvlist_t fsprops)
	{
	zfs_cmd_t zc = {"\0"};
	nvlist_t *zc_fsprops = NULL;
	nvlist_t *zc_props = NULL;
	nvlist_t *hidden_args = NULL;
	uint8_t *wkeydata = NULL;
	uint_t wkeylen = 0;
	char errbuf[ERRBUFLEN];
	int ret = -1;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot create '%s'"), pool);

	if (!zpool_name_valid(hdl, B_FALSE, pool))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	zcmd_write_conf_nvlist(hdl, &zc, nvroot);

	if (props) {
	prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };

	if ((zc_props = zpool_valid_proplist(hdl, pool, props,
	SPA_VERSION_1, flags, errbuf)) == NULL) {
	goto create_failed;
	}
	}

	if (fsprops) {
	uint64_t zoned;
	char *zonestr;

	zoned = ((nvlist_lookup_string(fsprops,
	zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
	strcmp(zonestr, "on") == 0);

	if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
	fsprops, zoned, NULL, NULL, B_TRUE, errbuf)) == NULL) {
	goto create_failed;
	}

	if (nvlist_exists(zc_fsprops,
	zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
	!zpool_has_special_vdev(nvroot)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"%s property requires a special vdev"),
	zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto create_failed;
	}

	if (!zc_props &&
	(nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
	goto create_failed;
	}
	if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
	&wkeydata, &wkeylen) != 0) {
	zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	goto create_failed;
	}
	if (nvlist_add_nvlist(zc_props,
	ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
	goto create_failed;
	}
	if (wkeydata != NULL) {
	if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
	goto create_failed;

	if (nvlist_add_uint8_array(hidden_args, "wkeydata",
	wkeydata, wkeylen) != 0)
	goto create_failed;

	if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
	hidden_args) != 0)
	goto create_failed;
	}
	}

	if (zc_props)
	zcmd_write_src_nvlist(hdl, &zc, zc_props);

	(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));

	if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {

	zcmd_free_nvlists(&zc);
	nvlist_free(zc_props);
	nvlist_free(zc_fsprops);
	nvlist_free(hidden_args);
	if (wkeydata != NULL)
	free(wkeydata);

	switch (errno) {
	case EBUSY:
	/*
	* This can happen if the user has specified the same
	* device multiple times. We can't reliably detect this
	* until we try to add it and see we already have a
	* label. This can also happen under if the device is
	* part of an active md or lvm device.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more vdevs refer to the same device, or "
	"one of\nthe devices is part of an active md or "
	"lvm device"));
	return (zfs_error(hdl, EZFS_BADDEV, errbuf));

	case ERANGE:
	/*
	* This happens if the record size is smaller or larger
	* than the allowed size range, or not a power of 2.
	*
	* NOTE: although zfs_valid_proplist is called earlier,
	* this case may have slipped through since the
	* pool does not exist yet and it is therefore
	* impossible to read properties e.g. max blocksize
	* from the pool.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"record size invalid"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));

	case EOVERFLOW:
	/*
	* This occurs when one of the devices is below
	* SPA_MINDEVSIZE. Unfortunately, we can't detect which
	* device was the problem device since there's no
	* reliable way to determine device size from userland.
	*/
	{
	char buf[64];

	zfs_nicebytes(SPA_MINDEVSIZE, buf,
	sizeof (buf));

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices is less than the "
	"minimum size (%s)"), buf);
	}
	return (zfs_error(hdl, EZFS_BADDEV, errbuf));

	case ENOSPC:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices is out of space"));
	return (zfs_error(hdl, EZFS_BADDEV, errbuf));

	case EINVAL:
	if (zpool_has_draid_vdev(nvroot) &&
	zfeature_lookup_name("draid", NULL) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dRAID vdevs are unsupported by the "
	"kernel"));
	return (zfs_error(hdl, EZFS_BADDEV, errbuf));
	} else {
	return (zpool_standard_error(hdl, errno,
	errbuf));
	}

	default:
	return (zpool_standard_error(hdl, errno, errbuf));
	}
	}

	create_failed:
	zcmd_free_nvlists(&zc);
	nvlist_free(zc_props);
	nvlist_free(zc_fsprops);
	nvlist_free(hidden_args);
	if (wkeydata != NULL)
	free(wkeydata);
	return (ret);
	}

	/*
	* Destroy the given pool. It is up to the caller to ensure that there are no
	* datasets left in the pool.
	*/
	int
	zpool_destroy(zpool_handle_t zhp, const char log_str)
	{
	zfs_cmd_t zc = {"\0"};
	zfs_handle_t *zfp = NULL;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	char errbuf[ERRBUFLEN];

	if (zhp->zpool_state == POOL_STATE_ACTIVE &&
	(zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
	return (-1);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_history = (uint64_t)(uintptr_t)log_str;

	if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot destroy '%s'"), zhp->zpool_name);

	if (errno == EROFS) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices is read only"));
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	} else {
	(void) zpool_standard_error(hdl, errno, errbuf);
	}

	if (zfp)
	zfs_close(zfp);
	return (-1);
	}

	if (zfp) {
	remove_mountpoint(zfp);
	zfs_close(zfp);
	}

	return (0);
	}

	/*
	* Create a checkpoint in the given pool.
	*/
	int
	zpool_checkpoint(zpool_handle_t *zhp)
	{
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	char errbuf[ERRBUFLEN];
	int error;

	error = lzc_pool_checkpoint(zhp->zpool_name);
	if (error != 0) {
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot checkpoint '%s'"), zhp->zpool_name);
	(void) zpool_standard_error(hdl, error, errbuf);
	return (-1);
	}

	return (0);
	}

	/*
	* Discard the checkpoint from the given pool.
	*/
	int
	zpool_discard_checkpoint(zpool_handle_t *zhp)
	{
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	char errbuf[ERRBUFLEN];
	int error;

	error = lzc_pool_checkpoint_discard(zhp->zpool_name);
	if (error != 0) {
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot discard checkpoint in '%s'"), zhp->zpool_name);
	(void) zpool_standard_error(hdl, error, errbuf);
	return (-1);
	}

	return (0);
	}

	/*
	* Add the given vdevs to the pool. The caller must have already performed the
	* necessary verification to ensure that the vdev specification is well-formed.
	*/
	int
	zpool_add(zpool_handle_t zhp, nvlist_t nvroot)
	{
	zfs_cmd_t zc = {"\0"};
	int ret;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	char errbuf[ERRBUFLEN];
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot add to '%s'"), zhp->zpool_name);

	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
	SPA_VERSION_SPARES &&
	nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
	"upgraded to add hot spares"));
	return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
	}

	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
	SPA_VERSION_L2CACHE &&
	nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
	"upgraded to add cache devices"));
	return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
	}

	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
	switch (errno) {
	case EBUSY:
	/*
	* This can happen if the user has specified the same
	* device multiple times. We can't reliably detect this
	* until we try to add it and see we already have a
	* label.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more vdevs refer to the same device"));
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case EINVAL:

	if (zpool_has_draid_vdev(nvroot) &&
	zfeature_lookup_name("draid", NULL) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dRAID vdevs are unsupported by the "
	"kernel"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid config; a pool with removing/"
	"removed vdevs does not support adding "
	"raidz or dRAID vdevs"));
	}

	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case EOVERFLOW:
	/*
	* This occurs when one of the devices is below
	* SPA_MINDEVSIZE. Unfortunately, we can't detect which
	* device was the problem device since there's no
	* reliable way to determine device size from userland.
	*/
	{
	char buf[64];

	zfs_nicebytes(SPA_MINDEVSIZE, buf,
	sizeof (buf));

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"device is less than the minimum "
	"size (%s)"), buf);
	}
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to add these vdevs"));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;

	default:
	(void) zpool_standard_error(hdl, errno, errbuf);
	}

	ret = -1;
	} else {
	ret = 0;
	}

	zcmd_free_nvlists(&zc);

	return (ret);
	}

	/*
	* Exports the pool from the system. The caller must ensure that there are no
	* mounted datasets in the pool.
	*/
	static int
	zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
	const char *log_str)
	{
	zfs_cmd_t zc = {"\0"};

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_cookie = force;
	zc.zc_guid = hardforce;
	zc.zc_history = (uint64_t)(uintptr_t)log_str;

	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
	switch (errno) {
	case EXDEV:
	zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
	"use '-f' to override the following errors:\n"
	"'%s' has an active shared spare which could be"
	" used by other pools once '%s' is exported."),
	zhp->zpool_name, zhp->zpool_name);
	return (zfs_error_fmt(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
	dgettext(TEXT_DOMAIN, "cannot export '%s'"),
	zhp->zpool_name));
	default:
	return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot export '%s'"),
	zhp->zpool_name));
	}
	}

	return (0);
	}

	int
	zpool_export(zpool_handle_t zhp, boolean_t force, const char log_str)
	{
	return (zpool_export_common(zhp, force, B_FALSE, log_str));
	}

	int
	zpool_export_force(zpool_handle_t zhp, const char log_str)
	{
	return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
	}

	static void
	zpool_rewind_exclaim(libzfs_handle_t hdl, const char name, boolean_t dryrun,
	nvlist_t *config)
	{
	nvlist_t *nv = NULL;
	uint64_t rewindto;
	int64_t loss = -1;
	struct tm t;
	char timestr[128];

	if (!hdl->libzfs_printerr \|\| config == NULL)
	return;

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 \|\|
	nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
	return;
	}

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
	return;
	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);

	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
	strftime(timestr, 128, "%c", &t) != 0) {
	if (dryrun) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"Would be able to return %s "
	"to its state as of %s.\n"),
	name, timestr);
	} else {
	(void) printf(dgettext(TEXT_DOMAIN,
	"Pool %s returned to its state as of %s.\n"),
	name, timestr);
	}
	if (loss > 120) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"%s approximately %lld "),
	dryrun ? "Would discard" : "Discarded",
	((longlong_t)loss + 30) / 60);
	(void) printf(dgettext(TEXT_DOMAIN,
	"minutes of transactions.\n"));
	} else if (loss > 0) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"%s approximately %lld "),
	dryrun ? "Would discard" : "Discarded",
	(longlong_t)loss);
	(void) printf(dgettext(TEXT_DOMAIN,
	"seconds of transactions.\n"));
	}
	}
	}

	void
	zpool_explain_recover(libzfs_handle_t hdl, const char name, int reason,
	nvlist_t *config)
	{
	nvlist_t *nv = NULL;
	int64_t loss = -1;
	uint64_t edata = UINT64_MAX;
	uint64_t rewindto;
	struct tm t;
	char timestr[128];

	if (!hdl->libzfs_printerr)
	return;

	if (reason >= 0)
	(void) printf(dgettext(TEXT_DOMAIN, "action: "));
	else
	(void) printf(dgettext(TEXT_DOMAIN, "\t"));

	/* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 \|\|
	nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 \|\|
	nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
	goto no_info;

	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
	&edata);

	(void) printf(dgettext(TEXT_DOMAIN,
	"Recovery is possible, but will result in some data loss.\n"));

	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
	strftime(timestr, 128, "%c", &t) != 0) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"\tReturning the pool to its state as of %s\n"
	"\tshould correct the problem. "),
	timestr);
	} else {
	(void) printf(dgettext(TEXT_DOMAIN,
	"\tReverting the pool to an earlier state "
	"should correct the problem.\n\t"));
	}

	if (loss > 120) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"Approximately %lld minutes of data\n"
	"\tmust be discarded, irreversibly. "),
	((longlong_t)loss + 30) / 60);
	} else if (loss > 0) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"Approximately %lld seconds of data\n"
	"\tmust be discarded, irreversibly. "),
	(longlong_t)loss);
	}
	if (edata != 0 && edata != UINT64_MAX) {
	if (edata == 1) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"After rewind, at least\n"
	"\tone persistent user-data error will remain. "));
	} else {
	(void) printf(dgettext(TEXT_DOMAIN,
	"After rewind, several\n"
	"\tpersistent user-data errors will remain. "));
	}
	}
	(void) printf(dgettext(TEXT_DOMAIN,
	"Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
	reason >= 0 ? "clear" : "import", name);

	(void) printf(dgettext(TEXT_DOMAIN,
	"A scrub of the pool\n"
	"\tis strongly recommended after recovery.\n"));
	return;

	no_info:
	(void) printf(dgettext(TEXT_DOMAIN,
	"Destroy and re-create the pool from\n\ta backup source.\n"));
	}

	/*
	* zpool_import() is a contracted interface. Should be kept the same
	* if possible.
	*
	* Applications should use zpool_import_props() to import a pool with
	* new properties value to be set.
	*/
	int
	zpool_import(libzfs_handle_t hdl, nvlist_t config, const char *newname,
	char *altroot)
	{
	nvlist_t *props = NULL;
	int ret;

	if (altroot != NULL) {
	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
	return (zfs_error_fmt(hdl, EZFS_NOMEM,
	dgettext(TEXT_DOMAIN, "cannot import '%s'"),
	newname));
	}

	if (nvlist_add_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 \|\|
	nvlist_add_string(props,
	zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
	nvlist_free(props);
	return (zfs_error_fmt(hdl, EZFS_NOMEM,
	dgettext(TEXT_DOMAIN, "cannot import '%s'"),
	newname));
	}
	}

	ret = zpool_import_props(hdl, config, newname, props,
	ZFS_IMPORT_NORMAL);
	nvlist_free(props);
	return (ret);
	}

	static void
	print_vdev_tree(libzfs_handle_t hdl, const char name, nvlist_t *nv,
	int indent)
	{
	nvlist_t **child;
	uint_t c, children;
	char *vname;
	uint64_t is_log = 0;

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
	&is_log);

	if (name != NULL)
	(void) printf("\t%*s%s%s\n", indent, "", name,
	is_log ? " [log]" : "");

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return;

	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
	print_vdev_tree(hdl, vname, child[c], indent + 2);
	free(vname);
	}
	}

	void
	zpool_print_unsup_feat(nvlist_t *config)
	{
	nvlist_t nvinfo, unsup_feat;

	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	unsup_feat = fnvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT);

	for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL);
	nvp != NULL; nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
	char *desc = fnvpair_value_string(nvp);
	if (strlen(desc) > 0)
	(void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
	else
	(void) printf("\t%s\n", nvpair_name(nvp));
	}
	}

	/*
	* Import the given pool using the known configuration and a list of
	* properties to be set. The configuration should have come from
	* zpool_find_import(). The 'newname' parameters control whether the pool
	* is imported with a different name.
	*/
	int
	zpool_import_props(libzfs_handle_t hdl, nvlist_t config, const char *newname,
	nvlist_t *props, int flags)
	{
	zfs_cmd_t zc = {"\0"};
	zpool_load_policy_t policy;
	nvlist_t *nv = NULL;
	nvlist_t *nvinfo = NULL;
	nvlist_t *missing = NULL;
	const char *thename;
	char *origname;
	int ret;
	int error = 0;
	char errbuf[ERRBUFLEN];

	origname = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot import pool '%s'"), origname);

	if (newname != NULL) {
	if (!zpool_name_valid(hdl, B_FALSE, newname))
	return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
	dgettext(TEXT_DOMAIN, "cannot import '%s'"),
	newname));
	thename = newname;
	} else {
	thename = origname;
	}

	if (props != NULL) {
	uint64_t version;
	prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };

	version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);

	if ((props = zpool_valid_proplist(hdl, origname,
	props, version, flags, errbuf)) == NULL)
	return (-1);
	zcmd_write_src_nvlist(hdl, &zc, props);
	nvlist_free(props);
	}

	(void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));

	zc.zc_guid = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID);

	zcmd_write_conf_nvlist(hdl, &zc, config);
	zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2);

	zc.zc_cookie = flags;
	while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
	errno == ENOMEM)
	zcmd_expand_dst_nvlist(hdl, &zc);
	if (ret != 0)
	error = errno;

	(void) zcmd_read_dst_nvlist(hdl, &zc, &nv);

	zcmd_free_nvlists(&zc);

	zpool_get_load_policy(config, &policy);

	if (error) {
	char desc[1024];
	char aux[256];

	/*
	* Dry-run failed, but we print out what success
	* looks like if we found a best txg
	*/
	if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
	zpool_rewind_exclaim(hdl, newname ? origname : thename,
	B_TRUE, nv);
	nvlist_free(nv);
	return (-1);
	}

	if (newname == NULL)
	(void) snprintf(desc, sizeof (desc),
	dgettext(TEXT_DOMAIN, "cannot import '%s'"),
	thename);
	else
	(void) snprintf(desc, sizeof (desc),
	dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
	origname, thename);

	switch (error) {
	case ENOTSUP:
	if (nv != NULL && nvlist_lookup_nvlist(nv,
	ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
	nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
	(void) printf(dgettext(TEXT_DOMAIN, "This "
	"pool uses the following feature(s) not "
	"supported by this system:\n"));
	zpool_print_unsup_feat(nv);
	if (nvlist_exists(nvinfo,
	ZPOOL_CONFIG_CAN_RDONLY)) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"All unsupported features are only "
	"required for writing to the pool."
	"\nThe pool can be imported using "
	"'-o readonly=on'.\n"));
	}
	}
	/*
	* Unsupported version.
	*/
	(void) zfs_error(hdl, EZFS_BADVERSION, desc);
	break;

	case EREMOTEIO:
	if (nv != NULL && nvlist_lookup_nvlist(nv,
	ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
	const char *hostname = "<unknown>";
	uint64_t hostid = 0;
	mmp_state_t mmp_state;

	mmp_state = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_STATE);

	if (nvlist_exists(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME);

	if (nvlist_exists(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID);

	if (mmp_state == MMP_STATE_ACTIVE) {
	(void) snprintf(aux, sizeof (aux),
	dgettext(TEXT_DOMAIN, "pool is imp"
	"orted on host '%s' (hostid=%lx).\n"
	"Export the pool on the other "
	"system, then run 'zpool import'."),
	hostname, (unsigned long) hostid);
	} else if (mmp_state == MMP_STATE_NO_HOSTID) {
	(void) snprintf(aux, sizeof (aux),
	dgettext(TEXT_DOMAIN, "pool has "
	"the multihost property on and "
	"the\nsystem's hostid is not set. "
	"Set a unique system hostid with "
	"the zgenhostid(8) command.\n"));
	}

	(void) zfs_error_aux(hdl, "%s", aux);
	}
	(void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
	break;

	case EINVAL:
	(void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
	break;

	case EROFS:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices is read only"));
	(void) zfs_error(hdl, EZFS_BADDEV, desc);
	break;

	case ENXIO:
	if (nv && nvlist_lookup_nvlist(nv,
	ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
	nvlist_lookup_nvlist(nvinfo,
	ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
	(void) printf(dgettext(TEXT_DOMAIN,
	"The devices below are missing or "
	"corrupted, use '-m' to import the pool "
	"anyway:\n"));
	print_vdev_tree(hdl, NULL, missing, 2);
	(void) printf("\n");
	}
	(void) zpool_standard_error(hdl, error, desc);
	break;

	case EEXIST:
	(void) zpool_standard_error(hdl, error, desc);
	break;

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices are already in use\n"));
	(void) zfs_error(hdl, EZFS_BADDEV, desc);
	break;
	case ENAMETOOLONG:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"new name of at least one dataset is longer than "
	"the maximum allowable length"));
	(void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
	break;
	default:
	(void) zpool_standard_error(hdl, error, desc);
	zpool_explain_recover(hdl,
	newname ? origname : thename, -error, nv);
	break;
	}

	nvlist_free(nv);
	ret = -1;
	} else {
	zpool_handle_t *zhp;

	/*
	* This should never fail, but play it safe anyway.
	*/
	if (zpool_open_silent(hdl, thename, &zhp) != 0)
	ret = -1;
	else if (zhp != NULL)
	zpool_close(zhp);
	if (policy.zlp_rewind &
	(ZPOOL_DO_REWIND \| ZPOOL_TRY_REWIND)) {
	zpool_rewind_exclaim(hdl, newname ? origname : thename,
	((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
	}
	nvlist_free(nv);
	return (0);
	}

	return (ret);
	}

	/*
	* Translate vdev names to guids. If a vdev_path is determined to be
	* unsuitable then a vd_errlist is allocated and the vdev path and errno
	* are added to it.
	*/
	static int
	zpool_translate_vdev_guids(zpool_handle_t zhp, nvlist_t vds,
	nvlist_t vdev_guids, nvlist_t guids_to_paths, nvlist_t **vd_errlist)
	{
	nvlist_t *errlist = NULL;
	int error = 0;

	for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
	elem = nvlist_next_nvpair(vds, elem)) {
	boolean_t spare, cache;

	char *vd_path = nvpair_name(elem);
	nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
	NULL);

	if ((tgt == NULL) \|\| cache \|\| spare) {
	if (errlist == NULL) {
	errlist = fnvlist_alloc();
	error = EINVAL;
	}

	uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
	(spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
	fnvlist_add_int64(errlist, vd_path, err);
	continue;
	}

	uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
	fnvlist_add_uint64(vdev_guids, vd_path, guid);

	char msg[MAXNAMELEN];
	(void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
	fnvlist_add_string(guids_to_paths, msg, vd_path);
	}

	if (error != 0) {
	verify(errlist != NULL);
	if (vd_errlist != NULL)
	*vd_errlist = errlist;
	else
	fnvlist_free(errlist);
	}

	return (error);
	}

	static int
	xlate_init_err(int err)
	{
	switch (err) {
	case ENODEV:
	return (EZFS_NODEVICE);
	case EINVAL:
	case EROFS:
	return (EZFS_BADDEV);
	case EBUSY:
	return (EZFS_INITIALIZING);
	case ESRCH:
	return (EZFS_NO_INITIALIZE);
	}
	return (err);
	}

	/*
	* Begin, suspend, or cancel the initialization (initializing of all free
	* blocks) for the given vdevs in the given pool.
	*/
	static int
	zpool_initialize_impl(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
	nvlist_t *vds, boolean_t wait)
	{
	int err;

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *guids_to_paths = fnvlist_alloc();
	nvlist_t *vd_errlist = NULL;
	nvlist_t *errlist;
	nvpair_t *elem;

	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
	guids_to_paths, &vd_errlist);

	if (err != 0) {
	verify(vd_errlist != NULL);
	goto list_errors;
	}

	err = lzc_initialize(zhp->zpool_name, cmd_type,
	vdev_guids, &errlist);

	if (err != 0) {
	if (errlist != NULL) {
	vd_errlist = fnvlist_lookup_nvlist(errlist,
	ZPOOL_INITIALIZE_VDEVS);
	goto list_errors;
	}
	(void) zpool_standard_error(zhp->zpool_hdl, err,
	dgettext(TEXT_DOMAIN, "operation failed"));
	goto out;
	}

	if (wait) {
	for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
	elem = nvlist_next_nvpair(vdev_guids, elem)) {

	uint64_t guid = fnvpair_value_uint64(elem);

	err = lzc_wait_tag(zhp->zpool_name,
	ZPOOL_WAIT_INITIALIZE, guid, NULL);
	if (err != 0) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl,
	err, dgettext(TEXT_DOMAIN, "error "
	"waiting for '%s' to initialize"),
	nvpair_name(elem));

	goto out;
	}
	}
	}
	goto out;

	list_errors:
	for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
	elem = nvlist_next_nvpair(vd_errlist, elem)) {
	int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
	char *path;

	if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
	&path) != 0)
	path = nvpair_name(elem);

	(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
	"cannot initialize '%s'", path);
	}

	out:
	fnvlist_free(vdev_guids);
	fnvlist_free(guids_to_paths);

	if (vd_errlist != NULL)
	fnvlist_free(vd_errlist);

	return (err == 0 ? 0 : -1);
	}

	int
	zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
	nvlist_t *vds)
	{
	return (zpool_initialize_impl(zhp, cmd_type, vds, B_FALSE));
	}

	int
	zpool_initialize_wait(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
	nvlist_t *vds)
	{
	return (zpool_initialize_impl(zhp, cmd_type, vds, B_TRUE));
	}

	static int
	xlate_trim_err(int err)
	{
	switch (err) {
	case ENODEV:
	return (EZFS_NODEVICE);
	case EINVAL:
	case EROFS:
	return (EZFS_BADDEV);
	case EBUSY:
	return (EZFS_TRIMMING);
	case ESRCH:
	return (EZFS_NO_TRIM);
	case EOPNOTSUPP:
	return (EZFS_TRIM_NOTSUP);
	}
	return (err);
	}

	static int
	zpool_trim_wait(zpool_handle_t zhp, nvlist_t vdev_guids)
	{
	int err;
	nvpair_t *elem;

	for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
	elem = nvlist_next_nvpair(vdev_guids, elem)) {

	uint64_t guid = fnvpair_value_uint64(elem);

	err = lzc_wait_tag(zhp->zpool_name,
	ZPOOL_WAIT_TRIM, guid, NULL);
	if (err != 0) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl,
	err, dgettext(TEXT_DOMAIN, "error "
	"waiting to trim '%s'"), nvpair_name(elem));

	return (err);
	}
	}
	return (0);
	}

	/*
	* Check errlist and report any errors, omitting ones which should be
	* suppressed. Returns B_TRUE if any errors were reported.
	*/
	static boolean_t
	check_trim_errs(zpool_handle_t zhp, trimflags_t trim_flags,
	nvlist_t guids_to_paths, nvlist_t vds, nvlist_t *errlist)
	{
	nvpair_t *elem;
	boolean_t reported_errs = B_FALSE;
	int num_vds = 0;
	int num_suppressed_errs = 0;

	for (elem = nvlist_next_nvpair(vds, NULL);
	elem != NULL; elem = nvlist_next_nvpair(vds, elem)) {
	num_vds++;
	}

	for (elem = nvlist_next_nvpair(errlist, NULL);
	elem != NULL; elem = nvlist_next_nvpair(errlist, elem)) {
	int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
	char *path;

	/*
	* If only the pool was specified, and it was not a secure
	* trim then suppress warnings for individual vdevs which
	* do not support trimming.
	*/
	if (vd_error == EZFS_TRIM_NOTSUP &&
	trim_flags->fullpool &&
	!trim_flags->secure) {
	num_suppressed_errs++;
	continue;
	}

	reported_errs = B_TRUE;
	if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
	&path) != 0)
	path = nvpair_name(elem);

	(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
	"cannot trim '%s'", path);
	}

	if (num_suppressed_errs == num_vds) {
	(void) zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
	"no devices in pool support trim operations"));
	(void) (zfs_error(zhp->zpool_hdl, EZFS_TRIM_NOTSUP,
	dgettext(TEXT_DOMAIN, "cannot trim")));
	reported_errs = B_TRUE;
	}

	return (reported_errs);
	}

	/*
	* Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
	* the given vdevs in the given pool.
	*/
	int
	zpool_trim(zpool_handle_t zhp, pool_trim_func_t cmd_type, nvlist_t vds,
	trimflags_t *trim_flags)
	{
	int err;
	int retval = 0;

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *guids_to_paths = fnvlist_alloc();
	nvlist_t *errlist = NULL;

	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
	guids_to_paths, &errlist);
	if (err != 0) {
	check_trim_errs(zhp, trim_flags, guids_to_paths, vds, errlist);
	retval = -1;
	goto out;
	}

	err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
	trim_flags->secure, vdev_guids, &errlist);
	if (err != 0) {
	nvlist_t *vd_errlist;
	if (errlist != NULL && nvlist_lookup_nvlist(errlist,
	ZPOOL_TRIM_VDEVS, &vd_errlist) == 0) {
	if (check_trim_errs(zhp, trim_flags, guids_to_paths,
	vds, vd_errlist)) {
	retval = -1;
	goto out;
	}
	} else {
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "operation failed"));
	zpool_standard_error(zhp->zpool_hdl, err, errbuf);
	retval = -1;
	goto out;
	}
	}


	if (trim_flags->wait)
	retval = zpool_trim_wait(zhp, vdev_guids);

	out:
	if (errlist != NULL)
	fnvlist_free(errlist);
	fnvlist_free(vdev_guids);
	fnvlist_free(guids_to_paths);
	return (retval);
	}

	/*
	* Scan the pool.
	*/
	int
	zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	int err;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_cookie = func;
	zc.zc_flags = cmd;

	if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
	return (0);

	err = errno;

	/* ECANCELED on a scrub means we resumed a paused scrub */
	if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
	cmd == POOL_SCRUB_NORMAL)
	return (0);

	if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
	return (0);

	if (func == POOL_SCAN_SCRUB) {
	if (cmd == POOL_SCRUB_PAUSE) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot pause scrubbing %s"),
	zc.zc_name);
	} else {
	assert(cmd == POOL_SCRUB_NORMAL);
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot scrub %s"),
	zc.zc_name);
	}
	} else if (func == POOL_SCAN_RESILVER) {
	assert(cmd == POOL_SCRUB_NORMAL);
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot restart resilver on %s"), zc.zc_name);
	} else if (func == POOL_SCAN_NONE) {
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot cancel scrubbing %s"), zc.zc_name);
	} else {
	assert(!"unexpected result");
	}

	if (err == EBUSY) {
	nvlist_t *nvroot;
	pool_scan_stat_t *ps = NULL;
	uint_t psc;

	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
	ZPOOL_CONFIG_VDEV_TREE);
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
	if (ps && ps->pss_func == POOL_SCAN_SCRUB &&
	ps->pss_state == DSS_SCANNING) {
	if (cmd == POOL_SCRUB_PAUSE)
	return (zfs_error(hdl, EZFS_SCRUB_PAUSED,
	errbuf));
	else
	return (zfs_error(hdl, EZFS_SCRUBBING, errbuf));
	} else {
	return (zfs_error(hdl, EZFS_RESILVERING, errbuf));
	}
	} else if (err == ENOENT) {
	return (zfs_error(hdl, EZFS_NO_SCRUB, errbuf));
	} else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
	return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, errbuf));
	} else {
	return (zpool_standard_error(hdl, err, errbuf));
	}
	}

	/*
	* Find a vdev that matches the search criteria specified. We use the
	* the nvpair name to determine how we should look for the device.
	* 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
	* spare; but FALSE if its an INUSE spare.
	*/
	static nvlist_t *
	vdev_to_nvlist_iter(nvlist_t nv, nvlist_t search, boolean_t *avail_spare,
	boolean_t l2cache, boolean_t log)
	{
	uint_t c, children;
	nvlist_t **child;
	nvlist_t *ret;
	uint64_t is_log;
	char *srchkey;
	nvpair_t *pair = nvlist_next_nvpair(search, NULL);

	/* Nothing to look for */
	if (search == NULL \|\| pair == NULL)
	return (NULL);

	/* Obtain the key we will use to search */
	srchkey = nvpair_name(pair);

	switch (nvpair_type(pair)) {
	case DATA_TYPE_UINT64:
	if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
	uint64_t srchval = fnvpair_value_uint64(pair);
	uint64_t theguid = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_GUID);
	if (theguid == srchval)
	return (nv);
	}
	break;

	case DATA_TYPE_STRING: {
	char srchval, val;

	srchval = fnvpair_value_string(pair);
	if (nvlist_lookup_string(nv, srchkey, &val) != 0)
	break;

	/*
	* Search for the requested value. Special cases:
	*
	* - ZPOOL_CONFIG_PATH for whole disk entries. These end in
	* "-part1", or "p1". The suffix is hidden from the user,
	* but included in the string, so this matches around it.
	* - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
	* is used to check all possible expanded paths.
	* - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
	*
	* Otherwise, all other searches are simple string compares.
	*/
	if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
	uint64_t wholedisk = 0;

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
	&wholedisk);
	if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
	return (nv);

	} else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
	char type, idx, end, p;
	uint64_t id, vdev_id;

	/*
	* Determine our vdev type, keeping in mind
	* that the srchval is composed of a type and
	* vdev id pair (i.e. mirror-4).
	*/
	if ((type = strdup(srchval)) == NULL)
	return (NULL);

	if ((p = strrchr(type, '-')) == NULL) {
	free(type);
	break;
	}
	idx = p + 1;
	*p = '\0';

	/*
	* If the types don't match then keep looking.
	*/
	if (strncmp(val, type, strlen(val)) != 0) {
	free(type);
	break;
	}

	verify(zpool_vdev_is_interior(type));

	id = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID);
	errno = 0;
	vdev_id = strtoull(idx, &end, 10);

	/*
	* If we are looking for a raidz and a parity is
	* specified, make sure it matches.
	*/
	int rzlen = strlen(VDEV_TYPE_RAIDZ);
	assert(rzlen == strlen(VDEV_TYPE_DRAID));
	int typlen = strlen(type);
	if ((strncmp(type, VDEV_TYPE_RAIDZ, rzlen) == 0 \|\|
	strncmp(type, VDEV_TYPE_DRAID, rzlen) == 0) &&
	typlen != rzlen) {
	uint64_t vdev_parity;
	int parity = *(type + rzlen) - '0';

	if (parity <= 0 \|\| parity > 3 \|\|
	(typlen - rzlen) != 1) {
	/*
	* Nonsense parity specified, can
	* never match
	*/
	free(type);
	return (NULL);
	}
	vdev_parity = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_NPARITY);
	if ((int)vdev_parity != parity) {
	free(type);
	break;
	}
	}

	free(type);
	if (errno != 0)
	return (NULL);

	/*
	* Now verify that we have the correct vdev id.
	*/
	if (vdev_id == id)
	return (nv);
	}

	/*
	* Common case
	*/
	if (strcmp(srchval, val) == 0)
	return (nv);
	break;
	}

	default:
	break;
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return (NULL);

	for (c = 0; c < children; c++) {
	if ((ret = vdev_to_nvlist_iter(child[c], search,
	avail_spare, l2cache, NULL)) != NULL) {
	/*
	* The 'is_log' value is only set for the toplevel
	* vdev, not the leaf vdevs. So we always lookup the
	* log device from the root of the vdev tree (where
	* 'log' is non-NULL).
	*/
	if (log != NULL &&
	nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
	is_log) {
	*log = B_TRUE;
	}
	return (ret);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	if ((ret = vdev_to_nvlist_iter(child[c], search,
	avail_spare, l2cache, NULL)) != NULL) {
	*avail_spare = B_TRUE;
	return (ret);
	}
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	if ((ret = vdev_to_nvlist_iter(child[c], search,
	avail_spare, l2cache, NULL)) != NULL) {
	*l2cache = B_TRUE;
	return (ret);
	}
	}
	}

	return (NULL);
	}

	/*
	* Given a physical path or guid, find the associated vdev.
	*/
	nvlist_t *
	zpool_find_vdev_by_physpath(zpool_handle_t zhp, const char ppath,
	boolean_t avail_spare, boolean_t l2cache, boolean_t *log)
	{
	nvlist_t search, nvroot, *ret;
	uint64_t guid;
	char *end;

	search = fnvlist_alloc();

	guid = strtoull(ppath, &end, 0);
	if (guid != 0 && *end == '\0') {
	fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
	} else {
	fnvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath);
	}

	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
	ZPOOL_CONFIG_VDEV_TREE);

	*avail_spare = B_FALSE;
	*l2cache = B_FALSE;
	if (log != NULL)
	*log = B_FALSE;
	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
	fnvlist_free(search);

	return (ret);
	}

	/*
	* Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
	*/
	static boolean_t
	zpool_vdev_is_interior(const char *name)
	{
	if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 \|\|
	strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 \|\|
	strncmp(name,
	VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 \|\|
	strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
	return (B_TRUE);

	if (strncmp(name, VDEV_TYPE_DRAID, strlen(VDEV_TYPE_DRAID)) == 0 &&
	!zpool_is_draid_spare(name))
	return (B_TRUE);

	return (B_FALSE);
	}

	nvlist_t *
	zpool_find_vdev(zpool_handle_t zhp, const char path, boolean_t *avail_spare,
	boolean_t l2cache, boolean_t log)
	{
	char *end;
	nvlist_t nvroot, search, *ret;
	uint64_t guid;

	search = fnvlist_alloc();

	guid = strtoull(path, &end, 0);
	if (guid != 0 && *end == '\0') {
	fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
	} else if (zpool_vdev_is_interior(path)) {
	fnvlist_add_string(search, ZPOOL_CONFIG_TYPE, path);
	} else {
	fnvlist_add_string(search, ZPOOL_CONFIG_PATH, path);
	}

	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
	ZPOOL_CONFIG_VDEV_TREE);

	*avail_spare = B_FALSE;
	*l2cache = B_FALSE;
	if (log != NULL)
	*log = B_FALSE;
	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
	fnvlist_free(search);

	return (ret);
	}

	/*
	* Convert a vdev path to a GUID. Returns GUID or 0 on error.
	*
	* If is_spare, is_l2cache, or is_log is non-NULL, then store within it
	* if the VDEV is a spare, l2cache, or log device. If they're NULL then
	* ignore them.
	*/
	static uint64_t
	zpool_vdev_path_to_guid_impl(zpool_handle_t zhp, const char path,
	boolean_t is_spare, boolean_t is_l2cache, boolean_t *is_log)
	{
	boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
	nvlist_t *tgt;

	if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
	&log)) == NULL)
	return (0);

	if (is_spare != NULL)
	*is_spare = spare;
	if (is_l2cache != NULL)
	*is_l2cache = l2cache;
	if (is_log != NULL)
	*is_log = log;

	return (fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID));
	}

	/* Convert a vdev path to a GUID. Returns GUID or 0 on error. */
	uint64_t
	zpool_vdev_path_to_guid(zpool_handle_t zhp, const char path)
	{
	return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
	}

	/*
	* Bring the specified vdev online. The 'flags' parameter is a set of the
	* ZFS_ONLINE_* flags.
	*/
	int
	zpool_vdev_online(zpool_handle_t zhp, const char path, int flags,
	vdev_state_t *newstate)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache, islog;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	if (flags & ZFS_ONLINE_EXPAND) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
	} else {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot online %s"), path);
	}

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	&islog)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);

	if (avail_spare)
	return (zfs_error(hdl, EZFS_ISSPARE, errbuf));

	#ifndef __FreeBSD__
	char *pathname;
	if ((flags & ZFS_ONLINE_EXPAND \|\|
	zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
	nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
	uint64_t wholedisk = 0;

	(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
	&wholedisk);

	/*
	* XXX - L2ARC 1.0 devices can't support expansion.
	*/
	if (l2cache) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot expand cache devices"));
	return (zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf));
	}

	if (wholedisk) {
	const char *fullpath = path;
	char buf[MAXPATHLEN];
	int error;

	if (path[0] != '/') {
	error = zfs_resolve_shortname(path, buf,
	sizeof (buf));
	if (error != 0)
	return (zfs_error(hdl, EZFS_NODEVICE,
	errbuf));

	fullpath = buf;
	}

	error = zpool_relabel_disk(hdl, fullpath, errbuf);
	if (error != 0)
	return (error);
	}
	}
	#endif

	zc.zc_cookie = VDEV_STATE_ONLINE;
	zc.zc_obj = flags;

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
	if (errno == EINVAL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
	"from this pool into a new one. Use '%s' "
	"instead"), "zpool detach");
	return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, errbuf));
	}
	return (zpool_standard_error(hdl, errno, errbuf));
	}

	*newstate = zc.zc_cookie;
	return (0);
	}

	/*
	* Take the specified vdev offline
	*/
	int
	zpool_vdev_offline(zpool_handle_t zhp, const char path, boolean_t istmp)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot offline %s"), path);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	NULL)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);

	if (avail_spare)
	return (zfs_error(hdl, EZFS_ISSPARE, errbuf));

	zc.zc_cookie = VDEV_STATE_OFFLINE;
	zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
	return (0);

	switch (errno) {
	case EBUSY:

	/*
	* There are no other replicas of this device.
	*/
	return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));

	case EEXIST:
	/*
	* The log device has unplayed logs
	*/
	return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, errbuf));

	default:
	return (zpool_standard_error(hdl, errno, errbuf));
	}
	}

	+/*
	+ * Remove the specified vdev asynchronously from the configuration, so
	+ * that it may come ONLINE if reinserted. This is called from zed on
	+ * Udev remove event.
	+ * Note: We also have a similar function zpool_vdev_remove() that
	+ * removes the vdev from the pool.
	+ */
	+int
	+zpool_vdev_remove_wanted(zpool_handle_t zhp, const char path)
	+{
	+ zfs_cmd_t zc = {"\0"};
	+ char errbuf[ERRBUFLEN];
	+ nvlist_t *tgt;
	+ boolean_t avail_spare, l2cache;
	+ libzfs_handle_t *hdl = zhp->zpool_hdl;
	+
	+ (void) snprintf(errbuf, sizeof (errbuf),
	+ dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
	+
	+ (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	+ if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	+ NULL)) == NULL)
	+ return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
	+
	+ zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
	+
	+ if (avail_spare)
	+ return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
	+
	+ zc.zc_cookie = VDEV_STATE_REMOVED;
	+
	+ if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
	+ return (0);
	+
	+ return (zpool_standard_error(hdl, errno, errbuf));
	+}
	+
	/*
	* Mark the given vdev faulted.
	*/
	int
	zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_guid = guid;
	zc.zc_cookie = VDEV_STATE_FAULTED;
	zc.zc_obj = aux;

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
	return (0);

	switch (errno) {
	case EBUSY:

	/*
	* There are no other replicas of this device.
	*/
	return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));

	default:
	return (zpool_standard_error(hdl, errno, errbuf));
	}

	}

	/*
	* Mark the given vdev degraded.
	*/
	int
	zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_guid = guid;
	zc.zc_cookie = VDEV_STATE_DEGRADED;
	zc.zc_obj = aux;

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
	return (0);

	return (zpool_standard_error(hdl, errno, errbuf));
	}

	/*
	* Returns TRUE if the given nvlist is a vdev that was originally swapped in as
	* a hot spare.
	*/
	static boolean_t
	is_replacing_spare(nvlist_t search, nvlist_t tgt, int which)
	{
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
	&children) == 0) {
	char *type = fnvlist_lookup_string(search, ZPOOL_CONFIG_TYPE);
	if ((strcmp(type, VDEV_TYPE_SPARE) == 0 \|\|
	strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0) &&
	children == 2 && child[which] == tgt)
	return (B_TRUE);

	for (c = 0; c < children; c++)
	if (is_replacing_spare(child[c], tgt, which))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Attach new_disk (fully described by nvroot) to old_disk.
	* If 'replacing' is specified, the new disk will replace the old one.
	*/
	int
	zpool_vdev_attach(zpool_handle_t zhp, const char old_disk,
	const char new_disk, nvlist_t nvroot, int replacing, boolean_t rebuild)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	int ret;
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache, islog;
	uint64_t val;
	char *newname;
	nvlist_t **child;
	uint_t children;
	nvlist_t *config_root;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	if (replacing)
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot replace %s with %s"), old_disk, new_disk);
	else
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot attach %s to %s"), new_disk, old_disk);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
	&islog)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	if (avail_spare)
	return (zfs_error(hdl, EZFS_ISSPARE, errbuf));

	if (l2cache)
	return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
	zc.zc_cookie = replacing;
	zc.zc_simple = rebuild;

	if (rebuild &&
	zfeature_lookup_guid("org.openzfs:device_rebuild", NULL) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"the loaded zfs module doesn't support device rebuilds"));
	return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
	}

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0 \|\| children != 1) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"new device must be a single disk"));
	return (zfs_error(hdl, EZFS_INVALCONFIG, errbuf));
	}

	config_root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
	ZPOOL_CONFIG_VDEV_TREE);

	if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
	return (-1);

	/*
	* If the target is a hot spare that has been swapped in, we can only
	* replace it with another hot spare.
	*/
	if (replacing &&
	nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
	(zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
	NULL) == NULL \|\| !avail_spare) &&
	is_replacing_spare(config_root, tgt, 1)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"can only be replaced by another hot spare"));
	free(newname);
	return (zfs_error(hdl, EZFS_BADTARGET, errbuf));
	}

	free(newname);

	zcmd_write_conf_nvlist(hdl, &zc, nvroot);

	ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);

	zcmd_free_nvlists(&zc);

	if (ret == 0)
	return (0);

	switch (errno) {
	case ENOTSUP:
	/*
	* Can't attach to or replace this type of vdev.
	*/
	if (replacing) {
	uint64_t version = zpool_get_prop_int(zhp,
	ZPOOL_PROP_VERSION, NULL);

	if (islog) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot replace a log with a spare"));
	} else if (rebuild) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"only mirror and dRAID vdevs support "
	"sequential reconstruction"));
	} else if (zpool_is_draid_spare(new_disk)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dRAID spares can only replace child "
	"devices in their parent's dRAID vdev"));
	} else if (version >= SPA_VERSION_MULTI_REPLACE) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"already in replacing/spare config; wait "
	"for completion or use 'zpool detach'"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot replace a replacing device"));
	}
	} else {
	char status[64] = {0};
	zpool_prop_get_feature(zhp,
	"feature@device_rebuild", status, 63);
	if (rebuild &&
	strncmp(status, ZFS_FEATURE_DISABLED, 64) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"device_rebuild feature must be enabled "
	"in order to use sequential "
	"reconstruction"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"can only attach to mirrors and top-level "
	"disks"));
	}
	}
	(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
	break;

	case EINVAL:
	/*
	* The new device must be a single disk.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"new device must be a single disk"));
	(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
	break;

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
	"or device removal is in progress"),
	new_disk);
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case EOVERFLOW:
	/*
	* The new device is too small.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"device is too small"));
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case EDOM:
	/*
	* The new device has a different optimal sector size.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"new device has a different optimal sector size; use the "
	"option '-o ashift=N' to override the optimal size"));
	(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
	break;

	case ENAMETOOLONG:
	/*
	* The resulting top-level vdev spec won't fit in the label.
	*/
	(void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
	break;

	default:
	(void) zpool_standard_error(hdl, errno, errbuf);
	}

	return (-1);
	}

	/*
	* Detach the specified device.
	*/
	int
	zpool_vdev_detach(zpool_handle_t zhp, const char path)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot detach %s"), path);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	NULL)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	if (avail_spare)
	return (zfs_error(hdl, EZFS_ISSPARE, errbuf));

	if (l2cache)
	return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
	return (0);

	switch (errno) {

	case ENOTSUP:
	/*
	* Can't detach from this type of vdev.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
	"applicable to mirror and replacing vdevs"));
	(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
	break;

	case EBUSY:
	/*
	* There are no other replicas of this device.
	*/
	(void) zfs_error(hdl, EZFS_NOREPLICAS, errbuf);
	break;

	default:
	(void) zpool_standard_error(hdl, errno, errbuf);
	}

	return (-1);
	}

	/*
	* Find a mirror vdev in the source nvlist.
	*
	* The mchild array contains a list of disks in one of the top-level mirrors
	* of the source pool. The schild array contains a list of disks that the
	* user specified on the command line. We loop over the mchild array to
	* see if any entry in the schild array matches.
	*
	* If a disk in the mchild array is found in the schild array, we return
	* the index of that entry. Otherwise we return -1.
	*/
	static int
	find_vdev_entry(zpool_handle_t zhp, nvlist_t *mchild, uint_t mchildren,
	nvlist_t **schild, uint_t schildren)
	{
	uint_t mc;

	for (mc = 0; mc < mchildren; mc++) {
	uint_t sc;
	char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
	mchild[mc], 0);

	for (sc = 0; sc < schildren; sc++) {
	char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
	schild[sc], 0);
	boolean_t result = (strcmp(mpath, spath) == 0);

	free(spath);
	if (result) {
	free(mpath);
	return (mc);
	}
	}

	free(mpath);
	}

	return (-1);
	}

	/*
	* Split a mirror pool. If newroot points to null, then a new nvlist
	* is generated and it is the responsibility of the caller to free it.
	*/
	int
	zpool_vdev_split(zpool_handle_t zhp, char newname, nvlist_t **newroot,
	nvlist_t *props, splitflags_t flags)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN], *bias;
	nvlist_t tree, config, child, newchild, *newconfig = NULL;
	nvlist_t *varray = NULL, zc_props = NULL;
	uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	uint64_t vers, readonly = B_FALSE;
	boolean_t freelist = B_FALSE, memory_err = B_TRUE;
	int retval = 0;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);

	if (!zpool_name_valid(hdl, B_FALSE, newname))
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));

	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
	(void) fprintf(stderr, gettext("Internal error: unable to "
	"retrieve pool configuration\n"));
	return (-1);
	}

	tree = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	vers = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);

	if (props) {
	prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
	if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
	props, vers, flags, errbuf)) == NULL)
	return (-1);
	(void) nvlist_lookup_uint64(zc_props,
	zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
	if (readonly) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property %s can only be set at import time"),
	zpool_prop_to_name(ZPOOL_PROP_READONLY));
	return (-1);
	}
	}

	if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
	&children) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Source pool is missing vdev tree"));
	nvlist_free(zc_props);
	return (-1);
	}

	varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
	vcount = 0;

	if (*newroot == NULL \|\|
	nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
	&newchild, &newchildren) != 0)
	newchildren = 0;

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE, is_hole = B_FALSE;
	boolean_t is_special = B_FALSE, is_dedup = B_FALSE;
	char *type;
	nvlist_t *mchild, vdev;
	uint_t mchildren;
	int entry;

	/*
	* Unlike cache & spares, slogs are stored in the
	* ZPOOL_CONFIG_CHILDREN array. We filter them out here.
	*/
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&is_hole);
	if (is_log \|\| is_hole) {
	/*
	* Create a hole vdev and put it in the config.
	*/
	if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
	goto out;
	if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_HOLE) != 0)
	goto out;
	if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
	1) != 0)
	goto out;
	if (lastlog == 0)
	lastlog = vcount;
	varray[vcount++] = vdev;
	continue;
	}
	lastlog = 0;
	type = fnvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE);

	if (strcmp(type, VDEV_TYPE_INDIRECT) == 0) {
	vdev = child[c];
	if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
	goto out;
	continue;
	} else if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Source pool must be composed only of mirrors\n"));
	retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
	goto out;
	}

	if (nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0) {
	if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
	is_special = B_TRUE;
	else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
	is_dedup = B_TRUE;
	}
	verify(nvlist_lookup_nvlist_array(child[c],
	ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);

	/* find or add an entry for this top-level vdev */
	if (newchildren > 0 &&
	(entry = find_vdev_entry(zhp, mchild, mchildren,
	newchild, newchildren)) >= 0) {
	/* We found a disk that the user specified. */
	vdev = mchild[entry];
	++found;
	} else {
	/* User didn't specify a disk for this vdev. */
	vdev = mchild[mchildren - 1];
	}

	if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
	goto out;

	if (flags.dryrun != 0) {
	if (is_dedup == B_TRUE) {
	if (nvlist_add_string(varray[vcount - 1],
	ZPOOL_CONFIG_ALLOCATION_BIAS,
	VDEV_ALLOC_BIAS_DEDUP) != 0)
	goto out;
	} else if (is_special == B_TRUE) {
	if (nvlist_add_string(varray[vcount - 1],
	ZPOOL_CONFIG_ALLOCATION_BIAS,
	VDEV_ALLOC_BIAS_SPECIAL) != 0)
	goto out;
	}
	}
	}

	/* did we find every disk the user specified? */
	if (found != newchildren) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
	"include at most one disk from each mirror"));
	retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
	goto out;
	}

	/* Prepare the nvlist for populating. */
	if (*newroot == NULL) {
	if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
	goto out;
	freelist = B_TRUE;
	if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_ROOT) != 0)
	goto out;
	} else {
	verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
	}

	/* Add all the children we found */
	if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)varray, lastlog == 0 ? vcount : lastlog) != 0)
	goto out;

	/*
	* If we're just doing a dry run, exit now with success.
	*/
	if (flags.dryrun) {
	memory_err = B_FALSE;
	freelist = B_FALSE;
	goto out;
	}

	/* now build up the config list & call the ioctl */
	if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
	goto out;

	if (nvlist_add_nvlist(newconfig,
	ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 \|\|
	nvlist_add_string(newconfig,
	ZPOOL_CONFIG_POOL_NAME, newname) != 0 \|\|
	nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
	goto out;

	/*
	* The new pool is automatically part of the namespace unless we
	* explicitly export it.
	*/
	if (!flags.import)
	zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	(void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
	zcmd_write_conf_nvlist(hdl, &zc, newconfig);
	if (zc_props != NULL)
	zcmd_write_src_nvlist(hdl, &zc, zc_props);

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
	retval = zpool_standard_error(hdl, errno, errbuf);
	goto out;
	}

	freelist = B_FALSE;
	memory_err = B_FALSE;

	out:
	if (varray != NULL) {
	int v;

	for (v = 0; v < vcount; v++)
	nvlist_free(varray[v]);
	free(varray);
	}
	zcmd_free_nvlists(&zc);
	nvlist_free(zc_props);
	nvlist_free(newconfig);
	if (freelist) {
	nvlist_free(*newroot);
	*newroot = NULL;
	}

	if (retval != 0)
	return (retval);

	if (memory_err)
	return (no_memory(hdl));

	return (0);
	}

	/*
	* Remove the given device.
	*/
	int
	zpool_vdev_remove(zpool_handle_t zhp, const char path)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache, islog;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	uint64_t version;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot remove %s"), path);

	if (zpool_is_draid_spare(path)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dRAID spares cannot be removed"));
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
	}

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	&islog)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
	if (islog && version < SPA_VERSION_HOLES) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to support log removal"));
	return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
	}

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
	return (0);

	switch (errno) {

	case EINVAL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid config; all top-level vdevs must "
	"have the same sector size and not be raidz."));
	(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
	break;

	case EBUSY:
	if (islog) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Mount encrypted datasets to replay logs."));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Pool busy; removal may already be in progress"));
	}
	(void) zfs_error(hdl, EZFS_BUSY, errbuf);
	break;

	case EACCES:
	if (islog) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Mount encrypted datasets to replay logs."));
	(void) zfs_error(hdl, EZFS_BUSY, errbuf);
	} else {
	(void) zpool_standard_error(hdl, errno, errbuf);
	}
	break;

	default:
	(void) zpool_standard_error(hdl, errno, errbuf);
	}
	return (-1);
	}

	int
	zpool_vdev_remove_cancel(zpool_handle_t *zhp)
	{
	zfs_cmd_t zc = {{0}};
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot cancel removal"));

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_cookie = 1;

	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
	return (0);

	return (zpool_standard_error(hdl, errno, errbuf));
	}

	int
	zpool_vdev_indirect_size(zpool_handle_t zhp, const char path,
	uint64_t *sizep)
	{
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache, islog;
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
	path);

	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
	&islog)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	if (avail_spare \|\| l2cache \|\| islog) {
	*sizep = 0;
	return (0);
	}

	if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"indirect size not available"));
	return (zfs_error(hdl, EINVAL, errbuf));
	}
	return (0);
	}

	/*
	* Clear the errors for the pool, or the particular device if specified.
	*/
	int
	zpool_clear(zpool_handle_t zhp, const char path, nvlist_t *rewindnvl)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	nvlist_t *tgt;
	zpool_load_policy_t policy;
	boolean_t avail_spare, l2cache;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	nvlist_t *nvi = NULL;
	int error;

	if (path)
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
	path);
	else
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
	zhp->zpool_name);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if (path) {
	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
	&l2cache, NULL)) == NULL)
	return (zfs_error(hdl, EZFS_NODEVICE, errbuf));

	/*
	* Don't allow error clearing for hot spares. Do allow
	* error clearing for l2cache devices.
	*/
	if (avail_spare)
	return (zfs_error(hdl, EZFS_ISSPARE, errbuf));

	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
	}

	zpool_get_load_policy(rewindnvl, &policy);
	zc.zc_cookie = policy.zlp_rewind;

	zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2);
	zcmd_write_src_nvlist(hdl, &zc, rewindnvl);

	while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
	errno == ENOMEM)
	zcmd_expand_dst_nvlist(hdl, &zc);

	if (!error \|\| ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
	errno != EPERM && errno != EACCES)) {
	if (policy.zlp_rewind &
	(ZPOOL_DO_REWIND \| ZPOOL_TRY_REWIND)) {
	(void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
	zpool_rewind_exclaim(hdl, zc.zc_name,
	((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
	nvi);
	nvlist_free(nvi);
	}
	zcmd_free_nvlists(&zc);
	return (0);
	}

	zcmd_free_nvlists(&zc);
	return (zpool_standard_error(hdl, errno, errbuf));
	}

	/*
	* Similar to zpool_clear(), but takes a GUID (used by fmd).
	*/
	int
	zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
	{
	zfs_cmd_t zc = {"\0"};
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
	(u_longlong_t)guid);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_guid = guid;
	zc.zc_cookie = ZPOOL_NO_REWIND;

	if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
	return (0);

	return (zpool_standard_error(hdl, errno, errbuf));
	}

	/*
	* Change the GUID for a pool.
	*/
	int
	zpool_reguid(zpool_handle_t *zhp)
	{
	char errbuf[ERRBUFLEN];
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	zfs_cmd_t zc = {"\0"};

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
	return (0);

	return (zpool_standard_error(hdl, errno, errbuf));
	}

	/*
	* Reopen the pool.
	*/
	int
	zpool_reopen_one(zpool_handle_t zhp, void data)
	{
	libzfs_handle_t *hdl = zpool_get_handle(zhp);
	const char *pool_name = zpool_get_name(zhp);
	boolean_t *scrub_restart = data;
	int error;

	error = lzc_reopen(pool_name, *scrub_restart);
	if (error) {
	return (zpool_standard_error_fmt(hdl, error,
	dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
	}

	return (0);
	}

	/* call into libzfs_core to execute the sync IOCTL per pool */
	int
	zpool_sync_one(zpool_handle_t zhp, void data)
	{
	int ret;
	libzfs_handle_t *hdl = zpool_get_handle(zhp);
	const char *pool_name = zpool_get_name(zhp);
	boolean_t *force = data;
	nvlist_t *innvl = fnvlist_alloc();

	fnvlist_add_boolean_value(innvl, "force", *force);
	if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
	nvlist_free(innvl);
	return (zpool_standard_error_fmt(hdl, ret,
	dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
	}
	nvlist_free(innvl);

	return (0);
	}

	#define PATH_BUF_LEN 64

	/*
	* Given a vdev, return the name to display in iostat. If the vdev has a path,
	* we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
	* We also check if this is a whole disk, in which case we strip off the
	* trailing 's0' slice name.
	*
	* This routine is also responsible for identifying when disks have been
	* reconfigured in a new location. The kernel will have opened the device by
	* devid, but the path will still refer to the old location. To catch this, we
	* first do a path -> devid translation (which is fast for the common case). If
	* the devid matches, we're done. If not, we do a reverse devid -> path
	* translation and issue the appropriate ioctl() to update the path of the vdev.
	* If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
	* of these checks.
	*/
	char *
	zpool_vdev_name(libzfs_handle_t hdl, zpool_handle_t zhp, nvlist_t *nv,
	int name_flags)
	{
	char type, tpath;
	const char *path;
	uint64_t value;
	char buf[PATH_BUF_LEN];
	char tmpbuf[PATH_BUF_LEN * 2];

	/*
	* vdev_name will be "root"/"root-0" for the root vdev, but it is the
	* zpool name that will be displayed to the user.
	*/
	type = fnvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE);
	if (zhp != NULL && strcmp(type, "root") == 0)
	return (zfs_strdup(hdl, zpool_get_name(zhp)));

	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_PATH"))
	name_flags \|= VDEV_NAME_PATH;
	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_GUID"))
	name_flags \|= VDEV_NAME_GUID;
	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_FOLLOW_LINKS"))
	name_flags \|= VDEV_NAME_FOLLOW_LINKS;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 \|\|
	name_flags & VDEV_NAME_GUID) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
	(void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
	path = buf;
	} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &tpath) == 0) {
	path = tpath;

	if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
	char *rp = realpath(path, NULL);
	if (rp) {
	strlcpy(buf, rp, sizeof (buf));
	path = buf;
	free(rp);
	}
	}

	/*
	* For a block device only use the name.
	*/
	if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
	!(name_flags & VDEV_NAME_PATH)) {
	path = zfs_strip_path(path);
	}

	/*
	* Remove the partition from the path if this is a whole disk.
	*/
	if (strcmp(type, VDEV_TYPE_DRAID_SPARE) != 0 &&
	nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
	== 0 && value && !(name_flags & VDEV_NAME_PATH)) {
	return (zfs_strip_partition(path));
	}
	} else {
	path = type;

	/*
	* If it's a raidz device, we need to stick in the parity level.
	*/
	if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
	value = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY);
	(void) snprintf(buf, sizeof (buf), "%s%llu", path,
	(u_longlong_t)value);
	path = buf;
	}

	/*
	* If it's a dRAID device, we add parity, groups, and spares.
	*/
	if (strcmp(path, VDEV_TYPE_DRAID) == 0) {
	uint64_t ndata, nparity, nspares;
	nvlist_t **child;
	uint_t children;

	verify(nvlist_lookup_nvlist_array(nv,
	ZPOOL_CONFIG_CHILDREN, &child, &children) == 0);
	nparity = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_NPARITY);
	ndata = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_DRAID_NDATA);
	nspares = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_DRAID_NSPARES);

	path = zpool_draid_name(buf, sizeof (buf), ndata,
	nparity, nspares, children);
	}

	/*
	* We identify each top-level vdev by using a <type-id>
	* naming convention.
	*/
	if (name_flags & VDEV_NAME_TYPE_ID) {
	uint64_t id = fnvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_ID);
	(void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
	path, (u_longlong_t)id);
	path = tmpbuf;
	}
	}

	return (zfs_strdup(hdl, path));
	}

	static int
	zbookmark_mem_compare(const void a, const void b)
	{
	return (memcmp(a, b, sizeof (zbookmark_phys_t)));
	}

	/*
	* Retrieve the persistent error log, uniquify the members, and return to the
	* caller.
	*/
	int
	zpool_get_errlog(zpool_handle_t zhp, nvlist_t *nverrlistp)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	uint64_t count;
	zbookmark_phys_t *zb = NULL;
	int i;

	/*
	* Retrieve the raw error list from the kernel. If the number of errors
	* has increased, allocate more space and continue until we get the
	* entire list.
	*/
	count = fnvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT);
	if (count == 0)
	return (0);
	zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
	count * sizeof (zbookmark_phys_t));
	zc.zc_nvlist_dst_size = count;
	(void) strcpy(zc.zc_name, zhp->zpool_name);
	for (;;) {
	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_ERROR_LOG,
	&zc) != 0) {
	free((void *)(uintptr_t)zc.zc_nvlist_dst);
	if (errno == ENOMEM) {
	void *dst;

	count = zc.zc_nvlist_dst_size;
	dst = zfs_alloc(zhp->zpool_hdl, count *
	sizeof (zbookmark_phys_t));
	zc.zc_nvlist_dst = (uintptr_t)dst;
	} else {
	return (zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN, "errors: List of "
	"errors unavailable")));
	}
	} else {
	break;
	}
	}

	/*
	* Sort the resulting bookmarks. This is a little confusing due to the
	* implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
	* to first, and 'zc_nvlist_dst_size' indicates the number of bookmarks
	* _not_ copied as part of the process. So we point the start of our
	* array appropriate and decrement the total number of elements.
	*/
	zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
	zc.zc_nvlist_dst_size;
	count -= zc.zc_nvlist_dst_size;

	qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);

	verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);

	/*
	* Fill in the nverrlistp with nvlist's of dataset and object numbers.
	*/
	for (i = 0; i < count; i++) {
	nvlist_t *nv;

	/* ignoring zb_blkid and zb_level for now */
	if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
	zb[i-1].zb_object == zb[i].zb_object)
	continue;

	if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
	goto nomem;
	if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
	zb[i].zb_objset) != 0) {
	nvlist_free(nv);
	goto nomem;
	}
	if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
	zb[i].zb_object) != 0) {
	nvlist_free(nv);
	goto nomem;
	}
	if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
	nvlist_free(nv);
	goto nomem;
	}
	nvlist_free(nv);
	}

	free((void *)(uintptr_t)zc.zc_nvlist_dst);
	return (0);

	nomem:
	free((void *)(uintptr_t)zc.zc_nvlist_dst);
	return (no_memory(zhp->zpool_hdl));
	}

	/*
	* Upgrade a ZFS pool to the latest on-disk version.
	*/
	int
	zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) strcpy(zc.zc_name, zhp->zpool_name);
	zc.zc_cookie = new_version;

	if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
	return (zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
	zhp->zpool_name));
	return (0);
	}

	void
	zfs_save_arguments(int argc, char *argv, char string, int len)
	{
	int i;

	(void) strlcpy(string, zfs_basename(argv[0]), len);
	for (i = 1; i < argc; i++) {
	(void) strlcat(string, " ", len);
	(void) strlcat(string, argv[i], len);
	}
	}

	int
	zpool_log_history(libzfs_handle_t hdl, const char message)
	{
	zfs_cmd_t zc = {"\0"};
	nvlist_t *args;

	args = fnvlist_alloc();
	fnvlist_add_string(args, "message", message);
	zcmd_write_src_nvlist(hdl, &zc, args);
	int err = zfs_ioctl(hdl, ZFS_IOC_LOG_HISTORY, &zc);
	nvlist_free(args);
	zcmd_free_nvlists(&zc);
	return (err);
	}

	/*
	* Perform ioctl to get some command history of a pool.
	*
	* 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
	* logical offset of the history buffer to start reading from.
	*
	* Upon return, 'off' is the next logical offset to read from and
	* 'len' is the actual amount of bytes read into 'buf'.
	*/
	static int
	get_history(zpool_handle_t zhp, char buf, uint64_t off, uint64_t len)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zpool_hdl;

	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));

	zc.zc_history = (uint64_t)(uintptr_t)buf;
	zc.zc_history_len = *len;
	zc.zc_history_offset = *off;

	if (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
	switch (errno) {
	case EPERM:
	return (zfs_error_fmt(hdl, EZFS_PERM,
	dgettext(TEXT_DOMAIN,
	"cannot show history for pool '%s'"),
	zhp->zpool_name));
	case ENOENT:
	return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
	dgettext(TEXT_DOMAIN, "cannot get history for pool "
	"'%s'"), zhp->zpool_name));
	case ENOTSUP:
	return (zfs_error_fmt(hdl, EZFS_BADVERSION,
	dgettext(TEXT_DOMAIN, "cannot get history for pool "
	"'%s', pool must be upgraded"), zhp->zpool_name));
	default:
	return (zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN,
	"cannot get history for '%s'"), zhp->zpool_name));
	}
	}

	*len = zc.zc_history_len;
	*off = zc.zc_history_offset;

	return (0);
	}

	/*
	* Retrieve the command history of a pool.
	*/
	int
	zpool_get_history(zpool_handle_t zhp, nvlist_t nvhisp, uint64_t off,
	boolean_t *eof)
	{
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	char *buf;
	int buflen = 128 * 1024;
	nvlist_t **records = NULL;
	uint_t numrecords = 0;
	int err = 0, i;
	uint64_t start = *off;

	buf = zfs_alloc(hdl, buflen);

	/* process about 1MiB a time */
	while (off - start < 1024 1024) {
	uint64_t bytes_read = buflen;
	uint64_t leftover;

	if ((err = get_history(zhp, buf, off, &bytes_read)) != 0)
	break;

	/* if nothing else was read in, we're at EOF, just return */
	if (!bytes_read) {
	*eof = B_TRUE;
	break;
	}

	if ((err = zpool_history_unpack(buf, bytes_read,
	&leftover, &records, &numrecords)) != 0) {
	zpool_standard_error_fmt(hdl, err,
	dgettext(TEXT_DOMAIN,
	"cannot get history for '%s'"), zhp->zpool_name);
	break;
	}
	*off -= leftover;
	if (leftover == bytes_read) {
	/*
	* no progress made, because buffer is not big enough
	* to hold this record; resize and retry.
	*/
	buflen *= 2;
	free(buf);
	buf = zfs_alloc(hdl, buflen);
	}
	}

	free(buf);

	if (!err) {
	*nvhisp = fnvlist_alloc();
	fnvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
	(const nvlist_t **)records, numrecords);
	}
	for (i = 0; i < numrecords; i++)
	nvlist_free(records[i]);
	free(records);

	return (err);
	}

	/*
	* Retrieve the next event given the passed 'zevent_fd' file descriptor.
	* If there is a new event available 'nvp' will contain a newly allocated
	* nvlist and 'dropped' will be set to the number of missed events since
	* the last call to this function. When 'nvp' is set to NULL it indicates
	* no new events are available. In either case the function returns 0 and
	* it is up to the caller to free 'nvp'. In the case of a fatal error the
	* function will return a non-zero value. When the function is called in
	* blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
	* it will not return until a new event is available.
	*/
	int
	zpool_events_next(libzfs_handle_t hdl, nvlist_t *nvp,
	int *dropped, unsigned flags, int zevent_fd)
	{
	zfs_cmd_t zc = {"\0"};
	int error = 0;

	*nvp = NULL;
	*dropped = 0;
	zc.zc_cleanup_fd = zevent_fd;

	if (flags & ZEVENT_NONBLOCK)
	zc.zc_guid = ZEVENT_NONBLOCK;

	zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE);

	retry:
	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
	switch (errno) {
	case ESHUTDOWN:
	error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
	dgettext(TEXT_DOMAIN, "zfs shutdown"));
	goto out;
	case ENOENT:
	/* Blocking error case should not occur */
	if (!(flags & ZEVENT_NONBLOCK))
	error = zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot get event"));

	goto out;
	case ENOMEM:
	zcmd_expand_dst_nvlist(hdl, &zc);
	goto retry;
	default:
	error = zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot get event"));
	goto out;
	}
	}

	error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
	if (error != 0)
	goto out;

	*dropped = (int)zc.zc_cookie;
	out:
	zcmd_free_nvlists(&zc);

	return (error);
	}

	/*
	* Clear all events.
	*/
	int
	zpool_events_clear(libzfs_handle_t hdl, int count)
	{
	zfs_cmd_t zc = {"\0"};

	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
	return (zpool_standard_error(hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot clear events")));

	if (count != NULL)
	count = (int)zc.zc_cookie; / # of events cleared */

	return (0);
	}

	/*
	* Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
	* the passed zevent_fd file handle. On success zero is returned,
	* otherwise -1 is returned and hdl->libzfs_error is set to the errno.
	*/
	int
	zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
	{
	zfs_cmd_t zc = {"\0"};
	int error = 0;

	zc.zc_guid = eid;
	zc.zc_cleanup_fd = zevent_fd;

	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
	switch (errno) {
	case ENOENT:
	error = zfs_error_fmt(hdl, EZFS_NOENT,
	dgettext(TEXT_DOMAIN, "cannot get event"));
	break;

	case ENOMEM:
	error = zfs_error_fmt(hdl, EZFS_NOMEM,
	dgettext(TEXT_DOMAIN, "cannot get event"));
	break;

	default:
	error = zpool_standard_error_fmt(hdl, errno,
	dgettext(TEXT_DOMAIN, "cannot get event"));
	break;
	}
	}

	return (error);
	}

	static void
	zpool_obj_to_path_impl(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
	char *pathname, size_t len, boolean_t always_unmounted)
	{
	zfs_cmd_t zc = {"\0"};
	boolean_t mounted = B_FALSE;
	char *mntpnt = NULL;
	char dsname[ZFS_MAX_DATASET_NAME_LEN];

	if (dsobj == 0) {
	/* special case for the MOS */
	(void) snprintf(pathname, len, "<metadata>:<0x%llx>",
	(longlong_t)obj);
	return;
	}

	/* get the dataset's name */
	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
	zc.zc_obj = dsobj;
	if (zfs_ioctl(zhp->zpool_hdl,
	ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
	/* just write out a path of two object numbers */
	(void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
	(longlong_t)dsobj, (longlong_t)obj);
	return;
	}
	(void) strlcpy(dsname, zc.zc_value, sizeof (dsname));

	/* find out if the dataset is mounted */
	mounted = !always_unmounted && is_mounted(zhp->zpool_hdl, dsname,
	&mntpnt);

	/* get the corrupted object's path */
	(void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
	zc.zc_obj = obj;
	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_OBJ_TO_PATH,
	&zc) == 0) {
	if (mounted) {
	(void) snprintf(pathname, len, "%s%s", mntpnt,
	zc.zc_value);
	} else {
	(void) snprintf(pathname, len, "%s:%s",
	dsname, zc.zc_value);
	}
	} else {
	(void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
	(longlong_t)obj);
	}
	free(mntpnt);
	}

	void
	zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
	char *pathname, size_t len)
	{
	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_FALSE);
	}

	void
	zpool_obj_to_path_ds(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
	char *pathname, size_t len)
	{
	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_TRUE);
	}
	/*
	* Wait while the specified activity is in progress in the pool.
	*/
	int
	zpool_wait(zpool_handle_t *zhp, zpool_wait_activity_t activity)
	{
	boolean_t missing;

	int error = zpool_wait_status(zhp, activity, &missing, NULL);

	if (missing) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl, ENOENT,
	dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
	zhp->zpool_name);
	return (ENOENT);
	} else {
	return (error);
	}
	}

	/*
	* Wait for the given activity and return the status of the wait (whether or not
	* any waiting was done) in the 'waited' parameter. Non-existent pools are
	* reported via the 'missing' parameter, rather than by printing an error
	* message. This is convenient when this function is called in a loop over a
	* long period of time (as it is, for example, by zpool's wait cmd). In that
	* scenario, a pool being exported or destroyed should be considered a normal
	* event, so we don't want to print an error when we find that the pool doesn't
	* exist.
	*/
	int
	zpool_wait_status(zpool_handle_t *zhp, zpool_wait_activity_t activity,
	boolean_t missing, boolean_t waited)
	{
	int error = lzc_wait(zhp->zpool_name, activity, waited);
	*missing = (error == ENOENT);
	if (*missing)
	return (0);

	if (error != 0) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
	dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
	zhp->zpool_name);
	}

	return (error);
	}

	int
	zpool_set_bootenv(zpool_handle_t zhp, const nvlist_t envmap)
	{
	int error = lzc_set_bootenv(zhp->zpool_name, envmap);
	if (error != 0) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
	dgettext(TEXT_DOMAIN,
	"error setting bootenv in pool '%s'"), zhp->zpool_name);
	}

	return (error);
	}

	int
	zpool_get_bootenv(zpool_handle_t zhp, nvlist_t *nvlp)
	{
	nvlist_t *nvl;
	int error;

	nvl = NULL;
	error = lzc_get_bootenv(zhp->zpool_name, &nvl);
	if (error != 0) {
	(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
	dgettext(TEXT_DOMAIN,
	"error getting bootenv in pool '%s'"), zhp->zpool_name);
	} else {
	*nvlp = nvl;
	}

	return (error);
	}

	/*
	* Attempt to read and parse feature file(s) (from "compatibility" property).
	* Files contain zpool feature names, comma or whitespace-separated.
	* Comments (# character to next newline) are discarded.
	*
	* Arguments:
	* compatibility : string containing feature filenames
	* features : either NULL or pointer to array of boolean
	* report : either NULL or pointer to string buffer
	* rlen : length of "report" buffer
	*
	* compatibility is NULL (unset), "", "off", "legacy", or list of
	* comma-separated filenames. filenames should either be absolute,
	* or relative to:
	* 1) ZPOOL_SYSCONF_COMPAT_D (eg: /etc/zfs/compatibility.d) or
	* 2) ZPOOL_DATA_COMPAT_D (eg: /usr/share/zfs/compatibility.d).
	* (Unset), "" or "off" => enable all features
	* "legacy" => disable all features
	*
	* Any feature names read from files which match unames in spa_feature_table
	* will have the corresponding boolean set in the features array (if non-NULL).
	* If more than one feature set specified, only features present in all of
	* them will be set.
	*
	* "report" if not NULL will be populated with a suitable status message.
	*
	* Return values:
	* ZPOOL_COMPATIBILITY_OK : files read and parsed ok
	* ZPOOL_COMPATIBILITY_BADFILE : file too big or not a text file
	* ZPOOL_COMPATIBILITY_BADTOKEN : SYSCONF file contains invalid feature name
	* ZPOOL_COMPATIBILITY_WARNTOKEN : DATA file contains invalid feature name
	* ZPOOL_COMPATIBILITY_NOFILES : no feature files found
	*/
	zpool_compat_status_t
	zpool_load_compat(const char compat, boolean_t features, char *report,
	size_t rlen)
	{
	int sdirfd, ddirfd, featfd;
	struct stat fs;
	char *fc;
	char ps, ls, *ws;
	char file, line, *word;

	char l_compat[ZFS_MAXPROPLEN];

	boolean_t ret_nofiles = B_TRUE;
	boolean_t ret_badfile = B_FALSE;
	boolean_t ret_badtoken = B_FALSE;
	boolean_t ret_warntoken = B_FALSE;

	/* special cases (unset), "" and "off" => enable all features */
	if (compat == NULL \|\| compat[0] == '\0' \|\|
	strcmp(compat, ZPOOL_COMPAT_OFF) == 0) {
	if (features != NULL)
	for (uint_t i = 0; i < SPA_FEATURES; i++)
	features[i] = B_TRUE;
	if (report != NULL)
	strlcpy(report, gettext("all features enabled"), rlen);
	return (ZPOOL_COMPATIBILITY_OK);
	}

	/* Final special case "legacy" => disable all features */
	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
	if (features != NULL)
	for (uint_t i = 0; i < SPA_FEATURES; i++)
	features[i] = B_FALSE;
	if (report != NULL)
	strlcpy(report, gettext("all features disabled"), rlen);
	return (ZPOOL_COMPATIBILITY_OK);
	}

	/*
	* Start with all true; will be ANDed with results from each file
	*/
	if (features != NULL)
	for (uint_t i = 0; i < SPA_FEATURES; i++)
	features[i] = B_TRUE;

	char err_badfile[ZFS_MAXPROPLEN] = "";
	char err_badtoken[ZFS_MAXPROPLEN] = "";

	/*
	* We ignore errors from the directory open()
	* as they're only needed if the filename is relative
	* which will be checked during the openat().
	*/

	/* O_PATH safer than O_RDONLY if system allows it */
	#if defined(O_PATH)
	#define ZC_DIR_FLAGS (O_DIRECTORY \| O_CLOEXEC \| O_PATH)
	#else
	#define ZC_DIR_FLAGS (O_DIRECTORY \| O_CLOEXEC \| O_RDONLY)
	#endif

	sdirfd = open(ZPOOL_SYSCONF_COMPAT_D, ZC_DIR_FLAGS);
	ddirfd = open(ZPOOL_DATA_COMPAT_D, ZC_DIR_FLAGS);

	(void) strlcpy(l_compat, compat, ZFS_MAXPROPLEN);

	for (file = strtok_r(l_compat, ",", &ps);
	file != NULL;
	file = strtok_r(NULL, ",", &ps)) {

	boolean_t l_features[SPA_FEATURES];

	enum { Z_SYSCONF, Z_DATA } source;

	/* try sysconfdir first, then datadir */
	source = Z_SYSCONF;
	if ((featfd = openat(sdirfd, file, O_RDONLY \| O_CLOEXEC)) < 0) {
	featfd = openat(ddirfd, file, O_RDONLY \| O_CLOEXEC);
	source = Z_DATA;
	}

	/* File readable and correct size? */
	if (featfd < 0 \|\|
	fstat(featfd, &fs) < 0 \|\|
	fs.st_size < 1 \|\|
	fs.st_size > ZPOOL_COMPAT_MAXSIZE) {
	(void) close(featfd);
	strlcat(err_badfile, file, ZFS_MAXPROPLEN);
	strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
	ret_badfile = B_TRUE;
	continue;
	}

	/* Prefault the file if system allows */
	#if defined(MAP_POPULATE)
	#define ZC_MMAP_FLAGS (MAP_PRIVATE \| MAP_POPULATE)
	#elif defined(MAP_PREFAULT_READ)
	#define ZC_MMAP_FLAGS (MAP_PRIVATE \| MAP_PREFAULT_READ)
	#else
	#define ZC_MMAP_FLAGS (MAP_PRIVATE)
	#endif

	/* private mmap() so we can strtok safely */
	fc = (char *)mmap(NULL, fs.st_size, PROT_READ \| PROT_WRITE,
	ZC_MMAP_FLAGS, featfd, 0);
	(void) close(featfd);

	/* map ok, and last character == newline? */
	if (fc == MAP_FAILED \|\| fc[fs.st_size - 1] != '\n') {
	(void) munmap((void *) fc, fs.st_size);
	strlcat(err_badfile, file, ZFS_MAXPROPLEN);
	strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
	ret_badfile = B_TRUE;
	continue;
	}

	ret_nofiles = B_FALSE;

	for (uint_t i = 0; i < SPA_FEATURES; i++)
	l_features[i] = B_FALSE;

	/* replace final newline with NULL to ensure string ends */
	fc[fs.st_size - 1] = '\0';

	for (line = strtok_r(fc, "\n", &ls);
	line != NULL;
	line = strtok_r(NULL, "\n", &ls)) {
	/* discard comments */
	char *r = strchr(line, '#');
	if (r != NULL)
	*r = '\0';

	for (word = strtok_r(line, ", \t", &ws);
	word != NULL;
	word = strtok_r(NULL, ", \t", &ws)) {
	/* Find matching feature name */
	uint_t f;
	for (f = 0; f < SPA_FEATURES; f++) {
	zfeature_info_t *fi =
	&spa_feature_table[f];
	if (strcmp(word, fi->fi_uname) == 0) {
	l_features[f] = B_TRUE;
	break;
	}
	}
	if (f < SPA_FEATURES)
	continue;

	/* found an unrecognized word */
	/* lightly sanitize it */
	if (strlen(word) > 32)
	word[32] = '\0';
	for (char c = word; c != '\0'; c++)
	if (!isprint(*c))
	*c = '?';

	strlcat(err_badtoken, word, ZFS_MAXPROPLEN);
	strlcat(err_badtoken, " ", ZFS_MAXPROPLEN);
	if (source == Z_SYSCONF)
	ret_badtoken = B_TRUE;
	else
	ret_warntoken = B_TRUE;
	}
	}
	(void) munmap((void *) fc, fs.st_size);

	if (features != NULL)
	for (uint_t i = 0; i < SPA_FEATURES; i++)
	features[i] &= l_features[i];
	}
	(void) close(sdirfd);
	(void) close(ddirfd);

	/* Return the most serious error */
	if (ret_badfile) {
	if (report != NULL)
	snprintf(report, rlen, gettext("could not read/"
	"parse feature file(s): %s"), err_badfile);
	return (ZPOOL_COMPATIBILITY_BADFILE);
	}
	if (ret_nofiles) {
	if (report != NULL)
	strlcpy(report,
	gettext("no valid compatibility files specified"),
	rlen);
	return (ZPOOL_COMPATIBILITY_NOFILES);
	}
	if (ret_badtoken) {
	if (report != NULL)
	snprintf(report, rlen, gettext("invalid feature "
	"name(s) in local compatibility files: %s"),
	err_badtoken);
	return (ZPOOL_COMPATIBILITY_BADTOKEN);
	}
	if (ret_warntoken) {
	if (report != NULL)
	snprintf(report, rlen, gettext("unrecognized feature "
	"name(s) in distribution compatibility files: %s"),
	err_badtoken);
	return (ZPOOL_COMPATIBILITY_WARNTOKEN);
	}
	if (report != NULL)
	strlcpy(report, gettext("compatibility set ok"), rlen);
	return (ZPOOL_COMPATIBILITY_OK);
	}

	static int
	zpool_vdev_guid(zpool_handle_t zhp, const char vdevname, uint64_t *vdev_guid)
	{
	nvlist_t *tgt;
	boolean_t avail_spare, l2cache;

	verify(zhp != NULL);
	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "pool is in an unavailable state"));
	return (zfs_error(zhp->zpool_hdl, EZFS_POOLUNAVAIL, errbuf));
	}

	if ((tgt = zpool_find_vdev(zhp, vdevname, &avail_spare, &l2cache,
	NULL)) == NULL) {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "can not find %s in %s"),
	vdevname, zhp->zpool_name);
	return (zfs_error(zhp->zpool_hdl, EZFS_NODEVICE, errbuf));
	}

	*vdev_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
	return (0);
	}

	/*
	* Get a vdev property value for 'prop' and return the value in
	* a pre-allocated buffer.
	*/
	int
	zpool_get_vdev_prop_value(nvlist_t nvprop, vdev_prop_t prop, char prop_name,
	char buf, size_t len, zprop_source_t srctype, boolean_t literal)
	{
	nvlist_t *nv;
	const char *strval;
	uint64_t intval;
	zprop_source_t src = ZPROP_SRC_NONE;

	if (prop == VDEV_PROP_USERPROP) {
	/* user property, prop_name must contain the property name */
	assert(prop_name != NULL);
	if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
	src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
	} else {
	/* user prop not found */
	return (-1);
	}
	(void) strlcpy(buf, strval, len);
	if (srctype)
	*srctype = src;
	return (0);
	}

	if (prop_name == NULL)
	prop_name = (char *)vdev_prop_to_name(prop);

	switch (vdev_prop_get_type(prop)) {
	case PROP_TYPE_STRING:
	if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
	src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
	} else {
	src = ZPROP_SRC_DEFAULT;
	if ((strval = vdev_prop_default_string(prop)) == NULL)
	strval = "-";
	}
	(void) strlcpy(buf, strval, len);
	break;

	case PROP_TYPE_NUMBER:
	if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
	src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
	} else {
	src = ZPROP_SRC_DEFAULT;
	intval = vdev_prop_default_numeric(prop);
	}

	switch (prop) {
	case VDEV_PROP_ASIZE:
	case VDEV_PROP_PSIZE:
	case VDEV_PROP_SIZE:
	case VDEV_PROP_BOOTSIZE:
	case VDEV_PROP_ALLOCATED:
	case VDEV_PROP_FREE:
	case VDEV_PROP_READ_ERRORS:
	case VDEV_PROP_WRITE_ERRORS:
	case VDEV_PROP_CHECKSUM_ERRORS:
	case VDEV_PROP_INITIALIZE_ERRORS:
	case VDEV_PROP_OPS_NULL:
	case VDEV_PROP_OPS_READ:
	case VDEV_PROP_OPS_WRITE:
	case VDEV_PROP_OPS_FREE:
	case VDEV_PROP_OPS_CLAIM:
	case VDEV_PROP_OPS_TRIM:
	case VDEV_PROP_BYTES_NULL:
	case VDEV_PROP_BYTES_READ:
	case VDEV_PROP_BYTES_WRITE:
	case VDEV_PROP_BYTES_FREE:
	case VDEV_PROP_BYTES_CLAIM:
	case VDEV_PROP_BYTES_TRIM:
	if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) zfs_nicenum(intval, buf, len);
	}
	break;
	case VDEV_PROP_EXPANDSZ:
	if (intval == 0) {
	(void) strlcpy(buf, "-", len);
	} else if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) zfs_nicenum(intval, buf, len);
	}
	break;
	case VDEV_PROP_CAPACITY:
	if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) snprintf(buf, len, "%llu%%",
	(u_longlong_t)intval);
	}
	break;
	case VDEV_PROP_FRAGMENTATION:
	if (intval == UINT64_MAX) {
	(void) strlcpy(buf, "-", len);
	} else {
	(void) snprintf(buf, len, "%llu%%",
	(u_longlong_t)intval);
	}
	break;
	case VDEV_PROP_STATE:
	if (literal) {
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	} else {
	(void) strlcpy(buf, zpool_state_to_name(intval,
	VDEV_AUX_NONE), len);
	}
	break;
	default:
	(void) snprintf(buf, len, "%llu",
	(u_longlong_t)intval);
	}
	break;

	case PROP_TYPE_INDEX:
	if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
	src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
	intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
	} else {
	src = ZPROP_SRC_DEFAULT;
	intval = vdev_prop_default_numeric(prop);
	}
	if (vdev_prop_index_to_string(prop, intval,
	(const char **)&strval) != 0)
	return (-1);
	(void) strlcpy(buf, strval, len);
	break;

	default:
	abort();
	}

	if (srctype)
	*srctype = src;

	return (0);
	}

	/*
	* Get a vdev property value for 'prop_name' and return the value in
	* a pre-allocated buffer.
	*/
	int
	zpool_get_vdev_prop(zpool_handle_t zhp, const char vdevname, vdev_prop_t prop,
	char prop_name, char buf, size_t len, zprop_source_t *srctype,
	boolean_t literal)
	{
	nvlist_t reqnvl, reqprops;
	nvlist_t *retprops = NULL;
	uint64_t vdev_guid = 0;
	int ret;

	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
	return (ret);

	if (nvlist_alloc(&reqnvl, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));
	if (nvlist_alloc(&reqprops, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));

	fnvlist_add_uint64(reqnvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);

	if (prop != VDEV_PROP_USERPROP) {
	/* prop_name overrides prop value */
	if (prop_name != NULL)
	prop = vdev_name_to_prop(prop_name);
	else
	prop_name = (char *)vdev_prop_to_name(prop);
	assert(prop < VDEV_NUM_PROPS);
	}

	assert(prop_name != NULL);
	if (nvlist_add_uint64(reqprops, prop_name, prop) != 0) {
	nvlist_free(reqnvl);
	nvlist_free(reqprops);
	return (no_memory(zhp->zpool_hdl));
	}

	fnvlist_add_nvlist(reqnvl, ZPOOL_VDEV_PROPS_GET_PROPS, reqprops);

	ret = lzc_get_vdev_prop(zhp->zpool_name, reqnvl, &retprops);

	if (ret == 0) {
	ret = zpool_get_vdev_prop_value(retprops, prop, prop_name, buf,
	len, srctype, literal);
	} else {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot get vdev property %s from"
	" %s in %s"), prop_name, vdevname, zhp->zpool_name);
	(void) zpool_standard_error(zhp->zpool_hdl, ret, errbuf);
	}

	nvlist_free(reqnvl);
	nvlist_free(reqprops);
	nvlist_free(retprops);

	return (ret);
	}

	/*
	* Get all vdev properties
	*/
	int
	zpool_get_all_vdev_props(zpool_handle_t zhp, const char vdevname,
	nvlist_t **outnvl)
	{
	nvlist_t *nvl = NULL;
	uint64_t vdev_guid = 0;
	int ret;

	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
	return (ret);

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));

	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);

	ret = lzc_get_vdev_prop(zhp->zpool_name, nvl, outnvl);

	nvlist_free(nvl);

	if (ret) {
	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot get vdev properties for"
	" %s in %s"), vdevname, zhp->zpool_name);
	(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
	}

	return (ret);
	}

	/*
	* Set vdev property
	*/
	int
	zpool_set_vdev_prop(zpool_handle_t zhp, const char vdevname,
	const char propname, const char propval)
	{
	int ret;
	nvlist_t *nvl = NULL;
	nvlist_t *outnvl = NULL;
	nvlist_t *props;
	nvlist_t *realprops;
	prop_flags_t flags = { 0 };
	uint64_t version;
	uint64_t vdev_guid;

	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
	return (ret);

	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));
	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	return (no_memory(zhp->zpool_hdl));

	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_SET_VDEV, vdev_guid);

	if (nvlist_add_string(props, propname, propval) != 0) {
	nvlist_free(props);
	return (no_memory(zhp->zpool_hdl));
	}

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot set property %s for %s on %s"),
	propname, vdevname, zhp->zpool_name);

	flags.vdevprop = 1;
	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
	zhp->zpool_name, props, version, flags, errbuf)) == NULL) {
	nvlist_free(props);
	nvlist_free(nvl);
	return (-1);
	}

	nvlist_free(props);
	props = realprops;

	fnvlist_add_nvlist(nvl, ZPOOL_VDEV_PROPS_SET_PROPS, props);

	ret = lzc_set_vdev_prop(zhp->zpool_name, nvl, &outnvl);

	nvlist_free(props);
	nvlist_free(nvl);
	nvlist_free(outnvl);

	if (ret)
	(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);

	return (ret);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c b/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	index 577ebf6aad42..d63a9e1a4e0a 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	@@ -1,5426 +1,5470 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	* Copyright (c) 2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
	* All rights reserved
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2019 Datto Inc.
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <errno.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <stddef.h>
	#include <fcntl.h>
	#include <sys/mount.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/avl.h>
	#include <sys/debug.h>
	#include <sys/stat.h>
	#include <pthread.h>
	#include <umem.h>
	#include <time.h>

	#include <libzfs.h>
	#include <libzfs_core.h>
	#include <libzutil.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "zfs_fletcher.h"
	#include "libzfs_impl.h"
	#include <cityhash.h>
	#include <zlib.h>
	#include <sys/zio_checksum.h>
	#include <sys/dsl_crypt.h>
	#include <sys/ddt.h>
	#include <sys/socket.h>
	#include <sys/sha2.h>

	static int zfs_receive_impl(libzfs_handle_t , const char , const char *,
	recvflags_t , int, const char , nvlist_t , avl_tree_t , char **,
	const char , nvlist_t );
	static int guid_to_name_redact_snaps(libzfs_handle_t hdl, const char parent,
	uint64_t guid, boolean_t bookmark_ok, uint64_t *redact_snap_guids,
	uint64_t num_redact_snaps, char *name);
	static int guid_to_name(libzfs_handle_t , const char ,
	uint64_t, boolean_t, char *);

	typedef struct progress_arg {
	zfs_handle_t *pa_zhp;
	int pa_fd;
	boolean_t pa_parsable;
	boolean_t pa_estimate;
	int pa_verbosity;
	} progress_arg_t;

	static int
	dump_record(dmu_replay_record_t drr, void payload, size_t payload_len,
	zio_cksum_t *zc, int outfd)
	{
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	fletcher_4_incremental_native(drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc);
	if (drr->drr_type != DRR_BEGIN) {
	ASSERT(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.
	drr_checksum.drr_checksum));
	drr->drr_u.drr_checksum.drr_checksum = *zc;
	}
	fletcher_4_incremental_native(&drr->drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), zc);
	if (write(outfd, drr, sizeof (*drr)) == -1)
	return (errno);
	if (payload_len != 0) {
	fletcher_4_incremental_native(payload, payload_len, zc);
	if (write(outfd, payload, payload_len) == -1)
	return (errno);
	}
	return (0);
	}

	/*
	* Routines for dealing with the AVL tree of fs-nvlists
	*/
	typedef struct fsavl_node {
	avl_node_t fn_node;
	nvlist_t *fn_nvfs;
	char *fn_snapname;
	uint64_t fn_guid;
	} fsavl_node_t;

	static int
	fsavl_compare(const void arg1, const void arg2)
	{
	const fsavl_node_t fn1 = (const fsavl_node_t )arg1;
	const fsavl_node_t fn2 = (const fsavl_node_t )arg2;

	return (TREE_CMP(fn1->fn_guid, fn2->fn_guid));
	}

	/*
	* Given the GUID of a snapshot, find its containing filesystem and
	* (optionally) name.
	*/
	static nvlist_t *
	fsavl_find(avl_tree_t avl, uint64_t snapguid, char *snapname)
	{
	fsavl_node_t fn_find;
	fsavl_node_t *fn;

	fn_find.fn_guid = snapguid;

	fn = avl_find(avl, &fn_find, NULL);
	if (fn) {
	if (snapname)
	*snapname = fn->fn_snapname;
	return (fn->fn_nvfs);
	}
	return (NULL);
	}

	static void
	fsavl_destroy(avl_tree_t *avl)
	{
	fsavl_node_t *fn;
	void *cookie;

	if (avl == NULL)
	return;

	cookie = NULL;
	while ((fn = avl_destroy_nodes(avl, &cookie)) != NULL)
	free(fn);
	avl_destroy(avl);
	free(avl);
	}

	/*
	* Given an nvlist, produce an avl tree of snapshots, ordered by guid
	*/
	static avl_tree_t *
	fsavl_create(nvlist_t *fss)
	{
	avl_tree_t *fsavl;
	nvpair_t *fselem = NULL;

	if ((fsavl = malloc(sizeof (avl_tree_t))) == NULL)
	return (NULL);

	avl_create(fsavl, fsavl_compare, sizeof (fsavl_node_t),
	offsetof(fsavl_node_t, fn_node));

	while ((fselem = nvlist_next_nvpair(fss, fselem)) != NULL) {
	nvlist_t nvfs, snaps;
	nvpair_t *snapelem = NULL;

	nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(nvfs, "snaps");

	while ((snapelem =
	nvlist_next_nvpair(snaps, snapelem)) != NULL) {
	fsavl_node_t *fn;

	if ((fn = malloc(sizeof (fsavl_node_t))) == NULL) {
	fsavl_destroy(fsavl);
	return (NULL);
	}
	fn->fn_nvfs = nvfs;
	fn->fn_snapname = nvpair_name(snapelem);
	fn->fn_guid = fnvpair_value_uint64(snapelem);

	/*
	* Note: if there are multiple snaps with the
	* same GUID, we ignore all but one.
	*/
	avl_index_t where = 0;
	if (avl_find(fsavl, fn, &where) == NULL)
	avl_insert(fsavl, fn, where);
	else
	free(fn);
	}
	}

	return (fsavl);
	}

	/*
	* Routines for dealing with the giant nvlist of fs-nvlists, etc.
	*/
	typedef struct send_data {
	/*
	* assigned inside every recursive call,
	* restored from *_save on return:
	*
	* guid of fromsnap snapshot in parent dataset
	* txg of fromsnap snapshot in current dataset
	* txg of tosnap snapshot in current dataset
	*/

	uint64_t parent_fromsnap_guid;
	uint64_t fromsnap_txg;
	uint64_t tosnap_txg;

	/* the nvlists get accumulated during depth-first traversal */
	nvlist_t *parent_snaps;
	nvlist_t *fss;
	nvlist_t *snapprops;
	nvlist_t snapholds; / user holds */

	/* send-receive configuration, does not change during traversal */
	const char *fsname;
	const char *fromsnap;
	const char *tosnap;
	boolean_t recursive;
	boolean_t raw;
	boolean_t doall;
	boolean_t replicate;
	boolean_t skipmissing;
	boolean_t verbose;
	boolean_t backup;
	boolean_t seenfrom;
	boolean_t seento;
	boolean_t holds; /* were holds requested with send -h */
	boolean_t props;

	/*
	* The header nvlist is of the following format:
	* {
	* "tosnap" -> string
	* "fromsnap" -> string (if incremental)
	* "fss" -> {
	* id -> {
	*
	* "name" -> string (full name; for debugging)
	* "parentfromsnap" -> number (guid of fromsnap in parent)
	*
	* "props" -> { name -> value (only if set here) }
	* "snaps" -> { name (lastname) -> number (guid) }
	* "snapprops" -> { name (lastname) -> { name -> value } }
	* "snapholds" -> { name (lastname) -> { holdname -> crtime } }
	*
	* "origin" -> number (guid) (if clone)
	* "is_encroot" -> boolean
	* "sent" -> boolean (not on-disk)
	* }
	* }
	* }
	*
	*/
	} send_data_t;

	static void
	send_iterate_prop(zfs_handle_t zhp, boolean_t received_only, nvlist_t nv);

	/*
	* Collect guid, valid props, optionally holds, etc. of a snapshot.
	* This interface is intended for use as a zfs_iter_snapshots_sorted visitor.
	*/
	static int
	send_iterate_snap(zfs_handle_t zhp, void arg)
	{
	send_data_t *sd = arg;
	uint64_t guid = zhp->zfs_dmustats.dds_guid;
	uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
	boolean_t isfromsnap, istosnap, istosnapwithnofrom;
	char *snapname;
	const char *from = sd->fromsnap;
	const char *to = sd->tosnap;

	snapname = strrchr(zhp->zfs_name, '@');
	assert(snapname != NULL);
	++snapname;

	isfromsnap = (from != NULL && strcmp(from, snapname) == 0);
	istosnap = (to != NULL && strcmp(to, snapname) == 0);
	istosnapwithnofrom = (istosnap && from == NULL);

	if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
	if (sd->verbose) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"skipping snapshot %s because it was created "
	"after the destination snapshot (%s)\n"),
	zhp->zfs_name, to);
	}
	zfs_close(zhp);
	return (0);
	}

	fnvlist_add_uint64(sd->parent_snaps, snapname, guid);

	/*
	* NB: if there is no fromsnap here (it's a newly created fs in
	* an incremental replication), we will substitute the tosnap.
	*/
	if (isfromsnap \|\| (sd->parent_fromsnap_guid == 0 && istosnap))
	sd->parent_fromsnap_guid = guid;

	if (!sd->recursive) {
	/*
	* To allow a doall stream to work properly
	* with a NULL fromsnap
	*/
	if (sd->doall && from == NULL && !sd->seenfrom)
	sd->seenfrom = B_TRUE;

	if (!sd->seenfrom && isfromsnap) {
	sd->seenfrom = B_TRUE;
	zfs_close(zhp);
	return (0);
	}

	if ((sd->seento \|\| !sd->seenfrom) && !istosnapwithnofrom) {
	zfs_close(zhp);
	return (0);
	}

	if (istosnap)
	sd->seento = B_TRUE;
	}

	nvlist_t *nv = fnvlist_alloc();
	send_iterate_prop(zhp, sd->backup, nv);
	fnvlist_add_nvlist(sd->snapprops, snapname, nv);
	fnvlist_free(nv);

	if (sd->holds) {
	nvlist_t *holds;
	if (lzc_get_holds(zhp->zfs_name, &holds) == 0) {
	fnvlist_add_nvlist(sd->snapholds, snapname, holds);
	fnvlist_free(holds);
	}
	}

	zfs_close(zhp);
	return (0);
	}

	/*
	* Collect all valid props from the handle snap into an nvlist.
	*/
	static void
	send_iterate_prop(zfs_handle_t zhp, boolean_t received_only, nvlist_t nv)
	{
	nvlist_t *props;

	if (received_only)
	props = zfs_get_recvd_props(zhp);
	else
	props = zhp->zfs_props;

	nvpair_t *elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	char *propname = nvpair_name(elem);
	zfs_prop_t prop = zfs_name_to_prop(propname);

	if (!zfs_prop_user(propname)) {
	/*
	* Realistically, this should never happen. However,
	* we want the ability to add DSL properties without
	* needing to make incompatible version changes. We
	* need to ignore unknown properties to allow older
	* software to still send datasets containing these
	* properties, with the unknown properties elided.
	*/
	if (prop == ZPROP_INVAL)
	continue;

	if (zfs_prop_readonly(prop))
	continue;
	}

	nvlist_t *propnv = fnvpair_value_nvlist(elem);

	boolean_t isspacelimit = (prop == ZFS_PROP_QUOTA \|\|
	prop == ZFS_PROP_RESERVATION \|\|
	prop == ZFS_PROP_REFQUOTA \|\|
	prop == ZFS_PROP_REFRESERVATION);
	if (isspacelimit && zhp->zfs_type == ZFS_TYPE_SNAPSHOT)
	continue;

	char *source;
	if (nvlist_lookup_string(propnv, ZPROP_SOURCE, &source) == 0) {
	if (strcmp(source, zhp->zfs_name) != 0 &&
	strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)
	continue;
	} else {
	/*
	* May have no source before SPA_VERSION_RECVD_PROPS,
	* but is still modifiable.
	*/
	if (!isspacelimit)
	continue;
	}

	if (zfs_prop_user(propname) \|\|
	zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
	char *value;
	value = fnvlist_lookup_string(propnv, ZPROP_VALUE);
	fnvlist_add_string(nv, propname, value);
	} else {
	uint64_t value;
	value = fnvlist_lookup_uint64(propnv, ZPROP_VALUE);
	fnvlist_add_uint64(nv, propname, value);
	}
	}
	}

	/*
	* returns snapshot guid
	* and returns 0 if the snapshot does not exist
	*/
	static uint64_t
	get_snap_guid(libzfs_handle_t hdl, const char fs, const char *snap)
	{
	char name[MAXPATHLEN + 1];
	uint64_t guid = 0;

	if (fs == NULL \|\| fs[0] == '\0' \|\| snap == NULL \|\| snap[0] == '\0')
	return (guid);

	(void) snprintf(name, sizeof (name), "%s@%s", fs, snap);
	zfs_handle_t *zhp = zfs_open(hdl, name, ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	guid = zfs_prop_get_int(zhp, ZFS_PROP_GUID);
	zfs_close(zhp);
	}

	return (guid);
	}

	/*
	* returns snapshot creation txg
	* and returns 0 if the snapshot does not exist
	*/
	static uint64_t
	get_snap_txg(libzfs_handle_t hdl, const char fs, const char *snap)
	{
	char name[ZFS_MAX_DATASET_NAME_LEN];
	uint64_t txg = 0;

	if (fs == NULL \|\| fs[0] == '\0' \|\| snap == NULL \|\| snap[0] == '\0')
	return (txg);

	(void) snprintf(name, sizeof (name), "%s@%s", fs, snap);
	if (zfs_dataset_exists(hdl, name, ZFS_TYPE_SNAPSHOT)) {
	zfs_handle_t *zhp = zfs_open(hdl, name, ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	txg = zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG);
	zfs_close(zhp);
	}
	}

	return (txg);
	}

	/*
	* Recursively generate nvlists describing datasets. See comment
	* for the data structure send_data_t above for description of contents
	* of the nvlist.
	*/
	static int
	send_iterate_fs(zfs_handle_t zhp, void arg)
	{
	send_data_t *sd = arg;
	nvlist_t nvfs = NULL, nv = NULL;
	int rv = 0;
	uint64_t min_txg = 0, max_txg = 0;
	uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
	uint64_t guid = zhp->zfs_dmustats.dds_guid;
	uint64_t fromsnap_txg, tosnap_txg;
	char guidstring[64];

	/* These fields are restored on return from a recursive call. */
	uint64_t parent_fromsnap_guid_save = sd->parent_fromsnap_guid;
	uint64_t fromsnap_txg_save = sd->fromsnap_txg;
	uint64_t tosnap_txg_save = sd->tosnap_txg;

	fromsnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->fromsnap);
	if (fromsnap_txg != 0)
	sd->fromsnap_txg = fromsnap_txg;

	tosnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->tosnap);
	if (tosnap_txg != 0)
	sd->tosnap_txg = tosnap_txg;

	/*
	* On the send side, if the current dataset does not have tosnap,
	* perform two additional checks:
	*
	* - Skip sending the current dataset if it was created later than
	* the parent tosnap.
	* - Return error if the current dataset was created earlier than
	* the parent tosnap, unless --skip-missing specified. Then
	* just print a warning.
	*/
	if (sd->tosnap != NULL && tosnap_txg == 0) {
	if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
	if (sd->verbose) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"skipping dataset %s: snapshot %s does "
	"not exist\n"), zhp->zfs_name, sd->tosnap);
	}
	} else if (sd->skipmissing) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: skipping dataset %s and its children:"
	" snapshot %s does not exist\n"),
	zhp->zfs_name, sd->tosnap);
	} else {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"cannot send %s@%s%s: snapshot %s@%s does not "
	"exist\n"), sd->fsname, sd->tosnap, sd->recursive ?
	dgettext(TEXT_DOMAIN, " recursively") : "",
	zhp->zfs_name, sd->tosnap);
	rv = EZFS_NOENT;
	}
	goto out;
	}

	nvfs = fnvlist_alloc();
	fnvlist_add_string(nvfs, "name", zhp->zfs_name);
	fnvlist_add_uint64(nvfs, "parentfromsnap", sd->parent_fromsnap_guid);

	if (zhp->zfs_dmustats.dds_origin[0] != '\0') {
	zfs_handle_t *origin = zfs_open(zhp->zfs_hdl,
	zhp->zfs_dmustats.dds_origin, ZFS_TYPE_SNAPSHOT);
	if (origin == NULL) {
	rv = -1;
	goto out;
	}
	fnvlist_add_uint64(nvfs, "origin",
	origin->zfs_dmustats.dds_guid);
	zfs_close(origin);
	}

	/* Iterate over props. */
	if (sd->props \|\| sd->backup \|\| sd->recursive) {
	nv = fnvlist_alloc();
	send_iterate_prop(zhp, sd->backup, nv);
	fnvlist_add_nvlist(nvfs, "props", nv);
	}
	if (zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF) {
	boolean_t encroot;

	/* Determine if this dataset is an encryption root. */
	if (zfs_crypto_get_encryption_root(zhp, &encroot, NULL) != 0) {
	rv = -1;
	goto out;
	}

	if (encroot)
	fnvlist_add_boolean(nvfs, "is_encroot");

	/*
	* Encrypted datasets can only be sent with properties if
	* the raw flag is specified because the receive side doesn't
	* currently have a mechanism for recursively asking the user
	* for new encryption parameters.
	*/
	if (!sd->raw) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"cannot send %s@%s: encrypted dataset %s may not "
	"be sent with properties without the raw flag\n"),
	sd->fsname, sd->tosnap, zhp->zfs_name);
	rv = -1;
	goto out;
	}

	}

	/*
	* Iterate over snaps, and set sd->parent_fromsnap_guid.
	*
	* If this is a "doall" send, a replicate send or we're just trying
	* to gather a list of previous snapshots, iterate through all the
	* snaps in the txg range. Otherwise just look at the one we're
	* interested in.
	*/
	sd->parent_fromsnap_guid = 0;
	sd->parent_snaps = fnvlist_alloc();
	sd->snapprops = fnvlist_alloc();
	if (sd->holds)
	sd->snapholds = fnvlist_alloc();
	if (sd->doall \|\| sd->replicate \|\| sd->tosnap == NULL) {
	if (!sd->replicate && fromsnap_txg != 0)
	min_txg = fromsnap_txg;
	if (!sd->replicate && tosnap_txg != 0)
	max_txg = tosnap_txg;
	(void) zfs_iter_snapshots_sorted(zhp, send_iterate_snap, sd,
	min_txg, max_txg);
	} else {
	char snapname[MAXPATHLEN];
	zfs_handle_t *snap;

	(void) snprintf(snapname, sizeof (snapname), "%s@%s",
	zhp->zfs_name, sd->tosnap);
	if (sd->fromsnap != NULL)
	sd->seenfrom = B_TRUE;
	snap = zfs_open(zhp->zfs_hdl, snapname, ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	(void) send_iterate_snap(snap, sd);
	}

	fnvlist_add_nvlist(nvfs, "snaps", sd->parent_snaps);
	fnvlist_free(sd->parent_snaps);
	fnvlist_add_nvlist(nvfs, "snapprops", sd->snapprops);
	fnvlist_free(sd->snapprops);
	if (sd->holds) {
	fnvlist_add_nvlist(nvfs, "snapholds", sd->snapholds);
	fnvlist_free(sd->snapholds);
	}

	/* Do not allow the size of the properties list to exceed the limit */
	if ((fnvlist_size(nvfs) + fnvlist_size(sd->fss)) >
	zhp->zfs_hdl->libzfs_max_nvlist) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"warning: cannot send %s@%s: the size of the list of "
	"snapshots and properties is too large to be received "
	"successfully.\n"
	"Select a smaller number of snapshots to send.\n"),
	zhp->zfs_name, sd->tosnap);
	rv = EZFS_NOSPC;
	goto out;
	}
	/* Add this fs to nvlist. */
	(void) snprintf(guidstring, sizeof (guidstring),
	"0x%llx", (longlong_t)guid);
	fnvlist_add_nvlist(sd->fss, guidstring, nvfs);

	/* Iterate over children. */
	if (sd->recursive)
	rv = zfs_iter_filesystems(zhp, send_iterate_fs, sd);

	out:
	/* Restore saved fields. */
	sd->parent_fromsnap_guid = parent_fromsnap_guid_save;
	sd->fromsnap_txg = fromsnap_txg_save;
	sd->tosnap_txg = tosnap_txg_save;

	fnvlist_free(nv);
	fnvlist_free(nvfs);

	zfs_close(zhp);
	return (rv);
	}

	static int
	gather_nvlist(libzfs_handle_t hdl, const char fsname, const char *fromsnap,
	const char *tosnap, boolean_t recursive, boolean_t raw, boolean_t doall,
	boolean_t replicate, boolean_t skipmissing, boolean_t verbose,
	boolean_t backup, boolean_t holds, boolean_t props, nvlist_t **nvlp,
	avl_tree_t **avlp)
	{
	zfs_handle_t *zhp;
	send_data_t sd = { 0 };
	int error;

	zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return (EZFS_BADTYPE);

	sd.fss = fnvlist_alloc();
	sd.fsname = fsname;
	sd.fromsnap = fromsnap;
	sd.tosnap = tosnap;
	sd.recursive = recursive;
	sd.raw = raw;
	sd.doall = doall;
	sd.replicate = replicate;
	sd.skipmissing = skipmissing;
	sd.verbose = verbose;
	sd.backup = backup;
	sd.holds = holds;
	sd.props = props;

	if ((error = send_iterate_fs(zhp, &sd)) != 0) {
	fnvlist_free(sd.fss);
	if (avlp != NULL)
	*avlp = NULL;
	*nvlp = NULL;
	return (error);
	}

	if (avlp != NULL && (*avlp = fsavl_create(sd.fss)) == NULL) {
	fnvlist_free(sd.fss);
	*nvlp = NULL;
	return (EZFS_NOMEM);
	}

	*nvlp = sd.fss;
	return (0);
	}

	/*
	* Routines specific to "zfs send"
	*/
	typedef struct send_dump_data {
	/* these are all just the short snapname (the part after the @) */
	const char *fromsnap;
	const char *tosnap;
	char prevsnap[ZFS_MAX_DATASET_NAME_LEN];
	uint64_t prevsnap_obj;
	boolean_t seenfrom, seento, replicate, doall, fromorigin;
	boolean_t dryrun, parsable, progress, embed_data, std_out;
	boolean_t large_block, compress, raw, holds;
	int outfd;
	boolean_t err;
	nvlist_t *fss;
	nvlist_t *snapholds;
	avl_tree_t *fsavl;
	snapfilter_cb_t *filter_cb;
	void *filter_cb_arg;
	nvlist_t *debugnv;
	char holdtag[ZFS_MAX_DATASET_NAME_LEN];
	int cleanup_fd;
	int verbosity;
	uint64_t size;
	} send_dump_data_t;

	static int
	zfs_send_space(zfs_handle_t zhp, const char snapname, const char *from,
	enum lzc_send_flags flags, uint64_t *spacep)
	{
	assert(snapname != NULL);

	int error = lzc_send_space(snapname, from, flags, spacep);
	if (error == 0)
	return (0);

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot estimate space for '%s'"), snapname);

	libzfs_handle_t *hdl = zhp->zfs_hdl;
	switch (error) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case ENOENT:
	if (zfs_dataset_exists(hdl, snapname,
	ZFS_TYPE_SNAPSHOT)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (%s) does not exist"),
	snapname);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	case EINVAL:
	zfs_error_aux(hdl, "%s", strerror(error));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, error, errbuf));
	}
	}

	/*
	* Dumps a backup of the given snapshot (incremental from fromsnap if it's not
	* NULL) to the file descriptor specified by outfd.
	*/
	static int
	dump_ioctl(zfs_handle_t zhp, const char fromsnap, uint64_t fromsnap_obj,
	boolean_t fromorigin, int outfd, enum lzc_send_flags flags,
	nvlist_t *debugnv)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t *thisdbg;

	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
	assert(fromsnap_obj == 0 \|\| !fromorigin);

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	zc.zc_cookie = outfd;
	zc.zc_obj = fromorigin;
	zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zc.zc_fromobj = fromsnap_obj;
	zc.zc_flags = flags;

	if (debugnv != NULL) {
	thisdbg = fnvlist_alloc();
	if (fromsnap != NULL && fromsnap[0] != '\0')
	fnvlist_add_string(thisdbg, "fromsnap", fromsnap);
	}

	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) {
	char errbuf[ERRBUFLEN];
	int error = errno;

	(void) snprintf(errbuf, sizeof (errbuf), "%s '%s'",
	dgettext(TEXT_DOMAIN, "warning: cannot send"),
	zhp->zfs_name);

	if (debugnv != NULL) {
	fnvlist_add_uint64(thisdbg, "error", error);
	fnvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg);
	fnvlist_free(thisdbg);
	}

	switch (error) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"source key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));

	case ENOENT:
	if (zfs_dataset_exists(hdl, zc.zc_name,
	ZFS_TYPE_SNAPSHOT)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (@%s) does not exist"),
	zc.zc_value);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	case EINVAL:
	zfs_error_aux(hdl, "%s", strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	}

	if (debugnv != NULL) {
	fnvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg);
	fnvlist_free(thisdbg);
	}

	return (0);
	}

	static void
	gather_holds(zfs_handle_t zhp, send_dump_data_t sdd)
	{
	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);

	/*
	* zfs_send() only sets snapholds for sends that need them,
	* e.g. replication and doall.
	*/
	if (sdd->snapholds == NULL)
	return;

	fnvlist_add_string(sdd->snapholds, zhp->zfs_name, sdd->holdtag);
	}

	int
	zfs_send_progress(zfs_handle_t zhp, int fd, uint64_t bytes_written,
	uint64_t *blocks_visited)
	{
	zfs_cmd_t zc = {"\0"};

	if (bytes_written != NULL)
	*bytes_written = 0;
	if (blocks_visited != NULL)
	*blocks_visited = 0;
	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	zc.zc_cookie = fd;
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND_PROGRESS, &zc) != 0)
	return (errno);
	if (bytes_written != NULL)
	*bytes_written = zc.zc_cookie;
	if (blocks_visited != NULL)
	*blocks_visited = zc.zc_objset_type;
	return (0);
	}

	static void *
	send_progress_thread(void *arg)
	{
	progress_arg_t *pa = arg;
	zfs_handle_t *zhp = pa->pa_zhp;
	uint64_t bytes;
	uint64_t blocks;
	char buf[16];
	time_t t;
	struct tm tm;
	int err;

	if (!pa->pa_parsable) {
	(void) fprintf(stderr,
	"TIME %s %sSNAPSHOT %s\n",
	pa->pa_estimate ? "BYTES" : " SENT",
	pa->pa_verbosity >= 2 ? " BLOCKS " : "",
	zhp->zfs_name);
	}

	/*
	* Print the progress from ZFS_IOC_SEND_PROGRESS every second.
	*/
	for (;;) {
	(void) sleep(1);
	if ((err = zfs_send_progress(zhp, pa->pa_fd, &bytes,
	&blocks)) != 0) {
	if (err == EINTR \|\| err == ENOENT)
	return ((void *)0);
	return ((void *)(uintptr_t)err);
	}

	(void) time(&t);
	localtime_r(&t, &tm);

	if (pa->pa_verbosity >= 2 && pa->pa_parsable) {
	(void) fprintf(stderr,
	"%02d:%02d:%02d\t%llu\t%llu\t%s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	(u_longlong_t)bytes, (u_longlong_t)blocks,
	zhp->zfs_name);
	} else if (pa->pa_verbosity >= 2) {
	zfs_nicenum(bytes, buf, sizeof (buf));
	(void) fprintf(stderr,
	"%02d:%02d:%02d %5s %8llu %s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	buf, (u_longlong_t)blocks, zhp->zfs_name);
	} else if (pa->pa_parsable) {
	(void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	(u_longlong_t)bytes, zhp->zfs_name);
	} else {
	zfs_nicebytes(bytes, buf, sizeof (buf));
	(void) fprintf(stderr, "%02d:%02d:%02d %5s %s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	buf, zhp->zfs_name);
	}
	}
	}

	static boolean_t
	send_progress_thread_exit(libzfs_handle_t *hdl, pthread_t ptid)
	{
	void *status = NULL;
	(void) pthread_cancel(ptid);
	(void) pthread_join(ptid, &status);
	int error = (int)(uintptr_t)status;
	if (error != 0 && status != PTHREAD_CANCELED)
	return (zfs_standard_error(hdl, error,
	dgettext(TEXT_DOMAIN, "progress thread exited nonzero")));
	else
	return (B_FALSE);
	}

	static void
	send_print_verbose(FILE fout, const char tosnap, const char *fromsnap,
	uint64_t size, boolean_t parsable)
	{
	if (parsable) {
	if (fromsnap != NULL) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"incremental\t%s\t%s"), fromsnap, tosnap);
	} else {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"full\t%s"), tosnap);
	}
	(void) fprintf(fout, "\t%llu", (longlong_t)size);
	} else {
	if (fromsnap != NULL) {
	if (strchr(fromsnap, '@') == NULL &&
	strchr(fromsnap, '#') == NULL) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"send from @%s to %s"), fromsnap, tosnap);
	} else {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"send from %s to %s"), fromsnap, tosnap);
	}
	} else {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"full send of %s"), tosnap);
	}
	if (size != 0) {
	char buf[16];
	zfs_nicebytes(size, buf, sizeof (buf));
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	" estimated size is %s"), buf);
	}
	}
	(void) fprintf(fout, "\n");
	}

	/*
	* Send a single filesystem snapshot, updating the send dump data.
	* This interface is intended for use as a zfs_iter_snapshots_sorted visitor.
	*/
	static int
	dump_snapshot(zfs_handle_t zhp, void arg)
	{
	send_dump_data_t *sdd = arg;
	progress_arg_t pa = { 0 };
	pthread_t tid;
	char *thissnap;
	enum lzc_send_flags flags = 0;
	int err;
	boolean_t isfromsnap, istosnap, fromorigin;
	boolean_t exclude = B_FALSE;
	FILE *fout = sdd->std_out ? stdout : stderr;

	err = 0;
	thissnap = strchr(zhp->zfs_name, '@') + 1;
	isfromsnap = (sdd->fromsnap != NULL &&
	strcmp(sdd->fromsnap, thissnap) == 0);

	if (!sdd->seenfrom && isfromsnap) {
	gather_holds(zhp, sdd);
	sdd->seenfrom = B_TRUE;
	(void) strlcpy(sdd->prevsnap, thissnap, sizeof (sdd->prevsnap));
	sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zfs_close(zhp);
	return (0);
	}

	if (sdd->seento \|\| !sdd->seenfrom) {
	zfs_close(zhp);
	return (0);
	}

	istosnap = (strcmp(sdd->tosnap, thissnap) == 0);
	if (istosnap)
	sdd->seento = B_TRUE;

	if (sdd->large_block)
	flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (sdd->embed_data)
	flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (sdd->compress)
	flags \|= LZC_SEND_FLAG_COMPRESS;
	if (sdd->raw)
	flags \|= LZC_SEND_FLAG_RAW;

	if (!sdd->doall && !isfromsnap && !istosnap) {
	if (sdd->replicate) {
	char *snapname;
	nvlist_t *snapprops;
	/*
	* Filter out all intermediate snapshots except origin
	* snapshots needed to replicate clones.
	*/
	nvlist_t *nvfs = fsavl_find(sdd->fsavl,
	zhp->zfs_dmustats.dds_guid, &snapname);

	if (nvfs != NULL) {
	snapprops = fnvlist_lookup_nvlist(nvfs,
	"snapprops");
	snapprops = fnvlist_lookup_nvlist(snapprops,
	thissnap);
	exclude = !nvlist_exists(snapprops,
	"is_clone_origin");
	}
	} else {
	exclude = B_TRUE;
	}
	}

	/*
	* If a filter function exists, call it to determine whether
	* this snapshot will be sent.
	*/
	if (exclude \|\| (sdd->filter_cb != NULL &&
	sdd->filter_cb(zhp, sdd->filter_cb_arg) == B_FALSE)) {
	/*
	* This snapshot is filtered out. Don't send it, and don't
	* set prevsnap_obj, so it will be as if this snapshot didn't
	* exist, and the next accepted snapshot will be sent as
	* an incremental from the last accepted one, or as the
	* first (and full) snapshot in the case of a replication,
	* non-incremental send.
	*/
	zfs_close(zhp);
	return (0);
	}

	gather_holds(zhp, sdd);
	fromorigin = sdd->prevsnap[0] == '\0' &&
	(sdd->fromorigin \|\| sdd->replicate);

	if (sdd->verbosity != 0) {
	uint64_t size = 0;
	char fromds[ZFS_MAX_DATASET_NAME_LEN];

	if (sdd->prevsnap[0] != '\0') {
	(void) strlcpy(fromds, zhp->zfs_name, sizeof (fromds));
	*(strchr(fromds, '@') + 1) = '\0';
	(void) strlcat(fromds, sdd->prevsnap, sizeof (fromds));
	}
	if (zfs_send_space(zhp, zhp->zfs_name,
	sdd->prevsnap[0] ? fromds : NULL, flags, &size) == 0) {
	send_print_verbose(fout, zhp->zfs_name,
	sdd->prevsnap[0] ? sdd->prevsnap : NULL,
	size, sdd->parsable);
	sdd->size += size;
	}
	}

	if (!sdd->dryrun) {
	/*
	* If progress reporting is requested, spawn a new thread to
	* poll ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	if (sdd->progress) {
	pa.pa_zhp = zhp;
	pa.pa_fd = sdd->outfd;
	pa.pa_parsable = sdd->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = sdd->verbosity;

	if ((err = pthread_create(&tid, NULL,
	send_progress_thread, &pa)) != 0) {
	zfs_close(zhp);
	return (err);
	}
	}

	err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj,
	fromorigin, sdd->outfd, flags, sdd->debugnv);

	if (sdd->progress &&
	send_progress_thread_exit(zhp->zfs_hdl, tid))
	return (-1);
	}

	- (void) strcpy(sdd->prevsnap, thissnap);
	+ (void) strlcpy(sdd->prevsnap, thissnap, sizeof (sdd->prevsnap));
	sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zfs_close(zhp);
	return (err);
	}

	/*
	* Send all snapshots for a filesystem, updating the send dump data.
	*/
	static int
	dump_filesystem(zfs_handle_t zhp, send_dump_data_t sdd)
	{
	int rv = 0;
	boolean_t missingfrom = B_FALSE;
	zfs_cmd_t zc = {"\0"};
	uint64_t min_txg = 0, max_txg = 0;

	/*
	* Make sure the tosnap exists.
	*/
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
	zhp->zfs_name, sdd->tosnap);
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s: does not exist\n"),
	zhp->zfs_name, sdd->tosnap);
	sdd->err = B_TRUE;
	return (0);
	}

	/*
	* If this fs does not have fromsnap, and we're doing
	* recursive, we need to send a full stream from the
	* beginning (or an incremental from the origin if this
	* is a clone). If we're doing non-recursive, then let
	* them get the error.
	*/
	if (sdd->replicate && sdd->fromsnap) {
	/*
	* Make sure the fromsnap exists.
	*/
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
	zhp->zfs_name, sdd->fromsnap);
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0)
	missingfrom = B_TRUE;
	}

	sdd->seenfrom = sdd->seento = B_FALSE;
	sdd->prevsnap[0] = '\0';
	sdd->prevsnap_obj = 0;
	if (sdd->fromsnap == NULL \|\| missingfrom)
	sdd->seenfrom = B_TRUE;

	/*
	* Iterate through all snapshots and process the ones we will be
	* sending. If we only have a "from" and "to" snapshot to deal
	* with, we can avoid iterating through all the other snapshots.
	*/
	if (sdd->doall \|\| sdd->replicate \|\| sdd->tosnap == NULL) {
	if (!sdd->replicate) {
	if (sdd->fromsnap != NULL) {
	min_txg = get_snap_txg(zhp->zfs_hdl,
	zhp->zfs_name, sdd->fromsnap);
	}
	if (sdd->tosnap != NULL) {
	max_txg = get_snap_txg(zhp->zfs_hdl,
	zhp->zfs_name, sdd->tosnap);
	}
	}
	rv = zfs_iter_snapshots_sorted(zhp, dump_snapshot, sdd,
	min_txg, max_txg);
	} else {
	char snapname[MAXPATHLEN] = { 0 };
	zfs_handle_t *snap;

	/* Dump fromsnap. */
	if (!sdd->seenfrom) {
	(void) snprintf(snapname, sizeof (snapname),
	"%s@%s", zhp->zfs_name, sdd->fromsnap);
	snap = zfs_open(zhp->zfs_hdl, snapname,
	ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	rv = dump_snapshot(snap, sdd);
	else
	rv = -1;
	}

	/* Dump tosnap. */
	if (rv == 0) {
	(void) snprintf(snapname, sizeof (snapname),
	"%s@%s", zhp->zfs_name, sdd->tosnap);
	snap = zfs_open(zhp->zfs_hdl, snapname,
	ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	rv = dump_snapshot(snap, sdd);
	else
	rv = -1;
	}
	}

	if (!sdd->seenfrom) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s:\n"
	"incremental source (%s@%s) does not exist\n"),
	zhp->zfs_name, sdd->tosnap,
	zhp->zfs_name, sdd->fromsnap);
	sdd->err = B_TRUE;
	} else if (!sdd->seento) {
	if (sdd->fromsnap) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s:\n"
	"incremental source (%s@%s) "
	"is not earlier than it\n"),
	zhp->zfs_name, sdd->tosnap,
	zhp->zfs_name, sdd->fromsnap);
	} else {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: "
	"could not send %s@%s: does not exist\n"),
	zhp->zfs_name, sdd->tosnap);
	}
	sdd->err = B_TRUE;
	}

	return (rv);
	}

	/*
	* Send all snapshots for all filesystems in sdd.
	*/
	static int
	dump_filesystems(zfs_handle_t rzhp, send_dump_data_t sdd)
	{
	nvpair_t *fspair;
	boolean_t needagain, progress;

	if (!sdd->replicate)
	return (dump_filesystem(rzhp, sdd));

	/* Mark the clone origin snapshots. */
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t *nvfs;
	uint64_t origin_guid = 0;

	nvfs = fnvpair_value_nvlist(fspair);
	(void) nvlist_lookup_uint64(nvfs, "origin", &origin_guid);
	if (origin_guid != 0) {
	char *snapname;
	nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
	origin_guid, &snapname);
	if (origin_nv != NULL) {
	nvlist_t *snapprops;
	snapprops = fnvlist_lookup_nvlist(origin_nv,
	"snapprops");
	snapprops = fnvlist_lookup_nvlist(snapprops,
	snapname);
	fnvlist_add_boolean(snapprops,
	"is_clone_origin");
	}
	}
	}
	again:
	needagain = progress = B_FALSE;
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t fslist, parent_nv;
	char *fsname;
	zfs_handle_t *zhp;
	int err;
	uint64_t origin_guid = 0;
	uint64_t parent_guid = 0;

	fslist = fnvpair_value_nvlist(fspair);
	if (nvlist_lookup_boolean(fslist, "sent") == 0)
	continue;

	fsname = fnvlist_lookup_string(fslist, "name");
	(void) nvlist_lookup_uint64(fslist, "origin", &origin_guid);
	(void) nvlist_lookup_uint64(fslist, "parentfromsnap",
	&parent_guid);

	if (parent_guid != 0) {
	parent_nv = fsavl_find(sdd->fsavl, parent_guid, NULL);
	if (!nvlist_exists(parent_nv, "sent")) {
	/* Parent has not been sent; skip this one. */
	needagain = B_TRUE;
	continue;
	}
	}

	if (origin_guid != 0) {
	nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
	origin_guid, NULL);
	if (origin_nv != NULL &&
	!nvlist_exists(origin_nv, "sent")) {
	/*
	* Origin has not been sent yet;
	* skip this clone.
	*/
	needagain = B_TRUE;
	continue;
	}
	}

	zhp = zfs_open(rzhp->zfs_hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (-1);
	err = dump_filesystem(zhp, sdd);
	fnvlist_add_boolean(fslist, "sent");
	progress = B_TRUE;
	zfs_close(zhp);
	if (err)
	return (err);
	}
	if (needagain) {
	assert(progress);
	goto again;
	}

	/* Clean out the sent flags in case we reuse this fss. */
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t *fslist;

	fslist = fnvpair_value_nvlist(fspair);
	(void) nvlist_remove_all(fslist, "sent");
	}

	return (0);
	}

	nvlist_t *
	zfs_send_resume_token_to_nvlist(libzfs_handle_t hdl, const char token)
	{
	unsigned int version;
	int nread, i;
	unsigned long long checksum, packed_len;

	/*
	* Decode token header, which is:
	* <token version>-<checksum of payload>-<uncompressed payload length>
	* Note that the only supported token version is 1.
	*/
	nread = sscanf(token, "%u-%llx-%llx-",
	&version, &checksum, &packed_len);
	if (nread != 3) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (invalid format)"));
	return (NULL);
	}

	if (version != ZFS_SEND_RESUME_TOKEN_VERSION) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (invalid version %u)"),
	version);
	return (NULL);
	}

	/* Convert hexadecimal representation to binary. */
	token = strrchr(token, '-') + 1;
	int len = strlen(token) / 2;
	unsigned char *compressed = zfs_alloc(hdl, len);
	for (i = 0; i < len; i++) {
	nread = sscanf(token + i * 2, "%2hhx", compressed + i);
	if (nread != 1) {
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt "
	"(payload is not hex-encoded)"));
	return (NULL);
	}
	}

	/* Verify checksum. */
	zio_cksum_t cksum;
	fletcher_4_native_varsize(compressed, len, &cksum);
	if (cksum.zc_word[0] != checksum) {
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (incorrect checksum)"));
	return (NULL);
	}

	/* Uncompress. */
	void *packed = zfs_alloc(hdl, packed_len);
	uLongf packed_len_long = packed_len;
	if (uncompress(packed, &packed_len_long, compressed, len) != Z_OK \|\|
	packed_len_long != packed_len) {
	free(packed);
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (decompression failed)"));
	return (NULL);
	}

	/* Unpack nvlist. */
	nvlist_t *nv;
	int error = nvlist_unpack(packed, packed_len, &nv, KM_SLEEP);
	free(packed);
	free(compressed);
	if (error != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (nvlist_unpack failed)"));
	return (NULL);
	}
	return (nv);
	}

	static enum lzc_send_flags
	lzc_flags_from_sendflags(const sendflags_t *flags)
	{
	enum lzc_send_flags lzc_flags = 0;

	if (flags->largeblock)
	lzc_flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (flags->embed_data)
	lzc_flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (flags->compress)
	lzc_flags \|= LZC_SEND_FLAG_COMPRESS;
	if (flags->raw)
	lzc_flags \|= LZC_SEND_FLAG_RAW;
	if (flags->saved)
	lzc_flags \|= LZC_SEND_FLAG_SAVED;

	return (lzc_flags);
	}

	static int
	estimate_size(zfs_handle_t zhp, const char from, int fd, sendflags_t *flags,
	uint64_t resumeobj, uint64_t resumeoff, uint64_t bytes,
	const char redactbook, char errbuf)
	{
	uint64_t size;
	FILE *fout = flags->dryrun ? stdout : stderr;
	progress_arg_t pa = { 0 };
	int err = 0;
	pthread_t ptid;

	if (flags->progress) {
	pa.pa_zhp = zhp;
	pa.pa_fd = fd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_TRUE;
	pa.pa_verbosity = flags->verbosity;

	err = pthread_create(&ptid, NULL,
	send_progress_thread, &pa);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(errno));
	return (zfs_error(zhp->zfs_hdl,
	EZFS_THREADCREATEFAILED, errbuf));
	}
	}

	err = lzc_send_space_resume_redacted(zhp->zfs_name, from,
	lzc_flags_from_sendflags(flags), resumeobj, resumeoff, bytes,
	redactbook, fd, &size);

	if (flags->progress && send_progress_thread_exit(zhp->zfs_hdl, ptid))
	return (-1);

	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	send_print_verbose(fout, zhp->zfs_name, from, size,
	flags->parsable);

	if (flags->parsable) {
	(void) fprintf(fout, "size\t%llu\n", (longlong_t)size);
	} else {
	char buf[16];
	zfs_nicenum(size, buf, sizeof (buf));
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"total estimated size is %s\n"), buf);
	}
	return (0);
	}

	static boolean_t
	redact_snaps_contains(const uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
	{
	for (int i = 0; i < num_snaps; i++) {
	if (snaps[i] == guid)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static boolean_t
	redact_snaps_equal(const uint64_t *snaps1, uint64_t num_snaps1,
	const uint64_t *snaps2, uint64_t num_snaps2)
	{
	if (num_snaps1 != num_snaps2)
	return (B_FALSE);
	for (int i = 0; i < num_snaps1; i++) {
	if (!redact_snaps_contains(snaps2, num_snaps2, snaps1[i]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	static int
	get_bookmarks(const char path, nvlist_t *bmarksp)
	{
	nvlist_t *props = fnvlist_alloc();
	int error;

	fnvlist_add_boolean(props, "redact_complete");
	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	error = lzc_get_bookmarks(path, props, bmarksp);
	fnvlist_free(props);
	return (error);
	}

	static nvpair_t *
	find_redact_pair(nvlist_t bmarks, const uint64_t redact_snap_guids,
	int num_redact_snaps)
	{
	nvpair_t *pair;

	for (pair = nvlist_next_nvpair(bmarks, NULL); pair;
	pair = nvlist_next_nvpair(bmarks, pair)) {

	nvlist_t *bmark = fnvpair_value_nvlist(pair);
	nvlist_t *vallist = fnvlist_lookup_nvlist(bmark,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	uint_t len = 0;
	uint64_t *bmarksnaps = fnvlist_lookup_uint64_array(vallist,
	ZPROP_VALUE, &len);
	if (redact_snaps_equal(redact_snap_guids,
	num_redact_snaps, bmarksnaps, len)) {
	break;
	}
	}
	return (pair);
	}

	static boolean_t
	get_redact_complete(nvpair_t *pair)
	{
	nvlist_t *bmark = fnvpair_value_nvlist(pair);
	nvlist_t *vallist = fnvlist_lookup_nvlist(bmark, "redact_complete");
	boolean_t complete = fnvlist_lookup_boolean_value(vallist,
	ZPROP_VALUE);

	return (complete);
	}

	/*
	* Check that the list of redaction snapshots in the bookmark matches the send
	* we're resuming, and return whether or not it's complete.
	*
	* Note that the caller needs to free the contents of *bookname with free() if
	* this function returns successfully.
	*/
	static int
	find_redact_book(libzfs_handle_t hdl, const char path,
	const uint64_t *redact_snap_guids, int num_redact_snaps,
	char **bookname)
	{
	char errbuf[ERRBUFLEN];
	nvlist_t *bmarks;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	int error = get_bookmarks(path, &bmarks);
	if (error != 0) {
	if (error == ESRCH) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"nonexistent redaction bookmark provided"));
	} else if (error == ENOENT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset to be sent no longer exists"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"unknown error: %s"), strerror(error));
	}
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	nvpair_t *pair = find_redact_pair(bmarks, redact_snap_guids,
	num_redact_snaps);
	if (pair == NULL) {
	fnvlist_free(bmarks);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no appropriate redaction bookmark exists"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	boolean_t complete = get_redact_complete(pair);
	if (!complete) {
	fnvlist_free(bmarks);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incomplete redaction bookmark provided"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	*bookname = strndup(nvpair_name(pair), ZFS_MAX_DATASET_NAME_LEN);
	ASSERT3P(*bookname, !=, NULL);
	fnvlist_free(bmarks);
	return (0);
	}

	static enum lzc_send_flags
	lzc_flags_from_resume_nvl(nvlist_t *resume_nvl)
	{
	enum lzc_send_flags lzc_flags = 0;

	if (nvlist_exists(resume_nvl, "largeblockok"))
	lzc_flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (nvlist_exists(resume_nvl, "embedok"))
	lzc_flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (nvlist_exists(resume_nvl, "compressok"))
	lzc_flags \|= LZC_SEND_FLAG_COMPRESS;
	if (nvlist_exists(resume_nvl, "rawok"))
	lzc_flags \|= LZC_SEND_FLAG_RAW;
	if (nvlist_exists(resume_nvl, "savedok"))
	lzc_flags \|= LZC_SEND_FLAG_SAVED;

	return (lzc_flags);
	}

	static int
	zfs_send_resume_impl_cb_impl(libzfs_handle_t hdl, sendflags_t flags,
	int outfd, nvlist_t *resume_nvl)
	{
	char errbuf[ERRBUFLEN];
	char *toname;
	char *fromname = NULL;
	uint64_t resumeobj, resumeoff, toguid, fromguid, bytes;
	zfs_handle_t *zhp;
	int error = 0;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	FILE *fout = (flags->verbosity > 0 && flags->dryrun) ? stdout : stderr;
	uint64_t *redact_snap_guids = NULL;
	int num_redact_snaps = 0;
	char *redact_book = NULL;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	if (flags->verbosity != 0) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"resume token contents:\n"));
	nvlist_print(fout, resume_nvl);
	}

	if (nvlist_lookup_string(resume_nvl, "toname", &toname) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "object", &resumeobj) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "offset", &resumeoff) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "toguid", &toguid) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt"));
	return (zfs_error(hdl, EZFS_FAULT, errbuf));
	}
	fromguid = 0;
	(void) nvlist_lookup_uint64(resume_nvl, "fromguid", &fromguid);

	if (flags->saved) {
	- (void) strcpy(name, toname);
	+ (void) strlcpy(name, toname, sizeof (name));
	} else {
	error = guid_to_name(hdl, toname, toguid, B_FALSE, name);
	if (error != 0) {
	if (zfs_dataset_exists(hdl, toname, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is no longer the same snapshot "
	"used in the initial send"), toname);
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' used in the initial send no "
	"longer exists"), toname);
	}
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}
	}

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"unable to access '%s'"), name);
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}

	if (nvlist_lookup_uint64_array(resume_nvl, "book_redact_snaps",
	&redact_snap_guids, (uint_t *)&num_redact_snaps) != 0) {
	num_redact_snaps = -1;
	}

	if (fromguid != 0) {
	if (guid_to_name_redact_snaps(hdl, toname, fromguid, B_TRUE,
	redact_snap_guids, num_redact_snaps, name) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source %#llx no longer exists"),
	(longlong_t)fromguid);
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}
	fromname = name;
	}

	redact_snap_guids = NULL;

	if (nvlist_lookup_uint64_array(resume_nvl,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS), &redact_snap_guids,
	(uint_t *)&num_redact_snaps) == 0) {
	char path[ZFS_MAX_DATASET_NAME_LEN];

	(void) strlcpy(path, toname, sizeof (path));
	char *at = strchr(path, '@');
	ASSERT3P(at, !=, NULL);

	*at = '\0';

	if ((error = find_redact_book(hdl, path, redact_snap_guids,
	num_redact_snaps, &redact_book)) != 0) {
	return (error);
	}
	}

	enum lzc_send_flags lzc_flags = lzc_flags_from_sendflags(flags) \|
	lzc_flags_from_resume_nvl(resume_nvl);

	if (flags->verbosity != 0) {
	/*
	* Some of these may have come from the resume token, set them
	* here for size estimate purposes.
	*/
	sendflags_t tmpflags = *flags;
	if (lzc_flags & LZC_SEND_FLAG_LARGE_BLOCK)
	tmpflags.largeblock = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_COMPRESS)
	tmpflags.compress = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_EMBED_DATA)
	tmpflags.embed_data = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_RAW)
	tmpflags.raw = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_SAVED)
	tmpflags.saved = B_TRUE;
	error = estimate_size(zhp, fromname, outfd, &tmpflags,
	resumeobj, resumeoff, bytes, redact_book, errbuf);
	}

	if (!flags->dryrun) {
	progress_arg_t pa = { 0 };
	pthread_t tid;
	/*
	* If progress reporting is requested, spawn a new thread to
	* poll ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	if (flags->progress) {
	pa.pa_zhp = zhp;
	pa.pa_fd = outfd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = flags->verbosity;

	error = pthread_create(&tid, NULL,
	send_progress_thread, &pa);
	if (error != 0) {
	if (redact_book != NULL)
	free(redact_book);
	zfs_close(zhp);
	return (error);
	}
	}

	error = lzc_send_resume_redacted(zhp->zfs_name, fromname, outfd,
	lzc_flags, resumeobj, resumeoff, redact_book);
	if (redact_book != NULL)
	free(redact_book);

	if (flags->progress && send_progress_thread_exit(hdl, tid))
	return (-1);

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), zhp->zfs_name);

	zfs_close(zhp);

	switch (error) {
	case 0:
	return (0);
	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"source key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
	case ESRCH:
	if (lzc_exists(zhp->zfs_name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source could not be found"));
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EXDEV:
	case ENOENT:
	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	zfs_error_aux(hdl, "%s", strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	} else {
	if (redact_book != NULL)
	free(redact_book);
	}

	zfs_close(zhp);

	return (error);
	}

	struct zfs_send_resume_impl {
	libzfs_handle_t *hdl;
	sendflags_t *flags;
	nvlist_t *resume_nvl;
	};

	static int
	zfs_send_resume_impl_cb(int outfd, void *arg)
	{
	struct zfs_send_resume_impl *zsri = arg;
	return (zfs_send_resume_impl_cb_impl(zsri->hdl, zsri->flags, outfd,
	zsri->resume_nvl));
	}

	static int
	zfs_send_resume_impl(libzfs_handle_t hdl, sendflags_t flags, int outfd,
	nvlist_t *resume_nvl)
	{
	struct zfs_send_resume_impl zsri = {
	.hdl = hdl,
	.flags = flags,
	.resume_nvl = resume_nvl,
	};
	return (lzc_send_wrapper(zfs_send_resume_impl_cb, outfd, &zsri));
	}

	int
	zfs_send_resume(libzfs_handle_t hdl, sendflags_t flags, int outfd,
	const char *resume_token)
	{
	int ret;
	char errbuf[ERRBUFLEN];
	nvlist_t *resume_nvl;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	resume_nvl = zfs_send_resume_token_to_nvlist(hdl, resume_token);
	if (resume_nvl == NULL) {
	/*
	* zfs_error_aux has already been set by
	* zfs_send_resume_token_to_nvlist()
	*/
	return (zfs_error(hdl, EZFS_FAULT, errbuf));
	}

	ret = zfs_send_resume_impl(hdl, flags, outfd, resume_nvl);
	fnvlist_free(resume_nvl);

	return (ret);
	}

	int
	zfs_send_saved(zfs_handle_t zhp, sendflags_t flags, int outfd,
	const char *resume_token)
	{
	int ret;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t saved_nvl = NULL, resume_nvl = NULL;
	uint64_t saved_guid = 0, resume_guid = 0;
	uint64_t obj = 0, off = 0, bytes = 0;
	char token_buf[ZFS_MAXPROPLEN];
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"saved send failed"));

	ret = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	token_buf, sizeof (token_buf), NULL, NULL, 0, B_TRUE);
	if (ret != 0)
	goto out;

	saved_nvl = zfs_send_resume_token_to_nvlist(hdl, token_buf);
	if (saved_nvl == NULL) {
	/*
	* zfs_error_aux has already been set by
	* zfs_send_resume_token_to_nvlist()
	*/
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	/*
	* If a resume token is provided we use the object and offset
	* from that instead of the default, which starts from the
	* beginning.
	*/
	if (resume_token != NULL) {
	resume_nvl = zfs_send_resume_token_to_nvlist(hdl,
	resume_token);
	if (resume_nvl == NULL) {
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (nvlist_lookup_uint64(resume_nvl, "object", &obj) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "offset", &off) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "toguid",
	&resume_guid) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"provided resume token is corrupt"));
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (nvlist_lookup_uint64(saved_nvl, "toguid",
	&saved_guid)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset's resume token is corrupt"));
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (resume_guid != saved_guid) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"provided resume token does not match dataset"));
	ret = zfs_error(hdl, EZFS_BADBACKUP, errbuf);
	goto out;
	}
	}

	(void) nvlist_remove_all(saved_nvl, "object");
	fnvlist_add_uint64(saved_nvl, "object", obj);

	(void) nvlist_remove_all(saved_nvl, "offset");
	fnvlist_add_uint64(saved_nvl, "offset", off);

	(void) nvlist_remove_all(saved_nvl, "bytes");
	fnvlist_add_uint64(saved_nvl, "bytes", bytes);

	(void) nvlist_remove_all(saved_nvl, "toname");
	fnvlist_add_string(saved_nvl, "toname", zhp->zfs_name);

	ret = zfs_send_resume_impl(hdl, flags, outfd, saved_nvl);

	out:
	fnvlist_free(saved_nvl);
	fnvlist_free(resume_nvl);
	return (ret);
	}

	/*
	* This function informs the target system that the recursive send is complete.
	* The record is also expected in the case of a send -p.
	*/
	static int
	send_conclusion_record(int fd, zio_cksum_t *zc)
	{
	dmu_replay_record_t drr = { 0 };
	drr.drr_type = DRR_END;
	if (zc != NULL)
	drr.drr_u.drr_end.drr_checksum = *zc;
	if (write(fd, &drr, sizeof (drr)) == -1) {
	return (errno);
	}
	return (0);
	}

	/*
	* This function is responsible for sending the records that contain the
	* necessary information for the target system's libzfs to be able to set the
	* properties of the filesystem being received, or to be able to prepare for
	* a recursive receive.
	*
	* The "zhp" argument is the handle of the snapshot we are sending
	* (the "tosnap"). The "from" argument is the short snapshot name (the part
	* after the @) of the incremental source.
	*/
	static int
	send_prelim_records(zfs_handle_t zhp, const char from, int fd,
	boolean_t gather_props, boolean_t recursive, boolean_t verbose,
	boolean_t dryrun, boolean_t raw, boolean_t replicate, boolean_t skipmissing,
	boolean_t backup, boolean_t holds, boolean_t props, boolean_t doall,
	nvlist_t fssp, avl_tree_t fsavlp)
	{
	int err = 0;
	char *packbuf = NULL;
	size_t buflen = 0;
	zio_cksum_t zc = { {0} };
	int featureflags = 0;
	/* name of filesystem/volume that contains snapshot we are sending */
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	/* short name of snap we are sending */
	const char *tosnap = "";

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), zhp->zfs_name);
	if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM && zfs_prop_get_int(zhp,
	ZFS_PROP_VERSION) >= ZPL_VERSION_SA) {
	featureflags \|= DMU_BACKUP_FEATURE_SA_SPILL;
	}

	if (holds)
	featureflags \|= DMU_BACKUP_FEATURE_HOLDS;

	(void) strlcpy(tofs, zhp->zfs_name, ZFS_MAX_DATASET_NAME_LEN);
	char *at = strchr(tofs, '@');
	if (at != NULL) {
	*at = '\0';
	tosnap = at + 1;
	}

	if (gather_props) {
	nvlist_t *hdrnv = fnvlist_alloc();
	nvlist_t *fss = NULL;

	if (from != NULL)
	fnvlist_add_string(hdrnv, "fromsnap", from);
	fnvlist_add_string(hdrnv, "tosnap", tosnap);
	if (!recursive)
	fnvlist_add_boolean(hdrnv, "not_recursive");

	if (raw) {
	fnvlist_add_boolean(hdrnv, "raw");
	}

	if ((err = gather_nvlist(zhp->zfs_hdl, tofs,
	from, tosnap, recursive, raw, doall, replicate, skipmissing,
	verbose, backup, holds, props, &fss, fsavlp)) != 0) {
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	/*
	* Do not allow the size of the properties list to exceed
	* the limit
	*/
	if ((fnvlist_size(fss) + fnvlist_size(hdrnv)) >
	zhp->zfs_hdl->libzfs_max_nvlist) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "warning: cannot send '%s': "
	"the size of the list of snapshots and properties "
	"is too large to be received successfully.\n"
	"Select a smaller number of snapshots to send.\n"),
	zhp->zfs_name);
	return (zfs_error(zhp->zfs_hdl, EZFS_NOSPC,
	errbuf));
	}
	fnvlist_add_nvlist(hdrnv, "fss", fss);
	VERIFY0(nvlist_pack(hdrnv, &packbuf, &buflen, NV_ENCODE_XDR,
	0));
	if (fssp != NULL) {
	*fssp = fss;
	} else {
	fnvlist_free(fss);
	}
	fnvlist_free(hdrnv);
	}

	if (!dryrun) {
	dmu_replay_record_t drr = { 0 };
	/* write first begin record */
	drr.drr_type = DRR_BEGIN;
	drr.drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC;
	DMU_SET_STREAM_HDRTYPE(drr.drr_u.drr_begin.
	drr_versioninfo, DMU_COMPOUNDSTREAM);
	DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin.
	drr_versioninfo, featureflags);
	if (snprintf(drr.drr_u.drr_begin.drr_toname,
	sizeof (drr.drr_u.drr_begin.drr_toname), "%s@%s", tofs,
	tosnap) >= sizeof (drr.drr_u.drr_begin.drr_toname)) {
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	drr.drr_payloadlen = buflen;

	err = dump_record(&drr, packbuf, buflen, &zc, fd);
	free(packbuf);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	err = send_conclusion_record(fd, &zc);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	}
	return (0);
	}

	/*
	* Generate a send stream. The "zhp" argument is the filesystem/volume
	* that contains the snapshot to send. The "fromsnap" argument is the
	* short name (the part after the '@') of the snapshot that is the
	* incremental source to send from (if non-NULL). The "tosnap" argument
	* is the short name of the snapshot to send.
	*
	* The content of the send stream is the snapshot identified by
	* 'tosnap'. Incremental streams are requested in two ways:
	* - from the snapshot identified by "fromsnap" (if non-null) or
	* - from the origin of the dataset identified by zhp, which must
	* be a clone. In this case, "fromsnap" is null and "fromorigin"
	* is TRUE.
	*
	* The send stream is recursive (i.e. dumps a hierarchy of snapshots) and
	* uses a special header (with a hdrtype field of DMU_COMPOUNDSTREAM)
	* if "replicate" is set. If "doall" is set, dump all the intermediate
	* snapshots. The DMU_COMPOUNDSTREAM header is used in the "doall"
	* case too. If "props" is set, send properties.
	*
	* Pre-wrapped (cf. lzc_send_wrapper()).
	*/
	static int
	zfs_send_cb_impl(zfs_handle_t zhp, const char fromsnap, const char *tosnap,
	sendflags_t *flags, int outfd, snapfilter_cb_t filter_func,
	void cb_arg, nvlist_t *debugnvp)
	{
	char errbuf[ERRBUFLEN];
	send_dump_data_t sdd = { 0 };
	int err = 0;
	nvlist_t *fss = NULL;
	avl_tree_t *fsavl = NULL;
	static uint64_t holdseq;
	int spa_version;
	FILE *fout;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot send '%s'"), zhp->zfs_name);

	if (fromsnap && fromsnap[0] == '\0') {
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"zero-length incremental source"));
	return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));
	}

	if (fromsnap) {
	char full_fromsnap_name[ZFS_MAX_DATASET_NAME_LEN];
	if (snprintf(full_fromsnap_name, sizeof (full_fromsnap_name),
	"%s@%s", zhp->zfs_name, fromsnap) >=
	sizeof (full_fromsnap_name)) {
	err = EINVAL;
	goto stderr_out;
	}
	zfs_handle_t *fromsnapn = zfs_open(zhp->zfs_hdl,
	full_fromsnap_name, ZFS_TYPE_SNAPSHOT);
	if (fromsnapn == NULL) {
	err = -1;
	goto err_out;
	}
	zfs_close(fromsnapn);
	}

	if (flags->replicate \|\| flags->doall \|\| flags->props \|\|
	flags->holds \|\| flags->backup) {
	char full_tosnap_name[ZFS_MAX_DATASET_NAME_LEN];
	if (snprintf(full_tosnap_name, sizeof (full_tosnap_name),
	"%s@%s", zhp->zfs_name, tosnap) >=
	sizeof (full_tosnap_name)) {
	err = EINVAL;
	goto stderr_out;
	}
	zfs_handle_t *tosnap = zfs_open(zhp->zfs_hdl,
	full_tosnap_name, ZFS_TYPE_SNAPSHOT);
	if (tosnap == NULL) {
	err = -1;
	goto err_out;
	}
	err = send_prelim_records(tosnap, fromsnap, outfd,
	flags->replicate \|\| flags->props \|\| flags->holds,
	flags->replicate, flags->verbosity > 0, flags->dryrun,
	flags->raw, flags->replicate, flags->skipmissing,
	flags->backup, flags->holds, flags->props, flags->doall,
	&fss, &fsavl);
	zfs_close(tosnap);
	if (err != 0)
	goto err_out;
	}

	/* dump each stream */
	sdd.fromsnap = fromsnap;
	sdd.tosnap = tosnap;
	sdd.outfd = outfd;
	sdd.replicate = flags->replicate;
	sdd.doall = flags->doall;
	sdd.fromorigin = flags->fromorigin;
	sdd.fss = fss;
	sdd.fsavl = fsavl;
	sdd.verbosity = flags->verbosity;
	sdd.parsable = flags->parsable;
	sdd.progress = flags->progress;
	sdd.dryrun = flags->dryrun;
	sdd.large_block = flags->largeblock;
	sdd.embed_data = flags->embed_data;
	sdd.compress = flags->compress;
	sdd.raw = flags->raw;
	sdd.holds = flags->holds;
	sdd.filter_cb = filter_func;
	sdd.filter_cb_arg = cb_arg;
	if (debugnvp)
	sdd.debugnv = *debugnvp;
	if (sdd.verbosity != 0 && sdd.dryrun)
	sdd.std_out = B_TRUE;
	fout = sdd.std_out ? stdout : stderr;

	/*
	* Some flags require that we place user holds on the datasets that are
	* being sent so they don't get destroyed during the send. We can skip
	* this step if the pool is imported read-only since the datasets cannot
	* be destroyed.
	*/
	if (!flags->dryrun && !zpool_get_prop_int(zfs_get_pool_handle(zhp),
	ZPOOL_PROP_READONLY, NULL) &&
	zfs_spa_version(zhp, &spa_version) == 0 &&
	spa_version >= SPA_VERSION_USERREFS &&
	(flags->doall \|\| flags->replicate)) {
	++holdseq;
	(void) snprintf(sdd.holdtag, sizeof (sdd.holdtag),
	".send-%d-%llu", getpid(), (u_longlong_t)holdseq);
	sdd.cleanup_fd = open(ZFS_DEV, O_RDWR \| O_CLOEXEC);
	if (sdd.cleanup_fd < 0) {
	err = errno;
	goto stderr_out;
	}
	sdd.snapholds = fnvlist_alloc();
	} else {
	sdd.cleanup_fd = -1;
	sdd.snapholds = NULL;
	}

	if (flags->verbosity != 0 \|\| sdd.snapholds != NULL) {
	/*
	* Do a verbose no-op dry run to get all the verbose output
	* or to gather snapshot hold's before generating any data,
	* then do a non-verbose real run to generate the streams.
	*/
	sdd.dryrun = B_TRUE;
	err = dump_filesystems(zhp, &sdd);

	if (err != 0)
	goto stderr_out;

	if (flags->verbosity != 0) {
	if (flags->parsable) {
	(void) fprintf(fout, "size\t%llu\n",
	(longlong_t)sdd.size);
	} else {
	char buf[16];
	zfs_nicebytes(sdd.size, buf, sizeof (buf));
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"total estimated size is %s\n"), buf);
	}
	}

	/* Ensure no snaps found is treated as an error. */
	if (!sdd.seento) {
	err = ENOENT;
	goto err_out;
	}

	/* Skip the second run if dryrun was requested. */
	if (flags->dryrun)
	goto err_out;

	if (sdd.snapholds != NULL) {
	err = zfs_hold_nvl(zhp, sdd.cleanup_fd, sdd.snapholds);
	if (err != 0)
	goto stderr_out;

	fnvlist_free(sdd.snapholds);
	sdd.snapholds = NULL;
	}

	sdd.dryrun = B_FALSE;
	sdd.verbosity = 0;
	}

	err = dump_filesystems(zhp, &sdd);
	fsavl_destroy(fsavl);
	fnvlist_free(fss);

	/* Ensure no snaps found is treated as an error. */
	if (err == 0 && !sdd.seento)
	err = ENOENT;

	if (sdd.cleanup_fd != -1) {
	VERIFY(0 == close(sdd.cleanup_fd));
	sdd.cleanup_fd = -1;
	}

	if (!flags->dryrun && (flags->replicate \|\| flags->doall \|\|
	flags->props \|\| flags->backup \|\| flags->holds)) {
	/*
	* write final end record. NB: want to do this even if
	* there was some error, because it might not be totally
	* failed.
	*/
	int err2 = send_conclusion_record(outfd, NULL);
	if (err2 != 0)
	return (zfs_standard_error(zhp->zfs_hdl, err2, errbuf));
	}

	return (err \|\| sdd.err);

	stderr_out:
	err = zfs_standard_error(zhp->zfs_hdl, err, errbuf);
	err_out:
	fsavl_destroy(fsavl);
	fnvlist_free(fss);
	fnvlist_free(sdd.snapholds);

	if (sdd.cleanup_fd != -1)
	VERIFY(0 == close(sdd.cleanup_fd));
	return (err);
	}

	struct zfs_send {
	zfs_handle_t *zhp;
	const char *fromsnap;
	const char *tosnap;
	sendflags_t *flags;
	snapfilter_cb_t *filter_func;
	void *cb_arg;
	nvlist_t **debugnvp;
	};

	static int
	zfs_send_cb(int outfd, void *arg)
	{
	struct zfs_send *zs = arg;
	return (zfs_send_cb_impl(zs->zhp, zs->fromsnap, zs->tosnap, zs->flags,
	outfd, zs->filter_func, zs->cb_arg, zs->debugnvp));
	}

	int
	zfs_send(zfs_handle_t zhp, const char fromsnap, const char *tosnap,
	sendflags_t *flags, int outfd, snapfilter_cb_t filter_func,
	void cb_arg, nvlist_t *debugnvp)
	{
	struct zfs_send arg = {
	.zhp = zhp,
	.fromsnap = fromsnap,
	.tosnap = tosnap,
	.flags = flags,
	.filter_func = filter_func,
	.cb_arg = cb_arg,
	.debugnvp = debugnvp,
	};
	return (lzc_send_wrapper(zfs_send_cb, outfd, &arg));
	}


	static zfs_handle_t *
	name_to_dir_handle(libzfs_handle_t hdl, const char snapname)
	{
	char dirname[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(dirname, snapname, ZFS_MAX_DATASET_NAME_LEN);
	char *c = strchr(dirname, '@');
	if (c != NULL)
	*c = '\0';
	return (zfs_open(hdl, dirname, ZFS_TYPE_DATASET));
	}

	/*
	* Returns B_TRUE if earlier is an earlier snapshot in later's timeline; either
	* an earlier snapshot in the same filesystem, or a snapshot before later's
	* origin, or it's origin's origin, etc.
	*/
	static boolean_t
	snapshot_is_before(zfs_handle_t earlier, zfs_handle_t later)
	{
	boolean_t ret;
	uint64_t later_txg =
	(later->zfs_type == ZFS_TYPE_FILESYSTEM \|\|
	later->zfs_type == ZFS_TYPE_VOLUME ?
	UINT64_MAX : zfs_prop_get_int(later, ZFS_PROP_CREATETXG));
	uint64_t earlier_txg = zfs_prop_get_int(earlier, ZFS_PROP_CREATETXG);

	if (earlier_txg >= later_txg)
	return (B_FALSE);

	zfs_handle_t *earlier_dir = name_to_dir_handle(earlier->zfs_hdl,
	earlier->zfs_name);
	zfs_handle_t *later_dir = name_to_dir_handle(later->zfs_hdl,
	later->zfs_name);

	if (strcmp(earlier_dir->zfs_name, later_dir->zfs_name) == 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	return (B_TRUE);
	}

	char clonename[ZFS_MAX_DATASET_NAME_LEN];
	if (zfs_prop_get(later_dir, ZFS_PROP_ORIGIN, clonename,
	ZFS_MAX_DATASET_NAME_LEN, NULL, NULL, 0, B_TRUE) != 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	return (B_FALSE);
	}

	zfs_handle_t *origin = zfs_open(earlier->zfs_hdl, clonename,
	ZFS_TYPE_DATASET);
	uint64_t origin_txg = zfs_prop_get_int(origin, ZFS_PROP_CREATETXG);

	/*
	* If "earlier" is exactly the origin, then
	* snapshot_is_before(earlier, origin) will return false (because
	* they're the same).
	*/
	if (origin_txg == earlier_txg &&
	strcmp(origin->zfs_name, earlier->zfs_name) == 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	zfs_close(origin);
	return (B_TRUE);
	}
	zfs_close(earlier_dir);
	zfs_close(later_dir);

	ret = snapshot_is_before(earlier, origin);
	zfs_close(origin);
	return (ret);
	}

	/*
	* The "zhp" argument is the handle of the dataset to send (typically a
	* snapshot). The "from" argument is the full name of the snapshot or
	* bookmark that is the incremental source.
	*
	* Pre-wrapped (cf. lzc_send_wrapper()).
	*/
	static int
	zfs_send_one_cb_impl(zfs_handle_t zhp, const char from, int fd,
	sendflags_t flags, const char redactbook)
	{
	int err;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char *name = zhp->zfs_name;
	pthread_t ptid;
	progress_arg_t pa = { 0 };

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), name);

	if (from != NULL && strchr(from, '@')) {
	zfs_handle_t *from_zhp = zfs_open(hdl, from,
	ZFS_TYPE_DATASET);
	if (from_zhp == NULL)
	return (-1);
	if (!snapshot_is_before(from_zhp, zhp)) {
	zfs_close(from_zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
	}
	zfs_close(from_zhp);
	}

	if (redactbook != NULL) {
	char bookname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *redact_snaps;
	zfs_handle_t *book_zhp;
	char at, pound;
	int dsnamelen;

	pound = strchr(redactbook, '#');
	if (pound != NULL)
	redactbook = pound + 1;
	at = strchr(name, '@');
	if (at == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot do a redacted send to a filesystem"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}
	dsnamelen = at - name;
	if (snprintf(bookname, sizeof (bookname), "%.*s#%s",
	dsnamelen, name, redactbook)
	>= sizeof (bookname)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid bookmark name"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}
	book_zhp = zfs_open(hdl, bookname, ZFS_TYPE_BOOKMARK);
	if (book_zhp == NULL)
	return (-1);
	if (nvlist_lookup_nvlist(book_zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS),
	&redact_snaps) != 0 \|\| redact_snaps == NULL) {
	zfs_close(book_zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not a redaction bookmark"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}
	zfs_close(book_zhp);
	}

	/*
	* Send fs properties
	*/
	if (flags->props \|\| flags->holds \|\| flags->backup) {
	/*
	* Note: the header generated by send_prelim_records()
	* assumes that the incremental source is in the same
	* filesystem/volume as the target (which is a requirement
	* when doing "zfs send -R"). But that isn't always the
	* case here (e.g. send from snap in origin, or send from
	* bookmark). We pass from=NULL, which will omit this
	* information from the prelim records; it isn't used
	* when receiving this type of stream.
	*/
	err = send_prelim_records(zhp, NULL, fd, B_TRUE, B_FALSE,
	flags->verbosity > 0, flags->dryrun, flags->raw,
	flags->replicate, B_FALSE, flags->backup, flags->holds,
	flags->props, flags->doall, NULL, NULL);
	if (err != 0)
	return (err);
	}

	/*
	* Perform size estimate if verbose was specified.
	*/
	if (flags->verbosity != 0) {
	err = estimate_size(zhp, from, fd, flags, 0, 0, 0, redactbook,
	errbuf);
	if (err != 0)
	return (err);
	}

	if (flags->dryrun)
	return (0);

	/*
	* If progress reporting is requested, spawn a new thread to poll
	* ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	if (flags->progress) {
	pa.pa_zhp = zhp;
	pa.pa_fd = fd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = flags->verbosity;

	err = pthread_create(&ptid, NULL,
	send_progress_thread, &pa);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", strerror(errno));
	return (zfs_error(zhp->zfs_hdl,
	EZFS_THREADCREATEFAILED, errbuf));
	}
	}

	err = lzc_send_redacted(name, from, fd,
	lzc_flags_from_sendflags(flags), redactbook);

	if (flags->progress && send_progress_thread_exit(hdl, ptid))
	return (-1);

	if (err == 0 && (flags->props \|\| flags->holds \|\| flags->backup)) {
	/* Write the final end record. */
	err = send_conclusion_record(fd, NULL);
	if (err != 0)
	return (zfs_standard_error(hdl, err, errbuf));
	}
	if (err != 0) {
	switch (errno) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case ENOENT:
	case ESRCH:
	if (lzc_exists(name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (%s) does not exist"),
	from);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"target is busy; if a filesystem, "
	"it must not be mounted"));
	return (zfs_error(hdl, EZFS_BUSY, errbuf));

	case EDQUOT:
	case EFAULT:
	case EFBIG:
	case EINVAL:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EROFS:
	zfs_error_aux(hdl, "%s", strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	}
	return (err != 0);
	}

	struct zfs_send_one {
	zfs_handle_t *zhp;
	const char *from;
	sendflags_t *flags;
	const char *redactbook;
	};

	static int
	zfs_send_one_cb(int fd, void *arg)
	{
	struct zfs_send_one *zso = arg;
	return (zfs_send_one_cb_impl(zso->zhp, zso->from, fd, zso->flags,
	zso->redactbook));
	}

	int
	zfs_send_one(zfs_handle_t zhp, const char from, int fd, sendflags_t *flags,
	const char *redactbook)
	{
	struct zfs_send_one zso = {
	.zhp = zhp,
	.from = from,
	.flags = flags,
	.redactbook = redactbook,
	};
	return (lzc_send_wrapper(zfs_send_one_cb, fd, &zso));
	}

	/*
	* Routines specific to "zfs recv"
	*/

	static int
	recv_read(libzfs_handle_t hdl, int fd, void buf, int ilen,
	boolean_t byteswap, zio_cksum_t *zc)
	{
	char *cp = buf;
	int rv;
	int len = ilen;

	do {
	rv = read(fd, cp, len);
	cp += rv;
	len -= rv;
	} while (rv > 0);

	if (rv < 0 \|\| len != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to read from stream"));
	return (zfs_error(hdl, EZFS_BADSTREAM, dgettext(TEXT_DOMAIN,
	"cannot receive")));
	}

	if (zc) {
	if (byteswap)
	fletcher_4_incremental_byteswap(buf, ilen, zc);
	else
	fletcher_4_incremental_native(buf, ilen, zc);
	}
	return (0);
	}

	static int
	recv_read_nvlist(libzfs_handle_t hdl, int fd, int len, nvlist_t *nvp,
	boolean_t byteswap, zio_cksum_t *zc)
	{
	char *buf;
	int err;

	buf = zfs_alloc(hdl, len);

	if (len > hdl->libzfs_max_nvlist) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "nvlist too large"));
	free(buf);
	return (ENOMEM);
	}

	err = recv_read(hdl, fd, buf, len, byteswap, zc);
	if (err != 0) {
	free(buf);
	return (err);
	}

	err = nvlist_unpack(buf, len, nvp, 0);
	free(buf);
	if (err != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (malformed nvlist)"));
	return (EINVAL);
	}
	return (0);
	}

	/*
	* Returns the grand origin (origin of origin of origin...) of a given handle.
	* If this dataset is not a clone, it simply returns a copy of the original
	* handle.
	*/
	static zfs_handle_t *
	recv_open_grand_origin(zfs_handle_t *zhp)
	{
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	zprop_source_t src;
	zfs_handle_t *ozhp = zfs_handle_dup(zhp);

	while (ozhp != NULL) {
	if (zfs_prop_get(ozhp, ZFS_PROP_ORIGIN, origin,
	sizeof (origin), &src, NULL, 0, B_FALSE) != 0)
	break;

	(void) zfs_close(ozhp);
	ozhp = zfs_open(zhp->zfs_hdl, origin, ZFS_TYPE_FILESYSTEM);
	}

	return (ozhp);
	}

	static int
	recv_rename_impl(zfs_handle_t zhp, const char name, const char *newname)
	{
	int err;
	zfs_handle_t *ozhp = NULL;

	/*
	* Attempt to rename the dataset. If it fails with EACCES we have
	* attempted to rename the dataset outside of its encryption root.
	* Force the dataset to become an encryption root and try again.
	*/
	err = lzc_rename(name, newname);
	if (err == EACCES) {
	ozhp = recv_open_grand_origin(zhp);
	if (ozhp == NULL) {
	err = ENOENT;
	goto out;
	}

	err = lzc_change_key(ozhp->zfs_name, DCP_CMD_FORCE_NEW_KEY,
	NULL, NULL, 0);
	if (err != 0)
	goto out;

	err = lzc_rename(name, newname);
	}

	out:
	if (ozhp != NULL)
	zfs_close(ozhp);
	return (err);
	}

	static int
	recv_rename(libzfs_handle_t hdl, const char name, const char *tryname,
	int baselen, char newname, recvflags_t flags)
	{
	static int seq;
	int err;
	prop_changelist_t *clp = NULL;
	zfs_handle_t *zhp = NULL;

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = -1;
	goto out;
	}
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->force ? MS_FORCE : 0);
	if (clp == NULL) {
	err = -1;
	goto out;
	}
	err = changelist_prefix(clp);
	if (err)
	goto out;

	if (tryname) {
	- (void) strcpy(newname, tryname);
	+ (void) strlcpy(newname, tryname, ZFS_MAX_DATASET_NAME_LEN);
	if (flags->verbose) {
	(void) printf("attempting rename %s to %s\n",
	name, newname);
	}
	err = recv_rename_impl(zhp, name, newname);
	if (err == 0)
	changelist_rename(clp, name, tryname);
	} else {
	err = ENOENT;
	}

	if (err != 0 && strncmp(name + baselen, "recv-", 5) != 0) {
	seq++;

	(void) snprintf(newname, ZFS_MAX_DATASET_NAME_LEN,
	"%.*srecv-%u-%u", baselen, name, getpid(), seq);

	if (flags->verbose) {
	(void) printf("failed - trying rename %s to %s\n",
	name, newname);
	}
	err = recv_rename_impl(zhp, name, newname);
	if (err == 0)
	changelist_rename(clp, name, newname);
	if (err && flags->verbose) {
	(void) printf("failed (%u) - "
	"will try again on next pass\n", errno);
	}
	err = EAGAIN;
	} else if (flags->verbose) {
	if (err == 0)
	(void) printf("success\n");
	else
	(void) printf("failed (%u)\n", errno);
	}

	(void) changelist_postfix(clp);

	out:
	if (clp != NULL)
	changelist_free(clp);
	if (zhp != NULL)
	zfs_close(zhp);

	return (err);
	}

	static int
	recv_promote(libzfs_handle_t hdl, const char fsname,
	const char origin_fsname, recvflags_t flags)
	{
	int err;
	zfs_cmd_t zc = {"\0"};
	zfs_handle_t zhp = NULL, ozhp = NULL;

	if (flags->verbose)
	(void) printf("promoting %s\n", fsname);

	(void) strlcpy(zc.zc_value, origin_fsname, sizeof (zc.zc_value));
	(void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_name));

	/*
	* Attempt to promote the dataset. If it fails with EACCES the
	* promotion would cause this dataset to leave its encryption root.
	* Force the origin to become an encryption root and try again.
	*/
	err = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc);
	if (err == EACCES) {
	zhp = zfs_open(hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = -1;
	goto out;
	}

	ozhp = recv_open_grand_origin(zhp);
	if (ozhp == NULL) {
	err = -1;
	goto out;
	}

	err = lzc_change_key(ozhp->zfs_name, DCP_CMD_FORCE_NEW_KEY,
	NULL, NULL, 0);
	if (err != 0)
	goto out;

	err = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc);
	}

	out:
	if (zhp != NULL)
	zfs_close(zhp);
	if (ozhp != NULL)
	zfs_close(ozhp);

	return (err);
	}

	static int
	recv_destroy(libzfs_handle_t hdl, const char name, int baselen,
	char newname, recvflags_t flags)
	{
	int err = 0;
	prop_changelist_t *clp;
	zfs_handle_t *zhp;
	boolean_t defer = B_FALSE;
	int spa_version;

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (-1);
	zfs_type_t type = zfs_get_type(zhp);
	if (type == ZFS_TYPE_SNAPSHOT &&
	zfs_spa_version(zhp, &spa_version) == 0 &&
	spa_version >= SPA_VERSION_USERREFS)
	defer = B_TRUE;
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->force ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp == NULL)
	return (-1);

	err = changelist_prefix(clp);
	if (err)
	return (err);

	if (flags->verbose)
	(void) printf("attempting destroy %s\n", name);
	if (type == ZFS_TYPE_SNAPSHOT) {
	nvlist_t *nv = fnvlist_alloc();
	fnvlist_add_boolean(nv, name);
	err = lzc_destroy_snaps(nv, defer, NULL);
	fnvlist_free(nv);
	} else {
	err = lzc_destroy(name);
	}
	if (err == 0) {
	if (flags->verbose)
	(void) printf("success\n");
	changelist_remove(clp, name);
	}

	(void) changelist_postfix(clp);
	changelist_free(clp);

	/*
	* Deferred destroy might destroy the snapshot or only mark it to be
	* destroyed later, and it returns success in either case.
	*/
	if (err != 0 \|\| (defer && zfs_dataset_exists(hdl, name,
	ZFS_TYPE_SNAPSHOT))) {
	err = recv_rename(hdl, name, NULL, baselen, newname, flags);
	}

	return (err);
	}

	typedef struct guid_to_name_data {
	uint64_t guid;
	boolean_t bookmark_ok;
	char *name;
	char *skip;
	uint64_t *redact_snap_guids;
	uint64_t num_redact_snaps;
	} guid_to_name_data_t;

	static boolean_t
	redact_snaps_match(zfs_handle_t zhp, guid_to_name_data_t gtnd)
	{
	uint64_t *bmark_snaps;
	uint_t bmark_num_snaps;
	nvlist_t *nvl;
	if (zhp->zfs_type != ZFS_TYPE_BOOKMARK)
	return (B_FALSE);

	nvl = fnvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	bmark_snaps = fnvlist_lookup_uint64_array(nvl, ZPROP_VALUE,
	&bmark_num_snaps);
	if (bmark_num_snaps != gtnd->num_redact_snaps)
	return (B_FALSE);
	int i = 0;
	for (; i < bmark_num_snaps; i++) {
	int j = 0;
	for (; j < bmark_num_snaps; j++) {
	if (bmark_snaps[i] == gtnd->redact_snap_guids[j])
	break;
	}
	if (j == bmark_num_snaps)
	break;
	}
	return (i == bmark_num_snaps);
	}

	static int
	guid_to_name_cb(zfs_handle_t zhp, void arg)
	{
	guid_to_name_data_t *gtnd = arg;
	const char *slash;
	int err;

	if (gtnd->skip != NULL &&
	(slash = strrchr(zhp->zfs_name, '/')) != NULL &&
	strcmp(slash + 1, gtnd->skip) == 0) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_GUID) == gtnd->guid &&
	(gtnd->num_redact_snaps == -1 \|\| redact_snaps_match(zhp, gtnd))) {
	(void) strcpy(gtnd->name, zhp->zfs_name);
	zfs_close(zhp);
	return (EEXIST);
	}

	err = zfs_iter_children(zhp, guid_to_name_cb, gtnd);
	if (err != EEXIST && gtnd->bookmark_ok)
	err = zfs_iter_bookmarks(zhp, guid_to_name_cb, gtnd);
	zfs_close(zhp);
	return (err);
	}

	/*
	* Attempt to find the local dataset associated with this guid. In the case of
	* multiple matches, we attempt to find the "best" match by searching
	* progressively larger portions of the hierarchy. This allows one to send a
	* tree of datasets individually and guarantee that we will find the source
	* guid within that hierarchy, even if there are multiple matches elsewhere.
	*
	* If num_redact_snaps is not -1, we attempt to find a redaction bookmark with
	* the specified number of redaction snapshots. If num_redact_snaps isn't 0 or
	* -1, then redact_snap_guids will be an array of the guids of the snapshots the
	* redaction bookmark was created with. If num_redact_snaps is -1, then we will
	* attempt to find a snapshot or bookmark (if bookmark_ok is passed) with the
	* given guid. Note that a redaction bookmark can be returned if
	* num_redact_snaps == -1.
	*/
	static int
	guid_to_name_redact_snaps(libzfs_handle_t hdl, const char parent,
	uint64_t guid, boolean_t bookmark_ok, uint64_t *redact_snap_guids,
	uint64_t num_redact_snaps, char *name)
	{
	char pname[ZFS_MAX_DATASET_NAME_LEN];
	guid_to_name_data_t gtnd;

	gtnd.guid = guid;
	gtnd.bookmark_ok = bookmark_ok;
	gtnd.name = name;
	gtnd.skip = NULL;
	gtnd.redact_snap_guids = redact_snap_guids;
	gtnd.num_redact_snaps = num_redact_snaps;

	/*
	* Search progressively larger portions of the hierarchy, starting
	* with the filesystem specified by 'parent'. This will
	* select the "most local" version of the origin snapshot in the case
	* that there are multiple matching snapshots in the system.
	*/
	(void) strlcpy(pname, parent, sizeof (pname));
	char *cp = strrchr(pname, '@');
	if (cp == NULL)
	cp = strchr(pname, '\0');
	for (; cp != NULL; cp = strrchr(pname, '/')) {
	/* Chop off the last component and open the parent */
	*cp = '\0';
	zfs_handle_t *zhp = make_dataset_handle(hdl, pname);

	if (zhp == NULL)
	continue;
	int err = guid_to_name_cb(zfs_handle_dup(zhp), &gtnd);
	if (err != EEXIST)
	err = zfs_iter_children(zhp, guid_to_name_cb, &gtnd);
	if (err != EEXIST && bookmark_ok)
	err = zfs_iter_bookmarks(zhp, guid_to_name_cb, &gtnd);
	zfs_close(zhp);
	if (err == EEXIST)
	return (0);

	/*
	* Remember the last portion of the dataset so we skip it next
	* time through (as we've already searched that portion of the
	* hierarchy).
	*/
	gtnd.skip = strrchr(pname, '/') + 1;
	}

	return (ENOENT);
	}

	static int
	guid_to_name(libzfs_handle_t hdl, const char parent, uint64_t guid,
	boolean_t bookmark_ok, char *name)
	{
	return (guid_to_name_redact_snaps(hdl, parent, guid, bookmark_ok, NULL,
	-1, name));
	}

	/*
	* Return +1 if guid1 is before guid2, 0 if they are the same, and -1 if
	* guid1 is after guid2.
	*/
	static int
	created_before(libzfs_handle_t hdl, avl_tree_t avl,
	uint64_t guid1, uint64_t guid2)
	{
	nvlist_t *nvfs;
	char fsname = NULL, snapname = NULL;
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	int rv;
	zfs_handle_t guid1hdl, guid2hdl;
	uint64_t create1, create2;

	if (guid2 == 0)
	return (0);
	if (guid1 == 0)
	return (1);

	nvfs = fsavl_find(avl, guid1, &snapname);
	fsname = fnvlist_lookup_string(nvfs, "name");
	(void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
	guid1hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
	if (guid1hdl == NULL)
	return (-1);

	nvfs = fsavl_find(avl, guid2, &snapname);
	fsname = fnvlist_lookup_string(nvfs, "name");
	(void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
	guid2hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
	if (guid2hdl == NULL) {
	zfs_close(guid1hdl);
	return (-1);
	}

	create1 = zfs_prop_get_int(guid1hdl, ZFS_PROP_CREATETXG);
	create2 = zfs_prop_get_int(guid2hdl, ZFS_PROP_CREATETXG);

	if (create1 < create2)
	rv = -1;
	else if (create1 > create2)
	rv = +1;
	else
	rv = 0;

	zfs_close(guid1hdl);
	zfs_close(guid2hdl);

	return (rv);
	}

	/*
	* This function reestablishes the hierarchy of encryption roots after a
	* recursive incremental receive has completed. This must be done after the
	* second call to recv_incremental_replication() has renamed and promoted all
	* sent datasets to their final locations in the dataset hierarchy.
	*/
	static int
	recv_fix_encryption_hierarchy(libzfs_handle_t hdl, const char top_zfs,
	nvlist_t *stream_nv)
	{
	int err;
	nvpair_t *fselem = NULL;
	nvlist_t *stream_fss;

	stream_fss = fnvlist_lookup_nvlist(stream_nv, "fss");

	while ((fselem = nvlist_next_nvpair(stream_fss, fselem)) != NULL) {
	zfs_handle_t *zhp = NULL;
	uint64_t crypt;
	nvlist_t snaps, props, *stream_nvfs = NULL;
	nvpair_t *snapel = NULL;
	boolean_t is_encroot, is_clone, stream_encroot;
	char *cp;
	char *stream_keylocation = NULL;
	char keylocation[MAXNAMELEN];
	char fsname[ZFS_MAX_DATASET_NAME_LEN];

	keylocation[0] = '\0';
	stream_nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(stream_nvfs, "snaps");
	props = fnvlist_lookup_nvlist(stream_nvfs, "props");
	stream_encroot = nvlist_exists(stream_nvfs, "is_encroot");

	/* find a snapshot from the stream that exists locally */
	err = ENOENT;
	while ((snapel = nvlist_next_nvpair(snaps, snapel)) != NULL) {
	uint64_t guid;

	guid = fnvpair_value_uint64(snapel);
	err = guid_to_name(hdl, top_zfs, guid, B_FALSE,
	fsname);
	if (err == 0)
	break;
	}

	if (err != 0)
	continue;

	cp = strchr(fsname, '@');
	if (cp != NULL)
	*cp = '\0';

	zhp = zfs_open(hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = ENOENT;
	goto error;
	}

	crypt = zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION);
	is_clone = zhp->zfs_dmustats.dds_origin[0] != '\0';
	(void) zfs_crypto_get_encryption_root(zhp, &is_encroot, NULL);

	/* we don't need to do anything for unencrypted datasets */
	if (crypt == ZIO_CRYPT_OFF) {
	zfs_close(zhp);
	continue;
	}

	/*
	* If the dataset is flagged as an encryption root, was not
	* received as a clone and is not currently an encryption root,
	* force it to become one. Fixup the keylocation if necessary.
	*/
	if (stream_encroot) {
	if (!is_clone && !is_encroot) {
	err = lzc_change_key(fsname,
	DCP_CMD_FORCE_NEW_KEY, NULL, NULL, 0);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}

	stream_keylocation = fnvlist_lookup_string(props,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION));

	/*
	* Refresh the properties in case the call to
	* lzc_change_key() changed the value.
	*/
	zfs_refresh_properties(zhp);
	err = zfs_prop_get(zhp, ZFS_PROP_KEYLOCATION,
	keylocation, sizeof (keylocation), NULL, NULL,
	0, B_TRUE);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}

	if (strcmp(keylocation, stream_keylocation) != 0) {
	err = zfs_prop_set(zhp,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION),
	stream_keylocation);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}
	}

	/*
	* If the dataset is not flagged as an encryption root and is
	* currently an encryption root, force it to inherit from its
	* parent. The root of a raw send should never be
	* force-inherited.
	*/
	if (!stream_encroot && is_encroot &&
	strcmp(top_zfs, fsname) != 0) {
	err = lzc_change_key(fsname, DCP_CMD_FORCE_INHERIT,
	NULL, NULL, 0);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}

	zfs_close(zhp);
	}

	return (0);

	error:
	return (err);
	}

	static int
	recv_incremental_replication(libzfs_handle_t hdl, const char tofs,
	recvflags_t flags, nvlist_t stream_nv, avl_tree_t *stream_avl,
	nvlist_t *renamed)
	{
	nvlist_t local_nv, deleted = NULL;
	avl_tree_t *local_avl;
	nvpair_t fselem, nextfselem;
	char *fromsnap;
	char newname[ZFS_MAX_DATASET_NAME_LEN];
	char guidname[32];
	int error;
	boolean_t needagain, progress, recursive;
	char s1, s2;

	fromsnap = fnvlist_lookup_string(stream_nv, "fromsnap");

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);

	if (flags->dryrun)
	return (0);

	again:
	needagain = progress = B_FALSE;

	deleted = fnvlist_alloc();

	if ((error = gather_nvlist(hdl, tofs, fromsnap, NULL,
	recursive, B_TRUE, B_FALSE, recursive, B_FALSE, B_FALSE, B_FALSE,
	B_FALSE, B_TRUE, &local_nv, &local_avl)) != 0)
	return (error);

	/*
	* Process deletes and renames
	*/
	for (fselem = nvlist_next_nvpair(local_nv, NULL);
	fselem; fselem = nextfselem) {
	nvlist_t nvfs, snaps;
	nvlist_t *stream_nvfs = NULL;
	nvpair_t snapelem, nextsnapelem;
	uint64_t fromguid = 0;
	uint64_t originguid = 0;
	uint64_t stream_originguid = 0;
	uint64_t parent_fromsnap_guid, stream_parent_fromsnap_guid;
	char fsname, stream_fsname;

	nextfselem = nvlist_next_nvpair(local_nv, fselem);

	nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(nvfs, "snaps");
	fsname = fnvlist_lookup_string(nvfs, "name");
	parent_fromsnap_guid = fnvlist_lookup_uint64(nvfs,
	"parentfromsnap");
	(void) nvlist_lookup_uint64(nvfs, "origin", &originguid);

	/*
	* First find the stream's fs, so we can check for
	* a different origin (due to "zfs promote")
	*/
	for (snapelem = nvlist_next_nvpair(snaps, NULL);
	snapelem; snapelem = nvlist_next_nvpair(snaps, snapelem)) {
	uint64_t thisguid;

	thisguid = fnvpair_value_uint64(snapelem);
	stream_nvfs = fsavl_find(stream_avl, thisguid, NULL);

	if (stream_nvfs != NULL)
	break;
	}

	/* check for promote */
	(void) nvlist_lookup_uint64(stream_nvfs, "origin",
	&stream_originguid);
	if (stream_nvfs && originguid != stream_originguid) {
	switch (created_before(hdl, local_avl,
	stream_originguid, originguid)) {
	case 1: {
	/* promote it! */
	nvlist_t *origin_nvfs;
	char *origin_fsname;

	origin_nvfs = fsavl_find(local_avl, originguid,
	NULL);
	origin_fsname = fnvlist_lookup_string(
	origin_nvfs, "name");
	error = recv_promote(hdl, fsname, origin_fsname,
	flags);
	if (error == 0)
	progress = B_TRUE;
	break;
	}
	default:
	break;
	case -1:
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);
	return (-1);
	}
	/*
	* We had/have the wrong origin, therefore our
	* list of snapshots is wrong. Need to handle
	* them on the next pass.
	*/
	needagain = B_TRUE;
	continue;
	}

	for (snapelem = nvlist_next_nvpair(snaps, NULL);
	snapelem; snapelem = nextsnapelem) {
	uint64_t thisguid;
	char *stream_snapname;
	nvlist_t found, props;

	nextsnapelem = nvlist_next_nvpair(snaps, snapelem);

	thisguid = fnvpair_value_uint64(snapelem);
	found = fsavl_find(stream_avl, thisguid,
	&stream_snapname);

	/* check for delete */
	if (found == NULL) {
	char name[ZFS_MAX_DATASET_NAME_LEN];

	if (!flags->force)
	continue;

	(void) snprintf(name, sizeof (name), "%s@%s",
	fsname, nvpair_name(snapelem));

	error = recv_destroy(hdl, name,
	strlen(fsname)+1, newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	sprintf(guidname, "%llu",
	(u_longlong_t)thisguid);
	nvlist_add_boolean(deleted, guidname);
	continue;
	}

	stream_nvfs = found;

	if (0 == nvlist_lookup_nvlist(stream_nvfs, "snapprops",
	&props) && 0 == nvlist_lookup_nvlist(props,
	stream_snapname, &props)) {
	zfs_cmd_t zc = {"\0"};

	zc.zc_cookie = B_TRUE; /* received */
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name),
	"%s@%s", fsname, nvpair_name(snapelem));
	zcmd_write_src_nvlist(hdl, &zc, props);
	(void) zfs_ioctl(hdl,
	ZFS_IOC_SET_PROP, &zc);
	zcmd_free_nvlists(&zc);
	}

	/* check for different snapname */
	if (strcmp(nvpair_name(snapelem),
	stream_snapname) != 0) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	char tryname[ZFS_MAX_DATASET_NAME_LEN];

	(void) snprintf(name, sizeof (name), "%s@%s",
	fsname, nvpair_name(snapelem));
	(void) snprintf(tryname, sizeof (name), "%s@%s",
	fsname, stream_snapname);

	error = recv_rename(hdl, name, tryname,
	strlen(fsname)+1, newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	}

	if (strcmp(stream_snapname, fromsnap) == 0)
	fromguid = thisguid;
	}

	/* check for delete */
	if (stream_nvfs == NULL) {
	if (!flags->force)
	continue;

	error = recv_destroy(hdl, fsname, strlen(tofs)+1,
	newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	sprintf(guidname, "%llu",
	(u_longlong_t)parent_fromsnap_guid);
	nvlist_add_boolean(deleted, guidname);
	continue;
	}

	if (fromguid == 0) {
	if (flags->verbose) {
	(void) printf("local fs %s does not have "
	"fromsnap (%s in stream); must have "
	"been deleted locally; ignoring\n",
	fsname, fromsnap);
	}
	continue;
	}

	stream_fsname = fnvlist_lookup_string(stream_nvfs, "name");
	stream_parent_fromsnap_guid = fnvlist_lookup_uint64(
	stream_nvfs, "parentfromsnap");

	s1 = strrchr(fsname, '/');
	s2 = strrchr(stream_fsname, '/');

	/*
	* Check if we're going to rename based on parent guid change
	* and the current parent guid was also deleted. If it was then
	* rename will fail and is likely unneeded, so avoid this and
	* force an early retry to determine the new
	* parent_fromsnap_guid.
	*/
	if (stream_parent_fromsnap_guid != 0 &&
	parent_fromsnap_guid != 0 &&
	stream_parent_fromsnap_guid != parent_fromsnap_guid) {
	sprintf(guidname, "%llu",
	(u_longlong_t)parent_fromsnap_guid);
	if (nvlist_exists(deleted, guidname)) {
	progress = B_TRUE;
	needagain = B_TRUE;
	goto doagain;
	}
	}

	/*
	* Check for rename. If the exact receive path is specified, it
	* does not count as a rename, but we still need to check the
	* datasets beneath it.
	*/
	if ((stream_parent_fromsnap_guid != 0 &&
	parent_fromsnap_guid != 0 &&
	stream_parent_fromsnap_guid != parent_fromsnap_guid) \|\|
	((flags->isprefix \|\| strcmp(tofs, fsname) != 0) &&
	(s1 != NULL) && (s2 != NULL) && strcmp(s1, s2) != 0)) {
	nvlist_t *parent;
	char tryname[ZFS_MAX_DATASET_NAME_LEN];

	parent = fsavl_find(local_avl,
	stream_parent_fromsnap_guid, NULL);
	/*
	* NB: parent might not be found if we used the
	* tosnap for stream_parent_fromsnap_guid,
	* because the parent is a newly-created fs;
	* we'll be able to rename it after we recv the
	* new fs.
	*/
	if (parent != NULL) {
	char *pname;

	pname = fnvlist_lookup_string(parent, "name");
	(void) snprintf(tryname, sizeof (tryname),
	"%s%s", pname, strrchr(stream_fsname, '/'));
	} else {
	tryname[0] = '\0';
	if (flags->verbose) {
	(void) printf("local fs %s new parent "
	"not found\n", fsname);
	}
	}

	newname[0] = '\0';

	error = recv_rename(hdl, fsname, tryname,
	strlen(tofs)+1, newname, flags);

	if (renamed != NULL && newname[0] != '\0') {
	fnvlist_add_boolean(renamed, newname);
	}

	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	}
	}

	doagain:
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);
	fnvlist_free(deleted);

	if (needagain && progress) {
	/* do another pass to fix up temporary names */
	if (flags->verbose)
	(void) printf("another pass:\n");
	goto again;
	}

	return (needagain \|\| error != 0);
	}

	static int
	zfs_receive_package(libzfs_handle_t hdl, int fd, const char destname,
	recvflags_t flags, dmu_replay_record_t drr, zio_cksum_t *zc,
	char *top_zfs, nvlist_t cmdprops)
	{
	nvlist_t *stream_nv = NULL;
	avl_tree_t *stream_avl = NULL;
	char *fromsnap = NULL;
	char *sendsnap = NULL;
	char *cp;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	char sendfs[ZFS_MAX_DATASET_NAME_LEN];
	char errbuf[ERRBUFLEN];
	dmu_replay_record_t drre;
	int error;
	boolean_t anyerr = B_FALSE;
	boolean_t softerr = B_FALSE;
	boolean_t recursive, raw;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	assert(drr->drr_type == DRR_BEGIN);
	assert(drr->drr_u.drr_begin.drr_magic == DMU_BACKUP_MAGIC);
	assert(DMU_GET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo) ==
	DMU_COMPOUNDSTREAM);

	/*
	* Read in the nvlist from the stream.
	*/
	if (drr->drr_payloadlen != 0) {
	error = recv_read_nvlist(hdl, fd, drr->drr_payloadlen,
	&stream_nv, flags->byteswap, zc);
	if (error) {
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	}

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);
	raw = (nvlist_lookup_boolean(stream_nv, "raw") == 0);

	if (recursive && strchr(destname, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot specify snapshot name for multi-snapshot stream"));
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}

	/*
	* Read in the end record and verify checksum.
	*/
	if (0 != (error = recv_read(hdl, fd, &drre, sizeof (drre),
	flags->byteswap, NULL)))
	goto out;
	if (flags->byteswap) {
	drre.drr_type = BSWAP_32(drre.drr_type);
	drre.drr_u.drr_end.drr_checksum.zc_word[0] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[0]);
	drre.drr_u.drr_end.drr_checksum.zc_word[1] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[1]);
	drre.drr_u.drr_end.drr_checksum.zc_word[2] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[2]);
	drre.drr_u.drr_end.drr_checksum.zc_word[3] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[3]);
	}
	if (drre.drr_type != DRR_END) {
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	if (!ZIO_CHECKSUM_EQUAL(drre.drr_u.drr_end.drr_checksum, *zc)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incorrect header checksum"));
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}

	(void) nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap);

	if (drr->drr_payloadlen != 0) {
	nvlist_t *stream_fss;

	stream_fss = fnvlist_lookup_nvlist(stream_nv, "fss");
	if ((stream_avl = fsavl_create(stream_fss)) == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"couldn't allocate avl tree"));
	error = zfs_error(hdl, EZFS_NOMEM, errbuf);
	goto out;
	}

	if (fromsnap != NULL && recursive) {
	nvlist_t *renamed = NULL;
	nvpair_t *pair = NULL;

	(void) strlcpy(tofs, destname, sizeof (tofs));
	if (flags->isprefix) {
	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	int i;

	if (flags->istail) {
	cp = strrchr(drrb->drr_toname, '/');
	if (cp == NULL) {
	(void) strlcat(tofs, "/",
	sizeof (tofs));
	i = 0;
	} else {
	i = (cp - drrb->drr_toname);
	}
	} else {
	i = strcspn(drrb->drr_toname, "/@");
	}
	/* zfs_receive_one() will create_parents() */
	(void) strlcat(tofs, &drrb->drr_toname[i],
	sizeof (tofs));
	*strchr(tofs, '@') = '\0';
	}

	if (!flags->dryrun && !flags->nomount) {
	renamed = fnvlist_alloc();
	}

	softerr = recv_incremental_replication(hdl, tofs, flags,
	stream_nv, stream_avl, renamed);

	/* Unmount renamed filesystems before receiving. */
	while ((pair = nvlist_next_nvpair(renamed,
	pair)) != NULL) {
	zfs_handle_t *zhp;
	prop_changelist_t *clp = NULL;

	zhp = zfs_open(hdl, nvpair_name(pair),
	ZFS_TYPE_FILESYSTEM);
	if (zhp != NULL) {
	clp = changelist_gather(zhp,
	ZFS_PROP_MOUNTPOINT, 0,
	flags->forceunmount ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp != NULL) {
	softerr \|=
	changelist_prefix(clp);
	changelist_free(clp);
	}
	}
	}

	fnvlist_free(renamed);
	}
	}

	/*
	* Get the fs specified by the first path in the stream (the top level
	* specified by 'zfs send') and pass it to each invocation of
	* zfs_receive_one().
	*/
	(void) strlcpy(sendfs, drr->drr_u.drr_begin.drr_toname,
	sizeof (sendfs));
	if ((cp = strchr(sendfs, '@')) != NULL) {
	*cp = '\0';
	/*
	* Find the "sendsnap", the final snapshot in a replication
	* stream. zfs_receive_one() handles certain errors
	* differently, depending on if the contained stream is the
	* last one or not.
	*/
	sendsnap = (cp + 1);
	}

	/* Finally, receive each contained stream */
	do {
	/*
	* we should figure out if it has a recoverable
	* error, in which case do a recv_skip() and drive on.
	* Note, if we fail due to already having this guid,
	* zfs_receive_one() will take care of it (ie,
	* recv_skip() and return 0).
	*/
	error = zfs_receive_impl(hdl, destname, NULL, flags, fd,
	sendfs, stream_nv, stream_avl, top_zfs, sendsnap, cmdprops);
	if (error == ENODATA) {
	error = 0;
	break;
	}
	anyerr \|= error;
	} while (error == 0);

	if (drr->drr_payloadlen != 0 && recursive && fromsnap != NULL) {
	/*
	* Now that we have the fs's they sent us, try the
	* renames again.
	*/
	softerr = recv_incremental_replication(hdl, tofs, flags,
	stream_nv, stream_avl, NULL);
	}

	if (raw && softerr == 0 && *top_zfs != NULL) {
	softerr = recv_fix_encryption_hierarchy(hdl, *top_zfs,
	stream_nv);
	}

	out:
	fsavl_destroy(stream_avl);
	fnvlist_free(stream_nv);
	if (softerr)
	error = -2;
	if (anyerr)
	error = -1;
	return (error);
	}

	static void
	trunc_prop_errs(int truncated)
	{
	ASSERT(truncated != 0);

	if (truncated == 1)
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"1 more property could not be set\n"));
	else
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"%d more properties could not be set\n"), truncated);
	}

	static int
	recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap)
	{
	dmu_replay_record_t *drr;
	void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);
	uint64_t payload_size;
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	/* XXX would be great to use lseek if possible... */
	drr = buf;

	while (recv_read(hdl, fd, drr, sizeof (dmu_replay_record_t),
	byteswap, NULL) == 0) {
	if (byteswap)
	drr->drr_type = BSWAP_32(drr->drr_type);

	switch (drr->drr_type) {
	case DRR_BEGIN:
	if (drr->drr_payloadlen != 0) {
	(void) recv_read(hdl, fd, buf,
	drr->drr_payloadlen, B_FALSE, NULL);
	}
	break;

	case DRR_END:
	free(buf);
	return (0);

	case DRR_OBJECT:
	if (byteswap) {
	drr->drr_u.drr_object.drr_bonuslen =
	BSWAP_32(drr->drr_u.drr_object.
	drr_bonuslen);
	drr->drr_u.drr_object.drr_raw_bonuslen =
	BSWAP_32(drr->drr_u.drr_object.
	drr_raw_bonuslen);
	}

	payload_size =
	DRR_OBJECT_PAYLOAD_SIZE(&drr->drr_u.drr_object);
	(void) recv_read(hdl, fd, buf, payload_size,
	B_FALSE, NULL);
	break;

	case DRR_WRITE:
	if (byteswap) {
	drr->drr_u.drr_write.drr_logical_size =
	BSWAP_64(
	drr->drr_u.drr_write.drr_logical_size);
	drr->drr_u.drr_write.drr_compressed_size =
	BSWAP_64(
	drr->drr_u.drr_write.drr_compressed_size);
	}
	payload_size =
	DRR_WRITE_PAYLOAD_SIZE(&drr->drr_u.drr_write);
	assert(payload_size <= SPA_MAXBLOCKSIZE);
	(void) recv_read(hdl, fd, buf,
	payload_size, B_FALSE, NULL);
	break;
	case DRR_SPILL:
	if (byteswap) {
	drr->drr_u.drr_spill.drr_length =
	BSWAP_64(drr->drr_u.drr_spill.drr_length);
	drr->drr_u.drr_spill.drr_compressed_size =
	BSWAP_64(drr->drr_u.drr_spill.
	drr_compressed_size);
	}

	payload_size =
	DRR_SPILL_PAYLOAD_SIZE(&drr->drr_u.drr_spill);
	(void) recv_read(hdl, fd, buf, payload_size,
	B_FALSE, NULL);
	break;
	case DRR_WRITE_EMBEDDED:
	if (byteswap) {
	drr->drr_u.drr_write_embedded.drr_psize =
	BSWAP_32(drr->drr_u.drr_write_embedded.
	drr_psize);
	}
	(void) recv_read(hdl, fd, buf,
	P2ROUNDUP(drr->drr_u.drr_write_embedded.drr_psize,
	8), B_FALSE, NULL);
	break;
	case DRR_OBJECT_RANGE:
	case DRR_WRITE_BYREF:
	case DRR_FREEOBJECTS:
	case DRR_FREE:
	break;

	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid record type"));
	free(buf);
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}
	}

	free(buf);
	return (-1);
	}

	static void
	recv_ecksum_set_aux(libzfs_handle_t hdl, const char target_snap,
	boolean_t resumable, boolean_t checksum)
	{
	char target_fs[ZFS_MAX_DATASET_NAME_LEN];

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, (checksum ?
	"checksum mismatch" : "incomplete stream")));

	if (!resumable)
	return;
	(void) strlcpy(target_fs, target_snap, sizeof (target_fs));
	*strchr(target_fs, '@') = '\0';
	zfs_handle_t *zhp = zfs_open(hdl, target_fs,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return;

	char token_buf[ZFS_MAXPROPLEN];
	int error = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	token_buf, sizeof (token_buf),
	NULL, NULL, 0, B_TRUE);
	if (error == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"checksum mismatch or incomplete stream.\n"
	"Partially received snapshot is saved.\n"
	"A resuming stream can be generated on the sending "
	"system by running:\n"
	" zfs send -t %s"),
	token_buf);
	}
	zfs_close(zhp);
	}

	/*
	* Prepare a new nvlist of properties that are to override (-o) or be excluded
	* (-x) from the received dataset
	* recvprops: received properties from the send stream
	* cmdprops: raw input properties from command line
	* origprops: properties, both locally-set and received, currently set on the
	* target dataset if it exists, NULL otherwise.
	* oxprops: valid output override (-o) and excluded (-x) properties
	*/
	static int
	zfs_setup_cmdline_props(libzfs_handle_t *hdl, zfs_type_t type,
	char *fsname, boolean_t zoned, boolean_t recursive, boolean_t newfs,
	boolean_t raw, boolean_t toplevel, nvlist_t recvprops, nvlist_t cmdprops,
	nvlist_t origprops, nvlist_t oxprops, uint8_t *wkeydata_out,
	uint_t wkeylen_out, const char errbuf)
	{
	nvpair_t *nvp;
	nvlist_t oprops, voprops;
	zfs_handle_t *zhp = NULL;
	zpool_handle_t *zpool_hdl = NULL;
	char *cp;
	int ret = 0;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];

	if (nvlist_empty(cmdprops))
	return (0); /* No properties to override or exclude */

	*oxprops = fnvlist_alloc();
	oprops = fnvlist_alloc();

	strlcpy(namebuf, fsname, ZFS_MAX_DATASET_NAME_LEN);

	/*
	* Get our dataset handle. The target dataset may not exist yet.
	*/
	if (zfs_dataset_exists(hdl, namebuf, ZFS_TYPE_DATASET)) {
	zhp = zfs_open(hdl, namebuf, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	ret = -1;
	goto error;
	}
	}

	/* open the zpool handle */
	cp = strchr(namebuf, '/');
	if (cp != NULL)
	*cp = '\0';
	zpool_hdl = zpool_open(hdl, namebuf);
	if (zpool_hdl == NULL) {
	ret = -1;
	goto error;
	}

	/* restore namebuf to match fsname for later use */
	if (cp != NULL)
	*cp = '/';

	/*
	* first iteration: process excluded (-x) properties now and gather
	* added (-o) properties to be later processed by zfs_valid_proplist()
	*/
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(cmdprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	zfs_prop_t prop = zfs_name_to_prop(name);

	/*
	* It turns out, if we don't normalize "aliased" names
	* e.g. compress= against the "real" names (e.g. compression)
	* here, then setting/excluding them does not work as
	* intended.
	*
	* But since user-defined properties wouldn't have a valid
	* mapping here, we do this conditional dance.
	*/
	const char *newname = name;
	if (prop >= ZFS_PROP_TYPE)
	newname = zfs_prop_to_name(prop);

	/* "origin" is processed separately, don't handle it here */
	if (prop == ZFS_PROP_ORIGIN)
	continue;

	/* raw streams can't override encryption properties */
	if ((zfs_prop_encryption_key_param(prop) \|\|
	prop == ZFS_PROP_ENCRYPTION) && raw) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption property '%s' cannot "
	"be set or excluded for raw streams."), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	/* incremental streams can only exclude encryption properties */
	if ((zfs_prop_encryption_key_param(prop) \|\|
	prop == ZFS_PROP_ENCRYPTION) && !newfs &&
	nvpair_type(nvp) != DATA_TYPE_BOOLEAN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption property '%s' cannot "
	"be set for incremental streams."), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	switch (nvpair_type(nvp)) {
	case DATA_TYPE_BOOLEAN: /* -x property */
	/*
	* DATA_TYPE_BOOLEAN is the way we're asked to "exclude"
	* a property: this is done by forcing an explicit
	* inherit on the destination so the effective value is
	* not the one we received from the send stream.
	*/
	if (!zfs_prop_valid_for_type(prop, type, B_FALSE) &&
	!zfs_prop_user(name)) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"Warning: %s: property '%s' does not "
	"apply to datasets of this type\n"),
	fsname, name);
	continue;
	}
	/*
	* We do this only if the property is not already
	* locally-set, in which case its value will take
	* priority over the received anyway.
	*/
	if (nvlist_exists(origprops, newname)) {
	nvlist_t *attrs;
	char *source = NULL;

	attrs = fnvlist_lookup_nvlist(origprops,
	newname);
	if (nvlist_lookup_string(attrs,
	ZPROP_SOURCE, &source) == 0 &&
	strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)
	continue;
	}
	/*
	* We can't force an explicit inherit on non-inheritable
	* properties: if we're asked to exclude this kind of
	* values we remove them from "recvprops" input nvlist.
	*/
	if (!zfs_prop_user(name) && /* can be inherited too */
	!zfs_prop_inheritable(prop) &&
	nvlist_exists(recvprops, newname))
	fnvlist_remove(recvprops, newname);
	else
	fnvlist_add_boolean(*oxprops, newname);
	break;
	case DATA_TYPE_STRING: /* -o property=value */
	/*
	* we're trying to override a property that does not
	* make sense for this type of dataset, but we don't
	* want to fail if the receive is recursive: this comes
	* in handy when the send stream contains, for
	* instance, a child ZVOL and we're trying to receive
	* it with "-o atime=on"
	*/
	if (!zfs_prop_valid_for_type(prop, type, B_FALSE) &&
	!zfs_prop_user(name)) {
	if (recursive)
	continue;
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' does not apply to datasets "
	"of this type"), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	fnvlist_add_string(oprops, newname,
	fnvpair_value_string(nvp));
	break;
	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' must be a string or boolean"), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	}

	if (toplevel) {
	/* convert override strings properties to native */
	if ((voprops = zfs_valid_proplist(hdl, ZFS_TYPE_DATASET,
	oprops, zoned, zhp, zpool_hdl, B_FALSE, errbuf)) == NULL) {
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	/*
	* zfs_crypto_create() requires the parent name. Get it
	* by truncating the fsname copy stored in namebuf.
	*/
	cp = strrchr(namebuf, '/');
	if (cp != NULL)
	*cp = '\0';

	if (!raw && zfs_crypto_create(hdl, namebuf, voprops, NULL,
	B_FALSE, wkeydata_out, wkeylen_out) != 0) {
	fnvlist_free(voprops);
	ret = zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	goto error;
	}

	/* second pass: process "-o" properties */
	fnvlist_merge(*oxprops, voprops);
	fnvlist_free(voprops);
	} else {
	/* override props on child dataset are inherited */
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(oprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	fnvlist_add_boolean(*oxprops, name);
	}
	}

	error:
	if (zhp != NULL)
	zfs_close(zhp);
	if (zpool_hdl != NULL)
	zpool_close(zpool_hdl);
	fnvlist_free(oprops);
	return (ret);
	}

	/*
	* Restores a backup of tosnap from the file descriptor specified by infd.
	*/
	static int
	zfs_receive_one(libzfs_handle_t hdl, int infd, const char tosnap,
	const char originsnap, recvflags_t flags, dmu_replay_record_t *drr,
	dmu_replay_record_t drr_noswap, const char sendfs, nvlist_t *stream_nv,
	avl_tree_t stream_avl, char *top_zfs,
	const char finalsnap, nvlist_t cmdprops)
	{
	struct timespec begin_time;
	int ioctl_err, ioctl_errno, err;
	char *cp;
	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	char errbuf[ERRBUFLEN];
	const char *chopprefix;
	boolean_t newfs = B_FALSE;
	boolean_t stream_wantsnewfs, stream_resumingnewfs;
	boolean_t newprops = B_FALSE;
	uint64_t read_bytes = 0;
	uint64_t errflags = 0;
	uint64_t parent_snapguid = 0;
	prop_changelist_t *clp = NULL;
	nvlist_t *snapprops_nvlist = NULL;
	nvlist_t *snapholds_nvlist = NULL;
	zprop_errflags_t prop_errflags;
	nvlist_t *prop_errors = NULL;
	boolean_t recursive;
	char *snapname = NULL;
	char destsnap[MAXPATHLEN * 2];
	char origin[MAXNAMELEN] = {0};
	char name[MAXPATHLEN];
	char tmp_keylocation[MAXNAMELEN] = {0};
	nvlist_t rcvprops = NULL; / props received from the send stream */
	nvlist_t oxprops = NULL; / override (-o) and exclude (-x) props */
	nvlist_t origprops = NULL; / original props (if destination exists) */
	zfs_type_t type = ZFS_TYPE_INVALID;
	boolean_t toplevel = B_FALSE;
	boolean_t zoned = B_FALSE;
	boolean_t hastoken = B_FALSE;
	boolean_t redacted;
	uint8_t *wkeydata = NULL;
	uint_t wkeylen = 0;

	#ifndef CLOCK_MONOTONIC_RAW
	#define CLOCK_MONOTONIC_RAW CLOCK_MONOTONIC
	#endif
	clock_gettime(CLOCK_MONOTONIC_RAW, &begin_time);

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);

	/* Did the user request holds be skipped via zfs recv -k? */
	boolean_t holds = flags->holds && !flags->skipholds;

	if (stream_avl != NULL) {
	char *keylocation = NULL;
	nvlist_t *lookup = NULL;
	nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid,
	&snapname);

	(void) nvlist_lookup_uint64(fs, "parentfromsnap",
	&parent_snapguid);
	err = nvlist_lookup_nvlist(fs, "props", &rcvprops);
	if (err) {
	rcvprops = fnvlist_alloc();
	newprops = B_TRUE;
	}

	/*
	* The keylocation property may only be set on encryption roots,
	* but this dataset might not become an encryption root until
	* recv_fix_encryption_hierarchy() is called. That function
	* will fixup the keylocation anyway, so we temporarily unset
	* the keylocation for now to avoid any errors from the receive
	* ioctl.
	*/
	err = nvlist_lookup_string(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), &keylocation);
	if (err == 0) {
	- strcpy(tmp_keylocation, keylocation);
	+ strlcpy(tmp_keylocation, keylocation, MAXNAMELEN);
	(void) nvlist_remove_all(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION));
	}

	if (flags->canmountoff) {
	fnvlist_add_uint64(rcvprops,
	zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0);
	} else if (newprops) { /* nothing in rcvprops, eliminate it */
	fnvlist_free(rcvprops);
	rcvprops = NULL;
	newprops = B_FALSE;
	}
	if (0 == nvlist_lookup_nvlist(fs, "snapprops", &lookup)) {
	snapprops_nvlist = fnvlist_lookup_nvlist(lookup,
	snapname);
	}
	if (holds) {
	if (0 == nvlist_lookup_nvlist(fs, "snapholds",
	&lookup)) {
	snapholds_nvlist = fnvlist_lookup_nvlist(
	lookup, snapname);
	}
	}
	}

	cp = NULL;

	/*
	* Determine how much of the snapshot name stored in the stream
	* we are going to tack on to the name they specified on the
	* command line, and how much we are going to chop off.
	*
	* If they specified a snapshot, chop the entire name stored in
	* the stream.
	*/
	if (flags->istail) {
	/*
	* A filesystem was specified with -e. We want to tack on only
	* the tail of the sent snapshot path.
	*/
	if (strchr(tosnap, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"argument - snapshot not allowed with -e"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	chopprefix = strrchr(sendfs, '/');

	if (chopprefix == NULL) {
	/*
	* The tail is the poolname, so we need to
	* prepend a path separator.
	*/
	int len = strlen(drrb->drr_toname);
	cp = malloc(len + 2);
	cp[0] = '/';
	(void) strcpy(&cp[1], drrb->drr_toname);
	chopprefix = cp;
	} else {
	chopprefix = drrb->drr_toname + (chopprefix - sendfs);
	}
	} else if (flags->isprefix) {
	/*
	* A filesystem was specified with -d. We want to tack on
	* everything but the first element of the sent snapshot path
	* (all but the pool name).
	*/
	if (strchr(tosnap, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"argument - snapshot not allowed with -d"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	chopprefix = strchr(drrb->drr_toname, '/');
	if (chopprefix == NULL)
	chopprefix = strchr(drrb->drr_toname, '@');
	} else if (strchr(tosnap, '@') == NULL) {
	/*
	* If a filesystem was specified without -d or -e, we want to
	* tack on everything after the fs specified by 'zfs send'.
	*/
	chopprefix = drrb->drr_toname + strlen(sendfs);
	} else {
	/* A snapshot was specified as an exact path (no -d or -e). */
	if (recursive) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot specify snapshot name for multi-snapshot "
	"stream"));
	err = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	chopprefix = drrb->drr_toname + strlen(drrb->drr_toname);
	}

	ASSERT(strstr(drrb->drr_toname, sendfs) == drrb->drr_toname);
	ASSERT(chopprefix > drrb->drr_toname \|\| strchr(sendfs, '/') == NULL);
	ASSERT(chopprefix <= drrb->drr_toname + strlen(drrb->drr_toname) \|\|
	strchr(sendfs, '/') == NULL);
	ASSERT(chopprefix[0] == '/' \|\| chopprefix[0] == '@' \|\|
	chopprefix[0] == '\0');

	/*
	* Determine name of destination snapshot.
	*/
	(void) strlcpy(destsnap, tosnap, sizeof (destsnap));
	(void) strlcat(destsnap, chopprefix, sizeof (destsnap));
	free(cp);
	if (!zfs_name_valid(destsnap, ZFS_TYPE_SNAPSHOT)) {
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	/*
	* Determine the name of the origin snapshot.
	*/
	if (originsnap) {
	(void) strlcpy(origin, originsnap, sizeof (origin));
	if (flags->verbose)
	(void) printf("using provided clone origin %s\n",
	origin);
	} else if (drrb->drr_flags & DRR_FLAG_CLONE) {
	if (guid_to_name(hdl, destsnap,
	drrb->drr_fromguid, B_FALSE, origin) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"local origin for clone %s does not exist"),
	destsnap);
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	goto out;
	}
	if (flags->verbose)
	(void) printf("found clone origin %s\n", origin);
	}

	if ((DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_DEDUP)) {
	(void) fprintf(stderr,
	gettext("ERROR: \"zfs receive\" no longer supports "
	"deduplicated send streams. Use\n"
	"the \"zstream redup\" command to convert this stream "
	"to a regular,\n"
	"non-deduplicated stream.\n"));
	err = zfs_error(hdl, EZFS_NOTSUP, errbuf);
	goto out;
	}

	boolean_t resuming = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RESUMING;
	boolean_t raw = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RAW;
	boolean_t embedded = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_EMBED_DATA;
	stream_wantsnewfs = (drrb->drr_fromguid == 0 \|\|
	(drrb->drr_flags & DRR_FLAG_CLONE) \|\| originsnap) && !resuming;
	stream_resumingnewfs = (drrb->drr_fromguid == 0 \|\|
	(drrb->drr_flags & DRR_FLAG_CLONE) \|\| originsnap) && resuming;

	if (stream_wantsnewfs) {
	/*
	* if the parent fs does not exist, look for it based on
	* the parent snap GUID
	*/
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive new filesystem stream"));

	- (void) strcpy(name, destsnap);
	+ (void) strlcpy(name, destsnap, sizeof (name));
	cp = strrchr(name, '/');
	if (cp)
	*cp = '\0';
	if (cp &&
	!zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	char suffix[ZFS_MAX_DATASET_NAME_LEN];
	- (void) strcpy(suffix, strrchr(destsnap, '/'));
	+ (void) strlcpy(suffix, strrchr(destsnap, '/'),
	+ sizeof (suffix));
	if (guid_to_name(hdl, name, parent_snapguid,
	B_FALSE, destsnap) == 0) {
	*strchr(destsnap, '@') = '\0';
	- (void) strcat(destsnap, suffix);
	+ (void) strlcat(destsnap, suffix,
	+ sizeof (destsnap) - strlen(destsnap));
	}
	}
	} else {
	/*
	* If the fs does not exist, look for it based on the
	* fromsnap GUID.
	*/
	if (resuming) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive resume stream"));
	} else {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive incremental stream"));
	}

	- (void) strcpy(name, destsnap);
	+ (void) strlcpy(name, destsnap, sizeof (name));
	*strchr(name, '@') = '\0';

	/*
	* If the exact receive path was specified and this is the
	* topmost path in the stream, then if the fs does not exist we
	* should look no further.
	*/
	if ((flags->isprefix \|\| (*(chopprefix = drrb->drr_toname +
	strlen(sendfs)) != '\0' && *chopprefix != '@')) &&
	!zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	char snap[ZFS_MAX_DATASET_NAME_LEN];
	- (void) strcpy(snap, strchr(destsnap, '@'));
	+ (void) strlcpy(snap, strchr(destsnap, '@'),
	+ sizeof (snap));
	if (guid_to_name(hdl, name, drrb->drr_fromguid,
	B_FALSE, destsnap) == 0) {
	*strchr(destsnap, '@') = '\0';
	- (void) strcat(destsnap, snap);
	+ (void) strlcat(destsnap, snap,
	+ sizeof (destsnap) - strlen(destsnap));
	}
	}
	}

	- (void) strcpy(name, destsnap);
	+ (void) strlcpy(name, destsnap, sizeof (name));
	*strchr(name, '@') = '\0';

	redacted = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_REDACTED;

	if (flags->heal) {
	if (flags->isprefix \|\| flags->istail \|\| flags->force \|\|
	flags->canmountoff \|\| flags->resumable \|\| flags->nomount \|\|
	flags->skipholds) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective recv can not be used when combined with"
	" this flag"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}
	uint64_t guid =
	get_snap_guid(hdl, name, strchr(destsnap, '@') + 1);
	if (guid == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective recv must specify an existing snapshot"
	" to heal"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	} else if (guid != drrb->drr_toguid) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"local snapshot doesn't match the snapshot"
	" in the provided stream"));
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	goto out;
	}
	} else if (zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	zfs_cmd_t zc = {"\0"};
	zfs_handle_t *zhp = NULL;
	boolean_t encrypted;

	(void) strcpy(zc.zc_name, name);

	/*
	* Destination fs exists. It must be one of these cases:
	* - an incremental send stream
	* - the stream specifies a new fs (full stream or clone)
	* and they want us to blow away the existing fs (and
	* have therefore specified -F and removed any snapshots)
	* - we are resuming a failed receive.
	*/
	if (stream_wantsnewfs) {
	boolean_t is_volume = drrb->drr_type == DMU_OST_ZVOL;
	if (!flags->force) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' exists\n"
	"must specify -F to overwrite it"), name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (zfs_ioctl(hdl, ZFS_IOC_SNAPSHOT_LIST_NEXT,
	&zc) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination has snapshots (eg. %s)\n"
	"must destroy them to overwrite it"),
	zc.zc_name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (is_volume && strrchr(name, '/') == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s is the root dataset\n"
	"cannot overwrite with a ZVOL"),
	name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (is_volume &&
	zfs_ioctl(hdl, ZFS_IOC_DATASET_LIST_NEXT,
	&zc) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination has children (eg. %s)\n"
	"cannot overwrite with a ZVOL"),
	zc.zc_name);
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	goto out;
	}
	}

	if ((zhp = zfs_open(hdl, name,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME)) == NULL) {
	err = -1;
	goto out;
	}

	+ /*
	+ * When receiving full/newfs on existing dataset, then it
	+ * should be done with "-F" flag. Its enforced for initial
	+ * receive in previous checks in this function.
	+ * Similarly, on resuming full/newfs recv on existing dataset,
	+ * it should be done with "-F" flag.
	+ *
	+ * When dataset doesn't exist, then full/newfs recv is done on
	+ * newly created dataset and it's marked INCONSISTENT. But
	+ * When receiving on existing dataset, recv is first done on
	+ * %recv and its marked INCONSISTENT. Existing dataset is not
	+ * marked INCONSISTENT.
	+ * Resume of full/newfs receive with dataset not INCONSISTENT
	+ * indicates that its resuming newfs on existing dataset. So,
	+ * enforce "-F" flag in this case.
	+ */
	+ if (stream_resumingnewfs &&
	+ !zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) &&
	+ !flags->force) {
	+ zfs_close(zhp);
	+ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	+ "Resuming recv on existing destination '%s'\n"
	+ "must specify -F to overwrite it"), name);
	+ err = zfs_error(hdl, EZFS_RESUME_EXISTS, errbuf);
	+ goto out;
	+ }
	+
	if (stream_wantsnewfs &&
	zhp->zfs_dmustats.dds_origin[0]) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' is a clone\n"
	"must destroy it to overwrite it"), name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}

	/*
	* Raw sends can not be performed as an incremental on top
	* of existing unencrypted datasets. zfs recv -F can't be
	* used to blow away an existing encrypted filesystem. This
	* is because it would require the dsl dir to point to the
	* new key (or lack of a key) and the old key at the same
	* time. The -F flag may still be used for deleting
	* intermediate snapshots that would otherwise prevent the
	* receive from working.
	*/
	encrypted = zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) !=
	ZIO_CRYPT_OFF;
	if (!stream_wantsnewfs && !encrypted && raw) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot perform raw receive on top of "
	"existing unencrypted dataset"));
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	if (stream_wantsnewfs && flags->force &&
	((raw && !encrypted) \|\| encrypted)) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"zfs receive -F cannot be used to destroy an "
	"encrypted filesystem or overwrite an "
	"unencrypted one with an encrypted one"));
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
	(stream_wantsnewfs \|\| stream_resumingnewfs)) {
	/* We can't do online recv in this case */
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->forceunmount ? MS_FORCE : 0);
	if (clp == NULL) {
	zfs_close(zhp);
	err = -1;
	goto out;
	}
	if (changelist_prefix(clp) != 0) {
	changelist_free(clp);
	zfs_close(zhp);
	err = -1;
	goto out;
	}
	}

	/*
	* If we are resuming a newfs, set newfs here so that we will
	* mount it if the recv succeeds this time. We can tell
	* that it was a newfs on the first recv because the fs
	* itself will be inconsistent (if the fs existed when we
	* did the first recv, we would have received it into
	* .../%recv).
	*/
	if (resuming && zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT))
	newfs = B_TRUE;

	/* we want to know if we're zoned when validating -o\|-x props */
	zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);

	/* may need this info later, get it now we have zhp around */
	if (zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN, NULL, 0,
	NULL, NULL, 0, B_TRUE) == 0)
	hastoken = B_TRUE;

	/* gather existing properties on destination */
	origprops = fnvlist_alloc();
	fnvlist_merge(origprops, zhp->zfs_props);
	fnvlist_merge(origprops, zhp->zfs_user_props);

	zfs_close(zhp);
	} else {
	zfs_handle_t *zhp;

	/*
	* Destination filesystem does not exist. Therefore we better
	* be creating a new filesystem (either from a full backup, or
	* a clone). It would therefore be invalid if the user
	* specified only the pool name (i.e. if the destination name
	* contained no slash character).
	*/
	cp = strrchr(name, '/');

	if (!stream_wantsnewfs \|\| cp == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' does not exist"), name);
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	goto out;
	}

	/*
	* Trim off the final dataset component so we perform the
	* recvbackup ioctl to the filesystems's parent.
	*/
	*cp = '\0';

	if (flags->isprefix && !flags->istail && !flags->dryrun &&
	create_parents(hdl, destsnap, strlen(tosnap)) != 0) {
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	/* validate parent */
	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}
	if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"parent '%s' is not a filesystem"), name);
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	zfs_close(zhp);
	goto out;
	}

	zfs_close(zhp);

	newfs = B_TRUE;
	*cp = '/';
	}

	if (flags->verbose) {
	(void) printf("%s %s%s stream of %s into %s\n",
	flags->dryrun ? "would receive" : "receiving",
	flags->heal ? " corrective" : "",
	drrb->drr_fromguid ? "incremental" : "full",
	drrb->drr_toname, destsnap);
	(void) fflush(stdout);
	}

	/*
	* If this is the top-level dataset, record it so we can use it
	* for recursive operations later.
	*/
	if (top_zfs != NULL &&
	(top_zfs == NULL \|\| strcmp(top_zfs, name) == 0)) {
	toplevel = B_TRUE;
	if (*top_zfs == NULL)
	*top_zfs = zfs_strdup(hdl, name);
	}

	if (drrb->drr_type == DMU_OST_ZVOL) {
	type = ZFS_TYPE_VOLUME;
	} else if (drrb->drr_type == DMU_OST_ZFS) {
	type = ZFS_TYPE_FILESYSTEM;
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid record type: 0x%d"), drrb->drr_type);
	err = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	if ((err = zfs_setup_cmdline_props(hdl, type, name, zoned, recursive,
	stream_wantsnewfs, raw, toplevel, rcvprops, cmdprops, origprops,
	&oxprops, &wkeydata, &wkeylen, errbuf)) != 0)
	goto out;

	/*
	* When sending with properties (zfs send -p), the encryption property
	* is not included because it is a SETONCE property and therefore
	* treated as read only. However, we are always able to determine its
	* value because raw sends will include it in the DRR_BDEGIN payload
	* and non-raw sends with properties are not allowed for encrypted
	* datasets. Therefore, if this is a non-raw properties stream, we can
	* infer that the value should be ZIO_CRYPT_OFF and manually add that
	* to the received properties.
	*/
	if (stream_wantsnewfs && !raw && rcvprops != NULL &&
	!nvlist_exists(cmdprops, zfs_prop_to_name(ZFS_PROP_ENCRYPTION))) {
	if (oxprops == NULL)
	oxprops = fnvlist_alloc();
	fnvlist_add_uint64(oxprops,
	zfs_prop_to_name(ZFS_PROP_ENCRYPTION), ZIO_CRYPT_OFF);
	}

	if (flags->dryrun) {
	void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);

	/*
	* We have read the DRR_BEGIN record, but we have
	* not yet read the payload. For non-dryrun sends
	* this will be done by the kernel, so we must
	* emulate that here, before attempting to read
	* more records.
	*/
	err = recv_read(hdl, infd, buf, drr->drr_payloadlen,
	flags->byteswap, NULL);
	free(buf);
	if (err != 0)
	goto out;

	err = recv_skip(hdl, infd, flags->byteswap);
	goto out;
	}

	if (flags->heal) {
	err = ioctl_err = lzc_receive_with_heal(destsnap, rcvprops,
	oxprops, wkeydata, wkeylen, origin, flags->force,
	flags->heal, flags->resumable, raw, infd, drr_noswap, -1,
	&read_bytes, &errflags, NULL, &prop_errors);
	} else {
	err = ioctl_err = lzc_receive_with_cmdprops(destsnap, rcvprops,
	oxprops, wkeydata, wkeylen, origin, flags->force,
	flags->resumable, raw, infd, drr_noswap, -1, &read_bytes,
	&errflags, NULL, &prop_errors);
	}
	ioctl_errno = ioctl_err;
	prop_errflags = errflags;

	if (err == 0) {
	nvpair_t *prop_err = NULL;

	while ((prop_err = nvlist_next_nvpair(prop_errors,
	prop_err)) != NULL) {
	char tbuf[1024];
	zfs_prop_t prop;
	int intval;

	prop = zfs_name_to_prop(nvpair_name(prop_err));
	(void) nvpair_value_int32(prop_err, &intval);
	if (strcmp(nvpair_name(prop_err),
	ZPROP_N_MORE_ERRORS) == 0) {
	trunc_prop_errs(intval);
	break;
	} else if (snapname == NULL \|\| finalsnap == NULL \|\|
	strcmp(finalsnap, snapname) == 0 \|\|
	strcmp(nvpair_name(prop_err),
	zfs_prop_to_name(ZFS_PROP_REFQUOTA)) != 0) {
	/*
	* Skip the special case of, for example,
	* "refquota", errors on intermediate
	* snapshots leading up to a final one.
	* That's why we have all of the checks above.
	*
	* See zfs_ioctl.c's extract_delay_props() for
	* a list of props which can fail on
	* intermediate snapshots, but shouldn't
	* affect the overall receive.
	*/
	(void) snprintf(tbuf, sizeof (tbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive %s property on %s"),
	nvpair_name(prop_err), name);
	zfs_setprop_error(hdl, prop, intval, tbuf);
	}
	}
	}

	if (err == 0 && snapprops_nvlist) {
	zfs_cmd_t zc = {"\0"};

	- (void) strcpy(zc.zc_name, destsnap);
	+ (void) strlcpy(zc.zc_name, destsnap, sizeof (zc.zc_name));
	zc.zc_cookie = B_TRUE; /* received */
	zcmd_write_src_nvlist(hdl, &zc, snapprops_nvlist);
	(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
	zcmd_free_nvlists(&zc);
	}
	if (err == 0 && snapholds_nvlist) {
	nvpair_t *pair;
	nvlist_t holds, errors = NULL;
	int cleanup_fd = -1;

	VERIFY(0 == nvlist_alloc(&holds, 0, KM_SLEEP));
	for (pair = nvlist_next_nvpair(snapholds_nvlist, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(snapholds_nvlist, pair)) {
	fnvlist_add_string(holds, destsnap, nvpair_name(pair));
	}
	(void) lzc_hold(holds, cleanup_fd, &errors);
	fnvlist_free(snapholds_nvlist);
	fnvlist_free(holds);
	}

	if (err && (ioctl_errno == ENOENT \|\| ioctl_errno == EEXIST)) {
	/*
	* It may be that this snapshot already exists,
	* in which case we want to consume & ignore it
	* rather than failing.
	*/
	avl_tree_t *local_avl;
	nvlist_t local_nv, fs;
	cp = strchr(destsnap, '@');

	/*
	* XXX Do this faster by just iterating over snaps in
	* this fs. Also if zc_value does not exist, we will
	* get a strange "does not exist" error message.
	*/
	*cp = '\0';
	if (gather_nvlist(hdl, destsnap, NULL, NULL, B_FALSE, B_TRUE,
	B_FALSE, B_FALSE, B_FALSE, B_FALSE, B_FALSE, B_FALSE,
	B_TRUE, &local_nv, &local_avl) == 0) {
	*cp = '@';
	fs = fsavl_find(local_avl, drrb->drr_toguid, NULL);
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);

	if (fs != NULL) {
	if (flags->verbose) {
	(void) printf("snap %s already exists; "
	"ignoring\n", destsnap);
	}
	err = ioctl_err = recv_skip(hdl, infd,
	flags->byteswap);
	}
	}
	*cp = '@';
	}

	if (ioctl_err != 0) {
	switch (ioctl_errno) {
	case ENODEV:
	cp = strchr(destsnap, '@');
	*cp = '\0';
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"most recent snapshot of %s does not\n"
	"match incremental source"), destsnap);
	(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	*cp = '@';
	break;
	case ETXTBSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s has been modified\n"
	"since most recent snapshot"), name);
	(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	break;
	case EACCES:
	if (flags->heal) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"key must be loaded to do a non-raw "
	"corrective recv on an encrypted "
	"dataset."));
	} else if (raw && stream_wantsnewfs) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to create encryption key"));
	} else if (raw && !stream_wantsnewfs) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption key does not match "
	"existing key"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"inherited key must be loaded"));
	}
	(void) zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	break;
	case EEXIST:
	cp = strchr(destsnap, '@');
	if (newfs) {
	/* it's the containing fs that exists */
	*cp = '\0';
	}
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination already exists"));
	(void) zfs_error_fmt(hdl, EZFS_EXISTS,
	dgettext(TEXT_DOMAIN, "cannot restore to %s"),
	destsnap);
	*cp = '@';
	break;
	case EINVAL:
	if (flags->resumable) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"kernel modules must be upgraded to "
	"receive this stream."));
	} else if (embedded && !raw) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incompatible embedded data stream "
	"feature with encrypted receive."));
	}
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ECKSUM:
	case ZFS_ERR_STREAM_TRUNCATED:
	if (flags->heal)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective receive was not able to "
	"reconstruct the data needed for "
	"healing."));
	else
	recv_ecksum_set_aux(hdl, destsnap,
	flags->resumable, ioctl_err == ECKSUM);
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental send stream requires -L "
	"(--large-block), to match previous receive."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ENOTSUP:
	if (flags->heal)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream is not compatible with the "
	"data in the pool."));
	else
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to receive this "
	"stream."));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case EDQUOT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s space quota exceeded."), name);
	(void) zfs_error(hdl, EZFS_NOSPC, errbuf);
	break;
	case ZFS_ERR_FROM_IVSET_GUID_MISSING:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"IV set guid missing. See errata %u at "
	"https://openzfs.github.io/openzfs-docs/msg/"
	"ZFS-8000-ER."),
	ZPOOL_ERRATA_ZOL_8308_ENCRYPTION);
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_FROM_IVSET_GUID_MISMATCH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"IV set guid mismatch. See the 'zfs receive' "
	"man page section\n discussing the limitations "
	"of raw encrypted send streams."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_SPILL_BLOCK_FLAG_MISSING:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Spill block flag missing for raw send.\n"
	"The zfs software on the sending system must "
	"be updated."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	+ case ZFS_ERR_RESUME_EXISTS:
	+ cp = strchr(destsnap, '@');
	+ if (newfs) {
	+ /* it's the containing fs that exists */
	+ *cp = '\0';
	+ }
	+ zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	+ "Resuming recv on existing dataset without force"));
	+ (void) zfs_error_fmt(hdl, EZFS_RESUME_EXISTS,
	+ dgettext(TEXT_DOMAIN, "cannot resume recv %s"),
	+ destsnap);
	+ *cp = '@';
	+ break;
	case EBUSY:
	if (hastoken) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s contains "
	"partially-complete state from "
	"\"zfs receive -s\"."), name);
	(void) zfs_error(hdl, EZFS_BUSY, errbuf);
	break;
	}
	zfs_fallthrough;
	default:
	(void) zfs_standard_error(hdl, ioctl_errno, errbuf);
	}
	}

	/*
	* Mount the target filesystem (if created). Also mount any
	* children of the target filesystem if we did a replication
	* receive (indicated by stream_avl being non-NULL).
	*/
	if (clp) {
	if (!flags->nomount)
	err \|= changelist_postfix(clp);
	changelist_free(clp);
	}

	if ((newfs \|\| stream_avl) && type == ZFS_TYPE_FILESYSTEM && !redacted)
	flags->domount = B_TRUE;

	if (prop_errflags & ZPROP_ERR_NOCLEAR) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
	"failed to clear unreceived properties on %s"), name);
	(void) fprintf(stderr, "\n");
	}
	if (prop_errflags & ZPROP_ERR_NORESTORE) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
	"failed to restore original properties on %s"), name);
	(void) fprintf(stderr, "\n");
	}

	if (err \|\| ioctl_err) {
	err = -1;
	goto out;
	}

	if (flags->verbose) {
	char buf1[64];
	char buf2[64];
	uint64_t bytes = read_bytes;
	struct timespec delta;
	clock_gettime(CLOCK_MONOTONIC_RAW, &delta);
	if (begin_time.tv_nsec > delta.tv_nsec) {
	delta.tv_nsec =
	1000000000 + delta.tv_nsec - begin_time.tv_nsec;
	delta.tv_sec -= 1;
	} else
	delta.tv_nsec -= begin_time.tv_nsec;
	delta.tv_sec -= begin_time.tv_sec;
	if (delta.tv_sec == 0 && delta.tv_nsec == 0)
	delta.tv_nsec = 1;
	double delta_f = delta.tv_sec + (delta.tv_nsec / 1e9);
	zfs_nicebytes(bytes, buf1, sizeof (buf1));
	zfs_nicebytes(bytes / delta_f, buf2, sizeof (buf2));

	(void) printf("received %s stream in %.2f seconds (%s/sec)\n",
	buf1, delta_f, buf2);
	}

	err = 0;
	out:
	if (prop_errors != NULL)
	fnvlist_free(prop_errors);

	if (tmp_keylocation[0] != '\0') {
	fnvlist_add_string(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), tmp_keylocation);
	}

	if (newprops)
	fnvlist_free(rcvprops);

	fnvlist_free(oxprops);
	fnvlist_free(origprops);

	return (err);
	}

	/*
	* Check properties we were asked to override (both -o\|-x)
	*/
	static boolean_t
	zfs_receive_checkprops(libzfs_handle_t hdl, nvlist_t props,
	const char *errbuf)
	{
	nvpair_t *nvp = NULL;
	zfs_prop_t prop;
	const char *name;

	while ((nvp = nvlist_next_nvpair(props, nvp)) != NULL) {
	name = nvpair_name(nvp);
	prop = zfs_name_to_prop(name);

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"%s: invalid property '%s'"), errbuf, name);
	return (B_FALSE);
	}
	continue;
	}
	/*
	* "origin" is readonly but is used to receive datasets as
	* clones so we don't raise an error here
	*/
	if (prop == ZFS_PROP_ORIGIN)
	continue;

	/* encryption params have their own verification later */
	if (prop == ZFS_PROP_ENCRYPTION \|\|
	zfs_prop_encryption_key_param(prop))
	continue;

	/*
	* cannot override readonly, set-once and other specific
	* settable properties
	*/
	if (zfs_prop_readonly(prop) \|\| prop == ZFS_PROP_VERSION \|\|
	prop == ZFS_PROP_VOLSIZE) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"%s: invalid property '%s'"), errbuf, name);
	return (B_FALSE);
	}
	}

	return (B_TRUE);
	}

	static int
	zfs_receive_impl(libzfs_handle_t hdl, const char tosnap,
	const char originsnap, recvflags_t flags, int infd, const char *sendfs,
	nvlist_t stream_nv, avl_tree_t stream_avl, char **top_zfs,
	const char finalsnap, nvlist_t cmdprops)
	{
	int err;
	dmu_replay_record_t drr, drr_noswap;
	struct drr_begin *drrb = &drr.drr_u.drr_begin;
	char errbuf[ERRBUFLEN];
	zio_cksum_t zcksum = { { 0 } };
	uint64_t featureflags;
	int hdrtype;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	/* check cmdline props, raise an error if they cannot be received */
	if (!zfs_receive_checkprops(hdl, cmdprops, errbuf))
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));

	if (flags->isprefix &&
	!zfs_dataset_exists(hdl, tosnap, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified fs "
	"(%s) does not exist"), tosnap);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}
	if (originsnap &&
	!zfs_dataset_exists(hdl, originsnap, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified origin fs "
	"(%s) does not exist"), originsnap);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}

	/* read in the BEGIN record */
	if (0 != (err = recv_read(hdl, infd, &drr, sizeof (drr), B_FALSE,
	&zcksum)))
	return (err);

	if (drr.drr_type == DRR_END \|\| drr.drr_type == BSWAP_32(DRR_END)) {
	/* It's the double end record at the end of a package */
	return (ENODATA);
	}

	/* the kernel needs the non-byteswapped begin record */
	drr_noswap = drr;

	flags->byteswap = B_FALSE;
	if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
	/*
	* We computed the checksum in the wrong byteorder in
	* recv_read() above; do it again correctly.
	*/
	memset(&zcksum, 0, sizeof (zio_cksum_t));
	fletcher_4_incremental_byteswap(&drr, sizeof (drr), &zcksum);
	flags->byteswap = B_TRUE;

	drr.drr_type = BSWAP_32(drr.drr_type);
	drr.drr_payloadlen = BSWAP_32(drr.drr_payloadlen);
	drrb->drr_magic = BSWAP_64(drrb->drr_magic);
	drrb->drr_versioninfo = BSWAP_64(drrb->drr_versioninfo);
	drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time);
	drrb->drr_type = BSWAP_32(drrb->drr_type);
	drrb->drr_flags = BSWAP_32(drrb->drr_flags);
	drrb->drr_toguid = BSWAP_64(drrb->drr_toguid);
	drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid);
	}

	if (drrb->drr_magic != DMU_BACKUP_MAGIC \|\| drr.drr_type != DRR_BEGIN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (bad magic number)"));
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	hdrtype = DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo);

	if (!DMU_STREAM_SUPPORTED(featureflags) \|\|
	(hdrtype != DMU_SUBSTREAM && hdrtype != DMU_COMPOUNDSTREAM)) {
	/*
	* Let's be explicit about this one, since rather than
	* being a new feature we can't know, it's an old
	* feature we dropped.
	*/
	if (featureflags & DMU_BACKUP_FEATURE_DEDUP) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream has deprecated feature: dedup, try "
	"'zstream redup [send in a file] \| zfs recv "
	"[...]'"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream has unsupported feature, feature flags = "
	"%llx (unknown flags = %llx)"),
	(u_longlong_t)featureflags,
	(u_longlong_t)((featureflags) &
	~DMU_BACKUP_FEATURE_MASK));
	}
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	/* Holds feature is set once in the compound stream header. */
	if (featureflags & DMU_BACKUP_FEATURE_HOLDS)
	flags->holds = B_TRUE;

	if (strchr(drrb->drr_toname, '@') == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (bad snapshot name)"));
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_SUBSTREAM) {
	char nonpackage_sendfs[ZFS_MAX_DATASET_NAME_LEN];
	if (sendfs == NULL) {
	/*
	* We were not called from zfs_receive_package(). Get
	* the fs specified by 'zfs send'.
	*/
	char *cp;
	(void) strlcpy(nonpackage_sendfs,
	drr.drr_u.drr_begin.drr_toname,
	sizeof (nonpackage_sendfs));
	if ((cp = strchr(nonpackage_sendfs, '@')) != NULL)
	*cp = '\0';
	sendfs = nonpackage_sendfs;
	VERIFY(finalsnap == NULL);
	}
	return (zfs_receive_one(hdl, infd, tosnap, originsnap, flags,
	&drr, &drr_noswap, sendfs, stream_nv, stream_avl, top_zfs,
	finalsnap, cmdprops));
	} else {
	assert(DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM);
	return (zfs_receive_package(hdl, infd, tosnap, flags, &drr,
	&zcksum, top_zfs, cmdprops));
	}
	}

	/*
	* Restores a backup of tosnap from the file descriptor specified by infd.
	* Return 0 on total success, -2 if some things couldn't be
	* destroyed/renamed/promoted, -1 if some things couldn't be received.
	* (-1 will override -2, if -1 and the resumable flag was specified the
	* transfer can be resumed if the sending side supports it).
	*/
	int
	zfs_receive(libzfs_handle_t hdl, const char tosnap, nvlist_t *props,
	recvflags_t flags, int infd, avl_tree_t stream_avl)
	{
	char *top_zfs = NULL;
	int err;
	struct stat sb;
	char *originsnap = NULL;

	/*
	* The only way fstat can fail is if we do not have a valid file
	* descriptor.
	*/
	if (fstat(infd, &sb) == -1) {
	perror("fstat");
	return (-2);
	}

	if (props) {
	err = nvlist_lookup_string(props, "origin", &originsnap);
	if (err && err != ENOENT)
	return (err);
	}

	err = zfs_receive_impl(hdl, tosnap, originsnap, flags, infd, NULL, NULL,
	stream_avl, &top_zfs, NULL, props);

	if (err == 0 && !flags->nomount && flags->domount && top_zfs) {
	zfs_handle_t *zhp = NULL;
	prop_changelist_t *clp = NULL;

	zhp = zfs_open(hdl, top_zfs,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	err = -1;
	goto out;
	} else {
	if (zhp->zfs_type == ZFS_TYPE_VOLUME) {
	zfs_close(zhp);
	goto out;
	}

	clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT,
	CL_GATHER_MOUNT_ALWAYS,
	flags->forceunmount ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp == NULL) {
	err = -1;
	goto out;
	}

	/* mount and share received datasets */
	err = changelist_postfix(clp);
	changelist_free(clp);
	if (err != 0)
	err = -1;
	}
	}

	out:
	if (top_zfs)
	free(top_zfs);

	return (err);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_util.c b/sys/contrib/openzfs/lib/libzfs/libzfs_util.c
	index bc00a8dffd81..28dd4f426a96 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_util.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_util.c
	@@ -1,2039 +1,2042 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2020 Joyent, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright (c) 2017 Datto Inc.
	* Copyright (c) 2020 The FreeBSD Foundation
	*
	* Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	*/

	/*
	* Internal utility routines for the ZFS library.
	*/

	#include <errno.h>
	#include <fcntl.h>
	#include <libintl.h>
	#include <stdarg.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <strings.h>
	#include <unistd.h>
	#include <math.h>
	#if LIBFETCH_DYNAMIC
	#include <dlfcn.h>
	#endif
	#include <sys/stat.h>
	#include <sys/mnttab.h>
	#include <sys/mntent.h>
	#include <sys/types.h>
	#include <sys/wait.h>

	#include <libzfs.h>
	#include <libzfs_core.h>

	#include "libzfs_impl.h"
	#include "zfs_prop.h"
	#include "zfeature_common.h"
	#include <zfs_fletcher.h>
	#include <libzutil.h>

	/*
	* We only care about the scheme in order to match the scheme
	* with the handler. Each handler should validate the full URI
	* as necessary.
	*/
	#define URI_REGEX "^\$[A-Za-z][A-Za-z0-9+.\\-]*\$:"

	int
	libzfs_errno(libzfs_handle_t *hdl)
	{
	return (hdl->libzfs_error);
	}

	const char *
	libzfs_error_action(libzfs_handle_t *hdl)
	{
	return (hdl->libzfs_action);
	}

	const char *
	libzfs_error_description(libzfs_handle_t *hdl)
	{
	if (hdl->libzfs_desc[0] != '\0')
	return (hdl->libzfs_desc);

	switch (hdl->libzfs_error) {
	case EZFS_NOMEM:
	return (dgettext(TEXT_DOMAIN, "out of memory"));
	case EZFS_BADPROP:
	return (dgettext(TEXT_DOMAIN, "invalid property value"));
	case EZFS_PROPREADONLY:
	return (dgettext(TEXT_DOMAIN, "read-only property"));
	case EZFS_PROPTYPE:
	return (dgettext(TEXT_DOMAIN, "property doesn't apply to "
	"datasets of this type"));
	case EZFS_PROPNONINHERIT:
	return (dgettext(TEXT_DOMAIN, "property cannot be inherited"));
	case EZFS_PROPSPACE:
	return (dgettext(TEXT_DOMAIN, "invalid quota or reservation"));
	case EZFS_BADTYPE:
	return (dgettext(TEXT_DOMAIN, "operation not applicable to "
	"datasets of this type"));
	case EZFS_BUSY:
	return (dgettext(TEXT_DOMAIN, "pool or dataset is busy"));
	case EZFS_EXISTS:
	return (dgettext(TEXT_DOMAIN, "pool or dataset exists"));
	case EZFS_NOENT:
	return (dgettext(TEXT_DOMAIN, "no such pool or dataset"));
	case EZFS_BADSTREAM:
	return (dgettext(TEXT_DOMAIN, "invalid backup stream"));
	case EZFS_DSREADONLY:
	return (dgettext(TEXT_DOMAIN, "dataset is read-only"));
	case EZFS_VOLTOOBIG:
	return (dgettext(TEXT_DOMAIN, "volume size exceeds limit for "
	"this system"));
	case EZFS_INVALIDNAME:
	return (dgettext(TEXT_DOMAIN, "invalid name"));
	case EZFS_BADRESTORE:
	return (dgettext(TEXT_DOMAIN, "unable to restore to "
	"destination"));
	case EZFS_BADBACKUP:
	return (dgettext(TEXT_DOMAIN, "backup failed"));
	case EZFS_BADTARGET:
	return (dgettext(TEXT_DOMAIN, "invalid target vdev"));
	case EZFS_NODEVICE:
	return (dgettext(TEXT_DOMAIN, "no such device in pool"));
	case EZFS_BADDEV:
	return (dgettext(TEXT_DOMAIN, "invalid device"));
	case EZFS_NOREPLICAS:
	return (dgettext(TEXT_DOMAIN, "no valid replicas"));
	case EZFS_RESILVERING:
	return (dgettext(TEXT_DOMAIN, "currently resilvering"));
	case EZFS_BADVERSION:
	return (dgettext(TEXT_DOMAIN, "unsupported version or "
	"feature"));
	case EZFS_POOLUNAVAIL:
	return (dgettext(TEXT_DOMAIN, "pool is unavailable"));
	case EZFS_DEVOVERFLOW:
	return (dgettext(TEXT_DOMAIN, "too many devices in one vdev"));
	case EZFS_BADPATH:
	return (dgettext(TEXT_DOMAIN, "must be an absolute path"));
	case EZFS_CROSSTARGET:
	return (dgettext(TEXT_DOMAIN, "operation crosses datasets or "
	"pools"));
	case EZFS_ZONED:
	return (dgettext(TEXT_DOMAIN, "dataset in use by local zone"));
	case EZFS_MOUNTFAILED:
	return (dgettext(TEXT_DOMAIN, "mount failed"));
	case EZFS_UMOUNTFAILED:
	return (dgettext(TEXT_DOMAIN, "unmount failed"));
	case EZFS_UNSHARENFSFAILED:
	return (dgettext(TEXT_DOMAIN, "NFS share removal failed"));
	case EZFS_SHARENFSFAILED:
	return (dgettext(TEXT_DOMAIN, "NFS share creation failed"));
	case EZFS_UNSHARESMBFAILED:
	return (dgettext(TEXT_DOMAIN, "SMB share removal failed"));
	case EZFS_SHARESMBFAILED:
	return (dgettext(TEXT_DOMAIN, "SMB share creation failed"));
	case EZFS_PERM:
	return (dgettext(TEXT_DOMAIN, "permission denied"));
	case EZFS_NOSPC:
	return (dgettext(TEXT_DOMAIN, "out of space"));
	case EZFS_FAULT:
	return (dgettext(TEXT_DOMAIN, "bad address"));
	case EZFS_IO:
	return (dgettext(TEXT_DOMAIN, "I/O error"));
	case EZFS_INTR:
	return (dgettext(TEXT_DOMAIN, "signal received"));
	case EZFS_CKSUM:
	return (dgettext(TEXT_DOMAIN, "insufficient replicas"));
	case EZFS_ISSPARE:
	return (dgettext(TEXT_DOMAIN, "device is reserved as a hot "
	"spare"));
	case EZFS_INVALCONFIG:
	return (dgettext(TEXT_DOMAIN, "invalid vdev configuration"));
	case EZFS_RECURSIVE:
	return (dgettext(TEXT_DOMAIN, "recursive dataset dependency"));
	case EZFS_NOHISTORY:
	return (dgettext(TEXT_DOMAIN, "no history available"));
	case EZFS_POOLPROPS:
	return (dgettext(TEXT_DOMAIN, "failed to retrieve "
	"pool properties"));
	case EZFS_POOL_NOTSUP:
	return (dgettext(TEXT_DOMAIN, "operation not supported "
	"on this type of pool"));
	case EZFS_POOL_INVALARG:
	return (dgettext(TEXT_DOMAIN, "invalid argument for "
	"this pool operation"));
	case EZFS_NAMETOOLONG:
	return (dgettext(TEXT_DOMAIN, "dataset name is too long"));
	case EZFS_OPENFAILED:
	return (dgettext(TEXT_DOMAIN, "open failed"));
	case EZFS_NOCAP:
	return (dgettext(TEXT_DOMAIN,
	"disk capacity information could not be retrieved"));
	case EZFS_LABELFAILED:
	return (dgettext(TEXT_DOMAIN, "write of label failed"));
	case EZFS_BADWHO:
	return (dgettext(TEXT_DOMAIN, "invalid user/group"));
	case EZFS_BADPERM:
	return (dgettext(TEXT_DOMAIN, "invalid permission"));
	case EZFS_BADPERMSET:
	return (dgettext(TEXT_DOMAIN, "invalid permission set name"));
	case EZFS_NODELEGATION:
	return (dgettext(TEXT_DOMAIN, "delegated administration is "
	"disabled on pool"));
	case EZFS_BADCACHE:
	return (dgettext(TEXT_DOMAIN, "invalid or missing cache file"));
	case EZFS_ISL2CACHE:
	return (dgettext(TEXT_DOMAIN, "device is in use as a cache"));
	case EZFS_VDEVNOTSUP:
	return (dgettext(TEXT_DOMAIN, "vdev specification is not "
	"supported"));
	case EZFS_NOTSUP:
	return (dgettext(TEXT_DOMAIN, "operation not supported "
	"on this dataset"));
	case EZFS_IOC_NOTSUPPORTED:
	return (dgettext(TEXT_DOMAIN, "operation not supported by "
	"zfs kernel module"));
	case EZFS_ACTIVE_SPARE:
	return (dgettext(TEXT_DOMAIN, "pool has active shared spare "
	"device"));
	case EZFS_UNPLAYED_LOGS:
	return (dgettext(TEXT_DOMAIN, "log device has unplayed intent "
	"logs"));
	case EZFS_REFTAG_RELE:
	return (dgettext(TEXT_DOMAIN, "no such tag on this dataset"));
	case EZFS_REFTAG_HOLD:
	return (dgettext(TEXT_DOMAIN, "tag already exists on this "
	"dataset"));
	case EZFS_TAGTOOLONG:
	return (dgettext(TEXT_DOMAIN, "tag too long"));
	case EZFS_PIPEFAILED:
	return (dgettext(TEXT_DOMAIN, "pipe create failed"));
	case EZFS_THREADCREATEFAILED:
	return (dgettext(TEXT_DOMAIN, "thread create failed"));
	case EZFS_POSTSPLIT_ONLINE:
	return (dgettext(TEXT_DOMAIN, "disk was split from this pool "
	"into a new one"));
	case EZFS_SCRUB_PAUSED:
	return (dgettext(TEXT_DOMAIN, "scrub is paused; "
	"use 'zpool scrub' to resume"));
	case EZFS_SCRUBBING:
	return (dgettext(TEXT_DOMAIN, "currently scrubbing; "
	"use 'zpool scrub -s' to cancel current scrub"));
	case EZFS_NO_SCRUB:
	return (dgettext(TEXT_DOMAIN, "there is no active scrub"));
	case EZFS_DIFF:
	return (dgettext(TEXT_DOMAIN, "unable to generate diffs"));
	case EZFS_DIFFDATA:
	return (dgettext(TEXT_DOMAIN, "invalid diff data"));
	case EZFS_POOLREADONLY:
	return (dgettext(TEXT_DOMAIN, "pool is read-only"));
	case EZFS_NO_PENDING:
	return (dgettext(TEXT_DOMAIN, "operation is not "
	"in progress"));
	case EZFS_CHECKPOINT_EXISTS:
	return (dgettext(TEXT_DOMAIN, "checkpoint exists"));
	case EZFS_DISCARDING_CHECKPOINT:
	return (dgettext(TEXT_DOMAIN, "currently discarding "
	"checkpoint"));
	case EZFS_NO_CHECKPOINT:
	return (dgettext(TEXT_DOMAIN, "checkpoint does not exist"));
	case EZFS_DEVRM_IN_PROGRESS:
	return (dgettext(TEXT_DOMAIN, "device removal in progress"));
	case EZFS_VDEV_TOO_BIG:
	return (dgettext(TEXT_DOMAIN, "device exceeds supported size"));
	case EZFS_ACTIVE_POOL:
	return (dgettext(TEXT_DOMAIN, "pool is imported on a "
	"different host"));
	case EZFS_CRYPTOFAILED:
	return (dgettext(TEXT_DOMAIN, "encryption failure"));
	case EZFS_TOOMANY:
	return (dgettext(TEXT_DOMAIN, "argument list too long"));
	case EZFS_INITIALIZING:
	return (dgettext(TEXT_DOMAIN, "currently initializing"));
	case EZFS_NO_INITIALIZE:
	return (dgettext(TEXT_DOMAIN, "there is no active "
	"initialization"));
	case EZFS_WRONG_PARENT:
	return (dgettext(TEXT_DOMAIN, "invalid parent dataset"));
	case EZFS_TRIMMING:
	return (dgettext(TEXT_DOMAIN, "currently trimming"));
	case EZFS_NO_TRIM:
	return (dgettext(TEXT_DOMAIN, "there is no active trim"));
	case EZFS_TRIM_NOTSUP:
	return (dgettext(TEXT_DOMAIN, "trim operations are not "
	"supported by this device"));
	case EZFS_NO_RESILVER_DEFER:
	return (dgettext(TEXT_DOMAIN, "this action requires the "
	"resilver_defer feature"));
	case EZFS_EXPORT_IN_PROGRESS:
	return (dgettext(TEXT_DOMAIN, "pool export in progress"));
	case EZFS_REBUILDING:
	return (dgettext(TEXT_DOMAIN, "currently sequentially "
	"resilvering"));
	case EZFS_VDEV_NOTSUP:
	return (dgettext(TEXT_DOMAIN, "operation not supported "
	"on this type of vdev"));
	case EZFS_NOT_USER_NAMESPACE:
	return (dgettext(TEXT_DOMAIN, "the provided file "
	"was not a user namespace file"));
	+ case EZFS_RESUME_EXISTS:
	+ return (dgettext(TEXT_DOMAIN, "Resuming recv on existing "
	+ "dataset without force"));
	case EZFS_UNKNOWN:
	return (dgettext(TEXT_DOMAIN, "unknown error"));
	default:
	assert(hdl->libzfs_error == 0);
	return (dgettext(TEXT_DOMAIN, "no error"));
	}
	}

	void
	zfs_error_aux(libzfs_handle_t hdl, const char fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	(void) vsnprintf(hdl->libzfs_desc, sizeof (hdl->libzfs_desc),
	fmt, ap);
	hdl->libzfs_desc_active = 1;

	va_end(ap);
	}

	static void
	zfs_verror(libzfs_handle_t hdl, int error, const char fmt, va_list ap)
	{
	(void) vsnprintf(hdl->libzfs_action, sizeof (hdl->libzfs_action),
	fmt, ap);
	hdl->libzfs_error = error;

	if (hdl->libzfs_desc_active)
	hdl->libzfs_desc_active = 0;
	else
	hdl->libzfs_desc[0] = '\0';

	if (hdl->libzfs_printerr) {
	if (error == EZFS_UNKNOWN) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "internal "
	"error: %s: %s\n"), hdl->libzfs_action,
	libzfs_error_description(hdl));
	abort();
	}

	(void) fprintf(stderr, "%s: %s\n", hdl->libzfs_action,
	libzfs_error_description(hdl));
	if (error == EZFS_NOMEM)
	exit(1);
	}
	}

	int
	zfs_error(libzfs_handle_t hdl, int error, const char msg)
	{
	return (zfs_error_fmt(hdl, error, "%s", msg));
	}

	int
	zfs_error_fmt(libzfs_handle_t hdl, int error, const char fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	zfs_verror(hdl, error, fmt, ap);

	va_end(ap);

	return (-1);
	}

	static int
	zfs_common_error(libzfs_handle_t hdl, int error, const char fmt,
	va_list ap)
	{
	switch (error) {
	case EPERM:
	case EACCES:
	zfs_verror(hdl, EZFS_PERM, fmt, ap);
	return (-1);

	case ECANCELED:
	zfs_verror(hdl, EZFS_NODELEGATION, fmt, ap);
	return (-1);

	case EIO:
	zfs_verror(hdl, EZFS_IO, fmt, ap);
	return (-1);

	case EFAULT:
	zfs_verror(hdl, EZFS_FAULT, fmt, ap);
	return (-1);

	case EINTR:
	zfs_verror(hdl, EZFS_INTR, fmt, ap);
	return (-1);

	case ECKSUM:
	zfs_verror(hdl, EZFS_CKSUM, fmt, ap);
	return (-1);
	}

	return (0);
	}

	int
	zfs_standard_error(libzfs_handle_t hdl, int error, const char msg)
	{
	return (zfs_standard_error_fmt(hdl, error, "%s", msg));
	}

	int
	zfs_standard_error_fmt(libzfs_handle_t hdl, int error, const char fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	if (zfs_common_error(hdl, error, fmt, ap) != 0) {
	va_end(ap);
	return (-1);
	}

	switch (error) {
	case ENXIO:
	case ENODEV:
	case EPIPE:
	zfs_verror(hdl, EZFS_IO, fmt, ap);
	break;

	case ENOENT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset does not exist"));
	zfs_verror(hdl, EZFS_NOENT, fmt, ap);
	break;

	case ENOSPC:
	case EDQUOT:
	zfs_verror(hdl, EZFS_NOSPC, fmt, ap);
	break;

	case EEXIST:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset already exists"));
	zfs_verror(hdl, EZFS_EXISTS, fmt, ap);
	break;

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset is busy"));
	zfs_verror(hdl, EZFS_BUSY, fmt, ap);
	break;
	case EROFS:
	zfs_verror(hdl, EZFS_POOLREADONLY, fmt, ap);
	break;
	case ENAMETOOLONG:
	zfs_verror(hdl, EZFS_NAMETOOLONG, fmt, ap);
	break;
	case ENOTSUP:
	zfs_verror(hdl, EZFS_BADVERSION, fmt, ap);
	break;
	case EAGAIN:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool I/O is currently suspended"));
	zfs_verror(hdl, EZFS_POOLUNAVAIL, fmt, ap);
	break;
	case EREMOTEIO:
	zfs_verror(hdl, EZFS_ACTIVE_POOL, fmt, ap);
	break;
	case ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE:
	case ZFS_ERR_IOC_CMD_UNAVAIL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
	"module does not support this operation. A reboot may "
	"be required to enable this operation."));
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	case ZFS_ERR_IOC_ARG_UNAVAIL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
	"module does not support an option for this operation. "
	"A reboot may be required to enable this option."));
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	case ZFS_ERR_IOC_ARG_REQUIRED:
	case ZFS_ERR_IOC_ARG_BADTYPE:
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	case ZFS_ERR_WRONG_PARENT:
	zfs_verror(hdl, EZFS_WRONG_PARENT, fmt, ap);
	break;
	case ZFS_ERR_BADPROP:
	zfs_verror(hdl, EZFS_BADPROP, fmt, ap);
	break;
	case ZFS_ERR_NOT_USER_NAMESPACE:
	zfs_verror(hdl, EZFS_NOT_USER_NAMESPACE, fmt, ap);
	break;
	default:
	zfs_error_aux(hdl, "%s", strerror(error));
	zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);
	break;
	}

	va_end(ap);
	return (-1);
	}

	void
	zfs_setprop_error(libzfs_handle_t *hdl, zfs_prop_t prop, int err,
	char *errbuf)
	{
	switch (err) {

	case ENOSPC:
	/*
	* For quotas and reservations, ENOSPC indicates
	* something different; setting a quota or reservation
	* doesn't use any disk space.
	*/
	switch (prop) {
	case ZFS_PROP_QUOTA:
	case ZFS_PROP_REFQUOTA:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"size is less than current used or "
	"reserved space"));
	(void) zfs_error(hdl, EZFS_PROPSPACE, errbuf);
	break;

	case ZFS_PROP_RESERVATION:
	case ZFS_PROP_REFRESERVATION:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"size is greater than available space"));
	(void) zfs_error(hdl, EZFS_PROPSPACE, errbuf);
	break;

	default:
	(void) zfs_standard_error(hdl, err, errbuf);
	break;
	}
	break;

	case EBUSY:
	(void) zfs_standard_error(hdl, EBUSY, errbuf);
	break;

	case EROFS:
	(void) zfs_error(hdl, EZFS_DSREADONLY, errbuf);
	break;

	case E2BIG:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property value too long"));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	break;

	case ENOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool and or dataset must be upgraded to set this "
	"property or value"));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;

	case ERANGE:
	if (prop == ZFS_PROP_COMPRESSION \|\|
	prop == ZFS_PROP_DNODESIZE \|\|
	prop == ZFS_PROP_RECORDSIZE) {
	(void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property setting is not allowed on "
	"bootable datasets"));
	(void) zfs_error(hdl, EZFS_NOTSUP, errbuf);
	} else if (prop == ZFS_PROP_CHECKSUM \|\|
	prop == ZFS_PROP_DEDUP) {
	(void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property setting is not allowed on "
	"root pools"));
	(void) zfs_error(hdl, EZFS_NOTSUP, errbuf);
	} else {
	(void) zfs_standard_error(hdl, err, errbuf);
	}
	break;

	case EINVAL:
	if (prop == ZPROP_INVAL) {
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	} else {
	(void) zfs_standard_error(hdl, err, errbuf);
	}
	break;

	case ZFS_ERR_BADPROP:
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	break;

	case EACCES:
	if (prop == ZFS_PROP_KEYLOCATION) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"keylocation may only be set on encryption roots"));
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	} else {
	(void) zfs_standard_error(hdl, err, errbuf);
	}
	break;

	case EOVERFLOW:
	/*
	* This platform can't address a volume this big.
	*/
	#ifdef _ILP32
	if (prop == ZFS_PROP_VOLSIZE) {
	(void) zfs_error(hdl, EZFS_VOLTOOBIG, errbuf);
	break;
	}
	zfs_fallthrough;
	#endif
	default:
	(void) zfs_standard_error(hdl, err, errbuf);
	}
	}

	int
	zpool_standard_error(libzfs_handle_t hdl, int error, const char msg)
	{
	return (zpool_standard_error_fmt(hdl, error, "%s", msg));
	}

	int
	zpool_standard_error_fmt(libzfs_handle_t hdl, int error, const char fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	if (zfs_common_error(hdl, error, fmt, ap) != 0) {
	va_end(ap);
	return (-1);
	}

	switch (error) {
	case ENODEV:
	zfs_verror(hdl, EZFS_NODEVICE, fmt, ap);
	break;

	case ENOENT:
	zfs_error_aux(hdl,
	dgettext(TEXT_DOMAIN, "no such pool or dataset"));
	zfs_verror(hdl, EZFS_NOENT, fmt, ap);
	break;

	case EEXIST:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool already exists"));
	zfs_verror(hdl, EZFS_EXISTS, fmt, ap);
	break;

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool is busy"));
	zfs_verror(hdl, EZFS_BUSY, fmt, ap);
	break;

	/* There is no pending operation to cancel */
	case ENOTACTIVE:
	zfs_verror(hdl, EZFS_NO_PENDING, fmt, ap);
	break;

	case ENXIO:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"one or more devices is currently unavailable"));
	zfs_verror(hdl, EZFS_BADDEV, fmt, ap);
	break;

	case ENAMETOOLONG:
	zfs_verror(hdl, EZFS_DEVOVERFLOW, fmt, ap);
	break;

	case ENOTSUP:
	zfs_verror(hdl, EZFS_POOL_NOTSUP, fmt, ap);
	break;

	case EINVAL:
	zfs_verror(hdl, EZFS_POOL_INVALARG, fmt, ap);
	break;

	case ENOSPC:
	case EDQUOT:
	zfs_verror(hdl, EZFS_NOSPC, fmt, ap);
	break;

	case EAGAIN:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool I/O is currently suspended"));
	zfs_verror(hdl, EZFS_POOLUNAVAIL, fmt, ap);
	break;

	case EROFS:
	zfs_verror(hdl, EZFS_POOLREADONLY, fmt, ap);
	break;
	case EDOM:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"block size out of range or does not match"));
	zfs_verror(hdl, EZFS_BADPROP, fmt, ap);
	break;
	case EREMOTEIO:
	zfs_verror(hdl, EZFS_ACTIVE_POOL, fmt, ap);
	break;
	case ZFS_ERR_CHECKPOINT_EXISTS:
	zfs_verror(hdl, EZFS_CHECKPOINT_EXISTS, fmt, ap);
	break;
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	zfs_verror(hdl, EZFS_DISCARDING_CHECKPOINT, fmt, ap);
	break;
	case ZFS_ERR_NO_CHECKPOINT:
	zfs_verror(hdl, EZFS_NO_CHECKPOINT, fmt, ap);
	break;
	case ZFS_ERR_DEVRM_IN_PROGRESS:
	zfs_verror(hdl, EZFS_DEVRM_IN_PROGRESS, fmt, ap);
	break;
	case ZFS_ERR_VDEV_TOO_BIG:
	zfs_verror(hdl, EZFS_VDEV_TOO_BIG, fmt, ap);
	break;
	case ZFS_ERR_EXPORT_IN_PROGRESS:
	zfs_verror(hdl, EZFS_EXPORT_IN_PROGRESS, fmt, ap);
	break;
	case ZFS_ERR_RESILVER_IN_PROGRESS:
	zfs_verror(hdl, EZFS_RESILVERING, fmt, ap);
	break;
	case ZFS_ERR_REBUILD_IN_PROGRESS:
	zfs_verror(hdl, EZFS_REBUILDING, fmt, ap);
	break;
	case ZFS_ERR_BADPROP:
	zfs_verror(hdl, EZFS_BADPROP, fmt, ap);
	break;
	case ZFS_ERR_VDEV_NOTSUP:
	zfs_verror(hdl, EZFS_VDEV_NOTSUP, fmt, ap);
	break;
	case ZFS_ERR_IOC_CMD_UNAVAIL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
	"module does not support this operation. A reboot may "
	"be required to enable this operation."));
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	case ZFS_ERR_IOC_ARG_UNAVAIL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
	"module does not support an option for this operation. "
	"A reboot may be required to enable this option."));
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	case ZFS_ERR_IOC_ARG_REQUIRED:
	case ZFS_ERR_IOC_ARG_BADTYPE:
	zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
	break;
	default:
	zfs_error_aux(hdl, "%s", strerror(error));
	zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);
	}

	va_end(ap);
	return (-1);
	}

	/*
	* Display an out of memory error message and abort the current program.
	*/
	int
	no_memory(libzfs_handle_t *hdl)
	{
	return (zfs_error(hdl, EZFS_NOMEM, "internal error"));
	}

	/*
	* A safe form of malloc() which will die if the allocation fails.
	*/
	void *
	zfs_alloc(libzfs_handle_t *hdl, size_t size)
	{
	void *data;

	if ((data = calloc(1, size)) == NULL)
	(void) no_memory(hdl);

	return (data);
	}

	/*
	* A safe form of asprintf() which will die if the allocation fails.
	*/
	char *
	zfs_asprintf(libzfs_handle_t hdl, const char fmt, ...)
	{
	va_list ap;
	char *ret;
	int err;

	va_start(ap, fmt);

	err = vasprintf(&ret, fmt, ap);

	va_end(ap);

	if (err < 0) {
	(void) no_memory(hdl);
	ret = NULL;
	}

	return (ret);
	}

	/*
	* A safe form of realloc(), which also zeroes newly allocated space.
	*/
	void *
	zfs_realloc(libzfs_handle_t hdl, void ptr, size_t oldsize, size_t newsize)
	{
	void *ret;

	if ((ret = realloc(ptr, newsize)) == NULL) {
	(void) no_memory(hdl);
	return (NULL);
	}

	memset((char *)ret + oldsize, 0, newsize - oldsize);
	return (ret);
	}

	/*
	* A safe form of strdup() which will die if the allocation fails.
	*/
	char *
	zfs_strdup(libzfs_handle_t hdl, const char str)
	{
	char *ret;

	if ((ret = strdup(str)) == NULL)
	(void) no_memory(hdl);

	return (ret);
	}

	void
	libzfs_print_on_error(libzfs_handle_t *hdl, boolean_t printerr)
	{
	hdl->libzfs_printerr = printerr;
	}

	/*
	* Read lines from an open file descriptor and store them in an array of
	* strings until EOF. lines[] will be allocated and populated with all the
	* lines read. All newlines are replaced with NULL terminators for
	* convenience. lines[] must be freed after use with libzfs_free_str_array().
	*
	* Returns the number of lines read.
	*/
	static int
	libzfs_read_stdout_from_fd(int fd, char **lines[])
	{

	FILE *fp;
	int lines_cnt = 0;
	size_t len = 0;
	char *line = NULL;
	char tmp_lines = NULL, tmp;

	fp = fdopen(fd, "r");
	if (fp == NULL) {
	close(fd);
	return (0);
	}
	while (getline(&line, &len, fp) != -1) {
	tmp = realloc(tmp_lines, sizeof (tmp_lines) (lines_cnt + 1));
	if (tmp == NULL) {
	/* Return the lines we were able to process */
	break;
	}
	tmp_lines = tmp;

	/* Remove newline if not EOF */
	if (line[strlen(line) - 1] == '\n')
	line[strlen(line) - 1] = '\0';

	tmp_lines[lines_cnt] = strdup(line);
	if (tmp_lines[lines_cnt] == NULL)
	break;
	++lines_cnt;
	}
	free(line);
	fclose(fp);
	*lines = tmp_lines;
	return (lines_cnt);
	}

	static int
	libzfs_run_process_impl(const char path, char argv[], char *env[], int flags,
	char *lines[], int lines_cnt)
	{
	pid_t pid;
	int error, devnull_fd;
	int link[2];

	/*
	* Setup a pipe between our child and parent process if we're
	* reading stdout.
	*/
	if (lines != NULL && pipe2(link, O_NONBLOCK \| O_CLOEXEC) == -1)
	return (-EPIPE);

	pid = fork();
	if (pid == 0) {
	/* Child process */
	devnull_fd = open("/dev/null", O_WRONLY \| O_CLOEXEC);

	if (devnull_fd < 0)
	_exit(-1);

	if (!(flags & STDOUT_VERBOSE) && (lines == NULL))
	(void) dup2(devnull_fd, STDOUT_FILENO);
	else if (lines != NULL) {
	/* Save the output to lines[] */
	dup2(link[1], STDOUT_FILENO);
	}

	if (!(flags & STDERR_VERBOSE))
	(void) dup2(devnull_fd, STDERR_FILENO);

	if (flags & NO_DEFAULT_PATH) {
	if (env == NULL)
	execv(path, argv);
	else
	execve(path, argv, env);
	} else {
	if (env == NULL)
	execvp(path, argv);
	else
	execvpe(path, argv, env);
	}

	_exit(-1);
	} else if (pid > 0) {
	/* Parent process */
	int status;

	while ((error = waitpid(pid, &status, 0)) == -1 &&
	errno == EINTR)
	;
	if (error < 0 \|\| !WIFEXITED(status))
	return (-1);

	if (lines != NULL) {
	close(link[1]);
	*lines_cnt = libzfs_read_stdout_from_fd(link[0], lines);
	}
	return (WEXITSTATUS(status));
	}

	return (-1);
	}

	int
	libzfs_run_process(const char path, char argv[], int flags)
	{
	return (libzfs_run_process_impl(path, argv, NULL, flags, NULL, NULL));
	}

	/*
	* Run a command and store its stdout lines in an array of strings (lines[]).
	* lines[] is allocated and populated for you, and the number of lines is set in
	* lines_cnt. lines[] must be freed after use with libzfs_free_str_array().
	* All newlines (\n) in lines[] are terminated for convenience.
	*/
	int
	libzfs_run_process_get_stdout(const char path, char argv[], char *env[],
	char *lines[], int lines_cnt)
	{
	return (libzfs_run_process_impl(path, argv, env, 0, lines, lines_cnt));
	}

	/*
	* Same as libzfs_run_process_get_stdout(), but run without $PATH set. This
	* means that *path needs to be the full path to the executable.
	*/
	int
	libzfs_run_process_get_stdout_nopath(const char path, char argv[],
	char env[], char lines[], int lines_cnt)
	{
	return (libzfs_run_process_impl(path, argv, env, NO_DEFAULT_PATH,
	lines, lines_cnt));
	}

	/*
	* Free an array of strings. Free both the strings contained in the array and
	* the array itself.
	*/
	void
	libzfs_free_str_array(char **strs, int count)
	{
	while (--count >= 0)
	free(strs[count]);

	free(strs);
	}

	/*
	* Returns 1 if environment variable is set to "YES", "yes", "ON", "on", or
	* a non-zero number.
	*
	* Returns 0 otherwise.
	*/
	boolean_t
	libzfs_envvar_is_set(const char *envvar)
	{
	char *env = getenv(envvar);
	return (env && (strtoul(env, NULL, 0) > 0 \|\|
	(!strncasecmp(env, "YES", 3) && strnlen(env, 4) == 3) \|\|
	(!strncasecmp(env, "ON", 2) && strnlen(env, 3) == 2)));
	}

	libzfs_handle_t *
	libzfs_init(void)
	{
	libzfs_handle_t *hdl;
	int error;
	char *env;

	if ((error = libzfs_load_module()) != 0) {
	errno = error;
	return (NULL);
	}

	if ((hdl = calloc(1, sizeof (libzfs_handle_t))) == NULL) {
	return (NULL);
	}

	if (regcomp(&hdl->libzfs_urire, URI_REGEX, 0) != 0) {
	free(hdl);
	return (NULL);
	}

	if ((hdl->libzfs_fd = open(ZFS_DEV, O_RDWR\|O_EXCL\|O_CLOEXEC)) < 0) {
	free(hdl);
	return (NULL);
	}

	if (libzfs_core_init() != 0) {
	(void) close(hdl->libzfs_fd);
	free(hdl);
	return (NULL);
	}

	zfs_prop_init();
	zpool_prop_init();
	zpool_feature_init();
	vdev_prop_init();
	libzfs_mnttab_init(hdl);
	fletcher_4_init();

	if (getenv("ZFS_PROP_DEBUG") != NULL) {
	hdl->libzfs_prop_debug = B_TRUE;
	}
	if ((env = getenv("ZFS_SENDRECV_MAX_NVLIST")) != NULL) {
	if ((error = zfs_nicestrtonum(hdl, env,
	&hdl->libzfs_max_nvlist))) {
	errno = error;
	(void) close(hdl->libzfs_fd);
	free(hdl);
	return (NULL);
	}
	} else {
	hdl->libzfs_max_nvlist = (SPA_MAXBLOCKSIZE * 4);
	}

	/*
	* For testing, remove some settable properties and features
	*/
	if (libzfs_envvar_is_set("ZFS_SYSFS_PROP_SUPPORT_TEST")) {
	zprop_desc_t *proptbl;

	proptbl = zpool_prop_get_table();
	proptbl[ZPOOL_PROP_COMMENT].pd_zfs_mod_supported = B_FALSE;

	proptbl = zfs_prop_get_table();
	proptbl[ZFS_PROP_DNODESIZE].pd_zfs_mod_supported = B_FALSE;

	zfeature_info_t *ftbl = spa_feature_table;
	ftbl[SPA_FEATURE_LARGE_BLOCKS].fi_zfs_mod_supported = B_FALSE;
	}

	return (hdl);
	}

	void
	libzfs_fini(libzfs_handle_t *hdl)
	{
	(void) close(hdl->libzfs_fd);
	zpool_free_handles(hdl);
	namespace_clear(hdl);
	libzfs_mnttab_fini(hdl);
	libzfs_core_fini();
	regfree(&hdl->libzfs_urire);
	fletcher_4_fini();
	#if LIBFETCH_DYNAMIC
	if (hdl->libfetch != (void *)-1 && hdl->libfetch != NULL)
	(void) dlclose(hdl->libfetch);
	free(hdl->libfetch_load_error);
	#endif
	free(hdl);
	}

	libzfs_handle_t *
	zpool_get_handle(zpool_handle_t *zhp)
	{
	return (zhp->zpool_hdl);
	}

	libzfs_handle_t *
	zfs_get_handle(zfs_handle_t *zhp)
	{
	return (zhp->zfs_hdl);
	}

	zpool_handle_t *
	zfs_get_pool_handle(const zfs_handle_t *zhp)
	{
	return (zhp->zpool_hdl);
	}

	/*
	* Given a name, determine whether or not it's a valid path
	* (starts with '/' or "./"). If so, walk the mnttab trying
	* to match the device number. If not, treat the path as an
	* fs/vol/snap/bkmark name.
	*/
	zfs_handle_t *
	zfs_path_to_zhandle(libzfs_handle_t hdl, const char path, zfs_type_t argtype)
	{
	struct stat64 statbuf;
	struct extmnttab entry;

	if (path[0] != '/' && strncmp(path, "./", strlen("./")) != 0) {
	/*
	* It's not a valid path, assume it's a name of type 'argtype'.
	*/
	return (zfs_open(hdl, path, argtype));
	}

	if (getextmntent(path, &entry, &statbuf) != 0)
	return (NULL);

	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) {
	(void) fprintf(stderr, gettext("'%s': not a ZFS filesystem\n"),
	path);
	return (NULL);
	}

	return (zfs_open(hdl, entry.mnt_special, ZFS_TYPE_FILESYSTEM));
	}

	/*
	* Initialize the zc_nvlist_dst member to prepare for receiving an nvlist from
	* an ioctl().
	*/
	void
	zcmd_alloc_dst_nvlist(libzfs_handle_t hdl, zfs_cmd_t zc, size_t len)
	{
	if (len == 0)
	len = 256 * 1024;
	zc->zc_nvlist_dst_size = len;
	zc->zc_nvlist_dst =
	(uint64_t)(uintptr_t)zfs_alloc(hdl, zc->zc_nvlist_dst_size);
	}

	/*
	* Called when an ioctl() which returns an nvlist fails with ENOMEM. This will
	* expand the nvlist to the size specified in 'zc_nvlist_dst_size', which was
	* filled in by the kernel to indicate the actual required size.
	*/
	void
	zcmd_expand_dst_nvlist(libzfs_handle_t hdl, zfs_cmd_t zc)
	{
	free((void *)(uintptr_t)zc->zc_nvlist_dst);
	zc->zc_nvlist_dst =
	(uint64_t)(uintptr_t)zfs_alloc(hdl, zc->zc_nvlist_dst_size);
	}

	/*
	* Called to free the src and dst nvlists stored in the command structure.
	*/
	void
	zcmd_free_nvlists(zfs_cmd_t *zc)
	{
	free((void *)(uintptr_t)zc->zc_nvlist_conf);
	free((void *)(uintptr_t)zc->zc_nvlist_src);
	free((void *)(uintptr_t)zc->zc_nvlist_dst);
	zc->zc_nvlist_conf = 0;
	zc->zc_nvlist_src = 0;
	zc->zc_nvlist_dst = 0;
	}

	static void
	zcmd_write_nvlist_com(libzfs_handle_t hdl, uint64_t outnv, uint64_t *outlen,
	nvlist_t *nvl)
	{
	char *packed;

	size_t len = fnvlist_size(nvl);
	packed = zfs_alloc(hdl, len);

	verify(nvlist_pack(nvl, &packed, &len, NV_ENCODE_NATIVE, 0) == 0);

	*outnv = (uint64_t)(uintptr_t)packed;
	*outlen = len;
	}

	void
	zcmd_write_conf_nvlist(libzfs_handle_t hdl, zfs_cmd_t zc, nvlist_t *nvl)
	{
	zcmd_write_nvlist_com(hdl, &zc->zc_nvlist_conf,
	&zc->zc_nvlist_conf_size, nvl);
	}

	void
	zcmd_write_src_nvlist(libzfs_handle_t hdl, zfs_cmd_t zc, nvlist_t *nvl)
	{
	zcmd_write_nvlist_com(hdl, &zc->zc_nvlist_src,
	&zc->zc_nvlist_src_size, nvl);
	}

	/*
	* Unpacks an nvlist from the ZFS ioctl command structure.
	*/
	int
	zcmd_read_dst_nvlist(libzfs_handle_t hdl, zfs_cmd_t zc, nvlist_t **nvlp)
	{
	if (nvlist_unpack((void *)(uintptr_t)zc->zc_nvlist_dst,
	zc->zc_nvlist_dst_size, nvlp, 0) != 0)
	return (no_memory(hdl));

	return (0);
	}

	/*
	* ================================================================
	* API shared by zfs and zpool property management
	* ================================================================
	*/

	static void
	zprop_print_headers(zprop_get_cbdata_t *cbp, zfs_type_t type)
	{
	zprop_list_t *pl = cbp->cb_proplist;
	int i;
	char *title;
	size_t len;

	cbp->cb_first = B_FALSE;
	if (cbp->cb_scripted)
	return;

	/*
	* Start with the length of the column headers.
	*/
	cbp->cb_colwidths[GET_COL_NAME] = strlen(dgettext(TEXT_DOMAIN, "NAME"));
	cbp->cb_colwidths[GET_COL_PROPERTY] = strlen(dgettext(TEXT_DOMAIN,
	"PROPERTY"));
	cbp->cb_colwidths[GET_COL_VALUE] = strlen(dgettext(TEXT_DOMAIN,
	"VALUE"));
	cbp->cb_colwidths[GET_COL_RECVD] = strlen(dgettext(TEXT_DOMAIN,
	"RECEIVED"));
	cbp->cb_colwidths[GET_COL_SOURCE] = strlen(dgettext(TEXT_DOMAIN,
	"SOURCE"));

	/* first property is always NAME */
	assert(cbp->cb_proplist->pl_prop ==
	((type == ZFS_TYPE_POOL) ? ZPOOL_PROP_NAME :
	((type == ZFS_TYPE_VDEV) ? VDEV_PROP_NAME : ZFS_PROP_NAME)));

	/*
	* Go through and calculate the widths for each column. For the
	* 'source' column, we kludge it up by taking the worst-case scenario of
	* inheriting from the longest name. This is acceptable because in the
	* majority of cases 'SOURCE' is the last column displayed, and we don't
	* use the width anyway. Note that the 'VALUE' column can be oversized,
	* if the name of the property is much longer than any values we find.
	*/
	for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) {
	/*
	* 'PROPERTY' column
	*/
	if (pl->pl_prop != ZPROP_USERPROP) {
	const char *propname = (type == ZFS_TYPE_POOL) ?
	zpool_prop_to_name(pl->pl_prop) :
	((type == ZFS_TYPE_VDEV) ?
	vdev_prop_to_name(pl->pl_prop) :
	zfs_prop_to_name(pl->pl_prop));

	assert(propname != NULL);
	len = strlen(propname);
	if (len > cbp->cb_colwidths[GET_COL_PROPERTY])
	cbp->cb_colwidths[GET_COL_PROPERTY] = len;
	} else {
	assert(pl->pl_user_prop != NULL);
	len = strlen(pl->pl_user_prop);
	if (len > cbp->cb_colwidths[GET_COL_PROPERTY])
	cbp->cb_colwidths[GET_COL_PROPERTY] = len;
	}

	/*
	* 'VALUE' column. The first property is always the 'name'
	* property that was tacked on either by /sbin/zfs's
	* zfs_do_get() or when calling zprop_expand_list(), so we
	* ignore its width. If the user specified the name property
	* to display, then it will be later in the list in any case.
	*/
	if (pl != cbp->cb_proplist &&
	pl->pl_width > cbp->cb_colwidths[GET_COL_VALUE])
	cbp->cb_colwidths[GET_COL_VALUE] = pl->pl_width;

	/* 'RECEIVED' column. */
	if (pl != cbp->cb_proplist &&
	pl->pl_recvd_width > cbp->cb_colwidths[GET_COL_RECVD])
	cbp->cb_colwidths[GET_COL_RECVD] = pl->pl_recvd_width;

	/*
	* 'NAME' and 'SOURCE' columns
	*/
	if (pl->pl_prop == ((type == ZFS_TYPE_POOL) ? ZPOOL_PROP_NAME :
	((type == ZFS_TYPE_VDEV) ? VDEV_PROP_NAME :
	ZFS_PROP_NAME)) && pl->pl_width >
	cbp->cb_colwidths[GET_COL_NAME]) {
	cbp->cb_colwidths[GET_COL_NAME] = pl->pl_width;
	cbp->cb_colwidths[GET_COL_SOURCE] = pl->pl_width +
	strlen(dgettext(TEXT_DOMAIN, "inherited from"));
	}
	}

	/*
	* Now go through and print the headers.
	*/
	for (i = 0; i < ZFS_GET_NCOLS; i++) {
	switch (cbp->cb_columns[i]) {
	case GET_COL_NAME:
	title = dgettext(TEXT_DOMAIN, "NAME");
	break;
	case GET_COL_PROPERTY:
	title = dgettext(TEXT_DOMAIN, "PROPERTY");
	break;
	case GET_COL_VALUE:
	title = dgettext(TEXT_DOMAIN, "VALUE");
	break;
	case GET_COL_RECVD:
	title = dgettext(TEXT_DOMAIN, "RECEIVED");
	break;
	case GET_COL_SOURCE:
	title = dgettext(TEXT_DOMAIN, "SOURCE");
	break;
	default:
	title = NULL;
	}

	if (title != NULL) {
	if (i == (ZFS_GET_NCOLS - 1) \|\|
	cbp->cb_columns[i + 1] == GET_COL_NONE)
	(void) printf("%s", title);
	else
	(void) printf("%-*s ",
	cbp->cb_colwidths[cbp->cb_columns[i]],
	title);
	}
	}
	(void) printf("\n");
	}

	/*
	* Display a single line of output, according to the settings in the callback
	* structure.
	*/
	void
	zprop_print_one_property(const char name, zprop_get_cbdata_t cbp,
	const char propname, const char value, zprop_source_t sourcetype,
	const char source, const char recvd_value)
	{
	int i;
	const char *str = NULL;
	char buf[128];

	/*
	* Ignore those source types that the user has chosen to ignore.
	*/
	if ((sourcetype & cbp->cb_sources) == 0)
	return;

	if (cbp->cb_first)
	zprop_print_headers(cbp, cbp->cb_type);

	for (i = 0; i < ZFS_GET_NCOLS; i++) {
	switch (cbp->cb_columns[i]) {
	case GET_COL_NAME:
	str = name;
	break;

	case GET_COL_PROPERTY:
	str = propname;
	break;

	case GET_COL_VALUE:
	str = value;
	break;

	case GET_COL_SOURCE:
	switch (sourcetype) {
	case ZPROP_SRC_NONE:
	str = "-";
	break;

	case ZPROP_SRC_DEFAULT:
	str = "default";
	break;

	case ZPROP_SRC_LOCAL:
	str = "local";
	break;

	case ZPROP_SRC_TEMPORARY:
	str = "temporary";
	break;

	case ZPROP_SRC_INHERITED:
	(void) snprintf(buf, sizeof (buf),
	"inherited from %s", source);
	str = buf;
	break;
	case ZPROP_SRC_RECEIVED:
	str = "received";
	break;

	default:
	str = NULL;
	assert(!"unhandled zprop_source_t");
	}
	break;

	case GET_COL_RECVD:
	str = (recvd_value == NULL ? "-" : recvd_value);
	break;

	default:
	continue;
	}

	if (i == (ZFS_GET_NCOLS - 1) \|\|
	cbp->cb_columns[i + 1] == GET_COL_NONE)
	(void) printf("%s", str);
	else if (cbp->cb_scripted)
	(void) printf("%s\t", str);
	else
	(void) printf("%-*s ",
	cbp->cb_colwidths[cbp->cb_columns[i]],
	str);
	}

	(void) printf("\n");
	}

	/*
	* Given a numeric suffix, convert the value into a number of bits that the
	* resulting value must be shifted.
	*/
	static int
	str2shift(libzfs_handle_t hdl, const char buf)
	{
	const char *ends = "BKMGTPEZ";
	int i;

	if (buf[0] == '\0')
	return (0);
	for (i = 0; i < strlen(ends); i++) {
	if (toupper(buf[0]) == ends[i])
	break;
	}
	if (i == strlen(ends)) {
	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid numeric suffix '%s'"), buf);
	return (-1);
	}

	/*
	* Allow 'G' = 'GB' = 'GiB', case-insensitively.
	* However, 'BB' and 'BiB' are disallowed.
	*/
	if (buf[1] == '\0' \|\|
	(toupper(buf[0]) != 'B' &&
	((toupper(buf[1]) == 'B' && buf[2] == '\0') \|\|
	(toupper(buf[1]) == 'I' && toupper(buf[2]) == 'B' &&
	buf[3] == '\0'))))
	return (10 * i);

	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid numeric suffix '%s'"), buf);
	return (-1);
	}

	/*
	* Convert a string of the form '100G' into a real number. Used when setting
	* properties or creating a volume. 'buf' is used to place an extended error
	* message for the caller to use.
	*/
	int
	zfs_nicestrtonum(libzfs_handle_t hdl, const char value, uint64_t *num)
	{
	char *end;
	int shift;

	*num = 0;

	/* Check to see if this looks like a number. */
	if ((value[0] < '0' \|\| value[0] > '9') && value[0] != '.') {
	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"bad numeric value '%s'"), value);
	return (-1);
	}

	/* Rely on strtoull() to process the numeric portion. */
	errno = 0;
	*num = strtoull(value, &end, 10);

	/*
	* Check for ERANGE, which indicates that the value is too large to fit
	* in a 64-bit value.
	*/
	if (errno == ERANGE) {
	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"numeric value is too large"));
	return (-1);
	}

	/*
	* If we have a decimal value, then do the computation with floating
	* point arithmetic. Otherwise, use standard arithmetic.
	*/
	if (*end == '.') {
	double fval = strtod(value, &end);

	if ((shift = str2shift(hdl, end)) == -1)
	return (-1);

	fval *= pow(2, shift);

	/*
	* UINT64_MAX is not exactly representable as a double.
	* The closest representation is UINT64_MAX + 1, so we
	* use a >= comparison instead of > for the bounds check.
	*/
	if (fval >= (double)UINT64_MAX) {
	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"numeric value is too large"));
	return (-1);
	}

	*num = (uint64_t)fval;
	} else {
	if ((shift = str2shift(hdl, end)) == -1)
	return (-1);

	/* Check for overflow */
	if (shift >= 64 \|\| (num << shift) >> shift != num) {
	if (hdl)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"numeric value is too large"));
	return (-1);
	}

	*num <<= shift;
	}

	return (0);
	}

	/*
	* Given a propname=value nvpair to set, parse any numeric properties
	* (index, boolean, etc) if they are specified as strings and add the
	* resulting nvpair to the returned nvlist.
	*
	* At the DSL layer, all properties are either 64-bit numbers or strings.
	* We want the user to be able to ignore this fact and specify properties
	* as native values (numbers, for example) or as strings (to simplify
	* command line utilities). This also handles converting index types
	* (compression, checksum, etc) from strings to their on-disk index.
	*/
	int
	zprop_parse_value(libzfs_handle_t hdl, nvpair_t elem, int prop,
	zfs_type_t type, nvlist_t ret, char svalp, uint64_t ivalp,
	const char *errbuf)
	{
	data_type_t datatype = nvpair_type(elem);
	zprop_type_t proptype;
	const char *propname;
	char *value;
	boolean_t isnone = B_FALSE;
	boolean_t isauto = B_FALSE;
	int err = 0;

	if (type == ZFS_TYPE_POOL) {
	proptype = zpool_prop_get_type(prop);
	propname = zpool_prop_to_name(prop);
	} else if (type == ZFS_TYPE_VDEV) {
	proptype = vdev_prop_get_type(prop);
	propname = vdev_prop_to_name(prop);
	} else {
	proptype = zfs_prop_get_type(prop);
	propname = zfs_prop_to_name(prop);
	}

	/*
	* Convert any properties to the internal DSL value types.
	*/
	*svalp = NULL;
	*ivalp = 0;

	switch (proptype) {
	case PROP_TYPE_STRING:
	if (datatype != DATA_TYPE_STRING) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a string"), nvpair_name(elem));
	goto error;
	}
	err = nvpair_value_string(elem, svalp);
	if (err != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is invalid"), nvpair_name(elem));
	goto error;
	}
	if (strlen(*svalp) >= ZFS_MAXPROPLEN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is too long"), nvpair_name(elem));
	goto error;
	}
	break;

	case PROP_TYPE_NUMBER:
	if (datatype == DATA_TYPE_STRING) {
	(void) nvpair_value_string(elem, &value);
	if (strcmp(value, "none") == 0) {
	isnone = B_TRUE;
	} else if (strcmp(value, "auto") == 0) {
	isauto = B_TRUE;
	} else if (zfs_nicestrtonum(hdl, value, ivalp) != 0) {
	goto error;
	}
	} else if (datatype == DATA_TYPE_UINT64) {
	(void) nvpair_value_uint64(elem, ivalp);
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a number"), nvpair_name(elem));
	goto error;
	}

	/*
	* Quota special: force 'none' and don't allow 0.
	*/
	if ((type & ZFS_TYPE_DATASET) && *ivalp == 0 && !isnone &&
	(prop == ZFS_PROP_QUOTA \|\| prop == ZFS_PROP_REFQUOTA)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"use 'none' to disable quota/refquota"));
	goto error;
	}

	/*
	* Special handling for "*_limit=none". In this case it's not
	* 0 but UINT64_MAX.
	*/
	if ((type & ZFS_TYPE_DATASET) && isnone &&
	(prop == ZFS_PROP_FILESYSTEM_LIMIT \|\|
	prop == ZFS_PROP_SNAPSHOT_LIMIT)) {
	*ivalp = UINT64_MAX;
	}

	/*
	* Special handling for setting 'refreservation' to 'auto'. Use
	* UINT64_MAX to tell the caller to use zfs_fix_auto_resv().
	* 'auto' is only allowed on volumes.
	*/
	if (isauto) {
	switch (prop) {
	case ZFS_PROP_REFRESERVATION:
	if ((type & ZFS_TYPE_VOLUME) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s=auto' only allowed on "
	"volumes"), nvpair_name(elem));
	goto error;
	}
	*ivalp = UINT64_MAX;
	break;
	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'auto' is invalid value for '%s'"),
	nvpair_name(elem));
	goto error;
	}
	}

	break;

	case PROP_TYPE_INDEX:
	if (datatype != DATA_TYPE_STRING) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be a string"), nvpair_name(elem));
	goto error;
	}

	(void) nvpair_value_string(elem, &value);

	if (zprop_string_to_index(prop, value, ivalp, type) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' must be one of '%s'"), propname,
	zprop_values(prop, type));
	goto error;
	}
	break;

	default:
	abort();
	}

	/*
	* Add the result to our return set of properties.
	*/
	if (*svalp != NULL) {
	if (nvlist_add_string(ret, propname, *svalp) != 0) {
	(void) no_memory(hdl);
	return (-1);
	}
	} else {
	if (nvlist_add_uint64(ret, propname, *ivalp) != 0) {
	(void) no_memory(hdl);
	return (-1);
	}
	}

	return (0);
	error:
	(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
	return (-1);
	}

	static int
	addlist(libzfs_handle_t hdl, const char propname, zprop_list_t **listp,
	zfs_type_t type)
	{
	int prop = zprop_name_to_prop(propname, type);
	if (prop != ZPROP_INVAL && !zprop_valid_for_type(prop, type, B_FALSE))
	prop = ZPROP_INVAL;

	/*
	* Return failure if no property table entry was found and this isn't
	* a user-defined property.
	*/
	if (prop == ZPROP_USERPROP && ((type == ZFS_TYPE_POOL &&
	!zpool_prop_feature(propname) &&
	!zpool_prop_unsupported(propname)) \|\|
	((type == ZFS_TYPE_DATASET) && !zfs_prop_user(propname) &&
	!zfs_prop_userquota(propname) && !zfs_prop_written(propname)) \|\|
	((type == ZFS_TYPE_VDEV) && !vdev_prop_user(propname)))) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid property '%s'"), propname);
	return (zfs_error(hdl, EZFS_BADPROP,
	dgettext(TEXT_DOMAIN, "bad property list")));
	}

	zprop_list_t entry = zfs_alloc(hdl, sizeof (entry));

	entry->pl_prop = prop;
	if (prop == ZPROP_USERPROP) {
	entry->pl_user_prop = zfs_strdup(hdl, propname);
	entry->pl_width = strlen(propname);
	} else {
	entry->pl_width = zprop_width(prop, &entry->pl_fixed,
	type);
	}

	*listp = entry;

	return (0);
	}

	/*
	* Given a comma-separated list of properties, construct a property list
	* containing both user-defined and native properties. This function will
	* return a NULL list if 'all' is specified, which can later be expanded
	* by zprop_expand_list().
	*/
	int
	zprop_get_list(libzfs_handle_t hdl, char props, zprop_list_t **listp,
	zfs_type_t type)
	{
	*listp = NULL;

	/*
	* If 'all' is specified, return a NULL list.
	*/
	if (strcmp(props, "all") == 0)
	return (0);

	/*
	* If no props were specified, return an error.
	*/
	if (props[0] == '\0') {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no properties specified"));
	return (zfs_error(hdl, EZFS_BADPROP, dgettext(TEXT_DOMAIN,
	"bad property list")));
	}

	for (char *p; (p = strsep(&props, ",")); )
	if (strcmp(p, "space") == 0) {
	static const char *const spaceprops[] = {
	"name", "avail", "used", "usedbysnapshots",
	"usedbydataset", "usedbyrefreservation",
	"usedbychildren"
	};

	for (int i = 0; i < ARRAY_SIZE(spaceprops); i++) {
	if (addlist(hdl, spaceprops[i], listp, type))
	return (-1);
	listp = &(*listp)->pl_next;
	}
	} else {
	if (addlist(hdl, p, listp, type))
	return (-1);
	listp = &(*listp)->pl_next;
	}

	return (0);
	}

	void
	zprop_free_list(zprop_list_t *pl)
	{
	zprop_list_t *next;

	while (pl != NULL) {
	next = pl->pl_next;
	free(pl->pl_user_prop);
	free(pl);
	pl = next;
	}
	}

	typedef struct expand_data {
	zprop_list_t **last;
	libzfs_handle_t *hdl;
	zfs_type_t type;
	} expand_data_t;

	static int
	zprop_expand_list_cb(int prop, void *cb)
	{
	zprop_list_t *entry;
	expand_data_t *edp = cb;

	entry = zfs_alloc(edp->hdl, sizeof (zprop_list_t));

	entry->pl_prop = prop;
	entry->pl_width = zprop_width(prop, &entry->pl_fixed, edp->type);
	entry->pl_all = B_TRUE;

	*(edp->last) = entry;
	edp->last = &entry->pl_next;

	return (ZPROP_CONT);
	}

	int
	zprop_expand_list(libzfs_handle_t hdl, zprop_list_t *plp, zfs_type_t type)
	{
	zprop_list_t *entry;
	zprop_list_t **last;
	expand_data_t exp;

	if (*plp == NULL) {
	/*
	* If this is the very first time we've been called for an 'all'
	* specification, expand the list to include all native
	* properties.
	*/
	last = plp;

	exp.last = last;
	exp.hdl = hdl;
	exp.type = type;

	if (zprop_iter_common(zprop_expand_list_cb, &exp, B_FALSE,
	B_FALSE, type) == ZPROP_INVAL)
	return (-1);

	/*
	* Add 'name' to the beginning of the list, which is handled
	* specially.
	*/
	entry = zfs_alloc(hdl, sizeof (zprop_list_t));
	entry->pl_prop = ((type == ZFS_TYPE_POOL) ? ZPOOL_PROP_NAME :
	((type == ZFS_TYPE_VDEV) ? VDEV_PROP_NAME : ZFS_PROP_NAME));
	entry->pl_width = zprop_width(entry->pl_prop,
	&entry->pl_fixed, type);
	entry->pl_all = B_TRUE;
	entry->pl_next = *plp;
	*plp = entry;
	}
	return (0);
	}

	int
	zprop_iter(zprop_func func, void *cb, boolean_t show_all, boolean_t ordered,
	zfs_type_t type)
	{
	return (zprop_iter_common(func, cb, show_all, ordered, type));
	}

	const char *
	zfs_version_userland(void)
	{
	return (ZFS_META_ALIAS);
	}

	/*
	* Prints both zfs userland and kernel versions
	* Returns 0 on success, and -1 on error
	*/
	int
	zfs_version_print(void)
	{
	(void) puts(ZFS_META_ALIAS);

	char *kver = zfs_version_kernel();
	if (kver == NULL) {
	fprintf(stderr, "zfs_version_kernel() failed: %s\n",
	strerror(errno));
	return (-1);
	}

	(void) printf("zfs-kmod-%s\n", kver);
	free(kver);
	return (0);
	}

	/*
	* Return 1 if the user requested ANSI color output, and our terminal supports
	* it. Return 0 for no color.
	*/
	static int
	use_color(void)
	{
	static int use_color = -1;
	char *term;

	/*
	* Optimization:
	*
	* For each zpool invocation, we do a single check to see if we should
	* be using color or not, and cache that value for the lifetime of the
	* the zpool command. That makes it cheap to call use_color() when
	* we're printing with color. We assume that the settings are not going
	* to change during the invocation of a zpool command (the user isn't
	* going to change the ZFS_COLOR value while zpool is running, for
	* example).
	*/
	if (use_color != -1) {
	/*
	* We've already figured out if we should be using color or
	* not. Return the cached value.
	*/
	return (use_color);
	}

	term = getenv("TERM");
	/*
	* The user sets the ZFS_COLOR env var set to enable zpool ANSI color
	* output. However if NO_COLOR is set (https://no-color.org/) then
	* don't use it. Also, don't use color if terminal doesn't support
	* it.
	*/
	if (libzfs_envvar_is_set("ZFS_COLOR") &&
	!libzfs_envvar_is_set("NO_COLOR") &&
	isatty(STDOUT_FILENO) && term && strcmp("dumb", term) != 0 &&
	strcmp("unknown", term) != 0) {
	/* Color supported */
	use_color = 1;
	} else {
	use_color = 0;
	}

	return (use_color);
	}

	/*
	* color_start() and color_end() are used for when you want to colorize a block
	* of text. For example:
	*
	* color_start(ANSI_RED_FG)
	* printf("hello");
	* printf("world");
	* color_end();
	*/
	void
	color_start(const char *color)
	{
	if (use_color())
	fputs(color, stdout);
	}

	void
	color_end(void)
	{
	if (use_color())
	fputs(ANSI_RESET, stdout);
	}

	/* printf() with a color. If color is NULL, then do a normal printf. */
	int
	printf_color(const char color, const char format, ...)
	{
	va_list aptr;
	int rc;

	if (color)
	color_start(color);

	va_start(aptr, format);
	rc = vprintf(format, aptr);
	va_end(aptr);

	if (color)
	color_end();

	return (rc);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_util_os.c b/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_util_os.c
	index 4a44aa1f1d0d..99faae66833e 100644
	--- a/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_util_os.c
	+++ b/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_util_os.c
	@@ -1,280 +1,280 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2021 Klara, Inc.
	*/

	#include <alloca.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <libintl.h>
	#include <math.h>
	#include <poll.h>
	#include <stdarg.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <strings.h>
	#include <sys/inotify.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/stat.h>
	#include <sys/timerfd.h>
	#include <sys/types.h>
	#include <sys/wait.h>
	#include <unistd.h>

	#include <libzfs.h>
	#include <libzfs_core.h>

	#include "../../libzfs_impl.h"
	#include "zfs_prop.h"
	#include <libzutil.h>
	#include <sys/zfs_sysfs.h>

	#define ZDIFF_SHARESDIR "/.zfs/shares/"

	int
	zfs_ioctl(libzfs_handle_t hdl, int request, zfs_cmd_t zc)
	{
	return (ioctl(hdl->libzfs_fd, request, zc));
	}

	const char *
	libzfs_error_init(int error)
	{
	switch (error) {
	case ENXIO:
	return (dgettext(TEXT_DOMAIN, "The ZFS modules are not "
	"loaded.\nTry running 'modprobe zfs' as root "
	"to load them."));
	case ENOENT:
	return (dgettext(TEXT_DOMAIN, "/dev/zfs and /proc/self/mounts "
	"are required.\nTry running 'udevadm trigger' and 'mount "
	"-t proc proc /proc' as root."));
	case ENOEXEC:
	return (dgettext(TEXT_DOMAIN, "The ZFS modules cannot be "
	"auto-loaded.\nTry running 'modprobe zfs' as "
	"root to manually load them."));
	case EACCES:
	return (dgettext(TEXT_DOMAIN, "Permission denied the "
	"ZFS utilities must be run as root."));
	default:
	return (dgettext(TEXT_DOMAIN, "Failed to initialize the "
	"libzfs library."));
	}
	}

	/*
	* zfs(4) is loaded by udev if there's a fstype=zfs device present,
	* but if there isn't, load them automatically;
	* always wait for ZFS_DEV to appear via udev.
	*
	* Environment variables:
	* - ZFS_MODULE_TIMEOUT="<seconds>" - Seconds to wait for ZFS_DEV,
	* defaults to 10, max. 10 min.
	*/
	int
	libzfs_load_module(void)
	{
	if (access(ZFS_DEV, F_OK) == 0)
	return (0);

	if (access(ZFS_SYSFS_DIR, F_OK) != 0) {
	char argv[] = {(char )"modprobe", (char *)"zfs", NULL};
	if (libzfs_run_process("modprobe", argv, 0))
	return (ENOEXEC);

	if (access(ZFS_SYSFS_DIR, F_OK) != 0)
	return (ENXIO);
	}

	const char *timeout_str = getenv("ZFS_MODULE_TIMEOUT");
	int seconds = 10;
	if (timeout_str)
	seconds = MIN(strtol(timeout_str, NULL, 0), 600);
	struct itimerspec timeout = {.it_value.tv_sec = MAX(seconds, 0)};

	int ino = inotify_init1(IN_CLOEXEC);
	if (ino == -1)
	return (ENOENT);
	inotify_add_watch(ino, ZFS_DEVDIR, IN_CREATE);

	if (access(ZFS_DEV, F_OK) == 0) {
	close(ino);
	return (0);
	} else if (seconds == 0) {
	close(ino);
	return (ENOENT);
	}

	size_t evsz = sizeof (struct inotify_event) + NAME_MAX + 1;
	struct inotify_event *ev = alloca(evsz);

	int tout = timerfd_create(CLOCK_MONOTONIC, TFD_CLOEXEC);
	if (tout == -1) {
	close(ino);
	return (ENOENT);
	}
	timerfd_settime(tout, 0, &timeout, NULL);

	int ret = ENOENT;
	struct pollfd pfds[] = {
	{.fd = ino, .events = POLLIN},
	{.fd = tout, .events = POLLIN},
	};
	while (poll(pfds, ARRAY_SIZE(pfds), -1) != -1) {
	if (pfds[0].revents & POLLIN) {
	verify(read(ino, ev, evsz) >
	sizeof (struct inotify_event));
	- if (strcmp(ev->name, &ZFS_DEV[sizeof (ZFS_DEVDIR)])
	- == 0) {
	+ if (strncmp(ev->name, &ZFS_DEV[sizeof (ZFS_DEVDIR)],
	+ ev->len) == 0) {
	ret = 0;
	break;
	}
	}
	if (pfds[1].revents & POLLIN)
	break;
	}
	close(tout);
	close(ino);
	return (ret);
	}

	int
	find_shares_object(differ_info_t *di)
	{
	char fullpath[MAXPATHLEN];
	struct stat64 sb = { 0 };

	(void) strlcpy(fullpath, di->dsmnt, MAXPATHLEN);
	(void) strlcat(fullpath, ZDIFF_SHARESDIR, MAXPATHLEN);

	if (stat64(fullpath, &sb) != 0) {
	(void) snprintf(di->errbuf, sizeof (di->errbuf),
	dgettext(TEXT_DOMAIN, "Cannot stat %s"), fullpath);
	return (zfs_error(di->zhp->zfs_hdl, EZFS_DIFF, di->errbuf));
	}

	di->shares = (uint64_t)sb.st_ino;
	return (0);
	}

	int
	zfs_destroy_snaps_nvl_os(libzfs_handle_t hdl, nvlist_t snaps)
	{
	(void) hdl, (void) snaps;
	return (0);
	}

	/*
	* Return allocated loaded module version, or NULL on error (with errno set)
	*/
	char *
	zfs_version_kernel(void)
	{
	FILE *f = fopen(ZFS_SYSFS_DIR "/version", "re");
	if (f == NULL)
	return (NULL);

	char *ret = NULL;
	size_t l;
	ssize_t read;
	if ((read = getline(&ret, &l, f)) == -1) {
	int err = errno;
	fclose(f);
	errno = err;
	return (NULL);
	}

	fclose(f);
	if (ret[read - 1] == '\n')
	ret[read - 1] = '\0';
	return (ret);
	}

	/*
	* Add or delete the given filesystem to/from the given user namespace.
	*/
	int
	zfs_userns(zfs_handle_t zhp, const char nspath, int attach)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	zfs_cmd_t zc = {"\0"};
	char errbuf[1024];
	unsigned long cmd;
	int ret;

	if (attach) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot add '%s' to namespace"),
	zhp->zfs_name);
	} else {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot remove '%s' from namespace"),
	zhp->zfs_name);
	}

	switch (zhp->zfs_type) {
	case ZFS_TYPE_VOLUME:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"volumes can not be namespaced"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_SNAPSHOT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"snapshots can not be namespaced"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_BOOKMARK:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"bookmarks can not be namespaced"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_VDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"vdevs can not be namespaced"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_INVALID:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid zfs_type_t: ZFS_TYPE_INVALID"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_POOL:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pools can not be namespaced"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	case ZFS_TYPE_FILESYSTEM:
	break;
	}
	assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM);

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	zc.zc_objset_type = DMU_OST_ZFS;
	zc.zc_cleanup_fd = open(nspath, O_RDONLY);
	if (zc.zc_cleanup_fd < 0) {
	return (zfs_error(hdl, EZFS_NOT_USER_NAMESPACE, errbuf));
	}

	cmd = attach ? ZFS_IOC_USERNS_ATTACH : ZFS_IOC_USERNS_DETACH;
	if ((ret = zfs_ioctl(hdl, cmd, &zc)) != 0)
	zfs_standard_error(hdl, errno, errbuf);

	(void) close(zc.zc_cleanup_fd);

	return (ret);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfsbootenv/lzbe_device.c b/sys/contrib/openzfs/lib/libzfsbootenv/lzbe_device.c
	index 39e33324b315..21007fd9ef50 100644
	--- a/sys/contrib/openzfs/lib/libzfsbootenv/lzbe_device.c
	+++ b/sys/contrib/openzfs/lib/libzfsbootenv/lzbe_device.c
	@@ -1,163 +1,164 @@
	/*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*/
	/*
	* Copyright 2020 Toomas Soome <tsoome@me.com>
	*/

	#include <sys/types.h>
	#include <string.h>
	#include <libzfs.h>
	#include <libzfsbootenv.h>
	#include <sys/zfs_bootenv.h>
	#include <sys/vdev_impl.h>

	/*
	* Store device name to zpool label bootenv area.
	* This call will set bootenv version to VB_NVLIST, if bootenv currently
	* does contain other version, then old data will be replaced.
	*/
	int
	lzbe_set_boot_device(const char pool, lzbe_flags_t flag, const char device)
	{
	libzfs_handle_t *hdl;
	zpool_handle_t *zphdl;
	nvlist_t *nv;
	char *descriptor;
	uint64_t version;
	int rv = -1;

	if (pool == NULL \|\| *pool == '\0')
	return (rv);

	if ((hdl = libzfs_init()) == NULL)
	return (rv);

	zphdl = zpool_open(hdl, pool);
	if (zphdl == NULL) {
	libzfs_fini(hdl);
	return (rv);
	}

	switch (flag) {
	case lzbe_add:
	rv = zpool_get_bootenv(zphdl, &nv);
	if (rv == 0) {
	/*
	* We got the nvlist, check for version.
	* if version is missing or is not VB_NVLIST,
	* create new list.
	*/
	rv = nvlist_lookup_uint64(nv, BOOTENV_VERSION,
	&version);
	if (rv == 0 && version == VB_NVLIST)
	break;

	/* Drop this nvlist */
	fnvlist_free(nv);
	}
	zfs_fallthrough;
	case lzbe_replace:
	nv = fnvlist_alloc();
	break;
	default:
	return (rv);
	}

	/* version is mandatory */
	fnvlist_add_uint64(nv, BOOTENV_VERSION, VB_NVLIST);

	+ rv = 0;
	/*
	* If device name is empty, remove boot device configuration.
	*/
	if ((device == NULL \|\| *device == '\0')) {
	if (nvlist_exists(nv, OS_BOOTONCE))
	fnvlist_remove(nv, OS_BOOTONCE);
	} else {
	/*
	* Use device name directly if it does start with
	* prefix "zfs:". Otherwise, add prefix and suffix.
	*/
	if (strncmp(device, "zfs:", 4) == 0) {
	fnvlist_add_string(nv, OS_BOOTONCE, device);
	} else {
	if (asprintf(&descriptor, "zfs:%s:", device) > 0) {
	fnvlist_add_string(nv, OS_BOOTONCE, descriptor);
	free(descriptor);
	} else
	rv = ENOMEM;
	}
	}
	-
	- rv = zpool_set_bootenv(zphdl, nv);
	+ if (rv == 0)
	+ rv = zpool_set_bootenv(zphdl, nv);
	if (rv != 0)
	fprintf(stderr, "%s\n", libzfs_error_description(hdl));

	fnvlist_free(nv);
	zpool_close(zphdl);
	libzfs_fini(hdl);
	return (rv);
	}

	/*
	* Return boot device name from bootenv, if set.
	*/
	int
	lzbe_get_boot_device(const char pool, char *device)
	{
	libzfs_handle_t *hdl;
	zpool_handle_t *zphdl;
	nvlist_t *nv;
	char *val;
	int rv = -1;

	if (pool == NULL \|\| *pool == '\0' \|\| device == NULL)
	return (rv);

	if ((hdl = libzfs_init()) == NULL)
	return (rv);

	zphdl = zpool_open(hdl, pool);
	if (zphdl == NULL) {
	libzfs_fini(hdl);
	return (rv);
	}

	rv = zpool_get_bootenv(zphdl, &nv);
	if (rv == 0) {
	rv = nvlist_lookup_string(nv, OS_BOOTONCE, &val);
	if (rv == 0) {
	/*
	* zfs device descriptor is in form of "zfs:dataset:",
	* we only do need dataset name.
	*/
	if (strncmp(val, "zfs:", 4) == 0) {
	val += 4;
	val = strdup(val);
	if (val != NULL) {
	size_t len = strlen(val);

	if (val[len - 1] == ':')
	val[len - 1] = '\0';
	*device = val;
	} else {
	rv = ENOMEM;
	}
	} else {
	rv = EINVAL;
	}
	}
	nvlist_free(nv);
	}

	zpool_close(zphdl);
	libzfs_fini(hdl);
	return (rv);
	}
	diff --git a/sys/contrib/openzfs/lib/libzpool/taskq.c b/sys/contrib/openzfs/lib/libzpool/taskq.c
	index 8d6f1c93d8c9..a2e457ef9e60 100644
	--- a/sys/contrib/openzfs/lib/libzpool/taskq.c
	+++ b/sys/contrib/openzfs/lib/libzpool/taskq.c
	@@ -1,384 +1,386 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2010 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/
	/*
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright 2012 Garrett D'Amore <garrett@damore.org>. All rights reserved.
	* Copyright (c) 2014 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>

	int taskq_now;
	taskq_t *system_taskq;
	taskq_t *system_delay_taskq;

	static pthread_key_t taskq_tsd;

	#define TASKQ_ACTIVE 0x00010000

	static taskq_ent_t *
	task_alloc(taskq_t *tq, int tqflags)
	{
	taskq_ent_t *t;
	int rv;

	again: if ((t = tq->tq_freelist) != NULL && tq->tq_nalloc >= tq->tq_minalloc) {
	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
	tq->tq_freelist = t->tqent_next;
	} else {
	if (tq->tq_nalloc >= tq->tq_maxalloc) {
	if (!(tqflags & KM_SLEEP))
	return (NULL);

	/*
	* We don't want to exceed tq_maxalloc, but we can't
	* wait for other tasks to complete (and thus free up
	* task structures) without risking deadlock with
	* the caller. So, we just delay for one second
	* to throttle the allocation rate. If we have tasks
	* complete before one second timeout expires then
	* taskq_ent_free will signal us and we will
	* immediately retry the allocation.
	*/
	tq->tq_maxalloc_wait++;
	rv = cv_timedwait(&tq->tq_maxalloc_cv,
	&tq->tq_lock, ddi_get_lbolt() + hz);
	tq->tq_maxalloc_wait--;
	if (rv > 0)
	goto again; /* signaled */
	}
	mutex_exit(&tq->tq_lock);

	t = kmem_alloc(sizeof (taskq_ent_t), tqflags);

	mutex_enter(&tq->tq_lock);
	if (t != NULL) {
	/* Make sure we start without any flags */
	t->tqent_flags = 0;
	tq->tq_nalloc++;
	}
	}
	return (t);
	}

	static void
	task_free(taskq_t tq, taskq_ent_t t)
	{
	if (tq->tq_nalloc <= tq->tq_minalloc) {
	t->tqent_next = tq->tq_freelist;
	tq->tq_freelist = t;
	} else {
	tq->tq_nalloc--;
	mutex_exit(&tq->tq_lock);
	kmem_free(t, sizeof (taskq_ent_t));
	mutex_enter(&tq->tq_lock);
	}

	if (tq->tq_maxalloc_wait)
	cv_signal(&tq->tq_maxalloc_cv);
	}

	taskqid_t
	taskq_dispatch(taskq_t tq, task_func_t func, void arg, uint_t tqflags)
	{
	taskq_ent_t *t;

	if (taskq_now) {
	func(arg);
	return (1);
	}

	mutex_enter(&tq->tq_lock);
	ASSERT(tq->tq_flags & TASKQ_ACTIVE);
	if ((t = task_alloc(tq, tqflags)) == NULL) {
	mutex_exit(&tq->tq_lock);
	return (0);
	}
	if (tqflags & TQ_FRONT) {
	t->tqent_next = tq->tq_task.tqent_next;
	t->tqent_prev = &tq->tq_task;
	} else {
	t->tqent_next = &tq->tq_task;
	t->tqent_prev = tq->tq_task.tqent_prev;
	}
	t->tqent_next->tqent_prev = t;
	t->tqent_prev->tqent_next = t;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_flags = 0;
	cv_signal(&tq->tq_dispatch_cv);
	mutex_exit(&tq->tq_lock);
	return (1);
	}

	taskqid_t
	taskq_dispatch_delay(taskq_t tq, task_func_t func, void arg, uint_t tqflags,
	clock_t expire_time)
	{
	(void) tq, (void) func, (void) arg, (void) tqflags, (void) expire_time;
	return (0);
	}

	int
	taskq_empty_ent(taskq_ent_t *t)
	{
	return (t->tqent_next == NULL);
	}

	void
	taskq_init_ent(taskq_ent_t *t)
	{
	t->tqent_next = NULL;
	t->tqent_prev = NULL;
	t->tqent_func = NULL;
	t->tqent_arg = NULL;
	t->tqent_flags = 0;
	}

	void
	taskq_dispatch_ent(taskq_t tq, task_func_t func, void arg, uint_t flags,
	taskq_ent_t *t)
	{
	ASSERT(func != NULL);

	/*
	* Mark it as a prealloc'd task. This is important
	* to ensure that we don't free it later.
	*/
	t->tqent_flags \|= TQENT_FLAG_PREALLOC;
	/*
	* Enqueue the task to the underlying queue.
	*/
	mutex_enter(&tq->tq_lock);

	if (flags & TQ_FRONT) {
	t->tqent_next = tq->tq_task.tqent_next;
	t->tqent_prev = &tq->tq_task;
	} else {
	t->tqent_next = &tq->tq_task;
	t->tqent_prev = tq->tq_task.tqent_prev;
	}
	t->tqent_next->tqent_prev = t;
	t->tqent_prev->tqent_next = t;
	t->tqent_func = func;
	t->tqent_arg = arg;
	cv_signal(&tq->tq_dispatch_cv);
	mutex_exit(&tq->tq_lock);
	}

	void
	taskq_wait(taskq_t *tq)
	{
	mutex_enter(&tq->tq_lock);
	while (tq->tq_task.tqent_next != &tq->tq_task \|\| tq->tq_active != 0)
	cv_wait(&tq->tq_wait_cv, &tq->tq_lock);
	mutex_exit(&tq->tq_lock);
	}

	void
	taskq_wait_id(taskq_t *tq, taskqid_t id)
	{
	(void) id;
	taskq_wait(tq);
	}

	void
	taskq_wait_outstanding(taskq_t *tq, taskqid_t id)
	{
	(void) id;
	taskq_wait(tq);
	}

	static __attribute__((noreturn)) void
	taskq_thread(void *arg)
	{
	taskq_t *tq = arg;
	taskq_ent_t *t;
	boolean_t prealloc;

	VERIFY0(pthread_setspecific(taskq_tsd, tq));

	mutex_enter(&tq->tq_lock);
	while (tq->tq_flags & TASKQ_ACTIVE) {
	if ((t = tq->tq_task.tqent_next) == &tq->tq_task) {
	if (--tq->tq_active == 0)
	cv_broadcast(&tq->tq_wait_cv);
	cv_wait(&tq->tq_dispatch_cv, &tq->tq_lock);
	tq->tq_active++;
	continue;
	}
	t->tqent_prev->tqent_next = t->tqent_next;
	t->tqent_next->tqent_prev = t->tqent_prev;
	t->tqent_next = NULL;
	t->tqent_prev = NULL;
	prealloc = t->tqent_flags & TQENT_FLAG_PREALLOC;
	mutex_exit(&tq->tq_lock);

	rw_enter(&tq->tq_threadlock, RW_READER);
	t->tqent_func(t->tqent_arg);
	rw_exit(&tq->tq_threadlock);

	mutex_enter(&tq->tq_lock);
	if (!prealloc)
	task_free(tq, t);
	}
	tq->tq_nthreads--;
	cv_broadcast(&tq->tq_wait_cv);
	mutex_exit(&tq->tq_lock);
	thread_exit();
	}

	taskq_t *
	taskq_create(const char *name, int nthreads, pri_t pri,
	int minalloc, int maxalloc, uint_t flags)
	{
	(void) pri;
	taskq_t *tq = kmem_zalloc(sizeof (taskq_t), KM_SLEEP);
	int t;

	if (flags & TASKQ_THREADS_CPU_PCT) {
	int pct;
	ASSERT3S(nthreads, >=, 0);
	ASSERT3S(nthreads, <=, 100);
	pct = MIN(nthreads, 100);
	pct = MAX(pct, 0);

	nthreads = (sysconf(_SC_NPROCESSORS_ONLN) * pct) / 100;
	nthreads = MAX(nthreads, 1); /* need at least 1 thread */
	} else {
	ASSERT3S(nthreads, >=, 1);
	}

	rw_init(&tq->tq_threadlock, NULL, RW_DEFAULT, NULL);
	mutex_init(&tq->tq_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&tq->tq_dispatch_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tq->tq_wait_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tq->tq_maxalloc_cv, NULL, CV_DEFAULT, NULL);
	- (void) strncpy(tq->tq_name, name, TASKQ_NAMELEN);
	+ (void) strlcpy(tq->tq_name, name, sizeof (tq->tq_name));
	tq->tq_flags = flags \| TASKQ_ACTIVE;
	tq->tq_active = nthreads;
	tq->tq_nthreads = nthreads;
	tq->tq_minalloc = minalloc;
	tq->tq_maxalloc = maxalloc;
	tq->tq_task.tqent_next = &tq->tq_task;
	tq->tq_task.tqent_prev = &tq->tq_task;
	tq->tq_threadlist = kmem_alloc(nthreads * sizeof (kthread_t *),
	KM_SLEEP);

	if (flags & TASKQ_PREPOPULATE) {
	mutex_enter(&tq->tq_lock);
	while (minalloc-- > 0)
	task_free(tq, task_alloc(tq, KM_SLEEP));
	mutex_exit(&tq->tq_lock);
	}

	for (t = 0; t < nthreads; t++)
	VERIFY((tq->tq_threadlist[t] = thread_create(NULL, 0,
	taskq_thread, tq, 0, &p0, TS_RUN, pri)) != NULL);

	return (tq);
	}

	void
	taskq_destroy(taskq_t *tq)
	{
	int nthreads = tq->tq_nthreads;

	taskq_wait(tq);

	mutex_enter(&tq->tq_lock);

	tq->tq_flags &= ~TASKQ_ACTIVE;
	cv_broadcast(&tq->tq_dispatch_cv);

	while (tq->tq_nthreads != 0)
	cv_wait(&tq->tq_wait_cv, &tq->tq_lock);

	tq->tq_minalloc = 0;
	while (tq->tq_nalloc != 0) {
	ASSERT(tq->tq_freelist != NULL);
	- task_free(tq, task_alloc(tq, KM_SLEEP));
	+ taskq_ent_t *tqent_nexttq = tq->tq_freelist->tqent_next;
	+ task_free(tq, tq->tq_freelist);
	+ tq->tq_freelist = tqent_nexttq;
	}

	mutex_exit(&tq->tq_lock);

	kmem_free(tq->tq_threadlist, nthreads * sizeof (kthread_t *));

	rw_destroy(&tq->tq_threadlock);
	mutex_destroy(&tq->tq_lock);
	cv_destroy(&tq->tq_dispatch_cv);
	cv_destroy(&tq->tq_wait_cv);
	cv_destroy(&tq->tq_maxalloc_cv);

	kmem_free(tq, sizeof (taskq_t));
	}

	int
	taskq_member(taskq_t tq, kthread_t t)
	{
	int i;

	if (taskq_now)
	return (1);

	for (i = 0; i < tq->tq_nthreads; i++)
	if (tq->tq_threadlist[i] == t)
	return (1);

	return (0);
	}

	taskq_t *
	taskq_of_curthread(void)
	{
	return (pthread_getspecific(taskq_tsd));
	}

	int
	taskq_cancel_id(taskq_t *tq, taskqid_t id)
	{
	(void) tq, (void) id;
	return (ENOENT);
	}

	void
	system_taskq_init(void)
	{
	VERIFY0(pthread_key_create(&taskq_tsd, NULL));
	system_taskq = taskq_create("system_taskq", 64, maxclsyspri, 4, 512,
	TASKQ_DYNAMIC \| TASKQ_PREPOPULATE);
	system_delay_taskq = taskq_create("delay_taskq", 4, maxclsyspri, 4,
	512, TASKQ_DYNAMIC \| TASKQ_PREPOPULATE);
	}

	void
	system_taskq_fini(void)
	{
	taskq_destroy(system_taskq);
	system_taskq = NULL; /* defensive */
	taskq_destroy(system_delay_taskq);
	system_delay_taskq = NULL;
	VERIFY0(pthread_key_delete(taskq_tsd));
	}
	diff --git a/sys/contrib/openzfs/lib/libzpool/util.c b/sys/contrib/openzfs/lib/libzpool/util.c
	index 833e51398b1c..0ce7822a3563 100644
	--- a/sys/contrib/openzfs/lib/libzpool/util.c
	+++ b/sys/contrib/openzfs/lib/libzpool/util.c
	@@ -1,355 +1,356 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2016 by Delphix. All rights reserved.
	* Copyright 2017 Jason King
	* Copyright (c) 2017, Intel Corporation.
	*/

	#include <assert.h>
	#include <sys/zfs_context.h>
	#include <sys/avl.h>
	#include <string.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <sys/spa.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_refcount.h>
	#include <sys/zfs_ioctl.h>
	#include <dlfcn.h>
	#include <libzutil.h>

	/*
	* Routines needed by more than one client of libzpool.
	*/

	static void
	show_vdev_stats(const char desc, const char ctype, nvlist_t *nv, int indent)
	{
	vdev_stat_t *vs;
	vdev_stat_t *v0 = { 0 };
	uint64_t sec;
	uint64_t is_log = 0;
	nvlist_t **child;
	uint_t c, children;
	char used[6], avail[6];
	char rops[6], wops[6], rbytes[6], wbytes[6], rerr[6], werr[6], cerr[6];

	v0 = umem_zalloc(sizeof (*v0), UMEM_NOFAIL);

	if (indent == 0 && desc != NULL) {
	(void) printf(" "
	" capacity operations bandwidth ---- errors ----\n");
	(void) printf("description "
	"used avail read write read write read write cksum\n");
	}

	if (desc != NULL) {
	const char *suffix = "";
	char *bias = NULL;
	char bias_suffix[32];

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log);
	(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS,
	&bias);
	if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) != 0)
	vs = v0;

	if (bias != NULL) {
	(void) snprintf(bias_suffix, sizeof (bias_suffix),
	" (%s)", bias);
	suffix = bias_suffix;
	} else if (is_log) {
	suffix = " (log)";
	}

	sec = MAX(1, vs->vs_timestamp / NANOSEC);

	nicenum(vs->vs_alloc, used, sizeof (used));
	nicenum(vs->vs_space - vs->vs_alloc, avail, sizeof (avail));
	nicenum(vs->vs_ops[ZIO_TYPE_READ] / sec, rops, sizeof (rops));
	nicenum(vs->vs_ops[ZIO_TYPE_WRITE] / sec, wops, sizeof (wops));
	nicenum(vs->vs_bytes[ZIO_TYPE_READ] / sec, rbytes,
	sizeof (rbytes));
	nicenum(vs->vs_bytes[ZIO_TYPE_WRITE] / sec, wbytes,
	sizeof (wbytes));
	nicenum(vs->vs_read_errors, rerr, sizeof (rerr));
	nicenum(vs->vs_write_errors, werr, sizeof (werr));
	nicenum(vs->vs_checksum_errors, cerr, sizeof (cerr));

	(void) printf("%s%s%s%s%s %5s %5s %5s %5s %5s %5s %5s\n",
	indent, "",
	desc,
	(int)(indent+strlen(desc)-25-(vs->vs_space ? 0 : 12)),
	suffix,
	vs->vs_space ? 6 : 0, vs->vs_space ? used : "",
	vs->vs_space ? 6 : 0, vs->vs_space ? avail : "",
	rops, wops, rbytes, wbytes, rerr, werr, cerr);
	}
	umem_free(v0, sizeof (*v0));

	if (nvlist_lookup_nvlist_array(nv, ctype, &child, &children) != 0)
	return;

	for (c = 0; c < children; c++) {
	nvlist_t *cnv = child[c];
	char cname = NULL, tname;
	uint64_t np;
	int len;
	if (nvlist_lookup_string(cnv, ZPOOL_CONFIG_PATH, &cname) &&
	nvlist_lookup_string(cnv, ZPOOL_CONFIG_TYPE, &cname))
	cname = (char *)"<unknown>";
	len = strlen(cname) + 2;
	tname = umem_zalloc(len, UMEM_NOFAIL);
	(void) strlcpy(tname, cname, len);
	if (nvlist_lookup_uint64(cnv, ZPOOL_CONFIG_NPARITY, &np) == 0)
	tname[strlen(tname)] = '0' + np;
	show_vdev_stats(tname, ctype, cnv, indent + 2);
	umem_free(tname, len);
	}
	}

	void
	show_pool_stats(spa_t *spa)
	{
	nvlist_t config, nvroot;
	char *name;

	VERIFY(spa_get_stats(spa_name(spa), &config, NULL, 0) == 0);

	VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&name) == 0);

	show_vdev_stats(name, ZPOOL_CONFIG_CHILDREN, nvroot, 0);
	show_vdev_stats(NULL, ZPOOL_CONFIG_L2CACHE, nvroot, 0);
	show_vdev_stats(NULL, ZPOOL_CONFIG_SPARES, nvroot, 0);

	nvlist_free(config);
	}

	/* k_out must be freed by the caller /
	static int
	set_global_var_parse_kv(const char arg, char k_out, u_longlong_t v_out)
	{
	int err;
	VERIFY(arg);
	char *d = strdup(arg);

	char *save = NULL;
	char *k = strtok_r(d, "=", &save);
	char *v_str = strtok_r(NULL, "=", &save);
	char *follow = strtok_r(NULL, "=", &save);
	if (k == NULL \|\| v_str == NULL \|\| follow != NULL) {
	err = EINVAL;
	goto err_free;
	}

	u_longlong_t val = strtoull(v_str, NULL, 0);
	if (val > UINT32_MAX) {
	fprintf(stderr, "Value for global variable '%s' must "
	"be a 32-bit unsigned integer, got '%s'\n", k, v_str);
	err = EOVERFLOW;
	goto err_free;
	}

	- *k_out = k;
	+ *k_out = strdup(k);
	*v_out = val;
	+ free(d);
	return (0);

	err_free:
	- free(k);
	+ free(d);

	return (err);
	}

	/*
	* Sets given global variable in libzpool to given unsigned 32-bit value.
	* arg: "<variable>=<value>"
	*/
	int
	set_global_var(char const *arg)
	{
	void *zpoolhdl;
	char *varname;
	u_longlong_t val;
	int ret;

	#ifndef _ZFS_LITTLE_ENDIAN
	/*
	* On big endian systems changing a 64-bit variable would set the high
	* 32 bits instead of the low 32 bits, which could cause unexpected
	* results.
	*/
	fprintf(stderr, "Setting global variables is only supported on "
	"little-endian systems\n");
	ret = ENOTSUP;
	goto out_ret;
	#endif

	if ((ret = set_global_var_parse_kv(arg, &varname, &val)) != 0) {
	goto out_ret;
	}

	zpoolhdl = dlopen("libzpool.so", RTLD_LAZY);
	if (zpoolhdl != NULL) {
	uint32_t *var;
	var = dlsym(zpoolhdl, varname);
	if (var == NULL) {
	fprintf(stderr, "Global variable '%s' does not exist "
	"in libzpool.so\n", varname);
	ret = EINVAL;
	goto out_dlclose;
	}
	*var = (uint32_t)val;

	} else {
	fprintf(stderr, "Failed to open libzpool.so to set global "
	"variable\n");
	ret = EIO;
	goto out_dlclose;
	}

	ret = 0;

	out_dlclose:
	dlclose(zpoolhdl);
	free(varname);
	out_ret:
	return (ret);
	}

	static nvlist_t *
	refresh_config(void unused, nvlist_t tryconfig)
	{
	(void) unused;
	return (spa_tryimport(tryconfig));
	}

	#if defined(__FreeBSD__)

	#include <sys/param.h>
	#include <sys/sysctl.h>
	#include <os/freebsd/zfs/sys/zfs_ioctl_compat.h>

	static int
	pool_active(void unused, const char name, uint64_t guid, boolean_t *isactive)
	{
	(void) unused, (void) guid;
	zfs_iocparm_t zp;
	zfs_cmd_t *zc = NULL;
	zfs_cmd_legacy_t *zcl = NULL;
	unsigned long request;
	int ret;

	int fd = open(ZFS_DEV, O_RDWR \| O_CLOEXEC);
	if (fd < 0)
	return (-1);

	/*
	* Use ZFS_IOC_POOL_STATS to check if the pool is active. We want to
	* avoid adding a dependency on libzfs_core solely for this ioctl(),
	* therefore we manually craft the stats command. Note that the command
	* ID is identical between the openzfs and legacy ioctl() formats.
	*/
	int ver = ZFS_IOCVER_NONE;
	size_t ver_size = sizeof (ver);

	sysctlbyname("vfs.zfs.version.ioctl", &ver, &ver_size, NULL, 0);

	switch (ver) {
	case ZFS_IOCVER_OZFS:
	zc = umem_zalloc(sizeof (zfs_cmd_t), UMEM_NOFAIL);

	(void) strlcpy(zc->zc_name, name, sizeof (zc->zc_name));
	zp.zfs_cmd = (uint64_t)(uintptr_t)zc;
	zp.zfs_cmd_size = sizeof (zfs_cmd_t);
	zp.zfs_ioctl_version = ZFS_IOCVER_OZFS;

	request = _IOWR('Z', ZFS_IOC_POOL_STATS, zfs_iocparm_t);
	ret = ioctl(fd, request, &zp);

	free((void *)(uintptr_t)zc->zc_nvlist_dst);
	umem_free(zc, sizeof (zfs_cmd_t));

	break;
	case ZFS_IOCVER_LEGACY:
	zcl = umem_zalloc(sizeof (zfs_cmd_legacy_t), UMEM_NOFAIL);

	(void) strlcpy(zcl->zc_name, name, sizeof (zcl->zc_name));
	zp.zfs_cmd = (uint64_t)(uintptr_t)zcl;
	zp.zfs_cmd_size = sizeof (zfs_cmd_legacy_t);
	zp.zfs_ioctl_version = ZFS_IOCVER_LEGACY;

	request = _IOWR('Z', ZFS_IOC_POOL_STATS, zfs_iocparm_t);
	ret = ioctl(fd, request, &zp);

	free((void *)(uintptr_t)zcl->zc_nvlist_dst);
	umem_free(zcl, sizeof (zfs_cmd_legacy_t));

	break;
	default:
	fprintf(stderr, "unrecognized zfs ioctl version %d", ver);
	exit(1);
	}

	(void) close(fd);

	*isactive = (ret == 0);

	return (0);
	}
	#else
	static int
	pool_active(void unused, const char name, uint64_t guid,
	boolean_t *isactive)
	{
	(void) unused, (void) guid;
	int fd = open(ZFS_DEV, O_RDWR \| O_CLOEXEC);
	if (fd < 0)
	return (-1);

	/*
	* Use ZFS_IOC_POOL_STATS to check if a pool is active.
	*/
	zfs_cmd_t *zcp = umem_zalloc(sizeof (zfs_cmd_t), UMEM_NOFAIL);
	(void) strlcpy(zcp->zc_name, name, sizeof (zcp->zc_name));

	int ret = ioctl(fd, ZFS_IOC_POOL_STATS, zcp);

	free((void *)(uintptr_t)zcp->zc_nvlist_dst);
	umem_free(zcp, sizeof (zfs_cmd_t));

	(void) close(fd);

	*isactive = (ret == 0);

	return (0);
	}
	#endif

	-const pool_config_ops_t libzpool_config_ops = {
	+pool_config_ops_t libzpool_config_ops = {
	.pco_refresh_config = refresh_config,
	.pco_pool_active = pool_active,
	};
	diff --git a/sys/contrib/openzfs/lib/libzutil/os/linux/zutil_device_path_os.c b/sys/contrib/openzfs/lib/libzutil/os/linux/zutil_device_path_os.c
	index 9f4c74f50f3c..05dbb39954fa 100644
	--- a/sys/contrib/openzfs/lib/libzutil/os/linux/zutil_device_path_os.c
	+++ b/sys/contrib/openzfs/lib/libzutil/os/linux/zutil_device_path_os.c
	@@ -1,691 +1,692 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	*/

	#include <ctype.h>
	#include <dirent.h>
	#include <fcntl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <sys/efi_partition.h>

	#ifdef HAVE_LIBUDEV
	#include <libudev.h>
	#endif

	#include <libzutil.h>

	/*
	* Append partition suffix to an otherwise fully qualified device path.
	* This is used to generate the name the full path as its stored in
	* ZPOOL_CONFIG_PATH for whole disk devices. On success the new length
	* of 'path' will be returned on error a negative value is returned.
	*/
	int
	zfs_append_partition(char *path, size_t max_len)
	{
	int len = strlen(path);

	if ((strncmp(path, UDISK_ROOT, strlen(UDISK_ROOT)) == 0) \|\|
	(strncmp(path, ZVOL_ROOT, strlen(ZVOL_ROOT)) == 0)) {
	if (len + 6 >= max_len)
	return (-1);

	(void) strcat(path, "-part1");
	len += 6;
	} else {
	if (len + 2 >= max_len)
	return (-1);

	if (isdigit(path[len-1])) {
	(void) strcat(path, "p1");
	len += 2;
	} else {
	(void) strcat(path, "1");
	len += 1;
	}
	}

	return (len);
	}

	/*
	* Remove partition suffix from a vdev path. Partition suffixes may take three
	* forms: "-partX", "pX", or "X", where X is a string of digits. The second
	* case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
	* third case only occurs when preceded by a string matching the regular
	* expression "^([hsv]\|xv)d[a-z]+", i.e. a scsi, ide, virtio or xen disk.
	*
	* caller must free the returned string
	*/
	char *
	zfs_strip_partition(const char *path)
	{
	char *tmp = strdup(path);
	char part = NULL, d = NULL;
	if (!tmp)
	return (NULL);

	if ((part = strstr(tmp, "-part")) && part != tmp) {
	d = part + 5;
	} else if ((part = strrchr(tmp, 'p')) &&
	part > tmp + 1 && isdigit(*(part-1))) {
	d = part + 1;
	} else if ((tmp[0] == 'h' \|\| tmp[0] == 's' \|\| tmp[0] == 'v') &&
	tmp[1] == 'd') {
	for (d = &tmp[2]; isalpha(*d); part = ++d) { }
	} else if (strncmp("xvd", tmp, 3) == 0) {
	for (d = &tmp[3]; isalpha(*d); part = ++d) { }
	}
	if (part && d && *d != '\0') {
	for (; isdigit(*d); d++) { }
	if (*d == '\0')
	*part = '\0';
	}

	return (tmp);
	}

	/*
	* Same as zfs_strip_partition, but allows "/dev/" to be in the pathname
	*
	* path: /dev/sda1
	* returns: /dev/sda
	*
	* Returned string must be freed.
	*/
	static char *
	zfs_strip_partition_path(const char *path)
	{
	char *newpath = strdup(path);
	char *sd_offset;
	char *new_sd;

	if (!newpath)
	return (NULL);

	/* Point to "sda1" part of "/dev/sda1" */
	sd_offset = strrchr(newpath, '/') + 1;

	/* Get our new name "sda" */
	new_sd = zfs_strip_partition(sd_offset);
	if (!new_sd) {
	free(newpath);
	return (NULL);
	}

	/* Paste the "sda" where "sda1" was */
	strlcpy(sd_offset, new_sd, strlen(sd_offset) + 1);

	/* Free temporary "sda" */
	free(new_sd);

	return (newpath);
	}

	/*
	* Strip the unwanted portion of a device path.
	*/
	const char *
	zfs_strip_path(const char *path)
	{
	size_t spath_count;
	const char const spaths = zpool_default_search_paths(&spath_count);

	for (size_t i = 0; i < spath_count; ++i)
	if (strncmp(path, spaths[i], strlen(spaths[i])) == 0 &&
	path[strlen(spaths[i])] == '/')
	return (path + strlen(spaths[i]) + 1);

	return (path);
	}

	/*
	* Read the contents of a sysfs file into an allocated buffer and remove the
	* last newline.
	*
	* This is useful for reading sysfs files that return a single string. Return
	* an allocated string pointer on success, NULL otherwise. Returned buffer
	* must be freed by the user.
	*/
	static char *
	zfs_read_sysfs_file(char *filepath)
	{
	char buf[4096]; /* all sysfs files report 4k size */
	char *str = NULL;

	FILE *fp = fopen(filepath, "r");
	if (fp == NULL) {
	return (NULL);
	}
	if (fgets(buf, sizeof (buf), fp) == buf) {
	/* success */

	/* Remove the last newline (if any) */
	size_t len = strlen(buf);
	if (buf[len - 1] == '\n') {
	buf[len - 1] = '\0';
	}
	str = strdup(buf);
	}

	fclose(fp);

	return (str);
	}

	/*
	* Given a dev name like "nvme0n1", return the full PCI slot sysfs path to
	* the drive (in /sys/bus/pci/slots).
	*
	* For example:
	* dev: "nvme0n1"
	* returns: "/sys/bus/pci/slots/0"
	*
	* 'dev' must be an NVMe device.
	*
	* Returned string must be freed. Returns NULL on error or no sysfs path.
	*/
	static char *
	zfs_get_pci_slots_sys_path(const char *dev_name)
	{
	DIR *dp = NULL;
	struct dirent *ep;
	char *address1 = NULL;
	char *address2 = NULL;
	char *path = NULL;
	char buf[MAXPATHLEN];
	char *tmp;

	/* If they preface 'dev' with a path (like "/dev") then strip it off */
	tmp = strrchr(dev_name, '/');
	if (tmp != NULL)
	dev_name = tmp + 1; /* +1 since we want the chr after '/' */

	if (strncmp("nvme", dev_name, 4) != 0)
	return (NULL);

	(void) snprintf(buf, sizeof (buf), "/sys/block/%s/device/address",
	dev_name);

	address1 = zfs_read_sysfs_file(buf);
	if (!address1)
	return (NULL);

	/*
	* /sys/block/nvme0n1/device/address format will
	* be "0000:01:00.0" while /sys/bus/pci/slots/0/address will be
	* "0000:01:00". Just NULL terminate at the '.' so they match.
	*/
	tmp = strrchr(address1, '.');
	if (tmp != NULL)
	*tmp = '\0';

	dp = opendir("/sys/bus/pci/slots/");
	if (dp == NULL) {
	free(address1);
	return (NULL);
	}

	/*
	* Look through all the /sys/bus/pci/slots/ subdirs
	*/
	while ((ep = readdir(dp))) {
	/*
	* We only care about directory names that are a single number.
	* Sometimes there's other directories like
	* "/sys/bus/pci/slots/0-3/" in there - skip those.
	*/
	if (!zfs_isnumber(ep->d_name))
	continue;

	(void) snprintf(buf, sizeof (buf),
	"/sys/bus/pci/slots/%s/address", ep->d_name);

	address2 = zfs_read_sysfs_file(buf);
	if (!address2)
	continue;

	if (strcmp(address1, address2) == 0) {
	/* Addresses match, we're all done */
	free(address2);
	if (asprintf(&path, "/sys/bus/pci/slots/%s",
	ep->d_name) == -1) {
	continue;
	}
	break;
	}
	free(address2);
	}

	closedir(dp);
	free(address1);

	return (path);
	}

	/*
	* Given a dev name like "sda", return the full enclosure sysfs path to
	* the disk. You can also pass in the name with "/dev" prepended
	* to it (like /dev/sda). This works for both JBODs and NVMe PCI devices.
	*
	* For example, disk "sda" in enclosure slot 1:
	* dev_name: "sda"
	* returns: "/sys/class/enclosure/1:0:3:0/Slot 1"
	*
	* Or:
	*
	* dev_name: "nvme0n1"
	* returns: "/sys/bus/pci/slots/0"
	*
	* 'dev' must be a non-devicemapper device.
	*
	* Returned string must be freed. Returns NULL on error.
	*/
	char *
	zfs_get_enclosure_sysfs_path(const char *dev_name)
	{
	DIR *dp = NULL;
	struct dirent *ep;
	char buf[MAXPATHLEN];
	char *tmp1 = NULL;
	char *tmp2 = NULL;
	char *tmp3 = NULL;
	char *path = NULL;
	size_t size;
	int tmpsize;

	if (dev_name == NULL)
	return (NULL);

	/* If they preface 'dev' with a path (like "/dev") then strip it off */
	tmp1 = strrchr(dev_name, '/');
	if (tmp1 != NULL)
	dev_name = tmp1 + 1; /* +1 since we want the chr after '/' */

	tmpsize = asprintf(&tmp1, "/sys/block/%s/device", dev_name);
	if (tmpsize == -1 \|\| tmp1 == NULL) {
	tmp1 = NULL;
	goto end;
	}

	dp = opendir(tmp1);
	if (dp == NULL)
	goto end;

	/*
	* Look though all sysfs entries in /sys/block/<dev>/device for
	* the enclosure symlink.
	*/
	while ((ep = readdir(dp))) {
	/* Ignore everything that's not our enclosure_device link */
	if (strstr(ep->d_name, "enclosure_device") == NULL)
	continue;

	+ if (tmp2 != NULL)
	+ free(tmp2);
	if (asprintf(&tmp2, "%s/%s", tmp1, ep->d_name) == -1) {
	tmp2 = NULL;
	break;
	}

	size = readlink(tmp2, buf, sizeof (buf));

	/* Did readlink fail or crop the link name? */
	if (size == -1 \|\| size >= sizeof (buf))
	break;

	/*
	* We got a valid link. readlink() doesn't terminate strings
	* so we have to do it.
	*/
	buf[size] = '\0';

	/*
	* Our link will look like:
	*
	* "../../../../port-11:1:2/..STUFF../enclosure/1:0:3:0/SLOT 1"
	*
	* We want to grab the "enclosure/1:0:3:0/SLOT 1" part
	*/
	tmp3 = strstr(buf, "enclosure");
	if (tmp3 == NULL)
	break;

	+ if (path != NULL)
	+ free(path);
	if (asprintf(&path, "/sys/class/%s", tmp3) == -1) {
	/* If asprintf() fails, 'path' is undefined */
	path = NULL;
	break;
	}
	-
	- if (path == NULL)
	- break;
	}

	end:
	free(tmp2);
	free(tmp1);

	if (dp != NULL)
	closedir(dp);

	if (!path) {
	/*
	* This particular disk isn't in a JBOD. It could be an NVMe
	* drive. If so, look up the NVMe device's path in
	* /sys/bus/pci/slots/. Within that directory is a 'attention'
	* file which controls the NVMe fault LED.
	*/
	path = zfs_get_pci_slots_sys_path(dev_name);
	}

	return (path);
	}

	/*
	* Allocate and return the underlying device name for a device mapper device.
	*
	* For example, dm_name = "/dev/dm-0" could return "/dev/sda". Symlinks to a
	* DM device (like /dev/disk/by-vdev/A0) are also allowed.
	*
	* If the DM device has multiple underlying devices (like with multipath
	* DM devices), then favor underlying devices that have a symlink back to their
	* back to their enclosure device in sysfs. This will be useful for the
	* zedlet scripts that toggle the fault LED.
	*
	* Returns an underlying device name, or NULL on error or no match. If dm_name
	* is not a DM device then return NULL.
	*
	* NOTE: The returned name string must be freed.
	*/
	static char *
	dm_get_underlying_path(const char *dm_name)
	{
	DIR *dp = NULL;
	struct dirent *ep;
	char *realp;
	char *tmp = NULL;
	char *path = NULL;
	char *dev_str;
	int size;
	char *first_path = NULL;
	char *enclosure_path;

	if (dm_name == NULL)
	return (NULL);

	/* dm name may be a symlink (like /dev/disk/by-vdev/A0) */
	realp = realpath(dm_name, NULL);
	if (realp == NULL)
	return (NULL);

	/*
	* If they preface 'dev' with a path (like "/dev") then strip it off.
	* We just want the 'dm-N' part.
	*/
	tmp = strrchr(realp, '/');
	if (tmp != NULL)
	dev_str = tmp + 1; /* +1 since we want the chr after '/' */
	else
	dev_str = tmp;

	if ((size = asprintf(&tmp, "/sys/block/%s/slaves/", dev_str)) == -1) {
	tmp = NULL;
	goto end;
	}

	dp = opendir(tmp);
	if (dp == NULL)
	goto end;

	/*
	* A device-mapper device can have multiple paths to it (multipath).
	* Favor paths that have a symlink back to their enclosure device.
	* We have to do this since some enclosures may only provide a symlink
	* back for one underlying path to a disk and not the other.
	*
	* If no paths have links back to their enclosure, then just return the
	* first path.
	*/
	while ((ep = readdir(dp))) {
	if (ep->d_type != DT_DIR) { /* skip "." and ".." dirs */
	if (!first_path)
	first_path = strdup(ep->d_name);

	enclosure_path =
	zfs_get_enclosure_sysfs_path(ep->d_name);

	if (!enclosure_path)
	continue;

	if ((size = asprintf(
	&path, "/dev/%s", ep->d_name)) == -1)
	path = NULL;
	free(enclosure_path);
	break;
	}
	}

	end:
	if (dp != NULL)
	closedir(dp);
	free(tmp);
	free(realp);

	if (!path && first_path) {
	/*
	* None of the underlying paths had a link back to their
	* enclosure devices. Throw up out hands and return the first
	* underlying path.
	*/
	if ((size = asprintf(&path, "/dev/%s", first_path)) == -1)
	path = NULL;
	}

	free(first_path);
	return (path);
	}

	/*
	* Return B_TRUE if device is a device mapper or multipath device.
	* Return B_FALSE if not.
	*/
	boolean_t
	zfs_dev_is_dm(const char *dev_name)
	{

	char *tmp;
	tmp = dm_get_underlying_path(dev_name);
	if (tmp == NULL)
	return (B_FALSE);

	free(tmp);
	return (B_TRUE);
	}

	/*
	* By "whole disk" we mean an entire physical disk (something we can
	* label, toggle the write cache on, etc.) as opposed to the full
	* capacity of a pseudo-device such as lofi or did. We act as if we
	* are labeling the disk, which should be a pretty good test of whether
	* it's a viable device or not. Returns B_TRUE if it is and B_FALSE if
	* it isn't.
	*/
	boolean_t
	zfs_dev_is_whole_disk(const char *dev_name)
	{
	struct dk_gpt *label = NULL;
	int fd;

	if ((fd = open(dev_name, O_RDONLY \| O_DIRECT \| O_CLOEXEC)) < 0)
	return (B_FALSE);

	if (efi_alloc_and_init(fd, EFI_NUMPAR, &label) != 0) {
	(void) close(fd);
	return (B_FALSE);
	}

	efi_free(label);
	(void) close(fd);

	return (B_TRUE);
	}

	/*
	* Lookup the underlying device for a device name
	*
	* Often you'll have a symlink to a device, a partition device,
	* or a multipath device, and want to look up the underlying device.
	* This function returns the underlying device name. If the device
	* name is already the underlying device, then just return the same
	* name. If the device is a DM device with multiple underlying devices
	* then return the first one.
	*
	* For example:
	*
	* 1. /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001 -> ../../sda
	* dev_name: /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001
	* returns: /dev/sda
	*
	* 2. /dev/mapper/mpatha (made up of /dev/sda and /dev/sdb)
	* dev_name: /dev/mapper/mpatha
	* returns: /dev/sda (first device)
	*
	* 3. /dev/sda (already the underlying device)
	* dev_name: /dev/sda
	* returns: /dev/sda
	*
	* 4. /dev/dm-3 (mapped to /dev/sda)
	* dev_name: /dev/dm-3
	* returns: /dev/sda
	*
	* 5. /dev/disk/by-id/scsi-0QEMU_drive-scsi0-0-0-0-part9 -> ../../sdb9
	* dev_name: /dev/disk/by-id/scsi-0QEMU_drive-scsi0-0-0-0-part9
	* returns: /dev/sdb
	*
	* 6. /dev/disk/by-uuid/5df030cf-3cd9-46e4-8e99-3ccb462a4e9a -> ../dev/sda2
	* dev_name: /dev/disk/by-uuid/5df030cf-3cd9-46e4-8e99-3ccb462a4e9a
	* returns: /dev/sda
	*
	* Returns underlying device name, or NULL on error or no match.
	*
	* NOTE: The returned name string must be freed.
	*/
	char *
	zfs_get_underlying_path(const char *dev_name)
	{
	char *name = NULL;
	char *tmp;

	if (dev_name == NULL)
	return (NULL);

	tmp = dm_get_underlying_path(dev_name);

	/* dev_name not a DM device, so just un-symlinkize it */
	if (tmp == NULL)
	tmp = realpath(dev_name, NULL);

	if (tmp != NULL) {
	name = zfs_strip_partition_path(tmp);
	free(tmp);
	}

	return (name);
	}


	#ifdef HAVE_LIBUDEV

	/*
	* A disk is considered a multipath whole disk when:
	* DEVNAME key value has "dm-"
	* DM_UUID key exists and starts with 'mpath-'
	* ID_PART_TABLE_TYPE key does not exist or is not gpt
	* ID_FS_LABEL key does not exist (disk isn't labeled)
	*/
	static boolean_t
	is_mpath_udev_sane(struct udev_device *dev)
	{
	const char devname, type, uuid, label;

	devname = udev_device_get_property_value(dev, "DEVNAME");
	type = udev_device_get_property_value(dev, "ID_PART_TABLE_TYPE");
	uuid = udev_device_get_property_value(dev, "DM_UUID");
	label = udev_device_get_property_value(dev, "ID_FS_LABEL");

	if ((devname != NULL && strncmp(devname, "/dev/dm-", 8) == 0) &&
	((type == NULL) \|\| (strcmp(type, "gpt") != 0)) &&
	((uuid != NULL) && (strncmp(uuid, "mpath-", 6) == 0)) &&
	(label == NULL)) {
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Check if a disk is a multipath "blank" disk:
	*
	* 1. The disk has udev values that suggest it's a multipath disk
	* 2. The disk is not currently labeled with a filesystem of any type
	* 3. There are no partitions on the disk
	*/
	boolean_t
	is_mpath_whole_disk(const char *path)
	{
	struct udev *udev;
	struct udev_device *dev = NULL;
	char nodepath[MAXPATHLEN];
	char *sysname;

	if (realpath(path, nodepath) == NULL)
	return (B_FALSE);
	sysname = strrchr(nodepath, '/') + 1;
	if (strncmp(sysname, "dm-", 3) != 0)
	return (B_FALSE);
	if ((udev = udev_new()) == NULL)
	return (B_FALSE);
	if ((dev = udev_device_new_from_subsystem_sysname(udev, "block",
	sysname)) == NULL) {
	udev_device_unref(dev);
	return (B_FALSE);
	}

	/* Sanity check some udev values */
	boolean_t is_sane = is_mpath_udev_sane(dev);
	udev_device_unref(dev);

	return (is_sane);
	}

	#else /* HAVE_LIBUDEV */

	boolean_t
	is_mpath_whole_disk(const char *path)
	{
	(void) path;
	return (B_FALSE);
	}

	#endif /* HAVE_LIBUDEV */
	diff --git a/sys/contrib/openzfs/lib/libzutil/zutil_import.c b/sys/contrib/openzfs/lib/libzutil/zutil_import.c
	index 252b0bac685c..e3f1c8942564 100644
	--- a/sys/contrib/openzfs/lib/libzutil/zutil_import.c
	+++ b/sys/contrib/openzfs/lib/libzutil/zutil_import.c
	@@ -1,1939 +1,1938 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright 2015 RackTop Systems.
	* Copyright (c) 2016, Intel Corporation.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	/*
	* Pool import support functions.
	*
	* Used by zpool, ztest, zdb, and zhack to locate importable configs. Since
	* these commands are expected to run in the global zone, we can assume
	* that the devices are all readable when called.
	*
	* To import a pool, we rely on reading the configuration information from the
	* ZFS label of each device. If we successfully read the label, then we
	* organize the configuration information in the following hierarchy:
	*
	* pool guid -> toplevel vdev guid -> label txg
	*
	* Duplicate entries matching this same tuple will be discarded. Once we have
	* examined every device, we pick the best label txg config for each toplevel
	* vdev. We then arrange these toplevel vdevs into a complete pool config, and
	* update any paths that have changed. Finally, we attempt to import the pool
	* using our derived config, and record the results.
	*/

	#ifdef HAVE_AIO_H
	#include <aio.h>
	#endif
	#include <ctype.h>
	#include <dirent.h>
	#include <errno.h>
	#include <libintl.h>
	#include <libgen.h>
	#include <stddef.h>
	#include <stdlib.h>
	#include <string.h>
	#include <sys/stat.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <sys/dktp/fdisk.h>
	#include <sys/vdev_impl.h>
	#include <sys/fs/zfs.h>

	#include <thread_pool.h>
	#include <libzutil.h>
	#include <libnvpair.h>

	#include "zutil_import.h"

	static __attribute__((format(printf, 2, 3))) void
	zutil_error_aux(libpc_handle_t hdl, const char fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	(void) vsnprintf(hdl->lpc_desc, sizeof (hdl->lpc_desc), fmt, ap);
	hdl->lpc_desc_active = B_TRUE;

	va_end(ap);
	}

	static void
	zutil_verror(libpc_handle_t hdl, const char error, const char *fmt,
	va_list ap)
	{
	char action[1024];

	(void) vsnprintf(action, sizeof (action), fmt, ap);

	if (hdl->lpc_desc_active)
	hdl->lpc_desc_active = B_FALSE;
	else
	hdl->lpc_desc[0] = '\0';

	if (hdl->lpc_printerr) {
	if (hdl->lpc_desc[0] != '\0')
	error = hdl->lpc_desc;

	(void) fprintf(stderr, "%s: %s\n", action, error);
	}
	}

	static __attribute__((format(printf, 3, 4))) int
	zutil_error_fmt(libpc_handle_t hdl, const char error, const char *fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);

	zutil_verror(hdl, error, fmt, ap);

	va_end(ap);

	return (-1);
	}

	static int
	zutil_error(libpc_handle_t hdl, const char error, const char *msg)
	{
	return (zutil_error_fmt(hdl, error, "%s", msg));
	}

	static int
	zutil_no_memory(libpc_handle_t *hdl)
	{
	zutil_error(hdl, EZFS_NOMEM, "internal error");
	exit(1);
	}

	void *
	zutil_alloc(libpc_handle_t *hdl, size_t size)
	{
	void *data;

	if ((data = calloc(1, size)) == NULL)
	(void) zutil_no_memory(hdl);

	return (data);
	}

	char *
	zutil_strdup(libpc_handle_t hdl, const char str)
	{
	char *ret;

	if ((ret = strdup(str)) == NULL)
	(void) zutil_no_memory(hdl);

	return (ret);
	}

	static char *
	zutil_strndup(libpc_handle_t hdl, const char str, size_t n)
	{
	char *ret;

	if ((ret = strndup(str, n)) == NULL)
	(void) zutil_no_memory(hdl);

	return (ret);
	}

	/*
	* Intermediate structures used to gather configuration information.
	*/
	typedef struct config_entry {
	uint64_t ce_txg;
	nvlist_t *ce_config;
	struct config_entry *ce_next;
	} config_entry_t;

	typedef struct vdev_entry {
	uint64_t ve_guid;
	config_entry_t *ve_configs;
	struct vdev_entry *ve_next;
	} vdev_entry_t;

	typedef struct pool_entry {
	uint64_t pe_guid;
	vdev_entry_t *pe_vdevs;
	struct pool_entry *pe_next;
	} pool_entry_t;

	typedef struct name_entry {
	char *ne_name;
	uint64_t ne_guid;
	uint64_t ne_order;
	uint64_t ne_num_labels;
	struct name_entry *ne_next;
	} name_entry_t;

	typedef struct pool_list {
	pool_entry_t *pools;
	name_entry_t *names;
	} pool_list_t;

	/*
	* Go through and fix up any path and/or devid information for the given vdev
	* configuration.
	*/
	static int
	fix_paths(libpc_handle_t hdl, nvlist_t nv, name_entry_t *names)
	{
	nvlist_t **child;
	uint_t c, children;
	uint64_t guid;
	name_entry_t ne, best;
	char *path;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	if (fix_paths(hdl, child[c], names) != 0)
	return (-1);
	return (0);
	}

	/*
	* This is a leaf (file or disk) vdev. In either case, go through
	* the name list and see if we find a matching guid. If so, replace
	* the path and see if we can calculate a new devid.
	*
	* There may be multiple names associated with a particular guid, in
	* which case we have overlapping partitions or multiple paths to the
	* same disk. In this case we prefer to use the path name which
	* matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
	* use the lowest order device which corresponds to the first match
	* while traversing the ZPOOL_IMPORT_PATH search path.
	*/
	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
	path = NULL;

	best = NULL;
	for (ne = names; ne != NULL; ne = ne->ne_next) {
	if (ne->ne_guid == guid) {
	if (path == NULL) {
	best = ne;
	break;
	}

	if ((strlen(path) == strlen(ne->ne_name)) &&
	strncmp(path, ne->ne_name, strlen(path)) == 0) {
	best = ne;
	break;
	}

	if (best == NULL) {
	best = ne;
	continue;
	}

	/* Prefer paths with move vdev labels. */
	if (ne->ne_num_labels > best->ne_num_labels) {
	best = ne;
	continue;
	}

	/* Prefer paths earlier in the search order. */
	if (ne->ne_num_labels == best->ne_num_labels &&
	ne->ne_order < best->ne_order) {
	best = ne;
	continue;
	}
	}
	}

	if (best == NULL)
	return (0);

	if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
	return (-1);

	update_vdev_config_dev_strs(nv);

	return (0);
	}

	/*
	* Add the given configuration to the list of known devices.
	*/
	static int
	add_config(libpc_handle_t hdl, pool_list_t pl, const char *path,
	int order, int num_labels, nvlist_t *config)
	{
	uint64_t pool_guid, vdev_guid, top_guid, txg, state;
	pool_entry_t *pe;
	vdev_entry_t *ve;
	config_entry_t *ce;
	name_entry_t *ne;

	/*
	* If this is a hot spare not currently in use or level 2 cache
	* device, add it to the list of names to translate, but don't do
	* anything else.
	*/
	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&state) == 0 &&
	(state == POOL_STATE_SPARE \|\| state == POOL_STATE_L2CACHE) &&
	nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
	if ((ne = zutil_alloc(hdl, sizeof (name_entry_t))) == NULL)
	return (-1);

	if ((ne->ne_name = zutil_strdup(hdl, path)) == NULL) {
	free(ne);
	return (-1);
	}
	ne->ne_guid = vdev_guid;
	ne->ne_order = order;
	ne->ne_num_labels = num_labels;
	ne->ne_next = pl->names;
	pl->names = ne;

	return (0);
	}

	/*
	* If we have a valid config but cannot read any of these fields, then
	* it means we have a half-initialized label. In vdev_label_init()
	* we write a label with txg == 0 so that we can identify the device
	* in case the user refers to the same disk later on. If we fail to
	* create the pool, we'll be left with a label in this state
	* which should not be considered part of a valid pool.
	*/
	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&pool_guid) != 0 \|\|
	nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
	&vdev_guid) != 0 \|\|
	nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
	&top_guid) != 0 \|\|
	nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
	&txg) != 0 \|\| txg == 0) {
	return (0);
	}

	/*
	* First, see if we know about this pool. If not, then add it to the
	* list of known pools.
	*/
	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
	if (pe->pe_guid == pool_guid)
	break;
	}

	if (pe == NULL) {
	if ((pe = zutil_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
	return (-1);
	}
	pe->pe_guid = pool_guid;
	pe->pe_next = pl->pools;
	pl->pools = pe;
	}

	/*
	* Second, see if we know about this toplevel vdev. Add it if its
	* missing.
	*/
	for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
	if (ve->ve_guid == top_guid)
	break;
	}

	if (ve == NULL) {
	if ((ve = zutil_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
	return (-1);
	}
	ve->ve_guid = top_guid;
	ve->ve_next = pe->pe_vdevs;
	pe->pe_vdevs = ve;
	}

	/*
	* Third, see if we have a config with a matching transaction group. If
	* so, then we do nothing. Otherwise, add it to the list of known
	* configs.
	*/
	for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
	if (ce->ce_txg == txg)
	break;
	}

	if (ce == NULL) {
	if ((ce = zutil_alloc(hdl, sizeof (config_entry_t))) == NULL) {
	return (-1);
	}
	ce->ce_txg = txg;
	ce->ce_config = fnvlist_dup(config);
	ce->ce_next = ve->ve_configs;
	ve->ve_configs = ce;
	}

	/*
	* At this point we've successfully added our config to the list of
	* known configs. The last thing to do is add the vdev guid -> path
	* mappings so that we can fix up the configuration as necessary before
	* doing the import.
	*/
	if ((ne = zutil_alloc(hdl, sizeof (name_entry_t))) == NULL)
	return (-1);

	if ((ne->ne_name = zutil_strdup(hdl, path)) == NULL) {
	free(ne);
	return (-1);
	}

	ne->ne_guid = vdev_guid;
	ne->ne_order = order;
	ne->ne_num_labels = num_labels;
	ne->ne_next = pl->names;
	pl->names = ne;

	return (0);
	}

	static int
	zutil_pool_active(libpc_handle_t hdl, const char name, uint64_t guid,
	boolean_t *isactive)
	{
	ASSERT(hdl->lpc_ops->pco_pool_active != NULL);

	int error = hdl->lpc_ops->pco_pool_active(hdl->lpc_lib_handle, name,
	guid, isactive);

	return (error);
	}

	static nvlist_t *
	zutil_refresh_config(libpc_handle_t hdl, nvlist_t tryconfig)
	{
	ASSERT(hdl->lpc_ops->pco_refresh_config != NULL);

	return (hdl->lpc_ops->pco_refresh_config(hdl->lpc_lib_handle,
	tryconfig));
	}

	/*
	* Determine if the vdev id is a hole in the namespace.
	*/
	static boolean_t
	vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
	{
	int c;

	for (c = 0; c < holes; c++) {

	/* Top-level is a hole */
	if (hole_array[c] == id)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Convert our list of pools into the definitive set of configurations. We
	* start by picking the best config for each toplevel vdev. Once that's done,
	* we assemble the toplevel vdevs into a full config for the pool. We make a
	* pass to fix up any incorrect paths, and then add it to the main list to
	* return to the user.
	*/
	static nvlist_t *
	get_configs(libpc_handle_t hdl, pool_list_t pl, boolean_t active_ok,
	nvlist_t *policy)
	{
	pool_entry_t *pe;
	vdev_entry_t *ve;
	config_entry_t *ce;
	nvlist_t ret = NULL, config = NULL, tmp = NULL, nvtop, *nvroot;
	nvlist_t spares, l2cache;
	uint_t i, nspares, nl2cache;
	boolean_t config_seen;
	uint64_t best_txg;
	char name, hostname = NULL;
	uint64_t guid;
	uint_t children = 0;
	nvlist_t **child = NULL;
	uint_t holes;
	uint64_t *hole_array, max_id;
	uint_t c;
	boolean_t isactive;
	uint64_t hostid;
	nvlist_t *nvl;
	boolean_t valid_top_config = B_FALSE;

	if (nvlist_alloc(&ret, 0, 0) != 0)
	goto nomem;

	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
	uint64_t id, max_txg = 0;

	if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
	goto nomem;
	config_seen = B_FALSE;

	/*
	* Iterate over all toplevel vdevs. Grab the pool configuration
	* from the first one we find, and then go through the rest and
	* add them as necessary to the 'vdevs' member of the config.
	*/
	for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {

	/*
	* Determine the best configuration for this vdev by
	* selecting the config with the latest transaction
	* group.
	*/
	best_txg = 0;
	for (ce = ve->ve_configs; ce != NULL;
	ce = ce->ce_next) {

	if (ce->ce_txg > best_txg) {
	tmp = ce->ce_config;
	best_txg = ce->ce_txg;
	}
	}

	/*
	* We rely on the fact that the max txg for the
	* pool will contain the most up-to-date information
	* about the valid top-levels in the vdev namespace.
	*/
	if (best_txg > max_txg) {
	(void) nvlist_remove(config,
	ZPOOL_CONFIG_VDEV_CHILDREN,
	DATA_TYPE_UINT64);
	(void) nvlist_remove(config,
	ZPOOL_CONFIG_HOLE_ARRAY,
	DATA_TYPE_UINT64_ARRAY);

	max_txg = best_txg;
	hole_array = NULL;
	holes = 0;
	max_id = 0;
	valid_top_config = B_FALSE;

	if (nvlist_lookup_uint64(tmp,
	ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
	verify(nvlist_add_uint64(config,
	ZPOOL_CONFIG_VDEV_CHILDREN,
	max_id) == 0);
	valid_top_config = B_TRUE;
	}

	if (nvlist_lookup_uint64_array(tmp,
	ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
	&holes) == 0) {
	verify(nvlist_add_uint64_array(config,
	ZPOOL_CONFIG_HOLE_ARRAY,
	hole_array, holes) == 0);
	}
	}

	if (!config_seen) {
	/*
	* Copy the relevant pieces of data to the pool
	* configuration:
	*
	* version
	* pool guid
	* name
	* comment (if available)
	* compatibility features (if available)
	* pool state
	* hostid (if available)
	* hostname (if available)
	*/
	uint64_t state, version;
	char *comment = NULL;
	char *compatibility = NULL;

	version = fnvlist_lookup_uint64(tmp,
	ZPOOL_CONFIG_VERSION);
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_VERSION, version);
	guid = fnvlist_lookup_uint64(tmp,
	ZPOOL_CONFIG_POOL_GUID);
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_POOL_GUID, guid);
	name = fnvlist_lookup_string(tmp,
	ZPOOL_CONFIG_POOL_NAME);
	fnvlist_add_string(config,
	ZPOOL_CONFIG_POOL_NAME, name);

	if (nvlist_lookup_string(tmp,
	ZPOOL_CONFIG_COMMENT, &comment) == 0)
	fnvlist_add_string(config,
	ZPOOL_CONFIG_COMMENT, comment);

	if (nvlist_lookup_string(tmp,
	ZPOOL_CONFIG_COMPATIBILITY,
	&compatibility) == 0)
	fnvlist_add_string(config,
	ZPOOL_CONFIG_COMPATIBILITY,
	compatibility);

	state = fnvlist_lookup_uint64(tmp,
	ZPOOL_CONFIG_POOL_STATE);
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_POOL_STATE, state);

	hostid = 0;
	if (nvlist_lookup_uint64(tmp,
	ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_HOSTID, hostid);
	hostname = fnvlist_lookup_string(tmp,
	ZPOOL_CONFIG_HOSTNAME);
	fnvlist_add_string(config,
	ZPOOL_CONFIG_HOSTNAME, hostname);
	}

	config_seen = B_TRUE;
	}

	/*
	* Add this top-level vdev to the child array.
	*/
	verify(nvlist_lookup_nvlist(tmp,
	ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
	verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
	&id) == 0);

	if (id >= children) {
	nvlist_t **newchild;

	newchild = zutil_alloc(hdl, (id + 1) *
	sizeof (nvlist_t *));
	if (newchild == NULL)
	goto nomem;

	for (c = 0; c < children; c++)
	newchild[c] = child[c];

	free(child);
	child = newchild;
	children = id + 1;
	}
	if (nvlist_dup(nvtop, &child[id], 0) != 0)
	goto nomem;

	}

	/*
	* If we have information about all the top-levels then
	* clean up the nvlist which we've constructed. This
	* means removing any extraneous devices that are
	* beyond the valid range or adding devices to the end
	* of our array which appear to be missing.
	*/
	if (valid_top_config) {
	if (max_id < children) {
	for (c = max_id; c < children; c++)
	nvlist_free(child[c]);
	children = max_id;
	} else if (max_id > children) {
	nvlist_t **newchild;

	newchild = zutil_alloc(hdl, (max_id) *
	sizeof (nvlist_t *));
	if (newchild == NULL)
	goto nomem;

	for (c = 0; c < children; c++)
	newchild[c] = child[c];

	free(child);
	child = newchild;
	children = max_id;
	}
	}

	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&guid) == 0);

	/*
	* The vdev namespace may contain holes as a result of
	* device removal. We must add them back into the vdev
	* tree before we process any missing devices.
	*/
	if (holes > 0) {
	ASSERT(valid_top_config);

	for (c = 0; c < children; c++) {
	nvlist_t *holey;

	if (child[c] != NULL \|\|
	!vdev_is_hole(hole_array, holes, c))
	continue;

	if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
	0) != 0)
	goto nomem;

	/*
	* Holes in the namespace are treated as
	* "hole" top-level vdevs and have a
	* special flag set on them.
	*/
	if (nvlist_add_string(holey,
	ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_HOLE) != 0 \|\|
	nvlist_add_uint64(holey,
	ZPOOL_CONFIG_ID, c) != 0 \|\|
	nvlist_add_uint64(holey,
	ZPOOL_CONFIG_GUID, 0ULL) != 0) {
	nvlist_free(holey);
	goto nomem;
	}
	child[c] = holey;
	}
	}

	/*
	* Look for any missing top-level vdevs. If this is the case,
	* create a faked up 'missing' vdev as a placeholder. We cannot
	* simply compress the child array, because the kernel performs
	* certain checks to make sure the vdev IDs match their location
	* in the configuration.
	*/
	for (c = 0; c < children; c++) {
	if (child[c] == NULL) {
	nvlist_t *missing;
	if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
	0) != 0)
	goto nomem;
	if (nvlist_add_string(missing,
	ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_MISSING) != 0 \|\|
	nvlist_add_uint64(missing,
	ZPOOL_CONFIG_ID, c) != 0 \|\|
	nvlist_add_uint64(missing,
	ZPOOL_CONFIG_GUID, 0ULL) != 0) {
	nvlist_free(missing);
	goto nomem;
	}
	child[c] = missing;
	}
	}

	/*
	* Put all of this pool's top-level vdevs into a root vdev.
	*/
	if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
	goto nomem;
	if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
	VDEV_TYPE_ROOT) != 0 \|\|
	nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 \|\|
	nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 \|\|
	nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, children) != 0) {
	nvlist_free(nvroot);
	goto nomem;
	}

	for (c = 0; c < children; c++)
	nvlist_free(child[c]);
	free(child);
	children = 0;
	child = NULL;

	/*
	* Go through and fix up any paths and/or devids based on our
	* known list of vdev GUID -> path mappings.
	*/
	if (fix_paths(hdl, nvroot, pl->names) != 0) {
	nvlist_free(nvroot);
	goto nomem;
	}

	/*
	* Add the root vdev to this pool's configuration.
	*/
	if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	nvroot) != 0) {
	nvlist_free(nvroot);
	goto nomem;
	}
	nvlist_free(nvroot);

	/*
	* zdb uses this path to report on active pools that were
	* imported or created using -R.
	*/
	if (active_ok)
	goto add_pool;

	/*
	* Determine if this pool is currently active, in which case we
	* can't actually import it.
	*/
	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&name) == 0);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&guid) == 0);

	if (zutil_pool_active(hdl, name, guid, &isactive) != 0)
	goto error;

	if (isactive) {
	nvlist_free(config);
	config = NULL;
	continue;
	}

	if (policy != NULL) {
	if (nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
	policy) != 0)
	goto nomem;
	}

	if ((nvl = zutil_refresh_config(hdl, config)) == NULL) {
	nvlist_free(config);
	config = NULL;
	continue;
	}

	nvlist_free(config);
	config = nvl;

	/*
	* Go through and update the paths for spares, now that we have
	* them.
	*/
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	for (i = 0; i < nspares; i++) {
	if (fix_paths(hdl, spares[i], pl->names) != 0)
	goto nomem;
	}
	}

	/*
	* Update the paths for l2cache devices.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	for (i = 0; i < nl2cache; i++) {
	if (fix_paths(hdl, l2cache[i], pl->names) != 0)
	goto nomem;
	}
	}

	/*
	* Restore the original information read from the actual label.
	*/
	(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
	DATA_TYPE_UINT64);
	(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
	DATA_TYPE_STRING);
	if (hostid != 0) {
	verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
	hostid) == 0);
	verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
	hostname) == 0);
	}

	add_pool:
	/*
	* Add this pool to the list of configs.
	*/
	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&name) == 0);

	if (nvlist_add_nvlist(ret, name, config) != 0)
	goto nomem;

	nvlist_free(config);
	config = NULL;
	}

	return (ret);

	nomem:
	(void) zutil_no_memory(hdl);
	error:
	nvlist_free(config);
	nvlist_free(ret);
	for (c = 0; c < children; c++)
	nvlist_free(child[c]);
	free(child);

	return (NULL);
	}

	/*
	* Return the offset of the given label.
	*/
	static uint64_t
	label_offset(uint64_t size, int l)
	{
	ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
	return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
	0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
	}

	/*
	* The same description applies as to zpool_read_label below,
	* except here we do it without aio, presumably because an aio call
	* errored out in a way we think not using it could circumvent.
	*/
	static int
	zpool_read_label_slow(int fd, nvlist_t *config, int num_labels)
	{
	struct stat64 statbuf;
	int l, count = 0;
	vdev_phys_t *label;
	nvlist_t *expected_config = NULL;
	uint64_t expected_guid = 0, size;
	int error;

	*config = NULL;

	if (fstat64_blk(fd, &statbuf) == -1)
	return (0);
	size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);

	error = posix_memalign((void *)&label, PAGESIZE, sizeof (label));
	if (error)
	return (-1);

	for (l = 0; l < VDEV_LABELS; l++) {
	uint64_t state, guid, txg;
	off_t offset = label_offset(size, l) + VDEV_SKIP_SIZE;

	if (pread64(fd, label, sizeof (vdev_phys_t),
	offset) != sizeof (vdev_phys_t))
	continue;

	if (nvlist_unpack(label->vp_nvlist,
	sizeof (label->vp_nvlist), config, 0) != 0)
	continue;

	if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
	&guid) != 0 \|\| guid == 0) {
	nvlist_free(*config);
	continue;
	}

	if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
	&state) != 0 \|\| state > POOL_STATE_L2CACHE) {
	nvlist_free(*config);
	continue;
	}

	if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
	(nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
	&txg) != 0 \|\| txg == 0)) {
	nvlist_free(*config);
	continue;
	}

	if (expected_guid) {
	if (expected_guid == guid)
	count++;

	nvlist_free(*config);
	} else {
	expected_config = *config;
	expected_guid = guid;
	count++;
	}
	}

	if (num_labels != NULL)
	*num_labels = count;

	free(label);
	*config = expected_config;

	return (0);
	}

	/*
	* Given a file descriptor, read the label information and return an nvlist
	* describing the configuration, if there is one. The number of valid
	* labels found will be returned in num_labels when non-NULL.
	*/
	int
	zpool_read_label(int fd, nvlist_t *config, int num_labels)
	{
	#ifndef HAVE_AIO_H
	return (zpool_read_label_slow(fd, config, num_labels));
	#else
	struct stat64 statbuf;
	struct aiocb aiocbs[VDEV_LABELS];
	struct aiocb *aiocbps[VDEV_LABELS];
	vdev_phys_t *labels;
	nvlist_t *expected_config = NULL;
	uint64_t expected_guid = 0, size;
	int error, l, count = 0;

	*config = NULL;

	if (fstat64_blk(fd, &statbuf) == -1)
	return (0);
	size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);

	error = posix_memalign((void **)&labels, PAGESIZE,
	VDEV_LABELS * sizeof (*labels));
	if (error)
	return (-1);

	memset(aiocbs, 0, sizeof (aiocbs));
	for (l = 0; l < VDEV_LABELS; l++) {
	off_t offset = label_offset(size, l) + VDEV_SKIP_SIZE;

	aiocbs[l].aio_fildes = fd;
	aiocbs[l].aio_offset = offset;
	aiocbs[l].aio_buf = &labels[l];
	aiocbs[l].aio_nbytes = sizeof (vdev_phys_t);
	aiocbs[l].aio_lio_opcode = LIO_READ;
	aiocbps[l] = &aiocbs[l];
	}

	if (lio_listio(LIO_WAIT, aiocbps, VDEV_LABELS, NULL) != 0) {
	int saved_errno = errno;
	boolean_t do_slow = B_FALSE;
	error = -1;

	if (errno == EAGAIN \|\| errno == EINTR \|\| errno == EIO) {
	/*
	* A portion of the requests may have been submitted.
	* Clean them up.
	*/
	for (l = 0; l < VDEV_LABELS; l++) {
	errno = 0;
	switch (aio_error(&aiocbs[l])) {
	case EINVAL:
	break;
	case EINPROGRESS:
	// This shouldn't be possible to
	// encounter, die if we do.
	ASSERT(B_FALSE);
	zfs_fallthrough;
	case EOPNOTSUPP:
	case ENOSYS:
	do_slow = B_TRUE;
	zfs_fallthrough;
	case 0:
	default:
	(void) aio_return(&aiocbs[l]);
	}
	}
	}
	if (do_slow) {
	/*
	* At least some IO involved access unsafe-for-AIO
	* files. Let's try again, without AIO this time.
	*/
	error = zpool_read_label_slow(fd, config, num_labels);
	saved_errno = errno;
	}
	free(labels);
	errno = saved_errno;
	return (error);
	}

	for (l = 0; l < VDEV_LABELS; l++) {
	uint64_t state, guid, txg;

	if (aio_return(&aiocbs[l]) != sizeof (vdev_phys_t))
	continue;

	if (nvlist_unpack(labels[l].vp_nvlist,
	sizeof (labels[l].vp_nvlist), config, 0) != 0)
	continue;

	if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
	&guid) != 0 \|\| guid == 0) {
	nvlist_free(*config);
	continue;
	}

	if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
	&state) != 0 \|\| state > POOL_STATE_L2CACHE) {
	nvlist_free(*config);
	continue;
	}

	if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
	(nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
	&txg) != 0 \|\| txg == 0)) {
	nvlist_free(*config);
	continue;
	}

	if (expected_guid) {
	if (expected_guid == guid)
	count++;

	nvlist_free(*config);
	} else {
	expected_config = *config;
	expected_guid = guid;
	count++;
	}
	}

	if (num_labels != NULL)
	*num_labels = count;

	free(labels);
	*config = expected_config;

	return (0);
	#endif
	}

	/*
	* Sorted by full path and then vdev guid to allow for multiple entries with
	* the same full path name. This is required because it's possible to
	* have multiple block devices with labels that refer to the same
	* ZPOOL_CONFIG_PATH yet have different vdev guids. In this case both
	* entries need to be added to the cache. Scenarios where this can occur
	* include overwritten pool labels, devices which are visible from multiple
	* hosts and multipath devices.
	*/
	int
	slice_cache_compare(const void arg1, const void arg2)
	{
	const char nm1 = ((rdsk_node_t )arg1)->rn_name;
	const char nm2 = ((rdsk_node_t )arg2)->rn_name;
	uint64_t guid1 = ((rdsk_node_t *)arg1)->rn_vdev_guid;
	uint64_t guid2 = ((rdsk_node_t *)arg2)->rn_vdev_guid;
	int rv;

	rv = TREE_ISIGN(strcmp(nm1, nm2));
	if (rv)
	return (rv);

	return (TREE_CMP(guid1, guid2));
	}

	static int
	label_paths_impl(libpc_handle_t hdl, nvlist_t nvroot, uint64_t pool_guid,
	uint64_t vdev_guid, char path, char devid)
	{
	nvlist_t **child;
	uint_t c, children;
	uint64_t guid;
	char *val;
	int error;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++) {
	error = label_paths_impl(hdl, child[c],
	pool_guid, vdev_guid, path, devid);
	if (error)
	return (error);
	}
	return (0);
	}

	if (nvroot == NULL)
	return (0);

	error = nvlist_lookup_uint64(nvroot, ZPOOL_CONFIG_GUID, &guid);
	if ((error != 0) \|\| (guid != vdev_guid))
	return (0);

	error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_PATH, &val);
	if (error == 0)
	*path = val;

	error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_DEVID, &val);
	if (error == 0)
	*devid = val;

	return (0);
	}

	/*
	* Given a disk label fetch the ZPOOL_CONFIG_PATH and ZPOOL_CONFIG_DEVID
	* and store these strings as config_path and devid_path respectively.
	* The returned pointers are only valid as long as label remains valid.
	*/
	int
	label_paths(libpc_handle_t hdl, nvlist_t label, char path, char devid)
	{
	nvlist_t *nvroot;
	uint64_t pool_guid;
	uint64_t vdev_guid;

	*path = NULL;
	*devid = NULL;

	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvroot) \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &pool_guid) \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &vdev_guid))
	return (ENOENT);

	return (label_paths_impl(hdl, nvroot, pool_guid, vdev_guid, path,
	devid));
	}

	static void
	zpool_find_import_scan_add_slice(libpc_handle_t hdl, pthread_mutex_t lock,
	avl_tree_t cache, const char path, const char *name, int order)
	{
	avl_index_t where;
	rdsk_node_t *slice;

	slice = zutil_alloc(hdl, sizeof (rdsk_node_t));
	if (asprintf(&slice->rn_name, "%s/%s", path, name) == -1) {
	free(slice);
	return;
	}
	slice->rn_vdev_guid = 0;
	slice->rn_lock = lock;
	slice->rn_avl = cache;
	slice->rn_hdl = hdl;
	slice->rn_order = order + IMPORT_ORDER_SCAN_OFFSET;
	slice->rn_labelpaths = B_FALSE;

	pthread_mutex_lock(lock);
	if (avl_find(cache, slice, &where)) {
	free(slice->rn_name);
	free(slice);
	} else {
	avl_insert(cache, slice, where);
	}
	pthread_mutex_unlock(lock);
	}

	static int
	zpool_find_import_scan_dir(libpc_handle_t hdl, pthread_mutex_t lock,
	avl_tree_t cache, const char dir, int order)
	{
	int error;
	char path[MAXPATHLEN];
	struct dirent64 *dp;
	DIR *dirp;

	if (realpath(dir, path) == NULL) {
	error = errno;
	if (error == ENOENT)
	return (0);

	zutil_error_aux(hdl, "%s", strerror(error));
	(void) zutil_error_fmt(hdl, EZFS_BADPATH, dgettext(
	TEXT_DOMAIN, "cannot resolve path '%s'"), dir);
	return (error);
	}

	dirp = opendir(path);
	if (dirp == NULL) {
	error = errno;
	zutil_error_aux(hdl, "%s", strerror(error));
	(void) zutil_error_fmt(hdl, EZFS_BADPATH,
	dgettext(TEXT_DOMAIN, "cannot open '%s'"), path);
	return (error);
	}

	while ((dp = readdir64(dirp)) != NULL) {
	const char *name = dp->d_name;
	if (strcmp(name, ".") == 0 \|\| strcmp(name, "..") == 0)
	continue;

	switch (dp->d_type) {
	case DT_UNKNOWN:
	case DT_BLK:
	case DT_LNK:
	#ifdef __FreeBSD__
	case DT_CHR:
	#endif
	case DT_REG:
	break;
	default:
	continue;
	}

	zpool_find_import_scan_add_slice(hdl, lock, cache, path, name,
	order);
	}

	(void) closedir(dirp);
	return (0);
	}

	static int
	zpool_find_import_scan_path(libpc_handle_t hdl, pthread_mutex_t lock,
	avl_tree_t cache, const char dir, int order)
	{
	int error = 0;
	char path[MAXPATHLEN];
	char *d = NULL;
	ssize_t dl;
	const char dpath, name;

	/*
	* Separate the directory and the basename.
	* We do this so that we can get the realpath of
	* the directory. We don't get the realpath on the
	* whole path because if it's a symlink, we want the
	* path of the symlink not where it points to.
	*/
	name = zfs_basename(dir);
	if ((dl = zfs_dirnamelen(dir)) == -1)
	dpath = ".";
	else
	dpath = d = zutil_strndup(hdl, dir, dl);

	if (realpath(dpath, path) == NULL) {
	error = errno;
	if (error == ENOENT) {
	error = 0;
	goto out;
	}

	zutil_error_aux(hdl, "%s", strerror(error));
	(void) zutil_error_fmt(hdl, EZFS_BADPATH, dgettext(
	TEXT_DOMAIN, "cannot resolve path '%s'"), dir);
	goto out;
	}

	zpool_find_import_scan_add_slice(hdl, lock, cache, path, name, order);

	out:
	free(d);
	return (error);
	}

	/*
	* Scan a list of directories for zfs devices.
	*/
	static int
	zpool_find_import_scan(libpc_handle_t hdl, pthread_mutex_t lock,
	avl_tree_t *slice_cache, const char const *dir, size_t dirs)
	{
	avl_tree_t *cache;
	rdsk_node_t *slice;
	void *cookie;
	int i, error;

	*slice_cache = NULL;
	cache = zutil_alloc(hdl, sizeof (avl_tree_t));
	avl_create(cache, slice_cache_compare, sizeof (rdsk_node_t),
	offsetof(rdsk_node_t, rn_node));

	for (i = 0; i < dirs; i++) {
	struct stat sbuf;

	if (stat(dir[i], &sbuf) != 0) {
	error = errno;
	if (error == ENOENT)
	continue;

	zutil_error_aux(hdl, "%s", strerror(error));
	(void) zutil_error_fmt(hdl, EZFS_BADPATH, dgettext(
	TEXT_DOMAIN, "cannot resolve path '%s'"), dir[i]);
	goto error;
	}

	/*
	* If dir[i] is a directory, we walk through it and add all
	* the entries to the cache. If it's not a directory, we just
	* add it to the cache.
	*/
	if (S_ISDIR(sbuf.st_mode)) {
	if ((error = zpool_find_import_scan_dir(hdl, lock,
	cache, dir[i], i)) != 0)
	goto error;
	} else {
	if ((error = zpool_find_import_scan_path(hdl, lock,
	cache, dir[i], i)) != 0)
	goto error;
	}
	}

	*slice_cache = cache;
	return (0);

	error:
	cookie = NULL;
	while ((slice = avl_destroy_nodes(cache, &cookie)) != NULL) {
	free(slice->rn_name);
	free(slice);
	}
	free(cache);

	return (error);
	}

	/*
	* Given a list of directories to search, find all pools stored on disk. This
	* includes partial pools which are not available to import. If no args are
	* given (argc is 0), then the default directory (/dev/dsk) is searched.
	* poolname or guid (but not both) are provided by the caller when trying
	* to import a specific pool.
	*/
	static nvlist_t *
	zpool_find_import_impl(libpc_handle_t hdl, importargs_t iarg,
	pthread_mutex_t lock, avl_tree_t cache)
	{
	(void) lock;
	nvlist_t *ret = NULL;
	pool_list_t pools = { 0 };
	pool_entry_t pe, penext;
	vdev_entry_t ve, venext;
	config_entry_t ce, cenext;
	name_entry_t ne, nenext;
	rdsk_node_t *slice;
	void *cookie;
	tpool_t *t;

	verify(iarg->poolname == NULL \|\| iarg->guid == 0);

	/*
	* Create a thread pool to parallelize the process of reading and
	* validating labels, a large number of threads can be used due to
	* minimal contention.
	*/
	t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN), 0, NULL);
	for (slice = avl_first(cache); slice;
	(slice = avl_walk(cache, slice, AVL_AFTER)))
	(void) tpool_dispatch(t, zpool_open_func, slice);

	tpool_wait(t);
	tpool_destroy(t);

	/*
	* Process the cache, filtering out any entries which are not
	* for the specified pool then adding matching label configs.
	*/
	cookie = NULL;
	while ((slice = avl_destroy_nodes(cache, &cookie)) != NULL) {
	if (slice->rn_config != NULL) {
	nvlist_t *config = slice->rn_config;
	boolean_t matched = B_TRUE;
	boolean_t aux = B_FALSE;
	int fd;

	/*
	* Check if it's a spare or l2cache device. If it is,
	* we need to skip the name and guid check since they
	* don't exist on aux device label.
	*/
	if (iarg->poolname != NULL \|\| iarg->guid != 0) {
	uint64_t state;
	aux = nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_POOL_STATE, &state) == 0 &&
	(state == POOL_STATE_SPARE \|\|
	state == POOL_STATE_L2CACHE);
	}

	if (iarg->poolname != NULL && !aux) {
	char *pname;

	matched = nvlist_lookup_string(config,
	ZPOOL_CONFIG_POOL_NAME, &pname) == 0 &&
	strcmp(iarg->poolname, pname) == 0;
	} else if (iarg->guid != 0 && !aux) {
	uint64_t this_guid;

	matched = nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_POOL_GUID, &this_guid) == 0 &&
	iarg->guid == this_guid;
	}
	if (matched) {
	/*
	* Verify all remaining entries can be opened
	* exclusively. This will prune all underlying
	* multipath devices which otherwise could
	* result in the vdev appearing as UNAVAIL.
	*
	* Under zdb, this step isn't required and
	* would prevent a zdb -e of active pools with
	* no cachefile.
	*/
	fd = open(slice->rn_name,
	O_RDONLY \| O_EXCL \| O_CLOEXEC);
	if (fd >= 0 \|\| iarg->can_be_active) {
	if (fd >= 0)
	close(fd);
	add_config(hdl, &pools,
	slice->rn_name, slice->rn_order,
	slice->rn_num_labels, config);
	}
	}
	nvlist_free(config);
	}
	free(slice->rn_name);
	free(slice);
	}
	avl_destroy(cache);
	free(cache);

	ret = get_configs(hdl, &pools, iarg->can_be_active, iarg->policy);

	for (pe = pools.pools; pe != NULL; pe = penext) {
	penext = pe->pe_next;
	for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
	venext = ve->ve_next;
	for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
	cenext = ce->ce_next;
	nvlist_free(ce->ce_config);
	free(ce);
	}
	free(ve);
	}
	free(pe);
	}

	for (ne = pools.names; ne != NULL; ne = nenext) {
	nenext = ne->ne_next;
	free(ne->ne_name);
	free(ne);
	}

	return (ret);
	}

	/*
	* Given a config, discover the paths for the devices which
	* exist in the config.
	*/
	static int
	discover_cached_paths(libpc_handle_t hdl, nvlist_t nv,
	avl_tree_t cache, pthread_mutex_t lock)
	{
	char *path = NULL;
	ssize_t dl;
	uint_t children;
	nvlist_t **child;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (int c = 0; c < children; c++) {
	discover_cached_paths(hdl, child[c], cache, lock);
	}
	}

	/*
	* Once we have the path, we need to add the directory to
	* our directory cache.
	*/
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
	if ((dl = zfs_dirnamelen(path)) == -1)
	path = (char *)".";
	else
	path[dl] = '\0';
	return (zpool_find_import_scan_dir(hdl, lock, cache,
	path, 0));
	}
	return (0);
	}

	/*
	* Given a cache file, return the contents as a list of importable pools.
	* poolname or guid (but not both) are provided by the caller when trying
	* to import a specific pool.
	*/
	static nvlist_t *
	zpool_find_import_cached(libpc_handle_t hdl, importargs_t iarg)
	{
	char *buf;
	int fd;
	struct stat64 statbuf;
	nvlist_t raw, src, *dst;
	nvlist_t *pools;
	nvpair_t *elem;
	char *name;
	uint64_t this_guid;
	boolean_t active;

	verify(iarg->poolname == NULL \|\| iarg->guid == 0);

	if ((fd = open(iarg->cachefile, O_RDONLY \| O_CLOEXEC)) < 0) {
	zutil_error_aux(hdl, "%s", strerror(errno));
	(void) zutil_error(hdl, EZFS_BADCACHE,
	dgettext(TEXT_DOMAIN, "failed to open cache file"));
	return (NULL);
	}

	if (fstat64(fd, &statbuf) != 0) {
	zutil_error_aux(hdl, "%s", strerror(errno));
	(void) close(fd);
	(void) zutil_error(hdl, EZFS_BADCACHE,
	dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
	return (NULL);
	}

	if ((buf = zutil_alloc(hdl, statbuf.st_size)) == NULL) {
	(void) close(fd);
	return (NULL);
	}

	if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
	(void) close(fd);
	free(buf);
	(void) zutil_error(hdl, EZFS_BADCACHE,
	dgettext(TEXT_DOMAIN,
	"failed to read cache file contents"));
	return (NULL);
	}

	(void) close(fd);

	if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
	free(buf);
	(void) zutil_error(hdl, EZFS_BADCACHE,
	dgettext(TEXT_DOMAIN,
	"invalid or corrupt cache file contents"));
	return (NULL);
	}

	free(buf);

	/*
	* Go through and get the current state of the pools and refresh their
	* state.
	*/
	if (nvlist_alloc(&pools, 0, 0) != 0) {
	(void) zutil_no_memory(hdl);
	nvlist_free(raw);
	return (NULL);
	}

	elem = NULL;
	while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
	src = fnvpair_value_nvlist(elem);

	name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
	if (iarg->poolname != NULL && strcmp(iarg->poolname, name) != 0)
	continue;

	this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
	if (iarg->guid != 0 && iarg->guid != this_guid)
	continue;

	if (zutil_pool_active(hdl, name, this_guid, &active) != 0) {
	nvlist_free(raw);
	nvlist_free(pools);
	return (NULL);
	}

	if (active)
	continue;

	if (iarg->scan) {
	uint64_t saved_guid = iarg->guid;
	const char *saved_poolname = iarg->poolname;
	pthread_mutex_t lock;

	/*
	* Create the device cache that will hold the
	* devices we will scan based on the cachefile.
	* This will get destroyed and freed by
	* zpool_find_import_impl.
	*/
	avl_tree_t *cache = zutil_alloc(hdl,
	sizeof (avl_tree_t));
	avl_create(cache, slice_cache_compare,
	sizeof (rdsk_node_t),
	offsetof(rdsk_node_t, rn_node));
	nvlist_t *nvroot = fnvlist_lookup_nvlist(src,
	ZPOOL_CONFIG_VDEV_TREE);

	/*
	* We only want to find the pool with this_guid.
	* We will reset these values back later.
	*/
	iarg->guid = this_guid;
	iarg->poolname = NULL;

	/*
	* We need to build up a cache of devices that exists
	* in the paths pointed to by the cachefile. This allows
	* us to preserve the device namespace that was
	* originally specified by the user but also lets us
	* scan devices in those directories in case they had
	* been renamed.
	*/
	pthread_mutex_init(&lock, NULL);
	discover_cached_paths(hdl, nvroot, cache, &lock);
	nvlist_t *nv = zpool_find_import_impl(hdl, iarg,
	&lock, cache);
	pthread_mutex_destroy(&lock);

	/*
	* zpool_find_import_impl will return back
	* a list of pools that it found based on the
	* device cache. There should only be one pool
	* since we're looking for a specific guid.
	* We will use that pool to build up the final
	* pool nvlist which is returned back to the
	* caller.
	*/
	nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
	if (pair == NULL)
	continue;
	fnvlist_add_nvlist(pools, nvpair_name(pair),
	fnvpair_value_nvlist(pair));

	VERIFY3P(nvlist_next_nvpair(nv, pair), ==, NULL);

	iarg->guid = saved_guid;
	iarg->poolname = saved_poolname;
	continue;
	}

	if (nvlist_add_string(src, ZPOOL_CONFIG_CACHEFILE,
	iarg->cachefile) != 0) {
	(void) zutil_no_memory(hdl);
	nvlist_free(raw);
	nvlist_free(pools);
	return (NULL);
	}

	update_vdevs_config_dev_sysfs_path(src);

	if ((dst = zutil_refresh_config(hdl, src)) == NULL) {
	nvlist_free(raw);
	nvlist_free(pools);
	return (NULL);
	}

	if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
	(void) zutil_no_memory(hdl);
	nvlist_free(dst);
	nvlist_free(raw);
	nvlist_free(pools);
	return (NULL);
	}
	nvlist_free(dst);
	}
	nvlist_free(raw);
	return (pools);
	}

	static nvlist_t *
	zpool_find_import(libpc_handle_t hdl, importargs_t iarg)
	{
	pthread_mutex_t lock;
	avl_tree_t *cache;
	nvlist_t *pools = NULL;

	verify(iarg->poolname == NULL \|\| iarg->guid == 0);
	pthread_mutex_init(&lock, NULL);

	/*
	* Locate pool member vdevs by blkid or by directory scanning.
	* On success a newly allocated AVL tree which is populated with an
	* entry for each discovered vdev will be returned in the cache.
	* It's the caller's responsibility to consume and destroy this tree.
	*/
	if (iarg->scan \|\| iarg->paths != 0) {
	size_t dirs = iarg->paths;
	const char * const dir = (const char const *)iarg->path;

	if (dirs == 0)
	dir = zpool_default_search_paths(&dirs);

	if (zpool_find_import_scan(hdl, &lock, &cache,
	dir, dirs) != 0) {
	pthread_mutex_destroy(&lock);
	return (NULL);
	}
	} else {
	if (zpool_find_import_blkid(hdl, &lock, &cache) != 0) {
	pthread_mutex_destroy(&lock);
	return (NULL);
	}
	}

	pools = zpool_find_import_impl(hdl, iarg, &lock, cache);
	pthread_mutex_destroy(&lock);
	return (pools);
	}


	nvlist_t *
	-zpool_search_import(void hdl, importargs_t import,
	- const pool_config_ops_t *pco)
	+zpool_search_import(void hdl, importargs_t import, pool_config_ops_t *pco)
	{
	libpc_handle_t handle = { 0 };
	nvlist_t *pools = NULL;

	handle.lpc_lib_handle = hdl;
	handle.lpc_ops = pco;
	handle.lpc_printerr = B_TRUE;

	verify(import->poolname == NULL \|\| import->guid == 0);

	if (import->cachefile != NULL)
	pools = zpool_find_import_cached(&handle, import);
	else
	pools = zpool_find_import(&handle, import);

	if ((pools == NULL \|\| nvlist_empty(pools)) &&
	handle.lpc_open_access_error && geteuid() != 0) {
	(void) zutil_error(&handle, EZFS_EACESS, dgettext(TEXT_DOMAIN,
	"no pools found"));
	}

	return (pools);
	}

	static boolean_t
	pool_match(nvlist_t cfg, char tgt)
	{
	uint64_t v, guid = strtoull(tgt, NULL, 0);
	char *s;

	if (guid != 0) {
	if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_GUID, &v) == 0)
	return (v == guid);
	} else {
	if (nvlist_lookup_string(cfg, ZPOOL_CONFIG_POOL_NAME, &s) == 0)
	return (strcmp(s, tgt) == 0);
	}
	return (B_FALSE);
	}

	int
	zpool_find_config(void hdl, const char target, nvlist_t **configp,
	- importargs_t args, const pool_config_ops_t pco)
	+ importargs_t args, pool_config_ops_t pco)
	{
	nvlist_t *pools;
	nvlist_t *match = NULL;
	nvlist_t *config = NULL;
	char *sepp = NULL;
	int count = 0;
	char *targetdup = strdup(target);

	*configp = NULL;

	if ((sepp = strpbrk(targetdup, "/@")) != NULL)
	*sepp = '\0';

	pools = zpool_search_import(hdl, args, pco);

	if (pools != NULL) {
	nvpair_t *elem = NULL;
	while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
	VERIFY0(nvpair_value_nvlist(elem, &config));
	if (pool_match(config, targetdup)) {
	count++;
	if (match != NULL) {
	/* multiple matches found */
	continue;
	} else {
	match = fnvlist_dup(config);
	}
	}
	}
	fnvlist_free(pools);
	}

	if (count == 0) {
	free(targetdup);
	return (ENOENT);
	}

	if (count > 1) {
	free(targetdup);
	fnvlist_free(match);
	return (EINVAL);
	}

	*configp = match;
	free(targetdup);

	return (0);
	}

	/*
	* Internal function for iterating over the vdevs.
	*
	* For each vdev, func() will be called and will be passed 'zhp' (which is
	* typically the zpool_handle_t cast as a void pointer), the vdev's nvlist, and
	* a user-defined data pointer).
	*
	* The return values from all the func() calls will be OR'd together and
	* returned.
	*/
	int
	for_each_vdev_cb(void zhp, nvlist_t nv, pool_vdev_iter_f func,
	void *data)
	{
	nvlist_t **child;
	uint_t c, children;
	int ret = 0;
	int i;
	char *type;

	const char *list[] = {
	ZPOOL_CONFIG_SPARES,
	ZPOOL_CONFIG_L2CACHE,
	ZPOOL_CONFIG_CHILDREN
	};

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
	return (ret);

	/* Don't run our function on root or indirect vdevs */
	if ((strcmp(type, VDEV_TYPE_ROOT) != 0) &&
	(strcmp(type, VDEV_TYPE_INDIRECT) != 0)) {
	ret \|= func(zhp, nv, data);
	}

	for (i = 0; i < ARRAY_SIZE(list); i++) {
	if (nvlist_lookup_nvlist_array(nv, list[i], &child,
	&children) == 0) {
	for (c = 0; c < children; c++) {
	uint64_t ishole = 0;

	(void) nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_HOLE, &ishole);

	if (ishole)
	continue;

	ret \|= for_each_vdev_cb(zhp, child[c],
	func, data);
	}
	}
	}

	return (ret);
	}

	/*
	* Given an ZPOOL_CONFIG_VDEV_TREE nvpair, iterate over all the vdevs, calling
	* func() for each one. func() is passed the vdev's nvlist and an optional
	* user-defined 'data' pointer.
	*/
	int
	for_each_vdev_in_nvlist(nvlist_t nvroot, pool_vdev_iter_f func, void data)
	{
	return (for_each_vdev_cb(NULL, nvroot, func, data));
	}
	diff --git a/sys/contrib/openzfs/lib/libzutil/zutil_import.h b/sys/contrib/openzfs/lib/libzutil/zutil_import.h
	index b68a7fcfcb78..482315e44130 100644
	--- a/sys/contrib/openzfs/lib/libzutil/zutil_import.h
	+++ b/sys/contrib/openzfs/lib/libzutil/zutil_import.h
	@@ -1,76 +1,76 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright 2015 RackTop Systems.
	* Copyright (c) 2016, Intel Corporation.
	*/
	#ifndef _LIBZUTIL_ZUTIL_IMPORT_H_
	#define _LIBZUTIL_ZUTIL_IMPORT_H_

	#define EZFS_BADCACHE "invalid or missing cache file"
	#define EZFS_BADPATH "must be an absolute path"
	#define EZFS_NOMEM "out of memory"
	#define EZFS_EACESS "some devices require root privileges"

	#define IMPORT_ORDER_PREFERRED_1 1
	#define IMPORT_ORDER_PREFERRED_2 2
	#define IMPORT_ORDER_SCAN_OFFSET 10
	#define IMPORT_ORDER_DEFAULT 100

	typedef struct libpc_handle {
	boolean_t lpc_printerr;
	boolean_t lpc_open_access_error;
	boolean_t lpc_desc_active;
	char lpc_desc[1024];
	- const pool_config_ops_t *lpc_ops;
	+ pool_config_ops_t *lpc_ops;
	void *lpc_lib_handle;
	} libpc_handle_t;


	int label_paths(libpc_handle_t hdl, nvlist_t label, char **path,
	char **devid);
	int zpool_find_import_blkid(libpc_handle_t hdl, pthread_mutex_t lock,
	avl_tree_t **slice_cache);

	void * zutil_alloc(libpc_handle_t *hdl, size_t size);
	char zutil_strdup(libpc_handle_t hdl, const char *str);

	typedef struct rdsk_node {
	char rn_name; / Full path to device */
	int rn_order; /* Preferred order (low to high) */
	int rn_num_labels; /* Number of valid labels */
	uint64_t rn_vdev_guid; /* Expected vdev guid when set */
	libpc_handle_t *rn_hdl;
	nvlist_t rn_config; / Label config */
	avl_tree_t *rn_avl;
	avl_node_t rn_node;
	pthread_mutex_t *rn_lock;
	boolean_t rn_labelpaths;
	} rdsk_node_t;

	int slice_cache_compare(const void , const void );

	void zpool_open_func(void *);

	#endif /* _LIBZUTIL_ZUTIL_IMPORT_H_ */
	diff --git a/sys/contrib/openzfs/man/man4/zfs.4 b/sys/contrib/openzfs/man/man4/zfs.4
	index cc8ff316791c..da1aa9c34287 100644
	--- a/sys/contrib/openzfs/man/man4/zfs.4
	+++ b/sys/contrib/openzfs/man/man4/zfs.4
	@@ -1,2579 +1,2585 @@
	.\"
	.\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
	.\" Copyright (c) 2019, 2021 by Delphix. All rights reserved.
	.\" Copyright (c) 2019 Datto Inc.
	.\" The contents of this file are subject to the terms of the Common Development
	.\" and Distribution License (the "License"). You may not use this file except
	.\" in compliance with the License. You can obtain a copy of the license at
	.\" usr/src/OPENSOLARIS.LICENSE or https://opensource.org/licenses/CDDL-1.0.
	.\"
	.\" See the License for the specific language governing permissions and
	.\" limitations under the License. When distributing Covered Code, include this
	.\" CDDL HEADER in each file and include the License file at
	.\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this
	.\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your
	.\" own identifying information:
	.\" Portions Copyright [yyyy] [name of copyright owner]
	.\"
	.Dd June 1, 2021
	.Dt ZFS 4
	.Os
	.
	.Sh NAME
	.Nm zfs
	.Nd tuning of the ZFS kernel module
	.
	.Sh DESCRIPTION
	The ZFS module supports these parameters:
	.Bl -tag -width Ds
	.It Sy dbuf_cache_max_bytes Ns = Ns Sy ULONG_MAX Ns B Pq ulong
	Maximum size in bytes of the dbuf cache.
	The target size is determined by the MIN versus
	.No 1/2^ Ns Sy dbuf_cache_shift Pq 1/32nd
	of the target ARC size.
	The behavior of the dbuf cache and its associated settings
	can be observed via the
	.Pa /proc/spl/kstat/zfs/dbufstats
	kstat.
	.
	.It Sy dbuf_metadata_cache_max_bytes Ns = Ns Sy ULONG_MAX Ns B Pq ulong
	Maximum size in bytes of the metadata dbuf cache.
	The target size is determined by the MIN versus
	.No 1/2^ Ns Sy dbuf_metadata_cache_shift Pq 1/64th
	of the target ARC size.
	The behavior of the metadata dbuf cache and its associated settings
	can be observed via the
	.Pa /proc/spl/kstat/zfs/dbufstats
	kstat.
	.
	.It Sy dbuf_cache_hiwater_pct Ns = Ns Sy 10 Ns % Pq uint
	The percentage over
	.Sy dbuf_cache_max_bytes
	when dbufs must be evicted directly.
	.
	.It Sy dbuf_cache_lowater_pct Ns = Ns Sy 10 Ns % Pq uint
	The percentage below
	.Sy dbuf_cache_max_bytes
	when the evict thread stops evicting dbufs.
	.
	-.It Sy dbuf_cache_shift Ns = Ns Sy 5 Pq int
	+.It Sy dbuf_cache_shift Ns = Ns Sy 5 Pq uint
	Set the size of the dbuf cache
	.Pq Sy dbuf_cache_max_bytes
	to a log2 fraction of the target ARC size.
	.
	-.It Sy dbuf_metadata_cache_shift Ns = Ns Sy 6 Pq int
	+.It Sy dbuf_metadata_cache_shift Ns = Ns Sy 6 Pq uint
	Set the size of the dbuf metadata cache
	.Pq Sy dbuf_metadata_cache_max_bytes
	to a log2 fraction of the target ARC size.
	.
	-.It Sy dmu_object_alloc_chunk_shift Ns = Ns Sy 7 Po 128 Pc Pq int
	+.It Sy dbuf_mutex_cache_shift Ns = Ns Sy 0 Pq uint
	+Set the size of the mutex array for the dbuf cache.
	+When set to
	+.Sy 0
	+the array is dynamically sized based on total system memory.
	+.
	+.It Sy dmu_object_alloc_chunk_shift Ns = Ns Sy 7 Po 128 Pc Pq uint
	dnode slots allocated in a single operation as a power of 2.
	The default value minimizes lock contention for the bulk operation performed.
	.
	-.It Sy dmu_prefetch_max Ns = Ns Sy 134217728 Ns B Po 128 MiB Pc Pq int
	+.It Sy dmu_prefetch_max Ns = Ns Sy 134217728 Ns B Po 128 MiB Pc Pq uint
	Limit the amount we can prefetch with one call to this amount in bytes.
	This helps to limit the amount of memory that can be used by prefetching.
	.
	.It Sy ignore_hole_birth Pq int
	Alias for
	.Sy send_holes_without_birth_time .
	.
	.It Sy l2arc_feed_again Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Turbo L2ARC warm-up.
	When the L2ARC is cold the fill interval will be set as fast as possible.
	.
	.It Sy l2arc_feed_min_ms Ns = Ns Sy 200 Pq ulong
	Min feed interval in milliseconds.
	Requires
	.Sy l2arc_feed_again Ns = Ns Ar 1
	and only applicable in related situations.
	.
	.It Sy l2arc_feed_secs Ns = Ns Sy 1 Pq ulong
	Seconds between L2ARC writing.
	.
	.It Sy l2arc_headroom Ns = Ns Sy 2 Pq ulong
	How far through the ARC lists to search for L2ARC cacheable content,
	expressed as a multiplier of
	.Sy l2arc_write_max .
	ARC persistence across reboots can be achieved with persistent L2ARC
	by setting this parameter to
	.Sy 0 ,
	allowing the full length of ARC lists to be searched for cacheable content.
	.
	.It Sy l2arc_headroom_boost Ns = Ns Sy 200 Ns % Pq ulong
	Scales
	.Sy l2arc_headroom
	by this percentage when L2ARC contents are being successfully compressed
	before writing.
	A value of
	.Sy 100
	disables this feature.
	.
	.It Sy l2arc_exclude_special Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Controls whether buffers present on special vdevs are eligible for caching
	into L2ARC.
	If set to 1, exclude dbufs on special vdevs from being cached to L2ARC.
	.
	.It Sy l2arc_mfuonly Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Controls whether only MFU metadata and data are cached from ARC into L2ARC.
	This may be desired to avoid wasting space on L2ARC when reading/writing large
	amounts of data that are not expected to be accessed more than once.
	.Pp
	The default is off,
	meaning both MRU and MFU data and metadata are cached.
	When turning off this feature, some MRU buffers will still be present
	in ARC and eventually cached on L2ARC.
	.No If Sy l2arc_noprefetch Ns = Ns Sy 0 ,
	some prefetched buffers will be cached to L2ARC, and those might later
	transition to MRU, in which case the
	.Sy l2arc_mru_asize No arcstat will not be Sy 0 .
	.Pp
	Regardless of
	.Sy l2arc_noprefetch ,
	some MFU buffers might be evicted from ARC,
	accessed later on as prefetches and transition to MRU as prefetches.
	If accessed again they are counted as MRU and the
	.Sy l2arc_mru_asize No arcstat will not be Sy 0 .
	.Pp
	The ARC status of L2ARC buffers when they were first cached in
	L2ARC can be seen in the
	.Sy l2arc_mru_asize , Sy l2arc_mfu_asize , No and Sy l2arc_prefetch_asize
	arcstats when importing the pool or onlining a cache
	device if persistent L2ARC is enabled.
	.Pp
	The
	.Sy evict_l2_eligible_mru
	arcstat does not take into account if this option is enabled as the information
	provided by the
	.Sy evict_l2_eligible_m[rf]u
	arcstats can be used to decide if toggling this option is appropriate
	for the current workload.
	.
	-.It Sy l2arc_meta_percent Ns = Ns Sy 33 Ns % Pq int
	+.It Sy l2arc_meta_percent Ns = Ns Sy 33 Ns % Pq uint
	Percent of ARC size allowed for L2ARC-only headers.
	Since L2ARC buffers are not evicted on memory pressure,
	too many headers on a system with an irrationally large L2ARC
	can render it slow or unusable.
	This parameter limits L2ARC writes and rebuilds to achieve the target.
	.
	.It Sy l2arc_trim_ahead Ns = Ns Sy 0 Ns % Pq ulong
	Trims ahead of the current write size
	.Pq Sy l2arc_write_max
	on L2ARC devices by this percentage of write size if we have filled the device.
	If set to
	.Sy 100
	we TRIM twice the space required to accommodate upcoming writes.
	A minimum of
	.Sy 64 MiB
	will be trimmed.
	It also enables TRIM of the whole L2ARC device upon creation
	or addition to an existing pool or if the header of the device is
	invalid upon importing a pool or onlining a cache device.
	A value of
	.Sy 0
	disables TRIM on L2ARC altogether and is the default as it can put significant
	stress on the underlying storage devices.
	This will vary depending of how well the specific device handles these commands.
	.
	.It Sy l2arc_noprefetch Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Do not write buffers to L2ARC if they were prefetched but not used by
	applications.
	In case there are prefetched buffers in L2ARC and this option
	is later set, we do not read the prefetched buffers from L2ARC.
	Unsetting this option is useful for caching sequential reads from the
	disks to L2ARC and serve those reads from L2ARC later on.
	This may be beneficial in case the L2ARC device is significantly faster
	in sequential reads than the disks of the pool.
	.Pp
	Use
	.Sy 1
	to disable and
	.Sy 0
	to enable caching/reading prefetches to/from L2ARC.
	.
	.It Sy l2arc_norw Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	No reads during writes.
	.
	.It Sy l2arc_write_boost Ns = Ns Sy 8388608 Ns B Po 8 MiB Pc Pq ulong
	Cold L2ARC devices will have
	.Sy l2arc_write_max
	increased by this amount while they remain cold.
	.
	.It Sy l2arc_write_max Ns = Ns Sy 8388608 Ns B Po 8 MiB Pc Pq ulong
	Max write bytes per interval.
	.
	.It Sy l2arc_rebuild_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Rebuild the L2ARC when importing a pool (persistent L2ARC).
	This can be disabled if there are problems importing a pool
	or attaching an L2ARC device (e.g. the L2ARC device is slow
	in reading stored log metadata, or the metadata
	has become somehow fragmented/unusable).
	.
	.It Sy l2arc_rebuild_blocks_min_l2size Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq ulong
	Mininum size of an L2ARC device required in order to write log blocks in it.
	The log blocks are used upon importing the pool to rebuild the persistent L2ARC.
	.Pp
	For L2ARC devices less than 1 GiB, the amount of data
	.Fn l2arc_evict
	evicts is significant compared to the amount of restored L2ARC data.
	In this case, do not write log blocks in L2ARC in order not to waste space.
	.
	.It Sy metaslab_aliquot Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
	Metaslab granularity, in bytes.
	This is roughly similar to what would be referred to as the "stripe size"
	in traditional RAID arrays.
	In normal operation, ZFS will try to write this amount of data to each disk
	before moving on to the next top-level vdev.
	.
	.It Sy metaslab_bias_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable metaslab group biasing based on their vdevs' over- or under-utilization
	relative to the pool.
	.
	.It Sy metaslab_force_ganging Ns = Ns Sy 16777217 Ns B Po 16 MiB + 1 B Pc Pq ulong
	Make some blocks above a certain size be gang blocks.
	This option is used by the test suite to facilitate testing.
	.
	.It Sy zfs_history_output_max Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	When attempting to log an output nvlist of an ioctl in the on-disk history,
	the output will not be stored if it is larger than this size (in bytes).
	This must be less than
	.Sy DMU_MAX_ACCESS Pq 64 MiB .
	This applies primarily to
	.Fn zfs_ioc_channel_program Pq cf. Xr zfs-program 8 .
	.
	.It Sy zfs_keep_log_spacemaps_at_export Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prevent log spacemaps from being destroyed during pool exports and destroys.
	.
	.It Sy zfs_metaslab_segment_weight_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable/disable segment-based metaslab selection.
	.
	.It Sy zfs_metaslab_switch_threshold Ns = Ns Sy 2 Pq int
	When using segment-based metaslab selection, continue allocating
	from the active metaslab until this option's
	worth of buckets have been exhausted.
	.
	.It Sy metaslab_debug_load Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Load all metaslabs during pool import.
	.
	.It Sy metaslab_debug_unload Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prevent metaslabs from being unloaded.
	.
	.It Sy metaslab_fragmentation_factor_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable use of the fragmentation metric in computing metaslab weights.
	.
	-.It Sy metaslab_df_max_search Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	+.It Sy metaslab_df_max_search Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	Maximum distance to search forward from the last offset.
	Without this limit, fragmented pools can see
	.Em >100`000
	iterations and
	.Fn metaslab_block_picker
	becomes the performance limiting factor on high-performance storage.
	.Pp
	With the default setting of
	.Sy 16 MiB ,
	we typically see less than
	.Em 500
	iterations, even with very fragmented
	.Sy ashift Ns = Ns Sy 9
	pools.
	The maximum number of iterations possible is
	.Sy metaslab_df_max_search / 2^(ashift+1) .
	With the default setting of
	.Sy 16 MiB
	this is
	.Em 16*1024 Pq with Sy ashift Ns = Ns Sy 9
	or
	.Em 2*1024 Pq with Sy ashift Ns = Ns Sy 12 .
	.
	.It Sy metaslab_df_use_largest_segment Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If not searching forward (due to
	.Sy metaslab_df_max_search , metaslab_df_free_pct ,
	.No or Sy metaslab_df_alloc_threshold ) ,
	this tunable controls which segment is used.
	If set, we will use the largest free segment.
	If unset, we will use a segment of at least the requested size.
	.
	.It Sy zfs_metaslab_max_size_cache_sec Ns = Ns Sy 3600 Ns s Po 1 hour Pc Pq ulong
	When we unload a metaslab, we cache the size of the largest free chunk.
	We use that cached size to determine whether or not to load a metaslab
	for a given allocation.
	As more frees accumulate in that metaslab while it's unloaded,
	the cached max size becomes less and less accurate.
	After a number of seconds controlled by this tunable,
	we stop considering the cached max size and start
	considering only the histogram instead.
	.
	-.It Sy zfs_metaslab_mem_limit Ns = Ns Sy 25 Ns % Pq int
	+.It Sy zfs_metaslab_mem_limit Ns = Ns Sy 25 Ns % Pq uint
	When we are loading a new metaslab, we check the amount of memory being used
	to store metaslab range trees.
	If it is over a threshold, we attempt to unload the least recently used metaslab
	to prevent the system from clogging all of its memory with range trees.
	This tunable sets the percentage of total system memory that is the threshold.
	.
	.It Sy zfs_metaslab_try_hard_before_gang Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	.Bl -item -compact
	.It
	If unset, we will first try normal allocation.
	.It
	If that fails then we will do a gang allocation.
	.It
	If that fails then we will do a "try hard" gang allocation.
	.It
	If that fails then we will have a multi-layer gang block.
	.El
	.Pp
	.Bl -item -compact
	.It
	If set, we will first try normal allocation.
	.It
	If that fails then we will do a "try hard" allocation.
	.It
	If that fails we will do a gang allocation.
	.It
	If that fails we will do a "try hard" gang allocation.
	.It
	If that fails then we will have a multi-layer gang block.
	.El
	.
	-.It Sy zfs_metaslab_find_max_tries Ns = Ns Sy 100 Pq int
	+.It Sy zfs_metaslab_find_max_tries Ns = Ns Sy 100 Pq uint
	When not trying hard, we only consider this number of the best metaslabs.
	This improves performance, especially when there are many metaslabs per vdev
	and the allocation can't actually be satisfied
	(so we would otherwise iterate all metaslabs).
	.
	-.It Sy zfs_vdev_default_ms_count Ns = Ns Sy 200 Pq int
	+.It Sy zfs_vdev_default_ms_count Ns = Ns Sy 200 Pq uint
	When a vdev is added, target this number of metaslabs per top-level vdev.
	.
	-.It Sy zfs_vdev_default_ms_shift Ns = Ns Sy 29 Po 512 MiB Pc Pq int
	+.It Sy zfs_vdev_default_ms_shift Ns = Ns Sy 29 Po 512 MiB Pc Pq uint
	Default limit for metaslab size.
	.
	.It Sy zfs_vdev_max_auto_ashift Ns = Ns Sy 14 Pq ulong
	Maximum ashift used when optimizing for logical \[->] physical sector size on new
	top-level vdevs.
	May be increased up to
	.Sy ASHIFT_MAX Po 16 Pc ,
	but this may negatively impact pool space efficiency.
	.
	.It Sy zfs_vdev_min_auto_ashift Ns = Ns Sy ASHIFT_MIN Po 9 Pc Pq ulong
	Minimum ashift used when creating new top-level vdevs.
	.
	-.It Sy zfs_vdev_min_ms_count Ns = Ns Sy 16 Pq int
	+.It Sy zfs_vdev_min_ms_count Ns = Ns Sy 16 Pq uint
	Minimum number of metaslabs to create in a top-level vdev.
	.
	.It Sy vdev_validate_skip Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Skip label validation steps during pool import.
	Changing is not recommended unless you know what you're doing
	and are recovering a damaged label.
	.
	-.It Sy zfs_vdev_ms_count_limit Ns = Ns Sy 131072 Po 128k Pc Pq int
	+.It Sy zfs_vdev_ms_count_limit Ns = Ns Sy 131072 Po 128k Pc Pq uint
	Practical upper limit of total metaslabs per top-level vdev.
	.
	.It Sy metaslab_preload_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable metaslab group preloading.
	.
	.It Sy metaslab_lba_weighting_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Give more weight to metaslabs with lower LBAs,
	assuming they have greater bandwidth,
	as is typically the case on a modern constant angular velocity disk drive.
	.
	-.It Sy metaslab_unload_delay Ns = Ns Sy 32 Pq int
	+.It Sy metaslab_unload_delay Ns = Ns Sy 32 Pq uint
	After a metaslab is used, we keep it loaded for this many TXGs, to attempt to
	reduce unnecessary reloading.
	Note that both this many TXGs and
	.Sy metaslab_unload_delay_ms
	milliseconds must pass before unloading will occur.
	.
	-.It Sy metaslab_unload_delay_ms Ns = Ns Sy 600000 Ns ms Po 10 min Pc Pq int
	+.It Sy metaslab_unload_delay_ms Ns = Ns Sy 600000 Ns ms Po 10 min Pc Pq uint
	After a metaslab is used, we keep it loaded for this many milliseconds,
	to attempt to reduce unnecessary reloading.
	Note, that both this many milliseconds and
	.Sy metaslab_unload_delay
	TXGs must pass before unloading will occur.
	.
	-.It Sy reference_history Ns = Ns Sy 3 Pq int
	+.It Sy reference_history Ns = Ns Sy 3 Pq uint
	Maximum reference holders being tracked when reference_tracking_enable is active.
	.
	.It Sy reference_tracking_enable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Track reference holders to
	.Sy refcount_t
	objects (debug builds only).
	.
	.It Sy send_holes_without_birth_time Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	When set, the
	.Sy hole_birth
	optimization will not be used, and all holes will always be sent during a
	.Nm zfs Cm send .
	This is useful if you suspect your datasets are affected by a bug in
	.Sy hole_birth .
	.
	.It Sy spa_config_path Ns = Ns Pa /etc/zfs/zpool.cache Pq charp
	SPA config file.
	.
	-.It Sy spa_asize_inflation Ns = Ns Sy 24 Pq int
	+.It Sy spa_asize_inflation Ns = Ns Sy 24 Pq uint
	Multiplication factor used to estimate actual disk consumption from the
	size of data being written.
	The default value is a worst case estimate,
	but lower values may be valid for a given pool depending on its configuration.
	Pool administrators who understand the factors involved
	may wish to specify a more realistic inflation factor,
	particularly if they operate close to quota or capacity limits.
	.
	.It Sy spa_load_print_vdev_tree Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Whether to print the vdev tree in the debugging message buffer during pool import.
	.
	.It Sy spa_load_verify_data Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Whether to traverse data blocks during an "extreme rewind"
	.Pq Fl X
	import.
	.Pp
	An extreme rewind import normally performs a full traversal of all
	blocks in the pool for verification.
	If this parameter is unset, the traversal skips non-metadata blocks.
	It can be toggled once the
	import has started to stop or start the traversal of non-metadata blocks.
	.
	.It Sy spa_load_verify_metadata Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Whether to traverse blocks during an "extreme rewind"
	.Pq Fl X
	pool import.
	.Pp
	An extreme rewind import normally performs a full traversal of all
	blocks in the pool for verification.
	If this parameter is unset, the traversal is not performed.
	It can be toggled once the import has started to stop or start the traversal.
	.
	-.It Sy spa_load_verify_shift Ns = Ns Sy 4 Po 1/16th Pc Pq int
	+.It Sy spa_load_verify_shift Ns = Ns Sy 4 Po 1/16th Pc Pq uint
	Sets the maximum number of bytes to consume during pool import to the log2
	fraction of the target ARC size.
	.
	.It Sy spa_slop_shift Ns = Ns Sy 5 Po 1/32nd Pc Pq int
	Normally, we don't allow the last
	.Sy 3.2% Pq Sy 1/2^spa_slop_shift
	of space in the pool to be consumed.
	This ensures that we don't run the pool completely out of space,
	due to unaccounted changes (e.g. to the MOS).
	It also limits the worst-case time to allocate space.
	If we have less than this amount of free space,
	most ZPL operations (e.g. write, create) will return
	.Sy ENOSPC .
	.
	.It Sy spa_upgrade_errlog_limit Ns = Ns Sy 0 Pq uint
	Limits the number of on-disk error log entries that will be converted to the
	new format when enabling the
	.Sy head_errlog
	feature.
	The default is to convert all log entries.
	.
	-.It Sy vdev_removal_max_span Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq int
	+.It Sy vdev_removal_max_span Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	During top-level vdev removal, chunks of data are copied from the vdev
	which may include free space in order to trade bandwidth for IOPS.
	This parameter determines the maximum span of free space, in bytes,
	which will be included as "unnecessary" data in a chunk of copied data.
	.Pp
	The default value here was chosen to align with
	.Sy zfs_vdev_read_gap_limit ,
	which is a similar concept when doing
	regular reads (but there's no reason it has to be the same).
	.
	.It Sy vdev_file_logical_ashift Ns = Ns Sy 9 Po 512 B Pc Pq ulong
	Logical ashift for file-based devices.
	.
	.It Sy vdev_file_physical_ashift Ns = Ns Sy 9 Po 512 B Pc Pq ulong
	Physical ashift for file-based devices.
	.
	.It Sy zap_iterate_prefetch Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If set, when we start iterating over a ZAP object,
	prefetch the entire object (all leaf blocks).
	However, this is limited by
	.Sy dmu_prefetch_max .
	.
	.It Sy zfetch_array_rd_sz Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
	If prefetching is enabled, disable prefetching for reads larger than this size.
	.
	.It Sy zfetch_min_distance Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
	Min bytes to prefetch per stream.
	Prefetch distance starts from the demand access size and quickly grows to
	this value, doubling on each hit.
	After that it may grow further by 1/8 per hit, but only if some prefetch
	since last time haven't completed in time to satisfy demand request, i.e.
	prefetch depth didn't cover the read latency or the pool got saturated.
	.
	.It Sy zfetch_max_distance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
	Max bytes to prefetch per stream.
	.
	.It Sy zfetch_max_idistance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
	Max bytes to prefetch indirects for per stream.
	.
	.It Sy zfetch_max_streams Ns = Ns Sy 8 Pq uint
	Max number of streams per zfetch (prefetch streams per file).
	.
	.It Sy zfetch_min_sec_reap Ns = Ns Sy 1 Pq uint
	Min time before inactive prefetch stream can be reclaimed
	.
	.It Sy zfetch_max_sec_reap Ns = Ns Sy 2 Pq uint
	Max time before inactive prefetch stream can be deleted
	.
	.It Sy zfs_abd_scatter_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enables ARC from using scatter/gather lists and forces all allocations to be
	linear in kernel memory.
	Disabling can improve performance in some code paths
	at the expense of fragmented kernel memory.
	.
	.It Sy zfs_abd_scatter_max_order Ns = Ns Sy MAX_ORDER\-1 Pq uint
	Maximum number of consecutive memory pages allocated in a single block for
	scatter/gather lists.
	.Pp
	The value of
	.Sy MAX_ORDER
	depends on kernel configuration.
	.
	.It Sy zfs_abd_scatter_min_size Ns = Ns Sy 1536 Ns B Po 1.5 KiB Pc Pq uint
	This is the minimum allocation size that will use scatter (page-based) ABDs.
	Smaller allocations will use linear ABDs.
	.
	.It Sy zfs_arc_dnode_limit Ns = Ns Sy 0 Ns B Pq ulong
	When the number of bytes consumed by dnodes in the ARC exceeds this number of
	bytes, try to unpin some of it in response to demand for non-metadata.
	This value acts as a ceiling to the amount of dnode metadata, and defaults to
	.Sy 0 ,
	which indicates that a percent which is based on
	.Sy zfs_arc_dnode_limit_percent
	of the ARC meta buffers that may be used for dnodes.
	.Pp
	Also see
	.Sy zfs_arc_meta_prune
	which serves a similar purpose but is used
	when the amount of metadata in the ARC exceeds
	.Sy zfs_arc_meta_limit
	rather than in response to overall demand for non-metadata.
	.
	.It Sy zfs_arc_dnode_limit_percent Ns = Ns Sy 10 Ns % Pq ulong
	Percentage that can be consumed by dnodes of ARC meta buffers.
	.Pp
	See also
	.Sy zfs_arc_dnode_limit ,
	which serves a similar purpose but has a higher priority if nonzero.
	.
	.It Sy zfs_arc_dnode_reduce_percent Ns = Ns Sy 10 Ns % Pq ulong
	Percentage of ARC dnodes to try to scan in response to demand for non-metadata
	when the number of bytes consumed by dnodes exceeds
	.Sy zfs_arc_dnode_limit .
	.
	-.It Sy zfs_arc_average_blocksize Ns = Ns Sy 8192 Ns B Po 8 KiB Pc Pq int
	+.It Sy zfs_arc_average_blocksize Ns = Ns Sy 8192 Ns B Po 8 KiB Pc Pq uint
	The ARC's buffer hash table is sized based on the assumption of an average
	block size of this value.
	This works out to roughly 1 MiB of hash table per 1 GiB of physical memory
	with 8-byte pointers.
	For configurations with a known larger average block size,
	this value can be increased to reduce the memory footprint.
	.
	-.It Sy zfs_arc_eviction_pct Ns = Ns Sy 200 Ns % Pq int
	+.It Sy zfs_arc_eviction_pct Ns = Ns Sy 200 Ns % Pq uint
	When
	.Fn arc_is_overflowing ,
	.Fn arc_get_data_impl
	waits for this percent of the requested amount of data to be evicted.
	For example, by default, for every
	.Em 2 KiB
	that's evicted,
	.Em 1 KiB
	of it may be "reused" by a new allocation.
	Since this is above
	.Sy 100 Ns % ,
	it ensures that progress is made towards getting
	.Sy arc_size No under Sy arc_c .
	Since this is finite, it ensures that allocations can still happen,
	even during the potentially long time that
	.Sy arc_size No is more than Sy arc_c .
	.
	-.It Sy zfs_arc_evict_batch_limit Ns = Ns Sy 10 Pq int
	+.It Sy zfs_arc_evict_batch_limit Ns = Ns Sy 10 Pq uint
	Number ARC headers to evict per sub-list before proceeding to another sub-list.
	This batch-style operation prevents entire sub-lists from being evicted at once
	but comes at a cost of additional unlocking and locking.
	.
	-.It Sy zfs_arc_grow_retry Ns = Ns Sy 0 Ns s Pq int
	+.It Sy zfs_arc_grow_retry Ns = Ns Sy 0 Ns s Pq uint
	If set to a non zero value, it will replace the
	.Sy arc_grow_retry
	value with this value.
	The
	.Sy arc_grow_retry
	.No value Pq default Sy 5 Ns s
	is the number of seconds the ARC will wait before
	trying to resume growth after a memory pressure event.
	.
	.It Sy zfs_arc_lotsfree_percent Ns = Ns Sy 10 Ns % Pq int
	Throttle I/O when free system memory drops below this percentage of total
	system memory.
	Setting this value to
	.Sy 0
	will disable the throttle.
	.
	.It Sy zfs_arc_max Ns = Ns Sy 0 Ns B Pq ulong
	Max size of ARC in bytes.
	If
	.Sy 0 ,
	then the max size of ARC is determined by the amount of system memory installed.
	Under Linux, half of system memory will be used as the limit.
	Under
	.Fx ,
	the larger of
	.Sy all_system_memory No \- Sy 1 GiB
	and
	.Sy 5/8 No \(mu Sy all_system_memory
	will be used as the limit.
	This value must be at least
	.Sy 67108864 Ns B Pq 64 MiB .
	.Pp
	This value can be changed dynamically, with some caveats.
	It cannot be set back to
	.Sy 0
	while running, and reducing it below the current ARC size will not cause
	the ARC to shrink without memory pressure to induce shrinking.
	.
	-.It Sy zfs_arc_meta_adjust_restarts Ns = Ns Sy 4096 Pq ulong
	+.It Sy zfs_arc_meta_adjust_restarts Ns = Ns Sy 4096 Pq uint
	The number of restart passes to make while scanning the ARC attempting
	the free buffers in order to stay below the
	.Sy fs_arc_meta_limit .
	This value should not need to be tuned but is available to facilitate
	performance analysis.
	.
	.It Sy zfs_arc_meta_limit Ns = Ns Sy 0 Ns B Pq ulong
	The maximum allowed size in bytes that metadata buffers are allowed to
	consume in the ARC.
	When this limit is reached, metadata buffers will be reclaimed,
	even if the overall
	.Sy arc_c_max
	has not been reached.
	It defaults to
	.Sy 0 ,
	which indicates that a percentage based on
	.Sy zfs_arc_meta_limit_percent
	of the ARC may be used for metadata.
	.Pp
	This value my be changed dynamically, except that must be set to an explicit value
	.Pq cannot be set back to Sy 0 .
	.
	.It Sy zfs_arc_meta_limit_percent Ns = Ns Sy 75 Ns % Pq ulong
	Percentage of ARC buffers that can be used for metadata.
	.Pp
	See also
	.Sy zfs_arc_meta_limit ,
	which serves a similar purpose but has a higher priority if nonzero.
	.
	.It Sy zfs_arc_meta_min Ns = Ns Sy 0 Ns B Pq ulong
	The minimum allowed size in bytes that metadata buffers may consume in
	the ARC.
	.
	.It Sy zfs_arc_meta_prune Ns = Ns Sy 10000 Pq int
	The number of dentries and inodes to be scanned looking for entries
	which can be dropped.
	This may be required when the ARC reaches the
	.Sy zfs_arc_meta_limit
	because dentries and inodes can pin buffers in the ARC.
	Increasing this value will cause to dentry and inode caches
	to be pruned more aggressively.
	Setting this value to
	.Sy 0
	will disable pruning the inode and dentry caches.
	.
	-.It Sy zfs_arc_meta_strategy Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	+.It Sy zfs_arc_meta_strategy Ns = Ns Sy 1 Ns \| Ns 0 Pq uint
	Define the strategy for ARC metadata buffer eviction (meta reclaim strategy):
	.Bl -tag -compact -offset 4n -width "0 (META_ONLY)"
	.It Sy 0 Pq META_ONLY
	evict only the ARC metadata buffers
	.It Sy 1 Pq BALANCED
	additional data buffers may be evicted if required
	to evict the required number of metadata buffers.
	.El
	.
	.It Sy zfs_arc_min Ns = Ns Sy 0 Ns B Pq ulong
	Min size of ARC in bytes.
	.No If set to Sy 0 , arc_c_min
	will default to consuming the larger of
	.Sy 32 MiB
	and
	.Sy all_system_memory No / Sy 32 .
	.
	-.It Sy zfs_arc_min_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 1s Pc Pq int
	+.It Sy zfs_arc_min_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 1s Pc Pq uint
	Minimum time prefetched blocks are locked in the ARC.
	.
	-.It Sy zfs_arc_min_prescient_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 6s Pc Pq int
	+.It Sy zfs_arc_min_prescient_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 6s Pc Pq uint
	Minimum time "prescient prefetched" blocks are locked in the ARC.
	These blocks are meant to be prefetched fairly aggressively ahead of
	the code that may use them.
	.
	.It Sy zfs_arc_prune_task_threads Ns = Ns Sy 1 Pq int
	Number of arc_prune threads.
	.Fx
	does not need more than one.
	Linux may theoretically use one per mount point up to number of CPUs,
	but that was not proven to be useful.
	.
	.It Sy zfs_max_missing_tvds Ns = Ns Sy 0 Pq int
	Number of missing top-level vdevs which will be allowed during
	pool import (only in read-only mode).
	.
	.It Sy zfs_max_nvlist_src_size Ns = Sy 0 Pq ulong
	Maximum size in bytes allowed to be passed as
	.Sy zc_nvlist_src_size
	for ioctls on
	.Pa /dev/zfs .
	This prevents a user from causing the kernel to allocate
	an excessive amount of memory.
	When the limit is exceeded, the ioctl fails with
	.Sy EINVAL
	and a description of the error is sent to the
	.Pa zfs-dbgmsg
	log.
	This parameter should not need to be touched under normal circumstances.
	If
	.Sy 0 ,
	equivalent to a quarter of the user-wired memory limit under
	.Fx
	and to
	.Sy 134217728 Ns B Pq 128 MiB
	under Linux.
	.
	-.It Sy zfs_multilist_num_sublists Ns = Ns Sy 0 Pq int
	+.It Sy zfs_multilist_num_sublists Ns = Ns Sy 0 Pq uint
	To allow more fine-grained locking, each ARC state contains a series
	of lists for both data and metadata objects.
	Locking is performed at the level of these "sub-lists".
	This parameters controls the number of sub-lists per ARC state,
	and also applies to other uses of the multilist data structure.
	.Pp
	If
	.Sy 0 ,
	equivalent to the greater of the number of online CPUs and
	.Sy 4 .
	.
	.It Sy zfs_arc_overflow_shift Ns = Ns Sy 8 Pq int
	The ARC size is considered to be overflowing if it exceeds the current
	ARC target size
	.Pq Sy arc_c
	by thresholds determined by this parameter.
	Exceeding by
	.Sy ( arc_c No >> Sy zfs_arc_overflow_shift ) No / Sy 2
	starts ARC reclamation process.
	If that appears insufficient, exceeding by
	.Sy ( arc_c No >> Sy zfs_arc_overflow_shift ) No \(mu Sy 1.5
	blocks new buffer allocation until the reclaim thread catches up.
	Started reclamation process continues till ARC size returns below the
	target size.
	.Pp
	The default value of
	.Sy 8
	causes the ARC to start reclamation if it exceeds the target size by
	.Em 0.2%
	of the target size, and block allocations by
	.Em 0.6% .
	.
	-.It Sy zfs_arc_p_min_shift Ns = Ns Sy 0 Pq int
	+.It Sy zfs_arc_p_min_shift Ns = Ns Sy 0 Pq uint
	If nonzero, this will update
	.Sy arc_p_min_shift Pq default Sy 4
	with the new value.
	.Sy arc_p_min_shift No is used as a shift of Sy arc_c
	when calculating the minumum
	.Sy arc_p No size.
	.
	.It Sy zfs_arc_p_dampener_disable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Disable
	.Sy arc_p
	adapt dampener, which reduces the maximum single adjustment to
	.Sy arc_p .
	.
	-.It Sy zfs_arc_shrink_shift Ns = Ns Sy 0 Pq int
	+.It Sy zfs_arc_shrink_shift Ns = Ns Sy 0 Pq uint
	If nonzero, this will update
	.Sy arc_shrink_shift Pq default Sy 7
	with the new value.
	.
	.It Sy zfs_arc_pc_percent Ns = Ns Sy 0 Ns % Po off Pc Pq uint
	Percent of pagecache to reclaim ARC to.
	.Pp
	This tunable allows the ZFS ARC to play more nicely
	with the kernel's LRU pagecache.
	It can guarantee that the ARC size won't collapse under scanning
	pressure on the pagecache, yet still allows the ARC to be reclaimed down to
	.Sy zfs_arc_min
	if necessary.
	This value is specified as percent of pagecache size (as measured by
	.Sy NR_FILE_PAGES ) ,
	where that percent may exceed
	.Sy 100 .
	This
	only operates during memory pressure/reclaim.
	.
	.It Sy zfs_arc_shrinker_limit Ns = Ns Sy 10000 Pq int
	This is a limit on how many pages the ARC shrinker makes available for
	eviction in response to one page allocation attempt.
	Note that in practice, the kernel's shrinker can ask us to evict
	up to about four times this for one allocation attempt.
	.Pp
	The default limit of
	.Sy 10000 Pq in practice, Em 160 MiB No per allocation attempt with 4 KiB pages
	limits the amount of time spent attempting to reclaim ARC memory to
	less than 100 ms per allocation attempt,
	even with a small average compressed block size of ~8 KiB.
	.Pp
	The parameter can be set to 0 (zero) to disable the limit,
	and only applies on Linux.
	.
	.It Sy zfs_arc_sys_free Ns = Ns Sy 0 Ns B Pq ulong
	The target number of bytes the ARC should leave as free memory on the system.
	If zero, equivalent to the bigger of
	.Sy 512 KiB No and Sy all_system_memory/64 .
	.
	.It Sy zfs_autoimport_disable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Disable pool import at module load by ignoring the cache file
	.Pq Sy spa_config_path .
	.
	.It Sy zfs_checksum_events_per_second Ns = Ns Sy 20 Ns /s Pq uint
	Rate limit checksum events to this many per second.
	Note that this should not be set below the ZED thresholds
	(currently 10 checksums over 10 seconds)
	or else the daemon may not trigger any action.
	.
	-.It Sy zfs_commit_timeout_pct Ns = Ns Sy 5 Ns % Pq int
	+.It Sy zfs_commit_timeout_pct Ns = Ns Sy 5 Ns % Pq uint
	This controls the amount of time that a ZIL block (lwb) will remain "open"
	when it isn't "full", and it has a thread waiting for it to be committed to
	stable storage.
	The timeout is scaled based on a percentage of the last lwb
	latency to avoid significantly impacting the latency of each individual
	transaction record (itx).
	.
	.It Sy zfs_condense_indirect_commit_entry_delay_ms Ns = Ns Sy 0 Ns ms Pq int
	Vdev indirection layer (used for device removal) sleeps for this many
	milliseconds during mapping generation.
	Intended for use with the test suite to throttle vdev removal speed.
	.
	-.It Sy zfs_condense_indirect_obsolete_pct Ns = Ns Sy 25 Ns % Pq int
	+.It Sy zfs_condense_indirect_obsolete_pct Ns = Ns Sy 25 Ns % Pq uint
	Minimum percent of obsolete bytes in vdev mapping required to attempt to condense
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	Intended for use with the test suite
	to facilitate triggering condensing as needed.
	.
	.It Sy zfs_condense_indirect_vdevs_enable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable condensing indirect vdev mappings.
	When set, attempt to condense indirect vdev mappings
	if the mapping uses more than
	.Sy zfs_condense_min_mapping_bytes
	bytes of memory and if the obsolete space map object uses more than
	.Sy zfs_condense_max_obsolete_bytes
	bytes on-disk.
	The condensing process is an attempt to save memory by removing obsolete mappings.
	.
	.It Sy zfs_condense_max_obsolete_bytes Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq ulong
	Only attempt to condense indirect vdev mappings if the on-disk size
	of the obsolete space map object is greater than this number of bytes
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	.
	.It Sy zfs_condense_min_mapping_bytes Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq ulong
	Minimum size vdev mapping to attempt to condense
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	.
	.It Sy zfs_dbgmsg_enable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Internally ZFS keeps a small log to facilitate debugging.
	The log is enabled by default, and can be disabled by unsetting this option.
	The contents of the log can be accessed by reading
	.Pa /proc/spl/kstat/zfs/dbgmsg .
	Writing
	.Sy 0
	to the file clears the log.
	.Pp
	This setting does not influence debug prints due to
	.Sy zfs_flags .
	.
	-.It Sy zfs_dbgmsg_maxsize Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq int
	+.It Sy zfs_dbgmsg_maxsize Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
	Maximum size of the internal ZFS debug log.
	.
	.It Sy zfs_dbuf_state_index Ns = Ns Sy 0 Pq int
	Historically used for controlling what reporting was available under
	.Pa /proc/spl/kstat/zfs .
	No effect.
	.
	.It Sy zfs_deadman_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	When a pool sync operation takes longer than
	.Sy zfs_deadman_synctime_ms ,
	or when an individual I/O operation takes longer than
	.Sy zfs_deadman_ziotime_ms ,
	then the operation is considered to be "hung".
	If
	.Sy zfs_deadman_enabled
	is set, then the deadman behavior is invoked as described by
	.Sy zfs_deadman_failmode .
	By default, the deadman is enabled and set to
	.Sy wait
	which results in "hung" I/O operations only being logged.
	The deadman is automatically disabled when a pool gets suspended.
	.
	.It Sy zfs_deadman_failmode Ns = Ns Sy wait Pq charp
	Controls the failure behavior when the deadman detects a "hung" I/O operation.
	Valid values are:
	.Bl -tag -compact -offset 4n -width "continue"
	.It Sy wait
	Wait for a "hung" operation to complete.
	For each "hung" operation a "deadman" event will be posted
	describing that operation.
	.It Sy continue
	Attempt to recover from a "hung" operation by re-dispatching it
	to the I/O pipeline if possible.
	.It Sy panic
	Panic the system.
	This can be used to facilitate automatic fail-over
	to a properly configured fail-over partner.
	.El
	.
	.It Sy zfs_deadman_checktime_ms Ns = Ns Sy 60000 Ns ms Po 1 min Pc Pq int
	Check time in milliseconds.
	This defines the frequency at which we check for hung I/O requests
	and potentially invoke the
	.Sy zfs_deadman_failmode
	behavior.
	.
	.It Sy zfs_deadman_synctime_ms Ns = Ns Sy 600000 Ns ms Po 10 min Pc Pq ulong
	Interval in milliseconds after which the deadman is triggered and also
	the interval after which a pool sync operation is considered to be "hung".
	Once this limit is exceeded the deadman will be invoked every
	.Sy zfs_deadman_checktime_ms
	milliseconds until the pool sync completes.
	.
	.It Sy zfs_deadman_ziotime_ms Ns = Ns Sy 300000 Ns ms Po 5 min Pc Pq ulong
	Interval in milliseconds after which the deadman is triggered and an
	individual I/O operation is considered to be "hung".
	As long as the operation remains "hung",
	the deadman will be invoked every
	.Sy zfs_deadman_checktime_ms
	milliseconds until the operation completes.
	.
	.It Sy zfs_dedup_prefetch Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Enable prefetching dedup-ed blocks which are going to be freed.
	.
	-.It Sy zfs_delay_min_dirty_percent Ns = Ns Sy 60 Ns % Pq int
	+.It Sy zfs_delay_min_dirty_percent Ns = Ns Sy 60 Ns % Pq uint
	Start to delay each transaction once there is this amount of dirty data,
	expressed as a percentage of
	.Sy zfs_dirty_data_max .
	This value should be at least
	.Sy zfs_vdev_async_write_active_max_dirty_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.
	.It Sy zfs_delay_scale Ns = Ns Sy 500000 Pq int
	This controls how quickly the transaction delay approaches infinity.
	Larger values cause longer delays for a given amount of dirty data.
	.Pp
	For the smoothest delay, this value should be about 1 billion divided
	by the maximum number of operations per second.
	This will smoothly handle between ten times and a tenth of this number.
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	.Sy zfs_delay_scale No \(mu Sy zfs_dirty_data_max Em must No be smaller than Sy 2^64 .
	.
	.It Sy zfs_disable_ivset_guid_check Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disables requirement for IVset GUIDs to be present and match when doing a raw
	receive of encrypted datasets.
	Intended for users whose pools were created with
	OpenZFS pre-release versions and now have compatibility issues.
	.
	.It Sy zfs_key_max_salt_uses Ns = Ns Sy 400000000 Po 4*10^8 Pc Pq ulong
	Maximum number of uses of a single salt value before generating a new one for
	encrypted datasets.
	The default value is also the maximum.
	.
	.It Sy zfs_object_mutex_size Ns = Ns Sy 64 Pq uint
	Size of the znode hashtable used for holds.
	.Pp
	Due to the need to hold locks on objects that may not exist yet, kernel mutexes
	are not created per-object and instead a hashtable is used where collisions
	will result in objects waiting when there is not actually contention on the
	same object.
	.
	.It Sy zfs_slow_io_events_per_second Ns = Ns Sy 20 Ns /s Pq int
	Rate limit delay and deadman zevents (which report slow I/O operations) to this many per
	second.
	.
	.It Sy zfs_unflushed_max_mem_amt Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq ulong
	Upper-bound limit for unflushed metadata changes to be held by the
	log spacemap in memory, in bytes.
	.
	.It Sy zfs_unflushed_max_mem_ppm Ns = Ns Sy 1000 Ns ppm Po 0.1% Pc Pq ulong
	Part of overall system memory that ZFS allows to be used
	for unflushed metadata changes by the log spacemap, in millionths.
	.
	.It Sy zfs_unflushed_log_block_max Ns = Ns Sy 131072 Po 128k Pc Pq ulong
	Describes the maximum number of log spacemap blocks allowed for each pool.
	The default value means that the space in all the log spacemaps
	can add up to no more than
	.Sy 131072
	blocks (which means
	.Em 16 GiB
	of logical space before compression and ditto blocks,
	assuming that blocksize is
	.Em 128 KiB ) .
	.Pp
	This tunable is important because it involves a trade-off between import
	time after an unclean export and the frequency of flushing metaslabs.
	The higher this number is, the more log blocks we allow when the pool is
	active which means that we flush metaslabs less often and thus decrease
	the number of I/O operations for spacemap updates per TXG.
	At the same time though, that means that in the event of an unclean export,
	there will be more log spacemap blocks for us to read, inducing overhead
	in the import time of the pool.
	The lower the number, the amount of flushing increases, destroying log
	blocks quicker as they become obsolete faster, which leaves less blocks
	to be read during import time after a crash.
	.Pp
	Each log spacemap block existing during pool import leads to approximately
	one extra logical I/O issued.
	This is the reason why this tunable is exposed in terms of blocks rather
	than space used.
	.
	.It Sy zfs_unflushed_log_block_min Ns = Ns Sy 1000 Pq ulong
	If the number of metaslabs is small and our incoming rate is high,
	we could get into a situation that we are flushing all our metaslabs every TXG.
	Thus we always allow at least this many log blocks.
	.
	.It Sy zfs_unflushed_log_block_pct Ns = Ns Sy 400 Ns % Pq ulong
	Tunable used to determine the number of blocks that can be used for
	the spacemap log, expressed as a percentage of the total number of
	unflushed metaslabs in the pool.
	.
	.It Sy zfs_unflushed_log_txg_max Ns = Ns Sy 1000 Pq ulong
	Tunable limiting maximum time in TXGs any metaslab may remain unflushed.
	It effectively limits maximum number of unflushed per-TXG spacemap logs
	that need to be read after unclean pool export.
	.
	.It Sy zfs_unlink_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	When enabled, files will not be asynchronously removed from the list of pending
	unlinks and the space they consume will be leaked.
	Once this option has been disabled and the dataset is remounted,
	the pending unlinks will be processed and the freed space returned to the pool.
	This option is used by the test suite.
	.
	.It Sy zfs_delete_blocks Ns = Ns Sy 20480 Pq ulong
	This is the used to define a large file for the purposes of deletion.
	Files containing more than
	.Sy zfs_delete_blocks
	will be deleted asynchronously, while smaller files are deleted synchronously.
	Decreasing this value will reduce the time spent in an
	.Xr unlink 2
	system call, at the expense of a longer delay before the freed space is available.
	.
	.It Sy zfs_dirty_data_max Ns = Pq int
	Determines the dirty space limit in bytes.
	Once this limit is exceeded, new writes are halted until space frees up.
	This parameter takes precedence over
	.Sy zfs_dirty_data_max_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Defaults to
	.Sy physical_ram/10 ,
	capped at
	.Sy zfs_dirty_data_max_max .
	.
	.It Sy zfs_dirty_data_max_max Ns = Pq int
	Maximum allowable value of
	.Sy zfs_dirty_data_max ,
	expressed in bytes.
	This limit is only enforced at module load time, and will be ignored if
	.Sy zfs_dirty_data_max
	is later changed.
	This parameter takes precedence over
	.Sy zfs_dirty_data_max_max_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Defaults to
	.Sy physical_ram/4 ,
	.
	-.It Sy zfs_dirty_data_max_max_percent Ns = Ns Sy 25 Ns % Pq int
	+.It Sy zfs_dirty_data_max_max_percent Ns = Ns Sy 25 Ns % Pq uint
	Maximum allowable value of
	.Sy zfs_dirty_data_max ,
	expressed as a percentage of physical RAM.
	This limit is only enforced at module load time, and will be ignored if
	.Sy zfs_dirty_data_max
	is later changed.
	The parameter
	.Sy zfs_dirty_data_max_max
	takes precedence over this one.
	.No See Sx ZFS TRANSACTION DELAY .
	.
	-.It Sy zfs_dirty_data_max_percent Ns = Ns Sy 10 Ns % Pq int
	+.It Sy zfs_dirty_data_max_percent Ns = Ns Sy 10 Ns % Pq uint
	Determines the dirty space limit, expressed as a percentage of all memory.
	Once this limit is exceeded, new writes are halted until space frees up.
	The parameter
	.Sy zfs_dirty_data_max
	takes precedence over this one.
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Subject to
	.Sy zfs_dirty_data_max_max .
	.
	-.It Sy zfs_dirty_data_sync_percent Ns = Ns Sy 20 Ns % Pq int
	+.It Sy zfs_dirty_data_sync_percent Ns = Ns Sy 20 Ns % Pq uint
	Start syncing out a transaction group if there's at least this much dirty data
	.Pq as a percentage of Sy zfs_dirty_data_max .
	This should be less than
	.Sy zfs_vdev_async_write_active_min_dirty_percent .
	.
	.It Sy zfs_wrlog_data_max Ns = Pq int
	The upper limit of write-transaction zil log data size in bytes.
	Write operations are throttled when approaching the limit until log data is
	cleared out after transaction group sync.
	Because of some overhead, it should be set at least 2 times the size of
	.Sy zfs_dirty_data_max
	.No to prevent harming normal write throughput.
	It also should be smaller than the size of the slog device if slog is present.
	.Pp
	Defaults to
	.Sy zfs_dirty_data_max*2
	.
	.It Sy zfs_fallocate_reserve_percent Ns = Ns Sy 110 Ns % Pq uint
	Since ZFS is a copy-on-write filesystem with snapshots, blocks cannot be
	preallocated for a file in order to guarantee that later writes will not
	run out of space.
	Instead,
	.Xr fallocate 2
	space preallocation only checks that sufficient space is currently available
	in the pool or the user's project quota allocation,
	and then creates a sparse file of the requested size.
	The requested space is multiplied by
	.Sy zfs_fallocate_reserve_percent
	to allow additional space for indirect blocks and other internal metadata.
	Setting this to
	.Sy 0
	disables support for
	.Xr fallocate 2
	and causes it to return
	.Sy EOPNOTSUPP .
	.
	.It Sy zfs_fletcher_4_impl Ns = Ns Sy fastest Pq string
	Select a fletcher 4 implementation.
	.Pp
	Supported selectors are:
	.Sy fastest , scalar , sse2 , ssse3 , avx2 , avx512f , avx512bw ,
	.No and Sy aarch64_neon .
	All except
	.Sy fastest No and Sy scalar
	require instruction set extensions to be available,
	and will only appear if ZFS detects that they are present at runtime.
	If multiple implementations of fletcher 4 are available, the
	.Sy fastest
	will be chosen using a micro benchmark.
	Selecting
	.Sy scalar
	results in the original CPU-based calculation being used.
	Selecting any option other than
	.Sy fastest No or Sy scalar
	results in vector instructions
	from the respective CPU instruction set being used.
	.
	.It Sy zfs_blake3_impl Ns = Ns Sy fastest Pq string
	Select a BLAKE3 implementation.
	.Pp
	Supported selectors are:
	.Sy cycle , fastest , generic , sse2 , sse41 , avx2 , avx512 .
	All except
	.Sy cycle , fastest No and Sy generic
	require instruction set extensions to be available,
	and will only appear if ZFS detects that they are present at runtime.
	If multiple implementations of BLAKE3 are available, the
	.Sy fastest will be chosen using a micro benchmark. You can see the
	benchmark results by reading this kstat file:
	.Pa /proc/spl/kstat/zfs/chksum_bench .
	.
	.It Sy zfs_free_bpobj_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable/disable the processing of the free_bpobj object.
	.
	.It Sy zfs_async_block_max_blocks Ns = Ns Sy ULONG_MAX Po unlimited Pc Pq ulong
	Maximum number of blocks freed in a single TXG.
	.
	.It Sy zfs_max_async_dedup_frees Ns = Ns Sy 100000 Po 10^5 Pc Pq ulong
	Maximum number of dedup blocks freed in a single TXG.
	.
	-.It Sy zfs_vdev_async_read_max_active Ns = Ns Sy 3 Pq int
	+.It Sy zfs_vdev_async_read_max_active Ns = Ns Sy 3 Pq uint
	Maximum asynchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_async_read_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_async_read_min_active Ns = Ns Sy 1 Pq uint
	Minimum asynchronous read I/O operation active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_async_write_active_max_dirty_percent Ns = Ns Sy 60 Ns % Pq int
	+.It Sy zfs_vdev_async_write_active_max_dirty_percent Ns = Ns Sy 60 Ns % Pq uint
	When the pool has more than this much dirty data, use
	.Sy zfs_vdev_async_write_max_active
	to limit active async writes.
	If the dirty data is between the minimum and maximum,
	the active I/O limit is linearly interpolated.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_async_write_active_min_dirty_percent Ns = Ns Sy 30 Ns % Pq int
	+.It Sy zfs_vdev_async_write_active_min_dirty_percent Ns = Ns Sy 30 Ns % Pq uint
	When the pool has less than this much dirty data, use
	.Sy zfs_vdev_async_write_min_active
	to limit active async writes.
	If the dirty data is between the minimum and maximum,
	the active I/O limit is linearly
	interpolated.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_async_write_max_active Ns = Ns Sy 30 Pq int
	+.It Sy zfs_vdev_async_write_max_active Ns = Ns Sy 30 Pq uint
	Maximum asynchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_async_write_min_active Ns = Ns Sy 2 Pq int
	+.It Sy zfs_vdev_async_write_min_active Ns = Ns Sy 2 Pq uint
	Minimum asynchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.Pp
	Lower values are associated with better latency on rotational media but poorer
	resilver performance.
	The default value of
	.Sy 2
	was chosen as a compromise.
	A value of
	.Sy 3
	has been shown to improve resilver performance further at a cost of
	further increasing latency.
	.
	-.It Sy zfs_vdev_initializing_max_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_initializing_max_active Ns = Ns Sy 1 Pq uint
	Maximum initializing I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_initializing_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_initializing_min_active Ns = Ns Sy 1 Pq uint
	Minimum initializing I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_max_active Ns = Ns Sy 1000 Pq int
	+.It Sy zfs_vdev_max_active Ns = Ns Sy 1000 Pq uint
	The maximum number of I/O operations active to each device.
	Ideally, this will be at least the sum of each queue's
	.Sy max_active .
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_rebuild_max_active Ns = Ns Sy 3 Pq int
	+.It Sy zfs_vdev_rebuild_max_active Ns = Ns Sy 3 Pq uint
	Maximum sequential resilver I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_rebuild_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_rebuild_min_active Ns = Ns Sy 1 Pq uint
	Minimum sequential resilver I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_removal_max_active Ns = Ns Sy 2 Pq int
	+.It Sy zfs_vdev_removal_max_active Ns = Ns Sy 2 Pq uint
	Maximum removal I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_removal_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_removal_min_active Ns = Ns Sy 1 Pq uint
	Minimum removal I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_scrub_max_active Ns = Ns Sy 2 Pq int
	+.It Sy zfs_vdev_scrub_max_active Ns = Ns Sy 2 Pq uint
	Maximum scrub I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_scrub_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_scrub_min_active Ns = Ns Sy 1 Pq uint
	Minimum scrub I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_sync_read_max_active Ns = Ns Sy 10 Pq int
	+.It Sy zfs_vdev_sync_read_max_active Ns = Ns Sy 10 Pq uint
	Maximum synchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_sync_read_min_active Ns = Ns Sy 10 Pq int
	+.It Sy zfs_vdev_sync_read_min_active Ns = Ns Sy 10 Pq uint
	Minimum synchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_sync_write_max_active Ns = Ns Sy 10 Pq int
	+.It Sy zfs_vdev_sync_write_max_active Ns = Ns Sy 10 Pq uint
	Maximum synchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_sync_write_min_active Ns = Ns Sy 10 Pq int
	+.It Sy zfs_vdev_sync_write_min_active Ns = Ns Sy 10 Pq uint
	Minimum synchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_trim_max_active Ns = Ns Sy 2 Pq int
	+.It Sy zfs_vdev_trim_max_active Ns = Ns Sy 2 Pq uint
	Maximum trim/discard I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_trim_min_active Ns = Ns Sy 1 Pq int
	+.It Sy zfs_vdev_trim_min_active Ns = Ns Sy 1 Pq uint
	Minimum trim/discard I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_nia_delay Ns = Ns Sy 5 Pq int
	+.It Sy zfs_vdev_nia_delay Ns = Ns Sy 5 Pq uint
	For non-interactive I/O (scrub, resilver, removal, initialize and rebuild),
	the number of concurrently-active I/O operations is limited to
	.Sy zfs_*_min_active ,
	unless the vdev is "idle".
	When there are no interactive I/O operations active (synchronous or otherwise),
	and
	.Sy zfs_vdev_nia_delay
	operations have completed since the last interactive operation,
	then the vdev is considered to be "idle",
	and the number of concurrently-active non-interactive operations is increased to
	.Sy zfs_*_max_active .
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_nia_credit Ns = Ns Sy 5 Pq int
	+.It Sy zfs_vdev_nia_credit Ns = Ns Sy 5 Pq uint
	Some HDDs tend to prioritize sequential I/O so strongly, that concurrent
	random I/O latency reaches several seconds.
	On some HDDs this happens even if sequential I/O operations
	are submitted one at a time, and so setting
	.Sy zfs_*_max_active Ns = Sy 1
	does not help.
	To prevent non-interactive I/O, like scrub,
	from monopolizing the device, no more than
	.Sy zfs_vdev_nia_credit operations can be sent
	while there are outstanding incomplete interactive operations.
	This enforced wait ensures the HDD services the interactive I/O
	within a reasonable amount of time.
	.No See Sx ZFS I/O SCHEDULER .
	.
	-.It Sy zfs_vdev_queue_depth_pct Ns = Ns Sy 1000 Ns % Pq int
	+.It Sy zfs_vdev_queue_depth_pct Ns = Ns Sy 1000 Ns % Pq uint
	Maximum number of queued allocations per top-level vdev expressed as
	a percentage of
	.Sy zfs_vdev_async_write_max_active ,
	which allows the system to detect devices that are more capable
	of handling allocations and to allocate more blocks to those devices.
	This allows for dynamic allocation distribution when devices are imbalanced,
	as fuller devices will tend to be slower than empty devices.
	.Pp
	Also see
	.Sy zio_dva_throttle_enabled .
	.
	.It Sy zfs_expire_snapshot Ns = Ns Sy 300 Ns s Pq int
	Time before expiring
	.Pa .zfs/snapshot .
	.
	.It Sy zfs_admin_snapshot Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow the creation, removal, or renaming of entries in the
	.Sy .zfs/snapshot
	directory to cause the creation, destruction, or renaming of snapshots.
	When enabled, this functionality works both locally and over NFS exports
	which have the
	.Em no_root_squash
	option set.
	.
	.It Sy zfs_flags Ns = Ns Sy 0 Pq int
	Set additional debugging flags.
	The following flags may be bitwise-ored together:
	.TS
	box;
	lbz r l l .
	Value Symbolic Name Description
	_
	1 ZFS_DEBUG_DPRINTF Enable dprintf entries in the debug log.
	* 2 ZFS_DEBUG_DBUF_VERIFY Enable extra dbuf verifications.
	* 4 ZFS_DEBUG_DNODE_VERIFY Enable extra dnode verifications.
	8 ZFS_DEBUG_SNAPNAMES Enable snapshot name verification.
	16 ZFS_DEBUG_MODIFY Check for illegally modified ARC buffers.
	64 ZFS_DEBUG_ZIO_FREE Enable verification of block frees.
	128 ZFS_DEBUG_HISTOGRAM_VERIFY Enable extra spacemap histogram verifications.
	256 ZFS_DEBUG_METASLAB_VERIFY Verify space accounting on disk matches in-memory \fBrange_trees\fP.
	512 ZFS_DEBUG_SET_ERROR Enable \fBSET_ERROR\fP and dprintf entries in the debug log.
	1024 ZFS_DEBUG_INDIRECT_REMAP Verify split blocks created by device removal.
	2048 ZFS_DEBUG_TRIM Verify TRIM ranges are always within the allocatable range tree.
	4096 ZFS_DEBUG_LOG_SPACEMAP Verify that the log summary is consistent with the spacemap log
	and enable \fBzfs_dbgmsgs\fP for metaslab loading and flushing.
	.TE
	.Sy \& * No Requires debug build.
	.
	.It Sy zfs_btree_verify_intensity Ns = Ns Sy 0 Pq uint
	Enables btree verification.
	The following settings are culminative:
	.TS
	box;
	lbz r l l .
	Value Description

	1 Verify height.
	2 Verify pointers from children to parent.
	3 Verify element counts.
	4 Verify element order. (expensive)
	* 5 Verify unused memory is poisoned. (expensive)
	.TE
	.Sy \& * No Requires debug build.
	.
	.It Sy zfs_free_leak_on_eio Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If destroy encounters an
	.Sy EIO
	while reading metadata (e.g. indirect blocks),
	space referenced by the missing metadata can not be freed.
	Normally this causes the background destroy to become "stalled",
	as it is unable to make forward progress.
	While in this stalled state, all remaining space to free
	from the error-encountering filesystem is "temporarily leaked".
	Set this flag to cause it to ignore the
	.Sy EIO ,
	permanently leak the space from indirect blocks that can not be read,
	and continue to free everything else that it can.
	.Pp
	The default "stalling" behavior is useful if the storage partially
	fails (i.e. some but not all I/O operations fail), and then later recovers.
	In this case, we will be able to continue pool operations while it is
	partially failed, and when it recovers, we can continue to free the
	space, with no leaks.
	Note, however, that this case is actually fairly rare.
	.Pp
	Typically pools either
	.Bl -enum -compact -offset 4n -width "1."
	.It
	fail completely (but perhaps temporarily,
	e.g. due to a top-level vdev going offline), or
	.It
	have localized, permanent errors (e.g. disk returns the wrong data
	due to bit flip or firmware bug).
	.El
	In the former case, this setting does not matter because the
	pool will be suspended and the sync thread will not be able to make
	forward progress regardless.
	In the latter, because the error is permanent, the best we can do
	is leak the minimum amount of space,
	which is what setting this flag will do.
	It is therefore reasonable for this flag to normally be set,
	but we chose the more conservative approach of not setting it,
	so that there is no possibility of
	leaking space in the "partial temporary" failure case.
	.
	-.It Sy zfs_free_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1s Pc Pq int
	+.It Sy zfs_free_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1s Pc Pq uint
	During a
	.Nm zfs Cm destroy
	operation using the
	.Sy async_destroy
	feature,
	a minimum of this much time will be spent working on freeing blocks per TXG.
	.
	-.It Sy zfs_obsolete_min_time_ms Ns = Ns Sy 500 Ns ms Pq int
	+.It Sy zfs_obsolete_min_time_ms Ns = Ns Sy 500 Ns ms Pq uint
	Similar to
	.Sy zfs_free_min_time_ms ,
	but for cleanup of old indirection records for removed vdevs.
	.
	.It Sy zfs_immediate_write_sz Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq long
	Largest data block to write to the ZIL.
	Larger blocks will be treated as if the dataset being written to had the
	.Sy logbias Ns = Ns Sy throughput
	property set.
	.
	.It Sy zfs_initialize_value Ns = Ns Sy 16045690984833335022 Po 0xDEADBEEFDEADBEEE Pc Pq ulong
	Pattern written to vdev free space by
	.Xr zpool-initialize 8 .
	.
	.It Sy zfs_initialize_chunk_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
	Size of writes used by
	.Xr zpool-initialize 8 .
	This option is used by the test suite.
	.
	.It Sy zfs_livelist_max_entries Ns = Ns Sy 500000 Po 5*10^5 Pc Pq ulong
	The threshold size (in block pointers) at which we create a new sub-livelist.
	Larger sublists are more costly from a memory perspective but the fewer
	sublists there are, the lower the cost of insertion.
	.
	.It Sy zfs_livelist_min_percent_shared Ns = Ns Sy 75 Ns % Pq int
	If the amount of shared space between a snapshot and its clone drops below
	this threshold, the clone turns off the livelist and reverts to the old
	deletion method.
	This is in place because livelists no long give us a benefit
	once a clone has been overwritten enough.
	.
	.It Sy zfs_livelist_condense_new_alloc Ns = Ns Sy 0 Pq int
	Incremented each time an extra ALLOC blkptr is added to a livelist entry while
	it is being condensed.
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_sync_cancel Ns = Ns Sy 0 Pq int
	Incremented each time livelist condensing is canceled while in
	.Fn spa_livelist_condense_sync .
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_sync_pause Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	When set, the livelist condense process pauses indefinitely before
	executing the synctask \(em
	.Fn spa_livelist_condense_sync .
	This option is used by the test suite to trigger race conditions.
	.
	.It Sy zfs_livelist_condense_zthr_cancel Ns = Ns Sy 0 Pq int
	Incremented each time livelist condensing is canceled while in
	.Fn spa_livelist_condense_cb .
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_zthr_pause Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	When set, the livelist condense process pauses indefinitely before
	executing the open context condensing work in
	.Fn spa_livelist_condense_cb .
	This option is used by the test suite to trigger race conditions.
	.
	.It Sy zfs_lua_max_instrlimit Ns = Ns Sy 100000000 Po 10^8 Pc Pq ulong
	The maximum execution time limit that can be set for a ZFS channel program,
	specified as a number of Lua instructions.
	.
	.It Sy zfs_lua_max_memlimit Ns = Ns Sy 104857600 Po 100 MiB Pc Pq ulong
	The maximum memory limit that can be set for a ZFS channel program, specified
	in bytes.
	.
	.It Sy zfs_max_dataset_nesting Ns = Ns Sy 50 Pq int
	The maximum depth of nested datasets.
	This value can be tuned temporarily to
	fix existing datasets that exceed the predefined limit.
	.
	.It Sy zfs_max_log_walking Ns = Ns Sy 5 Pq ulong
	The number of past TXGs that the flushing algorithm of the log spacemap
	feature uses to estimate incoming log blocks.
	.
	.It Sy zfs_max_logsm_summary_length Ns = Ns Sy 10 Pq ulong
	Maximum number of rows allowed in the summary of the spacemap log.
	.
	-.It Sy zfs_max_recordsize Ns = Ns Sy 16777216 Po 16 MiB Pc Pq int
	+.It Sy zfs_max_recordsize Ns = Ns Sy 16777216 Po 16 MiB Pc Pq uint
	We currently support block sizes from
	.Em 512 Po 512 B Pc No to Em 16777216 Po 16 MiB Pc .
	The benefits of larger blocks, and thus larger I/O,
	need to be weighed against the cost of COWing a giant block to modify one byte.
	Additionally, very large blocks can have an impact on I/O latency,
	and also potentially on the memory allocator.
	Therefore, we formerly forbade creating blocks larger than 1M.
	Larger blocks could be created by changing it,
	and pools with larger blocks can always be imported and used,
	regardless of this setting.
	.
	.It Sy zfs_allow_redacted_dataset_mount Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow datasets received with redacted send/receive to be mounted.
	Normally disabled because these datasets may be missing key data.
	.
	.It Sy zfs_min_metaslabs_to_flush Ns = Ns Sy 1 Pq ulong
	Minimum number of metaslabs to flush per dirty TXG.
	.
	-.It Sy zfs_metaslab_fragmentation_threshold Ns = Ns Sy 70 Ns % Pq int
	+.It Sy zfs_metaslab_fragmentation_threshold Ns = Ns Sy 70 Ns % Pq uint
	Allow metaslabs to keep their active state as long as their fragmentation
	percentage is no more than this value.
	An active metaslab that exceeds this threshold
	will no longer keep its active status allowing better metaslabs to be selected.
	.
	-.It Sy zfs_mg_fragmentation_threshold Ns = Ns Sy 95 Ns % Pq int
	+.It Sy zfs_mg_fragmentation_threshold Ns = Ns Sy 95 Ns % Pq uint
	Metaslab groups are considered eligible for allocations if their
	fragmentation metric (measured as a percentage) is less than or equal to
	this value.
	If a metaslab group exceeds this threshold then it will be
	skipped unless all metaslab groups within the metaslab class have also
	crossed this threshold.
	.
	-.It Sy zfs_mg_noalloc_threshold Ns = Ns Sy 0 Ns % Pq int
	+.It Sy zfs_mg_noalloc_threshold Ns = Ns Sy 0 Ns % Pq uint
	Defines a threshold at which metaslab groups should be eligible for allocations.
	The value is expressed as a percentage of free space
	beyond which a metaslab group is always eligible for allocations.
	If a metaslab group's free space is less than or equal to the
	threshold, the allocator will avoid allocating to that group
	unless all groups in the pool have reached the threshold.
	Once all groups have reached the threshold, all groups are allowed to accept
	allocations.
	The default value of
	.Sy 0
	disables the feature and causes all metaslab groups to be eligible for allocations.
	.Pp
	This parameter allows one to deal with pools having heavily imbalanced
	vdevs such as would be the case when a new vdev has been added.
	Setting the threshold to a non-zero percentage will stop allocations
	from being made to vdevs that aren't filled to the specified percentage
	and allow lesser filled vdevs to acquire more allocations than they
	otherwise would under the old
	.Sy zfs_mg_alloc_failures
	facility.
	.
	.It Sy zfs_ddt_data_is_special Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If enabled, ZFS will place DDT data into the special allocation class.
	.
	.It Sy zfs_user_indirect_is_special Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If enabled, ZFS will place user data indirect blocks
	into the special allocation class.
	.
	-.It Sy zfs_multihost_history Ns = Ns Sy 0 Pq int
	+.It Sy zfs_multihost_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest multihost updates will be available in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /multihost .
	.
	.It Sy zfs_multihost_interval Ns = Ns Sy 1000 Ns ms Po 1 s Pc Pq ulong
	Used to control the frequency of multihost writes which are performed when the
	.Sy multihost
	pool property is on.
	This is one of the factors used to determine the
	length of the activity check during import.
	.Pp
	The multihost write period is
	.Sy zfs_multihost_interval No / Sy leaf-vdevs .
	On average a multihost write will be issued for each leaf vdev
	every
	.Sy zfs_multihost_interval
	milliseconds.
	In practice, the observed period can vary with the I/O load
	and this observed value is the delay which is stored in the uberblock.
	.
	.It Sy zfs_multihost_import_intervals Ns = Ns Sy 20 Pq uint
	Used to control the duration of the activity test on import.
	Smaller values of
	.Sy zfs_multihost_import_intervals
	will reduce the import time but increase
	the risk of failing to detect an active pool.
	The total activity check time is never allowed to drop below one second.
	.Pp
	On import the activity check waits a minimum amount of time determined by
	.Sy zfs_multihost_interval No \(mu Sy zfs_multihost_import_intervals ,
	or the same product computed on the host which last had the pool imported,
	whichever is greater.
	The activity check time may be further extended if the value of MMP
	delay found in the best uberblock indicates actual multihost updates happened
	at longer intervals than
	.Sy zfs_multihost_interval .
	A minimum of
	.Em 100 ms
	is enforced.
	.Pp
	.Sy 0 No is equivalent to Sy 1 .
	.
	.It Sy zfs_multihost_fail_intervals Ns = Ns Sy 10 Pq uint
	Controls the behavior of the pool when multihost write failures or delays are
	detected.
	.Pp
	When
	.Sy 0 ,
	multihost write failures or delays are ignored.
	The failures will still be reported to the ZED which depending on
	its configuration may take action such as suspending the pool or offlining a
	device.
	.Pp
	Otherwise, the pool will be suspended if
	.Sy zfs_multihost_fail_intervals No \(mu Sy zfs_multihost_interval
	milliseconds pass without a successful MMP write.
	This guarantees the activity test will see MMP writes if the pool is imported.
	.Sy 1 No is equivalent to Sy 2 ;
	this is necessary to prevent the pool from being suspended
	due to normal, small I/O latency variations.
	.
	.It Sy zfs_no_scrub_io Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to disable scrub I/O.
	This results in scrubs not actually scrubbing data and
	simply doing a metadata crawl of the pool instead.
	.
	.It Sy zfs_no_scrub_prefetch Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to disable block prefetching for scrubs.
	.
	.It Sy zfs_nocacheflush Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable cache flush operations on disks when writing.
	Setting this will cause pool corruption on power loss
	if a volatile out-of-order write cache is enabled.
	.
	.It Sy zfs_nopwrite_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Allow no-operation writes.
	The occurrence of nopwrites will further depend on other pool properties
	.Pq i.a. the checksumming and compression algorithms .
	.
	.It Sy zfs_dmu_offset_next_sync Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable forcing TXG sync to find holes.
	When enabled forces ZFS to sync data when
	.Sy SEEK_HOLE No or Sy SEEK_DATA
	flags are used allowing holes in a file to be accurately reported.
	When disabled holes will not be reported in recently dirtied files.
	.
	.It Sy zfs_pd_bytes_max Ns = Ns Sy 52428800 Ns B Po 50 MiB Pc Pq int
	The number of bytes which should be prefetched during a pool traversal, like
	.Nm zfs Cm send
	or other data crawling operations.
	.
	-.It Sy zfs_traverse_indirect_prefetch_limit Ns = Ns Sy 32 Pq int
	+.It Sy zfs_traverse_indirect_prefetch_limit Ns = Ns Sy 32 Pq uint
	The number of blocks pointed by indirect (non-L0) block which should be
	prefetched during a pool traversal, like
	.Nm zfs Cm send
	or other data crawling operations.
	.
	.It Sy zfs_per_txg_dirty_frees_percent Ns = Ns Sy 30 Ns % Pq ulong
	Control percentage of dirtied indirect blocks from frees allowed into one TXG.
	After this threshold is crossed, additional frees will wait until the next TXG.
	.Sy 0 No disables this throttle.
	.
	.It Sy zfs_prefetch_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable predictive prefetch.
	Note that it leaves "prescient" prefetch
	.Pq for, e.g., Nm zfs Cm send
	intact.
	Unlike predictive prefetch, prescient prefetch never issues I/O
	that ends up not being needed, so it can't hurt performance.
	.
	.It Sy zfs_qat_checksum_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for SHA256 checksums.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_qat_compress_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for gzip compression.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_qat_encrypt_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for AES-GCM encryption.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_vnops_read_chunk_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq long
	Bytes to read per chunk.
	.
	-.It Sy zfs_read_history Ns = Ns Sy 0 Pq int
	+.It Sy zfs_read_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest reads will be available in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /reads .
	.
	.It Sy zfs_read_history_hits Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Include cache hits in read history
	.
	.It Sy zfs_rebuild_max_segment Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
	Maximum read segment size to issue when sequentially resilvering a
	top-level vdev.
	.
	.It Sy zfs_rebuild_scrub_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Automatically start a pool scrub when the last active sequential resilver
	completes in order to verify the checksums of all blocks which have been
	resilvered.
	This is enabled by default and strongly recommended.
	.
	.It Sy zfs_rebuild_vdev_limit Ns = Ns Sy 33554432 Ns B Po 32 MiB Pc Pq ulong
	Maximum amount of I/O that can be concurrently issued for a sequential
	resilver per leaf device, given in bytes.
	.
	.It Sy zfs_reconstruct_indirect_combinations_max Ns = Ns Sy 4096 Pq int
	If an indirect split block contains more than this many possible unique
	combinations when being reconstructed, consider it too computationally
	expensive to check them all.
	Instead, try at most this many randomly selected
	combinations each time the block is accessed.
	This allows all segment copies to participate fairly
	in the reconstruction when all combinations
	cannot be checked and prevents repeated use of one bad copy.
	.
	.It Sy zfs_recover Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to attempt to recover from fatal errors.
	This should only be used as a last resort,
	as it typically results in leaked space, or worse.
	.
	.It Sy zfs_removal_ignore_errors Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Ignore hard I/O errors during device removal.
	When set, if a device encounters a hard I/O error during the removal process
	the removal will not be cancelled.
	This can result in a normally recoverable block becoming permanently damaged
	and is hence not recommended.
	This should only be used as a last resort when the
	pool cannot be returned to a healthy state prior to removing the device.
	.
	-.It Sy zfs_removal_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	+.It Sy zfs_removal_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	This is used by the test suite so that it can ensure that certain actions
	happen while in the middle of a removal.
	.
	-.It Sy zfs_remove_max_segment Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	+.It Sy zfs_remove_max_segment Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The largest contiguous segment that we will attempt to allocate when removing
	a device.
	If there is a performance problem with attempting to allocate large blocks,
	consider decreasing this.
	The default value is also the maximum.
	.
	.It Sy zfs_resilver_disable_defer Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Ignore the
	.Sy resilver_defer
	feature, causing an operation that would start a resilver to
	immediately restart the one in progress.
	.
	-.It Sy zfs_resilver_min_time_ms Ns = Ns Sy 3000 Ns ms Po 3 s Pc Pq int
	+.It Sy zfs_resilver_min_time_ms Ns = Ns Sy 3000 Ns ms Po 3 s Pc Pq uint
	Resilvers are processed by the sync thread.
	While resilvering, it will spend at least this much time
	working on a resilver between TXG flushes.
	.
	.It Sy zfs_scan_ignore_errors Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If set, remove the DTL (dirty time list) upon completion of a pool scan (scrub),
	even if there were unrepairable errors.
	Intended to be used during pool repair or recovery to
	stop resilvering when the pool is next imported.
	.
	-.It Sy zfs_scrub_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1 s Pc Pq int
	+.It Sy zfs_scrub_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1 s Pc Pq uint
	Scrubs are processed by the sync thread.
	While scrubbing, it will spend at least this much time
	working on a scrub between TXG flushes.
	.
	-.It Sy zfs_scan_checkpoint_intval Ns = Ns Sy 7200 Ns s Po 2 hour Pc Pq int
	+.It Sy zfs_scan_checkpoint_intval Ns = Ns Sy 7200 Ns s Po 2 hour Pc Pq uint
	To preserve progress across reboots, the sequential scan algorithm periodically
	needs to stop metadata scanning and issue all the verification I/O to disk.
	The frequency of this flushing is determined by this tunable.
	.
	-.It Sy zfs_scan_fill_weight Ns = Ns Sy 3 Pq int
	+.It Sy zfs_scan_fill_weight Ns = Ns Sy 3 Pq uint
	This tunable affects how scrub and resilver I/O segments are ordered.
	A higher number indicates that we care more about how filled in a segment is,
	while a lower number indicates we care more about the size of the extent without
	considering the gaps within a segment.
	This value is only tunable upon module insertion.
	Changing the value afterwards will have no effect on scrub or resilver performance.
	.
	-.It Sy zfs_scan_issue_strategy Ns = Ns Sy 0 Pq int
	+.It Sy zfs_scan_issue_strategy Ns = Ns Sy 0 Pq uint
	Determines the order that data will be verified while scrubbing or resilvering:
	.Bl -tag -compact -offset 4n -width "a"
	.It Sy 1
	Data will be verified as sequentially as possible, given the
	amount of memory reserved for scrubbing
	.Pq see Sy zfs_scan_mem_lim_fact .
	This may improve scrub performance if the pool's data is very fragmented.
	.It Sy 2
	The largest mostly-contiguous chunk of found data will be verified first.
	By deferring scrubbing of small segments, we may later find adjacent data
	to coalesce and increase the segment size.
	.It Sy 0
	.No Use strategy Sy 1 No during normal verification
	.No and strategy Sy 2 No while taking a checkpoint.
	.El
	.
	.It Sy zfs_scan_legacy Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If unset, indicates that scrubs and resilvers will gather metadata in
	memory before issuing sequential I/O.
	Otherwise indicates that the legacy algorithm will be used,
	where I/O is initiated as soon as it is discovered.
	Unsetting will not affect scrubs or resilvers that are already in progress.
	.
	.It Sy zfs_scan_max_ext_gap Ns = Ns Sy 2097152 Ns B Po 2 MiB Pc Pq int
	Sets the largest gap in bytes between scrub/resilver I/O operations
	that will still be considered sequential for sorting purposes.
	Changing this value will not
	affect scrubs or resilvers that are already in progress.
	.
	-.It Sy zfs_scan_mem_lim_fact Ns = Ns Sy 20 Ns ^-1 Pq int
	+.It Sy zfs_scan_mem_lim_fact Ns = Ns Sy 20 Ns ^-1 Pq uint
	Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
	This tunable determines the hard limit for I/O sorting memory usage.
	When the hard limit is reached we stop scanning metadata and start issuing
	data verification I/O.
	This is done until we get below the soft limit.
	.
	-.It Sy zfs_scan_mem_lim_soft_fact Ns = Ns Sy 20 Ns ^-1 Pq int
	+.It Sy zfs_scan_mem_lim_soft_fact Ns = Ns Sy 20 Ns ^-1 Pq uint
	The fraction of the hard limit used to determined the soft limit for I/O sorting
	by the sequential scan algorithm.
	When we cross this limit from below no action is taken.
	When we cross this limit from above it is because we are issuing verification I/O.
	In this case (unless the metadata scan is done) we stop issuing verification I/O
	and start scanning metadata again until we get to the hard limit.
	.
	.It Sy zfs_scan_strict_mem_lim Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Enforce tight memory limits on pool scans when a sequential scan is in progress.
	When disabled, the memory limit may be exceeded by fast disks.
	.
	.It Sy zfs_scan_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Freezes a scrub/resilver in progress without actually pausing it.
	Intended for testing/debugging.
	.
	.It Sy zfs_scan_vdev_limit Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq int
	Maximum amount of data that can be concurrently issued at once for scrubs and
	resilvers per leaf device, given in bytes.
	.
	.It Sy zfs_send_corrupt_data Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow sending of corrupt data (ignore read/checksum errors when sending).
	.
	.It Sy zfs_send_unmodified_spill_blocks Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Include unmodified spill blocks in the send stream.
	Under certain circumstances, previous versions of ZFS could incorrectly
	remove the spill block from an existing object.
	Including unmodified copies of the spill blocks creates a backwards-compatible
	stream which will recreate a spill block if it was incorrectly removed.
	.
	-.It Sy zfs_send_no_prefetch_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq int
	+.It Sy zfs_send_no_prefetch_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm send
	internal queues.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	-.It Sy zfs_send_no_prefetch_queue_length Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	+.It Sy zfs_send_no_prefetch_queue_length Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	The maximum number of bytes allowed in
	.Nm zfs Cm send Ns 's
	internal queues.
	.
	-.It Sy zfs_send_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq int
	+.It Sy zfs_send_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm send
	prefetch queue.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	-.It Sy zfs_send_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	+.It Sy zfs_send_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The maximum number of bytes allowed that will be prefetched by
	.Nm zfs Cm send .
	This value must be at least twice the maximum block size in use.
	.
	-.It Sy zfs_recv_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq int
	+.It Sy zfs_recv_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm receive
	queue.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	-.It Sy zfs_recv_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	+.It Sy zfs_recv_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The maximum number of bytes allowed in the
	.Nm zfs Cm receive
	queue.
	This value must be at least twice the maximum block size in use.
	.
	-.It Sy zfs_recv_write_batch_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	+.It Sy zfs_recv_write_batch_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	The maximum amount of data, in bytes, that
	.Nm zfs Cm receive
	will write in one DMU transaction.
	This is the uncompressed size, even when receiving a compressed send stream.
	This setting will not reduce the write size below a single block.
	Capped at a maximum of
	.Sy 32 MiB .
	.
	.It Sy zfs_recv_best_effort_corrective Ns = Ns Sy 0 Pq int
	When this variable is set to non-zero a corrective receive:
	.Bl -enum -compact -offset 4n -width "1."
	.It
	Does not enforce the restriction of source & destination snapshot GUIDs
	matching.
	.It
	If there is an error during healing, the healing receive is not
	terminated instead it moves on to the next record.
	.El
	.
	-.It Sy zfs_override_estimate_recordsize Ns = Ns Sy 0 Ns \| Ns 1 Pq ulong
	+.It Sy zfs_override_estimate_recordsize Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Setting this variable overrides the default logic for estimating block
	sizes when doing a
	.Nm zfs Cm send .
	The default heuristic is that the average block size
	will be the current recordsize.
	Override this value if most data in your dataset is not of that size
	and you require accurate zfs send size estimates.
	.
	-.It Sy zfs_sync_pass_deferred_free Ns = Ns Sy 2 Pq int
	+.It Sy zfs_sync_pass_deferred_free Ns = Ns Sy 2 Pq uint
	Flushing of data to disk is done in passes.
	Defer frees starting in this pass.
	.
	.It Sy zfs_spa_discard_memory_limit Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	Maximum memory used for prefetching a checkpoint's space map on each
	vdev while discarding the checkpoint.
	.
	-.It Sy zfs_special_class_metadata_reserve_pct Ns = Ns Sy 25 Ns % Pq int
	+.It Sy zfs_special_class_metadata_reserve_pct Ns = Ns Sy 25 Ns % Pq uint
	Only allow small data blocks to be allocated on the special and dedup vdev
	types when the available free space percentage on these vdevs exceeds this value.
	This ensures reserved space is available for pool metadata as the
	special vdevs approach capacity.
	.
	-.It Sy zfs_sync_pass_dont_compress Ns = Ns Sy 8 Pq int
	+.It Sy zfs_sync_pass_dont_compress Ns = Ns Sy 8 Pq uint
	Starting in this sync pass, disable compression (including of metadata).
	With the default setting, in practice, we don't have this many sync passes,
	so this has no effect.
	.Pp
	The original intent was that disabling compression would help the sync passes
	to converge.
	However, in practice, disabling compression increases
	the average number of sync passes; because when we turn compression off,
	many blocks' size will change, and thus we have to re-allocate
	(not overwrite) them.
	It also increases the number of
	.Em 128 KiB
	allocations (e.g. for indirect blocks and spacemaps)
	because these will not be compressed.
	The
	.Em 128 KiB
	allocations are especially detrimental to performance
	on highly fragmented systems, which may have very few free segments of this size,
	and may need to load new metaslabs to satisfy these allocations.
	.
	-.It Sy zfs_sync_pass_rewrite Ns = Ns Sy 2 Pq int
	+.It Sy zfs_sync_pass_rewrite Ns = Ns Sy 2 Pq uint
	Rewrite new block pointers starting in this pass.
	.
	.It Sy zfs_sync_taskq_batch_pct Ns = Ns Sy 75 Ns % Pq int
	This controls the number of threads used by
	.Sy dp_sync_taskq .
	The default value of
	.Sy 75%
	will create a maximum of one thread per CPU.
	.
	.It Sy zfs_trim_extent_bytes_max Ns = Ns Sy 134217728 Ns B Po 128 MiB Pc Pq uint
	Maximum size of TRIM command.
	Larger ranges will be split into chunks no larger than this value before issuing.
	.
	.It Sy zfs_trim_extent_bytes_min Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	Minimum size of TRIM commands.
	TRIM ranges smaller than this will be skipped,
	unless they're part of a larger range which was chunked.
	This is done because it's common for these small TRIMs
	to negatively impact overall performance.
	.
	.It Sy zfs_trim_metaslab_skip Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Skip uninitialized metaslabs during the TRIM process.
	This option is useful for pools constructed from large thinly-provisioned devices
	where TRIM operations are slow.
	As a pool ages, an increasing fraction of the pool's metaslabs
	will be initialized, progressively degrading the usefulness of this option.
	This setting is stored when starting a manual TRIM and will
	persist for the duration of the requested TRIM.
	.
	.It Sy zfs_trim_queue_limit Ns = Ns Sy 10 Pq uint
	Maximum number of queued TRIMs outstanding per leaf vdev.
	The number of concurrent TRIM commands issued to the device is controlled by
	.Sy zfs_vdev_trim_min_active No and Sy zfs_vdev_trim_max_active .
	.
	.It Sy zfs_trim_txg_batch Ns = Ns Sy 32 Pq uint
	The number of transaction groups' worth of frees which should be aggregated
	before TRIM operations are issued to the device.
	This setting represents a trade-off between issuing larger,
	more efficient TRIM operations and the delay
	before the recently trimmed space is available for use by the device.
	.Pp
	Increasing this value will allow frees to be aggregated for a longer time.
	This will result is larger TRIM operations and potentially increased memory usage.
	Decreasing this value will have the opposite effect.
	The default of
	.Sy 32
	was determined to be a reasonable compromise.
	.
	-.It Sy zfs_txg_history Ns = Ns Sy 0 Pq int
	+.It Sy zfs_txg_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest TXGs will be available in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /TXGs .
	.
	-.It Sy zfs_txg_timeout Ns = Ns Sy 5 Ns s Pq int
	+.It Sy zfs_txg_timeout Ns = Ns Sy 5 Ns s Pq uint
	Flush dirty data to disk at least every this many seconds (maximum TXG duration).
	.
	-.It Sy zfs_vdev_aggregate_trim Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	+.It Sy zfs_vdev_aggregate_trim Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Allow TRIM I/O operations to be aggregated.
	This is normally not helpful because the extents to be trimmed
	will have been already been aggregated by the metaslab.
	This option is provided for debugging and performance analysis.
	.
	-.It Sy zfs_vdev_aggregation_limit Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	+.It Sy zfs_vdev_aggregation_limit Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	Max vdev I/O aggregation size.
	.
	-.It Sy zfs_vdev_aggregation_limit_non_rotating Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq int
	+.It Sy zfs_vdev_aggregation_limit_non_rotating Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	Max vdev I/O aggregation size for non-rotating media.
	.
	-.It Sy zfs_vdev_cache_bshift Ns = Ns Sy 16 Po 64 KiB Pc Pq int
	+.It Sy zfs_vdev_cache_bshift Ns = Ns Sy 16 Po 64 KiB Pc Pq uint
	Shift size to inflate reads to.
	.
	-.It Sy zfs_vdev_cache_max Ns = Ns Sy 16384 Ns B Po 16 KiB Pc Pq int
	+.It Sy zfs_vdev_cache_max Ns = Ns Sy 16384 Ns B Po 16 KiB Pc Pq uint
	Inflate reads smaller than this value to meet the
	.Sy zfs_vdev_cache_bshift
	size
	.Pq default Sy 64 KiB .
	.
	-.It Sy zfs_vdev_cache_size Ns = Ns Sy 0 Pq int
	+.It Sy zfs_vdev_cache_size Ns = Ns Sy 0 Pq uint
	Total size of the per-disk cache in bytes.
	.Pp
	Currently this feature is disabled, as it has been found to not be helpful
	for performance and in some cases harmful.
	.
	.It Sy zfs_vdev_mirror_rotating_inc Ns = Ns Sy 0 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation
	immediately follows its predecessor on rotational vdevs
	for the purpose of making decisions based on load.
	.
	.It Sy zfs_vdev_mirror_rotating_seek_inc Ns = Ns Sy 5 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation
	lacks locality as defined by
	.Sy zfs_vdev_mirror_rotating_seek_offset .
	Operations within this that are not immediately following the previous operation
	are incremented by half.
	.
	.It Sy zfs_vdev_mirror_rotating_seek_offset Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	The maximum distance for the last queued I/O operation in which
	the balancing algorithm considers an operation to have locality.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_mirror_non_rotating_inc Ns = Ns Sy 0 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member on non-rotational vdevs
	when I/O operations do not immediately follow one another.
	.
	.It Sy zfs_vdev_mirror_non_rotating_seek_inc Ns = Ns Sy 1 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation lacks
	locality as defined by the
	.Sy zfs_vdev_mirror_rotating_seek_offset .
	Operations within this that are not immediately following the previous operation
	are incremented by half.
	.
	-.It Sy zfs_vdev_read_gap_limit Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq int
	+.It Sy zfs_vdev_read_gap_limit Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	Aggregate read I/O operations if the on-disk gap between them is within this
	threshold.
	.
	-.It Sy zfs_vdev_write_gap_limit Ns = Ns Sy 4096 Ns B Po 4 KiB Pc Pq int
	+.It Sy zfs_vdev_write_gap_limit Ns = Ns Sy 4096 Ns B Po 4 KiB Pc Pq uint
	Aggregate write I/O operations if the on-disk gap between them is within this
	threshold.
	.
	.It Sy zfs_vdev_raidz_impl Ns = Ns Sy fastest Pq string
	Select the raidz parity implementation to use.
	.Pp
	Variants that don't depend on CPU-specific features
	may be selected on module load, as they are supported on all systems.
	The remaining options may only be set after the module is loaded,
	as they are available only if the implementations are compiled in
	and supported on the running system.
	.Pp
	Once the module is loaded,
	.Pa /sys/module/zfs/parameters/zfs_vdev_raidz_impl
	will show the available options,
	with the currently selected one enclosed in square brackets.
	.Pp
	.TS
	lb l l .
	fastest selected by built-in benchmark
	original original implementation
	scalar scalar implementation
	sse2 SSE2 instruction set 64-bit x86
	ssse3 SSSE3 instruction set 64-bit x86
	avx2 AVX2 instruction set 64-bit x86
	avx512f AVX512F instruction set 64-bit x86
	avx512bw AVX512F & AVX512BW instruction sets 64-bit x86
	aarch64_neon NEON Aarch64/64-bit ARMv8
	aarch64_neonx2 NEON with more unrolling Aarch64/64-bit ARMv8
	powerpc_altivec Altivec PowerPC
	.TE
	.
	.It Sy zfs_vdev_scheduler Pq charp
	.Sy DEPRECATED .
	Prints warning to kernel log for compatibility.
	.
	-.It Sy zfs_zevent_len_max Ns = Ns Sy 512 Pq int
	+.It Sy zfs_zevent_len_max Ns = Ns Sy 512 Pq uint
	Max event queue length.
	Events in the queue can be viewed with
	.Xr zpool-events 8 .
	.
	.It Sy zfs_zevent_retain_max Ns = Ns Sy 2000 Pq int
	Maximum recent zevent records to retain for duplicate checking.
	Setting this to
	.Sy 0
	disables duplicate detection.
	.
	.It Sy zfs_zevent_retain_expire_secs Ns = Ns Sy 900 Ns s Po 15 min Pc Pq int
	Lifespan for a recent ereport that was retained for duplicate checking.
	.
	.It Sy zfs_zil_clean_taskq_maxalloc Ns = Ns Sy 1048576 Pq int
	The maximum number of taskq entries that are allowed to be cached.
	When this limit is exceeded transaction records (itxs)
	will be cleaned synchronously.
	.
	.It Sy zfs_zil_clean_taskq_minalloc Ns = Ns Sy 1024 Pq int
	The number of taskq entries that are pre-populated when the taskq is first
	created and are immediately available for use.
	.
	.It Sy zfs_zil_clean_taskq_nthr_pct Ns = Ns Sy 100 Ns % Pq int
	This controls the number of threads used by
	.Sy dp_zil_clean_taskq .
	The default value of
	.Sy 100%
	will create a maximum of one thread per cpu.
	.
	-.It Sy zil_maxblocksize Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq int
	+.It Sy zil_maxblocksize Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	This sets the maximum block size used by the ZIL.
	On very fragmented pools, lowering this
	.Pq typically to Sy 36 KiB
	can improve performance.
	.
	.It Sy zil_nocacheflush Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable the cache flush commands that are normally sent to disk by
	the ZIL after an LWB write has completed.
	Setting this will cause ZIL corruption on power loss
	if a volatile out-of-order write cache is enabled.
	.
	.It Sy zil_replay_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable intent logging replay.
	Can be disabled for recovery from corrupted ZIL.
	.
	.It Sy zil_slog_bulk Ns = Ns Sy 786432 Ns B Po 768 KiB Pc Pq ulong
	Limit SLOG write size per commit executed with synchronous priority.
	Any writes above that will be executed with lower (asynchronous) priority
	to limit potential SLOG device abuse by single active ZIL writer.
	.
	.It Sy zfs_zil_saxattr Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Setting this tunable to zero disables ZIL logging of new
	.Sy xattr Ns = Ns Sy sa
	records if the
	.Sy org.openzfs:zilsaxattr
	feature is enabled on the pool.
	This would only be necessary to work around bugs in the ZIL logging or replay
	code for this record type.
	The tunable has no effect if the feature is disabled.
	.
	-.It Sy zfs_embedded_slog_min_ms Ns = Ns Sy 64 Pq int
	+.It Sy zfs_embedded_slog_min_ms Ns = Ns Sy 64 Pq uint
	Usually, one metaslab from each normal-class vdev is dedicated for use by
	the ZIL to log synchronous writes.
	However, if there are fewer than
	.Sy zfs_embedded_slog_min_ms
	metaslabs in the vdev, this functionality is disabled.
	This ensures that we don't set aside an unreasonable amount of space for the ZIL.
	.
	-.It Sy zstd_earlyabort_pass Ns = Ns Sy 1 Pq int
	+.It Sy zstd_earlyabort_pass Ns = Ns Sy 1 Pq uint
	Whether heuristic for detection of incompressible data with zstd levels >= 3
	using LZ4 and zstd-1 passes is enabled.
	.
	-.It Sy zstd_abort_size Ns = Ns Sy 131072 Pq int
	+.It Sy zstd_abort_size Ns = Ns Sy 131072 Pq uint
	Minimal uncompressed size (inclusive) of a record before the early abort
	heuristic will be attempted.
	.
	.It Sy zio_deadman_log_all Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If non-zero, the zio deadman will produce debugging messages
	.Pq see Sy zfs_dbgmsg_enable
	for all zios, rather than only for leaf zios possessing a vdev.
	This is meant to be used by developers to gain
	diagnostic information for hang conditions which don't involve a mutex
	or other locking primitive: typically conditions in which a thread in
	the zio pipeline is looping indefinitely.
	.
	.It Sy zio_slow_io_ms Ns = Ns Sy 30000 Ns ms Po 30 s Pc Pq int
	When an I/O operation takes more than this much time to complete,
	it's marked as slow.
	Each slow operation causes a delay zevent.
	Slow I/O counters can be seen with
	.Nm zpool Cm status Fl s .
	.
	.It Sy zio_dva_throttle_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Throttle block allocations in the I/O pipeline.
	This allows for dynamic allocation distribution when devices are imbalanced.
	When enabled, the maximum number of pending allocations per top-level vdev
	is limited by
	.Sy zfs_vdev_queue_depth_pct .
	.
	.It Sy zfs_xattr_compat Ns = Ns 0 Ns \| Ns 1 Pq int
	Control the naming scheme used when setting new xattrs in the user namespace.
	If
	.Sy 0
	.Pq the default on Linux ,
	user namespace xattr names are prefixed with the namespace, to be backwards
	compatible with previous versions of ZFS on Linux.
	If
	.Sy 1
	.Pq the default on Fx ,
	user namespace xattr names are not prefixed, to be backwards compatible with
	previous versions of ZFS on illumos and
	.Fx .
	.Pp
	Either naming scheme can be read on this and future versions of ZFS, regardless
	of this tunable, but legacy ZFS on illumos or
	.Fx
	are unable to read user namespace xattrs written in the Linux format, and
	legacy versions of ZFS on Linux are unable to read user namespace xattrs written
	in the legacy ZFS format.
	.Pp
	An existing xattr with the alternate naming scheme is removed when overwriting
	the xattr so as to not accumulate duplicates.
	.
	.It Sy zio_requeue_io_start_cut_in_line Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prioritize requeued I/O.
	.
	.It Sy zio_taskq_batch_pct Ns = Ns Sy 80 Ns % Pq uint
	Percentage of online CPUs which will run a worker thread for I/O.
	These workers are responsible for I/O work such as compression and
	checksum calculations.
	Fractional number of CPUs will be rounded down.
	.Pp
	The default value of
	.Sy 80%
	was chosen to avoid using all CPUs which can result in
	latency issues and inconsistent application performance,
	especially when slower compression and/or checksumming is enabled.
	.
	.It Sy zio_taskq_batch_tpq Ns = Ns Sy 0 Pq uint
	Number of worker threads per taskq.
	Lower values improve I/O ordering and CPU utilization,
	while higher reduces lock contention.
	.Pp
	If
	.Sy 0 ,
	generate a system-dependent value close to 6 threads per taskq.
	.
	.It Sy zvol_inhibit_dev Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Do not create zvol device nodes.
	This may slightly improve startup time on
	systems with a very large number of zvols.
	.
	.It Sy zvol_major Ns = Ns Sy 230 Pq uint
	Major number for zvol block devices.
	.
	.It Sy zvol_max_discard_blocks Ns = Ns Sy 16384 Pq ulong
	Discard (TRIM) operations done on zvols will be done in batches of this
	many blocks, where block size is determined by the
	.Sy volblocksize
	property of a zvol.
	.
	.It Sy zvol_prefetch_bytes Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	When adding a zvol to the system, prefetch this many bytes
	from the start and end of the volume.
	Prefetching these regions of the volume is desirable,
	because they are likely to be accessed immediately by
	.Xr blkid 8
	or the kernel partitioner.
	.
	.It Sy zvol_request_sync Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	When processing I/O requests for a zvol, submit them synchronously.
	This effectively limits the queue depth to
	.Em 1
	for each I/O submitter.
	When unset, requests are handled asynchronously by a thread pool.
	The number of requests which can be handled concurrently is controlled by
	.Sy zvol_threads .
	.Sy zvol_request_sync
	is ignored when running on a kernel that supports block multiqueue
	.Pq Li blk-mq .
	.
	.It Sy zvol_threads Ns = Ns Sy 0 Pq uint
	The number of system wide threads to use for processing zvol block IOs.
	If
	.Sy 0
	(the default) then internally set
	.Sy zvol_threads
	to the number of CPUs present or 32 (whichever is greater).
	.
	.It Sy zvol_blk_mq_threads Ns = Ns Sy 0 Pq uint
	The number of threads per zvol to use for queuing IO requests.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only read and assigned to a zvol at zvol load time.
	If
	.Sy 0
	(the default) then internally set
	.Sy zvol_blk_mq_threads
	to the number of CPUs present.
	.
	.It Sy zvol_use_blk_mq Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Set to
	.Sy 1
	to use the
	.Li blk-mq
	API for zvols.
	Set to
	.Sy 0
	(the default) to use the legacy zvol APIs.
	This setting can give better or worse zvol performance depending on
	the workload.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only read and assigned to a zvol at zvol load time.
	.
	.It Sy zvol_blk_mq_blocks_per_thread Ns = Ns Sy 8 Pq uint
	If
	.Sy zvol_use_blk_mq
	is enabled, then process this number of
	.Sy volblocksize Ns -sized blocks per zvol thread.
	This tunable can be use to favor better performance for zvol reads (lower
	values) or writes (higher values).
	If set to
	.Sy 0 ,
	then the zvol layer will process the maximum number of blocks
	per thread that it can.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only applied at each zvol's load time.
	.
	.It Sy zvol_blk_mq_queue_depth Ns = Ns Sy 0 Pq uint
	The queue_depth value for the zvol
	.Li blk-mq
	interface.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only applied at each zvol's load time.
	If
	.Sy 0
	(the default) then use the kernel's default queue depth.
	Values are clamped to the kernel's
	.Dv BLKDEV_MIN_RQ
	and
	.Dv BLKDEV_MAX_RQ Ns / Ns Dv BLKDEV_DEFAULT_RQ
	limits.
	.
	.It Sy zvol_volmode Ns = Ns Sy 1 Pq uint
	Defines zvol block devices behaviour when
	.Sy volmode Ns = Ns Sy default :
	.Bl -tag -compact -offset 4n -width "a"
	.It Sy 1
	.No equivalent to Sy full
	.It Sy 2
	.No equivalent to Sy dev
	.It Sy 3
	.No equivalent to Sy none
	.El
	.El
	.
	.Sh ZFS I/O SCHEDULER
	ZFS issues I/O operations to leaf vdevs to satisfy and complete I/O operations.
	The scheduler determines when and in what order those operations are issued.
	The scheduler divides operations into five I/O classes,
	prioritized in the following order: sync read, sync write, async read,
	async write, and scrub/resilver.
	Each queue defines the minimum and maximum number of concurrent operations
	that may be issued to the device.
	In addition, the device has an aggregate maximum,
	.Sy zfs_vdev_max_active .
	Note that the sum of the per-queue minima must not exceed the aggregate maximum.
	If the sum of the per-queue maxima exceeds the aggregate maximum,
	then the number of active operations may reach
	.Sy zfs_vdev_max_active ,
	in which case no further operations will be issued,
	regardless of whether all per-queue minima have been met.
	.Pp
	For many physical devices, throughput increases with the number of
	concurrent operations, but latency typically suffers.
	Furthermore, physical devices typically have a limit
	at which more concurrent operations have no
	effect on throughput or can actually cause it to decrease.
	.Pp
	The scheduler selects the next operation to issue by first looking for an
	I/O class whose minimum has not been satisfied.
	Once all are satisfied and the aggregate maximum has not been hit,
	the scheduler looks for classes whose maximum has not been satisfied.
	Iteration through the I/O classes is done in the order specified above.
	No further operations are issued
	if the aggregate maximum number of concurrent operations has been hit,
	or if there are no operations queued for an I/O class that has not hit its maximum.
	Every time an I/O operation is queued or an operation completes,
	the scheduler looks for new operations to issue.
	.Pp
	In general, smaller
	.Sy max_active Ns s
	will lead to lower latency of synchronous operations.
	Larger
	.Sy max_active Ns s
	may lead to higher overall throughput, depending on underlying storage.
	.Pp
	The ratio of the queues'
	.Sy max_active Ns s
	determines the balance of performance between reads, writes, and scrubs.
	For example, increasing
	.Sy zfs_vdev_scrub_max_active
	will cause the scrub or resilver to complete more quickly,
	but reads and writes to have higher latency and lower throughput.
	.Pp
	All I/O classes have a fixed maximum number of outstanding operations,
	except for the async write class.
	Asynchronous writes represent the data that is committed to stable storage
	during the syncing stage for transaction groups.
	Transaction groups enter the syncing state periodically,
	so the number of queued async writes will quickly burst up
	and then bleed down to zero.
	Rather than servicing them as quickly as possible,
	the I/O scheduler changes the maximum number of active async write operations
	according to the amount of dirty data in the pool.
	Since both throughput and latency typically increase with the number of
	concurrent operations issued to physical devices, reducing the
	burstiness in the number of concurrent operations also stabilizes the
	response time of operations from other – and in particular synchronous – queues.
	In broad strokes, the I/O scheduler will issue more concurrent operations
	from the async write queue as there's more dirty data in the pool.
	.
	.Ss Async Writes
	The number of concurrent operations issued for the async write I/O class
	follows a piece-wise linear function defined by a few adjustable points:
	.Bd -literal
	\| o---------\| <-- \fBzfs_vdev_async_write_max_active\fP
	^ \| /^ \|
	\| \| / \| \|
	active \| / \| \|
	I/O \| / \| \|
	count \| / \| \|
	\| / \| \|
	\|-------o \| \| <-- \fBzfs_vdev_async_write_min_active\fP
	0\|_______^______\|_________\|
	0% \| \| 100% of \fBzfs_dirty_data_max\fP
	\| \|
	\| `-- \fBzfs_vdev_async_write_active_max_dirty_percent\fP
	`--------- \fBzfs_vdev_async_write_active_min_dirty_percent\fP
	.Ed
	.Pp
	Until the amount of dirty data exceeds a minimum percentage of the dirty
	data allowed in the pool, the I/O scheduler will limit the number of
	concurrent operations to the minimum.
	As that threshold is crossed, the number of concurrent operations issued
	increases linearly to the maximum at the specified maximum percentage
	of the dirty data allowed in the pool.
	.Pp
	Ideally, the amount of dirty data on a busy pool will stay in the sloped
	part of the function between
	.Sy zfs_vdev_async_write_active_min_dirty_percent
	and
	.Sy zfs_vdev_async_write_active_max_dirty_percent .
	If it exceeds the maximum percentage,
	this indicates that the rate of incoming data is
	greater than the rate that the backend storage can handle.
	In this case, we must further throttle incoming writes,
	as described in the next section.
	.
	.Sh ZFS TRANSACTION DELAY
	We delay transactions when we've determined that the backend storage
	isn't able to accommodate the rate of incoming writes.
	.Pp
	If there is already a transaction waiting, we delay relative to when
	that transaction will finish waiting.
	This way the calculated delay time
	is independent of the number of threads concurrently executing transactions.
	.Pp
	If we are the only waiter, wait relative to when the transaction started,
	rather than the current time.
	This credits the transaction for "time already served",
	e.g. reading indirect blocks.
	.Pp
	The minimum time for a transaction to take is calculated as
	.D1 min_time = min( Ns Sy zfs_delay_scale No \(mu Po Sy dirty No \- Sy min Pc / Po Sy max No \- Sy dirty Pc , 100ms)
	.Pp
	The delay has two degrees of freedom that can be adjusted via tunables.
	The percentage of dirty data at which we start to delay is defined by
	.Sy zfs_delay_min_dirty_percent .
	This should typically be at or above
	.Sy zfs_vdev_async_write_active_max_dirty_percent ,
	so that we only start to delay after writing at full speed
	has failed to keep up with the incoming write rate.
	The scale of the curve is defined by
	.Sy zfs_delay_scale .
	Roughly speaking, this variable determines the amount of delay at the midpoint of the curve.
	.Bd -literal
	delay
	10ms +-------------------------------------------------------------*+
	\| *\|
	9ms + *+
	\| *\|
	8ms + *+
	\| * \|
	7ms + * +
	\| * \|
	6ms + * +
	\| * \|
	5ms + * +
	\| * \|
	4ms + * +
	\| * \|
	3ms + * +
	\| * \|
	2ms + (midpoint) * +
	\| \| ** \|
	1ms + v *** +
	\| \fBzfs_delay_scale\fP ----------> ******** \|
	0 +-------------------------------------*********----------------+
	0% <- \fBzfs_dirty_data_max\fP -> 100%
	.Ed
	.Pp
	Note, that since the delay is added to the outstanding time remaining on the
	most recent transaction it's effectively the inverse of IOPS.
	Here, the midpoint of
	.Em 500 us
	translates to
	.Em 2000 IOPS .
	The shape of the curve
	was chosen such that small changes in the amount of accumulated dirty data
	in the first three quarters of the curve yield relatively small differences
	in the amount of delay.
	.Pp
	The effects can be easier to understand when the amount of delay is
	represented on a logarithmic scale:
	.Bd -literal
	delay
	100ms +-------------------------------------------------------------++
	+ +
	\| \|
	+ *+
	10ms + *+
	+ ** +
	\| (midpoint) ** \|
	+ \| ** +
	1ms + v **** +
	+ \fBzfs_delay_scale\fP ----------> ***** +
	\| **** \|
	+ **** +
	100us + ** +
	+ * +
	\| * \|
	+ * +
	10us + * +
	+ +
	\| \|
	+ +
	+--------------------------------------------------------------+
	0% <- \fBzfs_dirty_data_max\fP -> 100%
	.Ed
	.Pp
	Note here that only as the amount of dirty data approaches its limit does
	the delay start to increase rapidly.
	The goal of a properly tuned system should be to keep the amount of dirty data
	out of that range by first ensuring that the appropriate limits are set
	for the I/O scheduler to reach optimal throughput on the back-end storage,
	and then by changing the value of
	.Sy zfs_delay_scale
	to increase the steepness of the curve.
	diff --git a/sys/contrib/openzfs/module/icp/algs/blake3/blake3_impl.c b/sys/contrib/openzfs/module/icp/algs/blake3/blake3_impl.c
	index 5276fd88fbb6..1692916cef97 100644
	--- a/sys/contrib/openzfs/module/icp/algs/blake3/blake3_impl.c
	+++ b/sys/contrib/openzfs/module/icp/algs/blake3/blake3_impl.c
	@@ -1,362 +1,361 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2021-2022 Tino Reichardt <milky-zfs@mcmilk.de>
	*/

	#include <sys/zfs_context.h>
	#include <sys/zio_checksum.h>

	#include "blake3_impl.h"

	static const blake3_ops_t *const blake3_impls[] = {
	&blake3_generic_impl,
	#if defined(__aarch64__) \|\| \
	(defined(__x86_64) && defined(HAVE_SSE2)) \|\| \
	(defined(__PPC64__) && defined(__LITTLE_ENDIAN__))
	&blake3_sse2_impl,
	#endif
	#if defined(__aarch64__) \|\| \
	(defined(__x86_64) && defined(HAVE_SSE4_1)) \|\| \
	(defined(__PPC64__) && defined(__LITTLE_ENDIAN__))
	&blake3_sse41_impl,
	#endif
	#if defined(__x86_64) && defined(HAVE_SSE4_1) && defined(HAVE_AVX2)
	&blake3_avx2_impl,
	#endif
	#if defined(__x86_64) && defined(HAVE_AVX512F) && defined(HAVE_AVX512VL)
	&blake3_avx512_impl,
	#endif
	};

	/* Select BLAKE3 implementation */
	#define IMPL_FASTEST (UINT32_MAX)
	#define IMPL_CYCLE (UINT32_MAX - 1)

	#define IMPL_READ(i) ((volatile uint32_t ) &(i))

	/* Indicate that benchmark has been done */
	static boolean_t blake3_initialized = B_FALSE;

	/* Implementation that contains the fastest methods */
	static blake3_ops_t blake3_fastest_impl = {
	.name = "fastest"
	};

	/* Hold all supported implementations */
	static const blake3_ops_t *blake3_supp_impls[ARRAY_SIZE(blake3_impls)];
	static uint32_t blake3_supp_impls_cnt = 0;

	/* Currently selected implementation */
	static uint32_t blake3_impl_chosen = IMPL_FASTEST;

	static struct blake3_impl_selector {
	const char *name;
	uint32_t sel;
	} blake3_impl_selectors[] = {
	{ "cycle", IMPL_CYCLE },
	{ "fastest", IMPL_FASTEST }
	};

	/* check the supported implementations */
	static void blake3_impl_init(void)
	{
	int i, c;

	/* init only once */
	if (likely(blake3_initialized))
	return;

	/* move supported implementations into blake3_supp_impls */
	for (i = 0, c = 0; i < ARRAY_SIZE(blake3_impls); i++) {
	const blake3_ops_t *impl = blake3_impls[i];

	if (impl->is_supported && impl->is_supported())
	blake3_supp_impls[c++] = impl;
	}
	blake3_supp_impls_cnt = c;

	/* first init generic impl, may be changed via set_fastest() */
	memcpy(&blake3_fastest_impl, blake3_impls[0],
	sizeof (blake3_fastest_impl));
	blake3_initialized = B_TRUE;
	}

	/* get number of supported implementations */
	uint32_t
	blake3_impl_getcnt(void)
	{
	blake3_impl_init();
	return (blake3_supp_impls_cnt);
	}

	/* get id of selected implementation */
	uint32_t
	blake3_impl_getid(void)
	{
	return (IMPL_READ(blake3_impl_chosen));
	}

	/* get name of selected implementation */
	const char *
	blake3_impl_getname(void)
	{
	uint32_t impl = IMPL_READ(blake3_impl_chosen);

	blake3_impl_init();
	switch (impl) {
	case IMPL_FASTEST:
	return ("fastest");
	case IMPL_CYCLE:
	return ("cycle");
	default:
	return (blake3_supp_impls[impl]->name);
	}
	}

	/* setup id as fastest implementation */
	void
	blake3_impl_set_fastest(uint32_t id)
	{
	/* setup fastest impl */
	memcpy(&blake3_fastest_impl, blake3_supp_impls[id],
	sizeof (blake3_fastest_impl));
	}

	/* set implementation by id */
	void
	blake3_impl_setid(uint32_t id)
	{
	blake3_impl_init();
	switch (id) {
	case IMPL_FASTEST:
	atomic_swap_32(&blake3_impl_chosen, IMPL_FASTEST);
	break;
	case IMPL_CYCLE:
	atomic_swap_32(&blake3_impl_chosen, IMPL_CYCLE);
	break;
	default:
	- ASSERT3U(id, >=, 0);
	ASSERT3U(id, <, blake3_supp_impls_cnt);
	atomic_swap_32(&blake3_impl_chosen, id);
	break;
	}
	}

	/* set implementation by name */
	int
	blake3_impl_setname(const char *val)
	{
	uint32_t impl = IMPL_READ(blake3_impl_chosen);
	size_t val_len;
	int i, err = -EINVAL;

	blake3_impl_init();
	val_len = strlen(val);
	while ((val_len > 0) && !!isspace(val[val_len-1])) /* trim '\n' */
	val_len--;

	/* check mandatory implementations */
	for (i = 0; i < ARRAY_SIZE(blake3_impl_selectors); i++) {
	const char *name = blake3_impl_selectors[i].name;

	if (val_len == strlen(name) &&
	strncmp(val, name, val_len) == 0) {
	impl = blake3_impl_selectors[i].sel;
	err = 0;
	break;
	}
	}

	if (err != 0 && blake3_initialized) {
	/* check all supported implementations */
	for (i = 0; i < blake3_supp_impls_cnt; i++) {
	const char *name = blake3_supp_impls[i]->name;

	if (val_len == strlen(name) &&
	strncmp(val, name, val_len) == 0) {
	impl = i;
	err = 0;
	break;
	}
	}
	}

	if (err == 0) {
	atomic_swap_32(&blake3_impl_chosen, impl);
	}

	return (err);
	}

	const blake3_ops_t *
	blake3_impl_get_ops(void)
	{
	const blake3_ops_t *ops = NULL;
	uint32_t impl = IMPL_READ(blake3_impl_chosen);

	blake3_impl_init();
	switch (impl) {
	case IMPL_FASTEST:
	ASSERT(blake3_initialized);
	ops = &blake3_fastest_impl;
	break;
	case IMPL_CYCLE:
	/* Cycle through supported implementations */
	ASSERT(blake3_initialized);
	ASSERT3U(blake3_supp_impls_cnt, >, 0);
	static uint32_t cycle_count = 0;
	uint32_t idx = (++cycle_count) % blake3_supp_impls_cnt;
	ops = blake3_supp_impls[idx];
	break;
	default:
	ASSERT3U(blake3_supp_impls_cnt, >, 0);
	ASSERT3U(impl, <, blake3_supp_impls_cnt);
	ops = blake3_supp_impls[impl];
	break;
	}

	ASSERT3P(ops, !=, NULL);
	return (ops);
	}

	#if defined(_KERNEL)

	void **blake3_per_cpu_ctx;

	void
	blake3_per_cpu_ctx_init(void)
	{
	/*
	* Create "The Godfather" ptr to hold all blake3 ctx
	*/
	blake3_per_cpu_ctx = kmem_alloc(max_ncpus * sizeof (void *), KM_SLEEP);
	for (int i = 0; i < max_ncpus; i++) {
	blake3_per_cpu_ctx[i] = kmem_alloc(sizeof (BLAKE3_CTX),
	KM_SLEEP);
	}

	/* init once in kernel mode */
	blake3_impl_init();
	}

	void
	blake3_per_cpu_ctx_fini(void)
	{
	for (int i = 0; i < max_ncpus; i++) {
	memset(blake3_per_cpu_ctx[i], 0, sizeof (BLAKE3_CTX));
	kmem_free(blake3_per_cpu_ctx[i], sizeof (BLAKE3_CTX));
	}
	memset(blake3_per_cpu_ctx, 0, max_ncpus * sizeof (void *));
	kmem_free(blake3_per_cpu_ctx, max_ncpus * sizeof (void *));
	}

	#define IMPL_FMT(impl, i) (((impl) == (i)) ? "[%s] " : "%s ")

	#if defined(__linux__)

	static int
	blake3_param_get(char buffer, zfs_kernel_param_t unused)
	{
	const uint32_t impl = IMPL_READ(blake3_impl_chosen);
	char *fmt;
	int cnt = 0;

	/* cycling */
	fmt = IMPL_FMT(impl, IMPL_CYCLE);
	cnt += sprintf(buffer + cnt, fmt, "cycle");

	/* list fastest */
	fmt = IMPL_FMT(impl, IMPL_FASTEST);
	cnt += sprintf(buffer + cnt, fmt, "fastest");

	/* list all supported implementations */
	for (uint32_t i = 0; i < blake3_supp_impls_cnt; ++i) {
	fmt = IMPL_FMT(impl, i);
	cnt += sprintf(buffer + cnt, fmt,
	blake3_supp_impls[i]->name);
	}

	return (cnt);
	}

	static int
	blake3_param_set(const char val, zfs_kernel_param_t unused)
	{
	(void) unused;
	return (blake3_impl_setname(val));
	}

	#elif defined(__FreeBSD__)

	#include <sys/sbuf.h>

	static int
	blake3_param(ZFS_MODULE_PARAM_ARGS)
	{
	int err;

	if (req->newptr == NULL) {
	const uint32_t impl = IMPL_READ(blake3_impl_chosen);
	const int init_buflen = 64;
	const char *fmt;
	struct sbuf *s;

	s = sbuf_new_for_sysctl(NULL, NULL, init_buflen, req);

	/* cycling */
	fmt = IMPL_FMT(impl, IMPL_CYCLE);
	(void) sbuf_printf(s, fmt, "cycle");

	/* list fastest */
	fmt = IMPL_FMT(impl, IMPL_FASTEST);
	(void) sbuf_printf(s, fmt, "fastest");

	/* list all supported implementations */
	for (uint32_t i = 0; i < blake3_supp_impls_cnt; ++i) {
	fmt = IMPL_FMT(impl, i);
	(void) sbuf_printf(s, fmt, blake3_supp_impls[i]->name);
	}

	err = sbuf_finish(s);
	sbuf_delete(s);

	return (err);
	}

	char buf[16];

	err = sysctl_handle_string(oidp, buf, sizeof (buf), req);
	if (err) {
	return (err);
	}

	return (-blake3_impl_setname(buf));
	}
	#endif

	#undef IMPL_FMT

	ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs, zfs_, blake3_impl,
	blake3_param_set, blake3_param_get, ZMOD_RW, \
	"Select BLAKE3 implementation.");
	#endif
	diff --git a/sys/contrib/openzfs/module/icp/io/skein_mod.c b/sys/contrib/openzfs/module/icp/io/skein_mod.c
	index a2ed6cedd8c6..221e1debd45b 100644
	--- a/sys/contrib/openzfs/module/icp/io/skein_mod.c
	+++ b/sys/contrib/openzfs/module/icp/io/skein_mod.c
	@@ -1,645 +1,656 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or http://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	*/

	#include <sys/crypto/common.h>
	#include <sys/crypto/icp.h>
	#include <sys/crypto/spi.h>
	#include <sys/sysmacros.h>
	#define SKEIN_MODULE_IMPL
	#include <sys/skein.h>

	static const crypto_mech_info_t skein_mech_info_tab[] = {
	{CKM_SKEIN_256, SKEIN_256_MECH_INFO_TYPE,
	CRYPTO_FG_DIGEST \| CRYPTO_FG_DIGEST_ATOMIC},
	{CKM_SKEIN_256_MAC, SKEIN_256_MAC_MECH_INFO_TYPE,
	CRYPTO_FG_MAC \| CRYPTO_FG_MAC_ATOMIC},
	{CKM_SKEIN_512, SKEIN_512_MECH_INFO_TYPE,
	CRYPTO_FG_DIGEST \| CRYPTO_FG_DIGEST_ATOMIC},
	{CKM_SKEIN_512_MAC, SKEIN_512_MAC_MECH_INFO_TYPE,
	CRYPTO_FG_MAC \| CRYPTO_FG_MAC_ATOMIC},
	{CKM_SKEIN1024, SKEIN1024_MECH_INFO_TYPE,
	CRYPTO_FG_DIGEST \| CRYPTO_FG_DIGEST_ATOMIC},
	{CKM_SKEIN1024_MAC, SKEIN1024_MAC_MECH_INFO_TYPE,
	CRYPTO_FG_MAC \| CRYPTO_FG_MAC_ATOMIC},
	};

	static int skein_digest_init(crypto_ctx_t , crypto_mechanism_t );
	static int skein_digest(crypto_ctx_t , crypto_data_t , crypto_data_t *);
	static int skein_update(crypto_ctx_t , crypto_data_t );
	static int skein_final(crypto_ctx_t , crypto_data_t );
	static int skein_digest_atomic(crypto_mechanism_t , crypto_data_t ,
	crypto_data_t *);

	static const crypto_digest_ops_t skein_digest_ops = {
	.digest_init = skein_digest_init,
	.digest = skein_digest,
	.digest_update = skein_update,
	.digest_final = skein_final,
	.digest_atomic = skein_digest_atomic
	};

	static int skein_mac_init(crypto_ctx_t , crypto_mechanism_t , crypto_key_t *,
	crypto_spi_ctx_template_t);
	static int skein_mac_atomic(crypto_mechanism_t , crypto_key_t ,
	crypto_data_t , crypto_data_t , crypto_spi_ctx_template_t);

	static const crypto_mac_ops_t skein_mac_ops = {
	.mac_init = skein_mac_init,
	.mac = NULL,
	.mac_update = skein_update, /* using regular digest update is OK here */
	.mac_final = skein_final, /* using regular digest final is OK here */
	.mac_atomic = skein_mac_atomic,
	.mac_verify_atomic = NULL
	};

	static int skein_create_ctx_template(crypto_mechanism_t , crypto_key_t ,
	crypto_spi_ctx_template_t , size_t );
	static int skein_free_context(crypto_ctx_t *);

	static const crypto_ctx_ops_t skein_ctx_ops = {
	.create_ctx_template = skein_create_ctx_template,
	.free_context = skein_free_context
	};

	static const crypto_ops_t skein_crypto_ops = {
	&skein_digest_ops,
	NULL,
	&skein_mac_ops,
	&skein_ctx_ops,
	};

	static const crypto_provider_info_t skein_prov_info = {
	"Skein Software Provider",
	&skein_crypto_ops,
	sizeof (skein_mech_info_tab) / sizeof (crypto_mech_info_t),
	skein_mech_info_tab
	};

	static crypto_kcf_provider_handle_t skein_prov_handle = 0;

	typedef struct skein_ctx {
	skein_mech_type_t sc_mech_type;
	size_t sc_digest_bitlen;
	/LINTED(E_ANONYMOUS_UNION_DECL)/
	union {
	Skein_256_Ctxt_t sc_256;
	Skein_512_Ctxt_t sc_512;
	Skein1024_Ctxt_t sc_1024;
	};
	} skein_ctx_t;
	#define SKEIN_CTX(_ctx_) ((skein_ctx_t *)((_ctx_)->cc_provider_private))
	#define SKEIN_CTX_LVALUE(_ctx_) (_ctx_)->cc_provider_private
	#define SKEIN_OP(_skein_ctx, _op, ...) \
	do { \
	skein_ctx_t *sc = (_skein_ctx); \
	switch (sc->sc_mech_type) { \
	case SKEIN_256_MECH_INFO_TYPE: \
	case SKEIN_256_MAC_MECH_INFO_TYPE: \
	(void) Skein_256_ ## _op(&sc->sc_256, __VA_ARGS__);\
	break; \
	case SKEIN_512_MECH_INFO_TYPE: \
	case SKEIN_512_MAC_MECH_INFO_TYPE: \
	(void) Skein_512_ ## _op(&sc->sc_512, __VA_ARGS__);\
	break; \
	case SKEIN1024_MECH_INFO_TYPE: \
	case SKEIN1024_MAC_MECH_INFO_TYPE: \
	(void) Skein1024_ ## _op(&sc->sc_1024, __VA_ARGS__);\
	break; \
	} \
	} while (0)

	static int
	skein_get_digest_bitlen(const crypto_mechanism_t mechanism, size_t result)
	{
	if (mechanism->cm_param != NULL) {
	/LINTED(E_BAD_PTR_CAST_ALIGN)/
	skein_param_t param = (skein_param_t )mechanism->cm_param;

	if (mechanism->cm_param_len != sizeof (*param) \|\|
	param->sp_digest_bitlen == 0) {
	return (CRYPTO_MECHANISM_PARAM_INVALID);
	}
	*result = param->sp_digest_bitlen;
	} else {
	switch (mechanism->cm_type) {
	case SKEIN_256_MECH_INFO_TYPE:
	*result = 256;
	break;
	case SKEIN_512_MECH_INFO_TYPE:
	*result = 512;
	break;
	case SKEIN1024_MECH_INFO_TYPE:
	*result = 1024;
	break;
	default:
	return (CRYPTO_MECHANISM_INVALID);
	}
	}
	return (CRYPTO_SUCCESS);
	}

	int
	skein_mod_init(void)
	{
	/*
	* Try to register with KCF - failure shouldn't unload us, since we
	* still may want to continue providing misc/skein functionality.
	*/
	(void) crypto_register_provider(&skein_prov_info, &skein_prov_handle);

	return (0);
	}

	int
	skein_mod_fini(void)
	{
	int ret = 0;

	if (skein_prov_handle != 0) {
	if ((ret = crypto_unregister_provider(skein_prov_handle)) !=
	CRYPTO_SUCCESS) {
	cmn_err(CE_WARN,
	"skein _fini: crypto_unregister_provider() "
	"failed (0x%x)", ret);
	return (EBUSY);
	}
	skein_prov_handle = 0;
	}

	return (0);
	}

	/*
	* General Skein hashing helper functions.
	*/

	/*
	* Performs an Update on a context with uio input data.
	*/
	static int
	skein_digest_update_uio(skein_ctx_t ctx, const crypto_data_t data)
	{
	off_t offset = data->cd_offset;
	size_t length = data->cd_length;
	uint_t vec_idx = 0;
	size_t cur_len;
	zfs_uio_t *uio = data->cd_uio;

	/* we support only kernel buffer */
	if (zfs_uio_segflg(uio) != UIO_SYSSPACE)
	return (CRYPTO_ARGUMENTS_BAD);

	/*
	* Jump to the first iovec containing data to be
	* digested.
	*/
	offset = zfs_uio_index_at_offset(uio, offset, &vec_idx);
	if (vec_idx == zfs_uio_iovcnt(uio)) {
	/*
	* The caller specified an offset that is larger than the
	* total size of the buffers it provided.
	*/
	return (CRYPTO_DATA_LEN_RANGE);
	}

	/*
	* Now do the digesting on the iovecs.
	*/
	while (vec_idx < zfs_uio_iovcnt(uio) && length > 0) {
	cur_len = MIN(zfs_uio_iovlen(uio, vec_idx) - offset, length);
	SKEIN_OP(ctx, Update, (uint8_t *)zfs_uio_iovbase(uio, vec_idx)
	+ offset, cur_len);
	length -= cur_len;
	vec_idx++;
	offset = 0;
	}

	if (vec_idx == zfs_uio_iovcnt(uio) && length > 0) {
	/*
	* The end of the specified iovec's was reached but
	* the length requested could not be processed, i.e.
	* The caller requested to digest more data than it provided.
	*/
	return (CRYPTO_DATA_LEN_RANGE);
	}

	return (CRYPTO_SUCCESS);
	}

	/*
	* Performs a Final on a context and writes to a uio digest output.
	*/
	static int
	skein_digest_final_uio(skein_ctx_t ctx, crypto_data_t digest)
	{
	off_t offset = digest->cd_offset;
	uint_t vec_idx = 0;
	zfs_uio_t *uio = digest->cd_uio;

	/* we support only kernel buffer */
	if (zfs_uio_segflg(uio) != UIO_SYSSPACE)
	return (CRYPTO_ARGUMENTS_BAD);

	/*
	* Jump to the first iovec containing ptr to the digest to be returned.
	*/
	offset = zfs_uio_index_at_offset(uio, offset, &vec_idx);
	if (vec_idx == zfs_uio_iovcnt(uio)) {
	/*
	* The caller specified an offset that is larger than the
	* total size of the buffers it provided.
	*/
	return (CRYPTO_DATA_LEN_RANGE);
	}
	if (offset + CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen) <=
	zfs_uio_iovlen(uio, vec_idx)) {
	/* The computed digest will fit in the current iovec. */
	SKEIN_OP(ctx, Final,
	(uchar_t *)zfs_uio_iovbase(uio, vec_idx) + offset);
	} else {
	uint8_t *digest_tmp;
	off_t scratch_offset = 0;
	size_t length = CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen);
	size_t cur_len;

	digest_tmp = kmem_alloc(CRYPTO_BITS2BYTES(
	ctx->sc_digest_bitlen), KM_SLEEP);
	if (digest_tmp == NULL)
	return (CRYPTO_HOST_MEMORY);
	SKEIN_OP(ctx, Final, digest_tmp);
	while (vec_idx < zfs_uio_iovcnt(uio) && length > 0) {
	cur_len = MIN(zfs_uio_iovlen(uio, vec_idx) - offset,
	length);
	memcpy(zfs_uio_iovbase(uio, vec_idx) + offset,
	digest_tmp + scratch_offset, cur_len);

	length -= cur_len;
	vec_idx++;
	scratch_offset += cur_len;
	offset = 0;
	}
	kmem_free(digest_tmp, CRYPTO_BITS2BYTES(ctx->sc_digest_bitlen));

	if (vec_idx == zfs_uio_iovcnt(uio) && length > 0) {
	/*
	* The end of the specified iovec's was reached but
	* the length requested could not be processed, i.e.
	* The caller requested to digest more data than it
	* provided.
	*/
	return (CRYPTO_DATA_LEN_RANGE);
	}
	}

	return (CRYPTO_SUCCESS);
	}

	/*
	* KCF software provider digest entry points.
	*/

	/*
	* Initializes a skein digest context to the configuration in `mechanism'.
	* The mechanism cm_type must be one of SKEIN_*_MECH_INFO_TYPE. The cm_param
	* field may contain a skein_param_t structure indicating the length of the
	* digest the algorithm should produce. Otherwise the default output lengths
	* are applied (32 bytes for Skein-256, 64 bytes for Skein-512 and 128 bytes
	* for Skein-1024).
	*/
	static int
	skein_digest_init(crypto_ctx_t ctx, crypto_mechanism_t mechanism)
	{
	int error = CRYPTO_SUCCESS;

	if (!VALID_SKEIN_DIGEST_MECH(mechanism->cm_type))
	return (CRYPTO_MECHANISM_INVALID);

	SKEIN_CTX_LVALUE(ctx) = kmem_alloc(sizeof (*SKEIN_CTX(ctx)), KM_SLEEP);
	if (SKEIN_CTX(ctx) == NULL)
	return (CRYPTO_HOST_MEMORY);

	SKEIN_CTX(ctx)->sc_mech_type = mechanism->cm_type;
	error = skein_get_digest_bitlen(mechanism,
	&SKEIN_CTX(ctx)->sc_digest_bitlen);
	if (error != CRYPTO_SUCCESS)
	goto errout;
	SKEIN_OP(SKEIN_CTX(ctx), Init, SKEIN_CTX(ctx)->sc_digest_bitlen);

	return (CRYPTO_SUCCESS);
	errout:
	memset(SKEIN_CTX(ctx), 0, sizeof (*SKEIN_CTX(ctx)));
	kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
	SKEIN_CTX_LVALUE(ctx) = NULL;
	return (error);
	}

	/*
	* Executes a skein_update and skein_digest on a pre-initialized crypto
	* context in a single step. See the documentation to these functions to
	* see what to pass here.
	*/
	static int
	skein_digest(crypto_ctx_t ctx, crypto_data_t data, crypto_data_t *digest)
	{
	int error = CRYPTO_SUCCESS;

	ASSERT(SKEIN_CTX(ctx) != NULL);

	if (digest->cd_length <
	CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen)) {
	digest->cd_length =
	CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	error = skein_update(ctx, data);
	if (error != CRYPTO_SUCCESS) {
	memset(SKEIN_CTX(ctx), 0, sizeof (*SKEIN_CTX(ctx)));
	kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
	SKEIN_CTX_LVALUE(ctx) = NULL;
	digest->cd_length = 0;
	return (error);
	}
	error = skein_final(ctx, digest);

	return (error);
	}

	/*
	* Performs a skein Update with the input message in `data' (successive calls
	* can push more data). This is used both for digest and MAC operation.
	* Supported input data formats are raw, uio and mblk.
	*/
	static int
	skein_update(crypto_ctx_t ctx, crypto_data_t data)
	{
	int error = CRYPTO_SUCCESS;

	ASSERT(SKEIN_CTX(ctx) != NULL);

	switch (data->cd_format) {
	case CRYPTO_DATA_RAW:
	SKEIN_OP(SKEIN_CTX(ctx), Update,
	(uint8_t *)data->cd_raw.iov_base + data->cd_offset,
	data->cd_length);
	break;
	case CRYPTO_DATA_UIO:
	error = skein_digest_update_uio(SKEIN_CTX(ctx), data);
	break;
	default:
	error = CRYPTO_ARGUMENTS_BAD;
	}

	return (error);
	}

	/*
	* Performs a skein Final, writing the output to `digest'. This is used both
	* for digest and MAC operation.
	* Supported output digest formats are raw, uio and mblk.
	*/
	static int
	-skein_final(crypto_ctx_t ctx, crypto_data_t digest)
	+skein_final_nofree(crypto_ctx_t ctx, crypto_data_t digest)
	{
	int error = CRYPTO_SUCCESS;

	ASSERT(SKEIN_CTX(ctx) != NULL);

	if (digest->cd_length <
	CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen)) {
	digest->cd_length =
	CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	switch (digest->cd_format) {
	case CRYPTO_DATA_RAW:
	SKEIN_OP(SKEIN_CTX(ctx), Final,
	(uint8_t *)digest->cd_raw.iov_base + digest->cd_offset);
	break;
	case CRYPTO_DATA_UIO:
	error = skein_digest_final_uio(SKEIN_CTX(ctx), digest);
	break;
	default:
	error = CRYPTO_ARGUMENTS_BAD;
	}

	if (error == CRYPTO_SUCCESS)
	digest->cd_length =
	CRYPTO_BITS2BYTES(SKEIN_CTX(ctx)->sc_digest_bitlen);
	else
	digest->cd_length = 0;

	+ return (error);
	+}
	+
	+static int
	+skein_final(crypto_ctx_t ctx, crypto_data_t digest)
	+{
	+ int error = skein_final_nofree(ctx, digest);
	+
	+ if (error == CRYPTO_BUFFER_TOO_SMALL)
	+ return (error);
	+
	memset(SKEIN_CTX(ctx), 0, sizeof (*SKEIN_CTX(ctx)));
	kmem_free(SKEIN_CTX(ctx), sizeof (*(SKEIN_CTX(ctx))));
	SKEIN_CTX_LVALUE(ctx) = NULL;

	return (error);
	}

	/*
	* Performs a full skein digest computation in a single call, configuring the
	* algorithm according to `mechanism', reading the input to be digested from
	* `data' and writing the output to `digest'.
	* Supported input/output formats are raw, uio and mblk.
	*/
	static int
	skein_digest_atomic(crypto_mechanism_t mechanism, crypto_data_t data,
	crypto_data_t *digest)
	{
	int error;
	skein_ctx_t skein_ctx;
	crypto_ctx_t ctx;
	SKEIN_CTX_LVALUE(&ctx) = &skein_ctx;

	/* Init */
	if (!VALID_SKEIN_DIGEST_MECH(mechanism->cm_type))
	return (CRYPTO_MECHANISM_INVALID);
	skein_ctx.sc_mech_type = mechanism->cm_type;
	error = skein_get_digest_bitlen(mechanism, &skein_ctx.sc_digest_bitlen);
	if (error != CRYPTO_SUCCESS)
	goto out;
	SKEIN_OP(&skein_ctx, Init, skein_ctx.sc_digest_bitlen);

	if ((error = skein_update(&ctx, data)) != CRYPTO_SUCCESS)
	goto out;
	- if ((error = skein_final(&ctx, data)) != CRYPTO_SUCCESS)
	+ if ((error = skein_final_nofree(&ctx, data)) != CRYPTO_SUCCESS)
	goto out;

	out:
	if (error == CRYPTO_SUCCESS)
	digest->cd_length =
	CRYPTO_BITS2BYTES(skein_ctx.sc_digest_bitlen);
	else
	digest->cd_length = 0;
	memset(&skein_ctx, 0, sizeof (skein_ctx));

	return (error);
	}

	/*
	* Helper function that builds a Skein MAC context from the provided
	* mechanism and key.
	*/
	static int
	skein_mac_ctx_build(skein_ctx_t ctx, crypto_mechanism_t mechanism,
	crypto_key_t *key)
	{
	int error;

	if (!VALID_SKEIN_MAC_MECH(mechanism->cm_type))
	return (CRYPTO_MECHANISM_INVALID);
	ctx->sc_mech_type = mechanism->cm_type;
	error = skein_get_digest_bitlen(mechanism, &ctx->sc_digest_bitlen);
	if (error != CRYPTO_SUCCESS)
	return (error);
	SKEIN_OP(ctx, InitExt, ctx->sc_digest_bitlen, 0, key->ck_data,
	CRYPTO_BITS2BYTES(key->ck_length));

	return (CRYPTO_SUCCESS);
	}

	/*
	* KCF software provide mac entry points.
	*/
	/*
	* Initializes a skein MAC context. You may pass a ctx_template, in which
	* case the template will be reused to make initialization more efficient.
	* Otherwise a new context will be constructed. The mechanism cm_type must
	* be one of SKEIN_*_MAC_MECH_INFO_TYPE. Same as in skein_digest_init, you
	* may pass a skein_param_t in cm_param to configure the length of the
	* digest. The key must be in raw format.
	*/
	static int
	skein_mac_init(crypto_ctx_t ctx, crypto_mechanism_t mechanism,
	crypto_key_t *key, crypto_spi_ctx_template_t ctx_template)
	{
	int error;

	SKEIN_CTX_LVALUE(ctx) = kmem_alloc(sizeof (*SKEIN_CTX(ctx)), KM_SLEEP);
	if (SKEIN_CTX(ctx) == NULL)
	return (CRYPTO_HOST_MEMORY);

	if (ctx_template != NULL) {
	memcpy(SKEIN_CTX(ctx), ctx_template,
	sizeof (*SKEIN_CTX(ctx)));
	} else {
	error = skein_mac_ctx_build(SKEIN_CTX(ctx), mechanism, key);
	if (error != CRYPTO_SUCCESS)
	goto errout;
	}

	return (CRYPTO_SUCCESS);
	errout:
	memset(SKEIN_CTX(ctx), 0, sizeof (*SKEIN_CTX(ctx)));
	kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
	return (error);
	}

	/*
	* The MAC update and final calls are reused from the regular digest code.
	*/

	/*
	* Same as skein_digest_atomic, performs an atomic Skein MAC operation in
	* one step. All the same properties apply to the arguments of this
	* function as to those of the partial operations above.
	*/
	static int
	skein_mac_atomic(crypto_mechanism_t *mechanism,
	crypto_key_t key, crypto_data_t data, crypto_data_t *mac,
	crypto_spi_ctx_template_t ctx_template)
	{
	/* faux crypto context just for skein_digest_{update,final} */
	int error;
	crypto_ctx_t ctx;
	skein_ctx_t skein_ctx;
	SKEIN_CTX_LVALUE(&ctx) = &skein_ctx;

	if (ctx_template != NULL) {
	memcpy(&skein_ctx, ctx_template, sizeof (skein_ctx));
	} else {
	error = skein_mac_ctx_build(&skein_ctx, mechanism, key);
	if (error != CRYPTO_SUCCESS)
	goto errout;
	}

	if ((error = skein_update(&ctx, data)) != CRYPTO_SUCCESS)
	goto errout;
	- if ((error = skein_final(&ctx, mac)) != CRYPTO_SUCCESS)
	+ if ((error = skein_final_nofree(&ctx, mac)) != CRYPTO_SUCCESS)
	goto errout;

	return (CRYPTO_SUCCESS);
	errout:
	memset(&skein_ctx, 0, sizeof (skein_ctx));
	return (error);
	}

	/*
	* KCF software provider context management entry points.
	*/

	/*
	* Constructs a context template for the Skein MAC algorithm. The same
	* properties apply to the arguments of this function as to those of
	* skein_mac_init.
	*/
	static int
	skein_create_ctx_template(crypto_mechanism_t mechanism, crypto_key_t key,
	crypto_spi_ctx_template_t ctx_template, size_t ctx_template_size)
	{
	int error;
	skein_ctx_t *ctx_tmpl;

	ctx_tmpl = kmem_alloc(sizeof (*ctx_tmpl), KM_SLEEP);
	if (ctx_tmpl == NULL)
	return (CRYPTO_HOST_MEMORY);
	error = skein_mac_ctx_build(ctx_tmpl, mechanism, key);
	if (error != CRYPTO_SUCCESS)
	goto errout;
	*ctx_template = ctx_tmpl;
	ctx_template_size = sizeof (ctx_tmpl);

	return (CRYPTO_SUCCESS);
	errout:
	memset(ctx_tmpl, 0, sizeof (*ctx_tmpl));
	kmem_free(ctx_tmpl, sizeof (*ctx_tmpl));
	return (error);
	}

	/*
	* Frees a skein context in a parent crypto context.
	*/
	static int
	skein_free_context(crypto_ctx_t *ctx)
	{
	if (SKEIN_CTX(ctx) != NULL) {
	memset(SKEIN_CTX(ctx), 0, sizeof (*SKEIN_CTX(ctx)));
	kmem_free(SKEIN_CTX(ctx), sizeof (*SKEIN_CTX(ctx)));
	SKEIN_CTX_LVALUE(ctx) = NULL;
	}

	return (CRYPTO_SUCCESS);
	}
	diff --git a/sys/contrib/openzfs/module/lua/ldo.c b/sys/contrib/openzfs/module/lua/ldo.c
	index b9368d9ceab6..24677596de12 100644
	--- a/sys/contrib/openzfs/module/lua/ldo.c
	+++ b/sys/contrib/openzfs/module/lua/ldo.c
	@@ -1,758 +1,758 @@
	/*
	** $Id: ldo.c,v 2.108.1.3 2013/11/08 18:22:50 roberto Exp $
	** Stack and Call structure of Lua
	** See Copyright Notice in lua.h
	*/


	#define ldo_c
	#define LUA_CORE

	#include <sys/lua/lua.h>

	#include "lapi.h"
	#include "ldebug.h"
	#include "ldo.h"
	#include "lfunc.h"
	#include "lgc.h"
	#include "lmem.h"
	#include "lobject.h"
	#include "lopcodes.h"
	#include "lparser.h"
	#include "lstate.h"
	#include "lstring.h"
	#include "ltable.h"
	#include "ltm.h"
	#include "lvm.h"
	#include "lzio.h"



	/* Return the number of bytes available on the stack. */
	#if defined (_KERNEL) && defined(__linux__)
	#include <asm/current.h>
	static intptr_t stack_remaining(void) {
	intptr_t local;
	local = (intptr_t)&local - (intptr_t)current->stack;
	return local;
	}
	#elif defined (_KERNEL) && defined(__FreeBSD__)
	#include <sys/pcpu.h>
	static intptr_t stack_remaining(void) {
	intptr_t local;
	local = (intptr_t)&local - (intptr_t)curthread->td_kstack;
	return local;
	}
	#else
	static intptr_t stack_remaining(void) {
	return INTPTR_MAX;
	}
	#endif

	/*
	** {======================================================
	** Error-recovery functions
	** =======================================================
	*/

	/*
	** LUAI_THROW/LUAI_TRY define how Lua does exception handling. By
	** default, Lua handles errors with exceptions when compiling as
	** C++ code, with _longjmp/_setjmp when asked to use them, and with
	** longjmp/setjmp otherwise.
	*/
	#if !defined(LUAI_THROW)

	#ifdef _KERNEL

	#ifdef __linux__
	#if defined(__i386__)
	#define JMP_BUF_CNT 6
	#elif defined(__x86_64__)
	#define JMP_BUF_CNT 8
	#elif defined(__sparc__) && defined(__arch64__)
	#define JMP_BUF_CNT 6
	#elif defined(__powerpc__)
	#define JMP_BUF_CNT 26
	#elif defined(__aarch64__)
	#define JMP_BUF_CNT 64
	#elif defined(__arm__)
	#define JMP_BUF_CNT 65
	#elif defined(__mips__)
	#define JMP_BUF_CNT 12
	#elif defined(__s390x__)
	#define JMP_BUF_CNT 18
	#elif defined(__riscv)
	#define JMP_BUF_CNT 64
	#else
	#define JMP_BUF_CNT 1
	#endif

	typedef struct _label_t { long long unsigned val[JMP_BUF_CNT]; } label_t;

	int setjmp(label_t *) __attribute__ ((__nothrow__));
	extern __attribute__((noreturn)) void longjmp(label_t *);

	#define LUAI_THROW(L,c) longjmp(&(c)->b)
	#define LUAI_TRY(L,c,a) if (setjmp(&(c)->b) == 0) { a }
	#define luai_jmpbuf label_t

	/* unsupported arches will build but not be able to run lua programs */
	#if JMP_BUF_CNT == 1
	int setjmp (label_t *buf) {
	return 1;
	}

	void longjmp (label_t * buf) {
	for (;;);
	}
	#endif
	#else
	#define LUAI_THROW(L,c) longjmp((c)->b, 1)
	#define LUAI_TRY(L,c,a) if (setjmp((c)->b) == 0) { a }
	#define luai_jmpbuf jmp_buf
	#endif

	#else /* _KERNEL */

	#if defined(__cplusplus) && !defined(LUA_USE_LONGJMP)
	/* C++ exceptions */
	#define LUAI_THROW(L,c) throw(c)
	#define LUAI_TRY(L,c,a) \
	try { a } catch(...) { if ((c)->status == 0) (c)->status = -1; }
	#define luai_jmpbuf int /* dummy variable */

	#elif defined(LUA_USE_ULONGJMP)
	/* in Unix, try _longjmp/_setjmp (more efficient) */
	#define LUAI_THROW(L,c) _longjmp((c)->b, 1)
	#define LUAI_TRY(L,c,a) if (_setjmp((c)->b) == 0) { a }
	#define luai_jmpbuf jmp_buf

	#else
	/* default handling with long jumps */
	#define LUAI_THROW(L,c) longjmp((c)->b, 1)
	#define LUAI_TRY(L,c,a) if (setjmp((c)->b) == 0) { a }
	#define luai_jmpbuf jmp_buf

	#endif

	#endif /* _KERNEL */

	#endif /* LUAI_THROW */


	/* chain list of long jump buffers */
	struct lua_longjmp {
	struct lua_longjmp *previous;
	luai_jmpbuf b;
	volatile int status; /* error code */
	};


	static void seterrorobj (lua_State *L, int errcode, StkId oldtop) {
	switch (errcode) {
	case LUA_ERRMEM: { /* memory error? */
	setsvalue2s(L, oldtop, G(L)->memerrmsg); /* reuse preregistered msg. */
	break;
	}
	case LUA_ERRERR: {
	setsvalue2s(L, oldtop, luaS_newliteral(L, "error in error handling"));
	break;
	}
	default: {
	setobjs2s(L, oldtop, L->top - 1); /* error message on current top */
	break;
	}
	}
	L->top = oldtop + 1;
	}

	/*
	* Silence infinite recursion warning which was added to -Wall in gcc 12.1
	*/
	#if defined(HAVE_INFINITE_RECURSION)
	#pragma GCC diagnostic push
	#pragma GCC diagnostic ignored "-Winfinite-recursion"
	#endif

	l_noret luaD_throw (lua_State *L, int errcode) {
	if (L->errorJmp) { /* thread has an error handler? */
	L->errorJmp->status = errcode; /* set status */
	LUAI_THROW(L, L->errorJmp); /* jump to it */
	}
	else { /* thread has no error handler */
	L->status = cast_byte(errcode); /* mark it as dead */
	if (G(L)->mainthread->errorJmp) { /* main thread has a handler? */
	setobjs2s(L, G(L)->mainthread->top++, L->top - 1); /* copy error obj. */
	luaD_throw(G(L)->mainthread, errcode); /* re-throw in main thread */
	}
	else { /* no handler at all; abort */
	if (G(L)->panic) { /* panic function? */
	lua_unlock(L);
	G(L)->panic(L); /* call it (last chance to jump out) */
	}
	panic("no error handler");
	}
	}
	}

	#if defined(HAVE_INFINITE_RECURSION)
	#pragma GCC diagnostic pop
	#endif


	int luaD_rawrunprotected (lua_State L, Pfunc f, void ud) {
	unsigned short oldnCcalls = L->nCcalls;
	struct lua_longjmp lj;
	lj.status = LUA_OK;
	lj.previous = L->errorJmp; /* chain new error handler */
	L->errorJmp = &lj;
	LUAI_TRY(L, &lj,
	(*f)(L, ud);
	);
	L->errorJmp = lj.previous; /* restore old error handler */
	L->nCcalls = oldnCcalls;
	return lj.status;
	}

	/* }====================================================== */


	static void correctstack (lua_State L, TValue oldstack) {
	CallInfo *ci;
	GCObject *up;
	L->top = (L->top - oldstack) + L->stack;
	for (up = L->openupval; up != NULL; up = up->gch.next)
	gco2uv(up)->v = (gco2uv(up)->v - oldstack) + L->stack;
	for (ci = L->ci; ci != NULL; ci = ci->previous) {
	ci->top = (ci->top - oldstack) + L->stack;
	ci->func = (ci->func - oldstack) + L->stack;
	if (isLua(ci))
	ci->u.l.base = (ci->u.l.base - oldstack) + L->stack;
	}
	}


	/* some space for error handling */
	#define ERRORSTACKSIZE (LUAI_MAXSTACK + 200)


	void luaD_reallocstack (lua_State *L, int newsize) {
	TValue *oldstack = L->stack;
	int lim = L->stacksize;
	lua_assert(newsize <= LUAI_MAXSTACK \|\| newsize == ERRORSTACKSIZE);
	lua_assert(L->stack_last - L->stack == L->stacksize - EXTRA_STACK);
	luaM_reallocvector(L, L->stack, L->stacksize, newsize, TValue);
	for (; lim < newsize; lim++)
	setnilvalue(L->stack + lim); /* erase new segment */
	L->stacksize = newsize;
	L->stack_last = L->stack + newsize - EXTRA_STACK;
	correctstack(L, oldstack);
	}


	void luaD_growstack (lua_State *L, int n) {
	int size = L->stacksize;
	if (size > LUAI_MAXSTACK) /* error after extra size? */
	luaD_throw(L, LUA_ERRERR);
	else {
	int needed = cast_int(L->top - L->stack) + n + EXTRA_STACK;
	int newsize = 2 * size;
	if (newsize > LUAI_MAXSTACK) newsize = LUAI_MAXSTACK;
	if (newsize < needed) newsize = needed;
	if (newsize > LUAI_MAXSTACK) { /* stack overflow? */
	luaD_reallocstack(L, ERRORSTACKSIZE);
	luaG_runerror(L, "stack overflow");
	}
	else
	luaD_reallocstack(L, newsize);
	}
	}


	static int stackinuse (lua_State *L) {
	CallInfo *ci;
	StkId lim = L->top;
	for (ci = L->ci; ci != NULL; ci = ci->previous) {
	lua_assert(ci->top <= L->stack_last);
	if (lim < ci->top) lim = ci->top;
	}
	return cast_int(lim - L->stack) + 1; /* part of stack in use */
	}


	void luaD_shrinkstack (lua_State *L) {
	int inuse = stackinuse(L);
	int goodsize = inuse + (inuse / 8) + 2*EXTRA_STACK;
	if (goodsize > LUAI_MAXSTACK) goodsize = LUAI_MAXSTACK;
	if (inuse > LUAI_MAXSTACK \|\| /* handling stack overflow? */
	goodsize >= L->stacksize) /* would grow instead of shrink? */
	condmovestack(L); /* don't change stack (change only for debugging) */
	else
	luaD_reallocstack(L, goodsize); /* shrink it */
	}


	void luaD_hook (lua_State *L, int event, int line) {
	lua_Hook hook = L->hook;
	if (hook && L->allowhook) {
	CallInfo *ci = L->ci;
	ptrdiff_t top = savestack(L, L->top);
	ptrdiff_t ci_top = savestack(L, ci->top);
	lua_Debug ar;
	ar.event = event;
	ar.currentline = line;
	ar.i_ci = ci;
	luaD_checkstack(L, LUA_MINSTACK); /* ensure minimum stack size */
	ci->top = L->top + LUA_MINSTACK;
	lua_assert(ci->top <= L->stack_last);
	L->allowhook = 0; /* cannot call hooks inside a hook */
	ci->callstatus \|= CIST_HOOKED;
	lua_unlock(L);
	(*hook)(L, &ar);
	lua_lock(L);
	lua_assert(!L->allowhook);
	L->allowhook = 1;
	ci->top = restorestack(L, ci_top);
	L->top = restorestack(L, top);
	ci->callstatus &= ~CIST_HOOKED;
	}
	}


	static void callhook (lua_State L, CallInfo ci) {
	int hook = LUA_HOOKCALL;
	ci->u.l.savedpc++; /* hooks assume 'pc' is already incremented */
	if (isLua(ci->previous) &&
	GET_OPCODE(*(ci->previous->u.l.savedpc - 1)) == OP_TAILCALL) {
	ci->callstatus \|= CIST_TAIL;
	hook = LUA_HOOKTAILCALL;
	}
	luaD_hook(L, hook, -1);
	ci->u.l.savedpc--; /* correct 'pc' */
	}


	static StkId adjust_varargs (lua_State L, Proto p, int actual) {
	int i;
	int nfixargs = p->numparams;
	StkId base, fixed;
	lua_assert(actual >= nfixargs);
	/* move fixed parameters to final position */
	luaD_checkstack(L, p->maxstacksize); /* check again for new 'base' */
	fixed = L->top - actual; /* first fixed argument */
	base = L->top; /* final position of first argument */
	for (i=0; i<nfixargs; i++) {
	setobjs2s(L, L->top++, fixed + i);
	setnilvalue(fixed + i);
	}
	return base;
	}


	static StkId tryfuncTM (lua_State *L, StkId func) {
	const TValue *tm = luaT_gettmbyobj(L, func, TM_CALL);
	StkId p;
	ptrdiff_t funcr = savestack(L, func);
	if (!ttisfunction(tm))
	luaG_typeerror(L, func, "call");
	/* Open a hole inside the stack at `func' */
	for (p = L->top; p > func; p--) setobjs2s(L, p, p-1);
	incr_top(L);
	func = restorestack(L, funcr); /* previous call may change stack */
	setobj2s(L, func, tm); /* tag method is the new function to be called */
	return func;
	}



	#define next_ci(L) (L->ci = (L->ci->next ? L->ci->next : luaE_extendCI(L)))


	/*
	** returns true if function has been executed (C function)
	*/
	int luaD_precall (lua_State *L, StkId func, int nresults) {
	lua_CFunction f;
	CallInfo *ci;
	int n; /* number of arguments (Lua) or returns (C) */
	ptrdiff_t funcr = savestack(L, func);
	switch (ttype(func)) {
	case LUA_TLCF: /* light C function */
	f = fvalue(func);
	goto Cfunc;
	case LUA_TCCL: { /* C closure */
	f = clCvalue(func)->f;
	Cfunc:
	luaD_checkstack(L, LUA_MINSTACK); /* ensure minimum stack size */
	ci = next_ci(L); /* now 'enter' new function */
	ci->nresults = nresults;
	ci->func = restorestack(L, funcr);
	ci->top = L->top + LUA_MINSTACK;
	lua_assert(ci->top <= L->stack_last);
	ci->callstatus = 0;
	luaC_checkGC(L); /* stack grow uses memory */
	if (L->hookmask & LUA_MASKCALL)
	luaD_hook(L, LUA_HOOKCALL, -1);
	lua_unlock(L);
	n = (f)(L); / do the actual call */
	lua_lock(L);
	api_checknelems(L, n);
	luaD_poscall(L, L->top - n);
	return 1;
	}
	case LUA_TLCL: { /* Lua function: prepare its call */
	StkId base;
	Proto *p = clLvalue(func)->p;
	n = cast_int(L->top - func) - 1; /* number of real arguments */
	- luaD_checkstack(L, p->maxstacksize);
	+ luaD_checkstack(L, p->maxstacksize + p->numparams);
	for (; n < p->numparams; n++)
	setnilvalue(L->top++); /* complete missing arguments */
	if (!p->is_vararg) {
	func = restorestack(L, funcr);
	base = func + 1;
	}
	else {
	base = adjust_varargs(L, p, n);
	func = restorestack(L, funcr); /* previous call can change stack */
	}
	ci = next_ci(L); /* now 'enter' new function */
	ci->nresults = nresults;
	ci->func = func;
	ci->u.l.base = base;
	ci->top = base + p->maxstacksize;
	lua_assert(ci->top <= L->stack_last);
	ci->u.l.savedpc = p->code; /* starting point */
	ci->callstatus = CIST_LUA;
	L->top = ci->top;
	luaC_checkGC(L); /* stack grow uses memory */
	if (L->hookmask & LUA_MASKCALL)
	callhook(L, ci);
	return 0;
	}
	default: { /* not a function */
	func = tryfuncTM(L, func); /* retry with 'function' tag method */
	return luaD_precall(L, func, nresults); /* now it must be a function */
	}
	}
	}


	int luaD_poscall (lua_State *L, StkId firstResult) {
	StkId res;
	int wanted, i;
	CallInfo *ci = L->ci;
	if (L->hookmask & (LUA_MASKRET \| LUA_MASKLINE)) {
	if (L->hookmask & LUA_MASKRET) {
	ptrdiff_t fr = savestack(L, firstResult); /* hook may change stack */
	luaD_hook(L, LUA_HOOKRET, -1);
	firstResult = restorestack(L, fr);
	}
	L->oldpc = ci->previous->u.l.savedpc; /* 'oldpc' for caller function */
	}
	res = ci->func; /* res == final position of 1st result */
	wanted = ci->nresults;
	L->ci = ci = ci->previous; /* back to caller */
	/* move results to correct place */
	for (i = wanted; i != 0 && firstResult < L->top; i--)
	setobjs2s(L, res++, firstResult++);
	while (i-- > 0)
	setnilvalue(res++);
	L->top = res;
	return (wanted - LUA_MULTRET); /* 0 iff wanted == LUA_MULTRET */
	}


	/*
	** Call a function (C or Lua). The function to be called is at *func.
	** The arguments are on the stack, right after the function.
	** When returns, all the results are on the stack, starting at the original
	** function position.
	*/
	void luaD_call (lua_State *L, StkId func, int nResults, int allowyield) {
	if (++L->nCcalls >= LUAI_MAXCCALLS) {
	if (L->nCcalls == LUAI_MAXCCALLS)
	luaG_runerror(L, "C stack overflow");
	else if (L->nCcalls >= (LUAI_MAXCCALLS + (LUAI_MAXCCALLS>>3)))
	luaD_throw(L, LUA_ERRERR); /* error while handling stack error */
	}
	intptr_t remaining = stack_remaining();
	if (L->runerror == 0 && remaining < LUAI_MINCSTACK)
	luaG_runerror(L, "C stack overflow");
	if (L->runerror != 0 && remaining < LUAI_MINCSTACK / 2)
	luaD_throw(L, LUA_ERRERR); /* error while handling stack error */
	if (!allowyield) L->nny++;
	if (!luaD_precall(L, func, nResults)) /* is a Lua function? */
	luaV_execute(L); /* call it */
	if (!allowyield) L->nny--;
	L->nCcalls--;
	}


	static void finishCcall (lua_State *L) {
	CallInfo *ci = L->ci;
	int n;
	lua_assert(ci->u.c.k != NULL); /* must have a continuation */
	lua_assert(L->nny == 0);
	if (ci->callstatus & CIST_YPCALL) { /* was inside a pcall? */
	ci->callstatus &= ~CIST_YPCALL; /* finish 'lua_pcall' */
	L->errfunc = ci->u.c.old_errfunc;
	}
	/* finish 'lua_callk'/'lua_pcall' */
	adjustresults(L, ci->nresults);
	/* call continuation function */
	if (!(ci->callstatus & CIST_STAT)) /* no call status? */
	ci->u.c.status = LUA_YIELD; /* 'default' status */
	lua_assert(ci->u.c.status != LUA_OK);
	ci->callstatus = (ci->callstatus & ~(CIST_YPCALL \| CIST_STAT)) \| CIST_YIELDED;
	lua_unlock(L);
	n = (*ci->u.c.k)(L);
	lua_lock(L);
	api_checknelems(L, n);
	/* finish 'luaD_precall' */
	luaD_poscall(L, L->top - n);
	}


	static void unroll (lua_State L, void ud) {
	UNUSED(ud);
	for (;;) {
	if (L->ci == &L->base_ci) /* stack is empty? */
	return; /* coroutine finished normally */
	if (!isLua(L->ci)) /* C function? */
	finishCcall(L);
	else { /* Lua function */
	luaV_finishOp(L); /* finish interrupted instruction */
	luaV_execute(L); /* execute down to higher C 'boundary' */
	}
	}
	}


	/*
	** check whether thread has a suspended protected call
	*/
	static CallInfo findpcall (lua_State L) {
	CallInfo *ci;
	for (ci = L->ci; ci != NULL; ci = ci->previous) { /* search for a pcall */
	if (ci->callstatus & CIST_YPCALL)
	return ci;
	}
	return NULL; /* no pending pcall */
	}


	static int recover (lua_State *L, int status) {
	StkId oldtop;
	CallInfo *ci = findpcall(L);
	if (ci == NULL) return 0; /* no recovery point */
	/* "finish" luaD_pcall */
	oldtop = restorestack(L, ci->extra);
	luaF_close(L, oldtop);
	seterrorobj(L, status, oldtop);
	L->ci = ci;
	L->allowhook = ci->u.c.old_allowhook;
	L->nny = 0; /* should be zero to be yieldable */
	luaD_shrinkstack(L);
	L->errfunc = ci->u.c.old_errfunc;
	ci->callstatus \|= CIST_STAT; /* call has error status */
	ci->u.c.status = status; /* (here it is) */
	return 1; /* continue running the coroutine */
	}


	/*
	** signal an error in the call to 'resume', not in the execution of the
	** coroutine itself. (Such errors should not be handled by any coroutine
	** error handler and should not kill the coroutine.)
	*/
	static l_noret resume_error (lua_State L, const char msg, StkId firstArg) {
	L->top = firstArg; /* remove args from the stack */
	setsvalue2s(L, L->top, luaS_new(L, msg)); /* push error message */
	api_incr_top(L);
	luaD_throw(L, -1); /* jump back to 'lua_resume' */
	}


	/*
	** do the work for 'lua_resume' in protected mode
	*/
	static void resume_cb (lua_State L, void ud) {
	int nCcalls = L->nCcalls;
	StkId firstArg = cast(StkId, ud);
	CallInfo *ci = L->ci;
	if (nCcalls >= LUAI_MAXCCALLS)
	resume_error(L, "C stack overflow", firstArg);
	if (L->status == LUA_OK) { /* may be starting a coroutine */
	if (ci != &L->base_ci) /* not in base level? */
	resume_error(L, "cannot resume non-suspended coroutine", firstArg);
	/* coroutine is in base level; start running it */
	if (!luaD_precall(L, firstArg - 1, LUA_MULTRET)) /* Lua function? */
	luaV_execute(L); /* call it */
	}
	else if (L->status != LUA_YIELD)
	resume_error(L, "cannot resume dead coroutine", firstArg);
	else { /* resuming from previous yield */
	L->status = LUA_OK;
	ci->func = restorestack(L, ci->extra);
	if (isLua(ci)) /* yielded inside a hook? */
	luaV_execute(L); /* just continue running Lua code */
	else { /* 'common' yield */
	if (ci->u.c.k != NULL) { /* does it have a continuation? */
	int n;
	ci->u.c.status = LUA_YIELD; /* 'default' status */
	ci->callstatus \|= CIST_YIELDED;
	lua_unlock(L);
	n = (ci->u.c.k)(L); / call continuation */
	lua_lock(L);
	api_checknelems(L, n);
	firstArg = L->top - n; /* yield results come from continuation */
	}
	luaD_poscall(L, firstArg); /* finish 'luaD_precall' */
	}
	unroll(L, NULL);
	}
	lua_assert(nCcalls == L->nCcalls);
	}


	LUA_API int lua_resume (lua_State L, lua_State from, int nargs) {
	int status;
	int oldnny = L->nny; /* save 'nny' */
	lua_lock(L);
	luai_userstateresume(L, nargs);
	L->nCcalls = (from) ? from->nCcalls + 1 : 1;
	L->nny = 0; /* allow yields */
	api_checknelems(L, (L->status == LUA_OK) ? nargs + 1 : nargs);
	status = luaD_rawrunprotected(L, resume_cb, L->top - nargs);
	if (status == -1) /* error calling 'lua_resume'? */
	status = LUA_ERRRUN;
	else { /* yield or regular error */
	while (status != LUA_OK && status != LUA_YIELD) { /* error? */
	if (recover(L, status)) /* recover point? */
	status = luaD_rawrunprotected(L, unroll, NULL); /* run continuation */
	else { /* unrecoverable error */
	L->status = cast_byte(status); /* mark thread as `dead' */
	seterrorobj(L, status, L->top);
	L->ci->top = L->top;
	break;
	}
	}
	lua_assert(status == L->status);
	}
	L->nny = oldnny; /* restore 'nny' */
	L->nCcalls--;
	lua_assert(L->nCcalls == ((from) ? from->nCcalls : 0));
	lua_unlock(L);
	return status;
	}


	LUA_API int lua_yieldk (lua_State *L, int nresults, int ctx, lua_CFunction k) {
	CallInfo *ci = L->ci;
	luai_userstateyield(L, nresults);
	lua_lock(L);
	api_checknelems(L, nresults);
	if (L->nny > 0) {
	if (L != G(L)->mainthread)
	luaG_runerror(L, "attempt to yield across a C-call boundary");
	else
	luaG_runerror(L, "attempt to yield from outside a coroutine");
	}
	L->status = LUA_YIELD;
	ci->extra = savestack(L, ci->func); /* save current 'func' */
	if (isLua(ci)) { /* inside a hook? */
	api_check(L, k == NULL, "hooks cannot continue after yielding");
	}
	else {
	if ((ci->u.c.k = k) != NULL) /* is there a continuation? */
	ci->u.c.ctx = ctx; /* save context */
	ci->func = L->top - nresults - 1; /* protect stack below results */
	luaD_throw(L, LUA_YIELD);
	}
	lua_assert(ci->callstatus & CIST_HOOKED); /* must be inside a hook */
	lua_unlock(L);
	return 0; /* return to 'luaD_hook' */
	}


	int luaD_pcall (lua_State L, Pfunc func, void u,
	ptrdiff_t old_top, ptrdiff_t ef) {
	int status;
	CallInfo *old_ci = L->ci;
	lu_byte old_allowhooks = L->allowhook;
	unsigned short old_nny = L->nny;
	ptrdiff_t old_errfunc = L->errfunc;
	L->errfunc = ef;
	status = luaD_rawrunprotected(L, func, u);
	if (status != LUA_OK) { /* an error occurred? */
	StkId oldtop = restorestack(L, old_top);
	luaF_close(L, oldtop); /* close possible pending closures */
	seterrorobj(L, status, oldtop);
	L->ci = old_ci;
	L->allowhook = old_allowhooks;
	L->nny = old_nny;
	luaD_shrinkstack(L);
	}
	L->errfunc = old_errfunc;
	return status;
	}



	/*
	** Execute a protected parser.
	*/
	struct SParser { /* data to `f_parser' */
	ZIO *z;
	Mbuffer buff; /* dynamic structure used by the scanner */
	Dyndata dyd; /* dynamic structures used by the parser */
	const char *mode;
	const char *name;
	};


	static void checkmode (lua_State L, const char mode, const char *x) {
	if (mode && strchr(mode, x[0]) == NULL) {
	luaO_pushfstring(L,
	"attempt to load a %s chunk (mode is " LUA_QS ")", x, mode);
	luaD_throw(L, LUA_ERRSYNTAX);
	}
	}


	static void f_parser (lua_State L, void ud) {
	int i;
	Closure *cl;
	struct SParser p = cast(struct SParser , ud);
	int c = zgetc(p->z); /* read first character */
	lua_assert(c != LUA_SIGNATURE[0]); /* binary not supported */
	checkmode(L, p->mode, "text");
	cl = luaY_parser(L, p->z, &p->buff, &p->dyd, p->name, c);
	lua_assert(cl->l.nupvalues == cl->l.p->sizeupvalues);
	for (i = 0; i < cl->l.nupvalues; i++) { /* initialize upvalues */
	UpVal *up = luaF_newupval(L);
	cl->l.upvals[i] = up;
	luaC_objbarrier(L, cl, up);
	}
	}


	int luaD_protectedparser (lua_State L, ZIO z, const char *name,
	const char *mode) {
	struct SParser p;
	int status;
	L->nny++; /* cannot yield during parsing */
	p.z = z; p.name = name; p.mode = mode;
	p.dyd.actvar.arr = NULL; p.dyd.actvar.size = 0;
	p.dyd.gt.arr = NULL; p.dyd.gt.size = 0;
	p.dyd.label.arr = NULL; p.dyd.label.size = 0;
	luaZ_initbuffer(L, &p.buff);
	status = luaD_pcall(L, f_parser, &p, savestack(L, L->top), L->errfunc);
	luaZ_freebuffer(L, &p.buff);
	luaM_freearray(L, p.dyd.actvar.arr, p.dyd.actvar.size);
	luaM_freearray(L, p.dyd.gt.arr, p.dyd.gt.size);
	luaM_freearray(L, p.dyd.label.arr, p.dyd.label.size);
	L->nny--;
	return status;
	}
	diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/callb.c b/sys/contrib/openzfs/module/os/freebsd/spl/callb.c
	index ba13ea887938..47f3ccc0c7fa 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/spl/callb.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/spl/callb.c
	@@ -1,373 +1,372 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/sysmacros.h>
	#include <sys/systm.h>
	#include <sys/proc.h>
	#include <sys/mutex.h>
	#include <sys/condvar.h>
	#include <sys/callb.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/debug.h>
	#include <sys/kobj.h>
	#include <sys/systm.h> /* for delay() */
	#include <sys/taskq.h> /* For TASKQ_NAMELEN */
	#include <sys/kernel.h>

	#define CB_MAXNAME TASKQ_NAMELEN

	/*
	* The callb mechanism provides generic event scheduling/echoing.
	* A callb function is registered and called on behalf of the event.
	*/
	typedef struct callb {
	struct callb c_next; / next in class or on freelist */
	kthread_id_t c_thread; /* ptr to caller's thread struct */
	char c_flag; /* info about the callb state */
	uchar_t c_class; /* this callb's class */
	kcondvar_t c_done_cv; /* signal callb completion */
	boolean_t (c_func)(void , int);
	/* cb function: returns true if ok */
	void c_arg; / arg to c_func */
	char c_name[CB_MAXNAME+1]; /* debug:max func name length */
	} callb_t;

	/*
	* callb c_flag bitmap definitions
	*/
	#define CALLB_FREE 0x0
	#define CALLB_TAKEN 0x1
	#define CALLB_EXECUTING 0x2

	/*
	* Basic structure for a callb table.
	* All callbs are organized into different class groups described
	* by ct_class array.
	* The callbs within a class are single-linked and normally run by a
	* serial execution.
	*/
	typedef struct callb_table {
	kmutex_t ct_lock; /* protect all callb states */
	callb_t ct_freelist; / free callb structures */
	boolean_t ct_busy; /* B_TRUE prevents additions */
	kcondvar_t ct_busy_cv; /* to wait for not busy */
	int ct_ncallb; /* num of callbs allocated */
	callb_t ct_first_cb[NCBCLASS]; / ptr to 1st callb in a class */
	} callb_table_t;

	int callb_timeout_sec = CPR_KTHREAD_TIMEOUT_SEC;

	static callb_id_t callb_add_common(boolean_t ()(void , int),
	void , int, char , kthread_id_t);

	static callb_table_t callb_table; /* system level callback table */
	static callb_table_t *ct = &callb_table;
	static kmutex_t callb_safe_mutex;
	callb_cpr_t callb_cprinfo_safe = {
	&callb_safe_mutex, CALLB_CPR_ALWAYS_SAFE, 0, {0, 0} };

	/*
	* Init all callb tables in the system.
	*/
	static void
	callb_init(void *dummy __unused)
	{
	callb_table.ct_busy = B_FALSE; /* mark table open for additions */
	mutex_init(&callb_safe_mutex, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&callb_table.ct_lock, NULL, MUTEX_DEFAULT, NULL);
	}

	static void
	callb_fini(void *dummy __unused)
	{
	callb_t *cp;
	int i;

	mutex_enter(&ct->ct_lock);
	for (i = 0; i < 16; i++) {
	while ((cp = ct->ct_freelist) != NULL) {
	ct->ct_freelist = cp->c_next;
	ct->ct_ncallb--;
	kmem_free(cp, sizeof (callb_t));
	}
	if (ct->ct_ncallb == 0)
	break;
	/* Not all callbacks finished, waiting for the rest. */
	mutex_exit(&ct->ct_lock);
	tsleep(ct, 0, "callb", hz / 4);
	mutex_enter(&ct->ct_lock);
	}
	if (ct->ct_ncallb > 0)
	printf("%s: Leaked %d callbacks!\n", __func__, ct->ct_ncallb);
	mutex_exit(&ct->ct_lock);
	mutex_destroy(&callb_safe_mutex);
	mutex_destroy(&callb_table.ct_lock);
	}

	/*
	* callout_add() is called to register func() be called later.
	*/
	static callb_id_t
	callb_add_common(boolean_t (func)(void arg, int code),
	void arg, int class, char name, kthread_id_t t)
	{
	callb_t *cp;

	ASSERT3S(class, <, NCBCLASS);

	mutex_enter(&ct->ct_lock);
	while (ct->ct_busy)
	cv_wait(&ct->ct_busy_cv, &ct->ct_lock);
	if ((cp = ct->ct_freelist) == NULL) {
	ct->ct_ncallb++;
	cp = (callb_t *)kmem_zalloc(sizeof (callb_t), KM_SLEEP);
	}
	ct->ct_freelist = cp->c_next;
	cp->c_thread = t;
	cp->c_func = func;
	cp->c_arg = arg;
	cp->c_class = (uchar_t)class;
	cp->c_flag \|= CALLB_TAKEN;
	#ifdef ZFS_DEBUG
	if (strlen(name) > CB_MAXNAME)
	cmn_err(CE_WARN, "callb_add: name of callback function '%s' "
	"too long -- truncated to %d chars",
	name, CB_MAXNAME);
	#endif
	- (void) strncpy(cp->c_name, name, CB_MAXNAME);
	- cp->c_name[CB_MAXNAME] = '\0';
	+ (void) strlcpy(cp->c_name, name, sizeof (cp->c_name));

	/*
	* Insert the new callb at the head of its class list.
	*/
	cp->c_next = ct->ct_first_cb[class];
	ct->ct_first_cb[class] = cp;

	mutex_exit(&ct->ct_lock);
	return ((callb_id_t)cp);
	}

	/*
	* The default function to add an entry to the callback table. Since
	* it uses curthread as the thread identifier to store in the table,
	* it should be used for the normal case of a thread which is calling
	* to add ITSELF to the table.
	*/
	callb_id_t
	callb_add(boolean_t (func)(void arg, int code),
	void arg, int class, char name)
	{
	return (callb_add_common(func, arg, class, name, curthread));
	}

	/*
	* A special version of callb_add() above for use by threads which
	* might be adding an entry to the table on behalf of some other
	* thread (for example, one which is constructed but not yet running).
	* In this version the thread id is an argument.
	*/
	callb_id_t
	callb_add_thread(boolean_t (func)(void arg, int code),
	void arg, int class, char name, kthread_id_t t)
	{
	return (callb_add_common(func, arg, class, name, t));
	}

	/*
	* callout_delete() is called to remove an entry identified by id
	* that was originally placed there by a call to callout_add().
	* return -1 if fail to delete a callb entry otherwise return 0.
	*/
	int
	callb_delete(callb_id_t id)
	{
	callb_t **pp;
	callb_t me = (callb_t )id;

	mutex_enter(&ct->ct_lock);

	for (;;) {
	pp = &ct->ct_first_cb[me->c_class];
	while (pp != NULL && pp != me)
	pp = &(*pp)->c_next;

	#ifdef ZFS_DEBUG
	if (*pp != me) {
	cmn_err(CE_WARN, "callb delete bogus entry 0x%p",
	(void *)me);
	mutex_exit(&ct->ct_lock);
	return (-1);
	}
	#endif /* DEBUG */

	/*
	* It is not allowed to delete a callb in the middle of
	* executing otherwise, the callb_execute() will be confused.
	*/
	if (!(me->c_flag & CALLB_EXECUTING))
	break;

	cv_wait(&me->c_done_cv, &ct->ct_lock);
	}
	/* relink the class list */
	*pp = me->c_next;

	/* clean up myself and return the free callb to the head of freelist */
	me->c_flag = CALLB_FREE;
	me->c_next = ct->ct_freelist;
	ct->ct_freelist = me;

	mutex_exit(&ct->ct_lock);
	return (0);
	}

	/*
	* class: indicates to execute all callbs in the same class;
	* code: optional argument for the callb functions.
	* return: = 0: success
	* != 0: ptr to string supplied when callback was registered
	*/
	void *
	callb_execute_class(int class, int code)
	{
	callb_t *cp;
	void *ret = NULL;

	ASSERT3S(class, <, NCBCLASS);

	mutex_enter(&ct->ct_lock);

	for (cp = ct->ct_first_cb[class];
	cp != NULL && ret == 0; cp = cp->c_next) {
	while (cp->c_flag & CALLB_EXECUTING)
	cv_wait(&cp->c_done_cv, &ct->ct_lock);
	/*
	* cont if the callb is deleted while we're sleeping
	*/
	if (cp->c_flag == CALLB_FREE)
	continue;
	cp->c_flag \|= CALLB_EXECUTING;

	#ifdef CALLB_DEBUG
	printf("callb_execute: name=%s func=%p arg=%p\n",
	cp->c_name, (void )cp->c_func, (void )cp->c_arg);
	#endif /* CALLB_DEBUG */

	mutex_exit(&ct->ct_lock);
	/* If callback function fails, pass back client's name */
	if (!(*cp->c_func)(cp->c_arg, code))
	ret = cp->c_name;
	mutex_enter(&ct->ct_lock);

	cp->c_flag &= ~CALLB_EXECUTING;
	cv_broadcast(&cp->c_done_cv);
	}
	mutex_exit(&ct->ct_lock);
	return (ret);
	}

	/*
	* callers make sure no recursive entries to this func.
	* dp->cc_lockp is registered by callb_add to protect callb_cpr_t structure.
	*
	* When calling to stop a kernel thread (code == CB_CODE_CPR_CHKPT) we
	* use a cv_timedwait() in case the kernel thread is blocked.
	*
	* Note that this is a generic callback handler for daemon CPR and
	* should NOT be changed to accommodate any specific requirement in a daemon.
	* Individual daemons that require changes to the handler shall write
	* callback routines in their own daemon modules.
	*/
	boolean_t
	callb_generic_cpr(void *arg, int code)
	{
	callb_cpr_t cp = (callb_cpr_t )arg;
	clock_t ret = 0; /* assume success */

	mutex_enter(cp->cc_lockp);

	switch (code) {
	case CB_CODE_CPR_CHKPT:
	cp->cc_events \|= CALLB_CPR_START;
	#ifdef CPR_NOT_THREAD_SAFE
	while (!(cp->cc_events & CALLB_CPR_SAFE))
	/* cv_timedwait() returns -1 if it times out. */
	if ((ret = cv_reltimedwait(&cp->cc_callb_cv,
	cp->cc_lockp, (callb_timeout_sec * hz),
	TR_CLOCK_TICK)) == -1)
	break;
	#endif
	break;

	case CB_CODE_CPR_RESUME:
	cp->cc_events &= ~CALLB_CPR_START;
	cv_signal(&cp->cc_stop_cv);
	break;
	}
	mutex_exit(cp->cc_lockp);
	return (ret != -1);
	}

	/*
	* The generic callback function associated with kernel threads which
	* are always considered safe.
	*/
	boolean_t
	callb_generic_cpr_safe(void *arg, int code)
	{
	(void) arg, (void) code;
	return (B_TRUE);
	}
	/*
	* Prevent additions to callback table.
	*/
	void
	callb_lock_table(void)
	{
	mutex_enter(&ct->ct_lock);
	ASSERT(!ct->ct_busy);
	ct->ct_busy = B_TRUE;
	mutex_exit(&ct->ct_lock);
	}

	/*
	* Allow additions to callback table.
	*/
	void
	callb_unlock_table(void)
	{
	mutex_enter(&ct->ct_lock);
	ASSERT(ct->ct_busy);
	ct->ct_busy = B_FALSE;
	cv_broadcast(&ct->ct_busy_cv);
	mutex_exit(&ct->ct_lock);
	}

	SYSINIT(sol_callb, SI_SUB_DRIVERS, SI_ORDER_FIRST, callb_init, NULL);
	SYSUNINIT(sol_callb, SI_SUB_DRIVERS, SI_ORDER_FIRST, callb_fini, NULL);
	diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	index e46271a039de..e3653167323b 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	@@ -1,109 +1,109 @@
	/*
	* Copyright (c) 2007 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/jail.h>
	#include <sys/kernel.h>
	#include <sys/libkern.h>
	#include <sys/limits.h>
	#include <sys/misc.h>
	#include <sys/sysctl.h>
	#include <sys/vnode.h>

	#include <sys/zfs_context.h>

	static struct opensolaris_utsname hw_utsname = {
	.machine = MACHINE
	};

	utsname_t *
	utsname(void)
	{
	return (&hw_utsname);
	}

	static void
	opensolaris_utsname_init(void *arg)
	{

	hw_utsname.sysname = ostype;
	hw_utsname.nodename = prison0.pr_hostname;
	hw_utsname.release = osrelease;
	snprintf(hw_utsname.version, sizeof (hw_utsname.version),
	"%d", osreldate);
	}

	char *
	kmem_strdup(const char *s)
	{
	char *buf;

	buf = kmem_alloc(strlen(s) + 1, KM_SLEEP);
	strcpy(buf, s);
	return (buf);
	}

	int
	ddi_copyin(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyin(from, to, len));
	}

	int
	ddi_copyout(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyout(from, to, len));
	}

	-int
	+void
	spl_panic(const char file, const char func, int line, const char *fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);
	vpanic(fmt, ap);
	va_end(ap);
	}


	SYSINIT(opensolaris_utsname_init, SI_SUB_TUNABLES, SI_ORDER_ANY,
	opensolaris_utsname_init, NULL);
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/arc_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/arc_os.c
	index 30e96a889e0d..dfe5c3d311c2 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/arc_os.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/arc_os.c
	@@ -1,260 +1,260 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/counter.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/arc_os.h>
	#include <sys/zfs_refcount.h>
	#include <sys/vdev.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_impl.h>
	#include <sys/dsl_pool.h>
	#include <sys/zio_checksum.h>
	#include <sys/multilist.h>
	#include <sys/abd.h>
	#include <sys/zil.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/eventhandler.h>
	#include <sys/callb.h>
	#include <sys/kstat.h>
	#include <sys/zthr.h>
	#include <zfs_fletcher.h>
	#include <sys/arc_impl.h>
	#include <sys/sdt.h>
	#include <sys/aggsum.h>
	#include <sys/vnode.h>
	#include <cityhash.h>
	#include <machine/vmparam.h>
	#include <sys/vm.h>
	#include <sys/vmmeter.h>

	#if __FreeBSD_version >= 1300139
	static struct sx arc_vnlru_lock;
	static struct vnode *arc_vnlru_marker;
	#endif

	extern struct vfsops zfs_vfsops;

	uint_t zfs_arc_free_target = 0;

	static void
	arc_free_target_init(void *unused __unused)
	{
	zfs_arc_free_target = vm_cnt.v_free_target;
	}
	SYSINIT(arc_free_target_init, SI_SUB_KTHREAD_PAGE, SI_ORDER_ANY,
	arc_free_target_init, NULL);

	/*
	* We don't have a tunable for arc_free_target due to the dependency on
	* pagedaemon initialisation.
	*/
	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, free_target,
	param_set_arc_free_target, 0, CTLFLAG_RW,
	"Desired number of free pages below which ARC triggers reclaim");
	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, no_grow_shift,
	param_set_arc_no_grow_shift, 0, ZMOD_RW,
	"log2(fraction of ARC which must be free to allow growing)");

	int64_t
	arc_available_memory(void)
	{
	int64_t lowest = INT64_MAX;
	int64_t n __unused;

	/*
	* Cooperate with pagedaemon when it's time for it to scan
	* and reclaim some pages.
	*/
	n = PAGESIZE * ((int64_t)freemem - zfs_arc_free_target);
	if (n < lowest) {
	lowest = n;
	}
	#if defined(__i386) \|\| !defined(UMA_MD_SMALL_ALLOC)
	/*
	* If we're on an i386 platform, it's possible that we'll exhaust the
	* kernel heap space before we ever run out of available physical
	* memory. Most checks of the size of the heap_area compare against
	* tune.t_minarmem, which is the minimum available real memory that we
	* can have in the system. However, this is generally fixed at 25 pages
	* which is so low that it's useless. In this comparison, we seek to
	* calculate the total heap-size, and reclaim if more than 3/4ths of the
	* heap is allocated. (Or, in the calculation, if less than 1/4th is
	* free)
	*/
	n = uma_avail() - (long)(uma_limit() / 4);
	if (n < lowest) {
	lowest = n;
	}
	#endif

	DTRACE_PROBE1(arc__available_memory, int64_t, lowest);
	return (lowest);
	}

	/*
	* Return a default max arc size based on the amount of physical memory.
	*/
	uint64_t
	arc_default_max(uint64_t min, uint64_t allmem)
	{
	uint64_t size;

	if (allmem >= 1 << 30)
	size = allmem - (1 << 30);
	else
	size = min;
	return (MAX(allmem * 5 / 8, size));
	}

	/*
	* Helper function for arc_prune_async() it is responsible for safely
	* handling the execution of a registered arc_prune_func_t.
	*/
	static void
	arc_prune_task(void *arg)
	{
	- int64_t nr_scan = (intptr_t)arg;
	+ uint64_t nr_scan = (uintptr_t)arg;

	arc_reduce_target_size(ptob(nr_scan));

	#ifndef __ILP32__
	if (nr_scan > INT_MAX)
	nr_scan = INT_MAX;
	#endif

	#if __FreeBSD_version >= 1300139
	sx_xlock(&arc_vnlru_lock);
	vnlru_free_vfsops(nr_scan, &zfs_vfsops, arc_vnlru_marker);
	sx_xunlock(&arc_vnlru_lock);
	#else
	vnlru_free(nr_scan, &zfs_vfsops);
	#endif
	}

	/*
	* Notify registered consumers they must drop holds on a portion of the ARC
	* buffered they reference. This provides a mechanism to ensure the ARC can
	* honor the arc_meta_limit and reclaim otherwise pinned ARC buffers. This
	* is analogous to dnlc_reduce_cache() but more generic.
	*
	* This operation is performed asynchronously so it may be safely called
	* in the context of the arc_reclaim_thread(). A reference is taken here
	* for each registered arc_prune_t and the arc_prune_task() is responsible
	* for releasing it once the registered arc_prune_func_t has completed.
	*/
	void
	-arc_prune_async(int64_t adjust)
	+arc_prune_async(uint64_t adjust)
	{

	#ifndef __LP64__
	- if (adjust > INTPTR_MAX)
	- adjust = INTPTR_MAX;
	+ if (adjust > UINTPTR_MAX)
	+ adjust = UINTPTR_MAX;
	#endif
	taskq_dispatch(arc_prune_taskq, arc_prune_task,
	(void *)(intptr_t)adjust, TQ_SLEEP);
	ARCSTAT_BUMP(arcstat_prune);
	}

	uint64_t
	arc_all_memory(void)
	{
	return (ptob(physmem));
	}

	int
	arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg)
	{
	return (0);
	}

	uint64_t
	arc_free_memory(void)
	{
	return (ptob(freemem));
	}

	static eventhandler_tag arc_event_lowmem = NULL;

	static void
	arc_lowmem(void *arg __unused, int howto __unused)
	{
	int64_t free_memory, to_free;

	arc_no_grow = B_TRUE;
	arc_warm = B_TRUE;
	arc_growtime = gethrtime() + SEC2NSEC(arc_grow_retry);
	free_memory = arc_available_memory();
	int64_t can_free = arc_c - arc_c_min;
	if (can_free <= 0)
	return;
	to_free = (can_free >> arc_shrink_shift) - MIN(free_memory, 0);
	DTRACE_PROBE2(arc__needfree, int64_t, free_memory, int64_t, to_free);
	arc_reduce_target_size(to_free);

	/*
	* It is unsafe to block here in arbitrary threads, because we can come
	* here from ARC itself and may hold ARC locks and thus risk a deadlock
	* with ARC reclaim thread.
	*/
	if (curproc == pageproc)
	arc_wait_for_eviction(to_free, B_FALSE);
	}

	void
	arc_lowmem_init(void)
	{
	arc_event_lowmem = EVENTHANDLER_REGISTER(vm_lowmem, arc_lowmem, NULL,
	EVENTHANDLER_PRI_FIRST);
	#if __FreeBSD_version >= 1300139
	arc_vnlru_marker = vnlru_alloc_marker();
	sx_init(&arc_vnlru_lock, "arc vnlru lock");
	#endif
	}

	void
	arc_lowmem_fini(void)
	{
	if (arc_event_lowmem != NULL)
	EVENTHANDLER_DEREGISTER(vm_lowmem, arc_event_lowmem);
	#if __FreeBSD_version >= 1300139
	if (arc_vnlru_marker != NULL) {
	vnlru_free_marker(arc_vnlru_marker);
	sx_destroy(&arc_vnlru_lock);
	}
	#endif
	}

	void
	arc_register_hotplug(void)
	{
	}

	void
	arc_unregister_hotplug(void)
	{
	}
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c
	index 4d908381c40a..980bb1c0f941 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/sysctl_os.c
	@@ -1,878 +1,878 @@
	/*
	* Copyright (c) 2020 iXsystems, Inc.
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*
	*/

	#include <sys/cdefs.h>
	__FBSDID("$FreeBSD$");

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/conf.h>
	#include <sys/kernel.h>
	#include <sys/lock.h>
	#include <sys/malloc.h>
	#include <sys/mutex.h>
	#include <sys/proc.h>
	#include <sys/errno.h>
	#include <sys/uio.h>
	#include <sys/buf.h>
	#include <sys/file.h>
	#include <sys/kmem.h>
	#include <sys/conf.h>
	#include <sys/cmn_err.h>
	#include <sys/stat.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_znode.h>
	#include <sys/zap.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/arc_os.h>
	#include <sys/dmu.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_deleg.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/sunddi.h>
	#include <sys/policy.h>
	#include <sys/zone.h>
	#include <sys/nvpair.h>
	#include <sys/mount.h>
	#include <sys/taskqueue.h>
	#include <sys/sdt.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_onexit.h>
	#include <sys/zvol.h>
	#include <sys/dsl_scan.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_send.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/dsl_userhold.h>
	#include <sys/zfeature.h>
	#include <sys/zcp.h>
	#include <sys/zio_checksum.h>
	#include <sys/vdev_removal.h>
	#include <sys/dsl_crypt.h>

	#include <sys/zfs_ioctl_compat.h>
	#include <sys/zfs_context.h>

	#include <sys/arc_impl.h>
	#include <sys/dsl_pool.h>

	#include <sys/vmmeter.h>

	SYSCTL_DECL(_vfs_zfs);
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, arc, CTLFLAG_RW, 0,
	"ZFS adaptive replacement cache");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, condense, CTLFLAG_RW, 0, "ZFS condense");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, dbuf, CTLFLAG_RW, 0, "ZFS disk buf cache");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, dbuf_cache, CTLFLAG_RW, 0,
	"ZFS disk buf cache");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, deadman, CTLFLAG_RW, 0, "ZFS deadman");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, dedup, CTLFLAG_RW, 0, "ZFS dedup");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, l2arc, CTLFLAG_RW, 0, "ZFS l2arc");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, livelist, CTLFLAG_RW, 0, "ZFS livelist");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, lua, CTLFLAG_RW, 0, "ZFS lua");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, metaslab, CTLFLAG_RW, 0, "ZFS metaslab");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, mg, CTLFLAG_RW, 0, "ZFS metaslab group");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, multihost, CTLFLAG_RW, 0,
	"ZFS multihost protection");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, prefetch, CTLFLAG_RW, 0, "ZFS prefetch");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, reconstruct, CTLFLAG_RW, 0, "ZFS reconstruct");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, recv, CTLFLAG_RW, 0, "ZFS receive");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, send, CTLFLAG_RW, 0, "ZFS send");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, spa, CTLFLAG_RW, 0, "ZFS space allocation");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, trim, CTLFLAG_RW, 0, "ZFS TRIM");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, txg, CTLFLAG_RW, 0, "ZFS transaction group");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, vdev, CTLFLAG_RW, 0, "ZFS VDEV");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, vnops, CTLFLAG_RW, 0, "ZFS VNOPS");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, zevent, CTLFLAG_RW, 0, "ZFS event");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, zil, CTLFLAG_RW, 0, "ZFS ZIL");
	SYSCTL_NODE(_vfs_zfs, OID_AUTO, zio, CTLFLAG_RW, 0, "ZFS ZIO");

	SYSCTL_NODE(_vfs_zfs_livelist, OID_AUTO, condense, CTLFLAG_RW, 0,
	"ZFS livelist condense");
	SYSCTL_NODE(_vfs_zfs_vdev, OID_AUTO, cache, CTLFLAG_RW, 0, "ZFS VDEV Cache");
	SYSCTL_NODE(_vfs_zfs_vdev, OID_AUTO, file, CTLFLAG_RW, 0, "ZFS VDEV file");
	SYSCTL_NODE(_vfs_zfs_vdev, OID_AUTO, mirror, CTLFLAG_RD, 0,
	"ZFS VDEV mirror");

	SYSCTL_DECL(_vfs_zfs_version);
	SYSCTL_CONST_STRING(_vfs_zfs_version, OID_AUTO, module, CTLFLAG_RD,
	(ZFS_META_VERSION "-" ZFS_META_RELEASE), "OpenZFS module version");

	/* arc.c */

	int
	param_set_arc_long(SYSCTL_HANDLER_ARGS)
	{
	int err;

	err = sysctl_handle_long(oidp, arg1, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	arc_tuning_update(B_TRUE);

	return (0);
	}

	int
	param_set_arc_int(SYSCTL_HANDLER_ARGS)
	{
	int err;

	err = sysctl_handle_int(oidp, arg1, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	arc_tuning_update(B_TRUE);

	return (0);
	}

	int
	param_set_arc_max(SYSCTL_HANDLER_ARGS)
	{
	unsigned long val;
	int err;

	val = zfs_arc_max;
	err = sysctl_handle_long(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (SET_ERROR(err));

	if (val != 0 && (val < MIN_ARC_MAX \|\| val <= arc_c_min \|\|
	val >= arc_all_memory()))
	return (SET_ERROR(EINVAL));

	zfs_arc_max = val;
	arc_tuning_update(B_TRUE);

	/* Update the sysctl to the tuned value */
	if (val != 0)
	zfs_arc_max = arc_c_max;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_max,
	CTLTYPE_ULONG \| CTLFLAG_RWTUN \| CTLFLAG_MPSAFE,
	NULL, 0, param_set_arc_max, "LU",
	"Maximum ARC size in bytes (LEGACY)");
	/* END CSTYLED */

	int
	param_set_arc_min(SYSCTL_HANDLER_ARGS)
	{
	unsigned long val;
	int err;

	val = zfs_arc_min;
	err = sysctl_handle_long(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (SET_ERROR(err));

	if (val != 0 && (val < 2ULL << SPA_MAXBLOCKSHIFT \|\| val > arc_c_max))
	return (SET_ERROR(EINVAL));

	zfs_arc_min = val;
	arc_tuning_update(B_TRUE);

	/* Update the sysctl to the tuned value */
	if (val != 0)
	zfs_arc_min = arc_c_min;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_min,
	CTLTYPE_ULONG \| CTLFLAG_RWTUN \| CTLFLAG_MPSAFE,
	NULL, 0, param_set_arc_min, "LU",
	"Minimum ARC size in bytes (LEGACY)");
	/* END CSTYLED */

	extern uint_t zfs_arc_free_target;

	int
	param_set_arc_free_target(SYSCTL_HANDLER_ARGS)
	{
	uint_t val;
	int err;

	val = zfs_arc_free_target;
	err = sysctl_handle_int(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	if (val < minfree)
	return (EINVAL);
	if (val > vm_cnt.v_page_count)
	return (EINVAL);

	zfs_arc_free_target = val;

	return (0);
	}

	/*
	* NOTE: This sysctl is CTLFLAG_RW not CTLFLAG_RWTUN due to its dependency on
	* pagedaemon initialization.
	*/
	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_free_target,
	CTLTYPE_UINT \| CTLFLAG_RW \| CTLFLAG_MPSAFE,
	NULL, 0, param_set_arc_free_target, "IU",
	"Desired number of free pages below which ARC triggers reclaim"
	" (LEGACY)");
	/* END CSTYLED */

	int
	param_set_arc_no_grow_shift(SYSCTL_HANDLER_ARGS)
	{
	int err, val;

	val = arc_no_grow_shift;
	err = sysctl_handle_int(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	if (val < 0 \|\| val >= arc_shrink_shift)
	return (EINVAL);

	arc_no_grow_shift = val;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, arc_no_grow_shift,
	CTLTYPE_INT \| CTLFLAG_RWTUN \| CTLFLAG_MPSAFE,
	NULL, 0, param_set_arc_no_grow_shift, "I",
	"log2(fraction of ARC which must be free to allow growing) (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_write_max;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_max,
	CTLFLAG_RWTUN, &l2arc_write_max, 0,
	"Max write bytes per interval (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_write_boost;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_write_boost,
	CTLFLAG_RWTUN, &l2arc_write_boost, 0,
	"Extra write bytes during device warmup (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_headroom;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_headroom,
	CTLFLAG_RWTUN, &l2arc_headroom, 0,
	"Number of max device writes to precache (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_headroom_boost;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_headroom_boost,
	CTLFLAG_RWTUN, &l2arc_headroom_boost, 0,
	"Compressed l2arc_headroom multiplier (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_feed_secs;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_secs,
	CTLFLAG_RWTUN, &l2arc_feed_secs, 0,
	"Seconds between L2ARC writing (LEGACY)");
	/* END CSTYLED */

	extern uint64_t l2arc_feed_min_ms;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2arc_feed_min_ms,
	CTLFLAG_RWTUN, &l2arc_feed_min_ms, 0,
	"Min feed interval in milliseconds (LEGACY)");
	/* END CSTYLED */

	extern int l2arc_noprefetch;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_noprefetch,
	CTLFLAG_RWTUN, &l2arc_noprefetch, 0,
	"Skip caching prefetched buffers (LEGACY)");
	/* END CSTYLED */

	extern int l2arc_feed_again;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_feed_again,
	CTLFLAG_RWTUN, &l2arc_feed_again, 0,
	"Turbo L2ARC warmup (LEGACY)");
	/* END CSTYLED */

	extern int l2arc_norw;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, l2arc_norw,
	CTLFLAG_RWTUN, &l2arc_norw, 0,
	"No reads during writes (LEGACY)");
	/* END CSTYLED */

	extern arc_state_t ARC_anon;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_size, CTLFLAG_RD,
	&ARC_anon.arcs_size.rc_count, 0, "size of anonymous state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_metadata_esize, CTLFLAG_RD,
	&ARC_anon.arcs_esize[ARC_BUFC_METADATA].rc_count, 0,
	"size of anonymous state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, anon_data_esize, CTLFLAG_RD,
	&ARC_anon.arcs_esize[ARC_BUFC_DATA].rc_count, 0,
	"size of anonymous state");
	/* END CSTYLED */

	extern arc_state_t ARC_mru;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_size, CTLFLAG_RD,
	&ARC_mru.arcs_size.rc_count, 0, "size of mru state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_metadata_esize, CTLFLAG_RD,
	&ARC_mru.arcs_esize[ARC_BUFC_METADATA].rc_count, 0,
	"size of metadata in mru state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_data_esize, CTLFLAG_RD,
	&ARC_mru.arcs_esize[ARC_BUFC_DATA].rc_count, 0,
	"size of data in mru state");
	/* END CSTYLED */

	extern arc_state_t ARC_mru_ghost;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_size, CTLFLAG_RD,
	&ARC_mru_ghost.arcs_size.rc_count, 0, "size of mru ghost state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_metadata_esize, CTLFLAG_RD,
	&ARC_mru_ghost.arcs_esize[ARC_BUFC_METADATA].rc_count, 0,
	"size of metadata in mru ghost state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mru_ghost_data_esize, CTLFLAG_RD,
	&ARC_mru_ghost.arcs_esize[ARC_BUFC_DATA].rc_count, 0,
	"size of data in mru ghost state");
	/* END CSTYLED */

	extern arc_state_t ARC_mfu;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_size, CTLFLAG_RD,
	&ARC_mfu.arcs_size.rc_count, 0, "size of mfu state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_metadata_esize, CTLFLAG_RD,
	&ARC_mfu.arcs_esize[ARC_BUFC_METADATA].rc_count, 0,
	"size of metadata in mfu state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_data_esize, CTLFLAG_RD,
	&ARC_mfu.arcs_esize[ARC_BUFC_DATA].rc_count, 0,
	"size of data in mfu state");
	/* END CSTYLED */

	extern arc_state_t ARC_mfu_ghost;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_size, CTLFLAG_RD,
	&ARC_mfu_ghost.arcs_size.rc_count, 0, "size of mfu ghost state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_metadata_esize, CTLFLAG_RD,
	&ARC_mfu_ghost.arcs_esize[ARC_BUFC_METADATA].rc_count, 0,
	"size of metadata in mfu ghost state");
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, mfu_ghost_data_esize, CTLFLAG_RD,
	&ARC_mfu_ghost.arcs_esize[ARC_BUFC_DATA].rc_count, 0,
	"size of data in mfu ghost state");
	/* END CSTYLED */

	extern arc_state_t ARC_l2c_only;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, l2c_only_size, CTLFLAG_RD,
	&ARC_l2c_only.arcs_size.rc_count, 0, "size of mru state");
	/* END CSTYLED */

	/* dbuf.c */

	/* dmu.c */

	/* dmu_zfetch.c */

	SYSCTL_NODE(_vfs_zfs, OID_AUTO, zfetch, CTLFLAG_RW, 0, "ZFS ZFETCH (LEGACY)");

	extern uint32_t zfetch_max_distance;

	/* BEGIN CSTYLED */
	SYSCTL_UINT(_vfs_zfs_zfetch, OID_AUTO, max_distance,
	CTLFLAG_RWTUN, &zfetch_max_distance, 0,
	"Max bytes to prefetch per stream (LEGACY)");
	/* END CSTYLED */

	extern uint32_t zfetch_max_idistance;

	/* BEGIN CSTYLED */
	SYSCTL_UINT(_vfs_zfs_zfetch, OID_AUTO, max_idistance,
	CTLFLAG_RWTUN, &zfetch_max_idistance, 0,
	"Max bytes to prefetch indirects for per stream (LEGACY)");
	/* END CSTYLED */

	/* dsl_pool.c */

	/* dnode.c */

	extern int zfs_default_bs;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, default_bs, CTLFLAG_RWTUN,
	&zfs_default_bs, 0, "Default dnode block shift");
	/* END CSTYLED */

	extern int zfs_default_ibs;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, default_ibs, CTLFLAG_RWTUN,
	&zfs_default_ibs, 0, "Default dnode indirect block shift");
	/* END CSTYLED */

	/* dsl_scan.c */

	/* metaslab.c */

	/*
	* In pools where the log space map feature is not enabled we touch
	* multiple metaslabs (and their respective space maps) with each
	* transaction group. Thus, we benefit from having a small space map
	* block size since it allows us to issue more I/O operations scattered
	* around the disk. So a sane default for the space map block size
	* is 8~16K.
	*/
	extern int zfs_metaslab_sm_blksz_no_log;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs_metaslab, OID_AUTO, sm_blksz_no_log,
	CTLFLAG_RDTUN, &zfs_metaslab_sm_blksz_no_log, 0,
	"Block size for space map in pools with log space map disabled. "
	"Power of 2 greater than 4096.");
	/* END CSTYLED */

	/*
	* When the log space map feature is enabled, we accumulate a lot of
	* changes per metaslab that are flushed once in a while so we benefit
	* from a bigger block size like 128K for the metaslab space maps.
	*/
	extern int zfs_metaslab_sm_blksz_with_log;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs_metaslab, OID_AUTO, sm_blksz_with_log,
	CTLFLAG_RDTUN, &zfs_metaslab_sm_blksz_with_log, 0,
	"Block size for space map in pools with log space map enabled. "
	"Power of 2 greater than 4096.");
	/* END CSTYLED */

	/*
	* The in-core space map representation is more compact than its on-disk form.
	* The zfs_condense_pct determines how much more compact the in-core
	* space map representation must be before we compact it on-disk.
	* Values should be greater than or equal to 100.
	*/
	-extern int zfs_condense_pct;
	+extern uint_t zfs_condense_pct;

	/* BEGIN CSTYLED */
	-SYSCTL_INT(_vfs_zfs, OID_AUTO, condense_pct,
	+SYSCTL_UINT(_vfs_zfs, OID_AUTO, condense_pct,
	CTLFLAG_RWTUN, &zfs_condense_pct, 0,
	"Condense on-disk spacemap when it is more than this many percents"
	" of in-memory counterpart");
	/* END CSTYLED */

	-extern int zfs_remove_max_segment;
	+extern uint_t zfs_remove_max_segment;

	/* BEGIN CSTYLED */
	-SYSCTL_INT(_vfs_zfs, OID_AUTO, remove_max_segment,
	+SYSCTL_UINT(_vfs_zfs, OID_AUTO, remove_max_segment,
	CTLFLAG_RWTUN, &zfs_remove_max_segment, 0,
	"Largest contiguous segment ZFS will attempt to allocate when removing"
	" a device");
	/* END CSTYLED */

	extern int zfs_removal_suspend_progress;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, removal_suspend_progress,
	CTLFLAG_RWTUN, &zfs_removal_suspend_progress, 0,
	"Ensures certain actions can happen while in the middle of a removal");
	/* END CSTYLED */

	/*
	* Minimum size which forces the dynamic allocator to change
	* it's allocation strategy. Once the space map cannot satisfy
	* an allocation of this size then it switches to using more
	* aggressive strategy (i.e search by size rather than offset).
	*/
	extern uint64_t metaslab_df_alloc_threshold;

	/* BEGIN CSTYLED */
	SYSCTL_QUAD(_vfs_zfs_metaslab, OID_AUTO, df_alloc_threshold,
	CTLFLAG_RWTUN, &metaslab_df_alloc_threshold, 0,
	"Minimum size which forces the dynamic allocator to change its"
	" allocation strategy");
	/* END CSTYLED */

	/*
	* The minimum free space, in percent, which must be available
	* in a space map to continue allocations in a first-fit fashion.
	* Once the space map's free space drops below this level we dynamically
	* switch to using best-fit allocations.
	*/
	-extern int metaslab_df_free_pct;
	+extern uint_t metaslab_df_free_pct;

	/* BEGIN CSTYLED */
	-SYSCTL_INT(_vfs_zfs_metaslab, OID_AUTO, df_free_pct,
	+SYSCTL_UINT(_vfs_zfs_metaslab, OID_AUTO, df_free_pct,
	CTLFLAG_RWTUN, &metaslab_df_free_pct, 0,
	"The minimum free space, in percent, which must be available in a"
	" space map to continue allocations in a first-fit fashion");
	/* END CSTYLED */

	/*
	* Percentage of all cpus that can be used by the metaslab taskq.
	*/
	extern int metaslab_load_pct;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs_metaslab, OID_AUTO, load_pct,
	CTLFLAG_RWTUN, &metaslab_load_pct, 0,
	"Percentage of cpus that can be used by the metaslab taskq");
	/* END CSTYLED */

	/*
	* Max number of metaslabs per group to preload.
	*/
	-extern int metaslab_preload_limit;
	+extern uint_t metaslab_preload_limit;

	/* BEGIN CSTYLED */
	-SYSCTL_INT(_vfs_zfs_metaslab, OID_AUTO, preload_limit,
	+SYSCTL_UINT(_vfs_zfs_metaslab, OID_AUTO, preload_limit,
	CTLFLAG_RWTUN, &metaslab_preload_limit, 0,
	"Max number of metaslabs per group to preload");
	/* END CSTYLED */

	/* mmp.c */

	int
	param_set_multihost_interval(SYSCTL_HANDLER_ARGS)
	{
	int err;

	err = sysctl_handle_long(oidp, &zfs_multihost_interval, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	if (spa_mode_global != SPA_MODE_UNINIT)
	mmp_signal_all_threads();

	return (0);
	}

	/* spa.c */

	extern int zfs_ccw_retry_interval;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, ccw_retry_interval,
	CTLFLAG_RWTUN, &zfs_ccw_retry_interval, 0,
	"Configuration cache file write, retry after failure, interval"
	" (seconds)");
	/* END CSTYLED */

	extern uint64_t zfs_max_missing_tvds_cachefile;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, max_missing_tvds_cachefile,
	CTLFLAG_RWTUN, &zfs_max_missing_tvds_cachefile, 0,
	"Allow importing pools with missing top-level vdevs in cache file");
	/* END CSTYLED */

	extern uint64_t zfs_max_missing_tvds_scan;

	/* BEGIN CSTYLED */
	SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, max_missing_tvds_scan,
	CTLFLAG_RWTUN, &zfs_max_missing_tvds_scan, 0,
	"Allow importing pools with missing top-level vdevs during scan");
	/* END CSTYLED */

	/* spa_misc.c */

	extern int zfs_flags;

	static int
	sysctl_vfs_zfs_debug_flags(SYSCTL_HANDLER_ARGS)
	{
	int err, val;

	val = zfs_flags;
	err = sysctl_handle_int(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	/*
	* ZFS_DEBUG_MODIFY must be enabled prior to boot so all
	* arc buffers in the system have the necessary additional
	* checksum data. However, it is safe to disable at any
	* time.
	*/
	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	val &= ~ZFS_DEBUG_MODIFY;
	zfs_flags = val;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, debugflags,
	CTLTYPE_UINT \| CTLFLAG_MPSAFE \| CTLFLAG_RWTUN, NULL, 0,
	sysctl_vfs_zfs_debug_flags, "IU", "Debug flags for ZFS testing.");
	/* END CSTYLED */

	int
	param_set_deadman_synctime(SYSCTL_HANDLER_ARGS)
	{
	unsigned long val;
	int err;

	val = zfs_deadman_synctime_ms;
	err = sysctl_handle_long(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);
	zfs_deadman_synctime_ms = val;

	spa_set_deadman_synctime(MSEC2NSEC(zfs_deadman_synctime_ms));

	return (0);
	}

	int
	param_set_deadman_ziotime(SYSCTL_HANDLER_ARGS)
	{
	unsigned long val;
	int err;

	val = zfs_deadman_ziotime_ms;
	err = sysctl_handle_long(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);
	zfs_deadman_ziotime_ms = val;

	spa_set_deadman_ziotime(MSEC2NSEC(zfs_deadman_synctime_ms));

	return (0);
	}

	int
	param_set_deadman_failmode(SYSCTL_HANDLER_ARGS)
	{
	char buf[16];
	int rc;

	if (req->newptr == NULL)
	strlcpy(buf, zfs_deadman_failmode, sizeof (buf));

	rc = sysctl_handle_string(oidp, buf, sizeof (buf), req);
	if (rc \|\| req->newptr == NULL)
	return (rc);
	if (strcmp(buf, zfs_deadman_failmode) == 0)
	return (0);
	if (strcmp(buf, "wait") == 0)
	zfs_deadman_failmode = "wait";
	if (strcmp(buf, "continue") == 0)
	zfs_deadman_failmode = "continue";
	if (strcmp(buf, "panic") == 0)
	zfs_deadman_failmode = "panic";

	return (-param_set_deadman_failmode_common(buf));
	}

	int
	param_set_slop_shift(SYSCTL_HANDLER_ARGS)
	{
	int val;
	int err;

	val = spa_slop_shift;
	err = sysctl_handle_int(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (err);

	if (val < 1 \|\| val > 31)
	return (EINVAL);

	spa_slop_shift = val;

	return (0);
	}

	/* spacemap.c */

	extern int space_map_ibs;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, space_map_ibs, CTLFLAG_RWTUN,
	&space_map_ibs, 0, "Space map indirect block shift");
	/* END CSTYLED */


	/* vdev.c */

	int
	param_set_min_auto_ashift(SYSCTL_HANDLER_ARGS)
	{
	uint64_t val;
	int err;

	val = zfs_vdev_min_auto_ashift;
	err = sysctl_handle_64(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (SET_ERROR(err));

	if (val < ASHIFT_MIN \|\| val > zfs_vdev_max_auto_ashift)
	return (SET_ERROR(EINVAL));

	zfs_vdev_min_auto_ashift = val;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, min_auto_ashift,
	CTLTYPE_U64 \| CTLFLAG_RWTUN \| CTLFLAG_MPSAFE,
	&zfs_vdev_min_auto_ashift, sizeof (zfs_vdev_min_auto_ashift),
	param_set_min_auto_ashift, "QU",
	"Min ashift used when creating new top-level vdev. (LEGACY)");
	/* END CSTYLED */

	int
	param_set_max_auto_ashift(SYSCTL_HANDLER_ARGS)
	{
	uint64_t val;
	int err;

	val = zfs_vdev_max_auto_ashift;
	err = sysctl_handle_64(oidp, &val, 0, req);
	if (err != 0 \|\| req->newptr == NULL)
	return (SET_ERROR(err));

	if (val > ASHIFT_MAX \|\| val < zfs_vdev_min_auto_ashift)
	return (SET_ERROR(EINVAL));

	zfs_vdev_max_auto_ashift = val;

	return (0);
	}

	/* BEGIN CSTYLED */
	SYSCTL_PROC(_vfs_zfs, OID_AUTO, max_auto_ashift,
	CTLTYPE_U64 \| CTLFLAG_RWTUN \| CTLFLAG_MPSAFE,
	&zfs_vdev_max_auto_ashift, sizeof (zfs_vdev_max_auto_ashift),
	param_set_max_auto_ashift, "QU",
	"Max ashift used when optimizing for logical -> physical sector size on"
	" new top-level vdevs. (LEGACY)");
	/* END CSTYLED */

	/*
	* Since the DTL space map of a vdev is not expected to have a lot of
	* entries, we default its block size to 4K.
	*/
	extern int zfs_vdev_dtl_sm_blksz;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, dtl_sm_blksz,
	CTLFLAG_RDTUN, &zfs_vdev_dtl_sm_blksz, 0,
	"Block size for DTL space map. Power of 2 greater than 4096.");
	/* END CSTYLED */

	/*
	* vdev-wide space maps that have lots of entries written to them at
	* the end of each transaction can benefit from a higher I/O bandwidth
	* (e.g. vdev_obsolete_sm), thus we default their block size to 128K.
	*/
	extern int zfs_vdev_standard_sm_blksz;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, standard_sm_blksz,
	CTLFLAG_RDTUN, &zfs_vdev_standard_sm_blksz, 0,
	"Block size for standard space map. Power of 2 greater than 4096.");
	/* END CSTYLED */

	extern int vdev_validate_skip;

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs, OID_AUTO, validate_skip,
	CTLFLAG_RDTUN, &vdev_validate_skip, 0,
	"Enable to bypass vdev_validate().");
	/* END CSTYLED */

	/* vdev_cache.c */

	/* vdev_mirror.c */

	/* vdev_queue.c */

	-extern uint32_t zfs_vdev_max_active;
	+extern uint_t zfs_vdev_max_active;

	/* BEGIN CSTYLED */
	SYSCTL_UINT(_vfs_zfs, OID_AUTO, top_maxinflight,
	CTLFLAG_RWTUN, &zfs_vdev_max_active, 0,
	"The maximum number of I/Os of all types active for each device."
	" (LEGACY)");
	/* END CSTYLED */

	-extern int zfs_vdev_def_queue_depth;
	+extern uint_t zfs_vdev_def_queue_depth;

	/* BEGIN CSTYLED */
	-SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, def_queue_depth,
	+SYSCTL_UINT(_vfs_zfs_vdev, OID_AUTO, def_queue_depth,
	CTLFLAG_RWTUN, &zfs_vdev_def_queue_depth, 0,
	"Default queue depth for each allocator");
	/* END CSTYLED */

	/* zio.c */

	/* BEGIN CSTYLED */
	SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, exclude_metadata,
	CTLFLAG_RDTUN, &zio_exclude_metadata, 0,
	"Exclude metadata buffers from dumps as well");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c
	index ad482ee9d49b..963102f3b62a 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_acl.c
	@@ -1,2674 +1,2674 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013 by Delphix. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc. All rights reserved.
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/systm.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <sys/vfs.h>
	#include <sys/vnode.h>
	#include <sys/file.h>
	#include <sys/stat.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/errno.h>
	#include <sys/unistd.h>
	#include <sys/sdt.h>
	#include <sys/fs/zfs.h>
	#include <sys/policy.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_fuid.h>
	#include <sys/zfs_acl.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/dmu.h>
	#include <sys/dnode.h>
	#include <sys/zap.h>
	#include <sys/sa.h>
	#include <acl/acl_common.h>


	#define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
	#define DENY ACE_ACCESS_DENIED_ACE_TYPE
	#define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
	#define MIN_ACE_TYPE ALLOW

	#define OWNING_GROUP (ACE_GROUP\|ACE_IDENTIFIER_GROUP)
	#define EVERYONE_ALLOW_MASK (ACE_READ_ACL\|ACE_READ_ATTRIBUTES \| \
	ACE_READ_NAMED_ATTRS\|ACE_SYNCHRONIZE)
	#define EVERYONE_DENY_MASK (ACE_WRITE_ACL\|ACE_WRITE_OWNER \| \
	ACE_WRITE_ATTRIBUTES\|ACE_WRITE_NAMED_ATTRS)
	#define OWNER_ALLOW_MASK (ACE_WRITE_ACL \| ACE_WRITE_OWNER \| \
	ACE_WRITE_ATTRIBUTES\|ACE_WRITE_NAMED_ATTRS)

	#define ZFS_CHECKED_MASKS (ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_READ_DATA\| \
	ACE_READ_NAMED_ATTRS\|ACE_WRITE_DATA\|ACE_WRITE_ATTRIBUTES\| \
	ACE_WRITE_NAMED_ATTRS\|ACE_APPEND_DATA\|ACE_EXECUTE\|ACE_WRITE_OWNER\| \
	ACE_WRITE_ACL\|ACE_DELETE\|ACE_DELETE_CHILD\|ACE_SYNCHRONIZE)

	#define WRITE_MASK_DATA (ACE_WRITE_DATA\|ACE_APPEND_DATA\|ACE_WRITE_NAMED_ATTRS)
	#define WRITE_MASK_ATTRS (ACE_WRITE_ACL\|ACE_WRITE_OWNER\|ACE_WRITE_ATTRIBUTES\| \
	ACE_DELETE\|ACE_DELETE_CHILD)
	#define WRITE_MASK (WRITE_MASK_DATA\|WRITE_MASK_ATTRS)

	#define OGE_CLEAR (ACE_READ_DATA\|ACE_LIST_DIRECTORY\|ACE_WRITE_DATA\| \
	ACE_ADD_FILE\|ACE_APPEND_DATA\|ACE_ADD_SUBDIRECTORY\|ACE_EXECUTE)

	#define OKAY_MASK_BITS (ACE_READ_DATA\|ACE_LIST_DIRECTORY\|ACE_WRITE_DATA\| \
	ACE_ADD_FILE\|ACE_APPEND_DATA\|ACE_ADD_SUBDIRECTORY\|ACE_EXECUTE)

	#define ALL_INHERIT (ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE \| \
	ACE_NO_PROPAGATE_INHERIT_ACE\|ACE_INHERIT_ONLY_ACE\|ACE_INHERITED_ACE)

	#define RESTRICTED_CLEAR (ACE_WRITE_ACL\|ACE_WRITE_OWNER)

	#define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT\|ZFS_ACL_DEFAULTED\|\
	ZFS_ACL_PROTECTED)

	#define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS\|ZFS_ACL_TRIVIAL\|ZFS_INHERIT_ACE\|\
	ZFS_ACL_OBJ_ACE)

	#define ALL_MODE_EXECS (S_IXUSR \| S_IXGRP \| S_IXOTH)

	static uint16_t
	zfs_ace_v0_get_type(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_type);
	}

	static uint16_t
	zfs_ace_v0_get_flags(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_flags);
	}

	static uint32_t
	zfs_ace_v0_get_mask(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_access_mask);
	}

	static uint64_t
	zfs_ace_v0_get_who(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_fuid);
	}

	static void
	zfs_ace_v0_set_type(void *acep, uint16_t type)
	{
	((zfs_oldace_t *)acep)->z_type = type;
	}

	static void
	zfs_ace_v0_set_flags(void *acep, uint16_t flags)
	{
	((zfs_oldace_t *)acep)->z_flags = flags;
	}

	static void
	zfs_ace_v0_set_mask(void *acep, uint32_t mask)
	{
	((zfs_oldace_t *)acep)->z_access_mask = mask;
	}

	static void
	zfs_ace_v0_set_who(void *acep, uint64_t who)
	{
	((zfs_oldace_t *)acep)->z_fuid = who;
	}

	static size_t
	zfs_ace_v0_size(void *acep)
	{
	(void) acep;
	return (sizeof (zfs_oldace_t));
	}

	static size_t
	zfs_ace_v0_abstract_size(void)
	{
	return (sizeof (zfs_oldace_t));
	}

	static int
	zfs_ace_v0_mask_off(void)
	{
	return (offsetof(zfs_oldace_t, z_access_mask));
	}

	static int
	zfs_ace_v0_data(void acep, void *datap)
	{
	(void) acep;
	*datap = NULL;
	return (0);
	}

	static const acl_ops_t zfs_acl_v0_ops = {
	zfs_ace_v0_get_mask,
	zfs_ace_v0_set_mask,
	zfs_ace_v0_get_flags,
	zfs_ace_v0_set_flags,
	zfs_ace_v0_get_type,
	zfs_ace_v0_set_type,
	zfs_ace_v0_get_who,
	zfs_ace_v0_set_who,
	zfs_ace_v0_size,
	zfs_ace_v0_abstract_size,
	zfs_ace_v0_mask_off,
	zfs_ace_v0_data
	};

	static uint16_t
	zfs_ace_fuid_get_type(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_type);
	}

	static uint16_t
	zfs_ace_fuid_get_flags(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_flags);
	}

	static uint32_t
	zfs_ace_fuid_get_mask(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_access_mask);
	}

	static uint64_t
	zfs_ace_fuid_get_who(void *args)
	{
	uint16_t entry_type;
	zfs_ace_t *acep = args;

	entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;

	if (entry_type == ACE_OWNER \|\| entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return (-1);
	return (((zfs_ace_t *)acep)->z_fuid);
	}

	static void
	zfs_ace_fuid_set_type(void *acep, uint16_t type)
	{
	((zfs_ace_hdr_t *)acep)->z_type = type;
	}

	static void
	zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
	{
	((zfs_ace_hdr_t *)acep)->z_flags = flags;
	}

	static void
	zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
	{
	((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
	}

	static void
	zfs_ace_fuid_set_who(void *arg, uint64_t who)
	{
	zfs_ace_t *acep = arg;

	uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;

	if (entry_type == ACE_OWNER \|\| entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return;
	acep->z_fuid = who;
	}

	static size_t
	zfs_ace_fuid_size(void *acep)
	{
	zfs_ace_hdr_t *zacep = acep;
	uint16_t entry_type;

	switch (zacep->z_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	return (sizeof (zfs_object_ace_t));
	case ALLOW:
	case DENY:
	entry_type =
	(((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
	if (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return (sizeof (zfs_ace_hdr_t));
	zfs_fallthrough;
	default:
	return (sizeof (zfs_ace_t));
	}
	}

	static size_t
	zfs_ace_fuid_abstract_size(void)
	{
	return (sizeof (zfs_ace_hdr_t));
	}

	static int
	zfs_ace_fuid_mask_off(void)
	{
	return (offsetof(zfs_ace_hdr_t, z_access_mask));
	}

	static int
	zfs_ace_fuid_data(void acep, void *datap)
	{
	zfs_ace_t *zacep = acep;
	zfs_object_ace_t *zobjp;

	switch (zacep->z_hdr.z_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	zobjp = acep;
	*datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
	return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
	default:
	*datap = NULL;
	return (0);
	}
	}

	static const acl_ops_t zfs_acl_fuid_ops = {
	zfs_ace_fuid_get_mask,
	zfs_ace_fuid_set_mask,
	zfs_ace_fuid_get_flags,
	zfs_ace_fuid_set_flags,
	zfs_ace_fuid_get_type,
	zfs_ace_fuid_set_type,
	zfs_ace_fuid_get_who,
	zfs_ace_fuid_set_who,
	zfs_ace_fuid_size,
	zfs_ace_fuid_abstract_size,
	zfs_ace_fuid_mask_off,
	zfs_ace_fuid_data
	};

	/*
	* The following three functions are provided for compatibility with
	* older ZPL version in order to determine if the file use to have
	* an external ACL and what version of ACL previously existed on the
	* file. Would really be nice to not need this, sigh.
	*/
	uint64_t
	zfs_external_acl(znode_t *zp)
	{
	zfs_acl_phys_t acl_phys;
	int error;

	if (zp->z_is_sa)
	return (0);

	/*
	* Need to deal with a potential
	* race where zfs_sa_upgrade could cause
	* z_isa_sa to change.
	*
	* If the lookup fails then the state of z_is_sa should have
	* changed.
	*/

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
	&acl_phys, sizeof (acl_phys))) == 0)
	return (acl_phys.z_acl_extern_obj);
	else {
	/*
	* after upgrade the SA_ZPL_ZNODE_ACL should have been
	* removed
	*/
	VERIFY(zp->z_is_sa);
	VERIFY3S(error, ==, ENOENT);
	return (0);
	}
	}

	/*
	* Determine size of ACL in bytes
	*
	* This is more complicated than it should be since we have to deal
	* with old external ACLs.
	*/
	static int
	zfs_acl_znode_info(znode_t zp, int aclsize, int *aclcount,
	zfs_acl_phys_t *aclphys)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	uint64_t acl_count;
	int size;
	int error;

	ASSERT(MUTEX_HELD(&zp->z_acl_lock));
	if (zp->z_is_sa) {
	if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
	&size)) != 0)
	return (error);
	*aclsize = size;
	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
	&acl_count, sizeof (acl_count))) != 0)
	return (error);
	*aclcount = acl_count;
	} else {
	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
	aclphys, sizeof (*aclphys))) != 0)
	return (error);

	if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
	*aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
	*aclcount = aclphys->z_acl_size;
	} else {
	*aclsize = aclphys->z_acl_size;
	*aclcount = aclphys->z_acl_count;
	}
	}
	return (0);
	}

	int
	zfs_znode_acl_version(znode_t *zp)
	{
	zfs_acl_phys_t acl_phys;

	if (zp->z_is_sa)
	return (ZFS_ACL_VERSION_FUID);
	else {
	int error;

	/*
	* Need to deal with a potential
	* race where zfs_sa_upgrade could cause
	* z_isa_sa to change.
	*
	* If the lookup fails then the state of z_is_sa should have
	* changed.
	*/
	if ((error = sa_lookup(zp->z_sa_hdl,
	SA_ZPL_ZNODE_ACL(zp->z_zfsvfs),
	&acl_phys, sizeof (acl_phys))) == 0)
	return (acl_phys.z_acl_version);
	else {
	/*
	* After upgrade SA_ZPL_ZNODE_ACL should have
	* been removed.
	*/
	VERIFY(zp->z_is_sa);
	VERIFY3S(error, ==, ENOENT);
	return (ZFS_ACL_VERSION_FUID);
	}
	}
	}

	static int
	zfs_acl_version(int version)
	{
	if (version < ZPL_VERSION_FUID)
	return (ZFS_ACL_VERSION_INITIAL);
	else
	return (ZFS_ACL_VERSION_FUID);
	}

	static int
	zfs_acl_version_zp(znode_t *zp)
	{
	return (zfs_acl_version(zp->z_zfsvfs->z_version));
	}

	zfs_acl_t *
	zfs_acl_alloc(int vers)
	{
	zfs_acl_t *aclp;

	aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
	list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
	offsetof(zfs_acl_node_t, z_next));
	aclp->z_version = vers;
	if (vers == ZFS_ACL_VERSION_FUID)
	aclp->z_ops = &zfs_acl_fuid_ops;
	else
	aclp->z_ops = &zfs_acl_v0_ops;
	return (aclp);
	}

	zfs_acl_node_t *
	zfs_acl_node_alloc(size_t bytes)
	{
	zfs_acl_node_t *aclnode;

	aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
	if (bytes) {
	aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
	aclnode->z_allocdata = aclnode->z_acldata;
	aclnode->z_allocsize = bytes;
	aclnode->z_size = bytes;
	}

	return (aclnode);
	}

	static void
	zfs_acl_node_free(zfs_acl_node_t *aclnode)
	{
	if (aclnode->z_allocsize)
	kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
	kmem_free(aclnode, sizeof (zfs_acl_node_t));
	}

	static void
	zfs_acl_release_nodes(zfs_acl_t *aclp)
	{
	zfs_acl_node_t *aclnode;

	while ((aclnode = list_head(&aclp->z_acl))) {
	list_remove(&aclp->z_acl, aclnode);
	zfs_acl_node_free(aclnode);
	}
	aclp->z_acl_count = 0;
	aclp->z_acl_bytes = 0;
	}

	void
	zfs_acl_free(zfs_acl_t *aclp)
	{
	zfs_acl_release_nodes(aclp);
	list_destroy(&aclp->z_acl);
	kmem_free(aclp, sizeof (zfs_acl_t));
	}

	static boolean_t
	zfs_acl_valid_ace_type(uint_t type, uint_t flags)
	{
	uint16_t entry_type;

	switch (type) {
	case ALLOW:
	case DENY:
	case ACE_SYSTEM_AUDIT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_ACE_TYPE:
	entry_type = flags & ACE_TYPE_FLAGS;
	return (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE \|\| entry_type == 0 \|\|
	entry_type == ACE_IDENTIFIER_GROUP);
	default:
	- if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
	+ if (type <= MAX_ACE_TYPE)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static boolean_t
	zfs_ace_valid(vtype_t obj_type, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
	{
	/*
	* first check type of entry
	*/

	if (!zfs_acl_valid_ace_type(type, iflags))
	return (B_FALSE);

	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	if (aclp->z_version < ZFS_ACL_VERSION_FUID)
	return (B_FALSE);
	aclp->z_hints \|= ZFS_ACL_OBJ_ACE;
	}

	/*
	* next check inheritance level flags
	*/

	if (obj_type == VDIR &&
	(iflags & (ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	if (iflags & (ACE_INHERIT_ONLY_ACE\|ACE_NO_PROPAGATE_INHERIT_ACE)) {
	if ((iflags & (ACE_FILE_INHERIT_ACE\|
	ACE_DIRECTORY_INHERIT_ACE)) == 0) {
	return (B_FALSE);
	}
	}

	return (B_TRUE);
	}

	static void *
	zfs_acl_next_ace(zfs_acl_t aclp, void start, uint64_t *who,
	uint32_t access_mask, uint16_t iflags, uint16_t *type)
	{
	zfs_acl_node_t *aclnode;

	ASSERT3P(aclp, !=, NULL);

	if (start == NULL) {
	aclnode = list_head(&aclp->z_acl);
	if (aclnode == NULL)
	return (NULL);

	aclp->z_next_ace = aclnode->z_acldata;
	aclp->z_curr_node = aclnode;
	aclnode->z_ace_idx = 0;
	}

	aclnode = aclp->z_curr_node;

	if (aclnode == NULL)
	return (NULL);

	if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
	aclnode = list_next(&aclp->z_acl, aclnode);
	if (aclnode == NULL)
	return (NULL);
	else {
	aclp->z_curr_node = aclnode;
	aclnode->z_ace_idx = 0;
	aclp->z_next_ace = aclnode->z_acldata;
	}
	}

	if (aclnode->z_ace_idx < aclnode->z_ace_count) {
	void *acep = aclp->z_next_ace;
	size_t ace_size;

	/*
	* Make sure we don't overstep our bounds
	*/
	ace_size = aclp->z_ops->ace_size(acep);

	if (((caddr_t)acep + ace_size) >
	((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
	return (NULL);
	}

	*iflags = aclp->z_ops->ace_flags_get(acep);
	*type = aclp->z_ops->ace_type_get(acep);
	*access_mask = aclp->z_ops->ace_mask_get(acep);
	*who = aclp->z_ops->ace_who_get(acep);
	aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
	aclnode->z_ace_idx++;

	return ((void *)acep);
	}
	return (NULL);
	}

	static uint64_t
	zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
	uint16_t flags, uint16_t type, uint32_t *mask)
	{
	(void) aclcnt;
	zfs_acl_t *aclp = datap;
	zfs_ace_hdr_t acep = (zfs_ace_hdr_t )(uintptr_t)cookie;
	uint64_t who;

	acep = zfs_acl_next_ace(aclp, acep, &who, mask,
	flags, type);
	return ((uint64_t)(uintptr_t)acep);
	}

	/*
	* Copy ACE to internal ZFS format.
	* While processing the ACL each ACE will be validated for correctness.
	* ACE FUIDs will be created later.
	*/
	static int
	zfs_copy_ace_2_fuid(zfsvfs_t zfsvfs, vtype_t obj_type, zfs_acl_t aclp,
	void datap, zfs_ace_t z_acl, uint64_t aclcnt, size_t *size,
	zfs_fuid_info_t *fuidp, cred_t cr)
	{
	int i;
	uint16_t entry_type;
	zfs_ace_t *aceptr = z_acl;
	ace_t *acep = datap;
	zfs_object_ace_t *zobjacep;
	ace_object_t *aceobjp;

	for (i = 0; i != aclcnt; i++) {
	aceptr->z_hdr.z_access_mask = acep->a_access_mask;
	aceptr->z_hdr.z_flags = acep->a_flags;
	aceptr->z_hdr.z_type = acep->a_type;
	entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
	if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
	entry_type != ACE_EVERYONE) {
	aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
	cr, (entry_type == 0) ?
	ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
	}

	/*
	* Make sure ACE is valid
	*/
	if (zfs_ace_valid(obj_type, aclp, aceptr->z_hdr.z_type,
	aceptr->z_hdr.z_flags) != B_TRUE)
	return (SET_ERROR(EINVAL));

	switch (acep->a_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	zobjacep = (zfs_object_ace_t *)aceptr;
	aceobjp = (ace_object_t *)acep;

	memcpy(zobjacep->z_object_type, aceobjp->a_obj_type,
	sizeof (aceobjp->a_obj_type));
	memcpy(zobjacep->z_inherit_type,
	aceobjp->a_inherit_obj_type,
	sizeof (aceobjp->a_inherit_obj_type));
	acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
	break;
	default:
	acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
	}

	aceptr = (zfs_ace_t *)((caddr_t)aceptr +
	aclp->z_ops->ace_size(aceptr));
	}

	*size = (caddr_t)aceptr - (caddr_t)z_acl;

	return (0);
	}

	/*
	* Copy ZFS ACEs to fixed size ace_t layout
	*/
	static void
	zfs_copy_fuid_2_ace(zfsvfs_t zfsvfs, zfs_acl_t aclp, cred_t *cr,
	void *datap, int filter)
	{
	uint64_t who;
	uint32_t access_mask;
	uint16_t iflags, type;
	zfs_ace_hdr_t *zacep = NULL;
	ace_t *acep = datap;
	ace_object_t *objacep;
	zfs_object_ace_t *zobjacep;
	size_t ace_size;
	uint16_t entry_type;

	while ((zacep = zfs_acl_next_ace(aclp, zacep,
	&who, &access_mask, &iflags, &type))) {

	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	if (filter) {
	continue;
	}
	zobjacep = (zfs_object_ace_t *)zacep;
	objacep = (ace_object_t *)acep;
	memcpy(objacep->a_obj_type,
	zobjacep->z_object_type,
	sizeof (zobjacep->z_object_type));
	memcpy(objacep->a_inherit_obj_type,
	zobjacep->z_inherit_type,
	sizeof (zobjacep->z_inherit_type));
	ace_size = sizeof (ace_object_t);
	break;
	default:
	ace_size = sizeof (ace_t);
	break;
	}

	entry_type = (iflags & ACE_TYPE_FLAGS);
	if ((entry_type != ACE_OWNER &&
	entry_type != OWNING_GROUP &&
	entry_type != ACE_EVERYONE)) {
	acep->a_who = zfs_fuid_map_id(zfsvfs, who,
	cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
	ZFS_ACE_GROUP : ZFS_ACE_USER);
	} else {
	acep->a_who = (uid_t)(int64_t)who;
	}
	acep->a_access_mask = access_mask;
	acep->a_flags = iflags;
	acep->a_type = type;
	acep = (ace_t *)((caddr_t)acep + ace_size);
	}
	}

	static int
	zfs_copy_ace_2_oldace(vtype_t obj_type, zfs_acl_t aclp, ace_t acep,
	zfs_oldace_t z_acl, int aclcnt, size_t size)
	{
	int i;
	zfs_oldace_t *aceptr = z_acl;

	for (i = 0; i != aclcnt; i++, aceptr++) {
	aceptr->z_access_mask = acep[i].a_access_mask;
	aceptr->z_type = acep[i].a_type;
	aceptr->z_flags = acep[i].a_flags;
	aceptr->z_fuid = acep[i].a_who;
	/*
	* Make sure ACE is valid
	*/
	if (zfs_ace_valid(obj_type, aclp, aceptr->z_type,
	aceptr->z_flags) != B_TRUE)
	return (SET_ERROR(EINVAL));
	}
	*size = (caddr_t)aceptr - (caddr_t)z_acl;
	return (0);
	}

	/*
	* convert old ACL format to new
	*/
	void
	zfs_acl_xform(znode_t zp, zfs_acl_t aclp, cred_t *cr)
	{
	zfs_oldace_t *oldaclp;
	int i;
	uint16_t type, iflags;
	uint32_t access_mask;
	uint64_t who;
	void *cookie = NULL;
	zfs_acl_node_t *newaclnode;

	ASSERT3U(aclp->z_version, ==, ZFS_ACL_VERSION_INITIAL);
	/*
	* First create the ACE in a contiguous piece of memory
	* for zfs_copy_ace_2_fuid().
	*
	* We only convert an ACL once, so this won't happen
	* everytime.
	*/
	oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
	KM_SLEEP);
	i = 0;
	while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
	&access_mask, &iflags, &type))) {
	oldaclp[i].z_flags = iflags;
	oldaclp[i].z_type = type;
	oldaclp[i].z_fuid = who;
	oldaclp[i++].z_access_mask = access_mask;
	}

	newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
	sizeof (zfs_object_ace_t));
	aclp->z_ops = &zfs_acl_fuid_ops;
	VERIFY0(zfs_copy_ace_2_fuid(zp->z_zfsvfs, ZTOV(zp)->v_type, aclp,
	oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
	&newaclnode->z_size, NULL, cr));
	newaclnode->z_ace_count = aclp->z_acl_count;
	aclp->z_version = ZFS_ACL_VERSION;
	kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));

	/*
	* Release all previous ACL nodes
	*/

	zfs_acl_release_nodes(aclp);

	list_insert_head(&aclp->z_acl, newaclnode);

	aclp->z_acl_bytes = newaclnode->z_size;
	aclp->z_acl_count = newaclnode->z_ace_count;

	}

	/*
	* Convert unix access mask to v4 access mask
	*/
	static uint32_t
	zfs_unix_to_v4(uint32_t access_mask)
	{
	uint32_t new_mask = 0;

	if (access_mask & S_IXOTH)
	new_mask \|= ACE_EXECUTE;
	if (access_mask & S_IWOTH)
	new_mask \|= ACE_WRITE_DATA;
	if (access_mask & S_IROTH)
	new_mask \|= ACE_READ_DATA;
	return (new_mask);
	}

	static void
	zfs_set_ace(zfs_acl_t aclp, void acep, uint32_t access_mask,
	uint16_t access_type, uint64_t fuid, uint16_t entry_type)
	{
	uint16_t type = entry_type & ACE_TYPE_FLAGS;

	aclp->z_ops->ace_mask_set(acep, access_mask);
	aclp->z_ops->ace_type_set(acep, access_type);
	aclp->z_ops->ace_flags_set(acep, entry_type);
	if ((type != ACE_OWNER && type != OWNING_GROUP &&
	type != ACE_EVERYONE))
	aclp->z_ops->ace_who_set(acep, fuid);
	}

	/*
	* Determine mode of file based on ACL.
	*/
	uint64_t
	zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
	uint64_t *pflags, uint64_t fuid, uint64_t fgid)
	{
	int entry_type;
	mode_t mode;
	mode_t seen = 0;
	zfs_ace_hdr_t *acep = NULL;
	uint64_t who;
	uint16_t iflags, type;
	uint32_t access_mask;
	boolean_t an_exec_denied = B_FALSE;

	mode = (fmode & (S_IFMT \| S_ISUID \| S_ISGID \| S_ISVTX));

	while ((acep = zfs_acl_next_ace(aclp, acep, &who,
	&access_mask, &iflags, &type))) {

	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	entry_type = (iflags & ACE_TYPE_FLAGS);

	/*
	* Skip over any inherit_only ACEs
	*/
	if (iflags & ACE_INHERIT_ONLY_ACE)
	continue;

	if (entry_type == ACE_OWNER \|\| (entry_type == 0 &&
	who == fuid)) {
	if ((access_mask & ACE_READ_DATA) &&
	(!(seen & S_IRUSR))) {
	seen \|= S_IRUSR;
	if (type == ALLOW) {
	mode \|= S_IRUSR;
	}
	}
	if ((access_mask & ACE_WRITE_DATA) &&
	(!(seen & S_IWUSR))) {
	seen \|= S_IWUSR;
	if (type == ALLOW) {
	mode \|= S_IWUSR;
	}
	}
	if ((access_mask & ACE_EXECUTE) &&
	(!(seen & S_IXUSR))) {
	seen \|= S_IXUSR;
	if (type == ALLOW) {
	mode \|= S_IXUSR;
	}
	}
	} else if (entry_type == OWNING_GROUP \|\|
	(entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
	if ((access_mask & ACE_READ_DATA) &&
	(!(seen & S_IRGRP))) {
	seen \|= S_IRGRP;
	if (type == ALLOW) {
	mode \|= S_IRGRP;
	}
	}
	if ((access_mask & ACE_WRITE_DATA) &&
	(!(seen & S_IWGRP))) {
	seen \|= S_IWGRP;
	if (type == ALLOW) {
	mode \|= S_IWGRP;
	}
	}
	if ((access_mask & ACE_EXECUTE) &&
	(!(seen & S_IXGRP))) {
	seen \|= S_IXGRP;
	if (type == ALLOW) {
	mode \|= S_IXGRP;
	}
	}
	} else if (entry_type == ACE_EVERYONE) {
	if ((access_mask & ACE_READ_DATA)) {
	if (!(seen & S_IRUSR)) {
	seen \|= S_IRUSR;
	if (type == ALLOW) {
	mode \|= S_IRUSR;
	}
	}
	if (!(seen & S_IRGRP)) {
	seen \|= S_IRGRP;
	if (type == ALLOW) {
	mode \|= S_IRGRP;
	}
	}
	if (!(seen & S_IROTH)) {
	seen \|= S_IROTH;
	if (type == ALLOW) {
	mode \|= S_IROTH;
	}
	}
	}
	if ((access_mask & ACE_WRITE_DATA)) {
	if (!(seen & S_IWUSR)) {
	seen \|= S_IWUSR;
	if (type == ALLOW) {
	mode \|= S_IWUSR;
	}
	}
	if (!(seen & S_IWGRP)) {
	seen \|= S_IWGRP;
	if (type == ALLOW) {
	mode \|= S_IWGRP;
	}
	}
	if (!(seen & S_IWOTH)) {
	seen \|= S_IWOTH;
	if (type == ALLOW) {
	mode \|= S_IWOTH;
	}
	}
	}
	if ((access_mask & ACE_EXECUTE)) {
	if (!(seen & S_IXUSR)) {
	seen \|= S_IXUSR;
	if (type == ALLOW) {
	mode \|= S_IXUSR;
	}
	}
	if (!(seen & S_IXGRP)) {
	seen \|= S_IXGRP;
	if (type == ALLOW) {
	mode \|= S_IXGRP;
	}
	}
	if (!(seen & S_IXOTH)) {
	seen \|= S_IXOTH;
	if (type == ALLOW) {
	mode \|= S_IXOTH;
	}
	}
	}
	} else {
	/*
	* Only care if this IDENTIFIER_GROUP or
	* USER ACE denies execute access to someone,
	* mode is not affected
	*/
	if ((access_mask & ACE_EXECUTE) && type == DENY)
	an_exec_denied = B_TRUE;
	}
	}

	/*
	* Failure to allow is effectively a deny, so execute permission
	* is denied if it was never mentioned or if we explicitly
	* weren't allowed it.
	*/
	if (!an_exec_denied &&
	((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS \|\|
	(mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
	an_exec_denied = B_TRUE;

	if (an_exec_denied)
	*pflags &= ~ZFS_NO_EXECS_DENIED;
	else
	*pflags \|= ZFS_NO_EXECS_DENIED;

	return (mode);
	}

	/*
	* Read an external acl object. If the intent is to modify, always
	* create a new acl and leave any cached acl in place.
	*/
	int
	zfs_acl_node_read(znode_t zp, boolean_t have_lock, zfs_acl_t *aclpp,
	boolean_t will_modify)
	{
	zfs_acl_t *aclp;
	int aclsize;
	int acl_count;
	zfs_acl_node_t *aclnode;
	zfs_acl_phys_t znode_acl;
	int version;
	int error;

	ASSERT(MUTEX_HELD(&zp->z_acl_lock));
	if (zp->z_zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(zp), __func__);

	if (zp->z_acl_cached && !will_modify) {
	*aclpp = zp->z_acl_cached;
	return (0);
	}

	version = zfs_znode_acl_version(zp);

	if ((error = zfs_acl_znode_info(zp, &aclsize,
	&acl_count, &znode_acl)) != 0) {
	goto done;
	}

	aclp = zfs_acl_alloc(version);

	aclp->z_acl_count = acl_count;
	aclp->z_acl_bytes = aclsize;

	aclnode = zfs_acl_node_alloc(aclsize);
	aclnode->z_ace_count = aclp->z_acl_count;
	aclnode->z_size = aclsize;

	if (!zp->z_is_sa) {
	if (znode_acl.z_acl_extern_obj) {
	error = dmu_read(zp->z_zfsvfs->z_os,
	znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
	aclnode->z_acldata, DMU_READ_PREFETCH);
	} else {
	memcpy(aclnode->z_acldata, znode_acl.z_ace_data,
	aclnode->z_size);
	}
	} else {
	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zp->z_zfsvfs),
	aclnode->z_acldata, aclnode->z_size);
	}

	if (error != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	/* convert checksum errors into IO errors */
	if (error == ECKSUM)
	error = SET_ERROR(EIO);
	goto done;
	}

	list_insert_head(&aclp->z_acl, aclnode);

	*aclpp = aclp;
	if (!will_modify)
	zp->z_acl_cached = aclp;
	done:
	return (error);
	}

	void
	zfs_acl_data_locator(void *dataptr, uint32_t length, uint32_t buflen,
	boolean_t start, void *userdata)
	{
	(void) buflen;
	zfs_acl_locator_cb_t cb = (zfs_acl_locator_cb_t )userdata;

	if (start) {
	cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
	} else {
	cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
	cb->cb_acl_node);
	}
	*dataptr = cb->cb_acl_node->z_acldata;
	*length = cb->cb_acl_node->z_size;
	}

	int
	zfs_acl_chown_setattr(znode_t *zp)
	{
	int error;
	zfs_acl_t *aclp;

	if (zp->z_zfsvfs->z_replay == B_FALSE) {
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	ASSERT_VOP_IN_SEQC(ZTOV(zp));
	}
	ASSERT(MUTEX_HELD(&zp->z_acl_lock));

	if ((error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE)) == 0)
	zp->z_mode = zfs_mode_compute(zp->z_mode, aclp,
	&zp->z_pflags, zp->z_uid, zp->z_gid);
	return (error);
	}

	/*
	* common code for setting ACLs.
	*
	* This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
	* zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
	* already checked the acl and knows whether to inherit.
	*/
	int
	zfs_aclset_common(znode_t zp, zfs_acl_t aclp, cred_t cr, dmu_tx_t tx)
	{
	int error;
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	dmu_object_type_t otype;
	zfs_acl_locator_cb_t locate = { 0 };
	uint64_t mode;
	sa_bulk_attr_t bulk[5];
	uint64_t ctime[2];
	int count = 0;
	zfs_acl_phys_t acl_phys;

	if (zp->z_zfsvfs->z_replay == B_FALSE) {
	ASSERT_VOP_IN_SEQC(ZTOV(zp));
	}

	mode = zp->z_mode;

	mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
	zp->z_uid, zp->z_gid);

	zp->z_mode = mode;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&mode, sizeof (mode));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, sizeof (zp->z_pflags));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, sizeof (ctime));

	if (zp->z_acl_cached) {
	zfs_acl_free(zp->z_acl_cached);
	zp->z_acl_cached = NULL;
	}

	/*
	* Upgrade needed?
	*/
	if (!zfsvfs->z_use_fuids) {
	otype = DMU_OT_OLDACL;
	} else {
	if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
	(zfsvfs->z_version >= ZPL_VERSION_FUID))
	zfs_acl_xform(zp, aclp, cr);
	ASSERT3U(aclp->z_version, >=, ZFS_ACL_VERSION_FUID);
	otype = DMU_OT_ACL;
	}

	/*
	* Arrgh, we have to handle old on disk format
	* as well as newer (preferred) SA format.
	*/

	if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
	locate.cb_aclp = aclp;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
	zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
	NULL, &aclp->z_acl_count, sizeof (uint64_t));
	} else { /* Painful legacy way */
	zfs_acl_node_t *aclnode;
	uint64_t off = 0;
	uint64_t aoid;

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
	&acl_phys, sizeof (acl_phys))) != 0)
	return (error);

	aoid = acl_phys.z_acl_extern_obj;

	if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	/*
	* If ACL was previously external and we are now
	* converting to new ACL format then release old
	* ACL object and create a new one.
	*/
	if (aoid &&
	aclp->z_version != acl_phys.z_acl_version) {
	error = dmu_object_free(zfsvfs->z_os, aoid, tx);
	if (error)
	return (error);
	aoid = 0;
	}
	if (aoid == 0) {
	aoid = dmu_object_alloc(zfsvfs->z_os,
	otype, aclp->z_acl_bytes,
	otype == DMU_OT_ACL ?
	DMU_OT_SYSACL : DMU_OT_NONE,
	otype == DMU_OT_ACL ?
	DN_OLD_MAX_BONUSLEN : 0, tx);
	} else {
	(void) dmu_object_set_blocksize(zfsvfs->z_os,
	aoid, aclp->z_acl_bytes, 0, tx);
	}
	acl_phys.z_acl_extern_obj = aoid;
	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	if (aclnode->z_ace_count == 0)
	continue;
	dmu_write(zfsvfs->z_os, aoid, off,
	aclnode->z_size, aclnode->z_acldata, tx);
	off += aclnode->z_size;
	}
	} else {
	void *start = acl_phys.z_ace_data;
	/*
	* Migrating back embedded?
	*/
	if (acl_phys.z_acl_extern_obj) {
	error = dmu_object_free(zfsvfs->z_os,
	acl_phys.z_acl_extern_obj, tx);
	if (error)
	return (error);
	acl_phys.z_acl_extern_obj = 0;
	}

	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	if (aclnode->z_ace_count == 0)
	continue;
	memcpy(start, aclnode->z_acldata,
	aclnode->z_size);
	start = (caddr_t)start + aclnode->z_size;
	}
	}
	/*
	* If Old version then swap count/bytes to match old
	* layout of znode_acl_phys_t.
	*/
	if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
	acl_phys.z_acl_size = aclp->z_acl_count;
	acl_phys.z_acl_count = aclp->z_acl_bytes;
	} else {
	acl_phys.z_acl_size = aclp->z_acl_bytes;
	acl_phys.z_acl_count = aclp->z_acl_count;
	}
	acl_phys.z_acl_version = aclp->z_version;

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
	&acl_phys, sizeof (acl_phys));
	}

	/*
	* Replace ACL wide bits, but first clear them.
	*/
	zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;

	zp->z_pflags \|= aclp->z_hints;

	if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
	zp->z_pflags \|= ZFS_ACL_TRIVIAL;

	zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
	return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
	}

	static void
	zfs_acl_chmod(vtype_t vtype, uint64_t mode, boolean_t split, boolean_t trim,
	zfs_acl_t *aclp)
	{
	void *acep = NULL;
	uint64_t who;
	int new_count, new_bytes;
	int ace_size;
	int entry_type;
	uint16_t iflags, type;
	uint32_t access_mask;
	zfs_acl_node_t *newnode;
	size_t abstract_size = aclp->z_ops->ace_abstract_size();
	void *zacep;
	boolean_t isdir;
	trivial_acl_t masks;

	new_count = new_bytes = 0;

	isdir = (vtype == VDIR);

	acl_trivial_access_masks((mode_t)mode, isdir, &masks);

	newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);

	zacep = newnode->z_acldata;
	if (masks.allow0) {
	zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}
	if (masks.deny1) {
	zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}
	if (masks.deny2) {
	zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}

	while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
	&iflags, &type))) {
	entry_type = (iflags & ACE_TYPE_FLAGS);
	/*
	* ACEs used to represent the file mode may be divided
	* into an equivalent pair of inherit-only and regular
	* ACEs, if they are inheritable.
	* Skip regular ACEs, which are replaced by the new mode.
	*/
	if (split && (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)) {
	if (!isdir \|\| !(iflags &
	(ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	continue;
	/*
	* We preserve owner@, group@, or @everyone
	* permissions, if they are inheritable, by
	* copying them to inherit_only ACEs. This
	* prevents inheritable permissions from being
	* altered along with the file mode.
	*/
	iflags \|= ACE_INHERIT_ONLY_ACE;
	}

	/*
	* If this ACL has any inheritable ACEs, mark that in
	* the hints (which are later masked into the pflags)
	* so create knows to do inheritance.
	*/
	if (isdir && (iflags &
	(ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	if ((type != ALLOW && type != DENY) \|\|
	(iflags & ACE_INHERIT_ONLY_ACE)) {
	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	aclp->z_hints \|= ZFS_ACL_OBJ_ACE;
	break;
	}
	} else {
	/*
	* Limit permissions granted by ACEs to be no greater
	* than permissions of the requested group mode.
	* Applies when the "aclmode" property is set to
	* "groupmask".
	*/
	if ((type == ALLOW) && trim)
	access_mask &= masks.group;
	}
	zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
	ace_size = aclp->z_ops->ace_size(acep);
	zacep = (void *)((uintptr_t)zacep + ace_size);
	new_count++;
	new_bytes += ace_size;
	}
	zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);

	new_count += 3;
	new_bytes += abstract_size * 3;
	zfs_acl_release_nodes(aclp);
	aclp->z_acl_count = new_count;
	aclp->z_acl_bytes = new_bytes;
	newnode->z_ace_count = new_count;
	newnode->z_size = new_bytes;
	list_insert_tail(&aclp->z_acl, newnode);
	}

	int
	zfs_acl_chmod_setattr(znode_t zp, zfs_acl_t *aclp, uint64_t mode)
	{
	int error = 0;

	mutex_enter(&zp->z_acl_lock);
	if (zp->z_zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_DISCARD)
	*aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
	else
	error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);

	if (error == 0) {
	(*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
	zfs_acl_chmod(ZTOV(zp)->v_type, mode, B_TRUE,
	(zp->z_zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
	}
	mutex_exit(&zp->z_acl_lock);

	return (error);
	}

	/*
	* Should ACE be inherited?
	*/
	static int
	zfs_ace_can_use(vtype_t vtype, uint16_t acep_flags)
	{
	int iflags = (acep_flags & 0xf);

	if ((vtype == VDIR) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
	return (1);
	else if (iflags & ACE_FILE_INHERIT_ACE)
	return (!((vtype == VDIR) &&
	(iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
	return (0);
	}

	/*
	* inherit inheritable ACEs from parent
	*/
	static zfs_acl_t *
	zfs_acl_inherit(zfsvfs_t zfsvfs, vtype_t vtype, zfs_acl_t paclp,
	uint64_t mode, boolean_t *need_chmod)
	{
	void *pacep = NULL;
	void *acep;
	zfs_acl_node_t *aclnode;
	zfs_acl_t *aclp = NULL;
	uint64_t who;
	uint32_t access_mask;
	uint16_t iflags, newflags, type;
	size_t ace_size;
	void data1, data2;
	size_t data1sz, data2sz;
	uint_t aclinherit;
	boolean_t isdir = (vtype == VDIR);
	boolean_t isreg = (vtype == VREG);

	*need_chmod = B_TRUE;

	aclp = zfs_acl_alloc(paclp->z_version);
	aclinherit = zfsvfs->z_acl_inherit;
	if (aclinherit == ZFS_ACL_DISCARD \|\| vtype == VLNK)
	return (aclp);

	while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
	&access_mask, &iflags, &type))) {

	/*
	* don't inherit bogus ACEs
	*/
	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	/*
	* Check if ACE is inheritable by this vnode
	*/
	if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) \|\|
	!zfs_ace_can_use(vtype, iflags))
	continue;

	/*
	* If owner@, group@, or everyone@ inheritable
	* then zfs_acl_chmod() isn't needed.
	*/
	if ((aclinherit == ZFS_ACL_PASSTHROUGH \|\|
	aclinherit == ZFS_ACL_PASSTHROUGH_X) &&
	((iflags & (ACE_OWNER\|ACE_EVERYONE)) \|\|
	((iflags & OWNING_GROUP) == OWNING_GROUP)) &&
	(isreg \|\| (isdir && (iflags & ACE_DIRECTORY_INHERIT_ACE))))
	*need_chmod = B_FALSE;

	/*
	* Strip inherited execute permission from file if
	* not in mode
	*/
	if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
	!isdir && ((mode & (S_IXUSR\|S_IXGRP\|S_IXOTH)) == 0)) {
	access_mask &= ~ACE_EXECUTE;
	}

	/*
	* Strip write_acl and write_owner from permissions
	* when inheriting an ACE
	*/
	if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
	access_mask &= ~RESTRICTED_CLEAR;
	}

	ace_size = aclp->z_ops->ace_size(pacep);
	aclnode = zfs_acl_node_alloc(ace_size);
	list_insert_tail(&aclp->z_acl, aclnode);
	acep = aclnode->z_acldata;

	zfs_set_ace(aclp, acep, access_mask, type,
	who, iflags\|ACE_INHERITED_ACE);

	/*
	* Copy special opaque data if any
	*/
	if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
	data2sz = aclp->z_ops->ace_data(acep, &data2);
	VERIFY3U(data2sz, ==, data1sz);
	memcpy(data2, data1, data2sz);
	}

	aclp->z_acl_count++;
	aclnode->z_ace_count++;
	aclp->z_acl_bytes += aclnode->z_size;
	newflags = aclp->z_ops->ace_flags_get(acep);

	/*
	* If ACE is not to be inherited further, or if the vnode is
	* not a directory, remove all inheritance flags
	*/
	if (!isdir \|\| (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
	newflags &= ~ALL_INHERIT;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	continue;
	}

	/*
	* This directory has an inheritable ACE
	*/
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	/*
	* If only FILE_INHERIT is set then turn on
	* inherit_only
	*/
	if ((iflags & (ACE_FILE_INHERIT_ACE \|
	ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
	newflags \|= ACE_INHERIT_ONLY_ACE;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	} else {
	newflags &= ~ACE_INHERIT_ONLY_ACE;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	}
	}
	if (zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
	aclp->z_acl_count != 0) {
	*need_chmod = B_FALSE;
	}

	return (aclp);
	}

	/*
	* Create file system object initial permissions
	* including inheritable ACEs.
	* Also, create FUIDs for owner and group.
	*/
	int
	zfs_acl_ids_create(znode_t dzp, int flag, vattr_t vap, cred_t *cr,
	vsecattr_t vsecp, zfs_acl_ids_t acl_ids)
	{
	int error;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zfs_acl_t *paclp;
	gid_t gid;
	boolean_t need_chmod = B_TRUE;
	boolean_t trim = B_FALSE;
	boolean_t inherited = B_FALSE;

	if ((flag & IS_ROOT_NODE) == 0) {
	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
	} else
	ASSERT3P(dzp->z_vnode, ==, NULL);
	memset(acl_ids, 0, sizeof (zfs_acl_ids_t));
	acl_ids->z_mode = MAKEIMODE(vap->va_type, vap->va_mode);

	if (vsecp)
	if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_type, vsecp, cr,
	&acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
	return (error);
	/*
	* Determine uid and gid.
	*/
	if ((flag & IS_ROOT_NODE) \|\| zfsvfs->z_replay \|\|
	((flag & IS_XATTR) && (vap->va_type == VDIR))) {
	acl_ids->z_fuid = zfs_fuid_create(zfsvfs,
	(uint64_t)vap->va_uid, cr,
	ZFS_OWNER, &acl_ids->z_fuidp);
	acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
	(uint64_t)vap->va_gid, cr,
	ZFS_GROUP, &acl_ids->z_fuidp);
	gid = vap->va_gid;
	} else {
	uid_t id = crgetuid(cr);
	if (IS_EPHEMERAL(id))
	id = UID_NOBODY;
	acl_ids->z_fuid = (uint64_t)id;
	acl_ids->z_fgid = 0;
	if (vap->va_mask & AT_GID) {
	acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
	(uint64_t)vap->va_gid,
	cr, ZFS_GROUP, &acl_ids->z_fuidp);
	gid = vap->va_gid;
	if (acl_ids->z_fgid != dzp->z_gid &&
	!groupmember(vap->va_gid, cr) &&
	secpolicy_vnode_create_gid(cr) != 0)
	acl_ids->z_fgid = 0;
	}
	if (acl_ids->z_fgid == 0) {
	const char *domain;
	uint32_t rid;

	acl_ids->z_fgid = dzp->z_gid;
	gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
	cr, ZFS_GROUP);

	if (zfsvfs->z_use_fuids &&
	IS_EPHEMERAL(acl_ids->z_fgid)) {
	domain =
	zfs_fuid_idx_domain(&zfsvfs->z_fuid_idx,
	FUID_INDEX(acl_ids->z_fgid));
	rid = FUID_RID(acl_ids->z_fgid);
	zfs_fuid_node_add(&acl_ids->z_fuidp,
	domain, rid, FUID_INDEX(acl_ids->z_fgid),
	acl_ids->z_fgid, ZFS_GROUP);
	}
	}
	}

	/*
	* If we're creating a directory, and the parent directory has the
	* set-GID bit set, set in on the new directory.
	* Otherwise, if the user is neither privileged nor a member of the
	* file's new group, clear the file's set-GID bit.
	*/

	if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
	(vap->va_type == VDIR)) {
	acl_ids->z_mode \|= S_ISGID;
	} else {
	if ((acl_ids->z_mode & S_ISGID) &&
	secpolicy_vnode_setids_setgids(ZTOV(dzp), cr, gid) != 0)
	acl_ids->z_mode &= ~S_ISGID;
	}

	if (acl_ids->z_aclp == NULL) {
	mutex_enter(&dzp->z_acl_lock);
	if (!(flag & IS_ROOT_NODE) &&
	(dzp->z_pflags & ZFS_INHERIT_ACE) &&
	!(dzp->z_pflags & ZFS_XATTR)) {
	VERIFY0(zfs_acl_node_read(dzp, B_TRUE,
	&paclp, B_FALSE));
	acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
	vap->va_type, paclp, acl_ids->z_mode, &need_chmod);
	inherited = B_TRUE;
	} else {
	acl_ids->z_aclp =
	zfs_acl_alloc(zfs_acl_version_zp(dzp));
	acl_ids->z_aclp->z_hints \|= ZFS_ACL_TRIVIAL;
	}
	mutex_exit(&dzp->z_acl_lock);

	if (need_chmod) {
	if (vap->va_type == VDIR)
	acl_ids->z_aclp->z_hints \|=
	ZFS_ACL_AUTO_INHERIT;

	if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
	zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
	zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
	trim = B_TRUE;
	zfs_acl_chmod(vap->va_type, acl_ids->z_mode, B_FALSE,
	trim, acl_ids->z_aclp);
	}
	}

	if (inherited \|\| vsecp) {
	acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
	acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
	acl_ids->z_fuid, acl_ids->z_fgid);
	if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
	acl_ids->z_aclp->z_hints \|= ZFS_ACL_TRIVIAL;
	}

	return (0);
	}

	/*
	* Free ACL and fuid_infop, but not the acl_ids structure
	*/
	void
	zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
	{
	if (acl_ids->z_aclp)
	zfs_acl_free(acl_ids->z_aclp);
	if (acl_ids->z_fuidp)
	zfs_fuid_info_free(acl_ids->z_fuidp);
	acl_ids->z_aclp = NULL;
	acl_ids->z_fuidp = NULL;
	}

	boolean_t
	zfs_acl_ids_overquota(zfsvfs_t zv, zfs_acl_ids_t acl_ids, uint64_t projid)
	{
	return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) \|\|
	zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) \|\|
	(projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
	zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
	}

	/*
	* Retrieve a file's ACL
	*/
	int
	zfs_getacl(znode_t zp, vsecattr_t vsecp, boolean_t skipaclchk, cred_t *cr)
	{
	zfs_acl_t *aclp;
	ulong_t mask;
	int error;
	int count = 0;
	int largeace = 0;

	mask = vsecp->vsa_mask & (VSA_ACE \| VSA_ACECNT \|
	VSA_ACE_ACLFLAGS \| VSA_ACE_ALLTYPES);

	if (mask == 0)
	return (SET_ERROR(ENOSYS));

	if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
	return (error);

	mutex_enter(&zp->z_acl_lock);

	if (zp->z_zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
	error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
	if (error != 0) {
	mutex_exit(&zp->z_acl_lock);
	return (error);
	}

	/*
	* Scan ACL to determine number of ACEs
	*/
	if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
	void *zacep = NULL;
	uint64_t who;
	uint32_t access_mask;
	uint16_t type, iflags;

	while ((zacep = zfs_acl_next_ace(aclp, zacep,
	&who, &access_mask, &iflags, &type))) {
	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	largeace++;
	continue;
	default:
	count++;
	}
	}
	vsecp->vsa_aclcnt = count;
	} else
	count = (int)aclp->z_acl_count;

	if (mask & VSA_ACECNT) {
	vsecp->vsa_aclcnt = count;
	}

	if (mask & VSA_ACE) {
	size_t aclsz;

	aclsz = count * sizeof (ace_t) +
	sizeof (ace_object_t) * largeace;

	vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
	vsecp->vsa_aclentsz = aclsz;

	if (aclp->z_version == ZFS_ACL_VERSION_FUID)
	zfs_copy_fuid_2_ace(zp->z_zfsvfs, aclp, cr,
	vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
	else {
	zfs_acl_node_t *aclnode;
	void *start = vsecp->vsa_aclentp;

	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	memcpy(start, aclnode->z_acldata,
	aclnode->z_size);
	start = (caddr_t)start + aclnode->z_size;
	}
	ASSERT3U((caddr_t)start - (caddr_t)vsecp->vsa_aclentp,
	==, aclp->z_acl_bytes);
	}
	}
	if (mask & VSA_ACE_ACLFLAGS) {
	vsecp->vsa_aclflags = 0;
	if (zp->z_pflags & ZFS_ACL_DEFAULTED)
	vsecp->vsa_aclflags \|= ACL_DEFAULTED;
	if (zp->z_pflags & ZFS_ACL_PROTECTED)
	vsecp->vsa_aclflags \|= ACL_PROTECTED;
	if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
	vsecp->vsa_aclflags \|= ACL_AUTO_INHERIT;
	}

	mutex_exit(&zp->z_acl_lock);

	return (0);
	}

	int
	zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_type,
	vsecattr_t vsecp, cred_t cr, zfs_fuid_info_t fuidp, zfs_acl_t zaclp)
	{
	zfs_acl_t *aclp;
	zfs_acl_node_t *aclnode;
	int aclcnt = vsecp->vsa_aclcnt;
	int error;

	if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES \|\| vsecp->vsa_aclcnt <= 0)
	return (SET_ERROR(EINVAL));

	aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));

	aclp->z_hints = 0;
	aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
	if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
	if ((error = zfs_copy_ace_2_oldace(obj_type, aclp,
	(ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
	aclcnt, &aclnode->z_size)) != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	return (error);
	}
	} else {
	if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_type, aclp,
	vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
	&aclnode->z_size, fuidp, cr)) != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	return (error);
	}
	}
	aclp->z_acl_bytes = aclnode->z_size;
	aclnode->z_ace_count = aclcnt;
	aclp->z_acl_count = aclcnt;
	list_insert_head(&aclp->z_acl, aclnode);

	/*
	* If flags are being set then add them to z_hints
	*/
	if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
	if (vsecp->vsa_aclflags & ACL_PROTECTED)
	aclp->z_hints \|= ZFS_ACL_PROTECTED;
	if (vsecp->vsa_aclflags & ACL_DEFAULTED)
	aclp->z_hints \|= ZFS_ACL_DEFAULTED;
	if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
	aclp->z_hints \|= ZFS_ACL_AUTO_INHERIT;
	}

	*zaclp = aclp;

	return (0);
	}

	/*
	* Set a file's ACL
	*/
	int
	zfs_setacl(znode_t zp, vsecattr_t vsecp, boolean_t skipaclchk, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zilog_t *zilog = zfsvfs->z_log;
	ulong_t mask = vsecp->vsa_mask & (VSA_ACE \| VSA_ACECNT);
	dmu_tx_t *tx;
	int error;
	zfs_acl_t *aclp;
	zfs_fuid_info_t *fuidp = NULL;
	boolean_t fuid_dirtied;
	uint64_t acl_obj;

	if (zp->z_zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	if (mask == 0)
	return (SET_ERROR(ENOSYS));

	if (zp->z_pflags & ZFS_IMMUTABLE)
	return (SET_ERROR(EPERM));

	if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
	return (error);

	error = zfs_vsec_2_aclp(zfsvfs, ZTOV(zp)->v_type, vsecp, cr, &fuidp,
	&aclp);
	if (error)
	return (error);

	/*
	* If ACL wide flags aren't being set then preserve any
	* existing flags.
	*/
	if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
	aclp->z_hints \|=
	(zp->z_pflags & V4_ACL_WIDE_FLAGS);
	}
	top:
	mutex_enter(&zp->z_acl_lock);

	tx = dmu_tx_create(zfsvfs->z_os);

	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);

	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);

	/*
	* If old version and ACL won't fit in bonus and we aren't
	* upgrading then take out necessary DMU holds
	*/

	if ((acl_obj = zfs_external_acl(zp)) != 0) {
	if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
	zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
	dmu_tx_hold_free(tx, acl_obj, 0,
	DMU_OBJECT_END);
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
	aclp->z_acl_bytes);
	} else {
	dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
	}
	} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
	}

	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_NOWAIT);
	if (error) {
	mutex_exit(&zp->z_acl_lock);

	if (error == ERESTART) {
	dmu_tx_wait(tx);
	dmu_tx_abort(tx);
	goto top;
	}
	dmu_tx_abort(tx);
	zfs_acl_free(aclp);
	return (error);
	}

	error = zfs_aclset_common(zp, aclp, cr, tx);
	ASSERT0(error);
	ASSERT3P(zp->z_acl_cached, ==, NULL);
	zp->z_acl_cached = aclp;

	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	zfs_log_acl(zilog, tx, zp, vsecp, fuidp);

	if (fuidp)
	zfs_fuid_info_free(fuidp);
	dmu_tx_commit(tx);
	mutex_exit(&zp->z_acl_lock);

	return (error);
	}

	/*
	* Check accesses of interest (AoI) against attributes of the dataset
	* such as read-only. Returns zero if no AoI conflict with dataset
	* attributes, otherwise an appropriate errno is returned.
	*/
	static int
	zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
	{
	if ((v4_mode & WRITE_MASK) &&
	(zp->z_zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
	(!IS_DEVVP(ZTOV(zp)) \|\|
	(IS_DEVVP(ZTOV(zp)) && (v4_mode & WRITE_MASK_ATTRS)))) {
	return (SET_ERROR(EROFS));
	}

	/*
	* Intentionally allow ZFS_READONLY through here.
	* See zfs_zaccess_common().
	*/
	if ((v4_mode & WRITE_MASK_DATA) &&
	(zp->z_pflags & ZFS_IMMUTABLE)) {
	return (SET_ERROR(EPERM));
	}

	/*
	* In FreeBSD we allow to modify directory's content is ZFS_NOUNLINK
	* (sunlnk) is set. We just don't allow directory removal, which is
	* handled in zfs_zaccess_delete().
	*/
	if ((v4_mode & ACE_DELETE) &&
	(zp->z_pflags & ZFS_NOUNLINK)) {
	return (EPERM);
	}

	if (((v4_mode & (ACE_READ_DATA\|ACE_EXECUTE)) &&
	(zp->z_pflags & ZFS_AV_QUARANTINED))) {
	return (SET_ERROR(EACCES));
	}

	return (0);
	}

	/*
	* The primary usage of this function is to loop through all of the
	* ACEs in the znode, determining what accesses of interest (AoI) to
	* the caller are allowed or denied. The AoI are expressed as bits in
	* the working_mode parameter. As each ACE is processed, bits covered
	* by that ACE are removed from the working_mode. This removal
	* facilitates two things. The first is that when the working mode is
	* empty (= 0), we know we've looked at all the AoI. The second is
	* that the ACE interpretation rules don't allow a later ACE to undo
	* something granted or denied by an earlier ACE. Removing the
	* discovered access or denial enforces this rule. At the end of
	* processing the ACEs, all AoI that were found to be denied are
	* placed into the working_mode, giving the caller a mask of denied
	* accesses. Returns:
	* 0 if all AoI granted
	* EACCESS if the denied mask is non-zero
	* other error if abnormal failure (e.g., IO error)
	*
	* A secondary usage of the function is to determine if any of the
	* AoI are granted. If an ACE grants any access in
	* the working_mode, we immediately short circuit out of the function.
	* This mode is chosen by setting anyaccess to B_TRUE. The
	* working_mode is not a denied access mask upon exit if the function
	* is used in this manner.
	*/
	static int
	zfs_zaccess_aces_check(znode_t zp, uint32_t working_mode,
	boolean_t anyaccess, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zfs_acl_t *aclp;
	int error;
	uid_t uid = crgetuid(cr);
	uint64_t who;
	uint16_t type, iflags;
	uint16_t entry_type;
	uint32_t access_mask;
	uint32_t deny_mask = 0;
	zfs_ace_hdr_t *acep = NULL;
	boolean_t checkit;
	uid_t gowner;
	uid_t fowner;

	zfs_fuid_map_ids(zp, cr, &fowner, &gowner);

	mutex_enter(&zp->z_acl_lock);

	if (zp->z_zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(zp), __func__);
	error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
	if (error != 0) {
	mutex_exit(&zp->z_acl_lock);
	return (error);
	}

	ASSERT3P(zp->z_acl_cached, !=, NULL);

	while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
	&iflags, &type))) {
	uint32_t mask_matched;

	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	if (ZTOV(zp)->v_type == VDIR && (iflags & ACE_INHERIT_ONLY_ACE))
	continue;

	/* Skip ACE if it does not affect any AoI */
	mask_matched = (access_mask & *working_mode);
	if (!mask_matched)
	continue;

	entry_type = (iflags & ACE_TYPE_FLAGS);

	checkit = B_FALSE;

	switch (entry_type) {
	case ACE_OWNER:
	if (uid == fowner)
	checkit = B_TRUE;
	break;
	case OWNING_GROUP:
	who = gowner;
	zfs_fallthrough;
	case ACE_IDENTIFIER_GROUP:
	checkit = zfs_groupmember(zfsvfs, who, cr);
	break;
	case ACE_EVERYONE:
	checkit = B_TRUE;
	break;

	/* USER Entry */
	default:
	if (entry_type == 0) {
	uid_t newid;

	newid = zfs_fuid_map_id(zfsvfs, who, cr,
	ZFS_ACE_USER);
	if (newid != UID_NOBODY &&
	uid == newid)
	checkit = B_TRUE;
	break;
	} else {
	mutex_exit(&zp->z_acl_lock);
	return (SET_ERROR(EIO));
	}
	}

	if (checkit) {
	if (type == DENY) {
	DTRACE_PROBE3(zfs__ace__denies,
	znode_t *, zp,
	zfs_ace_hdr_t *, acep,
	uint32_t, mask_matched);
	deny_mask \|= mask_matched;
	} else {
	DTRACE_PROBE3(zfs__ace__allows,
	znode_t *, zp,
	zfs_ace_hdr_t *, acep,
	uint32_t, mask_matched);
	if (anyaccess) {
	mutex_exit(&zp->z_acl_lock);
	return (0);
	}
	}
	*working_mode &= ~mask_matched;
	}

	/* Are we done? */
	if (*working_mode == 0)
	break;
	}

	mutex_exit(&zp->z_acl_lock);

	/* Put the found 'denies' back on the working mode */
	if (deny_mask) {
	*working_mode \|= deny_mask;
	return (SET_ERROR(EACCES));
	} else if (*working_mode) {
	return (-1);
	}

	return (0);
	}

	/*
	* Return true if any access whatsoever granted, we don't actually
	* care what access is granted.
	*/
	boolean_t
	zfs_has_access(znode_t zp, cred_t cr)
	{
	uint32_t have = ACE_ALL_PERMS;

	if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
	uid_t owner;

	owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);
	return (secpolicy_vnode_any_access(cr, ZTOV(zp), owner) == 0);
	}
	return (B_TRUE);
	}

	static int
	zfs_zaccess_common(znode_t zp, uint32_t v4_mode, uint32_t working_mode,
	boolean_t check_privs, boolean_t skipaclchk, cred_t cr)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int err;

	*working_mode = v4_mode;
	*check_privs = B_TRUE;

	/*
	* Short circuit empty requests
	*/
	if (v4_mode == 0 \|\| zfsvfs->z_replay) {
	*working_mode = 0;
	return (0);
	}

	if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
	*check_privs = B_FALSE;
	return (err);
	}

	/*
	* The caller requested that the ACL check be skipped. This
	* would only happen if the caller checked VOP_ACCESS() with a
	* 32 bit ACE mask and already had the appropriate permissions.
	*/
	if (skipaclchk) {
	*working_mode = 0;
	return (0);
	}

	/*
	* Note: ZFS_READONLY represents the "DOS R/O" attribute.
	* When that flag is set, we should behave as if write access
	* were not granted by anything in the ACL. In particular:
	* We _must_ allow writes after opening the file r/w, then
	* setting the DOS R/O attribute, and writing some more.
	* (Similar to how you can write after fchmod(fd, 0444).)
	*
	* Therefore ZFS_READONLY is ignored in the dataset check
	* above, and checked here as if part of the ACL check.
	* Also note: DOS R/O is ignored for directories.
	*/
	if ((v4_mode & WRITE_MASK_DATA) &&
	(ZTOV(zp)->v_type != VDIR) &&
	(zp->z_pflags & ZFS_READONLY)) {
	return (SET_ERROR(EPERM));
	}

	return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
	}

	static int
	zfs_zaccess_append(znode_t zp, uint32_t working_mode, boolean_t *check_privs,
	cred_t *cr)
	{
	if (*working_mode != ACE_WRITE_DATA)
	return (SET_ERROR(EACCES));

	return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
	check_privs, B_FALSE, cr));
	}

	/*
	* Check if VEXEC is allowed.
	*
	* This routine is based on zfs_fastaccesschk_execute which has slowpath
	* calling zfs_zaccess. This would be incorrect on FreeBSD (see
	* zfs_freebsd_access for the difference). Thus this variant let's the
	* caller handle the slowpath (if necessary).
	*
	* On top of that we perform a lockless check for ZFS_NO_EXECS_DENIED.
	*
	* Safe access to znode_t is provided by the vnode lock.
	*/
	int
	zfs_fastaccesschk_execute(znode_t zdp, cred_t cr)
	{
	boolean_t is_attr;

	if (zdp->z_pflags & ZFS_AV_QUARANTINED)
	return (1);

	is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
	(ZTOV(zdp)->v_type == VDIR));
	if (is_attr)
	return (1);

	if (zdp->z_pflags & ZFS_NO_EXECS_DENIED)
	return (0);

	return (1);
	}


	/*
	* Determine whether Access should be granted/denied.
	*
	* The least priv subsystem is always consulted as a basic privilege
	* can define any form of access.
	*/
	int
	zfs_zaccess(znode_t zp, int mode, int flags, boolean_t skipaclchk, cred_t cr)
	{
	uint32_t working_mode;
	int error;
	int is_attr;
	boolean_t check_privs;
	znode_t *xzp = NULL;
	znode_t *check_zp = zp;
	mode_t needed_bits;
	uid_t owner;

	is_attr = ((zp->z_pflags & ZFS_XATTR) && (ZTOV(zp)->v_type == VDIR));

	/*
	* In FreeBSD, we don't care about permissions of individual ADS.
	* Note that not checking them is not just an optimization - without
	* this shortcut, EA operations may bogusly fail with EACCES.
	*/
	if (zp->z_pflags & ZFS_XATTR)
	return (0);

	owner = zfs_fuid_map_id(zp->z_zfsvfs, zp->z_uid, cr, ZFS_OWNER);

	/*
	* Map the bits required to the standard vnode flags VREAD\|VWRITE\|VEXEC
	* in needed_bits. Map the bits mapped by working_mode (currently
	* missing) in missing_bits.
	* Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
	* needed_bits.
	*/
	needed_bits = 0;

	working_mode = mode;
	if ((working_mode & (ACE_READ_ACL\|ACE_READ_ATTRIBUTES)) &&
	owner == crgetuid(cr))
	working_mode &= ~(ACE_READ_ACL\|ACE_READ_ATTRIBUTES);

	if (working_mode & (ACE_READ_DATA\|ACE_READ_NAMED_ATTRS\|
	ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_SYNCHRONIZE))
	needed_bits \|= VREAD;
	if (working_mode & (ACE_WRITE_DATA\|ACE_WRITE_NAMED_ATTRS\|
	ACE_APPEND_DATA\|ACE_WRITE_ATTRIBUTES\|ACE_SYNCHRONIZE))
	needed_bits \|= VWRITE;
	if (working_mode & ACE_EXECUTE)
	needed_bits \|= VEXEC;

	if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
	&check_privs, skipaclchk, cr)) == 0) {
	if (is_attr)
	VN_RELE(ZTOV(xzp));
	return (secpolicy_vnode_access2(cr, ZTOV(zp), owner,
	needed_bits, needed_bits));
	}

	if (error && !check_privs) {
	if (is_attr)
	VN_RELE(ZTOV(xzp));
	return (error);
	}

	if (error && (flags & V_APPEND)) {
	error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
	}

	if (error && check_privs) {
	mode_t checkmode = 0;
	vnode_t *check_vp = ZTOV(check_zp);

	/*
	* First check for implicit owner permission on
	* read_acl/read_attributes
	*/

	error = 0;
	ASSERT3U(working_mode, !=, 0);

	if ((working_mode & (ACE_READ_ACL\|ACE_READ_ATTRIBUTES) &&
	owner == crgetuid(cr)))
	working_mode &= ~(ACE_READ_ACL\|ACE_READ_ATTRIBUTES);

	if (working_mode & (ACE_READ_DATA\|ACE_READ_NAMED_ATTRS\|
	ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_SYNCHRONIZE))
	checkmode \|= VREAD;
	if (working_mode & (ACE_WRITE_DATA\|ACE_WRITE_NAMED_ATTRS\|
	ACE_APPEND_DATA\|ACE_WRITE_ATTRIBUTES\|ACE_SYNCHRONIZE))
	checkmode \|= VWRITE;
	if (working_mode & ACE_EXECUTE)
	checkmode \|= VEXEC;

	error = secpolicy_vnode_access2(cr, check_vp, owner,
	needed_bits & ~checkmode, needed_bits);

	if (error == 0 && (working_mode & ACE_WRITE_OWNER))
	error = secpolicy_vnode_chown(check_vp, cr, owner);
	if (error == 0 && (working_mode & ACE_WRITE_ACL))
	error = secpolicy_vnode_setdac(check_vp, cr, owner);

	if (error == 0 && (working_mode &
	(ACE_DELETE\|ACE_DELETE_CHILD)))
	error = secpolicy_vnode_remove(check_vp, cr);

	if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
	error = secpolicy_vnode_chown(check_vp, cr, owner);
	}
	if (error == 0) {
	/*
	* See if any bits other than those already checked
	* for are still present. If so then return EACCES
	*/
	if (working_mode & ~(ZFS_CHECKED_MASKS)) {
	error = SET_ERROR(EACCES);
	}
	}
	} else if (error == 0) {
	error = secpolicy_vnode_access2(cr, ZTOV(zp), owner,
	needed_bits, needed_bits);
	}


	if (is_attr)
	VN_RELE(ZTOV(xzp));

	return (error);
	}

	/*
	* Translate traditional unix VREAD/VWRITE/VEXEC mode into
	* NFSv4-style ZFS ACL format and call zfs_zaccess()
	*/
	int
	zfs_zaccess_rwx(znode_t zp, mode_t mode, int flags, cred_t cr)
	{
	return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
	}

	/*
	* Access function for secpolicy_vnode_setattr
	*/
	int
	zfs_zaccess_unix(znode_t zp, mode_t mode, cred_t cr)
	{
	int v4_mode = zfs_unix_to_v4(mode >> 6);

	return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
	}

	static int
	zfs_delete_final_check(znode_t zp, znode_t dzp,
	mode_t available_perms, cred_t *cr)
	{
	int error;
	uid_t downer;

	downer = zfs_fuid_map_id(dzp->z_zfsvfs, dzp->z_uid, cr, ZFS_OWNER);

	error = secpolicy_vnode_access2(cr, ZTOV(dzp),
	downer, available_perms, VWRITE\|VEXEC);

	if (error == 0)
	error = zfs_sticky_remove_access(dzp, zp, cr);

	return (error);
	}

	/*
	* Determine whether Access should be granted/deny, without
	* consulting least priv subsystem.
	*
	* The following chart is the recommended NFSv4 enforcement for
	* ability to delete an object.
	*
	* -------------------------------------------------------
	* \| Parent Dir \| Target Object Permissions \|
	* \| permissions \| \|
	* -------------------------------------------------------
	* \| \| ACL Allows \| ACL Denies\| Delete \|
	* \| \| Delete \| Delete \| unspecified\|
	* -------------------------------------------------------
	* \| ACL Allows \| Permit \| Permit \| Permit \|
	* \| DELETE_CHILD \| \|
	* -------------------------------------------------------
	* \| ACL Denies \| Permit \| Deny \| Deny \|
	* \| DELETE_CHILD \| \| \| \|
	* -------------------------------------------------------
	* \| ACL specifies \| \| \| \|
	* \| only allow \| Permit \| Permit \| Permit \|
	* \| write and \| \| \| \|
	* \| execute \| \| \| \|
	* -------------------------------------------------------
	* \| ACL denies \| \| \| \|
	* \| write and \| Permit \| Deny \| Deny \|
	* \| execute \| \| \| \|
	* -------------------------------------------------------
	* ^
	* \|
	* No search privilege, can't even look up file?
	*
	*/
	int
	zfs_zaccess_delete(znode_t dzp, znode_t zp, cred_t *cr)
	{
	uint32_t dzp_working_mode = 0;
	uint32_t zp_working_mode = 0;
	int dzp_error, zp_error;
	mode_t available_perms;
	boolean_t dzpcheck_privs = B_TRUE;
	boolean_t zpcheck_privs = B_TRUE;

	/*
	* We want specific DELETE permissions to
	* take precedence over WRITE/EXECUTE. We don't
	* want an ACL such as this to mess us up.
	* user:joe:write_data:deny,user:joe:delete:allow
	*
	* However, deny permissions may ultimately be overridden
	* by secpolicy_vnode_access().
	*
	* We will ask for all of the necessary permissions and then
	* look at the working modes from the directory and target object
	* to determine what was found.
	*/

	if (zp->z_pflags & (ZFS_IMMUTABLE \| ZFS_NOUNLINK))
	return (SET_ERROR(EPERM));

	/*
	* First row
	* If the directory permissions allow the delete, we are done.
	*/
	if ((dzp_error = zfs_zaccess_common(dzp, ACE_DELETE_CHILD,
	&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr)) == 0)
	return (0);

	/*
	* If target object has delete permission then we are done
	*/
	if ((zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
	&zpcheck_privs, B_FALSE, cr)) == 0)
	return (0);

	ASSERT(dzp_error);
	ASSERT(zp_error);

	if (!dzpcheck_privs)
	return (dzp_error);
	if (!zpcheck_privs)
	return (zp_error);

	/*
	* Second row
	*
	* If directory returns EACCES then delete_child was denied
	* due to deny delete_child. In this case send the request through
	* secpolicy_vnode_remove(). We don't use zfs_delete_final_check()
	* since that could allow the delete based on write/execute permission
	* and we want delete permissions to override write/execute.
	*/

	if (dzp_error == EACCES) {
	/* XXXPJD: s/dzp/zp/ ? */
	return (secpolicy_vnode_remove(ZTOV(dzp), cr));
	}
	/*
	* Third Row
	* only need to see if we have write/execute on directory.
	*/

	dzp_error = zfs_zaccess_common(dzp, ACE_EXECUTE\|ACE_WRITE_DATA,
	&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);

	if (dzp_error != 0 && !dzpcheck_privs)
	return (dzp_error);

	/*
	* Fourth row
	*/

	available_perms = (dzp_working_mode & ACE_WRITE_DATA) ? 0 : VWRITE;
	available_perms \|= (dzp_working_mode & ACE_EXECUTE) ? 0 : VEXEC;

	return (zfs_delete_final_check(zp, dzp, available_perms, cr));

	}

	int
	zfs_zaccess_rename(znode_t sdzp, znode_t szp, znode_t *tdzp,
	znode_t tzp, cred_t cr)
	{
	int add_perm;
	int error;

	if (szp->z_pflags & ZFS_AV_QUARANTINED)
	return (SET_ERROR(EACCES));

	add_perm = (ZTOV(szp)->v_type == VDIR) ?
	ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;

	/*
	* Rename permissions are combination of delete permission +
	* add file/subdir permission.
	*
	* BSD operating systems also require write permission
	* on the directory being moved from one parent directory
	* to another.
	*/
	if (ZTOV(szp)->v_type == VDIR && ZTOV(sdzp) != ZTOV(tdzp)) {
	if ((error = zfs_zaccess(szp, ACE_WRITE_DATA, 0, B_FALSE, cr)))
	return (error);
	}

	/*
	* first make sure we do the delete portion.
	*
	* If that succeeds then check for add_file/add_subdir permissions
	*/

	if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
	return (error);

	/*
	* If we have a tzp, see if we can delete it?
	*/
	if (tzp && (error = zfs_zaccess_delete(tdzp, tzp, cr)))
	return (error);

	/*
	* Now check for add permissions
	*/
	error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);

	return (error);
	}
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	index 32fb5a872190..abb3c0033194 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	@@ -1,252 +1,252 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/kstat.h>

	typedef struct zfs_dbgmsg {
	list_node_t zdm_node;
	time_t zdm_timestamp;
	- int zdm_size;
	+ uint_t zdm_size;
	char zdm_msg[1]; /* variable length allocation */
	} zfs_dbgmsg_t;

	static list_t zfs_dbgmsgs;
	-static int zfs_dbgmsg_size = 0;
	+static uint_t zfs_dbgmsg_size = 0;
	static kmutex_t zfs_dbgmsgs_lock;
	-int zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
	+uint_t zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
	static kstat_t *zfs_dbgmsg_kstat;

	/*
	* Internal ZFS debug messages are enabled by default.
	*
	* # Print debug messages as they're logged
	* dtrace -n 'zfs-dbgmsg { print(stringof(arg0)); }'
	*
	* # Print all logged dbgmsg entries
	* sysctl kstat.zfs.misc.dbgmsg
	*
	* # Disable the kernel debug message log.
	* sysctl vfs.zfs.dbgmsg_enable=0
	*/
	int zfs_dbgmsg_enable = B_TRUE;

	static int
	zfs_dbgmsg_headers(char *buf, size_t size)
	{
	(void) snprintf(buf, size, "%-12s %-8s\n", "timestamp", "message");

	return (0);
	}

	static int
	zfs_dbgmsg_data(char buf, size_t size, void data)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )data;

	(void) snprintf(buf, size, "%-12llu %-s\n",
	(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);

	return (0);
	}

	static void *
	zfs_dbgmsg_addr(kstat_t *ksp, loff_t n)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )ksp->ks_private;

	ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));

	if (n == 0)
	ksp->ks_private = list_head(&zfs_dbgmsgs);
	else if (zdm)
	ksp->ks_private = list_next(&zfs_dbgmsgs, zdm);

	return (ksp->ks_private);
	}

	static void
	-zfs_dbgmsg_purge(int max_size)
	+zfs_dbgmsg_purge(uint_t max_size)
	{
	zfs_dbgmsg_t *zdm;
	- int size;
	+ uint_t size;

	ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));

	while (zfs_dbgmsg_size > max_size) {
	zdm = list_remove_head(&zfs_dbgmsgs);
	if (zdm == NULL)
	return;

	size = zdm->zdm_size;
	kmem_free(zdm, size);
	zfs_dbgmsg_size -= size;
	}
	}

	static int
	zfs_dbgmsg_update(kstat_t *ksp, int rw)
	{
	if (rw == KSTAT_WRITE)
	zfs_dbgmsg_purge(0);

	return (0);
	}

	void
	zfs_dbgmsg_init(void)
	{
	list_create(&zfs_dbgmsgs, sizeof (zfs_dbgmsg_t),
	offsetof(zfs_dbgmsg_t, zdm_node));
	mutex_init(&zfs_dbgmsgs_lock, NULL, MUTEX_DEFAULT, NULL);

	zfs_dbgmsg_kstat = kstat_create("zfs", 0, "dbgmsg", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
	if (zfs_dbgmsg_kstat) {
	zfs_dbgmsg_kstat->ks_lock = &zfs_dbgmsgs_lock;
	zfs_dbgmsg_kstat->ks_ndata = UINT32_MAX;
	zfs_dbgmsg_kstat->ks_private = NULL;
	zfs_dbgmsg_kstat->ks_update = zfs_dbgmsg_update;
	kstat_set_raw_ops(zfs_dbgmsg_kstat, zfs_dbgmsg_headers,
	zfs_dbgmsg_data, zfs_dbgmsg_addr);
	kstat_install(zfs_dbgmsg_kstat);
	}
	}

	void
	zfs_dbgmsg_fini(void)
	{
	if (zfs_dbgmsg_kstat)
	kstat_delete(zfs_dbgmsg_kstat);
	/*
	* TODO - decide how to make this permanent
	*/
	#ifdef _KERNEL
	mutex_enter(&zfs_dbgmsgs_lock);
	zfs_dbgmsg_purge(0);
	mutex_exit(&zfs_dbgmsgs_lock);
	mutex_destroy(&zfs_dbgmsgs_lock);
	#endif
	}

	void
	__zfs_dbgmsg(char *buf)
	{
	zfs_dbgmsg_t *zdm;
	- int size;
	+ uint_t size;

	DTRACE_PROBE1(zfs__dbgmsg, char *, buf);

	size = sizeof (zfs_dbgmsg_t) + strlen(buf);
	zdm = kmem_zalloc(size, KM_SLEEP);
	zdm->zdm_size = size;
	zdm->zdm_timestamp = gethrestime_sec();
	strcpy(zdm->zdm_msg, buf);

	mutex_enter(&zfs_dbgmsgs_lock);
	list_insert_tail(&zfs_dbgmsgs, zdm);
	zfs_dbgmsg_size += size;
	- zfs_dbgmsg_purge(MAX(zfs_dbgmsg_maxsize, 0));
	+ zfs_dbgmsg_purge(zfs_dbgmsg_maxsize);
	mutex_exit(&zfs_dbgmsgs_lock);
	}

	void
	__set_error(const char file, const char func, int line, int err)
	{
	/*
	* To enable this:
	*
	* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
	*/
	if (zfs_flags & ZFS_DEBUG_SET_ERROR)
	__dprintf(B_FALSE, file, func, line, "error %lu", (ulong_t)err);
	}

	#ifdef _KERNEL
	void
	__dprintf(boolean_t dprint, const char file, const char func,
	int line, const char *fmt, ...)
	{
	const char *newfile;
	va_list adx;
	size_t size;
	char *buf;
	char *nl;
	int i;

	size = 1024;
	buf = kmem_alloc(size, KM_SLEEP);

	/*
	* Get rid of annoying prefix to filename.
	*/
	newfile = strrchr(file, '/');
	if (newfile != NULL) {
	newfile = newfile + 1; /* Get rid of leading / */
	} else {
	newfile = file;
	}

	i = snprintf(buf, size, "%s:%d:%s(): ", newfile, line, func);

	if (i < size) {
	va_start(adx, fmt);
	(void) vsnprintf(buf + i, size - i, fmt, adx);
	va_end(adx);
	}

	/*
	* Get rid of trailing newline.
	*/
	nl = strrchr(buf, '\n');
	if (nl != NULL)
	*nl = '\0';

	__zfs_dbgmsg(buf);

	kmem_free(buf, size);
	}

	#else

	void
	zfs_dbgmsg_print(const char *tag)
	{
	zfs_dbgmsg_t *zdm;

	(void) printf("ZFS_DBGMSG(%s):\n", tag);
	mutex_enter(&zfs_dbgmsgs_lock);
	for (zdm = list_head(&zfs_dbgmsgs); zdm;
	zdm = list_next(&zfs_dbgmsgs, zdm))
	(void) printf("%s\n", zdm->zdm_msg);
	mutex_exit(&zfs_dbgmsgs_lock);
	}
	#endif /* _KERNEL */

	ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_enable, INT, ZMOD_RW,
	"Enable ZFS debug message log");

	-ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_maxsize, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_maxsize, UINT, ZMOD_RW,
	"Maximum ZFS debug log size");
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c
	index b290c36748ca..c65be4c134d5 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vfsops.c
	@@ -1,2318 +1,2317 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011 Pawel Jakub Dawidek <pawel@dawidek.net>.
	* All rights reserved.
	* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
	*/

	/* Portions Copyright 2010 Robert Milkowski */

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/kernel.h>
	#include <sys/sysmacros.h>
	#include <sys/kmem.h>
	#include <sys/acl.h>
	#include <sys/vnode.h>
	#include <sys/vfs.h>
	#include <sys/mntent.h>
	#include <sys/mount.h>
	#include <sys/cmn_err.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_vnops.h>
	#include <sys/zfs_dir.h>
	#include <sys/zil.h>
	#include <sys/fs/zfs.h>
	#include <sys/dmu.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_deleg.h>
	#include <sys/spa.h>
	#include <sys/zap.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/policy.h>
	#include <sys/atomic.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_fuid.h>
	#include <sys/sunddi.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dir.h>
	#include <sys/jail.h>
	#include <ufs/ufs/quota.h>
	#include <sys/zfs_quota.h>

	#include "zfs_comutil.h"

	#ifndef MNTK_VMSETSIZE_BUG
	#define MNTK_VMSETSIZE_BUG 0
	#endif
	#ifndef MNTK_NOMSYNC
	#define MNTK_NOMSYNC 8
	#endif

	struct mtx zfs_debug_mtx;
	MTX_SYSINIT(zfs_debug_mtx, &zfs_debug_mtx, "zfs_debug", MTX_DEF);

	SYSCTL_NODE(_vfs, OID_AUTO, zfs, CTLFLAG_RW, 0, "ZFS file system");

	int zfs_super_owner;
	SYSCTL_INT(_vfs_zfs, OID_AUTO, super_owner, CTLFLAG_RW, &zfs_super_owner, 0,
	"File system owners can perform privileged operation on file systems");

	int zfs_debug_level;
	SYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RWTUN, &zfs_debug_level, 0,
	"Debug level");

	SYSCTL_NODE(_vfs_zfs, OID_AUTO, version, CTLFLAG_RD, 0, "ZFS versions");
	static int zfs_version_acl = ZFS_ACL_VERSION;
	SYSCTL_INT(_vfs_zfs_version, OID_AUTO, acl, CTLFLAG_RD, &zfs_version_acl, 0,
	"ZFS_ACL_VERSION");
	static int zfs_version_spa = SPA_VERSION;
	SYSCTL_INT(_vfs_zfs_version, OID_AUTO, spa, CTLFLAG_RD, &zfs_version_spa, 0,
	"SPA_VERSION");
	static int zfs_version_zpl = ZPL_VERSION;
	SYSCTL_INT(_vfs_zfs_version, OID_AUTO, zpl, CTLFLAG_RD, &zfs_version_zpl, 0,
	"ZPL_VERSION");

	#if __FreeBSD_version >= 1400018
	static int zfs_quotactl(vfs_t vfsp, int cmds, uid_t id, void arg,
	bool *mp_busy);
	#else
	static int zfs_quotactl(vfs_t vfsp, int cmds, uid_t id, void arg);
	#endif
	static int zfs_mount(vfs_t *vfsp);
	static int zfs_umount(vfs_t *vfsp, int fflag);
	static int zfs_root(vfs_t vfsp, int flags, vnode_t *vpp);
	static int zfs_statfs(vfs_t vfsp, struct statfs statp);
	static int zfs_vget(vfs_t vfsp, ino_t ino, int flags, vnode_t *vpp);
	static int zfs_sync(vfs_t *vfsp, int waitfor);
	#if __FreeBSD_version >= 1300098
	static int zfs_checkexp(vfs_t vfsp, struct sockaddr nam, uint64_t *extflagsp,
	struct ucred *credanonp, int numsecflavors, int *secflavors);
	#else
	static int zfs_checkexp(vfs_t vfsp, struct sockaddr nam, int *extflagsp,
	struct ucred *credanonp, int numsecflavors, int **secflavors);
	#endif
	static int zfs_fhtovp(vfs_t vfsp, fid_t fidp, int flags, vnode_t **vpp);
	static void zfs_freevfs(vfs_t *vfsp);

	struct vfsops zfs_vfsops = {
	.vfs_mount = zfs_mount,
	.vfs_unmount = zfs_umount,
	#if __FreeBSD_version >= 1300049
	.vfs_root = vfs_cache_root,
	.vfs_cachedroot = zfs_root,
	#else
	.vfs_root = zfs_root,
	#endif
	.vfs_statfs = zfs_statfs,
	.vfs_vget = zfs_vget,
	.vfs_sync = zfs_sync,
	.vfs_checkexp = zfs_checkexp,
	.vfs_fhtovp = zfs_fhtovp,
	.vfs_quotactl = zfs_quotactl,
	};

	VFS_SET(zfs_vfsops, zfs, VFCF_JAIL \| VFCF_DELEGADMIN);

	/*
	* We need to keep a count of active fs's.
	* This is necessary to prevent our module
	* from being unloaded after a umount -f
	*/
	static uint32_t zfs_active_fs_count = 0;

	int
	zfs_get_temporary_prop(dsl_dataset_t ds, zfs_prop_t zfs_prop, uint64_t val,
	char *setpoint)
	{
	int error;
	zfsvfs_t *zfvp;
	vfs_t *vfsp;
	objset_t *os;
	uint64_t tmp = *val;

	error = dmu_objset_from_ds(ds, &os);
	if (error != 0)
	return (error);

	error = getzfsvfs_impl(os, &zfvp);
	if (error != 0)
	return (error);
	if (zfvp == NULL)
	return (ENOENT);
	vfsp = zfvp->z_vfs;
	switch (zfs_prop) {
	case ZFS_PROP_ATIME:
	if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL))
	tmp = 1;
	break;
	case ZFS_PROP_DEVICES:
	if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL))
	tmp = 1;
	break;
	case ZFS_PROP_EXEC:
	if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL))
	tmp = 1;
	break;
	case ZFS_PROP_SETUID:
	if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL))
	tmp = 1;
	break;
	case ZFS_PROP_READONLY:
	if (vfs_optionisset(vfsp, MNTOPT_RW, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_RO, NULL))
	tmp = 1;
	break;
	case ZFS_PROP_XATTR:
	if (zfvp->z_flags & ZSB_XATTR)
	tmp = zfvp->z_xattr;
	break;
	case ZFS_PROP_NBMAND:
	if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL))
	tmp = 0;
	if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL))
	tmp = 1;
	break;
	default:
	vfs_unbusy(vfsp);
	return (ENOENT);
	}

	vfs_unbusy(vfsp);
	if (tmp != *val) {
	(void) strcpy(setpoint, "temporary");
	*val = tmp;
	}
	return (0);
	}

	static int
	zfs_getquota(zfsvfs_t zfsvfs, uid_t id, int isgroup, struct dqblk64 dqp)
	{
	int error = 0;
	char buf[32];
	uint64_t usedobj, quotaobj;
	uint64_t quota, used = 0;
	timespec_t now;

	usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
	quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj;

	if (quotaobj == 0 \|\| zfsvfs->z_replay) {
	error = ENOENT;
	goto done;
	}
	(void) sprintf(buf, "%llx", (longlong_t)id);
	if ((error = zap_lookup(zfsvfs->z_os, quotaobj,
	buf, sizeof (quota), 1, &quota)) != 0) {
	dprintf("%s(%d): quotaobj lookup failed\n",
	__FUNCTION__, __LINE__);
	goto done;
	}
	/*
	* quota(8) uses bsoftlimit as "quoota", and hardlimit as "limit".
	* So we set them to be the same.
	*/
	dqp->dqb_bsoftlimit = dqp->dqb_bhardlimit = btodb(quota);
	error = zap_lookup(zfsvfs->z_os, usedobj, buf, sizeof (used), 1, &used);
	if (error && error != ENOENT) {
	dprintf("%s(%d): usedobj failed; %d\n",
	__FUNCTION__, __LINE__, error);
	goto done;
	}
	dqp->dqb_curblocks = btodb(used);
	dqp->dqb_ihardlimit = dqp->dqb_isoftlimit = 0;
	vfs_timestamp(&now);
	/*
	* Setting this to 0 causes FreeBSD quota(8) to print
	* the number of days since the epoch, which isn't
	* particularly useful.
	*/
	dqp->dqb_btime = dqp->dqb_itime = now.tv_sec;
	done:
	return (error);
	}

	static int
	#if __FreeBSD_version >= 1400018
	zfs_quotactl(vfs_t vfsp, int cmds, uid_t id, void arg, bool *mp_busy)
	#else
	zfs_quotactl(vfs_t vfsp, int cmds, uid_t id, void arg)
	#endif
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	struct thread *td;
	int cmd, type, error = 0;
	int bitsize;
	zfs_userquota_prop_t quota_type;
	struct dqblk64 dqblk = { 0 };

	td = curthread;
	cmd = cmds >> SUBCMDSHIFT;
	type = cmds & SUBCMDMASK;

	if ((error = zfs_enter(zfsvfs, FTAG)) != 0)
	return (error);
	if (id == -1) {
	switch (type) {
	case USRQUOTA:
	id = td->td_ucred->cr_ruid;
	break;
	case GRPQUOTA:
	id = td->td_ucred->cr_rgid;
	break;
	default:
	error = EINVAL;
	#if __FreeBSD_version < 1400018
	if (cmd == Q_QUOTAON \|\| cmd == Q_QUOTAOFF)
	vfs_unbusy(vfsp);
	#endif
	goto done;
	}
	}
	/*
	* Map BSD type to:
	* ZFS_PROP_USERUSED,
	* ZFS_PROP_USERQUOTA,
	* ZFS_PROP_GROUPUSED,
	* ZFS_PROP_GROUPQUOTA
	*/
	switch (cmd) {
	case Q_SETQUOTA:
	case Q_SETQUOTA32:
	if (type == USRQUOTA)
	quota_type = ZFS_PROP_USERQUOTA;
	else if (type == GRPQUOTA)
	quota_type = ZFS_PROP_GROUPQUOTA;
	else
	error = EINVAL;
	break;
	case Q_GETQUOTA:
	case Q_GETQUOTA32:
	if (type == USRQUOTA)
	quota_type = ZFS_PROP_USERUSED;
	else if (type == GRPQUOTA)
	quota_type = ZFS_PROP_GROUPUSED;
	else
	error = EINVAL;
	break;
	}

	/*
	* Depending on the cmd, we may need to get
	* the ruid and domain (see fuidstr_to_sid?),
	* the fuid (how?), or other information.
	* Create fuid using zfs_fuid_create(zfsvfs, id,
	* ZFS_OWNER or ZFS_GROUP, cr, &fuidp)?
	* I think I can use just the id?
	*
	* Look at zfs_id_overquota() to look up a quota.
	* zap_lookup(something, quotaobj, fuidstring,
	* sizeof (long long), 1, &quota)
	*
	* See zfs_set_userquota() to set a quota.
	*/
	if ((uint32_t)type >= MAXQUOTAS) {
	error = EINVAL;
	goto done;
	}

	switch (cmd) {
	case Q_GETQUOTASIZE:
	bitsize = 64;
	error = copyout(&bitsize, arg, sizeof (int));
	break;
	case Q_QUOTAON:
	// As far as I can tell, you can't turn quotas on or off on zfs
	error = 0;
	#if __FreeBSD_version < 1400018
	vfs_unbusy(vfsp);
	#endif
	break;
	case Q_QUOTAOFF:
	error = ENOTSUP;
	#if __FreeBSD_version < 1400018
	vfs_unbusy(vfsp);
	#endif
	break;
	case Q_SETQUOTA:
	error = copyin(arg, &dqblk, sizeof (dqblk));
	if (error == 0)
	error = zfs_set_userquota(zfsvfs, quota_type,
	"", id, dbtob(dqblk.dqb_bhardlimit));
	break;
	case Q_GETQUOTA:
	error = zfs_getquota(zfsvfs, id, type == GRPQUOTA, &dqblk);
	if (error == 0)
	error = copyout(&dqblk, arg, sizeof (dqblk));
	break;
	default:
	error = EINVAL;
	break;
	}
	done:
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}


	boolean_t
	zfs_is_readonly(zfsvfs_t *zfsvfs)
	{
	return (!!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY));
	}

	static int
	zfs_sync(vfs_t *vfsp, int waitfor)
	{

	/*
	* Data integrity is job one. We don't want a compromised kernel
	* writing to the storage pool, so we never sync during panic.
	*/
	if (panicstr)
	return (0);

	/*
	* Ignore the system syncher. ZFS already commits async data
	* at zfs_txg_timeout intervals.
	*/
	if (waitfor == MNT_LAZY)
	return (0);

	if (vfsp != NULL) {
	/*
	* Sync a specific filesystem.
	*/
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	dsl_pool_t *dp;
	int error;

	error = vfs_stdsync(vfsp, waitfor);
	if (error != 0)
	return (error);

	if ((error = zfs_enter(zfsvfs, FTAG)) != 0)
	return (error);
	dp = dmu_objset_pool(zfsvfs->z_os);

	/*
	* If the system is shutting down, then skip any
	* filesystems which may exist on a suspended pool.
	*/
	if (rebooting && spa_suspended(dp->dp_spa)) {
	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	if (zfsvfs->z_log != NULL)
	zil_commit(zfsvfs->z_log, 0);

	zfs_exit(zfsvfs, FTAG);
	} else {
	/*
	* Sync all ZFS filesystems. This is what happens when you
	* run sync(8). Unlike other filesystems, ZFS honors the
	* request by waiting for all pools to commit all dirty data.
	*/
	spa_sync_allpools();
	}

	return (0);
	}

	static void
	atime_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	if (newval == TRUE) {
	zfsvfs->z_atime = TRUE;
	zfsvfs->z_vfs->vfs_flag &= ~MNT_NOATIME;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0);
	} else {
	zfsvfs->z_atime = FALSE;
	zfsvfs->z_vfs->vfs_flag \|= MNT_NOATIME;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0);
	}
	}

	static void
	xattr_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	if (newval == ZFS_XATTR_OFF) {
	zfsvfs->z_flags &= ~ZSB_XATTR;
	} else {
	zfsvfs->z_flags \|= ZSB_XATTR;

	if (newval == ZFS_XATTR_SA)
	zfsvfs->z_xattr_sa = B_TRUE;
	else
	zfsvfs->z_xattr_sa = B_FALSE;
	}
	}

	static void
	blksz_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;
	ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os)));
	ASSERT3U(newval, >=, SPA_MINBLOCKSIZE);
	ASSERT(ISP2(newval));

	zfsvfs->z_max_blksz = newval;
	zfsvfs->z_vfs->mnt_stat.f_iosize = newval;
	}

	static void
	readonly_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	if (newval) {
	/* XXX locking on vfs_flag? */
	zfsvfs->z_vfs->vfs_flag \|= VFS_RDONLY;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0);
	} else {
	/* XXX locking on vfs_flag? */
	zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0);
	}
	}

	static void
	setuid_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	if (newval == FALSE) {
	zfsvfs->z_vfs->vfs_flag \|= VFS_NOSETUID;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0);
	} else {
	zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0);
	}
	}

	static void
	exec_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	if (newval == FALSE) {
	zfsvfs->z_vfs->vfs_flag \|= VFS_NOEXEC;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0);
	} else {
	zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC;
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0);
	}
	}

	/*
	* The nbmand mount option can be changed at mount time.
	* We can't allow it to be toggled on live file systems or incorrect
	* behavior may be seen from cifs clients
	*
	* This property isn't registered via dsl_prop_register(), but this callback
	* will be called when a file system is first mounted
	*/
	static void
	nbmand_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;
	if (newval == FALSE) {
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0);
	} else {
	vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND);
	vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0);
	}
	}

	static void
	snapdir_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	zfsvfs->z_show_ctldir = newval;
	}

	static void
	acl_mode_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	zfsvfs->z_acl_mode = newval;
	}

	static void
	acl_inherit_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	zfsvfs->z_acl_inherit = newval;
	}

	static void
	acl_type_changed_cb(void *arg, uint64_t newval)
	{
	zfsvfs_t *zfsvfs = arg;

	zfsvfs->z_acl_type = newval;
	}

	static int
	zfs_register_callbacks(vfs_t *vfsp)
	{
	struct dsl_dataset *ds = NULL;
	objset_t *os = NULL;
	zfsvfs_t *zfsvfs = NULL;
	uint64_t nbmand;
	boolean_t readonly = B_FALSE;
	boolean_t do_readonly = B_FALSE;
	boolean_t setuid = B_FALSE;
	boolean_t do_setuid = B_FALSE;
	boolean_t exec = B_FALSE;
	boolean_t do_exec = B_FALSE;
	boolean_t xattr = B_FALSE;
	boolean_t atime = B_FALSE;
	boolean_t do_atime = B_FALSE;
	boolean_t do_xattr = B_FALSE;
	int error = 0;

	ASSERT3P(vfsp, !=, NULL);
	zfsvfs = vfsp->vfs_data;
	ASSERT3P(zfsvfs, !=, NULL);
	os = zfsvfs->z_os;

	/*
	* This function can be called for a snapshot when we update snapshot's
	* mount point, which isn't really supported.
	*/
	if (dmu_objset_is_snapshot(os))
	return (EOPNOTSUPP);

	/*
	* The act of registering our callbacks will destroy any mount
	* options we may have. In order to enable temporary overrides
	* of mount options, we stash away the current values and
	* restore them after we register the callbacks.
	*/
	if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) \|\|
	!spa_writeable(dmu_objset_spa(os))) {
	readonly = B_TRUE;
	do_readonly = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
	readonly = B_FALSE;
	do_readonly = B_TRUE;
	}
	if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
	setuid = B_FALSE;
	do_setuid = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
	setuid = B_TRUE;
	do_setuid = B_TRUE;
	}
	if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
	exec = B_FALSE;
	do_exec = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
	exec = B_TRUE;
	do_exec = B_TRUE;
	}
	if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
	zfsvfs->z_xattr = xattr = ZFS_XATTR_OFF;
	do_xattr = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
	zfsvfs->z_xattr = xattr = ZFS_XATTR_DIR;
	do_xattr = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_DIRXATTR, NULL)) {
	zfsvfs->z_xattr = xattr = ZFS_XATTR_DIR;
	do_xattr = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_SAXATTR, NULL)) {
	zfsvfs->z_xattr = xattr = ZFS_XATTR_SA;
	do_xattr = B_TRUE;
	}
	if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) {
	atime = B_FALSE;
	do_atime = B_TRUE;
	} else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) {
	atime = B_TRUE;
	do_atime = B_TRUE;
	}

	/*
	* We need to enter pool configuration here, so that we can use
	* dsl_prop_get_int_ds() to handle the special nbmand property below.
	* dsl_prop_get_integer() can not be used, because it has to acquire
	* spa_namespace_lock and we can not do that because we already hold
	* z_teardown_lock. The problem is that spa_write_cachefile() is called
	* with spa_namespace_lock held and the function calls ZFS vnode
	* operations to write the cache file and thus z_teardown_lock is
	* acquired after spa_namespace_lock.
	*/
	ds = dmu_objset_ds(os);
	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);

	/*
	* nbmand is a special property. It can only be changed at
	* mount time.
	*
	* This is weird, but it is documented to only be changeable
	* at mount time.
	*/
	if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
	nbmand = B_FALSE;
	} else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) {
	nbmand = B_TRUE;
	} else if ((error = dsl_prop_get_int_ds(ds, "nbmand", &nbmand) != 0)) {
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	return (error);
	}

	/*
	* Register property callbacks.
	*
	* It would probably be fine to just check for i/o error from
	* the first prop_register(), but I guess I like to go
	* overboard...
	*/
	error = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_ACLTYPE), acl_type_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_ACLMODE), acl_mode_changed_cb, zfsvfs);
	error = error ? error : dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb,
	zfsvfs);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	if (error)
	goto unregister;

	/*
	* Invoke our callbacks to restore temporary mount options.
	*/
	if (do_readonly)
	readonly_changed_cb(zfsvfs, readonly);
	if (do_setuid)
	setuid_changed_cb(zfsvfs, setuid);
	if (do_exec)
	exec_changed_cb(zfsvfs, exec);
	if (do_xattr)
	xattr_changed_cb(zfsvfs, xattr);
	if (do_atime)
	atime_changed_cb(zfsvfs, atime);

	nbmand_changed_cb(zfsvfs, nbmand);

	return (0);

	unregister:
	dsl_prop_unregister_all(ds, zfsvfs);
	return (error);
	}

	/*
	* Associate this zfsvfs with the given objset, which must be owned.
	* This will cache a bunch of on-disk state from the objset in the
	* zfsvfs.
	*/
	static int
	zfsvfs_init(zfsvfs_t zfsvfs, objset_t os)
	{
	int error;
	uint64_t val;

	zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
	zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
	zfsvfs->z_os = os;

	error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
	if (error != 0)
	return (error);
	if (zfsvfs->z_version >
	zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
	(void) printf("Can't mount a version %lld file system "
	"on a version %lld pool\n. Pool must be upgraded to mount "
	"this file system.", (u_longlong_t)zfsvfs->z_version,
	(u_longlong_t)spa_version(dmu_objset_spa(os)));
	return (SET_ERROR(ENOTSUP));
	}
	error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val);
	if (error != 0)
	return (error);
	zfsvfs->z_norm = (int)val;

	error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val);
	if (error != 0)
	return (error);
	zfsvfs->z_utf8 = (val != 0);

	error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val);
	if (error != 0)
	return (error);
	zfsvfs->z_case = (uint_t)val;

	error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val);
	if (error != 0)
	return (error);
	zfsvfs->z_acl_type = (uint_t)val;

	/*
	* Fold case on file systems that are always or sometimes case
	* insensitive.
	*/
	if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE \|\|
	zfsvfs->z_case == ZFS_CASE_MIXED)
	zfsvfs->z_norm \|= U8_TEXTPREP_TOUPPER;

	zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
	zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os);

	uint64_t sa_obj = 0;
	if (zfsvfs->z_use_sa) {
	/* should either have both of these objects or none */
	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
	&sa_obj);
	if (error != 0)
	return (error);

	error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val);
	if (error == 0 && val == ZFS_XATTR_SA)
	zfsvfs->z_xattr_sa = B_TRUE;
	}

	error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
	&zfsvfs->z_attr_table);
	if (error != 0)
	return (error);

	if (zfsvfs->z_version >= ZPL_VERSION_SA)
	sa_register_update_callback(os, zfs_sa_upgrade);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
	&zfsvfs->z_root);
	if (error != 0)
	return (error);
	ASSERT3U(zfsvfs->z_root, !=, 0);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
	&zfsvfs->z_unlinkedobj);
	if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
	8, 1, &zfsvfs->z_userquota_obj);
	if (error == ENOENT)
	zfsvfs->z_userquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
	8, 1, &zfsvfs->z_groupquota_obj);
	if (error == ENOENT)
	zfsvfs->z_groupquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA],
	8, 1, &zfsvfs->z_projectquota_obj);
	if (error == ENOENT)
	zfsvfs->z_projectquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA],
	8, 1, &zfsvfs->z_userobjquota_obj);
	if (error == ENOENT)
	zfsvfs->z_userobjquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA],
	8, 1, &zfsvfs->z_groupobjquota_obj);
	if (error == ENOENT)
	zfsvfs->z_groupobjquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ,
	zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA],
	8, 1, &zfsvfs->z_projectobjquota_obj);
	if (error == ENOENT)
	zfsvfs->z_projectobjquota_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
	&zfsvfs->z_fuid_obj);
	if (error == ENOENT)
	zfsvfs->z_fuid_obj = 0;
	else if (error != 0)
	return (error);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
	&zfsvfs->z_shares_dir);
	if (error == ENOENT)
	zfsvfs->z_shares_dir = 0;
	else if (error != 0)
	return (error);

	/*
	* Only use the name cache if we are looking for a
	* name on a file system that does not require normalization
	* or case folding. We can also look there if we happen to be
	* on a non-normalizing, mixed sensitivity file system IF we
	* are looking for the exact name (which is always the case on
	* FreeBSD).
	*/
	zfsvfs->z_use_namecache = !zfsvfs->z_norm \|\|
	((zfsvfs->z_case == ZFS_CASE_MIXED) &&
	!(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER));

	return (0);
	}

	taskq_t *zfsvfs_taskq;

	static void
	zfsvfs_task_unlinked_drain(void *context, int pending __unused)
	{

	zfs_unlinked_drain((zfsvfs_t *)context);
	}

	int
	zfsvfs_create(const char osname, boolean_t readonly, zfsvfs_t *zfvp)
	{
	objset_t *os;
	zfsvfs_t *zfsvfs;
	int error;
	boolean_t ro = (readonly \|\| (strchr(osname, '@') != NULL));

	/*
	* XXX: Fix struct statfs so this isn't necessary!
	*
	* The 'osname' is used as the filesystem's special node, which means
	* it must fit in statfs.f_mntfromname, or else it can't be
	* enumerated, so libzfs_mnttab_find() returns NULL, which causes
	* 'zfs unmount' to think it's not mounted when it is.
	*/
	if (strlen(osname) >= MNAMELEN)
	return (SET_ERROR(ENAMETOOLONG));

	zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);

	error = dmu_objset_own(osname, DMU_OST_ZFS, ro, B_TRUE, zfsvfs,
	&os);
	if (error != 0) {
	kmem_free(zfsvfs, sizeof (zfsvfs_t));
	return (error);
	}

	error = zfsvfs_create_impl(zfvp, zfsvfs, os);

	return (error);
	}


	int
	zfsvfs_create_impl(zfsvfs_t *zfvp, zfsvfs_t zfsvfs, objset_t *os)
	{
	int error;

	zfsvfs->z_vfs = NULL;
	zfsvfs->z_parent = zfsvfs;

	mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
	offsetof(znode_t, z_link_node));
	TASK_INIT(&zfsvfs->z_unlinked_drain_task, 0,
	zfsvfs_task_unlinked_drain, zfsvfs);
	ZFS_TEARDOWN_INIT(zfsvfs);
	ZFS_TEARDOWN_INACTIVE_INIT(zfsvfs);
	rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
	for (int i = 0; i != ZFS_OBJ_MTX_SZ; i++)
	mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);

	error = zfsvfs_init(zfsvfs, os);
	if (error != 0) {
	dmu_objset_disown(os, B_TRUE, zfsvfs);
	*zfvp = NULL;
	kmem_free(zfsvfs, sizeof (zfsvfs_t));
	return (error);
	}

	*zfvp = zfsvfs;
	return (0);
	}

	static int
	zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
	{
	int error;

	/*
	* Check for a bad on-disk format version now since we
	* lied about owning the dataset readonly before.
	*/
	if (!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) &&
	dmu_objset_incompatible_encryption_version(zfsvfs->z_os))
	return (SET_ERROR(EROFS));

	error = zfs_register_callbacks(zfsvfs->z_vfs);
	if (error)
	return (error);

	/*
	* If we are not mounting (ie: online recv), then we don't
	* have to worry about replaying the log as we blocked all
	* operations out since we closed the ZIL.
	*/
	if (mounting) {
	boolean_t readonly;

	ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL);
	error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os);
	if (error)
	return (error);
	zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data,
	&zfsvfs->z_kstat.dk_zil_sums);

	/*
	* During replay we remove the read only flag to
	* allow replays to succeed.
	*/
	readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY;
	if (readonly != 0) {
	zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
	} else {
	dsl_dir_t *dd;
	zap_stats_t zs;

	if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj,
	&zs) == 0) {
	dataset_kstats_update_nunlinks_kstat(
	&zfsvfs->z_kstat, zs.zs_num_entries);
	dprintf_ds(zfsvfs->z_os->os_dsl_dataset,
	"num_entries in unlinked set: %llu",
	(u_longlong_t)zs.zs_num_entries);
	}

	zfs_unlinked_drain(zfsvfs);
	dd = zfsvfs->z_os->os_dsl_dataset->ds_dir;
	dd->dd_activity_cancelled = B_FALSE;
	}

	/*
	* Parse and replay the intent log.
	*
	* Because of ziltest, this must be done after
	* zfs_unlinked_drain(). (Further note: ziltest
	* doesn't use readonly mounts, where
	* zfs_unlinked_drain() isn't called.) This is because
	* ziltest causes spa_sync() to think it's committed,
	* but actually it is not, so the intent log contains
	* many txg's worth of changes.
	*
	* In particular, if object N is in the unlinked set in
	* the last txg to actually sync, then it could be
	* actually freed in a later txg and then reallocated
	* in a yet later txg. This would write a "create
	* object N" record to the intent log. Normally, this
	* would be fine because the spa_sync() would have
	* written out the fact that object N is free, before
	* we could write the "create object N" intent log
	* record.
	*
	* But when we are in ziltest mode, we advance the "open
	* txg" without actually spa_sync()-ing the changes to
	* disk. So we would see that object N is still
	* allocated and in the unlinked set, and there is an
	* intent log record saying to allocate it.
	*/
	if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) {
	if (zil_replay_disable) {
	zil_destroy(zfsvfs->z_log, B_FALSE);
	} else {
	boolean_t use_nc = zfsvfs->z_use_namecache;
	zfsvfs->z_use_namecache = B_FALSE;
	zfsvfs->z_replay = B_TRUE;
	zil_replay(zfsvfs->z_os, zfsvfs,
	zfs_replay_vector);
	zfsvfs->z_replay = B_FALSE;
	zfsvfs->z_use_namecache = use_nc;
	}
	}

	/* restore readonly bit */
	if (readonly != 0)
	zfsvfs->z_vfs->vfs_flag \|= VFS_RDONLY;
	} else {
	ASSERT3P(zfsvfs->z_kstat.dk_kstats, !=, NULL);
	zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data,
	&zfsvfs->z_kstat.dk_zil_sums);
	}

	/*
	* Set the objset user_ptr to track its zfsvfs.
	*/
	mutex_enter(&zfsvfs->z_os->os_user_ptr_lock);
	dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
	mutex_exit(&zfsvfs->z_os->os_user_ptr_lock);

	return (0);
	}

	void
	zfsvfs_free(zfsvfs_t *zfsvfs)
	{
	int i;

	zfs_fuid_destroy(zfsvfs);

	mutex_destroy(&zfsvfs->z_znodes_lock);
	mutex_destroy(&zfsvfs->z_lock);
	ASSERT3U(zfsvfs->z_nr_znodes, ==, 0);
	list_destroy(&zfsvfs->z_all_znodes);
	ZFS_TEARDOWN_DESTROY(zfsvfs);
	ZFS_TEARDOWN_INACTIVE_DESTROY(zfsvfs);
	rw_destroy(&zfsvfs->z_fuid_lock);
	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
	mutex_destroy(&zfsvfs->z_hold_mtx[i]);
	dataset_kstats_destroy(&zfsvfs->z_kstat);
	kmem_free(zfsvfs, sizeof (zfsvfs_t));
	}

	static void
	zfs_set_fuid_feature(zfsvfs_t *zfsvfs)
	{
	zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
	zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os);
	}

	static int
	zfs_domount(vfs_t vfsp, char osname)
	{
	uint64_t recordsize, fsid_guid;
	int error = 0;
	zfsvfs_t *zfsvfs;

	ASSERT3P(vfsp, !=, NULL);
	ASSERT3P(osname, !=, NULL);

	error = zfsvfs_create(osname, vfsp->mnt_flag & MNT_RDONLY, &zfsvfs);
	if (error)
	return (error);
	zfsvfs->z_vfs = vfsp;

	if ((error = dsl_prop_get_integer(osname,
	"recordsize", &recordsize, NULL)))
	goto out;
	zfsvfs->z_vfs->vfs_bsize = SPA_MINBLOCKSIZE;
	zfsvfs->z_vfs->mnt_stat.f_iosize = recordsize;

	vfsp->vfs_data = zfsvfs;
	vfsp->mnt_flag \|= MNT_LOCAL;
	vfsp->mnt_kern_flag \|= MNTK_LOOKUP_SHARED;
	vfsp->mnt_kern_flag \|= MNTK_SHARED_WRITES;
	vfsp->mnt_kern_flag \|= MNTK_EXTENDED_SHARED;
	/*
	* This can cause a loss of coherence between ARC and page cache
	* on ZoF - unclear if the problem is in FreeBSD or ZoF
	*/
	vfsp->mnt_kern_flag \|= MNTK_NO_IOPF; /* vn_io_fault can be used */
	vfsp->mnt_kern_flag \|= MNTK_NOMSYNC;
	vfsp->mnt_kern_flag \|= MNTK_VMSETSIZE_BUG;

	#if defined(_KERNEL) && !defined(KMEM_DEBUG)
	vfsp->mnt_kern_flag \|= MNTK_FPLOOKUP;
	#endif
	/*
	* The fsid is 64 bits, composed of an 8-bit fs type, which
	* separates our fsid from any other filesystem types, and a
	* 56-bit objset unique ID. The objset unique ID is unique to
	* all objsets open on this system, provided by unique_create().
	* The 8-bit fs type must be put in the low bits of fsid[1]
	* because that's where other Solaris filesystems put it.
	*/
	fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os);
	ASSERT3U((fsid_guid & ~((1ULL << 56) - 1)), ==, 0);
	vfsp->vfs_fsid.val[0] = fsid_guid;
	vfsp->vfs_fsid.val[1] = ((fsid_guid >> 32) << 8) \|
	(vfsp->mnt_vfc->vfc_typenum & 0xFF);

	/*
	* Set features for file system.
	*/
	zfs_set_fuid_feature(zfsvfs);

	if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
	uint64_t pval;

	atime_changed_cb(zfsvfs, B_FALSE);
	readonly_changed_cb(zfsvfs, B_TRUE);
	if ((error = dsl_prop_get_integer(osname,
	"xattr", &pval, NULL)))
	goto out;
	xattr_changed_cb(zfsvfs, pval);
	if ((error = dsl_prop_get_integer(osname,
	"acltype", &pval, NULL)))
	goto out;
	acl_type_changed_cb(zfsvfs, pval);
	zfsvfs->z_issnap = B_TRUE;
	zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED;

	mutex_enter(&zfsvfs->z_os->os_user_ptr_lock);
	dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
	mutex_exit(&zfsvfs->z_os->os_user_ptr_lock);
	} else {
	if ((error = zfsvfs_setup(zfsvfs, B_TRUE)))
	goto out;
	}

	vfs_mountedfrom(vfsp, osname);

	if (!zfsvfs->z_issnap)
	zfsctl_create(zfsvfs);
	out:
	if (error) {
	dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
	zfsvfs_free(zfsvfs);
	} else {
	atomic_inc_32(&zfs_active_fs_count);
	}

	return (error);
	}

	static void
	zfs_unregister_callbacks(zfsvfs_t *zfsvfs)
	{
	objset_t *os = zfsvfs->z_os;

	if (!dmu_objset_is_snapshot(os))
	dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs);
	}

	static int
	getpoolname(const char osname, char poolname)
	{
	char *p;

	p = strchr(osname, '/');
	if (p == NULL) {
	if (strlen(osname) >= MAXNAMELEN)
	return (ENAMETOOLONG);
	(void) strcpy(poolname, osname);
	} else {
	if (p - osname >= MAXNAMELEN)
	return (ENAMETOOLONG);
	- (void) strncpy(poolname, osname, p - osname);
	- poolname[p - osname] = '\0';
	+ (void) strlcpy(poolname, osname, p - osname + 1);
	}
	return (0);
	}

	static void
	fetch_osname_options(char name, bool checkpointrewind)
	{

	if (name[0] == '!') {
	*checkpointrewind = true;
	memmove(name, name + 1, strlen(name));
	} else {
	*checkpointrewind = false;
	}
	}

	static int
	zfs_mount(vfs_t *vfsp)
	{
	kthread_t *td = curthread;
	vnode_t *mvp = vfsp->mnt_vnodecovered;
	cred_t *cr = td->td_ucred;
	char *osname;
	int error = 0;
	int canwrite;
	bool checkpointrewind;

	if (vfs_getopt(vfsp->mnt_optnew, "from", (void **)&osname, NULL))
	return (SET_ERROR(EINVAL));

	/*
	* If full-owner-access is enabled and delegated administration is
	* turned on, we must set nosuid.
	*/
	if (zfs_super_owner &&
	dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != ECANCELED) {
	secpolicy_fs_mount_clearopts(cr, vfsp);
	}

	fetch_osname_options(osname, &checkpointrewind);

	/*
	* Check for mount privilege?
	*
	* If we don't have privilege then see if
	* we have local permission to allow it
	*/
	error = secpolicy_fs_mount(cr, mvp, vfsp);
	if (error) {
	if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != 0)
	goto out;

	if (!(vfsp->vfs_flag & MS_REMOUNT)) {
	vattr_t vattr;

	/*
	* Make sure user is the owner of the mount point
	* or has sufficient privileges.
	*/

	vattr.va_mask = AT_UID;

	vn_lock(mvp, LK_SHARED \| LK_RETRY);
	if (VOP_GETATTR(mvp, &vattr, cr)) {
	VOP_UNLOCK1(mvp);
	goto out;
	}

	if (secpolicy_vnode_owner(mvp, cr, vattr.va_uid) != 0 &&
	VOP_ACCESS(mvp, VWRITE, cr, td) != 0) {
	VOP_UNLOCK1(mvp);
	goto out;
	}
	VOP_UNLOCK1(mvp);
	}

	secpolicy_fs_mount_clearopts(cr, vfsp);
	}

	/*
	* Refuse to mount a filesystem if we are in a local zone and the
	* dataset is not visible.
	*/
	if (!INGLOBALZONE(curproc) &&
	(!zone_dataset_visible(osname, &canwrite) \|\| !canwrite)) {
	error = SET_ERROR(EPERM);
	goto out;
	}

	vfsp->vfs_flag \|= MNT_NFS4ACLS;

	/*
	* When doing a remount, we simply refresh our temporary properties
	* according to those options set in the current VFS options.
	*/
	if (vfsp->vfs_flag & MS_REMOUNT) {
	zfsvfs_t *zfsvfs = vfsp->vfs_data;

	/*
	* Refresh mount options with z_teardown_lock blocking I/O while
	* the filesystem is in an inconsistent state.
	* The lock also serializes this code with filesystem
	* manipulations between entry to zfs_suspend_fs() and return
	* from zfs_resume_fs().
	*/
	ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG);
	zfs_unregister_callbacks(zfsvfs);
	error = zfs_register_callbacks(vfsp);
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);
	goto out;
	}

	/* Initial root mount: try hard to import the requested root pool. */
	if ((vfsp->vfs_flag & MNT_ROOTFS) != 0 &&
	(vfsp->vfs_flag & MNT_UPDATE) == 0) {
	char pname[MAXNAMELEN];

	error = getpoolname(osname, pname);
	if (error == 0)
	error = spa_import_rootpool(pname, checkpointrewind);
	if (error)
	goto out;
	}
	DROP_GIANT();
	error = zfs_domount(vfsp, osname);
	PICKUP_GIANT();

	out:
	return (error);
	}

	static int
	zfs_statfs(vfs_t vfsp, struct statfs statp)
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	uint64_t refdbytes, availbytes, usedobjs, availobjs;
	int error;

	statp->f_version = STATFS_VERSION;

	if ((error = zfs_enter(zfsvfs, FTAG)) != 0)
	return (error);

	dmu_objset_space(zfsvfs->z_os,
	&refdbytes, &availbytes, &usedobjs, &availobjs);

	/*
	* The underlying storage pool actually uses multiple block sizes.
	* We report the fragsize as the smallest block size we support,
	* and we report our blocksize as the filesystem's maximum blocksize.
	*/
	statp->f_bsize = SPA_MINBLOCKSIZE;
	statp->f_iosize = zfsvfs->z_vfs->mnt_stat.f_iosize;

	/*
	* The following report "total" blocks of various kinds in the
	* file system, but reported in terms of f_frsize - the
	* "fragment" size.
	*/

	statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT;
	statp->f_bfree = availbytes / statp->f_bsize;
	statp->f_bavail = statp->f_bfree; /* no root reservation */

	/*
	* statvfs() should really be called statufs(), because it assumes
	* static metadata. ZFS doesn't preallocate files, so the best
	* we can do is report the max that could possibly fit in f_files,
	* and that minus the number actually used in f_ffree.
	* For f_ffree, report the smaller of the number of object available
	* and the number of blocks (each object will take at least a block).
	*/
	statp->f_ffree = MIN(availobjs, statp->f_bfree);
	statp->f_files = statp->f_ffree + usedobjs;

	/*
	* We're a zfs filesystem.
	*/
	strlcpy(statp->f_fstypename, "zfs",
	sizeof (statp->f_fstypename));

	strlcpy(statp->f_mntfromname, vfsp->mnt_stat.f_mntfromname,
	sizeof (statp->f_mntfromname));
	strlcpy(statp->f_mntonname, vfsp->mnt_stat.f_mntonname,
	sizeof (statp->f_mntonname));

	statp->f_namemax = MAXNAMELEN - 1;

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	static int
	zfs_root(vfs_t vfsp, int flags, vnode_t *vpp)
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	znode_t *rootzp;
	int error;

	if ((error = zfs_enter(zfsvfs, FTAG)) != 0)
	return (error);

	error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
	if (error == 0)
	*vpp = ZTOV(rootzp);

	zfs_exit(zfsvfs, FTAG);

	if (error == 0) {
	error = vn_lock(*vpp, flags);
	if (error != 0) {
	VN_RELE(*vpp);
	*vpp = NULL;
	}
	}
	return (error);
	}

	/*
	* Teardown the zfsvfs::z_os.
	*
	* Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock'
	* and 'z_teardown_inactive_lock' held.
	*/
	static int
	zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
	{
	znode_t *zp;
	dsl_dir_t *dd;

	/*
	* If someone has not already unmounted this file system,
	* drain the zrele_taskq to ensure all active references to the
	* zfsvfs_t have been handled only then can it be safely destroyed.
	*/
	if (zfsvfs->z_os) {
	/*
	* If we're unmounting we have to wait for the list to
	* drain completely.
	*
	* If we're not unmounting there's no guarantee the list
	* will drain completely, but zreles run from the taskq
	* may add the parents of dir-based xattrs to the taskq
	* so we want to wait for these.
	*
	* We can safely read z_nr_znodes without locking because the
	* VFS has already blocked operations which add to the
	* z_all_znodes list and thus increment z_nr_znodes.
	*/
	int round = 0;
	while (zfsvfs->z_nr_znodes > 0) {
	taskq_wait_outstanding(dsl_pool_zrele_taskq(
	dmu_objset_pool(zfsvfs->z_os)), 0);
	if (++round > 1 && !unmounting)
	break;
	}
	}
	ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG);

	if (!unmounting) {
	/*
	* We purge the parent filesystem's vfsp as the parent
	* filesystem and all of its snapshots have their vnode's
	* v_vfsp set to the parent's filesystem's vfsp. Note,
	* 'z_parent' is self referential for non-snapshots.
	*/
	#ifdef FREEBSD_NAMECACHE
	#if __FreeBSD_version >= 1300117
	cache_purgevfs(zfsvfs->z_parent->z_vfs);
	#else
	cache_purgevfs(zfsvfs->z_parent->z_vfs, true);
	#endif
	#endif
	}

	/*
	* Close the zil. NB: Can't close the zil while zfs_inactive
	* threads are blocked as zil_close can call zfs_inactive.
	*/
	if (zfsvfs->z_log) {
	zil_close(zfsvfs->z_log);
	zfsvfs->z_log = NULL;
	}

	ZFS_TEARDOWN_INACTIVE_ENTER_WRITE(zfsvfs);

	/*
	* If we are not unmounting (ie: online recv) and someone already
	* unmounted this file system while we were doing the switcheroo,
	* or a reopen of z_os failed then just bail out now.
	*/
	if (!unmounting && (zfsvfs->z_unmounted \|\| zfsvfs->z_os == NULL)) {
	ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs);
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);
	return (SET_ERROR(EIO));
	}

	/*
	* At this point there are no vops active, and any new vops will
	* fail with EIO since we have z_teardown_lock for writer (only
	* relevant for forced unmount).
	*
	* Release all holds on dbufs.
	*/
	mutex_enter(&zfsvfs->z_znodes_lock);
	for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL;
	zp = list_next(&zfsvfs->z_all_znodes, zp)) {
	if (zp->z_sa_hdl != NULL) {
	zfs_znode_dmu_fini(zp);
	}
	}
	mutex_exit(&zfsvfs->z_znodes_lock);

	/*
	* If we are unmounting, set the unmounted flag and let new vops
	* unblock. zfs_inactive will have the unmounted behavior, and all
	* other vops will fail with EIO.
	*/
	if (unmounting) {
	zfsvfs->z_unmounted = B_TRUE;
	ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs);
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);
	}

	/*
	* z_os will be NULL if there was an error in attempting to reopen
	* zfsvfs, so just return as the properties had already been
	* unregistered and cached data had been evicted before.
	*/
	if (zfsvfs->z_os == NULL)
	return (0);

	/*
	* Unregister properties.
	*/
	zfs_unregister_callbacks(zfsvfs);

	/*
	* Evict cached data
	*/
	if (!zfs_is_readonly(zfsvfs))
	txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
	dmu_objset_evict_dbufs(zfsvfs->z_os);
	dd = zfsvfs->z_os->os_dsl_dataset->ds_dir;
	dsl_dir_cancel_waiters(dd);

	return (0);
	}

	static int
	zfs_umount(vfs_t *vfsp, int fflag)
	{
	kthread_t *td = curthread;
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	objset_t *os;
	cred_t *cr = td->td_ucred;
	int ret;

	ret = secpolicy_fs_unmount(cr, vfsp);
	if (ret) {
	if (dsl_deleg_access((char *)vfsp->vfs_resource,
	ZFS_DELEG_PERM_MOUNT, cr))
	return (ret);
	}

	/*
	* Unmount any snapshots mounted under .zfs before unmounting the
	* dataset itself.
	*/
	if (zfsvfs->z_ctldir != NULL) {
	if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0)
	return (ret);
	}

	if (fflag & MS_FORCE) {
	/*
	* Mark file system as unmounted before calling
	* vflush(FORCECLOSE). This way we ensure no future vnops
	* will be called and risk operating on DOOMED vnodes.
	*/
	ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG);
	zfsvfs->z_unmounted = B_TRUE;
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);
	}

	/*
	* Flush all the files.
	*/
	ret = vflush(vfsp, 0, (fflag & MS_FORCE) ? FORCECLOSE : 0, td);
	if (ret != 0)
	return (ret);
	while (taskqueue_cancel(zfsvfs_taskq->tq_queue,
	&zfsvfs->z_unlinked_drain_task, NULL) != 0)
	taskqueue_drain(zfsvfs_taskq->tq_queue,
	&zfsvfs->z_unlinked_drain_task);

	VERIFY0(zfsvfs_teardown(zfsvfs, B_TRUE));
	os = zfsvfs->z_os;

	/*
	* z_os will be NULL if there was an error in
	* attempting to reopen zfsvfs.
	*/
	if (os != NULL) {
	/*
	* Unset the objset user_ptr.
	*/
	mutex_enter(&os->os_user_ptr_lock);
	dmu_objset_set_user(os, NULL);
	mutex_exit(&os->os_user_ptr_lock);

	/*
	* Finally release the objset
	*/
	dmu_objset_disown(os, B_TRUE, zfsvfs);
	}

	/*
	* We can now safely destroy the '.zfs' directory node.
	*/
	if (zfsvfs->z_ctldir != NULL)
	zfsctl_destroy(zfsvfs);
	zfs_freevfs(vfsp);

	return (0);
	}

	static int
	zfs_vget(vfs_t vfsp, ino_t ino, int flags, vnode_t *vpp)
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	znode_t *zp;
	int err;

	/*
	* zfs_zget() can't operate on virtual entries like .zfs/ or
	* .zfs/snapshot/ directories, that's why we return EOPNOTSUPP.
	* This will make NFS to switch to LOOKUP instead of using VGET.
	*/
	if (ino == ZFSCTL_INO_ROOT \|\| ino == ZFSCTL_INO_SNAPDIR \|\|
	(zfsvfs->z_shares_dir != 0 && ino == zfsvfs->z_shares_dir))
	return (EOPNOTSUPP);

	if ((err = zfs_enter(zfsvfs, FTAG)) != 0)
	return (err);
	err = zfs_zget(zfsvfs, ino, &zp);
	if (err == 0 && zp->z_unlinked) {
	vrele(ZTOV(zp));
	err = EINVAL;
	}
	if (err == 0)
	*vpp = ZTOV(zp);
	zfs_exit(zfsvfs, FTAG);
	if (err == 0) {
	err = vn_lock(*vpp, flags);
	if (err != 0)
	vrele(*vpp);
	}
	if (err != 0)
	*vpp = NULL;
	return (err);
	}

	static int
	#if __FreeBSD_version >= 1300098
	zfs_checkexp(vfs_t vfsp, struct sockaddr nam, uint64_t *extflagsp,
	struct ucred *credanonp, int numsecflavors, int *secflavors)
	#else
	zfs_checkexp(vfs_t vfsp, struct sockaddr nam, int *extflagsp,
	struct ucred *credanonp, int numsecflavors, int **secflavors)
	#endif
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;

	/*
	* If this is regular file system vfsp is the same as
	* zfsvfs->z_parent->z_vfs, but if it is snapshot,
	* zfsvfs->z_parent->z_vfs represents parent file system
	* which we have to use here, because only this file system
	* has mnt_export configured.
	*/
	return (vfs_stdcheckexp(zfsvfs->z_parent->z_vfs, nam, extflagsp,
	credanonp, numsecflavors, secflavors));
	}

	_Static_assert(sizeof (struct fid) >= SHORT_FID_LEN,
	"struct fid bigger than SHORT_FID_LEN");
	_Static_assert(sizeof (struct fid) >= LONG_FID_LEN,
	"struct fid bigger than LONG_FID_LEN");

	static int
	zfs_fhtovp(vfs_t vfsp, fid_t fidp, int flags, vnode_t **vpp)
	{
	struct componentname cn;
	zfsvfs_t *zfsvfs = vfsp->vfs_data;
	znode_t *zp;
	vnode_t *dvp;
	uint64_t object = 0;
	uint64_t fid_gen = 0;
	uint64_t setgen = 0;
	uint64_t gen_mask;
	uint64_t zp_gen;
	int i, err;

	*vpp = NULL;

	if ((err = zfs_enter(zfsvfs, FTAG)) != 0)
	return (err);

	/*
	* On FreeBSD we can get snapshot's mount point or its parent file
	* system mount point depending if snapshot is already mounted or not.
	*/
	if (zfsvfs->z_parent == zfsvfs && fidp->fid_len == LONG_FID_LEN) {
	zfid_long_t zlfid = (zfid_long_t )fidp;
	uint64_t objsetid = 0;

	for (i = 0; i < sizeof (zlfid->zf_setid); i++)
	objsetid \|= ((uint64_t)zlfid->zf_setid[i]) << (8 * i);

	for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
	setgen \|= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);

	zfs_exit(zfsvfs, FTAG);

	err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs);
	if (err)
	return (SET_ERROR(EINVAL));
	if ((err = zfs_enter(zfsvfs, FTAG)) != 0)
	return (err);
	}

	if (fidp->fid_len == SHORT_FID_LEN \|\| fidp->fid_len == LONG_FID_LEN) {
	zfid_short_t zfid = (zfid_short_t )fidp;

	for (i = 0; i < sizeof (zfid->zf_object); i++)
	object \|= ((uint64_t)zfid->zf_object[i]) << (8 * i);

	for (i = 0; i < sizeof (zfid->zf_gen); i++)
	fid_gen \|= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
	} else {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (fidp->fid_len == LONG_FID_LEN && (fid_gen > 1 \|\| setgen != 0)) {
	dprintf("snapdir fid: fid_gen (%llu) and setgen (%llu)\n",
	(u_longlong_t)fid_gen, (u_longlong_t)setgen);
	return (SET_ERROR(EINVAL));
	}

	/*
	* A zero fid_gen means we are in .zfs or the .zfs/snapshot
	* directory tree. If the object == zfsvfs->z_shares_dir, then
	* we are in the .zfs/shares directory tree.
	*/
	if ((fid_gen == 0 &&
	(object == ZFSCTL_INO_ROOT \|\| object == ZFSCTL_INO_SNAPDIR)) \|\|
	(zfsvfs->z_shares_dir != 0 && object == zfsvfs->z_shares_dir)) {
	zfs_exit(zfsvfs, FTAG);
	VERIFY0(zfsctl_root(zfsvfs, LK_SHARED, &dvp));
	if (object == ZFSCTL_INO_SNAPDIR) {
	cn.cn_nameptr = "snapshot";
	cn.cn_namelen = strlen(cn.cn_nameptr);
	cn.cn_nameiop = LOOKUP;
	cn.cn_flags = ISLASTCN \| LOCKLEAF;
	cn.cn_lkflags = flags;
	VERIFY0(VOP_LOOKUP(dvp, vpp, &cn));
	vput(dvp);
	} else if (object == zfsvfs->z_shares_dir) {
	/*
	* XXX This branch must not be taken,
	* if it is, then the lookup below will
	* explode.
	*/
	cn.cn_nameptr = "shares";
	cn.cn_namelen = strlen(cn.cn_nameptr);
	cn.cn_nameiop = LOOKUP;
	cn.cn_flags = ISLASTCN;
	cn.cn_lkflags = flags;
	VERIFY0(VOP_LOOKUP(dvp, vpp, &cn));
	vput(dvp);
	} else {
	*vpp = dvp;
	}
	return (err);
	}

	gen_mask = -1ULL >> (64 - 8 * i);

	dprintf("getting %llu [%llu mask %llx]\n", (u_longlong_t)object,
	(u_longlong_t)fid_gen,
	(u_longlong_t)gen_mask);
	if ((err = zfs_zget(zfsvfs, object, &zp))) {
	zfs_exit(zfsvfs, FTAG);
	return (err);
	}
	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen,
	sizeof (uint64_t));
	zp_gen = zp_gen & gen_mask;
	if (zp_gen == 0)
	zp_gen = 1;
	if (zp->z_unlinked \|\| zp_gen != fid_gen) {
	dprintf("znode gen (%llu) != fid gen (%llu)\n",
	(u_longlong_t)zp_gen, (u_longlong_t)fid_gen);
	vrele(ZTOV(zp));
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	*vpp = ZTOV(zp);
	zfs_exit(zfsvfs, FTAG);
	err = vn_lock(*vpp, flags);
	if (err == 0)
	vnode_create_vobject(*vpp, zp->z_size, curthread);
	else
	*vpp = NULL;
	return (err);
	}

	/*
	* Block out VOPs and close zfsvfs_t::z_os
	*
	* Note, if successful, then we return with the 'z_teardown_lock' and
	* 'z_teardown_inactive_lock' write held. We leave ownership of the underlying
	* dataset and objset intact so that they can be atomically handed off during
	* a subsequent rollback or recv operation and the resume thereafter.
	*/
	int
	zfs_suspend_fs(zfsvfs_t *zfsvfs)
	{
	int error;

	if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0)
	return (error);

	return (0);
	}

	/*
	* Rebuild SA and release VOPs. Note that ownership of the underlying dataset
	* is an invariant across any of the operations that can be performed while the
	* filesystem was suspended. Whether it succeeded or failed, the preconditions
	* are the same: the relevant objset and associated dataset are owned by
	* zfsvfs, held, and long held on entry.
	*/
	int
	zfs_resume_fs(zfsvfs_t zfsvfs, dsl_dataset_t ds)
	{
	int err;
	znode_t *zp;

	ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs));
	ASSERT(ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs));

	/*
	* We already own this, so just update the objset_t, as the one we
	* had before may have been evicted.
	*/
	objset_t *os;
	VERIFY3P(ds->ds_owner, ==, zfsvfs);
	VERIFY(dsl_dataset_long_held(ds));
	dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds));
	dsl_pool_config_enter(dp, FTAG);
	VERIFY0(dmu_objset_from_ds(ds, &os));
	dsl_pool_config_exit(dp, FTAG);

	err = zfsvfs_init(zfsvfs, os);
	if (err != 0)
	goto bail;

	ds->ds_dir->dd_activity_cancelled = B_FALSE;
	VERIFY0(zfsvfs_setup(zfsvfs, B_FALSE));

	zfs_set_fuid_feature(zfsvfs);

	/*
	* Attempt to re-establish all the active znodes with
	* their dbufs. If a zfs_rezget() fails, then we'll let
	* any potential callers discover that via zfs_enter_verify_zp
	* when they try to use their znode.
	*/
	mutex_enter(&zfsvfs->z_znodes_lock);
	for (zp = list_head(&zfsvfs->z_all_znodes); zp;
	zp = list_next(&zfsvfs->z_all_znodes, zp)) {
	(void) zfs_rezget(zp);
	}
	mutex_exit(&zfsvfs->z_znodes_lock);

	bail:
	/* release the VOPs */
	ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs);
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);

	if (err) {
	/*
	* Since we couldn't setup the sa framework, try to force
	* unmount this file system.
	*/
	if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0) {
	vfs_ref(zfsvfs->z_vfs);
	(void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread);
	}
	}
	return (err);
	}

	static void
	zfs_freevfs(vfs_t *vfsp)
	{
	zfsvfs_t *zfsvfs = vfsp->vfs_data;

	zfsvfs_free(zfsvfs);

	atomic_dec_32(&zfs_active_fs_count);
	}

	#ifdef __i386__
	static int desiredvnodes_backup;
	#include <sys/vmmeter.h>


	#include <vm/vm_page.h>
	#include <vm/vm_object.h>
	#include <vm/vm_kern.h>
	#include <vm/vm_map.h>
	#endif

	static void
	zfs_vnodes_adjust(void)
	{
	#ifdef __i386__
	int newdesiredvnodes;

	desiredvnodes_backup = desiredvnodes;

	/*
	* We calculate newdesiredvnodes the same way it is done in
	* vntblinit(). If it is equal to desiredvnodes, it means that
	* it wasn't tuned by the administrator and we can tune it down.
	*/
	newdesiredvnodes = min(maxproc + vm_cnt.v_page_count / 4, 2 *
	vm_kmem_size / (5 * (sizeof (struct vm_object) +
	sizeof (struct vnode))));
	if (newdesiredvnodes == desiredvnodes)
	desiredvnodes = (3 * newdesiredvnodes) / 4;
	#endif
	}

	static void
	zfs_vnodes_adjust_back(void)
	{

	#ifdef __i386__
	desiredvnodes = desiredvnodes_backup;
	#endif
	}

	void
	zfs_init(void)
	{

	printf("ZFS filesystem version: " ZPL_VERSION_STRING "\n");

	/*
	* Initialize .zfs directory structures
	*/
	zfsctl_init();

	/*
	* Initialize znode cache, vnode ops, etc...
	*/
	zfs_znode_init();

	/*
	* Reduce number of vnodes. Originally number of vnodes is calculated
	* with UFS inode in mind. We reduce it here, because it's too big for
	* ZFS/i386.
	*/
	zfs_vnodes_adjust();

	dmu_objset_register_type(DMU_OST_ZFS, zpl_get_file_info);

	zfsvfs_taskq = taskq_create("zfsvfs", 1, minclsyspri, 0, 0, 0);
	}

	void
	zfs_fini(void)
	{
	taskq_destroy(zfsvfs_taskq);
	zfsctl_fini();
	zfs_znode_fini();
	zfs_vnodes_adjust_back();
	}

	int
	zfs_busy(void)
	{
	return (zfs_active_fs_count != 0);
	}

	/*
	* Release VOPs and unmount a suspended filesystem.
	*/
	int
	zfs_end_fs(zfsvfs_t zfsvfs, dsl_dataset_t ds)
	{
	ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs));
	ASSERT(ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zfsvfs));

	/*
	* We already own this, so just hold and rele it to update the
	* objset_t, as the one we had before may have been evicted.
	*/
	objset_t *os;
	VERIFY3P(ds->ds_owner, ==, zfsvfs);
	VERIFY(dsl_dataset_long_held(ds));
	dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds));
	dsl_pool_config_enter(dp, FTAG);
	VERIFY0(dmu_objset_from_ds(ds, &os));
	dsl_pool_config_exit(dp, FTAG);
	zfsvfs->z_os = os;

	/* release the VOPs */
	ZFS_TEARDOWN_INACTIVE_EXIT_WRITE(zfsvfs);
	ZFS_TEARDOWN_EXIT(zfsvfs, FTAG);

	/*
	* Try to force unmount this file system.
	*/
	(void) zfs_umount(zfsvfs->z_vfs, 0);
	zfsvfs->z_unmounted = B_TRUE;
	return (0);
	}

	int
	zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers)
	{
	int error;
	objset_t *os = zfsvfs->z_os;
	dmu_tx_t *tx;

	if (newvers < ZPL_VERSION_INITIAL \|\| newvers > ZPL_VERSION)
	return (SET_ERROR(EINVAL));

	if (newvers < zfsvfs->z_version)
	return (SET_ERROR(EINVAL));

	if (zfs_spa_version_map(newvers) >
	spa_version(dmu_objset_spa(zfsvfs->z_os)))
	return (SET_ERROR(ENOTSUP));

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
	if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) {
	dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
	ZFS_SA_ATTRS);
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
	}
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	return (error);
	}

	error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
	8, 1, &newvers, tx);

	if (error) {
	dmu_tx_commit(tx);
	return (error);
	}

	if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) {
	uint64_t sa_obj;

	ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=,
	SPA_VERSION_SA);
	sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
	DMU_OT_NONE, 0, tx);

	error = zap_add(os, MASTER_NODE_OBJ,
	ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
	ASSERT0(error);

	VERIFY0(sa_set_sa_object(os, sa_obj));
	sa_register_update_callback(os, zfs_sa_upgrade);
	}

	spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx,
	"from %ju to %ju", (uintmax_t)zfsvfs->z_version,
	(uintmax_t)newvers);
	dmu_tx_commit(tx);

	zfsvfs->z_version = newvers;
	os->os_version = newvers;

	zfs_set_fuid_feature(zfsvfs);

	return (0);
	}

	/*
	* Read a property stored within the master node.
	*/
	int
	zfs_get_zplprop(objset_t os, zfs_prop_t prop, uint64_t value)
	{
	uint64_t *cached_copy = NULL;

	/*
	* Figure out where in the objset_t the cached copy would live, if it
	* is available for the requested property.
	*/
	if (os != NULL) {
	switch (prop) {
	case ZFS_PROP_VERSION:
	cached_copy = &os->os_version;
	break;
	case ZFS_PROP_NORMALIZE:
	cached_copy = &os->os_normalization;
	break;
	case ZFS_PROP_UTF8ONLY:
	cached_copy = &os->os_utf8only;
	break;
	case ZFS_PROP_CASE:
	cached_copy = &os->os_casesensitivity;
	break;
	default:
	break;
	}
	}
	if (cached_copy != NULL && *cached_copy != OBJSET_PROP_UNINITIALIZED) {
	value = cached_copy;
	return (0);
	}

	/*
	* If the property wasn't cached, look up the file system's value for
	* the property. For the version property, we look up a slightly
	* different string.
	*/
	const char *pname;
	int error = ENOENT;
	if (prop == ZFS_PROP_VERSION) {
	pname = ZPL_VERSION_STR;
	} else {
	pname = zfs_prop_to_name(prop);
	}

	if (os != NULL) {
	ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS);
	error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value);
	}

	if (error == ENOENT) {
	/* No value set, use the default value */
	switch (prop) {
	case ZFS_PROP_VERSION:
	*value = ZPL_VERSION;
	break;
	case ZFS_PROP_NORMALIZE:
	case ZFS_PROP_UTF8ONLY:
	*value = 0;
	break;
	case ZFS_PROP_CASE:
	*value = ZFS_CASE_SENSITIVE;
	break;
	case ZFS_PROP_ACLTYPE:
	*value = ZFS_ACLTYPE_NFSV4;
	break;
	default:
	return (error);
	}
	error = 0;
	}

	/*
	* If one of the methods for getting the property value above worked,
	* copy it into the objset_t's cache.
	*/
	if (error == 0 && cached_copy != NULL) {
	cached_copy = value;
	}

	return (error);
	}

	/*
	* Return true if the corresponding vfs's unmounted flag is set.
	* Otherwise return false.
	* If this function returns true we know VFS unmount has been initiated.
	*/
	boolean_t
	zfs_get_vfs_flag_unmounted(objset_t *os)
	{
	zfsvfs_t *zfvp;
	boolean_t unmounted = B_FALSE;

	ASSERT3U(dmu_objset_type(os), ==, DMU_OST_ZFS);

	mutex_enter(&os->os_user_ptr_lock);
	zfvp = dmu_objset_get_user(os);
	if (zfvp != NULL && zfvp->z_vfs != NULL &&
	(zfvp->z_vfs->mnt_kern_flag & MNTK_UNMOUNT))
	unmounted = B_TRUE;
	mutex_exit(&os->os_user_ptr_lock);

	return (unmounted);
	}

	#ifdef _KERNEL
	void
	zfsvfs_update_fromname(const char oldname, const char newname)
	{
	char tmpbuf[MAXPATHLEN];
	struct mount *mp;
	char *fromname;
	size_t oldlen;

	oldlen = strlen(oldname);

	mtx_lock(&mountlist_mtx);
	TAILQ_FOREACH(mp, &mountlist, mnt_list) {
	fromname = mp->mnt_stat.f_mntfromname;
	if (strcmp(fromname, oldname) == 0) {
	(void) strlcpy(fromname, newname,
	sizeof (mp->mnt_stat.f_mntfromname));
	continue;
	}
	if (strncmp(fromname, oldname, oldlen) == 0 &&
	(fromname[oldlen] == '/' \|\| fromname[oldlen] == '@')) {
	(void) snprintf(tmpbuf, sizeof (tmpbuf), "%s%s",
	newname, fromname + oldlen);
	(void) strlcpy(fromname, tmpbuf,
	sizeof (mp->mnt_stat.f_mntfromname));
	continue;
	}
	}
	mtx_unlock(&mountlist_mtx);
	}
	#endif
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-err.c b/sys/contrib/openzfs/module/os/linux/spl/spl-err.c
	index c84c39b56bf7..7d3f6127c4af 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-err.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-err.c
	@@ -1,125 +1,124 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Error Implementation.
	*/

	#include <sys/sysmacros.h>
	#include <sys/cmn_err.h>

	/*
	* It is often useful to actually have the panic crash the node so you
	* can then get notified of the event, get the crashdump for later
	* analysis and other such goodies.
	* But we would still default to the current default of not to do that.
	*/
	unsigned int spl_panic_halt;
	/* CSTYLED */
	module_param(spl_panic_halt, uint, 0644);
	MODULE_PARM_DESC(spl_panic_halt, "Cause kernel panic on assertion failures");

	void
	spl_dumpstack(void)
	{
	printk("Showing stack for process %d\n", current->pid);
	dump_stack();
	}
	EXPORT_SYMBOL(spl_dumpstack);

	-int
	+void
	spl_panic(const char file, const char func, int line, const char *fmt, ...)
	{
	const char *newfile;
	char msg[MAXMSGLEN];
	va_list ap;

	newfile = strrchr(file, '/');
	if (newfile != NULL)
	newfile = newfile + 1;
	else
	newfile = file;

	va_start(ap, fmt);
	(void) vsnprintf(msg, sizeof (msg), fmt, ap);
	va_end(ap);

	printk(KERN_EMERG "%s", msg);
	printk(KERN_EMERG "PANIC at %s:%d:%s()\n", newfile, line, func);
	if (spl_panic_halt)
	panic("%s", msg);

	spl_dumpstack();

	/* Halt the thread to facilitate further debugging */
	set_current_state(TASK_UNINTERRUPTIBLE);
	while (1)
	schedule();

	/* Unreachable */
	- return (1);
	}
	EXPORT_SYMBOL(spl_panic);

	void
	vcmn_err(int ce, const char *fmt, va_list ap)
	{
	char msg[MAXMSGLEN];

	vsnprintf(msg, MAXMSGLEN, fmt, ap);

	switch (ce) {
	case CE_IGNORE:
	break;
	case CE_CONT:
	printk("%s", msg);
	break;
	case CE_NOTE:
	printk(KERN_NOTICE "NOTICE: %s\n", msg);
	break;
	case CE_WARN:
	printk(KERN_WARNING "WARNING: %s\n", msg);
	break;
	case CE_PANIC:
	printk(KERN_EMERG "PANIC: %s\n", msg);
	if (spl_panic_halt)
	panic("%s", msg);

	spl_dumpstack();

	/* Halt the thread to facilitate further debugging */
	set_current_state(TASK_UNINTERRUPTIBLE);
	while (1)
	schedule();
	}
	} /* vcmn_err() */
	EXPORT_SYMBOL(vcmn_err);

	void
	cmn_err(int ce, const char *fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);
	vcmn_err(ce, fmt, ap);
	va_end(ap);
	} /* cmn_err() */
	EXPORT_SYMBOL(cmn_err);
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c b/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c
	index de91c44257aa..bc39ece9a427 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c
	@@ -1,834 +1,867 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Generic Implementation.
	*/

	#include <sys/sysmacros.h>
	#include <sys/systeminfo.h>
	#include <sys/vmsystm.h>
	#include <sys/kmem.h>
	#include <sys/kmem_cache.h>
	#include <sys/vmem.h>
	#include <sys/mutex.h>
	#include <sys/rwlock.h>
	#include <sys/taskq.h>
	#include <sys/tsd.h>
	#include <sys/zmod.h>
	#include <sys/debug.h>
	#include <sys/proc.h>
	#include <sys/kstat.h>
	#include <sys/file.h>
	#include <sys/sunddi.h>
	#include <linux/ctype.h>
	#include <sys/disp.h>
	#include <sys/random.h>
	#include <sys/string.h>
	#include <linux/kmod.h>
	#include <linux/mod_compat.h>
	#include <sys/cred.h>
	#include <sys/vnode.h>
	+#include <sys/misc.h>

	unsigned long spl_hostid = 0;
	EXPORT_SYMBOL(spl_hostid);

	/* CSTYLED */
	module_param(spl_hostid, ulong, 0644);
	MODULE_PARM_DESC(spl_hostid, "The system hostid.");

	proc_t p0;
	EXPORT_SYMBOL(p0);

	/*
	* Xorshift Pseudo Random Number Generator based on work by Sebastiano Vigna
	*
	* "Further scramblings of Marsaglia's xorshift generators"
	* http://vigna.di.unimi.it/ftp/papers/xorshiftplus.pdf
	*
	* random_get_pseudo_bytes() is an API function on Illumos whose sole purpose
	* is to provide bytes containing random numbers. It is mapped to /dev/urandom
	* on Illumos, which uses a "FIPS 186-2 algorithm". No user of the SPL's
	* random_get_pseudo_bytes() needs bytes that are of cryptographic quality, so
	* we can implement it using a fast PRNG that we seed using Linux' actual
	* equivalent to random_get_pseudo_bytes(). We do this by providing each CPU
	* with an independent seed so that all calls to random_get_pseudo_bytes() are
	* free of atomic instructions.
	*
	* A consequence of using a fast PRNG is that using random_get_pseudo_bytes()
	* to generate words larger than 128 bits will paradoxically be limited to
	* `2^128 - 1` possibilities. This is because we have a sequence of `2^128 - 1`
	* 128-bit words and selecting the first will implicitly select the second. If
	* a caller finds this behavior undesirable, random_get_bytes() should be used
	* instead.
	*
	* XXX: Linux interrupt handlers that trigger within the critical section
	* formed by `s[1] = xp[1];` and `xp[0] = s[0];` and call this function will
	* see the same numbers. Nothing in the code currently calls this in an
	* interrupt handler, so this is considered to be okay. If that becomes a
	* problem, we could create a set of per-cpu variables for interrupt handlers
	* and use them when in_interrupt() from linux/preempt_mask.h evaluates to
	* true.
	*/
	void __percpu *spl_pseudo_entropy;

	/*
	* spl_rand_next()/spl_rand_jump() are copied from the following CC-0 licensed
	* file:
	*
	* http://xorshift.di.unimi.it/xorshift128plus.c
	*/

	static inline uint64_t
	spl_rand_next(uint64_t *s)
	{
	uint64_t s1 = s[0];
	const uint64_t s0 = s[1];
	s[0] = s0;
	s1 ^= s1 << 23; // a
	s[1] = s1 ^ s0 ^ (s1 >> 18) ^ (s0 >> 5); // b, c
	return (s[1] + s0);
	}

	static inline void
	spl_rand_jump(uint64_t *s)
	{
	static const uint64_t JUMP[] =
	{ 0x8a5cd789635d2dff, 0x121fd2155c472f96 };

	uint64_t s0 = 0;
	uint64_t s1 = 0;
	int i, b;
	for (i = 0; i < sizeof (JUMP) / sizeof (*JUMP); i++)
	for (b = 0; b < 64; b++) {
	if (JUMP[i] & 1ULL << b) {
	s0 ^= s[0];
	s1 ^= s[1];
	}
	(void) spl_rand_next(s);
	}

	s[0] = s0;
	s[1] = s1;
	}

	int
	random_get_pseudo_bytes(uint8_t *ptr, size_t len)
	{
	uint64_t *xp, s[2];

	ASSERT(ptr);

	xp = get_cpu_ptr(spl_pseudo_entropy);

	s[0] = xp[0];
	s[1] = xp[1];

	while (len) {
	union {
	uint64_t ui64;
	uint8_t byte[sizeof (uint64_t)];
	}entropy;
	int i = MIN(len, sizeof (uint64_t));

	len -= i;
	entropy.ui64 = spl_rand_next(s);

	while (i--)
	*ptr++ = entropy.byte[i];
	}

	xp[0] = s[0];
	xp[1] = s[1];

	put_cpu_ptr(spl_pseudo_entropy);

	return (0);
	}


	EXPORT_SYMBOL(random_get_pseudo_bytes);

	#if BITS_PER_LONG == 32

	/*
	* Support 64/64 => 64 division on a 32-bit platform. While the kernel
	* provides a div64_u64() function for this we do not use it because the
	* implementation is flawed. There are cases which return incorrect
	* results as late as linux-2.6.35. Until this is fixed upstream the
	* spl must provide its own implementation.
	*
	* This implementation is a slightly modified version of the algorithm
	* proposed by the book 'Hacker's Delight'. The original source can be
	* found here and is available for use without restriction.
	*
	* http://www.hackersdelight.org/HDcode/newCode/divDouble.c
	*/

	/*
	* Calculate number of leading of zeros for a 64-bit value.
	*/
	static int
	nlz64(uint64_t x)
	{
	register int n = 0;

	if (x == 0)
	return (64);

	if (x <= 0x00000000FFFFFFFFULL) { n = n + 32; x = x << 32; }
	if (x <= 0x0000FFFFFFFFFFFFULL) { n = n + 16; x = x << 16; }
	if (x <= 0x00FFFFFFFFFFFFFFULL) { n = n + 8; x = x << 8; }
	if (x <= 0x0FFFFFFFFFFFFFFFULL) { n = n + 4; x = x << 4; }
	if (x <= 0x3FFFFFFFFFFFFFFFULL) { n = n + 2; x = x << 2; }
	if (x <= 0x7FFFFFFFFFFFFFFFULL) { n = n + 1; }

	return (n);
	}

	/*
	* Newer kernels have a div_u64() function but we define our own
	* to simplify portability between kernel versions.
	*/
	static inline uint64_t
	__div_u64(uint64_t u, uint32_t v)
	{
	(void) do_div(u, v);
	return (u);
	}

	/*
	* Turn off missing prototypes warning for these functions. They are
	* replacements for libgcc-provided functions and will never be called
	* directly.
	*/
	#pragma GCC diagnostic push
	#pragma GCC diagnostic ignored "-Wmissing-prototypes"

	/*
	* Implementation of 64-bit unsigned division for 32-bit machines.
	*
	* First the procedure takes care of the case in which the divisor is a
	* 32-bit quantity. There are two subcases: (1) If the left half of the
	* dividend is less than the divisor, one execution of do_div() is all that
	* is required (overflow is not possible). (2) Otherwise it does two
	* divisions, using the grade school method.
	*/
	uint64_t
	__udivdi3(uint64_t u, uint64_t v)
	{
	uint64_t u0, u1, v1, q0, q1, k;
	int n;

	if (v >> 32 == 0) { // If v < 2**32:
	if (u >> 32 < v) { // If u/v cannot overflow,
	return (__div_u64(u, v)); // just do one division.
	} else { // If u/v would overflow:
	u1 = u >> 32; // Break u into two halves.
	u0 = u & 0xFFFFFFFF;
	q1 = __div_u64(u1, v); // First quotient digit.
	k = u1 - q1 * v; // First remainder, < v.
	u0 += (k << 32);
	q0 = __div_u64(u0, v); // Seconds quotient digit.
	return ((q1 << 32) + q0);
	}
	} else { // If v >= 2**32:
	n = nlz64(v); // 0 <= n <= 31.
	v1 = (v << n) >> 32; // Normalize divisor, MSB is 1.
	u1 = u >> 1; // To ensure no overflow.
	q1 = __div_u64(u1, v1); // Get quotient from
	q0 = (q1 << n) >> 31; // Undo normalization and
	// division of u by 2.
	if (q0 != 0) // Make q0 correct or
	q0 = q0 - 1; // too small by 1.
	if ((u - q0 * v) >= v)
	q0 = q0 + 1; // Now q0 is correct.

	return (q0);
	}
	}
	EXPORT_SYMBOL(__udivdi3);

	#ifndef abs64
	/* CSTYLED */
	#define abs64(x) ({ uint64_t t = (x) >> 63; ((x) ^ t) - t; })
	#endif

	/*
	* Implementation of 64-bit signed division for 32-bit machines.
	*/
	int64_t
	__divdi3(int64_t u, int64_t v)
	{
	int64_t q, t;
	q = __udivdi3(abs64(u), abs64(v));
	t = (u ^ v) >> 63; // If u, v have different
	return ((q ^ t) - t); // signs, negate q.
	}
	EXPORT_SYMBOL(__divdi3);

	/*
	* Implementation of 64-bit unsigned modulo for 32-bit machines.
	*/
	uint64_t
	__umoddi3(uint64_t dividend, uint64_t divisor)
	{
	return (dividend - (divisor * __udivdi3(dividend, divisor)));
	}
	EXPORT_SYMBOL(__umoddi3);

	/* 64-bit signed modulo for 32-bit machines. */
	int64_t
	__moddi3(int64_t n, int64_t d)
	{
	int64_t q;
	boolean_t nn = B_FALSE;

	if (n < 0) {
	nn = B_TRUE;
	n = -n;
	}
	if (d < 0)
	d = -d;

	q = __umoddi3(n, d);

	return (nn ? -q : q);
	}
	EXPORT_SYMBOL(__moddi3);

	/*
	* Implementation of 64-bit unsigned division/modulo for 32-bit machines.
	*/
	uint64_t
	__udivmoddi4(uint64_t n, uint64_t d, uint64_t *r)
	{
	uint64_t q = __udivdi3(n, d);
	if (r)
	r = n - d q;
	return (q);
	}
	EXPORT_SYMBOL(__udivmoddi4);

	/*
	* Implementation of 64-bit signed division/modulo for 32-bit machines.
	*/
	int64_t
	__divmoddi4(int64_t n, int64_t d, int64_t *r)
	{
	int64_t q, rr;
	boolean_t nn = B_FALSE;
	boolean_t nd = B_FALSE;
	if (n < 0) {
	nn = B_TRUE;
	n = -n;
	}
	if (d < 0) {
	nd = B_TRUE;
	d = -d;
	}

	q = __udivmoddi4(n, d, (uint64_t *)&rr);

	if (nn != nd)
	q = -q;
	if (nn)
	rr = -rr;
	if (r)
	*r = rr;
	return (q);
	}
	EXPORT_SYMBOL(__divmoddi4);

	#if defined(__arm) \|\| defined(__arm__)
	/*
	* Implementation of 64-bit (un)signed division for 32-bit arm machines.
	*
	* Run-time ABI for the ARM Architecture (page 20). A pair of (unsigned)
	* long longs is returned in {{r0, r1}, {r2,r3}}, the quotient in {r0, r1},
	* and the remainder in {r2, r3}. The return type is specifically left
	* set to 'void' to ensure the compiler does not overwrite these registers
	* during the return. All results are in registers as per ABI
	*/
	void
	__aeabi_uldivmod(uint64_t u, uint64_t v)
	{
	uint64_t res;
	uint64_t mod;

	res = __udivdi3(u, v);
	mod = __umoddi3(u, v);
	{
	register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
	register uint32_t r1 asm("r1") = (res >> 32);
	register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
	register uint32_t r3 asm("r3") = (mod >> 32);

	asm volatile(""
	: "+r"(r0), "+r"(r1), "+r"(r2), "+r"(r3) /* output */
	: "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */

	return; /* r0; */
	}
	}
	EXPORT_SYMBOL(__aeabi_uldivmod);

	void
	__aeabi_ldivmod(int64_t u, int64_t v)
	{
	int64_t res;
	uint64_t mod;

	res = __divdi3(u, v);
	mod = __umoddi3(u, v);
	{
	register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
	register uint32_t r1 asm("r1") = (res >> 32);
	register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
	register uint32_t r3 asm("r3") = (mod >> 32);

	asm volatile(""
	: "+r"(r0), "+r"(r1), "+r"(r2), "+r"(r3) /* output */
	: "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */

	return; /* r0; */
	}
	}
	EXPORT_SYMBOL(__aeabi_ldivmod);
	#endif /* __arm \|\| __arm__ */

	#pragma GCC diagnostic pop

	#endif /* BITS_PER_LONG */

	/*
	* NOTE: The strtoxx behavior is solely based on my reading of the Solaris
	* ddi_strtol(9F) man page. I have not verified the behavior of these
	* functions against their Solaris counterparts. It is possible that I
	* may have misinterpreted the man page or the man page is incorrect.
	*/
	int ddi_strtol(const char , char , int, long );
	int ddi_strtoull(const char , char , int, unsigned long long );
	int ddi_strtoll(const char , char , int, long long );

	#define define_ddi_strtox(type, valtype) \
	int ddi_strto##type(const char str, char *endptr, \
	int base, valtype *result) \
	{ \
	valtype last_value, value = 0; \
	char ptr = (char )str; \
	int digit, minus = 0; \
	\
	while (strchr(" \t\n\r\f", *ptr)) \
	++ptr; \
	\
	if (strlen(ptr) == 0) \
	return (EINVAL); \
	\
	switch (*ptr) { \
	case '-': \
	minus = 1; \
	zfs_fallthrough; \
	case '+': \
	++ptr; \
	break; \
	} \
	\
	/* Auto-detect base based on prefix */ \
	if (!base) { \
	if (str[0] == '0') { \
	if (tolower(str[1]) == 'x' && isxdigit(str[2])) { \
	base = 16; /* hex */ \
	ptr += 2; \
	} else if (str[1] >= '0' && str[1] < 8) { \
	base = 8; /* octal */ \
	ptr += 1; \
	} else { \
	return (EINVAL); \
	} \
	} else { \
	base = 10; /* decimal */ \
	} \
	} \
	\
	while (1) { \
	if (isdigit(*ptr)) \
	digit = *ptr - '0'; \
	else if (isalpha(*ptr)) \
	digit = tolower(*ptr) - 'a' + 10; \
	else \
	break; \
	\
	if (digit >= base) \
	break; \
	\
	last_value = value; \
	value = value * base + digit; \
	if (last_value > value) /* Overflow */ \
	return (ERANGE); \
	\
	ptr++; \
	} \
	\
	*result = minus ? -value : value; \
	\
	if (endptr) \
	*endptr = ptr; \
	\
	return (0); \
	} \

	define_ddi_strtox(l, long)
	define_ddi_strtox(ull, unsigned long long)
	define_ddi_strtox(ll, long long)

	EXPORT_SYMBOL(ddi_strtol);
	EXPORT_SYMBOL(ddi_strtoll);
	EXPORT_SYMBOL(ddi_strtoull);

	int
	ddi_copyin(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyin(from, to, len));
	}
	EXPORT_SYMBOL(ddi_copyin);

	+/*
	+ * Post a uevent to userspace whenever a new vdev adds to the pool. It is
	+ * necessary to sync blkid information with udev, which zed daemon uses
	+ * during device hotplug to identify the vdev.
	+ */
	+void
	+spl_signal_kobj_evt(struct block_device *bdev)
	+{
	+#if defined(HAVE_BDEV_KOBJ) \|\| defined(HAVE_PART_TO_DEV)
	+#ifdef HAVE_BDEV_KOBJ
	+ struct kobject *disk_kobj = bdev_kobj(bdev);
	+#else
	+ struct kobject *disk_kobj = &part_to_dev(bdev->bd_part)->kobj;
	+#endif
	+ if (disk_kobj) {
	+ int ret = kobject_uevent(disk_kobj, KOBJ_CHANGE);
	+ if (ret) {
	+ pr_warn("ZFS: Sending event '%d' to kobject: '%s'"
	+ " (%p): failed(ret:%d)\n", KOBJ_CHANGE,
	+ kobject_name(disk_kobj), disk_kobj, ret);
	+ }
	+ }
	+#else
	+/*
	+ * This is encountered if neither bdev_kobj() nor part_to_dev() is available
	+ * in the kernel - likely due to an API change that needs to be chased down.
	+ */
	+#error "Unsupported kernel: unable to get struct kobj from bdev"
	+#endif
	+}
	+EXPORT_SYMBOL(spl_signal_kobj_evt);
	+
	int
	ddi_copyout(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyout(from, to, len));
	}
	EXPORT_SYMBOL(ddi_copyout);

	static ssize_t
	spl_kernel_read(struct file file, void buf, size_t count, loff_t *pos)
	{
	#if defined(HAVE_KERNEL_READ_PPOS)
	return (kernel_read(file, buf, count, pos));
	#else
	mm_segment_t saved_fs;
	ssize_t ret;

	saved_fs = get_fs();
	set_fs(KERNEL_DS);

	ret = vfs_read(file, (void __user *)buf, count, pos);

	set_fs(saved_fs);

	return (ret);
	#endif
	}

	static int
	spl_getattr(struct file filp, struct kstat stat)
	{
	int rc;

	ASSERT(filp);
	ASSERT(stat);

	#if defined(HAVE_4ARGS_VFS_GETATTR)
	rc = vfs_getattr(&filp->f_path, stat, STATX_BASIC_STATS,
	AT_STATX_SYNC_AS_STAT);
	#elif defined(HAVE_2ARGS_VFS_GETATTR)
	rc = vfs_getattr(&filp->f_path, stat);
	#elif defined(HAVE_3ARGS_VFS_GETATTR)
	rc = vfs_getattr(filp->f_path.mnt, filp->f_dentry, stat);
	#else
	#error "No available vfs_getattr()"
	#endif
	if (rc)
	return (-rc);

	return (0);
	}

	/*
	* Read the unique system identifier from the /etc/hostid file.
	*
	* The behavior of /usr/bin/hostid on Linux systems with the
	* regular eglibc and coreutils is:
	*
	* 1. Generate the value if the /etc/hostid file does not exist
	* or if the /etc/hostid file is less than four bytes in size.
	*
	* 2. If the /etc/hostid file is at least 4 bytes, then return
	* the first four bytes [0..3] in native endian order.
	*
	* 3. Always ignore bytes [4..] if they exist in the file.
	*
	* Only the first four bytes are significant, even on systems that
	* have a 64-bit word size.
	*
	* See:
	*
	* eglibc: sysdeps/unix/sysv/linux/gethostid.c
	* coreutils: src/hostid.c
	*
	* Notes:
	*
	* The /etc/hostid file on Solaris is a text file that often reads:
	*
	* # DO NOT EDIT
	* "0123456789"
	*
	* Directly copying this file to Linux results in a constant
	* hostid of 4f442023 because the default comment constitutes
	* the first four bytes of the file.
	*
	*/

	static char *spl_hostid_path = HW_HOSTID_PATH;
	module_param(spl_hostid_path, charp, 0444);
	MODULE_PARM_DESC(spl_hostid_path, "The system hostid file (/etc/hostid)");

	static int
	hostid_read(uint32_t *hostid)
	{
	uint64_t size;
	uint32_t value = 0;
	int error;
	loff_t off;
	struct file *filp;
	struct kstat stat;

	filp = filp_open(spl_hostid_path, 0, 0);

	if (IS_ERR(filp))
	return (ENOENT);

	error = spl_getattr(filp, &stat);
	if (error) {
	filp_close(filp, 0);
	return (error);
	}
	size = stat.size;
	// cppcheck-suppress sizeofwithnumericparameter
	if (size < sizeof (HW_HOSTID_MASK)) {
	filp_close(filp, 0);
	return (EINVAL);
	}

	off = 0;
	/*
	* Read directly into the variable like eglibc does.
	* Short reads are okay; native behavior is preserved.
	*/
	error = spl_kernel_read(filp, &value, sizeof (value), &off);
	if (error < 0) {
	filp_close(filp, 0);
	return (EIO);
	}

	/* Mask down to 32 bits like coreutils does. */
	*hostid = (value & HW_HOSTID_MASK);
	filp_close(filp, 0);

	return (0);
	}

	/*
	* Return the system hostid. Preferentially use the spl_hostid module option
	* when set, otherwise use the value in the /etc/hostid file.
	*/
	uint32_t
	zone_get_hostid(void *zone)
	{
	uint32_t hostid;

	ASSERT3P(zone, ==, NULL);

	if (spl_hostid != 0)
	return ((uint32_t)(spl_hostid & HW_HOSTID_MASK));

	if (hostid_read(&hostid) == 0)
	return (hostid);

	return (0);
	}
	EXPORT_SYMBOL(zone_get_hostid);

	static int
	spl_kvmem_init(void)
	{
	int rc = 0;

	rc = spl_kmem_init();
	if (rc)
	return (rc);

	rc = spl_vmem_init();
	if (rc) {
	spl_kmem_fini();
	return (rc);
	}

	return (rc);
	}

	/*
	* We initialize the random number generator with 128 bits of entropy from the
	* system random number generator. In the improbable case that we have a zero
	* seed, we fallback to the system jiffies, unless it is also zero, in which
	* situation we use a preprogrammed seed. We step forward by 2^64 iterations to
	* initialize each of the per-cpu seeds so that the sequences generated on each
	* CPU are guaranteed to never overlap in practice.
	*/
	static int __init
	spl_random_init(void)
	{
	uint64_t s[2];
	int i = 0;

	spl_pseudo_entropy = __alloc_percpu(2 * sizeof (uint64_t),
	sizeof (uint64_t));

	if (!spl_pseudo_entropy)
	return (-ENOMEM);

	get_random_bytes(s, sizeof (s));

	if (s[0] == 0 && s[1] == 0) {
	if (jiffies != 0) {
	s[0] = jiffies;
	s[1] = ~0 - jiffies;
	} else {
	(void) memcpy(s, "improbable seed", sizeof (s));
	}
	printk("SPL: get_random_bytes() returned 0 "
	"when generating random seed. Setting initial seed to "
	"0x%016llx%016llx.\n", cpu_to_be64(s[0]),
	cpu_to_be64(s[1]));
	}

	for_each_possible_cpu(i) {
	uint64_t *wordp = per_cpu_ptr(spl_pseudo_entropy, i);

	spl_rand_jump(s);

	wordp[0] = s[0];
	wordp[1] = s[1];
	}

	return (0);
	}

	static void
	spl_random_fini(void)
	{
	free_percpu(spl_pseudo_entropy);
	}

	static void
	spl_kvmem_fini(void)
	{
	spl_vmem_fini();
	spl_kmem_fini();
	}

	static int __init
	spl_init(void)
	{
	int rc = 0;

	if ((rc = spl_random_init()))
	goto out0;

	if ((rc = spl_kvmem_init()))
	goto out1;

	if ((rc = spl_tsd_init()))
	goto out2;

	if ((rc = spl_taskq_init()))
	goto out3;

	if ((rc = spl_kmem_cache_init()))
	goto out4;

	if ((rc = spl_proc_init()))
	goto out5;

	if ((rc = spl_kstat_init()))
	goto out6;

	if ((rc = spl_zlib_init()))
	goto out7;

	if ((rc = spl_zone_init()))
	goto out8;

	return (rc);

	out8:
	spl_zlib_fini();
	out7:
	spl_kstat_fini();
	out6:
	spl_proc_fini();
	out5:
	spl_kmem_cache_fini();
	out4:
	spl_taskq_fini();
	out3:
	spl_tsd_fini();
	out2:
	spl_kvmem_fini();
	out1:
	spl_random_fini();
	out0:
	return (rc);
	}

	static void __exit
	spl_fini(void)
	{
	spl_zone_fini();
	spl_zlib_fini();
	spl_kstat_fini();
	spl_proc_fini();
	spl_kmem_cache_fini();
	spl_taskq_fini();
	spl_tsd_fini();
	spl_kvmem_fini();
	spl_random_fini();
	}

	module_init(spl_init);
	module_exit(spl_fini);

	MODULE_DESCRIPTION("Solaris Porting Layer");
	MODULE_AUTHOR(ZFS_META_AUTHOR);
	MODULE_LICENSE("GPL");
	MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c
	index efb8d0c30330..e355e2bfc3a0 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-kmem-cache.c
	@@ -1,1465 +1,1465 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#include <linux/percpu_compat.h>
	#include <sys/kmem.h>
	#include <sys/kmem_cache.h>
	#include <sys/taskq.h>
	#include <sys/timer.h>
	#include <sys/vmem.h>
	#include <sys/wait.h>
	#include <linux/slab.h>
	#include <linux/swap.h>
	#include <linux/prefetch.h>

	/*
	* Within the scope of spl-kmem.c file the kmem_cache_* definitions
	* are removed to allow access to the real Linux slab allocator.
	*/
	#undef kmem_cache_destroy
	#undef kmem_cache_create
	#undef kmem_cache_alloc
	#undef kmem_cache_free


	/*
	* Linux 3.16 replaced smp_mb__{before,after}_{atomic,clear}_{dec,inc,bit}()
	* with smp_mb__{before,after}_atomic() because they were redundant. This is
	* only used inside our SLAB allocator, so we implement an internal wrapper
	* here to give us smp_mb__{before,after}_atomic() on older kernels.
	*/
	#ifndef smp_mb__before_atomic
	#define smp_mb__before_atomic(x) smp_mb__before_clear_bit(x)
	#endif

	#ifndef smp_mb__after_atomic
	#define smp_mb__after_atomic(x) smp_mb__after_clear_bit(x)
	#endif

	/* BEGIN CSTYLED */
	/*
	* Cache magazines are an optimization designed to minimize the cost of
	* allocating memory. They do this by keeping a per-cpu cache of recently
	* freed objects, which can then be reallocated without taking a lock. This
	* can improve performance on highly contended caches. However, because
	* objects in magazines will prevent otherwise empty slabs from being
	* immediately released this may not be ideal for low memory machines.
	*
	* For this reason spl_kmem_cache_magazine_size can be used to set a maximum
	* magazine size. When this value is set to 0 the magazine size will be
	* automatically determined based on the object size. Otherwise magazines
	* will be limited to 2-256 objects per magazine (i.e per cpu). Magazines
	* may never be entirely disabled in this implementation.
	*/
	static unsigned int spl_kmem_cache_magazine_size = 0;
	module_param(spl_kmem_cache_magazine_size, uint, 0444);
	MODULE_PARM_DESC(spl_kmem_cache_magazine_size,
	"Default magazine size (2-256), set automatically (0)");

	/*
	* The default behavior is to report the number of objects remaining in the
	* cache. This allows the Linux VM to repeatedly reclaim objects from the
	* cache when memory is low satisfy other memory allocations. Alternately,
	* setting this value to KMC_RECLAIM_ONCE limits how aggressively the cache
	* is reclaimed. This may increase the likelihood of out of memory events.
	*/
	static unsigned int spl_kmem_cache_reclaim = 0 /* KMC_RECLAIM_ONCE */;
	module_param(spl_kmem_cache_reclaim, uint, 0644);
	MODULE_PARM_DESC(spl_kmem_cache_reclaim, "Single reclaim pass (0x1)");

	static unsigned int spl_kmem_cache_obj_per_slab = SPL_KMEM_CACHE_OBJ_PER_SLAB;
	module_param(spl_kmem_cache_obj_per_slab, uint, 0644);
	MODULE_PARM_DESC(spl_kmem_cache_obj_per_slab, "Number of objects per slab");

	static unsigned int spl_kmem_cache_max_size = SPL_KMEM_CACHE_MAX_SIZE;
	module_param(spl_kmem_cache_max_size, uint, 0644);
	MODULE_PARM_DESC(spl_kmem_cache_max_size, "Maximum size of slab in MB");

	/*
	* For small objects the Linux slab allocator should be used to make the most
	* efficient use of the memory. However, large objects are not supported by
	* the Linux slab and therefore the SPL implementation is preferred. A cutoff
	* of 16K was determined to be optimal for architectures using 4K pages and
	* to also work well on architecutres using larger 64K page sizes.
	*/
	static unsigned int spl_kmem_cache_slab_limit = 16384;
	module_param(spl_kmem_cache_slab_limit, uint, 0644);
	MODULE_PARM_DESC(spl_kmem_cache_slab_limit,
	"Objects less than N bytes use the Linux slab");

	/*
	* The number of threads available to allocate new slabs for caches. This
	* should not need to be tuned but it is available for performance analysis.
	*/
	static unsigned int spl_kmem_cache_kmem_threads = 4;
	module_param(spl_kmem_cache_kmem_threads, uint, 0444);
	MODULE_PARM_DESC(spl_kmem_cache_kmem_threads,
	"Number of spl_kmem_cache threads");
	/* END CSTYLED */

	/*
	* Slab allocation interfaces
	*
	* While the Linux slab implementation was inspired by the Solaris
	* implementation I cannot use it to emulate the Solaris APIs. I
	* require two features which are not provided by the Linux slab.
	*
	* 1) Constructors AND destructors. Recent versions of the Linux
	* kernel have removed support for destructors. This is a deal
	* breaker for the SPL which contains particularly expensive
	* initializers for mutex's, condition variables, etc. We also
	* require a minimal level of cleanup for these data types unlike
	* many Linux data types which do need to be explicitly destroyed.
	*
	* 2) Virtual address space backed slab. Callers of the Solaris slab
	* expect it to work well for both small are very large allocations.
	* Because of memory fragmentation the Linux slab which is backed
	* by kmalloc'ed memory performs very badly when confronted with
	* large numbers of large allocations. Basing the slab on the
	* virtual address space removes the need for contiguous pages
	* and greatly improve performance for large allocations.
	*
	* For these reasons, the SPL has its own slab implementation with
	* the needed features. It is not as highly optimized as either the
	* Solaris or Linux slabs, but it should get me most of what is
	* needed until it can be optimized or obsoleted by another approach.
	*
	* One serious concern I do have about this method is the relatively
	* small virtual address space on 32bit arches. This will seriously
	* constrain the size of the slab caches and their performance.
	*/

	struct list_head spl_kmem_cache_list; /* List of caches */
	struct rw_semaphore spl_kmem_cache_sem; /* Cache list lock */
	taskq_t spl_kmem_cache_taskq; / Task queue for aging / reclaim */

	static void spl_cache_shrink(spl_kmem_cache_t skc, void obj);

	static void *
	kv_alloc(spl_kmem_cache_t *skc, int size, int flags)
	{
	gfp_t lflags = kmem_flags_convert(flags);
	void *ptr;

	ptr = spl_vmalloc(size, lflags \| __GFP_HIGHMEM);

	/* Resulting allocated memory will be page aligned */
	ASSERT(IS_P2ALIGNED(ptr, PAGE_SIZE));

	return (ptr);
	}

	static void
	kv_free(spl_kmem_cache_t skc, void ptr, int size)
	{
	ASSERT(IS_P2ALIGNED(ptr, PAGE_SIZE));

	/*
	* The Linux direct reclaim path uses this out of band value to
	* determine if forward progress is being made. Normally this is
	* incremented by kmem_freepages() which is part of the various
	* Linux slab implementations. However, since we are using none
	* of that infrastructure we are responsible for incrementing it.
	*/
	if (current->reclaim_state)
	current->reclaim_state->reclaimed_slab += size >> PAGE_SHIFT;

	vfree(ptr);
	}

	/*
	* Required space for each aligned sks.
	*/
	static inline uint32_t
	spl_sks_size(spl_kmem_cache_t *skc)
	{
	return (P2ROUNDUP_TYPED(sizeof (spl_kmem_slab_t),
	skc->skc_obj_align, uint32_t));
	}

	/*
	* Required space for each aligned object.
	*/
	static inline uint32_t
	spl_obj_size(spl_kmem_cache_t *skc)
	{
	uint32_t align = skc->skc_obj_align;

	return (P2ROUNDUP_TYPED(skc->skc_obj_size, align, uint32_t) +
	P2ROUNDUP_TYPED(sizeof (spl_kmem_obj_t), align, uint32_t));
	}

	uint64_t
	spl_kmem_cache_inuse(kmem_cache_t *cache)
	{
	return (cache->skc_obj_total);
	}
	EXPORT_SYMBOL(spl_kmem_cache_inuse);

	uint64_t
	spl_kmem_cache_entry_size(kmem_cache_t *cache)
	{
	return (cache->skc_obj_size);
	}
	EXPORT_SYMBOL(spl_kmem_cache_entry_size);

	/*
	* Lookup the spl_kmem_object_t for an object given that object.
	*/
	static inline spl_kmem_obj_t *
	spl_sko_from_obj(spl_kmem_cache_t skc, void obj)
	{
	return (obj + P2ROUNDUP_TYPED(skc->skc_obj_size,
	skc->skc_obj_align, uint32_t));
	}

	/*
	* It's important that we pack the spl_kmem_obj_t structure and the
	* actual objects in to one large address space to minimize the number
	* of calls to the allocator. It is far better to do a few large
	* allocations and then subdivide it ourselves. Now which allocator
	* we use requires balancing a few trade offs.
	*
	* For small objects we use kmem_alloc() because as long as you are
	* only requesting a small number of pages (ideally just one) its cheap.
	* However, when you start requesting multiple pages with kmem_alloc()
	* it gets increasingly expensive since it requires contiguous pages.
	* For this reason we shift to vmem_alloc() for slabs of large objects
	* which removes the need for contiguous pages. We do not use
	* vmem_alloc() in all cases because there is significant locking
	* overhead in __get_vm_area_node(). This function takes a single
	* global lock when acquiring an available virtual address range which
	* serializes all vmem_alloc()'s for all slab caches. Using slightly
	* different allocation functions for small and large objects should
	* give us the best of both worlds.
	*
	* +------------------------+
	* \| spl_kmem_slab_t --+-+ \|
	* \| skc_obj_size <-+ \| \|
	* \| spl_kmem_obj_t \| \|
	* \| skc_obj_size <---+ \|
	* \| spl_kmem_obj_t \| \|
	* \| ... v \|
	* +------------------------+
	*/
	static spl_kmem_slab_t *
	spl_slab_alloc(spl_kmem_cache_t *skc, int flags)
	{
	spl_kmem_slab_t *sks;
	void *base;
	uint32_t obj_size;

	base = kv_alloc(skc, skc->skc_slab_size, flags);
	if (base == NULL)
	return (NULL);

	sks = (spl_kmem_slab_t *)base;
	sks->sks_magic = SKS_MAGIC;
	sks->sks_objs = skc->skc_slab_objs;
	sks->sks_age = jiffies;
	sks->sks_cache = skc;
	INIT_LIST_HEAD(&sks->sks_list);
	INIT_LIST_HEAD(&sks->sks_free_list);
	sks->sks_ref = 0;
	obj_size = spl_obj_size(skc);

	for (int i = 0; i < sks->sks_objs; i++) {
	void obj = base + spl_sks_size(skc) + (i obj_size);

	ASSERT(IS_P2ALIGNED(obj, skc->skc_obj_align));
	spl_kmem_obj_t *sko = spl_sko_from_obj(skc, obj);
	sko->sko_addr = obj;
	sko->sko_magic = SKO_MAGIC;
	sko->sko_slab = sks;
	INIT_LIST_HEAD(&sko->sko_list);
	list_add_tail(&sko->sko_list, &sks->sks_free_list);
	}

	return (sks);
	}

	/*
	* Remove a slab from complete or partial list, it must be called with
	* the 'skc->skc_lock' held but the actual free must be performed
	* outside the lock to prevent deadlocking on vmem addresses.
	*/
	static void
	spl_slab_free(spl_kmem_slab_t *sks,
	struct list_head sks_list, struct list_head sko_list)
	{
	spl_kmem_cache_t *skc;

	ASSERT(sks->sks_magic == SKS_MAGIC);
	ASSERT(sks->sks_ref == 0);

	skc = sks->sks_cache;
	ASSERT(skc->skc_magic == SKC_MAGIC);

	/*
	* Update slab/objects counters in the cache, then remove the
	* slab from the skc->skc_partial_list. Finally add the slab
	* and all its objects in to the private work lists where the
	* destructors will be called and the memory freed to the system.
	*/
	skc->skc_obj_total -= sks->sks_objs;
	skc->skc_slab_total--;
	list_del(&sks->sks_list);
	list_add(&sks->sks_list, sks_list);
	list_splice_init(&sks->sks_free_list, sko_list);
	}

	/*
	* Reclaim empty slabs at the end of the partial list.
	*/
	static void
	spl_slab_reclaim(spl_kmem_cache_t *skc)
	{
	spl_kmem_slab_t sks = NULL, m = NULL;
	spl_kmem_obj_t sko = NULL, n = NULL;
	LIST_HEAD(sks_list);
	LIST_HEAD(sko_list);

	/*
	* Empty slabs and objects must be moved to a private list so they
	* can be safely freed outside the spin lock. All empty slabs are
	* at the end of skc->skc_partial_list, therefore once a non-empty
	* slab is found we can stop scanning.
	*/
	spin_lock(&skc->skc_lock);
	list_for_each_entry_safe_reverse(sks, m,
	&skc->skc_partial_list, sks_list) {

	if (sks->sks_ref > 0)
	break;

	spl_slab_free(sks, &sks_list, &sko_list);
	}
	spin_unlock(&skc->skc_lock);

	/*
	* The following two loops ensure all the object destructors are run,
	* and the slabs themselves are freed. This is all done outside the
	* skc->skc_lock since this allows the destructor to sleep, and
	* allows us to perform a conditional reschedule when a freeing a
	* large number of objects and slabs back to the system.
	*/

	list_for_each_entry_safe(sko, n, &sko_list, sko_list) {
	ASSERT(sko->sko_magic == SKO_MAGIC);
	}

	list_for_each_entry_safe(sks, m, &sks_list, sks_list) {
	ASSERT(sks->sks_magic == SKS_MAGIC);
	kv_free(skc, sks, skc->skc_slab_size);
	}
	}

	static spl_kmem_emergency_t *
	spl_emergency_search(struct rb_root root, void obj)
	{
	struct rb_node *node = root->rb_node;
	spl_kmem_emergency_t *ske;
	unsigned long address = (unsigned long)obj;

	while (node) {
	ske = container_of(node, spl_kmem_emergency_t, ske_node);

	if (address < ske->ske_obj)
	node = node->rb_left;
	else if (address > ske->ske_obj)
	node = node->rb_right;
	else
	return (ske);
	}

	return (NULL);
	}

	static int
	spl_emergency_insert(struct rb_root root, spl_kmem_emergency_t ske)
	{
	struct rb_node *new = &(root->rb_node), parent = NULL;
	spl_kmem_emergency_t *ske_tmp;
	unsigned long address = ske->ske_obj;

	while (*new) {
	ske_tmp = container_of(*new, spl_kmem_emergency_t, ske_node);

	parent = *new;
	if (address < ske_tmp->ske_obj)
	new = &((*new)->rb_left);
	else if (address > ske_tmp->ske_obj)
	new = &((*new)->rb_right);
	else
	return (0);
	}

	rb_link_node(&ske->ske_node, parent, new);
	rb_insert_color(&ske->ske_node, root);

	return (1);
	}

	/*
	* Allocate a single emergency object and track it in a red black tree.
	*/
	static int
	spl_emergency_alloc(spl_kmem_cache_t skc, int flags, void *obj)
	{
	gfp_t lflags = kmem_flags_convert(flags);
	spl_kmem_emergency_t *ske;
	int order = get_order(skc->skc_obj_size);
	int empty;

	/* Last chance use a partial slab if one now exists */
	spin_lock(&skc->skc_lock);
	empty = list_empty(&skc->skc_partial_list);
	spin_unlock(&skc->skc_lock);
	if (!empty)
	return (-EEXIST);

	ske = kmalloc(sizeof (*ske), lflags);
	if (ske == NULL)
	return (-ENOMEM);

	ske->ske_obj = __get_free_pages(lflags, order);
	if (ske->ske_obj == 0) {
	kfree(ske);
	return (-ENOMEM);
	}

	spin_lock(&skc->skc_lock);
	empty = spl_emergency_insert(&skc->skc_emergency_tree, ske);
	if (likely(empty)) {
	skc->skc_obj_total++;
	skc->skc_obj_emergency++;
	if (skc->skc_obj_emergency > skc->skc_obj_emergency_max)
	skc->skc_obj_emergency_max = skc->skc_obj_emergency;
	}
	spin_unlock(&skc->skc_lock);

	if (unlikely(!empty)) {
	free_pages(ske->ske_obj, order);
	kfree(ske);
	return (-EINVAL);
	}

	obj = (void )ske->ske_obj;

	return (0);
	}

	/*
	* Locate the passed object in the red black tree and free it.
	*/
	static int
	spl_emergency_free(spl_kmem_cache_t skc, void obj)
	{
	spl_kmem_emergency_t *ske;
	int order = get_order(skc->skc_obj_size);

	spin_lock(&skc->skc_lock);
	ske = spl_emergency_search(&skc->skc_emergency_tree, obj);
	if (ske) {
	rb_erase(&ske->ske_node, &skc->skc_emergency_tree);
	skc->skc_obj_emergency--;
	skc->skc_obj_total--;
	}
	spin_unlock(&skc->skc_lock);

	if (ske == NULL)
	return (-ENOENT);

	free_pages(ske->ske_obj, order);
	kfree(ske);

	return (0);
	}

	/*
	* Release objects from the per-cpu magazine back to their slab. The flush
	* argument contains the max number of entries to remove from the magazine.
	*/
	static void
	spl_cache_flush(spl_kmem_cache_t skc, spl_kmem_magazine_t skm, int flush)
	{
	spin_lock(&skc->skc_lock);

	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(skm->skm_magic == SKM_MAGIC);

	int count = MIN(flush, skm->skm_avail);
	for (int i = 0; i < count; i++)
	spl_cache_shrink(skc, skm->skm_objs[i]);

	skm->skm_avail -= count;
	memmove(skm->skm_objs, &(skm->skm_objs[count]),
	sizeof (void ) skm->skm_avail);

	spin_unlock(&skc->skc_lock);
	}

	/*
	* Size a slab based on the size of each aligned object plus spl_kmem_obj_t.
	* When on-slab we want to target spl_kmem_cache_obj_per_slab. However,
	* for very small objects we may end up with more than this so as not
	* to waste space in the minimal allocation of a single page.
	*/
	static int
	spl_slab_size(spl_kmem_cache_t skc, uint32_t objs, uint32_t *size)
	{
	uint32_t sks_size, obj_size, max_size, tgt_size, tgt_objs;

	sks_size = spl_sks_size(skc);
	obj_size = spl_obj_size(skc);
	max_size = (spl_kmem_cache_max_size * 1024 * 1024);
	tgt_size = (spl_kmem_cache_obj_per_slab * obj_size + sks_size);

	if (tgt_size <= max_size) {
	tgt_objs = (tgt_size - sks_size) / obj_size;
	} else {
	tgt_objs = (max_size - sks_size) / obj_size;
	tgt_size = (tgt_objs * obj_size) + sks_size;
	}

	if (tgt_objs == 0)
	return (-ENOSPC);

	*objs = tgt_objs;
	*size = tgt_size;

	return (0);
	}

	/*
	* Make a guess at reasonable per-cpu magazine size based on the size of
	* each object and the cost of caching N of them in each magazine. Long
	* term this should really adapt based on an observed usage heuristic.
	*/
	static int
	spl_magazine_size(spl_kmem_cache_t *skc)
	{
	uint32_t obj_size = spl_obj_size(skc);
	int size;

	if (spl_kmem_cache_magazine_size > 0)
	return (MAX(MIN(spl_kmem_cache_magazine_size, 256), 2));

	/* Per-magazine sizes below assume a 4Kib page size */
	if (obj_size > (PAGE_SIZE * 256))
	size = 4; /* Minimum 4Mib per-magazine */
	else if (obj_size > (PAGE_SIZE * 32))
	size = 16; /* Minimum 2Mib per-magazine */
	else if (obj_size > (PAGE_SIZE))
	size = 64; /* Minimum 256Kib per-magazine */
	else if (obj_size > (PAGE_SIZE / 4))
	size = 128; /* Minimum 128Kib per-magazine */
	else
	size = 256;

	return (size);
	}

	/*
	* Allocate a per-cpu magazine to associate with a specific core.
	*/
	static spl_kmem_magazine_t *
	spl_magazine_alloc(spl_kmem_cache_t *skc, int cpu)
	{
	spl_kmem_magazine_t *skm;
	int size = sizeof (spl_kmem_magazine_t) +
	sizeof (void ) skc->skc_mag_size;

	skm = kmalloc_node(size, GFP_KERNEL, cpu_to_node(cpu));
	if (skm) {
	skm->skm_magic = SKM_MAGIC;
	skm->skm_avail = 0;
	skm->skm_size = skc->skc_mag_size;
	skm->skm_refill = skc->skc_mag_refill;
	skm->skm_cache = skc;
	skm->skm_cpu = cpu;
	}

	return (skm);
	}

	/*
	* Free a per-cpu magazine associated with a specific core.
	*/
	static void
	spl_magazine_free(spl_kmem_magazine_t *skm)
	{
	ASSERT(skm->skm_magic == SKM_MAGIC);
	ASSERT(skm->skm_avail == 0);
	kfree(skm);
	}

	/*
	* Create all pre-cpu magazines of reasonable sizes.
	*/
	static int
	spl_magazine_create(spl_kmem_cache_t *skc)
	{
	int i = 0;

	ASSERT((skc->skc_flags & KMC_SLAB) == 0);

	skc->skc_mag = kzalloc(sizeof (spl_kmem_magazine_t )
	num_possible_cpus(), kmem_flags_convert(KM_SLEEP));
	skc->skc_mag_size = spl_magazine_size(skc);
	skc->skc_mag_refill = (skc->skc_mag_size + 1) / 2;

	for_each_possible_cpu(i) {
	skc->skc_mag[i] = spl_magazine_alloc(skc, i);
	if (!skc->skc_mag[i]) {
	for (i--; i >= 0; i--)
	spl_magazine_free(skc->skc_mag[i]);

	kfree(skc->skc_mag);
	return (-ENOMEM);
	}
	}

	return (0);
	}

	/*
	* Destroy all pre-cpu magazines.
	*/
	static void
	spl_magazine_destroy(spl_kmem_cache_t *skc)
	{
	spl_kmem_magazine_t *skm;
	int i = 0;

	ASSERT((skc->skc_flags & KMC_SLAB) == 0);

	for_each_possible_cpu(i) {
	skm = skc->skc_mag[i];
	spl_cache_flush(skc, skm, skm->skm_avail);
	spl_magazine_free(skm);
	}

	kfree(skc->skc_mag);
	}

	/*
	* Create a object cache based on the following arguments:
	* name cache name
	* size cache object size
	* align cache object alignment
	* ctor cache object constructor
	* dtor cache object destructor
	* reclaim cache object reclaim
	* priv cache private data for ctor/dtor/reclaim
	* vmp unused must be NULL
	* flags
	* KMC_KVMEM Force kvmem backed SPL cache
	* KMC_SLAB Force Linux slab backed cache
	* KMC_NODEBUG Disable debugging (unsupported)
	*/
	spl_kmem_cache_t *
	spl_kmem_cache_create(const char *name, size_t size, size_t align,
	spl_kmem_ctor_t ctor, spl_kmem_dtor_t dtor, void *reclaim,
	void priv, void vmp, int flags)
	{
	gfp_t lflags = kmem_flags_convert(KM_SLEEP);
	spl_kmem_cache_t *skc;
	int rc;

	/*
	* Unsupported flags
	*/
	ASSERT(vmp == NULL);
	ASSERT(reclaim == NULL);

	might_sleep();

	skc = kzalloc(sizeof (*skc), lflags);
	if (skc == NULL)
	return (NULL);

	skc->skc_magic = SKC_MAGIC;
	skc->skc_name_size = strlen(name) + 1;
	skc->skc_name = (char *)kmalloc(skc->skc_name_size, lflags);
	if (skc->skc_name == NULL) {
	kfree(skc);
	return (NULL);
	}
	- strncpy(skc->skc_name, name, skc->skc_name_size);
	+ strlcpy(skc->skc_name, name, skc->skc_name_size);

	skc->skc_ctor = ctor;
	skc->skc_dtor = dtor;
	skc->skc_private = priv;
	skc->skc_vmp = vmp;
	skc->skc_linux_cache = NULL;
	skc->skc_flags = flags;
	skc->skc_obj_size = size;
	skc->skc_obj_align = SPL_KMEM_CACHE_ALIGN;
	atomic_set(&skc->skc_ref, 0);

	INIT_LIST_HEAD(&skc->skc_list);
	INIT_LIST_HEAD(&skc->skc_complete_list);
	INIT_LIST_HEAD(&skc->skc_partial_list);
	skc->skc_emergency_tree = RB_ROOT;
	spin_lock_init(&skc->skc_lock);
	init_waitqueue_head(&skc->skc_waitq);
	skc->skc_slab_fail = 0;
	skc->skc_slab_create = 0;
	skc->skc_slab_destroy = 0;
	skc->skc_slab_total = 0;
	skc->skc_slab_alloc = 0;
	skc->skc_slab_max = 0;
	skc->skc_obj_total = 0;
	skc->skc_obj_alloc = 0;
	skc->skc_obj_max = 0;
	skc->skc_obj_deadlock = 0;
	skc->skc_obj_emergency = 0;
	skc->skc_obj_emergency_max = 0;

	rc = percpu_counter_init_common(&skc->skc_linux_alloc, 0,
	GFP_KERNEL);
	if (rc != 0) {
	kfree(skc);
	return (NULL);
	}

	/*
	* Verify the requested alignment restriction is sane.
	*/
	if (align) {
	VERIFY(ISP2(align));
	VERIFY3U(align, >=, SPL_KMEM_CACHE_ALIGN);
	VERIFY3U(align, <=, PAGE_SIZE);
	skc->skc_obj_align = align;
	}

	/*
	* When no specific type of slab is requested (kmem, vmem, or
	* linuxslab) then select a cache type based on the object size
	* and default tunables.
	*/
	if (!(skc->skc_flags & (KMC_SLAB \| KMC_KVMEM))) {
	if (spl_kmem_cache_slab_limit &&
	size <= (size_t)spl_kmem_cache_slab_limit) {
	/*
	* Objects smaller than spl_kmem_cache_slab_limit can
	* use the Linux slab for better space-efficiency.
	*/
	skc->skc_flags \|= KMC_SLAB;
	} else {
	/*
	* All other objects are considered large and are
	* placed on kvmem backed slabs.
	*/
	skc->skc_flags \|= KMC_KVMEM;
	}
	}

	/*
	* Given the type of slab allocate the required resources.
	*/
	if (skc->skc_flags & KMC_KVMEM) {
	rc = spl_slab_size(skc,
	&skc->skc_slab_objs, &skc->skc_slab_size);
	if (rc)
	goto out;

	rc = spl_magazine_create(skc);
	if (rc)
	goto out;
	} else {
	unsigned long slabflags = 0;

	if (size > (SPL_MAX_KMEM_ORDER_NR_PAGES * PAGE_SIZE)) {
	rc = EINVAL;
	goto out;
	}

	#if defined(SLAB_USERCOPY)
	/*
	* Required for PAX-enabled kernels if the slab is to be
	* used for copying between user and kernel space.
	*/
	slabflags \|= SLAB_USERCOPY;
	#endif

	#if defined(HAVE_KMEM_CACHE_CREATE_USERCOPY)
	/*
	* Newer grsec patchset uses kmem_cache_create_usercopy()
	* instead of SLAB_USERCOPY flag
	*/
	skc->skc_linux_cache = kmem_cache_create_usercopy(
	skc->skc_name, size, align, slabflags, 0, size, NULL);
	#else
	skc->skc_linux_cache = kmem_cache_create(
	skc->skc_name, size, align, slabflags, NULL);
	#endif
	if (skc->skc_linux_cache == NULL) {
	rc = ENOMEM;
	goto out;
	}
	}

	down_write(&spl_kmem_cache_sem);
	list_add_tail(&skc->skc_list, &spl_kmem_cache_list);
	up_write(&spl_kmem_cache_sem);

	return (skc);
	out:
	kfree(skc->skc_name);
	percpu_counter_destroy(&skc->skc_linux_alloc);
	kfree(skc);
	return (NULL);
	}
	EXPORT_SYMBOL(spl_kmem_cache_create);

	/*
	* Register a move callback for cache defragmentation.
	* XXX: Unimplemented but harmless to stub out for now.
	*/
	void
	spl_kmem_cache_set_move(spl_kmem_cache_t *skc,
	kmem_cbrc_t (move)(void , void , size_t, void *))
	{
	ASSERT(move != NULL);
	}
	EXPORT_SYMBOL(spl_kmem_cache_set_move);

	/*
	* Destroy a cache and all objects associated with the cache.
	*/
	void
	spl_kmem_cache_destroy(spl_kmem_cache_t *skc)
	{
	DECLARE_WAIT_QUEUE_HEAD(wq);
	taskqid_t id;

	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(skc->skc_flags & (KMC_KVMEM \| KMC_SLAB));

	down_write(&spl_kmem_cache_sem);
	list_del_init(&skc->skc_list);
	up_write(&spl_kmem_cache_sem);

	/* Cancel any and wait for any pending delayed tasks */
	VERIFY(!test_and_set_bit(KMC_BIT_DESTROY, &skc->skc_flags));

	spin_lock(&skc->skc_lock);
	id = skc->skc_taskqid;
	spin_unlock(&skc->skc_lock);

	taskq_cancel_id(spl_kmem_cache_taskq, id);

	/*
	* Wait until all current callers complete, this is mainly
	* to catch the case where a low memory situation triggers a
	* cache reaping action which races with this destroy.
	*/
	wait_event(wq, atomic_read(&skc->skc_ref) == 0);

	if (skc->skc_flags & KMC_KVMEM) {
	spl_magazine_destroy(skc);
	spl_slab_reclaim(skc);
	} else {
	ASSERT(skc->skc_flags & KMC_SLAB);
	kmem_cache_destroy(skc->skc_linux_cache);
	}

	spin_lock(&skc->skc_lock);

	/*
	* Validate there are no objects in use and free all the
	* spl_kmem_slab_t, spl_kmem_obj_t, and object buffers.
	*/
	ASSERT3U(skc->skc_slab_alloc, ==, 0);
	ASSERT3U(skc->skc_obj_alloc, ==, 0);
	ASSERT3U(skc->skc_slab_total, ==, 0);
	ASSERT3U(skc->skc_obj_total, ==, 0);
	ASSERT3U(skc->skc_obj_emergency, ==, 0);
	ASSERT(list_empty(&skc->skc_complete_list));

	ASSERT3U(percpu_counter_sum(&skc->skc_linux_alloc), ==, 0);
	percpu_counter_destroy(&skc->skc_linux_alloc);

	spin_unlock(&skc->skc_lock);

	kfree(skc->skc_name);
	kfree(skc);
	}
	EXPORT_SYMBOL(spl_kmem_cache_destroy);

	/*
	* Allocate an object from a slab attached to the cache. This is used to
	* repopulate the per-cpu magazine caches in batches when they run low.
	*/
	static void *
	spl_cache_obj(spl_kmem_cache_t skc, spl_kmem_slab_t sks)
	{
	spl_kmem_obj_t *sko;

	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(sks->sks_magic == SKS_MAGIC);

	sko = list_entry(sks->sks_free_list.next, spl_kmem_obj_t, sko_list);
	ASSERT(sko->sko_magic == SKO_MAGIC);
	ASSERT(sko->sko_addr != NULL);

	/* Remove from sks_free_list */
	list_del_init(&sko->sko_list);

	sks->sks_age = jiffies;
	sks->sks_ref++;
	skc->skc_obj_alloc++;

	/* Track max obj usage statistics */
	if (skc->skc_obj_alloc > skc->skc_obj_max)
	skc->skc_obj_max = skc->skc_obj_alloc;

	/* Track max slab usage statistics */
	if (sks->sks_ref == 1) {
	skc->skc_slab_alloc++;

	if (skc->skc_slab_alloc > skc->skc_slab_max)
	skc->skc_slab_max = skc->skc_slab_alloc;
	}

	return (sko->sko_addr);
	}

	/*
	* Generic slab allocation function to run by the global work queues.
	* It is responsible for allocating a new slab, linking it in to the list
	* of partial slabs, and then waking any waiters.
	*/
	static int
	__spl_cache_grow(spl_kmem_cache_t *skc, int flags)
	{
	spl_kmem_slab_t *sks;

	fstrans_cookie_t cookie = spl_fstrans_mark();
	sks = spl_slab_alloc(skc, flags);
	spl_fstrans_unmark(cookie);

	spin_lock(&skc->skc_lock);
	if (sks) {
	skc->skc_slab_total++;
	skc->skc_obj_total += sks->sks_objs;
	list_add_tail(&sks->sks_list, &skc->skc_partial_list);

	smp_mb__before_atomic();
	clear_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags);
	smp_mb__after_atomic();
	}
	spin_unlock(&skc->skc_lock);

	return (sks == NULL ? -ENOMEM : 0);
	}

	static void
	spl_cache_grow_work(void *data)
	{
	spl_kmem_alloc_t ska = (spl_kmem_alloc_t )data;
	spl_kmem_cache_t *skc = ska->ska_cache;

	int error = __spl_cache_grow(skc, ska->ska_flags);

	atomic_dec(&skc->skc_ref);
	smp_mb__before_atomic();
	clear_bit(KMC_BIT_GROWING, &skc->skc_flags);
	smp_mb__after_atomic();
	if (error == 0)
	wake_up_all(&skc->skc_waitq);

	kfree(ska);
	}

	/*
	* Returns non-zero when a new slab should be available.
	*/
	static int
	spl_cache_grow_wait(spl_kmem_cache_t *skc)
	{
	return (!test_bit(KMC_BIT_GROWING, &skc->skc_flags));
	}

	/*
	* No available objects on any slabs, create a new slab. Note that this
	* functionality is disabled for KMC_SLAB caches which are backed by the
	* Linux slab.
	*/
	static int
	spl_cache_grow(spl_kmem_cache_t skc, int flags, void *obj)
	{
	int remaining, rc = 0;

	ASSERT0(flags & ~KM_PUBLIC_MASK);
	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT((skc->skc_flags & KMC_SLAB) == 0);
	might_sleep();
	*obj = NULL;

	/*
	* Before allocating a new slab wait for any reaping to complete and
	* then return so the local magazine can be rechecked for new objects.
	*/
	if (test_bit(KMC_BIT_REAPING, &skc->skc_flags)) {
	rc = spl_wait_on_bit(&skc->skc_flags, KMC_BIT_REAPING,
	TASK_UNINTERRUPTIBLE);
	return (rc ? rc : -EAGAIN);
	}

	/*
	* Note: It would be nice to reduce the overhead of context switch
	* and improve NUMA locality, by trying to allocate a new slab in the
	* current process context with KM_NOSLEEP flag.
	*
	* However, this can't be applied to vmem/kvmem due to a bug that
	* spl_vmalloc() doesn't honor gfp flags in page table allocation.
	*/

	/*
	* This is handled by dispatching a work request to the global work
	* queue. This allows us to asynchronously allocate a new slab while
	* retaining the ability to safely fall back to a smaller synchronous
	* allocations to ensure forward progress is always maintained.
	*/
	if (test_and_set_bit(KMC_BIT_GROWING, &skc->skc_flags) == 0) {
	spl_kmem_alloc_t *ska;

	ska = kmalloc(sizeof (*ska), kmem_flags_convert(flags));
	if (ska == NULL) {
	clear_bit_unlock(KMC_BIT_GROWING, &skc->skc_flags);
	smp_mb__after_atomic();
	wake_up_all(&skc->skc_waitq);
	return (-ENOMEM);
	}

	atomic_inc(&skc->skc_ref);
	ska->ska_cache = skc;
	ska->ska_flags = flags;
	taskq_init_ent(&ska->ska_tqe);
	taskq_dispatch_ent(spl_kmem_cache_taskq,
	spl_cache_grow_work, ska, 0, &ska->ska_tqe);
	}

	/*
	* The goal here is to only detect the rare case where a virtual slab
	* allocation has deadlocked. We must be careful to minimize the use
	* of emergency objects which are more expensive to track. Therefore,
	* we set a very long timeout for the asynchronous allocation and if
	* the timeout is reached the cache is flagged as deadlocked. From
	* this point only new emergency objects will be allocated until the
	* asynchronous allocation completes and clears the deadlocked flag.
	*/
	if (test_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags)) {
	rc = spl_emergency_alloc(skc, flags, obj);
	} else {
	remaining = wait_event_timeout(skc->skc_waitq,
	spl_cache_grow_wait(skc), HZ / 10);

	if (!remaining) {
	spin_lock(&skc->skc_lock);
	if (test_bit(KMC_BIT_GROWING, &skc->skc_flags)) {
	set_bit(KMC_BIT_DEADLOCKED, &skc->skc_flags);
	skc->skc_obj_deadlock++;
	}
	spin_unlock(&skc->skc_lock);
	}

	rc = -ENOMEM;
	}

	return (rc);
	}

	/*
	* Refill a per-cpu magazine with objects from the slabs for this cache.
	* Ideally the magazine can be repopulated using existing objects which have
	* been released, however if we are unable to locate enough free objects new
	* slabs of objects will be created. On success NULL is returned, otherwise
	* the address of a single emergency object is returned for use by the caller.
	*/
	static void *
	spl_cache_refill(spl_kmem_cache_t skc, spl_kmem_magazine_t skm, int flags)
	{
	spl_kmem_slab_t *sks;
	int count = 0, rc, refill;
	void *obj = NULL;

	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(skm->skm_magic == SKM_MAGIC);

	refill = MIN(skm->skm_refill, skm->skm_size - skm->skm_avail);
	spin_lock(&skc->skc_lock);

	while (refill > 0) {
	/* No slabs available we may need to grow the cache */
	if (list_empty(&skc->skc_partial_list)) {
	spin_unlock(&skc->skc_lock);

	local_irq_enable();
	rc = spl_cache_grow(skc, flags, &obj);
	local_irq_disable();

	/* Emergency object for immediate use by caller */
	if (rc == 0 && obj != NULL)
	return (obj);

	if (rc)
	goto out;

	/* Rescheduled to different CPU skm is not local */
	if (skm != skc->skc_mag[smp_processor_id()])
	goto out;

	/*
	* Potentially rescheduled to the same CPU but
	* allocations may have occurred from this CPU while
	* we were sleeping so recalculate max refill.
	*/
	refill = MIN(refill, skm->skm_size - skm->skm_avail);

	spin_lock(&skc->skc_lock);
	continue;
	}

	/* Grab the next available slab */
	sks = list_entry((&skc->skc_partial_list)->next,
	spl_kmem_slab_t, sks_list);
	ASSERT(sks->sks_magic == SKS_MAGIC);
	ASSERT(sks->sks_ref < sks->sks_objs);
	ASSERT(!list_empty(&sks->sks_free_list));

	/*
	* Consume as many objects as needed to refill the requested
	* cache. We must also be careful not to overfill it.
	*/
	while (sks->sks_ref < sks->sks_objs && refill-- > 0 &&
	++count) {
	ASSERT(skm->skm_avail < skm->skm_size);
	ASSERT(count < skm->skm_size);
	skm->skm_objs[skm->skm_avail++] =
	spl_cache_obj(skc, sks);
	}

	/* Move slab to skc_complete_list when full */
	if (sks->sks_ref == sks->sks_objs) {
	list_del(&sks->sks_list);
	list_add(&sks->sks_list, &skc->skc_complete_list);
	}
	}

	spin_unlock(&skc->skc_lock);
	out:
	return (NULL);
	}

	/*
	* Release an object back to the slab from which it came.
	*/
	static void
	spl_cache_shrink(spl_kmem_cache_t skc, void obj)
	{
	spl_kmem_slab_t *sks = NULL;
	spl_kmem_obj_t *sko = NULL;

	ASSERT(skc->skc_magic == SKC_MAGIC);

	sko = spl_sko_from_obj(skc, obj);
	ASSERT(sko->sko_magic == SKO_MAGIC);
	sks = sko->sko_slab;
	ASSERT(sks->sks_magic == SKS_MAGIC);
	ASSERT(sks->sks_cache == skc);
	list_add(&sko->sko_list, &sks->sks_free_list);

	sks->sks_age = jiffies;
	sks->sks_ref--;
	skc->skc_obj_alloc--;

	/*
	* Move slab to skc_partial_list when no longer full. Slabs
	* are added to the head to keep the partial list is quasi-full
	* sorted order. Fuller at the head, emptier at the tail.
	*/
	if (sks->sks_ref == (sks->sks_objs - 1)) {
	list_del(&sks->sks_list);
	list_add(&sks->sks_list, &skc->skc_partial_list);
	}

	/*
	* Move empty slabs to the end of the partial list so
	* they can be easily found and freed during reclamation.
	*/
	if (sks->sks_ref == 0) {
	list_del(&sks->sks_list);
	list_add_tail(&sks->sks_list, &skc->skc_partial_list);
	skc->skc_slab_alloc--;
	}
	}

	/*
	* Allocate an object from the per-cpu magazine, or if the magazine
	* is empty directly allocate from a slab and repopulate the magazine.
	*/
	void *
	spl_kmem_cache_alloc(spl_kmem_cache_t *skc, int flags)
	{
	spl_kmem_magazine_t *skm;
	void *obj = NULL;

	ASSERT0(flags & ~KM_PUBLIC_MASK);
	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));

	/*
	* Allocate directly from a Linux slab. All optimizations are left
	* to the underlying cache we only need to guarantee that KM_SLEEP
	* callers will never fail.
	*/
	if (skc->skc_flags & KMC_SLAB) {
	struct kmem_cache *slc = skc->skc_linux_cache;
	do {
	obj = kmem_cache_alloc(slc, kmem_flags_convert(flags));
	} while ((obj == NULL) && !(flags & KM_NOSLEEP));

	if (obj != NULL) {
	/*
	* Even though we leave everything up to the
	* underlying cache we still keep track of
	* how many objects we've allocated in it for
	* better debuggability.
	*/
	percpu_counter_inc(&skc->skc_linux_alloc);
	}
	goto ret;
	}

	local_irq_disable();

	restart:
	/*
	* Safe to update per-cpu structure without lock, but
	* in the restart case we must be careful to reacquire
	* the local magazine since this may have changed
	* when we need to grow the cache.
	*/
	skm = skc->skc_mag[smp_processor_id()];
	ASSERT(skm->skm_magic == SKM_MAGIC);

	if (likely(skm->skm_avail)) {
	/* Object available in CPU cache, use it */
	obj = skm->skm_objs[--skm->skm_avail];
	} else {
	obj = spl_cache_refill(skc, skm, flags);
	if ((obj == NULL) && !(flags & KM_NOSLEEP))
	goto restart;

	local_irq_enable();
	goto ret;
	}

	local_irq_enable();
	ASSERT(obj);
	ASSERT(IS_P2ALIGNED(obj, skc->skc_obj_align));

	ret:
	/* Pre-emptively migrate object to CPU L1 cache */
	if (obj) {
	if (obj && skc->skc_ctor)
	skc->skc_ctor(obj, skc->skc_private, flags);
	else
	prefetchw(obj);
	}

	return (obj);
	}
	EXPORT_SYMBOL(spl_kmem_cache_alloc);

	/*
	* Free an object back to the local per-cpu magazine, there is no
	* guarantee that this is the same magazine the object was originally
	* allocated from. We may need to flush entire from the magazine
	* back to the slabs to make space.
	*/
	void
	spl_kmem_cache_free(spl_kmem_cache_t skc, void obj)
	{
	spl_kmem_magazine_t *skm;
	unsigned long flags;
	int do_reclaim = 0;
	int do_emergency = 0;

	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));

	/*
	* Run the destructor
	*/
	if (skc->skc_dtor)
	skc->skc_dtor(obj, skc->skc_private);

	/*
	* Free the object from the Linux underlying Linux slab.
	*/
	if (skc->skc_flags & KMC_SLAB) {
	kmem_cache_free(skc->skc_linux_cache, obj);
	percpu_counter_dec(&skc->skc_linux_alloc);
	return;
	}

	/*
	* While a cache has outstanding emergency objects all freed objects
	* must be checked. However, since emergency objects will never use
	* a virtual address these objects can be safely excluded as an
	* optimization.
	*/
	if (!is_vmalloc_addr(obj)) {
	spin_lock(&skc->skc_lock);
	do_emergency = (skc->skc_obj_emergency > 0);
	spin_unlock(&skc->skc_lock);

	if (do_emergency && (spl_emergency_free(skc, obj) == 0))
	return;
	}

	local_irq_save(flags);

	/*
	* Safe to update per-cpu structure without lock, but
	* no remote memory allocation tracking is being performed
	* it is entirely possible to allocate an object from one
	* CPU cache and return it to another.
	*/
	skm = skc->skc_mag[smp_processor_id()];
	ASSERT(skm->skm_magic == SKM_MAGIC);

	/*
	* Per-CPU cache full, flush it to make space for this object,
	* this may result in an empty slab which can be reclaimed once
	* interrupts are re-enabled.
	*/
	if (unlikely(skm->skm_avail >= skm->skm_size)) {
	spl_cache_flush(skc, skm, skm->skm_refill);
	do_reclaim = 1;
	}

	/* Available space in cache, use it */
	skm->skm_objs[skm->skm_avail++] = obj;

	local_irq_restore(flags);

	if (do_reclaim)
	spl_slab_reclaim(skc);
	}
	EXPORT_SYMBOL(spl_kmem_cache_free);

	/*
	* Depending on how many and which objects are released it may simply
	* repopulate the local magazine which will then need to age-out. Objects
	* which cannot fit in the magazine will be released back to their slabs
	* which will also need to age out before being released. This is all just
	* best effort and we do not want to thrash creating and destroying slabs.
	*/
	void
	spl_kmem_cache_reap_now(spl_kmem_cache_t *skc)
	{
	ASSERT(skc->skc_magic == SKC_MAGIC);
	ASSERT(!test_bit(KMC_BIT_DESTROY, &skc->skc_flags));

	if (skc->skc_flags & KMC_SLAB)
	return;

	atomic_inc(&skc->skc_ref);

	/*
	* Prevent concurrent cache reaping when contended.
	*/
	if (test_and_set_bit(KMC_BIT_REAPING, &skc->skc_flags))
	goto out;

	/* Reclaim from the magazine and free all now empty slabs. */
	unsigned long irq_flags;
	local_irq_save(irq_flags);
	spl_kmem_magazine_t *skm = skc->skc_mag[smp_processor_id()];
	spl_cache_flush(skc, skm, skm->skm_avail);
	local_irq_restore(irq_flags);

	spl_slab_reclaim(skc);
	clear_bit_unlock(KMC_BIT_REAPING, &skc->skc_flags);
	smp_mb__after_atomic();
	wake_up_bit(&skc->skc_flags, KMC_BIT_REAPING);
	out:
	atomic_dec(&skc->skc_ref);
	}
	EXPORT_SYMBOL(spl_kmem_cache_reap_now);

	/*
	* This is stubbed out for code consistency with other platforms. There
	* is existing logic to prevent concurrent reaping so while this is ugly
	* it should do no harm.
	*/
	int
	spl_kmem_cache_reap_active(void)
	{
	return (0);
	}
	EXPORT_SYMBOL(spl_kmem_cache_reap_active);

	/*
	* Reap all free slabs from all registered caches.
	*/
	void
	spl_kmem_reap(void)
	{
	spl_kmem_cache_t *skc = NULL;

	down_read(&spl_kmem_cache_sem);
	list_for_each_entry(skc, &spl_kmem_cache_list, skc_list) {
	spl_kmem_cache_reap_now(skc);
	}
	up_read(&spl_kmem_cache_sem);
	}
	EXPORT_SYMBOL(spl_kmem_reap);

	int
	spl_kmem_cache_init(void)
	{
	init_rwsem(&spl_kmem_cache_sem);
	INIT_LIST_HEAD(&spl_kmem_cache_list);
	spl_kmem_cache_taskq = taskq_create("spl_kmem_cache",
	spl_kmem_cache_kmem_threads, maxclsyspri,
	spl_kmem_cache_kmem_threads * 8, INT_MAX,
	TASKQ_PREPOPULATE \| TASKQ_DYNAMIC);

	if (spl_kmem_cache_taskq == NULL)
	return (-ENOMEM);

	return (0);
	}

	void
	spl_kmem_cache_fini(void)
	{
	taskq_destroy(spl_kmem_cache_taskq);
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c b/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c
	index c6d3c8f4413f..4308581147a9 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-kstat.c
	@@ -1,715 +1,715 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Kstat Implementation.
	*
	* Links to Illumos.org for more information on kstat function:
	* [1] https://illumos.org/man/1M/kstat
	* [2] https://illumos.org/man/9f/kstat_create
	*/

	#include <linux/seq_file.h>
	#include <sys/kstat.h>
	#include <sys/vmem.h>
	#include <sys/cmn_err.h>
	#include <sys/sysmacros.h>

	static kmutex_t kstat_module_lock;
	static struct list_head kstat_module_list;
	static kid_t kstat_id;

	static int
	kstat_resize_raw(kstat_t *ksp)
	{
	if (ksp->ks_raw_bufsize == KSTAT_RAW_MAX)
	return (ENOMEM);

	vmem_free(ksp->ks_raw_buf, ksp->ks_raw_bufsize);
	ksp->ks_raw_bufsize = MIN(ksp->ks_raw_bufsize * 2, KSTAT_RAW_MAX);
	ksp->ks_raw_buf = vmem_alloc(ksp->ks_raw_bufsize, KM_SLEEP);

	return (0);
	}

	static int
	kstat_seq_show_headers(struct seq_file *f)
	{
	kstat_t ksp = (kstat_t )f->private;
	int rc = 0;

	ASSERT(ksp->ks_magic == KS_MAGIC);

	seq_printf(f, "%d %d 0x%02x %d %d %lld %lld\n",
	ksp->ks_kid, ksp->ks_type, ksp->ks_flags,
	ksp->ks_ndata, (int)ksp->ks_data_size,
	ksp->ks_crtime, ksp->ks_snaptime);

	switch (ksp->ks_type) {
	case KSTAT_TYPE_RAW:
	restart:
	if (ksp->ks_raw_ops.headers) {
	rc = ksp->ks_raw_ops.headers(
	ksp->ks_raw_buf, ksp->ks_raw_bufsize);
	if (rc == ENOMEM && !kstat_resize_raw(ksp))
	goto restart;
	if (!rc)
	seq_puts(f, ksp->ks_raw_buf);
	} else {
	seq_printf(f, "raw data\n");
	}
	break;
	case KSTAT_TYPE_NAMED:
	seq_printf(f, "%-31s %-4s %s\n",
	"name", "type", "data");
	break;
	case KSTAT_TYPE_INTR:
	seq_printf(f, "%-8s %-8s %-8s %-8s %-8s\n",
	"hard", "soft", "watchdog",
	"spurious", "multsvc");
	break;
	case KSTAT_TYPE_IO:
	seq_printf(f,
	"%-8s %-8s %-8s %-8s %-8s %-8s "
	"%-8s %-8s %-8s %-8s %-8s %-8s\n",
	"nread", "nwritten", "reads", "writes",
	"wtime", "wlentime", "wupdate",
	"rtime", "rlentime", "rupdate",
	"wcnt", "rcnt");
	break;
	case KSTAT_TYPE_TIMER:
	seq_printf(f,
	"%-31s %-8s "
	"%-8s %-8s %-8s %-8s %-8s\n",
	"name", "events", "elapsed",
	"min", "max", "start", "stop");
	break;
	default:
	PANIC("Undefined kstat type %d\n", ksp->ks_type);
	}

	return (-rc);
	}

	static int
	kstat_seq_show_raw(struct seq_file f, unsigned char p, int l)
	{
	int i, j;

	for (i = 0; ; i++) {
	seq_printf(f, "%03x:", i);

	for (j = 0; j < 16; j++) {
	if (i * 16 + j >= l) {
	seq_printf(f, "\n");
	goto out;
	}

	seq_printf(f, " %02x", (unsigned char)p[i * 16 + j]);
	}
	seq_printf(f, "\n");
	}
	out:
	return (0);
	}

	static int
	kstat_seq_show_named(struct seq_file f, kstat_named_t knp)
	{
	seq_printf(f, "%-31s %-4d ", knp->name, knp->data_type);

	switch (knp->data_type) {
	case KSTAT_DATA_CHAR:
	knp->value.c[15] = '\0'; /* NULL terminate */
	seq_printf(f, "%-16s", knp->value.c);
	break;
	/*
	* NOTE - We need to be more careful able what tokens are
	* used for each arch, for now this is correct for x86_64.
	*/
	case KSTAT_DATA_INT32:
	seq_printf(f, "%d", knp->value.i32);
	break;
	case KSTAT_DATA_UINT32:
	seq_printf(f, "%u", knp->value.ui32);
	break;
	case KSTAT_DATA_INT64:
	seq_printf(f, "%lld", (signed long long)knp->value.i64);
	break;
	case KSTAT_DATA_UINT64:
	seq_printf(f, "%llu",
	(unsigned long long)knp->value.ui64);
	break;
	case KSTAT_DATA_LONG:
	seq_printf(f, "%ld", knp->value.l);
	break;
	case KSTAT_DATA_ULONG:
	seq_printf(f, "%lu", knp->value.ul);
	break;
	case KSTAT_DATA_STRING:
	KSTAT_NAMED_STR_PTR(knp)
	[KSTAT_NAMED_STR_BUFLEN(knp)-1] = '\0';
	seq_printf(f, "%s", KSTAT_NAMED_STR_PTR(knp));
	break;
	default:
	PANIC("Undefined kstat data type %d\n", knp->data_type);
	}

	seq_printf(f, "\n");

	return (0);
	}

	static int
	kstat_seq_show_intr(struct seq_file f, kstat_intr_t kip)
	{
	seq_printf(f, "%-8u %-8u %-8u %-8u %-8u\n",
	kip->intrs[KSTAT_INTR_HARD],
	kip->intrs[KSTAT_INTR_SOFT],
	kip->intrs[KSTAT_INTR_WATCHDOG],
	kip->intrs[KSTAT_INTR_SPURIOUS],
	kip->intrs[KSTAT_INTR_MULTSVC]);

	return (0);
	}

	static int
	kstat_seq_show_io(struct seq_file f, kstat_io_t kip)
	{
	/* though wlentime & friends are signed, they will never be negative */
	seq_printf(f,
	"%-8llu %-8llu %-8u %-8u %-8llu %-8llu "
	"%-8llu %-8llu %-8llu %-8llu %-8u %-8u\n",
	kip->nread, kip->nwritten,
	kip->reads, kip->writes,
	kip->wtime, kip->wlentime, kip->wlastupdate,
	kip->rtime, kip->rlentime, kip->rlastupdate,
	kip->wcnt, kip->rcnt);

	return (0);
	}

	static int
	kstat_seq_show_timer(struct seq_file f, kstat_timer_t ktp)
	{
	seq_printf(f,
	"%-31s %-8llu %-8llu %-8llu %-8llu %-8llu %-8llu\n",
	ktp->name, ktp->num_events, ktp->elapsed_time,
	ktp->min_time, ktp->max_time,
	ktp->start_time, ktp->stop_time);

	return (0);
	}

	static int
	kstat_seq_show(struct seq_file f, void p)
	{
	kstat_t ksp = (kstat_t )f->private;
	int rc = 0;

	ASSERT(ksp->ks_magic == KS_MAGIC);

	switch (ksp->ks_type) {
	case KSTAT_TYPE_RAW:
	restart:
	if (ksp->ks_raw_ops.data) {
	rc = ksp->ks_raw_ops.data(
	ksp->ks_raw_buf, ksp->ks_raw_bufsize, p);
	if (rc == ENOMEM && !kstat_resize_raw(ksp))
	goto restart;
	if (!rc)
	seq_puts(f, ksp->ks_raw_buf);
	} else {
	ASSERT(ksp->ks_ndata == 1);
	rc = kstat_seq_show_raw(f, ksp->ks_data,
	ksp->ks_data_size);
	}
	break;
	case KSTAT_TYPE_NAMED:
	rc = kstat_seq_show_named(f, (kstat_named_t *)p);
	break;
	case KSTAT_TYPE_INTR:
	rc = kstat_seq_show_intr(f, (kstat_intr_t *)p);
	break;
	case KSTAT_TYPE_IO:
	rc = kstat_seq_show_io(f, (kstat_io_t *)p);
	break;
	case KSTAT_TYPE_TIMER:
	rc = kstat_seq_show_timer(f, (kstat_timer_t *)p);
	break;
	default:
	PANIC("Undefined kstat type %d\n", ksp->ks_type);
	}

	return (-rc);
	}

	static int
	kstat_default_update(kstat_t *ksp, int rw)
	{
	ASSERT(ksp != NULL);

	if (rw == KSTAT_WRITE)
	return (EACCES);

	return (0);
	}

	static void *
	kstat_seq_data_addr(kstat_t *ksp, loff_t n)
	{
	void *rc = NULL;

	switch (ksp->ks_type) {
	case KSTAT_TYPE_RAW:
	if (ksp->ks_raw_ops.addr)
	rc = ksp->ks_raw_ops.addr(ksp, n);
	else
	rc = ksp->ks_data;
	break;
	case KSTAT_TYPE_NAMED:
	rc = ksp->ks_data + n * sizeof (kstat_named_t);
	break;
	case KSTAT_TYPE_INTR:
	rc = ksp->ks_data + n * sizeof (kstat_intr_t);
	break;
	case KSTAT_TYPE_IO:
	rc = ksp->ks_data + n * sizeof (kstat_io_t);
	break;
	case KSTAT_TYPE_TIMER:
	rc = ksp->ks_data + n * sizeof (kstat_timer_t);
	break;
	default:
	PANIC("Undefined kstat type %d\n", ksp->ks_type);
	}

	return (rc);
	}

	static void *
	kstat_seq_start(struct seq_file f, loff_t pos)
	{
	loff_t n = *pos;
	kstat_t ksp = (kstat_t )f->private;
	ASSERT(ksp->ks_magic == KS_MAGIC);

	mutex_enter(ksp->ks_lock);

	if (ksp->ks_type == KSTAT_TYPE_RAW) {
	ksp->ks_raw_bufsize = PAGE_SIZE;
	ksp->ks_raw_buf = vmem_alloc(ksp->ks_raw_bufsize, KM_SLEEP);
	}

	/* Dynamically update kstat, on error existing kstats are used */
	(void) ksp->ks_update(ksp, KSTAT_READ);

	ksp->ks_snaptime = gethrtime();

	if (!(ksp->ks_flags & KSTAT_FLAG_NO_HEADERS) && !n &&
	kstat_seq_show_headers(f))
	return (NULL);

	if (n >= ksp->ks_ndata)
	return (NULL);

	return (kstat_seq_data_addr(ksp, n));
	}

	static void *
	kstat_seq_next(struct seq_file f, void p, loff_t *pos)
	{
	kstat_t ksp = (kstat_t )f->private;
	ASSERT(ksp->ks_magic == KS_MAGIC);

	++*pos;
	if (*pos >= ksp->ks_ndata)
	return (NULL);

	return (kstat_seq_data_addr(ksp, *pos));
	}

	static void
	kstat_seq_stop(struct seq_file f, void v)
	{
	kstat_t ksp = (kstat_t )f->private;
	ASSERT(ksp->ks_magic == KS_MAGIC);

	if (ksp->ks_type == KSTAT_TYPE_RAW)
	vmem_free(ksp->ks_raw_buf, ksp->ks_raw_bufsize);

	mutex_exit(ksp->ks_lock);
	}

	static const struct seq_operations kstat_seq_ops = {
	.show = kstat_seq_show,
	.start = kstat_seq_start,
	.next = kstat_seq_next,
	.stop = kstat_seq_stop,
	};

	static kstat_module_t *
	kstat_find_module(char *name)
	{
	kstat_module_t *module = NULL;

	list_for_each_entry(module, &kstat_module_list, ksm_module_list) {
	if (strncmp(name, module->ksm_name, KSTAT_STRLEN) == 0)
	return (module);
	}

	return (NULL);
	}

	static kstat_module_t *
	kstat_create_module(char *name)
	{
	kstat_module_t *module;
	struct proc_dir_entry *pde;

	pde = proc_mkdir(name, proc_spl_kstat);
	if (pde == NULL)
	return (NULL);

	module = kmem_alloc(sizeof (kstat_module_t), KM_SLEEP);
	module->ksm_proc = pde;
	- strlcpy(module->ksm_name, name, KSTAT_STRLEN+1);
	+ strlcpy(module->ksm_name, name, KSTAT_STRLEN);
	INIT_LIST_HEAD(&module->ksm_kstat_list);
	list_add_tail(&module->ksm_module_list, &kstat_module_list);

	return (module);

	}

	static void
	kstat_delete_module(kstat_module_t *module)
	{
	ASSERT(list_empty(&module->ksm_kstat_list));
	remove_proc_entry(module->ksm_name, proc_spl_kstat);
	list_del(&module->ksm_module_list);
	kmem_free(module, sizeof (kstat_module_t));
	}

	static int
	proc_kstat_open(struct inode inode, struct file filp)
	{
	struct seq_file *f;
	int rc;

	rc = seq_open(filp, &kstat_seq_ops);
	if (rc)
	return (rc);

	f = filp->private_data;
	f->private = SPL_PDE_DATA(inode);

	return (0);
	}

	static ssize_t
	proc_kstat_write(struct file filp, const char __user buf, size_t len,
	loff_t *ppos)
	{
	struct seq_file *f = filp->private_data;
	kstat_t *ksp = f->private;
	int rc;

	ASSERT(ksp->ks_magic == KS_MAGIC);

	mutex_enter(ksp->ks_lock);
	rc = ksp->ks_update(ksp, KSTAT_WRITE);
	mutex_exit(ksp->ks_lock);

	if (rc)
	return (-rc);

	*ppos += len;
	return (len);
	}

	static const kstat_proc_op_t proc_kstat_operations = {
	#ifdef HAVE_PROC_OPS_STRUCT
	.proc_open = proc_kstat_open,
	.proc_write = proc_kstat_write,
	.proc_read = seq_read,
	.proc_lseek = seq_lseek,
	.proc_release = seq_release,
	#else
	.open = proc_kstat_open,
	.write = proc_kstat_write,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = seq_release,
	#endif
	};

	void
	__kstat_set_raw_ops(kstat_t *ksp,
	int (headers)(char buf, size_t size),
	int (data)(char buf, size_t size, void *data),
	void (addr)(kstat_t *ksp, loff_t index))
	{
	ksp->ks_raw_ops.headers = headers;
	ksp->ks_raw_ops.data = data;
	ksp->ks_raw_ops.addr = addr;
	}
	EXPORT_SYMBOL(__kstat_set_raw_ops);

	void
	kstat_proc_entry_init(kstat_proc_entry_t kpep, const char module,
	const char *name)
	{
	kpep->kpe_owner = NULL;
	kpep->kpe_proc = NULL;
	INIT_LIST_HEAD(&kpep->kpe_list);
	- strncpy(kpep->kpe_module, module, KSTAT_STRLEN);
	- strncpy(kpep->kpe_name, name, KSTAT_STRLEN);
	+ strlcpy(kpep->kpe_module, module, sizeof (kpep->kpe_module));
	+ strlcpy(kpep->kpe_name, name, sizeof (kpep->kpe_name));
	}
	EXPORT_SYMBOL(kstat_proc_entry_init);

	kstat_t *
	__kstat_create(const char ks_module, int ks_instance, const char ks_name,
	const char *ks_class, uchar_t ks_type, uint_t ks_ndata,
	uchar_t ks_flags)
	{
	kstat_t *ksp;

	ASSERT(ks_module);
	ASSERT(ks_instance == 0);
	ASSERT(ks_name);

	if ((ks_type == KSTAT_TYPE_INTR) \|\| (ks_type == KSTAT_TYPE_IO))
	ASSERT(ks_ndata == 1);

	ksp = kmem_zalloc(sizeof (*ksp), KM_SLEEP);
	if (ksp == NULL)
	return (ksp);

	mutex_enter(&kstat_module_lock);
	ksp->ks_kid = kstat_id;
	kstat_id++;
	mutex_exit(&kstat_module_lock);

	ksp->ks_magic = KS_MAGIC;
	mutex_init(&ksp->ks_private_lock, NULL, MUTEX_DEFAULT, NULL);
	ksp->ks_lock = &ksp->ks_private_lock;

	ksp->ks_crtime = gethrtime();
	ksp->ks_snaptime = ksp->ks_crtime;
	ksp->ks_instance = ks_instance;
	- strncpy(ksp->ks_class, ks_class, KSTAT_STRLEN);
	+ strlcpy(ksp->ks_class, ks_class, sizeof (ksp->ks_class));
	ksp->ks_type = ks_type;
	ksp->ks_flags = ks_flags;
	ksp->ks_update = kstat_default_update;
	ksp->ks_private = NULL;
	ksp->ks_raw_ops.headers = NULL;
	ksp->ks_raw_ops.data = NULL;
	ksp->ks_raw_ops.addr = NULL;
	ksp->ks_raw_buf = NULL;
	ksp->ks_raw_bufsize = 0;
	kstat_proc_entry_init(&ksp->ks_proc, ks_module, ks_name);

	switch (ksp->ks_type) {
	case KSTAT_TYPE_RAW:
	ksp->ks_ndata = 1;
	ksp->ks_data_size = ks_ndata;
	break;
	case KSTAT_TYPE_NAMED:
	ksp->ks_ndata = ks_ndata;
	ksp->ks_data_size = ks_ndata * sizeof (kstat_named_t);
	break;
	case KSTAT_TYPE_INTR:
	ksp->ks_ndata = ks_ndata;
	ksp->ks_data_size = ks_ndata * sizeof (kstat_intr_t);
	break;
	case KSTAT_TYPE_IO:
	ksp->ks_ndata = ks_ndata;
	ksp->ks_data_size = ks_ndata * sizeof (kstat_io_t);
	break;
	case KSTAT_TYPE_TIMER:
	ksp->ks_ndata = ks_ndata;
	ksp->ks_data_size = ks_ndata * sizeof (kstat_timer_t);
	break;
	default:
	PANIC("Undefined kstat type %d\n", ksp->ks_type);
	}

	if (ksp->ks_flags & KSTAT_FLAG_VIRTUAL) {
	ksp->ks_data = NULL;
	} else {
	ksp->ks_data = kmem_zalloc(ksp->ks_data_size, KM_SLEEP);
	if (ksp->ks_data == NULL) {
	kmem_free(ksp, sizeof (*ksp));
	ksp = NULL;
	}
	}

	return (ksp);
	}
	EXPORT_SYMBOL(__kstat_create);

	static int
	kstat_detect_collision(kstat_proc_entry_t *kpep)
	{
	kstat_module_t *module;
	kstat_proc_entry_t *tmp = NULL;
	char *parent;
	char *cp;

	parent = kmem_asprintf("%s", kpep->kpe_module);

	if ((cp = strrchr(parent, '/')) == NULL) {
	kmem_strfree(parent);
	return (0);
	}

	cp[0] = '\0';
	if ((module = kstat_find_module(parent)) != NULL) {
	list_for_each_entry(tmp, &module->ksm_kstat_list, kpe_list) {
	if (strncmp(tmp->kpe_name, cp+1, KSTAT_STRLEN) == 0) {
	kmem_strfree(parent);
	return (EEXIST);
	}
	}
	}

	kmem_strfree(parent);
	return (0);
	}

	/*
	* Add a file to the proc filesystem under the kstat namespace (i.e.
	* /proc/spl/kstat/). The file need not necessarily be implemented as a
	* kstat.
	*/
	void
	kstat_proc_entry_install(kstat_proc_entry_t *kpep, mode_t mode,
	const kstat_proc_op_t proc_ops, void data)
	{
	kstat_module_t *module;
	kstat_proc_entry_t *tmp = NULL;

	ASSERT(kpep);

	mutex_enter(&kstat_module_lock);

	module = kstat_find_module(kpep->kpe_module);
	if (module == NULL) {
	if (kstat_detect_collision(kpep) != 0) {
	cmn_err(CE_WARN, "kstat_create('%s', '%s'): namespace" \
	" collision", kpep->kpe_module, kpep->kpe_name);
	goto out;
	}
	module = kstat_create_module(kpep->kpe_module);
	if (module == NULL)
	goto out;
	}

	/*
	* Only one entry by this name per-module, on failure the module
	* shouldn't be deleted because we know it has at least one entry.
	*/
	list_for_each_entry(tmp, &module->ksm_kstat_list, kpe_list) {
	if (strncmp(tmp->kpe_name, kpep->kpe_name, KSTAT_STRLEN) == 0)
	goto out;
	}

	list_add_tail(&kpep->kpe_list, &module->ksm_kstat_list);

	kpep->kpe_owner = module;
	kpep->kpe_proc = proc_create_data(kpep->kpe_name, mode,
	module->ksm_proc, proc_ops, data);
	if (kpep->kpe_proc == NULL) {
	list_del_init(&kpep->kpe_list);
	if (list_empty(&module->ksm_kstat_list))
	kstat_delete_module(module);
	}
	out:
	mutex_exit(&kstat_module_lock);

	}
	EXPORT_SYMBOL(kstat_proc_entry_install);

	void
	__kstat_install(kstat_t *ksp)
	{
	ASSERT(ksp);
	mode_t mode;
	/* Specify permission modes for different kstats */
	if (strncmp(ksp->ks_proc.kpe_name, "dbufs", KSTAT_STRLEN) == 0) {
	mode = 0600;
	} else {
	mode = 0644;
	}
	kstat_proc_entry_install(
	&ksp->ks_proc, mode, &proc_kstat_operations, ksp);
	}
	EXPORT_SYMBOL(__kstat_install);

	void
	kstat_proc_entry_delete(kstat_proc_entry_t *kpep)
	{
	kstat_module_t *module = kpep->kpe_owner;
	if (kpep->kpe_proc)
	remove_proc_entry(kpep->kpe_name, module->ksm_proc);

	mutex_enter(&kstat_module_lock);
	list_del_init(&kpep->kpe_list);

	/*
	* Remove top level module directory if it wasn't empty before, but now
	* is.
	*/
	if (kpep->kpe_proc && list_empty(&module->ksm_kstat_list))
	kstat_delete_module(module);
	mutex_exit(&kstat_module_lock);

	}
	EXPORT_SYMBOL(kstat_proc_entry_delete);

	void
	__kstat_delete(kstat_t *ksp)
	{
	kstat_proc_entry_delete(&ksp->ks_proc);

	if (!(ksp->ks_flags & KSTAT_FLAG_VIRTUAL))
	kmem_free(ksp->ks_data, ksp->ks_data_size);

	ksp->ks_lock = NULL;
	mutex_destroy(&ksp->ks_private_lock);
	kmem_free(ksp, sizeof (*ksp));
	}
	EXPORT_SYMBOL(__kstat_delete);

	int
	spl_kstat_init(void)
	{
	mutex_init(&kstat_module_lock, NULL, MUTEX_DEFAULT, NULL);
	INIT_LIST_HEAD(&kstat_module_list);
	kstat_id = 0;
	return (0);
	}

	void
	spl_kstat_fini(void)
	{
	ASSERT(list_empty(&kstat_module_list));
	mutex_destroy(&kstat_module_lock);
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c b/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	index 3b0c29606c2e..abf4dca585b2 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	@@ -1,1433 +1,1435 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Task Queue Implementation.
	*/

	#include <sys/timer.h>
	#include <sys/taskq.h>
	#include <sys/kmem.h>
	#include <sys/tsd.h>
	#include <sys/trace_spl.h>
	#ifdef HAVE_CPU_HOTPLUG
	#include <linux/cpuhotplug.h>
	#endif

	static int spl_taskq_thread_bind = 0;
	module_param(spl_taskq_thread_bind, int, 0644);
	MODULE_PARM_DESC(spl_taskq_thread_bind, "Bind taskq thread to CPU by default");


	static int spl_taskq_thread_dynamic = 1;
	module_param(spl_taskq_thread_dynamic, int, 0444);
	MODULE_PARM_DESC(spl_taskq_thread_dynamic, "Allow dynamic taskq threads");

	static int spl_taskq_thread_priority = 1;
	module_param(spl_taskq_thread_priority, int, 0644);
	MODULE_PARM_DESC(spl_taskq_thread_priority,
	"Allow non-default priority for taskq threads");

	-static int spl_taskq_thread_sequential = 4;
	-module_param(spl_taskq_thread_sequential, int, 0644);
	+static uint_t spl_taskq_thread_sequential = 4;
	+/* BEGIN CSTYLED */
	+module_param(spl_taskq_thread_sequential, uint, 0644);
	+/* END CSTYLED */
	MODULE_PARM_DESC(spl_taskq_thread_sequential,
	"Create new taskq threads after N sequential tasks");

	/*
	* Global system-wide dynamic task queue available for all consumers. This
	* taskq is not intended for long-running tasks; instead, a dedicated taskq
	* should be created.
	*/
	taskq_t *system_taskq;
	EXPORT_SYMBOL(system_taskq);
	/* Global dynamic task queue for long delay */
	taskq_t *system_delay_taskq;
	EXPORT_SYMBOL(system_delay_taskq);

	/* Private dedicated taskq for creating new taskq threads on demand. */
	static taskq_t *dynamic_taskq;
	static taskq_thread_t taskq_thread_create(taskq_t );

	#ifdef HAVE_CPU_HOTPLUG
	/* Multi-callback id for cpu hotplugging. */
	static int spl_taskq_cpuhp_state;
	#endif

	/* List of all taskqs */
	LIST_HEAD(tq_list);
	struct rw_semaphore tq_list_sem;
	static uint_t taskq_tsd;

	static int
	task_km_flags(uint_t flags)
	{
	if (flags & TQ_NOSLEEP)
	return (KM_NOSLEEP);

	if (flags & TQ_PUSHPAGE)
	return (KM_PUSHPAGE);

	return (KM_SLEEP);
	}

	/*
	* taskq_find_by_name - Find the largest instance number of a named taskq.
	*/
	static int
	taskq_find_by_name(const char *name)
	{
	struct list_head *tql = NULL;
	taskq_t *tq;

	list_for_each_prev(tql, &tq_list) {
	tq = list_entry(tql, taskq_t, tq_taskqs);
	if (strcmp(name, tq->tq_name) == 0)
	return (tq->tq_instance);
	}
	return (-1);
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, returns a list_t which
	* is not attached to the free, work, or pending taskq lists.
	*/
	static taskq_ent_t *
	task_alloc(taskq_t tq, uint_t flags, unsigned long irqflags)
	{
	taskq_ent_t *t;
	int count = 0;

	ASSERT(tq);
	retry:
	/* Acquire taskq_ent_t's from free list if available */
	if (!list_empty(&tq->tq_free_list) && !(flags & TQ_NEW)) {
	t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
	ASSERT(!(t->tqent_flags & TQENT_FLAG_CANCEL));
	ASSERT(!timer_pending(&t->tqent_timer));

	list_del_init(&t->tqent_list);
	return (t);
	}

	/* Free list is empty and memory allocations are prohibited */
	if (flags & TQ_NOALLOC)
	return (NULL);

	/* Hit maximum taskq_ent_t pool size */
	if (tq->tq_nalloc >= tq->tq_maxalloc) {
	if (flags & TQ_NOSLEEP)
	return (NULL);

	/*
	* Sleep periodically polling the free list for an available
	* taskq_ent_t. Dispatching with TQ_SLEEP should always succeed
	* but we cannot block forever waiting for an taskq_ent_t to
	* show up in the free list, otherwise a deadlock can happen.
	*
	* Therefore, we need to allocate a new task even if the number
	* of allocated tasks is above tq->tq_maxalloc, but we still
	* end up delaying the task allocation by one second, thereby
	* throttling the task dispatch rate.
	*/
	spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
	schedule_timeout(HZ / 100);
	spin_lock_irqsave_nested(&tq->tq_lock, *irqflags,
	tq->tq_lock_class);
	if (count < 100) {
	count++;
	goto retry;
	}
	}

	spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
	t = kmem_alloc(sizeof (taskq_ent_t), task_km_flags(flags));
	spin_lock_irqsave_nested(&tq->tq_lock, *irqflags, tq->tq_lock_class);

	if (t) {
	taskq_init_ent(t);
	tq->tq_nalloc++;
	}

	return (t);
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, expects the taskq_ent_t
	* to already be removed from the free, work, or pending taskq lists.
	*/
	static void
	task_free(taskq_t tq, taskq_ent_t t)
	{
	ASSERT(tq);
	ASSERT(t);
	ASSERT(list_empty(&t->tqent_list));
	ASSERT(!timer_pending(&t->tqent_timer));

	kmem_free(t, sizeof (taskq_ent_t));
	tq->tq_nalloc--;
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, either destroys the
	* taskq_ent_t if too many exist or moves it to the free list for later use.
	*/
	static void
	task_done(taskq_t tq, taskq_ent_t t)
	{
	ASSERT(tq);
	ASSERT(t);

	/* Wake tasks blocked in taskq_wait_id() */
	wake_up_all(&t->tqent_waitq);

	list_del_init(&t->tqent_list);

	if (tq->tq_nalloc <= tq->tq_minalloc) {
	t->tqent_id = TASKQID_INVALID;
	t->tqent_func = NULL;
	t->tqent_arg = NULL;
	t->tqent_flags = 0;

	list_add_tail(&t->tqent_list, &tq->tq_free_list);
	} else {
	task_free(tq, t);
	}
	}

	/*
	* When a delayed task timer expires remove it from the delay list and
	* add it to the priority list in order for immediate processing.
	*/
	static void
	task_expire_impl(taskq_ent_t *t)
	{
	taskq_ent_t *w;
	taskq_t *tq = t->tqent_taskq;
	struct list_head *l = NULL;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (t->tqent_flags & TQENT_FLAG_CANCEL) {
	ASSERT(list_empty(&t->tqent_list));
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return;
	}

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	/*
	* The priority list must be maintained in strict task id order
	* from lowest to highest for lowest_id to be easily calculable.
	*/
	list_del(&t->tqent_list);
	list_for_each_prev(l, &tq->tq_prio_list) {
	w = list_entry(l, taskq_ent_t, tqent_list);
	if (w->tqent_id < t->tqent_id) {
	list_add(&t->tqent_list, l);
	break;
	}
	}
	if (l == &tq->tq_prio_list)
	list_add(&t->tqent_list, &tq->tq_prio_list);

	spin_unlock_irqrestore(&tq->tq_lock, flags);

	wake_up(&tq->tq_work_waitq);
	}

	static void
	task_expire(spl_timer_list_t tl)
	{
	struct timer_list tmr = (struct timer_list )tl;
	taskq_ent_t *t = from_timer(t, tmr, tqent_timer);
	task_expire_impl(t);
	}

	/*
	* Returns the lowest incomplete taskqid_t. The taskqid_t may
	* be queued on the pending list, on the priority list, on the
	* delay list, or on the work list currently being handled, but
	* it is not 100% complete yet.
	*/
	static taskqid_t
	taskq_lowest_id(taskq_t *tq)
	{
	taskqid_t lowest_id = tq->tq_next_id;
	taskq_ent_t *t;
	taskq_thread_t *tqt;

	if (!list_empty(&tq->tq_pend_list)) {
	t = list_entry(tq->tq_pend_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_prio_list)) {
	t = list_entry(tq->tq_prio_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_delay_list)) {
	t = list_entry(tq->tq_delay_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_active_list)) {
	tqt = list_entry(tq->tq_active_list.next, taskq_thread_t,
	tqt_active_list);
	ASSERT(tqt->tqt_id != TASKQID_INVALID);
	lowest_id = MIN(lowest_id, tqt->tqt_id);
	}

	return (lowest_id);
	}

	/*
	* Insert a task into a list keeping the list sorted by increasing taskqid.
	*/
	static void
	taskq_insert_in_order(taskq_t tq, taskq_thread_t tqt)
	{
	taskq_thread_t *w;
	struct list_head *l = NULL;

	ASSERT(tq);
	ASSERT(tqt);

	list_for_each_prev(l, &tq->tq_active_list) {
	w = list_entry(l, taskq_thread_t, tqt_active_list);
	if (w->tqt_id < tqt->tqt_id) {
	list_add(&tqt->tqt_active_list, l);
	break;
	}
	}
	if (l == &tq->tq_active_list)
	list_add(&tqt->tqt_active_list, &tq->tq_active_list);
	}

	/*
	* Find and return a task from the given list if it exists. The list
	* must be in lowest to highest task id order.
	*/
	static taskq_ent_t *
	taskq_find_list(taskq_t tq, struct list_head lh, taskqid_t id)
	{
	struct list_head *l = NULL;
	taskq_ent_t *t;

	list_for_each(l, lh) {
	t = list_entry(l, taskq_ent_t, tqent_list);

	if (t->tqent_id == id)
	return (t);

	if (t->tqent_id > id)
	break;
	}

	return (NULL);
	}

	/*
	* Find an already dispatched task given the task id regardless of what
	* state it is in. If a task is still pending it will be returned.
	* If a task is executing, then -EBUSY will be returned instead.
	* If the task has already been run then NULL is returned.
	*/
	static taskq_ent_t *
	taskq_find(taskq_t *tq, taskqid_t id)
	{
	taskq_thread_t *tqt;
	struct list_head *l = NULL;
	taskq_ent_t *t;

	t = taskq_find_list(tq, &tq->tq_delay_list, id);
	if (t)
	return (t);

	t = taskq_find_list(tq, &tq->tq_prio_list, id);
	if (t)
	return (t);

	t = taskq_find_list(tq, &tq->tq_pend_list, id);
	if (t)
	return (t);

	list_for_each(l, &tq->tq_active_list) {
	tqt = list_entry(l, taskq_thread_t, tqt_active_list);
	if (tqt->tqt_id == id) {
	/*
	* Instead of returning tqt_task, we just return a non
	* NULL value to prevent misuse, since tqt_task only
	* has two valid fields.
	*/
	return (ERR_PTR(-EBUSY));
	}
	}

	return (NULL);
	}

	/*
	* Theory for the taskq_wait_id(), taskq_wait_outstanding(), and
	* taskq_wait() functions below.
	*
	* Taskq waiting is accomplished by tracking the lowest outstanding task
	* id and the next available task id. As tasks are dispatched they are
	* added to the tail of the pending, priority, or delay lists. As worker
	* threads become available the tasks are removed from the heads of these
	* lists and linked to the worker threads. This ensures the lists are
	* kept sorted by lowest to highest task id.
	*
	* Therefore the lowest outstanding task id can be quickly determined by
	* checking the head item from all of these lists. This value is stored
	* with the taskq as the lowest id. It only needs to be recalculated when
	* either the task with the current lowest id completes or is canceled.
	*
	* By blocking until the lowest task id exceeds the passed task id the
	* taskq_wait_outstanding() function can be easily implemented. Similarly,
	* by blocking until the lowest task id matches the next task id taskq_wait()
	* can be implemented.
	*
	* Callers should be aware that when there are multiple worked threads it
	* is possible for larger task ids to complete before smaller ones. Also
	* when the taskq contains delay tasks with small task ids callers may
	* block for a considerable length of time waiting for them to expire and
	* execute.
	*/
	static int
	taskq_wait_id_check(taskq_t *tq, taskqid_t id)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (taskq_find(tq, id) == NULL);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait_id() function blocks until the passed task id completes.
	* This does not guarantee that all lower task ids have completed.
	*/
	void
	taskq_wait_id(taskq_t *tq, taskqid_t id)
	{
	wait_event(tq->tq_wait_waitq, taskq_wait_id_check(tq, id));
	}
	EXPORT_SYMBOL(taskq_wait_id);

	static int
	taskq_wait_outstanding_check(taskq_t *tq, taskqid_t id)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (id < tq->tq_lowest_id);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait_outstanding() function will block until all tasks with a
	* lower taskqid than the passed 'id' have been completed. Note that all
	* task id's are assigned monotonically at dispatch time. Zero may be
	* passed for the id to indicate all tasks dispatch up to this point,
	* but not after, should be waited for.
	*/
	void
	taskq_wait_outstanding(taskq_t *tq, taskqid_t id)
	{
	id = id ? id : tq->tq_next_id - 1;
	wait_event(tq->tq_wait_waitq, taskq_wait_outstanding_check(tq, id));
	}
	EXPORT_SYMBOL(taskq_wait_outstanding);

	static int
	taskq_wait_check(taskq_t *tq)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (tq->tq_lowest_id == tq->tq_next_id);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait() function will block until the taskq is empty.
	* This means that if a taskq re-dispatches work to itself taskq_wait()
	* callers will block indefinitely.
	*/
	void
	taskq_wait(taskq_t *tq)
	{
	wait_event(tq->tq_wait_waitq, taskq_wait_check(tq));
	}
	EXPORT_SYMBOL(taskq_wait);

	int
	taskq_member(taskq_t tq, kthread_t t)
	{
	return (tq == (taskq_t *)tsd_get_by_thread(taskq_tsd, t));
	}
	EXPORT_SYMBOL(taskq_member);

	taskq_t *
	taskq_of_curthread(void)
	{
	return (tsd_get(taskq_tsd));
	}
	EXPORT_SYMBOL(taskq_of_curthread);

	/*
	* Cancel an already dispatched task given the task id. Still pending tasks
	* will be immediately canceled, and if the task is active the function will
	* block until it completes. Preallocated tasks which are canceled must be
	* freed by the caller.
	*/
	int
	taskq_cancel_id(taskq_t *tq, taskqid_t id)
	{
	taskq_ent_t *t;
	int rc = ENOENT;
	unsigned long flags;

	ASSERT(tq);

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	t = taskq_find(tq, id);
	if (t && t != ERR_PTR(-EBUSY)) {
	list_del_init(&t->tqent_list);
	t->tqent_flags \|= TQENT_FLAG_CANCEL;

	/*
	* When canceling the lowest outstanding task id we
	* must recalculate the new lowest outstanding id.
	*/
	if (tq->tq_lowest_id == t->tqent_id) {
	tq->tq_lowest_id = taskq_lowest_id(tq);
	ASSERT3S(tq->tq_lowest_id, >, t->tqent_id);
	}

	/*
	* The task_expire() function takes the tq->tq_lock so drop
	* drop the lock before synchronously cancelling the timer.
	*/
	if (timer_pending(&t->tqent_timer)) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	del_timer_sync(&t->tqent_timer);
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	if (!(t->tqent_flags & TQENT_FLAG_PREALLOC))
	task_done(tq, t);

	rc = 0;
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	if (t == ERR_PTR(-EBUSY)) {
	taskq_wait_id(tq, id);
	rc = EBUSY;
	}

	return (rc);
	}
	EXPORT_SYMBOL(taskq_cancel_id);

	static int taskq_thread_spawn(taskq_t *tq);

	taskqid_t
	taskq_dispatch(taskq_t tq, task_func_t func, void arg, uint_t flags)
	{
	taskq_ent_t *t;
	taskqid_t rc = TASKQID_INVALID;
	unsigned long irqflags;

	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	/* Do not queue the task unless there is idle thread for it */
	ASSERT(tq->tq_nactive <= tq->tq_nthreads);
	if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
	/* Dynamic taskq may be able to spawn another thread */
	if (!(tq->tq_flags & TASKQ_DYNAMIC) \|\|
	taskq_thread_spawn(tq) == 0)
	goto out;
	}

	if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
	goto out;

	spin_lock(&t->tqent_lock);

	/* Queue to the front of the list to enforce TQ_NOQUEUE semantics */
	if (flags & TQ_NOQUEUE)
	list_add(&t->tqent_list, &tq->tq_prio_list);
	/* Queue to the priority list instead of the pending list */
	else if (flags & TQ_FRONT)
	list_add_tail(&t->tqent_list, &tq->tq_prio_list);
	else
	list_add_tail(&t->tqent_list, &tq->tq_pend_list);

	t->tqent_id = rc = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;
	t->tqent_timer.function = NULL;
	t->tqent_timer.expires = 0;

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	spin_unlock(&t->tqent_lock);

	wake_up(&tq->tq_work_waitq);
	out:
	/* Spawn additional taskq threads if required. */
	if (!(flags & TQ_NOQUEUE) && tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);

	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	return (rc);
	}
	EXPORT_SYMBOL(taskq_dispatch);

	taskqid_t
	taskq_dispatch_delay(taskq_t tq, task_func_t func, void arg,
	uint_t flags, clock_t expire_time)
	{
	taskqid_t rc = TASKQID_INVALID;
	taskq_ent_t *t;
	unsigned long irqflags;

	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
	goto out;

	spin_lock(&t->tqent_lock);

	/* Queue to the delay list for subsequent execution */
	list_add_tail(&t->tqent_list, &tq->tq_delay_list);

	t->tqent_id = rc = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;
	t->tqent_timer.function = task_expire;
	t->tqent_timer.expires = (unsigned long)expire_time;
	add_timer(&t->tqent_timer);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	spin_unlock(&t->tqent_lock);
	out:
	/* Spawn additional taskq threads if required. */
	if (tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);
	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	return (rc);
	}
	EXPORT_SYMBOL(taskq_dispatch_delay);

	void
	taskq_dispatch_ent(taskq_t tq, task_func_t func, void arg, uint_t flags,
	taskq_ent_t *t)
	{
	unsigned long irqflags;
	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
	tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE)) {
	t->tqent_id = TASKQID_INVALID;
	goto out;
	}

	if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
	/* Dynamic taskq may be able to spawn another thread */
	if (!(tq->tq_flags & TASKQ_DYNAMIC) \|\|
	taskq_thread_spawn(tq) == 0)
	goto out2;
	flags \|= TQ_FRONT;
	}

	spin_lock(&t->tqent_lock);

	/*
	* Make sure the entry is not on some other taskq; it is important to
	* ASSERT() under lock
	*/
	ASSERT(taskq_empty_ent(t));

	/*
	* Mark it as a prealloc'd task. This is important
	* to ensure that we don't free it later.
	*/
	t->tqent_flags \|= TQENT_FLAG_PREALLOC;

	/* Queue to the priority list instead of the pending list */
	if (flags & TQ_FRONT)
	list_add_tail(&t->tqent_list, &tq->tq_prio_list);
	else
	list_add_tail(&t->tqent_list, &tq->tq_pend_list);

	t->tqent_id = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	spin_unlock(&t->tqent_lock);

	wake_up(&tq->tq_work_waitq);
	out:
	/* Spawn additional taskq threads if required. */
	if (tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);
	out2:
	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	}
	EXPORT_SYMBOL(taskq_dispatch_ent);

	int
	taskq_empty_ent(taskq_ent_t *t)
	{
	return (list_empty(&t->tqent_list));
	}
	EXPORT_SYMBOL(taskq_empty_ent);

	void
	taskq_init_ent(taskq_ent_t *t)
	{
	spin_lock_init(&t->tqent_lock);
	init_waitqueue_head(&t->tqent_waitq);
	timer_setup(&t->tqent_timer, NULL, 0);
	INIT_LIST_HEAD(&t->tqent_list);
	t->tqent_id = 0;
	t->tqent_func = NULL;
	t->tqent_arg = NULL;
	t->tqent_flags = 0;
	t->tqent_taskq = NULL;
	}
	EXPORT_SYMBOL(taskq_init_ent);

	/*
	* Return the next pending task, preference is given to tasks on the
	* priority list which were dispatched with TQ_FRONT.
	*/
	static taskq_ent_t *
	taskq_next_ent(taskq_t *tq)
	{
	struct list_head *list;

	if (!list_empty(&tq->tq_prio_list))
	list = &tq->tq_prio_list;
	else if (!list_empty(&tq->tq_pend_list))
	list = &tq->tq_pend_list;
	else
	return (NULL);

	return (list_entry(list->next, taskq_ent_t, tqent_list));
	}

	/*
	* Spawns a new thread for the specified taskq.
	*/
	static void
	taskq_thread_spawn_task(void *arg)
	{
	taskq_t tq = (taskq_t )arg;
	unsigned long flags;

	if (taskq_thread_create(tq) == NULL) {
	/* restore spawning count if failed */
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	tq->tq_nspawn--;
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	}
	}

	/*
	* Spawn addition threads for dynamic taskqs (TASKQ_DYNAMIC) the current
	* number of threads is insufficient to handle the pending tasks. These
	* new threads must be created by the dedicated dynamic_taskq to avoid
	* deadlocks between thread creation and memory reclaim. The system_taskq
	* which is also a dynamic taskq cannot be safely used for this.
	*/
	static int
	taskq_thread_spawn(taskq_t *tq)
	{
	int spawning = 0;

	if (!(tq->tq_flags & TASKQ_DYNAMIC))
	return (0);

	if ((tq->tq_nthreads + tq->tq_nspawn < tq->tq_maxthreads) &&
	(tq->tq_flags & TASKQ_ACTIVE)) {
	spawning = (++tq->tq_nspawn);
	taskq_dispatch(dynamic_taskq, taskq_thread_spawn_task,
	tq, TQ_NOSLEEP);
	}

	return (spawning);
	}

	/*
	* Threads in a dynamic taskq should only exit once it has been completely
	* drained and no other threads are actively servicing tasks. This prevents
	* threads from being created and destroyed more than is required.
	*
	* The first thread is the thread list is treated as the primary thread.
	* There is nothing special about the primary thread but in order to avoid
	* all the taskq pids from changing we opt to make it long running.
	*/
	static int
	taskq_thread_should_stop(taskq_t tq, taskq_thread_t tqt)
	{
	if (!(tq->tq_flags & TASKQ_DYNAMIC))
	return (0);

	if (list_first_entry(&(tq->tq_thread_list), taskq_thread_t,
	tqt_thread_list) == tqt)
	return (0);

	return
	((tq->tq_nspawn == 0) && /* No threads are being spawned */
	(tq->tq_nactive == 0) && /* No threads are handling tasks */
	(tq->tq_nthreads > 1) && /* More than 1 thread is running */
	(!taskq_next_ent(tq)) && /* There are no pending tasks */
	(spl_taskq_thread_dynamic)); /* Dynamic taskqs are allowed */
	}

	static int
	taskq_thread(void *args)
	{
	DECLARE_WAITQUEUE(wait, current);
	sigset_t blocked;
	taskq_thread_t *tqt = args;
	taskq_t *tq;
	taskq_ent_t *t;
	int seq_tasks = 0;
	unsigned long flags;
	taskq_ent_t dup_task = {};

	ASSERT(tqt);
	ASSERT(tqt->tqt_tq);
	tq = tqt->tqt_tq;
	current->flags \|= PF_NOFREEZE;

	(void) spl_fstrans_mark();

	sigfillset(&blocked);
	sigprocmask(SIG_BLOCK, &blocked, NULL);
	flush_signals(current);

	tsd_set(taskq_tsd, tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	/*
	* If we are dynamically spawned, decrease spawning count. Note that
	* we could be created during taskq_create, in which case we shouldn't
	* do the decrement. But it's fine because taskq_create will reset
	* tq_nspawn later.
	*/
	if (tq->tq_flags & TASKQ_DYNAMIC)
	tq->tq_nspawn--;

	/* Immediately exit if more threads than allowed were created. */
	if (tq->tq_nthreads >= tq->tq_maxthreads)
	goto error;

	tq->tq_nthreads++;
	list_add_tail(&tqt->tqt_thread_list, &tq->tq_thread_list);
	wake_up(&tq->tq_wait_waitq);
	set_current_state(TASK_INTERRUPTIBLE);

	while (!kthread_should_stop()) {

	if (list_empty(&tq->tq_pend_list) &&
	list_empty(&tq->tq_prio_list)) {

	if (taskq_thread_should_stop(tq, tqt)) {
	wake_up_all(&tq->tq_wait_waitq);
	break;
	}

	add_wait_queue_exclusive(&tq->tq_work_waitq, &wait);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	schedule();
	seq_tasks = 0;

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	remove_wait_queue(&tq->tq_work_waitq, &wait);
	} else {
	__set_current_state(TASK_RUNNING);
	}

	if ((t = taskq_next_ent(tq)) != NULL) {
	list_del_init(&t->tqent_list);

	/*
	* A TQENT_FLAG_PREALLOC task may be reused or freed
	* during the task function call. Store tqent_id and
	* tqent_flags here.
	*
	* Also use an on stack taskq_ent_t for tqt_task
	* assignment in this case; we want to make sure
	* to duplicate all fields, so the values are
	* correct when it's accessed via DTRACE_PROBE*.
	*/
	tqt->tqt_id = t->tqent_id;
	tqt->tqt_flags = t->tqent_flags;

	if (t->tqent_flags & TQENT_FLAG_PREALLOC) {
	dup_task = *t;
	t = &dup_task;
	}
	tqt->tqt_task = t;

	taskq_insert_in_order(tq, tqt);
	tq->tq_nactive++;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	DTRACE_PROBE1(taskq_ent__start, taskq_ent_t *, t);

	/* Perform the requested task */
	t->tqent_func(t->tqent_arg);

	DTRACE_PROBE1(taskq_ent__finish, taskq_ent_t *, t);

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	tq->tq_nactive--;
	list_del_init(&tqt->tqt_active_list);
	tqt->tqt_task = NULL;

	/* For prealloc'd tasks, we don't free anything. */
	if (!(tqt->tqt_flags & TQENT_FLAG_PREALLOC))
	task_done(tq, t);

	/*
	* When the current lowest outstanding taskqid is
	* done calculate the new lowest outstanding id
	*/
	if (tq->tq_lowest_id == tqt->tqt_id) {
	tq->tq_lowest_id = taskq_lowest_id(tq);
	ASSERT3S(tq->tq_lowest_id, >, tqt->tqt_id);
	}

	/* Spawn additional taskq threads if required. */
	if ((++seq_tasks) > spl_taskq_thread_sequential &&
	taskq_thread_spawn(tq))
	seq_tasks = 0;

	tqt->tqt_id = TASKQID_INVALID;
	tqt->tqt_flags = 0;
	wake_up_all(&tq->tq_wait_waitq);
	} else {
	if (taskq_thread_should_stop(tq, tqt))
	break;
	}

	set_current_state(TASK_INTERRUPTIBLE);

	}

	__set_current_state(TASK_RUNNING);
	tq->tq_nthreads--;
	list_del_init(&tqt->tqt_thread_list);
	error:
	kmem_free(tqt, sizeof (taskq_thread_t));
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	tsd_set(taskq_tsd, NULL);
	thread_exit();

	return (0);
	}

	static taskq_thread_t *
	taskq_thread_create(taskq_t *tq)
	{
	static int last_used_cpu = 0;
	taskq_thread_t *tqt;

	tqt = kmem_alloc(sizeof (*tqt), KM_PUSHPAGE);
	INIT_LIST_HEAD(&tqt->tqt_thread_list);
	INIT_LIST_HEAD(&tqt->tqt_active_list);
	tqt->tqt_tq = tq;
	tqt->tqt_id = TASKQID_INVALID;

	tqt->tqt_thread = spl_kthread_create(taskq_thread, tqt,
	"%s", tq->tq_name);
	if (tqt->tqt_thread == NULL) {
	kmem_free(tqt, sizeof (taskq_thread_t));
	return (NULL);
	}

	if (spl_taskq_thread_bind) {
	last_used_cpu = (last_used_cpu + 1) % num_online_cpus();
	kthread_bind(tqt->tqt_thread, last_used_cpu);
	}

	if (spl_taskq_thread_priority)
	set_user_nice(tqt->tqt_thread, PRIO_TO_NICE(tq->tq_pri));

	wake_up_process(tqt->tqt_thread);

	return (tqt);
	}

	taskq_t *
	taskq_create(const char *name, int threads_arg, pri_t pri,
	int minalloc, int maxalloc, uint_t flags)
	{
	taskq_t *tq;
	taskq_thread_t *tqt;
	int count = 0, rc = 0, i;
	unsigned long irqflags;
	int nthreads = threads_arg;

	ASSERT(name != NULL);
	ASSERT(minalloc >= 0);
	ASSERT(maxalloc <= INT_MAX);
	ASSERT(!(flags & (TASKQ_CPR_SAFE))); /* Unsupported */

	/* Scale the number of threads using nthreads as a percentage */
	if (flags & TASKQ_THREADS_CPU_PCT) {
	ASSERT(nthreads <= 100);
	ASSERT(nthreads >= 0);
	nthreads = MIN(threads_arg, 100);
	nthreads = MAX(nthreads, 0);
	nthreads = MAX((num_online_cpus() * nthreads) /100, 1);
	}

	tq = kmem_alloc(sizeof (*tq), KM_PUSHPAGE);
	if (tq == NULL)
	return (NULL);

	tq->tq_hp_support = B_FALSE;
	#ifdef HAVE_CPU_HOTPLUG
	if (flags & TASKQ_THREADS_CPU_PCT) {
	tq->tq_hp_support = B_TRUE;
	if (cpuhp_state_add_instance_nocalls(spl_taskq_cpuhp_state,
	&tq->tq_hp_cb_node) != 0) {
	kmem_free(tq, sizeof (*tq));
	return (NULL);
	}
	}
	#endif

	spin_lock_init(&tq->tq_lock);
	INIT_LIST_HEAD(&tq->tq_thread_list);
	INIT_LIST_HEAD(&tq->tq_active_list);
	tq->tq_name = kmem_strdup(name);
	tq->tq_nactive = 0;
	tq->tq_nthreads = 0;
	tq->tq_nspawn = 0;
	tq->tq_maxthreads = nthreads;
	tq->tq_cpu_pct = threads_arg;
	tq->tq_pri = pri;
	tq->tq_minalloc = minalloc;
	tq->tq_maxalloc = maxalloc;
	tq->tq_nalloc = 0;
	tq->tq_flags = (flags \| TASKQ_ACTIVE);
	tq->tq_next_id = TASKQID_INITIAL;
	tq->tq_lowest_id = TASKQID_INITIAL;
	INIT_LIST_HEAD(&tq->tq_free_list);
	INIT_LIST_HEAD(&tq->tq_pend_list);
	INIT_LIST_HEAD(&tq->tq_prio_list);
	INIT_LIST_HEAD(&tq->tq_delay_list);
	init_waitqueue_head(&tq->tq_work_waitq);
	init_waitqueue_head(&tq->tq_wait_waitq);
	tq->tq_lock_class = TQ_LOCK_GENERAL;
	INIT_LIST_HEAD(&tq->tq_taskqs);

	if (flags & TASKQ_PREPOPULATE) {
	spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
	tq->tq_lock_class);

	for (i = 0; i < minalloc; i++)
	task_done(tq, task_alloc(tq, TQ_PUSHPAGE \| TQ_NEW,
	&irqflags));

	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	}

	if ((flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic)
	nthreads = 1;

	for (i = 0; i < nthreads; i++) {
	tqt = taskq_thread_create(tq);
	if (tqt == NULL)
	rc = 1;
	else
	count++;
	}

	/* Wait for all threads to be started before potential destroy */
	wait_event(tq->tq_wait_waitq, tq->tq_nthreads == count);
	/*
	* taskq_thread might have touched nspawn, but we don't want them to
	* because they're not dynamically spawned. So we reset it to 0
	*/
	tq->tq_nspawn = 0;

	if (rc) {
	taskq_destroy(tq);
	tq = NULL;
	} else {
	down_write(&tq_list_sem);
	tq->tq_instance = taskq_find_by_name(name) + 1;
	list_add_tail(&tq->tq_taskqs, &tq_list);
	up_write(&tq_list_sem);
	}

	return (tq);
	}
	EXPORT_SYMBOL(taskq_create);

	void
	taskq_destroy(taskq_t *tq)
	{
	struct task_struct *thread;
	taskq_thread_t *tqt;
	taskq_ent_t *t;
	unsigned long flags;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	tq->tq_flags &= ~TASKQ_ACTIVE;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	#ifdef HAVE_CPU_HOTPLUG
	if (tq->tq_hp_support) {
	VERIFY0(cpuhp_state_remove_instance_nocalls(
	spl_taskq_cpuhp_state, &tq->tq_hp_cb_node));
	}
	#endif
	/*
	* When TASKQ_ACTIVE is clear new tasks may not be added nor may
	* new worker threads be spawned for dynamic taskq.
	*/
	if (dynamic_taskq != NULL)
	taskq_wait_outstanding(dynamic_taskq, 0);

	taskq_wait(tq);

	/* remove taskq from global list used by the kstats */
	down_write(&tq_list_sem);
	list_del(&tq->tq_taskqs);
	up_write(&tq_list_sem);

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	/* wait for spawning threads to insert themselves to the list */
	while (tq->tq_nspawn) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	schedule_timeout_interruptible(1);
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	/*
	* Signal each thread to exit and block until it does. Each thread
	* is responsible for removing itself from the list and freeing its
	* taskq_thread_t. This allows for idle threads to opt to remove
	* themselves from the taskq. They can be recreated as needed.
	*/
	while (!list_empty(&tq->tq_thread_list)) {
	tqt = list_entry(tq->tq_thread_list.next,
	taskq_thread_t, tqt_thread_list);
	thread = tqt->tqt_thread;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kthread_stop(thread);

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	while (!list_empty(&tq->tq_free_list)) {
	t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	list_del_init(&t->tqent_list);
	task_free(tq, t);
	}

	ASSERT0(tq->tq_nthreads);
	ASSERT0(tq->tq_nalloc);
	ASSERT0(tq->tq_nspawn);
	ASSERT(list_empty(&tq->tq_thread_list));
	ASSERT(list_empty(&tq->tq_active_list));
	ASSERT(list_empty(&tq->tq_free_list));
	ASSERT(list_empty(&tq->tq_pend_list));
	ASSERT(list_empty(&tq->tq_prio_list));
	ASSERT(list_empty(&tq->tq_delay_list));

	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kmem_strfree(tq->tq_name);
	kmem_free(tq, sizeof (taskq_t));
	}
	EXPORT_SYMBOL(taskq_destroy);

	static unsigned int spl_taskq_kick = 0;

	/*
	* 2.6.36 API Change
	* module_param_cb is introduced to take kernel_param_ops and
	* module_param_call is marked as obsolete. Also set and get operations
	* were changed to take a 'const struct kernel_param *'.
	*/
	static int
	#ifdef module_param_cb
	param_set_taskq_kick(const char val, const struct kernel_param kp)
	#else
	param_set_taskq_kick(const char val, struct kernel_param kp)
	#endif
	{
	int ret;
	taskq_t *tq = NULL;
	taskq_ent_t *t;
	unsigned long flags;

	ret = param_set_uint(val, kp);
	if (ret < 0 \|\| !spl_taskq_kick)
	return (ret);
	/* reset value */
	spl_taskq_kick = 0;

	down_read(&tq_list_sem);
	list_for_each_entry(tq, &tq_list, tq_taskqs) {
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	/* Check if the first pending is older than 5 seconds */
	t = taskq_next_ent(tq);
	if (t && time_after(jiffies, t->tqent_birth + 5*HZ)) {
	(void) taskq_thread_spawn(tq);
	printk(KERN_INFO "spl: Kicked taskq %s/%d\n",
	tq->tq_name, tq->tq_instance);
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	}
	up_read(&tq_list_sem);
	return (ret);
	}

	#ifdef module_param_cb
	static const struct kernel_param_ops param_ops_taskq_kick = {
	.set = param_set_taskq_kick,
	.get = param_get_uint,
	};
	module_param_cb(spl_taskq_kick, &param_ops_taskq_kick, &spl_taskq_kick, 0644);
	#else
	module_param_call(spl_taskq_kick, param_set_taskq_kick, param_get_uint,
	&spl_taskq_kick, 0644);
	#endif
	MODULE_PARM_DESC(spl_taskq_kick,
	"Write nonzero to kick stuck taskqs to spawn more threads");

	#ifdef HAVE_CPU_HOTPLUG
	/*
	* This callback will be called exactly once for each core that comes online,
	* for each dynamic taskq. We attempt to expand taskqs that have
	* TASKQ_THREADS_CPU_PCT set. We need to redo the percentage calculation every
	* time, to correctly determine whether or not to add a thread.
	*/
	static int
	spl_taskq_expand(unsigned int cpu, struct hlist_node *node)
	{
	taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
	unsigned long flags;
	int err = 0;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (!(tq->tq_flags & TASKQ_ACTIVE)) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (err);
	}

	ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
	int nthreads = MIN(tq->tq_cpu_pct, 100);
	nthreads = MAX(((num_online_cpus() + 1) * nthreads) / 100, 1);
	tq->tq_maxthreads = nthreads;

	if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
	tq->tq_maxthreads > tq->tq_nthreads) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	taskq_thread_t *tqt = taskq_thread_create(tq);
	if (tqt == NULL)
	err = -1;
	return (err);
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (err);
	}

	/*
	* While we don't support offlining CPUs, it is possible that CPUs will fail
	* to online successfully. We do need to be able to handle this case
	* gracefully.
	*/
	static int
	spl_taskq_prepare_down(unsigned int cpu, struct hlist_node *node)
	{
	taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
	unsigned long flags;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
	int nthreads = MIN(tq->tq_cpu_pct, 100);
	nthreads = MAX(((num_online_cpus()) * nthreads) / 100, 1);
	tq->tq_maxthreads = nthreads;

	if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
	tq->tq_maxthreads < tq->tq_nthreads) {
	ASSERT3U(tq->tq_maxthreads, ==, tq->tq_nthreads - 1);
	taskq_thread_t *tqt = list_entry(tq->tq_thread_list.next,
	taskq_thread_t, tqt_thread_list);
	struct task_struct *thread = tqt->tqt_thread;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kthread_stop(thread);

	return (0);
	}

	out:
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (0);
	}
	#endif

	int
	spl_taskq_init(void)
	{
	init_rwsem(&tq_list_sem);
	tsd_create(&taskq_tsd, NULL);

	#ifdef HAVE_CPU_HOTPLUG
	spl_taskq_cpuhp_state = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN,
	"fs/spl_taskq:online", spl_taskq_expand, spl_taskq_prepare_down);
	#endif

	system_taskq = taskq_create("spl_system_taskq", MAX(boot_ncpus, 64),
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE\|TASKQ_DYNAMIC);
	if (system_taskq == NULL)
	return (-ENOMEM);

	system_delay_taskq = taskq_create("spl_delay_taskq", MAX(boot_ncpus, 4),
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE\|TASKQ_DYNAMIC);
	if (system_delay_taskq == NULL) {
	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	#endif
	taskq_destroy(system_taskq);
	return (-ENOMEM);
	}

	dynamic_taskq = taskq_create("spl_dynamic_taskq", 1,
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE);
	if (dynamic_taskq == NULL) {
	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	#endif
	taskq_destroy(system_taskq);
	taskq_destroy(system_delay_taskq);
	return (-ENOMEM);
	}

	/*
	* This is used to annotate tq_lock, so
	* taskq_dispatch -> taskq_thread_spawn -> taskq_dispatch
	* does not trigger a lockdep warning re: possible recursive locking
	*/
	dynamic_taskq->tq_lock_class = TQ_LOCK_DYNAMIC;

	return (0);
	}

	void
	spl_taskq_fini(void)
	{
	taskq_destroy(dynamic_taskq);
	dynamic_taskq = NULL;

	taskq_destroy(system_delay_taskq);
	system_delay_taskq = NULL;

	taskq_destroy(system_taskq);
	system_taskq = NULL;

	tsd_destroy(&taskq_tsd);

	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	spl_taskq_cpuhp_state = 0;
	#endif
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	index 32a2d34b1d93..b863945a1c59 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	@@ -1,207 +1,207 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Thread Implementation.
	*/

	#include <sys/thread.h>
	#include <sys/kmem.h>
	#include <sys/tsd.h>

	/*
	* Thread interfaces
	*/
	typedef struct thread_priv_s {
	unsigned long tp_magic; /* Magic */
	int tp_name_size; /* Name size */
	char tp_name; / Name (without _thread suffix) */
	void (tp_func)(void ); /* Registered function */
	void tp_args; / Args to be passed to function */
	size_t tp_len; /* Len to be passed to function */
	int tp_state; /* State to start thread at */
	pri_t tp_pri; /* Priority to start threat at */
	} thread_priv_t;

	static int
	thread_generic_wrapper(void *arg)
	{
	thread_priv_t tp = (thread_priv_t )arg;
	void (func)(void );
	void *args;

	ASSERT(tp->tp_magic == TP_MAGIC);
	func = tp->tp_func;
	args = tp->tp_args;
	set_current_state(tp->tp_state);
	set_user_nice((kthread_t *)current, PRIO_TO_NICE(tp->tp_pri));
	kmem_free(tp->tp_name, tp->tp_name_size);
	kmem_free(tp, sizeof (thread_priv_t));

	if (func)
	func(args);

	return (0);
	}

	/*
	* thread_create() may block forever if it cannot create a thread or
	* allocate memory. This is preferable to returning a NULL which Solaris
	* style callers likely never check for... since it can't fail.
	*/
	kthread_t *
	__thread_create(caddr_t stk, size_t stksize, thread_func_t func,
	const char name, void args, size_t len, proc_t *pp, int state, pri_t pri)
	{
	thread_priv_t *tp;
	struct task_struct *tsk;
	char *p;

	/* Option pp is simply ignored */
	/* Variable stack size unsupported */
	ASSERT(stk == NULL);

	tp = kmem_alloc(sizeof (thread_priv_t), KM_PUSHPAGE);
	if (tp == NULL)
	return (NULL);

	tp->tp_magic = TP_MAGIC;
	tp->tp_name_size = strlen(name) + 1;

	tp->tp_name = kmem_alloc(tp->tp_name_size, KM_PUSHPAGE);
	if (tp->tp_name == NULL) {
	kmem_free(tp, sizeof (thread_priv_t));
	return (NULL);
	}

	- strncpy(tp->tp_name, name, tp->tp_name_size);
	+ strlcpy(tp->tp_name, name, tp->tp_name_size);

	/*
	* Strip trailing "_thread" from passed name which will be the func
	* name since the exposed API has no parameter for passing a name.
	*/
	p = strstr(tp->tp_name, "_thread");
	if (p)
	p[0] = '\0';

	tp->tp_func = func;
	tp->tp_args = args;
	tp->tp_len = len;
	tp->tp_state = state;
	tp->tp_pri = pri;

	tsk = spl_kthread_create(thread_generic_wrapper, (void *)tp,
	"%s", tp->tp_name);
	if (IS_ERR(tsk))
	return (NULL);

	wake_up_process(tsk);
	return ((kthread_t *)tsk);
	}
	EXPORT_SYMBOL(__thread_create);

	/*
	* spl_kthread_create - Wrapper providing pre-3.13 semantics for
	* kthread_create() in which it is not killable and less likely
	* to return -ENOMEM.
	*/
	struct task_struct *
	spl_kthread_create(int (func)(void ), void *data, const char namefmt[], ...)
	{
	struct task_struct *tsk;
	va_list args;
	char name[TASK_COMM_LEN];

	va_start(args, namefmt);
	vsnprintf(name, sizeof (name), namefmt, args);
	va_end(args);
	do {
	tsk = kthread_create(func, data, "%s", name);
	if (IS_ERR(tsk)) {
	if (signal_pending(current)) {
	clear_thread_flag(TIF_SIGPENDING);
	continue;
	}
	if (PTR_ERR(tsk) == -ENOMEM)
	continue;
	return (NULL);
	} else {
	return (tsk);
	}
	} while (1);
	}
	EXPORT_SYMBOL(spl_kthread_create);

	/*
	* The "why" argument indicates the allowable side-effects of the call:
	*
	* FORREAL: Extract the next pending signal from p_sig into p_cursig;
	* stop the process if a stop has been requested or if a traced signal
	* is pending.
	*
	* JUSTLOOKING: Don't stop the process, just indicate whether or not
	* a signal might be pending (FORREAL is needed to tell for sure).
	*/
	int
	issig(int why)
	{
	ASSERT(why == FORREAL \|\| why == JUSTLOOKING);

	if (!signal_pending(current))
	return (0);

	if (why != FORREAL)
	return (1);

	struct task_struct *task = current;
	spl_kernel_siginfo_t __info;
	sigset_t set;
	siginitsetinv(&set, 1ULL << (SIGSTOP - 1) \| 1ULL << (SIGTSTP - 1));
	sigorsets(&set, &task->blocked, &set);

	spin_lock_irq(&task->sighand->siglock);
	int ret;
	#ifdef HAVE_DEQUEUE_SIGNAL_4ARG
	enum pid_type __type;
	if ((ret = dequeue_signal(task, &set, &__info, &__type)) != 0) {
	#else
	if ((ret = dequeue_signal(task, &set, &__info)) != 0) {
	#endif
	#ifdef HAVE_SIGNAL_STOP
	spin_unlock_irq(&task->sighand->siglock);
	kernel_signal_stop();
	#else
	if (current->jobctl & JOBCTL_STOP_DEQUEUED)
	spl_set_special_state(TASK_STOPPED);

	spin_unlock_irq(&current->sighand->siglock);

	schedule();
	#endif
	return (0);
	}

	spin_unlock_irq(&task->sighand->siglock);

	return (1);
	}

	EXPORT_SYMBOL(issig);
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c b/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c
	index 234ae7f6cd0c..9421f81bf0c8 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-zone.c
	@@ -1,424 +1,423 @@
	/*
	* Copyright (c) 2021 Klara Systems, Inc.
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*/

	#include <sys/types.h>
	#include <sys/mutex.h>
	#include <sys/sysmacros.h>
	#include <sys/kmem.h>
	#include <linux/file.h>
	#include <linux/magic.h>
	#include <sys/zone.h>

	#if defined(CONFIG_USER_NS)
	#include <linux/statfs.h>
	#include <linux/proc_ns.h>
	#endif

	static kmutex_t zone_datasets_lock;
	static struct list_head zone_datasets;

	typedef struct zone_datasets {
	struct list_head zds_list; /* zone_datasets linkage */
	struct user_namespace zds_userns; / namespace reference */
	struct list_head zds_datasets; /* datasets for the namespace */
	} zone_datasets_t;

	typedef struct zone_dataset {
	struct list_head zd_list; /* zone_dataset linkage */
	size_t zd_dsnamelen; /* length of name */
	char zd_dsname[0]; /* name of the member dataset */
	} zone_dataset_t;

	#if defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM)
	/*
	* Returns:
	* - 0 on success
	* - EBADF if it cannot open the provided file descriptor
	* - ENOTTY if the file itself is a not a user namespace file. We want to
	* intercept this error in the ZFS layer. We cannot just return one of the
	* ZFS_ERR_* errors here as we want to preserve the seperation of the ZFS
	* and the SPL layers.
	*/
	static int
	user_ns_get(int fd, struct user_namespace **userns)
	{
	struct kstatfs st;
	struct file *nsfile;
	struct ns_common *ns;
	int error;

	if ((nsfile = fget(fd)) == NULL)
	return (EBADF);
	if (vfs_statfs(&nsfile->f_path, &st) != 0) {
	error = ENOTTY;
	goto done;
	}
	if (st.f_type != NSFS_MAGIC) {
	error = ENOTTY;
	goto done;
	}
	ns = get_proc_ns(file_inode(nsfile));
	if (ns->ops->type != CLONE_NEWUSER) {
	error = ENOTTY;
	goto done;
	}
	*userns = container_of(ns, struct user_namespace, ns);

	error = 0;
	done:
	fput(nsfile);

	return (error);
	}
	#endif /* defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM) */

	static unsigned int
	user_ns_zoneid(struct user_namespace *user_ns)
	{
	unsigned int r;

	#if defined(HAVE_USER_NS_COMMON_INUM)
	r = user_ns->ns.inum;
	#else
	r = user_ns->proc_inum;
	#endif

	return (r);
	}

	static struct zone_datasets *
	zone_datasets_lookup(unsigned int nsinum)
	{
	zone_datasets_t *zds;

	list_for_each_entry(zds, &zone_datasets, zds_list) {
	if (user_ns_zoneid(zds->zds_userns) == nsinum)
	return (zds);
	}
	return (NULL);
	}

	#if defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM)
	static struct zone_dataset *
	zone_dataset_lookup(zone_datasets_t zds, const char dataset, size_t dsnamelen)
	{
	zone_dataset_t *zd;

	list_for_each_entry(zd, &zds->zds_datasets, zd_list) {
	if (zd->zd_dsnamelen != dsnamelen)
	continue;
	if (strncmp(zd->zd_dsname, dataset, dsnamelen) == 0)
	return (zd);
	}

	return (NULL);
	}

	static int
	zone_dataset_cred_check(cred_t *cred)
	{

	if (!uid_eq(cred->uid, GLOBAL_ROOT_UID))
	return (EPERM);

	return (0);
	}
	#endif /* defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM) */

	static int
	zone_dataset_name_check(const char dataset, size_t dsnamelen)
	{

	if (dataset[0] == '\0' \|\| dataset[0] == '/')
	return (ENOENT);

	*dsnamelen = strlen(dataset);
	/* Ignore trailing slash, if supplied. */
	if (dataset[*dsnamelen - 1] == '/')
	(*dsnamelen)--;

	return (0);
	}

	int
	zone_dataset_attach(cred_t cred, const char dataset, int userns_fd)
	{
	#if defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM)
	struct user_namespace *userns;
	zone_datasets_t *zds;
	zone_dataset_t *zd;
	int error;
	size_t dsnamelen;

	if ((error = zone_dataset_cred_check(cred)) != 0)
	return (error);
	if ((error = zone_dataset_name_check(dataset, &dsnamelen)) != 0)
	return (error);
	if ((error = user_ns_get(userns_fd, &userns)) != 0)
	return (error);

	mutex_enter(&zone_datasets_lock);
	zds = zone_datasets_lookup(user_ns_zoneid(userns));
	if (zds == NULL) {
	zds = kmem_alloc(sizeof (zone_datasets_t), KM_SLEEP);
	INIT_LIST_HEAD(&zds->zds_list);
	INIT_LIST_HEAD(&zds->zds_datasets);
	zds->zds_userns = userns;
	/*
	* Lock the namespace by incresing its refcount to prevent
	* the namespace ID from being reused.
	*/
	get_user_ns(userns);
	list_add_tail(&zds->zds_list, &zone_datasets);
	} else {
	zd = zone_dataset_lookup(zds, dataset, dsnamelen);
	if (zd != NULL) {
	mutex_exit(&zone_datasets_lock);
	return (EEXIST);
	}
	}

	zd = kmem_alloc(sizeof (zone_dataset_t) + dsnamelen + 1, KM_SLEEP);
	zd->zd_dsnamelen = dsnamelen;
	- strncpy(zd->zd_dsname, dataset, dsnamelen);
	- zd->zd_dsname[dsnamelen] = '\0';
	+ strlcpy(zd->zd_dsname, dataset, dsnamelen + 1);
	INIT_LIST_HEAD(&zd->zd_list);
	list_add_tail(&zd->zd_list, &zds->zds_datasets);

	mutex_exit(&zone_datasets_lock);
	return (0);
	#else
	return (ENXIO);
	#endif /* defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM) */
	}
	EXPORT_SYMBOL(zone_dataset_attach);

	int
	zone_dataset_detach(cred_t cred, const char dataset, int userns_fd)
	{
	#if defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM)
	struct user_namespace *userns;
	zone_datasets_t *zds;
	zone_dataset_t *zd;
	int error;
	size_t dsnamelen;

	if ((error = zone_dataset_cred_check(cred)) != 0)
	return (error);
	if ((error = zone_dataset_name_check(dataset, &dsnamelen)) != 0)
	return (error);
	if ((error = user_ns_get(userns_fd, &userns)) != 0)
	return (error);

	mutex_enter(&zone_datasets_lock);
	zds = zone_datasets_lookup(user_ns_zoneid(userns));
	if (zds != NULL)
	zd = zone_dataset_lookup(zds, dataset, dsnamelen);
	if (zds == NULL \|\| zd == NULL) {
	mutex_exit(&zone_datasets_lock);
	return (ENOENT);
	}

	list_del(&zd->zd_list);
	kmem_free(zd, sizeof (*zd) + zd->zd_dsnamelen + 1);

	/* Prune the namespace entry if it has no more delegations. */
	if (list_empty(&zds->zds_datasets)) {
	/*
	* Decrease the refcount now that the namespace is no longer
	* used. It is no longer necessary to prevent the namespace ID
	* from being reused.
	*/
	put_user_ns(userns);
	list_del(&zds->zds_list);
	kmem_free(zds, sizeof (*zds));
	}

	mutex_exit(&zone_datasets_lock);
	return (0);
	#else
	return (ENXIO);
	#endif /* defined(CONFIG_USER_NS) && defined(HAVE_USER_NS_COMMON_INUM) */
	}
	EXPORT_SYMBOL(zone_dataset_detach);

	/*
	* A dataset is visible if:
	* - It is a parent of a namespace entry.
	* - It is one of the namespace entries.
	* - It is a child of a namespace entry.
	*
	* A dataset is writable if:
	* - It is one of the namespace entries.
	* - It is a child of a namespace entry.
	*
	* The parent datasets of namespace entries are visible and
	* read-only to provide a path back to the root of the pool.
	*/
	int
	zone_dataset_visible(const char dataset, int write)
	{
	zone_datasets_t *zds;
	zone_dataset_t *zd;
	size_t dsnamelen, zd_len;
	int visible;

	/* Default to read-only, in case visible is returned. */
	if (write != NULL)
	*write = 0;
	if (zone_dataset_name_check(dataset, &dsnamelen) != 0)
	return (0);
	if (INGLOBALZONE(curproc)) {
	if (write != NULL)
	*write = 1;
	return (1);
	}

	mutex_enter(&zone_datasets_lock);
	zds = zone_datasets_lookup(crgetzoneid(curproc->cred));
	if (zds == NULL) {
	mutex_exit(&zone_datasets_lock);
	return (0);
	}

	visible = 0;
	list_for_each_entry(zd, &zds->zds_datasets, zd_list) {
	zd_len = strlen(zd->zd_dsname);
	if (zd_len > dsnamelen) {
	/*
	* The name of the namespace entry is longer than that
	* of the dataset, so it could be that the dataset is a
	* parent of the namespace entry.
	*/
	visible = memcmp(zd->zd_dsname, dataset,
	dsnamelen) == 0 &&
	zd->zd_dsname[dsnamelen] == '/';
	if (visible)
	break;
	} else if (zd_len == dsnamelen) {
	/*
	* The name of the namespace entry is as long as that
	* of the dataset, so perhaps the dataset itself is the
	* namespace entry.
	*/
	visible = memcmp(zd->zd_dsname, dataset, zd_len) == 0;
	if (visible) {
	if (write != NULL)
	*write = 1;
	break;
	}
	} else {
	/*
	* The name of the namespace entry is shorter than that
	* of the dataset, so perhaps the dataset is a child of
	* the namespace entry.
	*/
	visible = memcmp(zd->zd_dsname, dataset,
	zd_len) == 0 && dataset[zd_len] == '/';
	if (visible) {
	if (write != NULL)
	*write = 1;
	break;
	}
	}
	}

	mutex_exit(&zone_datasets_lock);
	return (visible);
	}
	EXPORT_SYMBOL(zone_dataset_visible);

	unsigned int
	global_zoneid(void)
	{
	unsigned int z = 0;

	#if defined(CONFIG_USER_NS)
	z = user_ns_zoneid(&init_user_ns);
	#endif

	return (z);
	}
	EXPORT_SYMBOL(global_zoneid);

	unsigned int
	crgetzoneid(const cred_t *cr)
	{
	unsigned int r = 0;

	#if defined(CONFIG_USER_NS)
	r = user_ns_zoneid(cr->user_ns);
	#endif

	return (r);
	}
	EXPORT_SYMBOL(crgetzoneid);

	boolean_t
	inglobalzone(proc_t *proc)
	{
	#if defined(CONFIG_USER_NS)
	return (proc->cred->user_ns == &init_user_ns);
	#else
	return (B_TRUE);
	#endif
	}
	EXPORT_SYMBOL(inglobalzone);

	int
	spl_zone_init(void)
	{
	mutex_init(&zone_datasets_lock, NULL, MUTEX_DEFAULT, NULL);
	INIT_LIST_HEAD(&zone_datasets);
	return (0);
	}

	void
	spl_zone_fini(void)
	{
	zone_datasets_t *zds;
	zone_dataset_t *zd;

	/*
	* It would be better to assert an empty zone_datasets, but since
	* there's no automatic mechanism for cleaning them up if the user
	* namespace is destroyed, just do it here, since spl is about to go
	* out of context.
	*/
	while (!list_empty(&zone_datasets)) {
	zds = list_entry(zone_datasets.next, zone_datasets_t, zds_list);
	while (!list_empty(&zds->zds_datasets)) {
	zd = list_entry(zds->zds_datasets.next,
	zone_dataset_t, zd_list);
	list_del(&zd->zd_list);
	kmem_free(zd, sizeof (*zd) + zd->zd_dsnamelen + 1);
	}
	put_user_ns(zds->zds_userns);
	list_del(&zds->zds_list);
	kmem_free(zds, sizeof (*zds));
	}
	mutex_destroy(&zone_datasets_lock);
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/arc_os.c b/sys/contrib/openzfs/module/os/linux/zfs/arc_os.c
	index d2a176d77813..eaaf7d0bb746 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/arc_os.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/arc_os.c
	@@ -1,540 +1,540 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2018, Joyent, Inc.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2014 by Saso Kiselkov. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc. All rights reserved.
	*/

	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/zfs_refcount.h>
	#include <sys/vdev.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_impl.h>
	#include <sys/dsl_pool.h>
	#include <sys/multilist.h>
	#include <sys/abd.h>
	#include <sys/zil.h>
	#include <sys/fm/fs/zfs.h>
	#ifdef _KERNEL
	#include <sys/shrinker.h>
	#include <sys/vmsystm.h>
	#include <sys/zpl.h>
	#include <linux/page_compat.h>
	#include <linux/notifier.h>
	#include <linux/memory.h>
	#endif
	#include <sys/callb.h>
	#include <sys/kstat.h>
	#include <sys/zthr.h>
	#include <zfs_fletcher.h>
	#include <sys/arc_impl.h>
	#include <sys/trace_zfs.h>
	#include <sys/aggsum.h>

	/*
	* This is a limit on how many pages the ARC shrinker makes available for
	* eviction in response to one page allocation attempt. Note that in
	* practice, the kernel's shrinker can ask us to evict up to about 4x this
	* for one allocation attempt.
	*
	* The default limit of 10,000 (in practice, 160MB per allocation attempt
	* with 4K pages) limits the amount of time spent attempting to reclaim ARC
	* memory to less than 100ms per allocation attempt, even with a small
	* average compressed block size of ~8KB.
	*
	* See also the comment in arc_shrinker_count().
	* Set to 0 to disable limit.
	*/
	int zfs_arc_shrinker_limit = 10000;

	#ifdef CONFIG_MEMORY_HOTPLUG
	static struct notifier_block arc_hotplug_callback_mem_nb;
	#endif

	/*
	* Return a default max arc size based on the amount of physical memory.
	*/
	uint64_t
	arc_default_max(uint64_t min, uint64_t allmem)
	{
	/* Default to 1/2 of all memory. */
	return (MAX(allmem / 2, min));
	}

	#ifdef _KERNEL
	/*
	* Return maximum amount of memory that we could possibly use. Reduced
	* to half of all memory in user space which is primarily used for testing.
	*/
	uint64_t
	arc_all_memory(void)
	{
	#ifdef CONFIG_HIGHMEM
	return (ptob(zfs_totalram_pages - zfs_totalhigh_pages));
	#else
	return (ptob(zfs_totalram_pages));
	#endif /* CONFIG_HIGHMEM */
	}

	/*
	* Return the amount of memory that is considered free. In user space
	* which is primarily used for testing we pretend that free memory ranges
	* from 0-20% of all memory.
	*/
	uint64_t
	arc_free_memory(void)
	{
	#ifdef CONFIG_HIGHMEM
	struct sysinfo si;
	si_meminfo(&si);
	return (ptob(si.freeram - si.freehigh));
	#else
	return (ptob(nr_free_pages() +
	nr_inactive_file_pages()));
	#endif /* CONFIG_HIGHMEM */
	}

	/*
	* Return the amount of memory that can be consumed before reclaim will be
	* needed. Positive if there is sufficient free memory, negative indicates
	* the amount of memory that needs to be freed up.
	*/
	int64_t
	arc_available_memory(void)
	{
	return (arc_free_memory() - arc_sys_free);
	}

	static uint64_t
	arc_evictable_memory(void)
	{
	int64_t asize = aggsum_value(&arc_sums.arcstat_size);
	uint64_t arc_clean =
	zfs_refcount_count(&arc_mru->arcs_esize[ARC_BUFC_DATA]) +
	zfs_refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) +
	zfs_refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_DATA]) +
	zfs_refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
	uint64_t arc_dirty = MAX((int64_t)asize - (int64_t)arc_clean, 0);

	/*
	* Scale reported evictable memory in proportion to page cache, cap
	* at specified min/max.
	*/
	uint64_t min = (ptob(nr_file_pages()) / 100) * zfs_arc_pc_percent;
	min = MAX(arc_c_min, MIN(arc_c_max, min));

	if (arc_dirty >= min)
	return (arc_clean);

	return (MAX((int64_t)asize - (int64_t)min, 0));
	}

	/*
	* The _count() function returns the number of free-able objects.
	* The _scan() function returns the number of objects that were freed.
	*/
	static unsigned long
	arc_shrinker_count(struct shrinker shrink, struct shrink_control sc)
	{
	/*
	* __GFP_FS won't be set if we are called from ZFS code (see
	* kmem_flags_convert(), which removes it). To avoid a deadlock, we
	* don't allow evicting in this case. We return 0 rather than
	* SHRINK_STOP so that the shrinker logic doesn't accumulate a
	* deficit against us.
	*/
	if (!(sc->gfp_mask & __GFP_FS)) {
	return (0);
	}

	/*
	* This code is reached in the "direct reclaim" case, where the
	* kernel (outside ZFS) is trying to allocate a page, and the system
	* is low on memory.
	*
	* The kernel's shrinker code doesn't understand how many pages the
	* ARC's callback actually frees, so it may ask the ARC to shrink a
	* lot for one page allocation. This is problematic because it may
	* take a long time, thus delaying the page allocation, and because
	* it may force the ARC to unnecessarily shrink very small.
	*
	* Therefore, we limit the amount of data that we say is evictable,
	* which limits the amount that the shrinker will ask us to evict for
	* one page allocation attempt.
	*
	* In practice, we may be asked to shrink 4x the limit to satisfy one
	* page allocation, before the kernel's shrinker code gives up on us.
	* When that happens, we rely on the kernel code to find the pages
	* that we freed before invoking the OOM killer. This happens in
	* __alloc_pages_slowpath(), which retries and finds the pages we
	* freed when it calls get_page_from_freelist().
	*
	* See also the comment above zfs_arc_shrinker_limit.
	*/
	int64_t limit = zfs_arc_shrinker_limit != 0 ?
	zfs_arc_shrinker_limit : INT64_MAX;
	return (MIN(limit, btop((int64_t)arc_evictable_memory())));
	}

	static unsigned long
	arc_shrinker_scan(struct shrinker shrink, struct shrink_control sc)
	{
	ASSERT((sc->gfp_mask & __GFP_FS) != 0);

	/* The arc is considered warm once reclaim has occurred */
	if (unlikely(arc_warm == B_FALSE))
	arc_warm = B_TRUE;

	/*
	* Evict the requested number of pages by reducing arc_c and waiting
	* for the requested amount of data to be evicted.
	*/
	arc_reduce_target_size(ptob(sc->nr_to_scan));
	arc_wait_for_eviction(ptob(sc->nr_to_scan), B_FALSE);
	if (current->reclaim_state != NULL)
	current->reclaim_state->reclaimed_slab += sc->nr_to_scan;

	/*
	* We are experiencing memory pressure which the arc_evict_zthr was
	* unable to keep up with. Set arc_no_grow to briefly pause arc
	* growth to avoid compounding the memory pressure.
	*/
	arc_no_grow = B_TRUE;

	/*
	* When direct reclaim is observed it usually indicates a rapid
	* increase in memory pressure. This occurs because the kswapd
	* threads were unable to asynchronously keep enough free memory
	* available.
	*/
	if (current_is_kswapd()) {
	ARCSTAT_BUMP(arcstat_memory_indirect_count);
	} else {
	ARCSTAT_BUMP(arcstat_memory_direct_count);
	}

	return (sc->nr_to_scan);
	}

	SPL_SHRINKER_DECLARE(arc_shrinker,
	arc_shrinker_count, arc_shrinker_scan, DEFAULT_SEEKS);

	int
	arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg)
	{
	uint64_t free_memory = arc_free_memory();

	if (free_memory > arc_all_memory() * arc_lotsfree_percent / 100)
	return (0);

	if (txg > spa->spa_lowmem_last_txg) {
	spa->spa_lowmem_last_txg = txg;
	spa->spa_lowmem_page_load = 0;
	}
	/*
	* If we are in pageout, we know that memory is already tight,
	* the arc is already going to be evicting, so we just want to
	* continue to let page writes occur as quickly as possible.
	*/
	if (current_is_kswapd()) {
	if (spa->spa_lowmem_page_load >
	MAX(arc_sys_free / 4, free_memory) / 4) {
	DMU_TX_STAT_BUMP(dmu_tx_memory_reclaim);
	return (SET_ERROR(ERESTART));
	}
	/* Note: reserve is inflated, so we deflate */
	atomic_add_64(&spa->spa_lowmem_page_load, reserve / 8);
	return (0);
	} else if (spa->spa_lowmem_page_load > 0 && arc_reclaim_needed()) {
	/* memory is low, delay before restarting */
	ARCSTAT_INCR(arcstat_memory_throttle_count, 1);
	DMU_TX_STAT_BUMP(dmu_tx_memory_reclaim);
	return (SET_ERROR(EAGAIN));
	}
	spa->spa_lowmem_page_load = 0;
	return (0);
	}

	static void
	arc_set_sys_free(uint64_t allmem)
	{
	/*
	* The ARC tries to keep at least this much memory available for the
	* system. This gives the ARC time to shrink in response to memory
	* pressure, before running completely out of memory and invoking the
	* direct-reclaim ARC shrinker.
	*
	* This should be more than twice high_wmark_pages(), so that
	* arc_wait_for_eviction() will wait until at least the
	* high_wmark_pages() are free (see arc_evict_state_impl()).
	*
	* Note: Even when the system is very low on memory, the kernel's
	* shrinker code may only ask for one "batch" of pages (512KB) to be
	* evicted. If concurrent allocations consume these pages, there may
	* still be insufficient free pages, and the OOM killer takes action.
	*
	* By setting arc_sys_free large enough, and having
	* arc_wait_for_eviction() wait until there is at least arc_sys_free/2
	* free memory, it is much less likely that concurrent allocations can
	* consume all the memory that was evicted before checking for
	* OOM.
	*
	* It's hard to iterate the zones from a linux kernel module, which
	* makes it difficult to determine the watermark dynamically. Instead
	* we compute the maximum high watermark for this system, based
	* on the amount of memory, assuming default parameters on Linux kernel
	* 5.3.
	*/

	/*
	* Base wmark_low is 4 * the square root of Kbytes of RAM.
	*/
	long wmark = 4 * int_sqrt(allmem/1024) * 1024;

	/*
	* Clamp to between 128K and 64MB.
	*/
	wmark = MAX(wmark, 128 * 1024);
	wmark = MIN(wmark, 64 * 1024 * 1024);

	/*
	* watermark_boost can increase the wmark by up to 150%.
	*/
	wmark += wmark * 150 / 100;

	/*
	* arc_sys_free needs to be more than 2x the watermark, because
	* arc_wait_for_eviction() waits for half of arc_sys_free. Bump this up
	* to 3x to ensure we're above it.
	*/
	arc_sys_free = wmark * 3 + allmem / 32;
	}

	void
	arc_lowmem_init(void)
	{
	uint64_t allmem = arc_all_memory();

	/*
	* Register a shrinker to support synchronous (direct) memory
	* reclaim from the arc. This is done to prevent kswapd from
	* swapping out pages when it is preferable to shrink the arc.
	*/
	spl_register_shrinker(&arc_shrinker);
	arc_set_sys_free(allmem);
	}

	void
	arc_lowmem_fini(void)
	{
	spl_unregister_shrinker(&arc_shrinker);
	}

	int
	param_set_arc_long(const char buf, zfs_kernel_param_t kp)
	{
	int error;

	error = param_set_long(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	arc_tuning_update(B_TRUE);

	return (0);
	}

	int
	param_set_arc_min(const char buf, zfs_kernel_param_t kp)
	{
	return (param_set_arc_long(buf, kp));
	}

	int
	param_set_arc_max(const char buf, zfs_kernel_param_t kp)
	{
	return (param_set_arc_long(buf, kp));
	}

	int
	param_set_arc_int(const char buf, zfs_kernel_param_t kp)
	{
	int error;

	error = param_set_int(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	arc_tuning_update(B_TRUE);

	return (0);
	}

	#ifdef CONFIG_MEMORY_HOTPLUG
	static int
	arc_hotplug_callback(struct notifier_block *self, unsigned long action,
	void *arg)
	{
	(void) self, (void) arg;
	uint64_t allmem = arc_all_memory();
	if (action != MEM_ONLINE)
	return (NOTIFY_OK);

	arc_set_limits(allmem);

	#ifdef __LP64__
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#else
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(1ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#endif

	arc_set_sys_free(allmem);
	return (NOTIFY_OK);
	}
	#endif

	void
	arc_register_hotplug(void)
	{
	#ifdef CONFIG_MEMORY_HOTPLUG
	arc_hotplug_callback_mem_nb.notifier_call = arc_hotplug_callback;
	/* There is no significance to the value 100 */
	arc_hotplug_callback_mem_nb.priority = 100;
	register_memory_notifier(&arc_hotplug_callback_mem_nb);
	#endif
	}

	void
	arc_unregister_hotplug(void)
	{
	#ifdef CONFIG_MEMORY_HOTPLUG
	unregister_memory_notifier(&arc_hotplug_callback_mem_nb);
	#endif
	}
	#else /* _KERNEL */
	int64_t
	arc_available_memory(void)
	{
	int64_t lowest = INT64_MAX;

	/* Every 100 calls, free a small amount */
	if (random_in_range(100) == 0)
	lowest = -1024;

	return (lowest);
	}

	int
	arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg)
	{
	(void) spa, (void) reserve, (void) txg;
	return (0);
	}

	uint64_t
	arc_all_memory(void)
	{
	return (ptob(physmem) / 2);
	}

	uint64_t
	arc_free_memory(void)
	{
	return (random_in_range(arc_all_memory() * 20 / 100));
	}

	void
	arc_register_hotplug(void)
	{
	}

	void
	arc_unregister_hotplug(void)
	{
	}
	#endif /* _KERNEL */

	/*
	* Helper function for arc_prune_async() it is responsible for safely
	* handling the execution of a registered arc_prune_func_t.
	*/
	static void
	arc_prune_task(void *ptr)
	{
	arc_prune_t ap = (arc_prune_t )ptr;
	arc_prune_func_t *func = ap->p_pfunc;

	if (func != NULL)
	func(ap->p_adjust, ap->p_private);

	zfs_refcount_remove(&ap->p_refcnt, func);
	}

	/*
	* Notify registered consumers they must drop holds on a portion of the ARC
	* buffered they reference. This provides a mechanism to ensure the ARC can
	* honor the arc_meta_limit and reclaim otherwise pinned ARC buffers. This
	* is analogous to dnlc_reduce_cache() but more generic.
	*
	* This operation is performed asynchronously so it may be safely called
	* in the context of the arc_reclaim_thread(). A reference is taken here
	* for each registered arc_prune_t and the arc_prune_task() is responsible
	* for releasing it once the registered arc_prune_func_t has completed.
	*/
	void
	-arc_prune_async(int64_t adjust)
	+arc_prune_async(uint64_t adjust)
	{
	arc_prune_t *ap;

	mutex_enter(&arc_prune_mtx);
	for (ap = list_head(&arc_prune_list); ap != NULL;
	ap = list_next(&arc_prune_list, ap)) {

	if (zfs_refcount_count(&ap->p_refcnt) >= 2)
	continue;

	zfs_refcount_add(&ap->p_refcnt, ap->p_pfunc);
	ap->p_adjust = adjust;
	if (taskq_dispatch(arc_prune_taskq, arc_prune_task,
	ap, TQ_SLEEP) == TASKQID_INVALID) {
	zfs_refcount_remove(&ap->p_refcnt, ap->p_pfunc);
	continue;
	}
	ARCSTAT_BUMP(arcstat_prune);
	}
	mutex_exit(&arc_prune_mtx);
	}

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, shrinker_limit, INT, ZMOD_RW,
	"Limit on number of pages that ARC shrinker can reclaim at once");
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	index ba7adcc1b576..0fed09df5203 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	@@ -1,1024 +1,1044 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
	* LLNL-CODE-403049.
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_disk.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_trim.h>
	#include <sys/abd.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio.h>
	#include <linux/blkpg.h>
	#include <linux/msdos_fs.h>
	#include <linux/vfs_compat.h>
	#ifdef HAVE_LINUX_BLK_CGROUP_HEADER
	#include <linux/blk-cgroup.h>
	#endif

	typedef struct vdev_disk {
	struct block_device *vd_bdev;
	krwlock_t vd_lock;
	} vdev_disk_t;

	/*
	* Unique identifier for the exclusive vdev holder.
	*/
	static void *zfs_vdev_holder = VDEV_HOLDER;

	/*
	* Wait up to zfs_vdev_open_timeout_ms milliseconds before determining the
	* device is missing. The missing path may be transient since the links
	* can be briefly removed and recreated in response to udev events.
	*/
	static unsigned zfs_vdev_open_timeout_ms = 1000;

	/*
	* Size of the "reserved" partition, in blocks.
	*/
	#define EFI_MIN_RESV_SIZE (16 * 1024)

	/*
	* Virtual device vector for disks.
	*/
	typedef struct dio_request {
	zio_t dr_zio; / Parent ZIO */
	atomic_t dr_ref; /* References */
	int dr_error; /* Bio error */
	int dr_bio_count; /* Count of bio's */
	struct bio dr_bio[0]; / Attached bio's */
	} dio_request_t;

	static fmode_t
	vdev_bdev_mode(spa_mode_t spa_mode)
	{
	fmode_t mode = 0;

	if (spa_mode & SPA_MODE_READ)
	mode \|= FMODE_READ;

	if (spa_mode & SPA_MODE_WRITE)
	mode \|= FMODE_WRITE;

	return (mode);
	}

	/*
	* Returns the usable capacity (in bytes) for the partition or disk.
	*/
	static uint64_t
	bdev_capacity(struct block_device *bdev)
	{
	return (i_size_read(bdev->bd_inode));
	}

	#if !defined(HAVE_BDEV_WHOLE)
	static inline struct block_device *
	bdev_whole(struct block_device *bdev)
	{
	return (bdev->bd_contains);
	}
	#endif

	#if defined(HAVE_BDEVNAME)
	#define vdev_bdevname(bdev, name) bdevname(bdev, name)
	#else
	static inline void
	vdev_bdevname(struct block_device bdev, char name)
	{
	snprintf(name, BDEVNAME_SIZE, "%pg", bdev);
	}
	#endif

	/*
	* Returns the maximum expansion capacity of the block device (in bytes).
	*
	* It is possible to expand a vdev when it has been created as a wholedisk
	* and the containing block device has increased in capacity. Or when the
	* partition containing the pool has been manually increased in size.
	*
	* This function is only responsible for calculating the potential expansion
	* size so it can be reported by 'zpool list'. The efi_use_whole_disk() is
	* responsible for verifying the expected partition layout in the wholedisk
	* case, and updating the partition table if appropriate. Once the partition
	* size has been increased the additional capacity will be visible using
	* bdev_capacity().
	*
	* The returned maximum expansion capacity is always expected to be larger, or
	* at the very least equal, to its usable capacity to prevent overestimating
	* the pool expandsize.
	*/
	static uint64_t
	bdev_max_capacity(struct block_device *bdev, uint64_t wholedisk)
	{
	uint64_t psize;
	int64_t available;

	if (wholedisk && bdev != bdev_whole(bdev)) {
	/*
	* When reporting maximum expansion capacity for a wholedisk
	* deduct any capacity which is expected to be lost due to
	* alignment restrictions. Over reporting this value isn't
	* harmful and would only result in slightly less capacity
	* than expected post expansion.
	* The estimated available space may be slightly smaller than
	* bdev_capacity() for devices where the number of sectors is
	* not a multiple of the alignment size and the partition layout
	* is keeping less than PARTITION_END_ALIGNMENT bytes after the
	* "reserved" EFI partition: in such cases return the device
	* usable capacity.
	*/
	available = i_size_read(bdev_whole(bdev)->bd_inode) -
	((EFI_MIN_RESV_SIZE + NEW_START_BLOCK +
	PARTITION_END_ALIGNMENT) << SECTOR_BITS);
	psize = MAX(available, bdev_capacity(bdev));
	} else {
	psize = bdev_capacity(bdev);
	}

	return (psize);
	}

	static void
	vdev_disk_error(zio_t *zio)
	{
	/*
	* This function can be called in interrupt context, for instance while
	* handling IRQs coming from a misbehaving disk device; use printk()
	* which is safe from any context.
	*/
	printk(KERN_WARNING "zio pool=%s vdev=%s error=%d type=%d "
	"offset=%llu size=%llu flags=%x\n", spa_name(zio->io_spa),
	zio->io_vd->vdev_path, zio->io_error, zio->io_type,
	(u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size,
	zio->io_flags);
	}

	+static void
	+vdev_disk_kobj_evt_post(vdev_t *v)
	+{
	+ vdev_disk_t *vd = v->vdev_tsd;
	+ if (vd && vd->vd_bdev) {
	+ spl_signal_kobj_evt(vd->vd_bdev);
	+ } else {
	+ vdev_dbgmsg(v, "vdev_disk_t is NULL for VDEV:%s\n",
	+ v->vdev_path);
	+ }
	+}
	+
	static int
	vdev_disk_open(vdev_t v, uint64_t psize, uint64_t *max_psize,
	uint64_t logical_ashift, uint64_t physical_ashift)
	{
	struct block_device *bdev;
	fmode_t mode = vdev_bdev_mode(spa_mode(v->vdev_spa));
	hrtime_t timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms);
	vdev_disk_t *vd;

	/* Must have a pathname and it must be absolute. */
	if (v->vdev_path == NULL \|\| v->vdev_path[0] != '/') {
	v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
	vdev_dbgmsg(v, "invalid vdev_path");
	return (SET_ERROR(EINVAL));
	}

	/*
	* Reopen the device if it is currently open. When expanding a
	* partition force re-scanning the partition table if userland
	* did not take care of this already. We need to do this while closed
	* in order to get an accurate updated block device size. Then
	* since udev may need to recreate the device links increase the
	* open retry timeout before reporting the device as unavailable.
	*/
	vd = v->vdev_tsd;
	if (vd) {
	char disk_name[BDEVNAME_SIZE + 6] = "/dev/";
	boolean_t reread_part = B_FALSE;

	rw_enter(&vd->vd_lock, RW_WRITER);
	bdev = vd->vd_bdev;
	vd->vd_bdev = NULL;

	if (bdev) {
	if (v->vdev_expanding && bdev != bdev_whole(bdev)) {
	vdev_bdevname(bdev_whole(bdev), disk_name + 5);
	/*
	* If userland has BLKPG_RESIZE_PARTITION,
	* then it should have updated the partition
	* table already. We can detect this by
	* comparing our current physical size
	* with that of the device. If they are
	* the same, then we must not have
	* BLKPG_RESIZE_PARTITION or it failed to
	* update the partition table online. We
	* fallback to rescanning the partition
	* table from the kernel below. However,
	* if the capacity already reflects the
	* updated partition, then we skip
	* rescanning the partition table here.
	*/
	if (v->vdev_psize == bdev_capacity(bdev))
	reread_part = B_TRUE;
	}

	blkdev_put(bdev, mode \| FMODE_EXCL);
	}

	if (reread_part) {
	bdev = blkdev_get_by_path(disk_name, mode \| FMODE_EXCL,
	zfs_vdev_holder);
	if (!IS_ERR(bdev)) {
	int error = vdev_bdev_reread_part(bdev);
	blkdev_put(bdev, mode \| FMODE_EXCL);
	if (error == 0) {
	timeout = MSEC2NSEC(
	zfs_vdev_open_timeout_ms * 2);
	}
	}
	}
	} else {
	vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);

	rw_init(&vd->vd_lock, NULL, RW_DEFAULT, NULL);
	rw_enter(&vd->vd_lock, RW_WRITER);
	}

	/*
	* Devices are always opened by the path provided at configuration
	* time. This means that if the provided path is a udev by-id path
	* then drives may be re-cabled without an issue. If the provided
	* path is a udev by-path path, then the physical location information
	* will be preserved. This can be critical for more complicated
	* configurations where drives are located in specific physical
	* locations to maximize the systems tolerance to component failure.
	*
	* Alternatively, you can provide your own udev rule to flexibly map
	* the drives as you see fit. It is not advised that you use the
	* /dev/[hd]d devices which may be reordered due to probing order.
	* Devices in the wrong locations will be detected by the higher
	* level vdev validation.
	*
	* The specified paths may be briefly removed and recreated in
	* response to udev events. This should be exceptionally unlikely
	* because the zpool command makes every effort to verify these paths
	* have already settled prior to reaching this point. Therefore,
	* a ENOENT failure at this point is highly likely to be transient
	* and it is reasonable to sleep and retry before giving up. In
	* practice delays have been observed to be on the order of 100ms.
	*
	* When ERESTARTSYS is returned it indicates the block device is
	* a zvol which could not be opened due to the deadlock detection
	* logic in zvol_open(). Extend the timeout and retry the open
	* subsequent attempts are expected to eventually succeed.
	*/
	hrtime_t start = gethrtime();
	bdev = ERR_PTR(-ENXIO);
	while (IS_ERR(bdev) && ((gethrtime() - start) < timeout)) {
	bdev = blkdev_get_by_path(v->vdev_path, mode \| FMODE_EXCL,
	zfs_vdev_holder);
	if (unlikely(PTR_ERR(bdev) == -ENOENT)) {
	+ /*
	+ * There is no point of waiting since device is removed
	+ * explicitly
	+ */
	+ if (v->vdev_removed)
	+ break;
	+
	schedule_timeout(MSEC_TO_TICK(10));
	} else if (unlikely(PTR_ERR(bdev) == -ERESTARTSYS)) {
	timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms * 10);
	continue;
	} else if (IS_ERR(bdev)) {
	break;
	}
	}

	if (IS_ERR(bdev)) {
	int error = -PTR_ERR(bdev);
	vdev_dbgmsg(v, "open error=%d timeout=%llu/%llu", error,
	(u_longlong_t)(gethrtime() - start),
	(u_longlong_t)timeout);
	vd->vd_bdev = NULL;
	v->vdev_tsd = vd;
	rw_exit(&vd->vd_lock);
	return (SET_ERROR(error));
	} else {
	vd->vd_bdev = bdev;
	v->vdev_tsd = vd;
	rw_exit(&vd->vd_lock);
	}

	/* Determine the physical block size */
	int physical_block_size = bdev_physical_block_size(vd->vd_bdev);

	/* Determine the logical block size */
	int logical_block_size = bdev_logical_block_size(vd->vd_bdev);

	/* Clear the nowritecache bit, causes vdev_reopen() to try again. */
	v->vdev_nowritecache = B_FALSE;

	/* Set when device reports it supports TRIM. */
	v->vdev_has_trim = bdev_discard_supported(vd->vd_bdev);

	/* Set when device reports it supports secure TRIM. */
	v->vdev_has_securetrim = bdev_secure_discard_supported(vd->vd_bdev);

	/* Inform the ZIO pipeline that we are non-rotational */
	v->vdev_nonrot = blk_queue_nonrot(bdev_get_queue(vd->vd_bdev));

	/* Physical volume size in bytes for the partition */
	*psize = bdev_capacity(vd->vd_bdev);

	/* Physical volume size in bytes including possible expansion space */
	*max_psize = bdev_max_capacity(vd->vd_bdev, v->vdev_wholedisk);

	/* Based on the minimum sector size set the block size */
	*physical_ashift = highbit64(MAX(physical_block_size,
	SPA_MINBLOCKSIZE)) - 1;

	*logical_ashift = highbit64(MAX(logical_block_size,
	SPA_MINBLOCKSIZE)) - 1;

	return (0);
	}

	static void
	vdev_disk_close(vdev_t *v)
	{
	vdev_disk_t *vd = v->vdev_tsd;

	if (v->vdev_reopening \|\| vd == NULL)
	return;

	if (vd->vd_bdev != NULL) {
	blkdev_put(vd->vd_bdev,
	vdev_bdev_mode(spa_mode(v->vdev_spa)) \| FMODE_EXCL);
	}

	rw_destroy(&vd->vd_lock);
	kmem_free(vd, sizeof (vdev_disk_t));
	v->vdev_tsd = NULL;
	}

	static dio_request_t *
	vdev_disk_dio_alloc(int bio_count)
	{
	dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) +
	sizeof (struct bio ) bio_count, KM_SLEEP);
	atomic_set(&dr->dr_ref, 0);
	dr->dr_bio_count = bio_count;
	dr->dr_error = 0;

	for (int i = 0; i < dr->dr_bio_count; i++)
	dr->dr_bio[i] = NULL;

	return (dr);
	}

	static void
	vdev_disk_dio_free(dio_request_t *dr)
	{
	int i;

	for (i = 0; i < dr->dr_bio_count; i++)
	if (dr->dr_bio[i])
	bio_put(dr->dr_bio[i]);

	kmem_free(dr, sizeof (dio_request_t) +
	sizeof (struct bio ) dr->dr_bio_count);
	}

	static void
	vdev_disk_dio_get(dio_request_t *dr)
	{
	atomic_inc(&dr->dr_ref);
	}

	static int
	vdev_disk_dio_put(dio_request_t *dr)
	{
	int rc = atomic_dec_return(&dr->dr_ref);

	/*
	* Free the dio_request when the last reference is dropped and
	* ensure zio_interpret is called only once with the correct zio
	*/
	if (rc == 0) {
	zio_t *zio = dr->dr_zio;
	int error = dr->dr_error;

	vdev_disk_dio_free(dr);

	if (zio) {
	zio->io_error = error;
	ASSERT3S(zio->io_error, >=, 0);
	if (zio->io_error)
	vdev_disk_error(zio);

	zio_delay_interrupt(zio);
	}
	}

	return (rc);
	}

	BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
	{
	dio_request_t *dr = bio->bi_private;
	int rc;

	if (dr->dr_error == 0) {
	#ifdef HAVE_1ARG_BIO_END_IO_T
	dr->dr_error = BIO_END_IO_ERROR(bio);
	#else
	if (error)
	dr->dr_error = -(error);
	else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
	dr->dr_error = EIO;
	#endif
	}

	/* Drop reference acquired by __vdev_disk_physio */
	rc = vdev_disk_dio_put(dr);
	}

	static inline void
	vdev_submit_bio_impl(struct bio *bio)
	{
	#ifdef HAVE_1ARG_SUBMIT_BIO
	(void) submit_bio(bio);
	#else
	(void) submit_bio(bio_data_dir(bio), bio);
	#endif
	}

	/*
	* preempt_schedule_notrace is GPL-only which breaks the ZFS build, so
	* replace it with preempt_schedule under the following condition:
	*/
	#if defined(CONFIG_ARM64) && \
	defined(CONFIG_PREEMPTION) && \
	defined(CONFIG_BLK_CGROUP)
	#define preempt_schedule_notrace(x) preempt_schedule(x)
	#endif

	/*
	* As for the Linux 5.18 kernel bio_alloc() expects a block_device struct
	* as an argument removing the need to set it with bio_set_dev(). This
	* removes the need for all of the following compatibility code.
	*/
	#if !defined(HAVE_BIO_ALLOC_4ARG)

	#ifdef HAVE_BIO_SET_DEV
	#if defined(CONFIG_BLK_CGROUP) && defined(HAVE_BIO_SET_DEV_GPL_ONLY)
	/*
	* The Linux 5.5 kernel updated percpu_ref_tryget() which is inlined by
	* blkg_tryget() to use rcu_read_lock() instead of rcu_read_lock_sched().
	* As a side effect the function was converted to GPL-only. Define our
	* own version when needed which uses rcu_read_lock_sched().
	*
	* The Linux 5.17 kernel split linux/blk-cgroup.h into a private and a public
	* part, moving blkg_tryget into the private one. Define our own version.
	*/
	#if defined(HAVE_BLKG_TRYGET_GPL_ONLY) \|\| !defined(HAVE_BLKG_TRYGET)
	static inline bool
	vdev_blkg_tryget(struct blkcg_gq *blkg)
	{
	struct percpu_ref *ref = &blkg->refcnt;
	unsigned long __percpu *count;
	bool rc;

	rcu_read_lock_sched();

	if (__ref_is_percpu(ref, &count)) {
	this_cpu_inc(*count);
	rc = true;
	} else {
	#ifdef ZFS_PERCPU_REF_COUNT_IN_DATA
	rc = atomic_long_inc_not_zero(&ref->data->count);
	#else
	rc = atomic_long_inc_not_zero(&ref->count);
	#endif
	}

	rcu_read_unlock_sched();

	return (rc);
	}
	#else
	#define vdev_blkg_tryget(bg) blkg_tryget(bg)
	#endif
	#ifdef HAVE_BIO_SET_DEV_MACRO
	/*
	* The Linux 5.0 kernel updated the bio_set_dev() macro so it calls the
	* GPL-only bio_associate_blkg() symbol thus inadvertently converting
	* the entire macro. Provide a minimal version which always assigns the
	* request queue's root_blkg to the bio.
	*/
	static inline void
	vdev_bio_associate_blkg(struct bio *bio)
	{
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif

	ASSERT3P(q, !=, NULL);
	ASSERT3P(bio->bi_blkg, ==, NULL);

	if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
	bio->bi_blkg = q->root_blkg;
	}

	#define bio_associate_blkg vdev_bio_associate_blkg
	#else
	static inline void
	vdev_bio_set_dev(struct bio bio, struct block_device bdev)
	{
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif
	bio_clear_flag(bio, BIO_REMAPPED);
	if (bio->bi_bdev != bdev)
	bio_clear_flag(bio, BIO_THROTTLED);
	bio->bi_bdev = bdev;

	ASSERT3P(q, !=, NULL);
	ASSERT3P(bio->bi_blkg, ==, NULL);

	if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
	bio->bi_blkg = q->root_blkg;
	}
	#define bio_set_dev vdev_bio_set_dev
	#endif
	#endif
	#else
	/*
	* Provide a bio_set_dev() helper macro for pre-Linux 4.14 kernels.
	*/
	static inline void
	bio_set_dev(struct bio bio, struct block_device bdev)
	{
	bio->bi_bdev = bdev;
	}
	#endif /* HAVE_BIO_SET_DEV */
	#endif /* !HAVE_BIO_ALLOC_4ARG */

	static inline void
	vdev_submit_bio(struct bio *bio)
	{
	struct bio_list *bio_list = current->bio_list;
	current->bio_list = NULL;
	vdev_submit_bio_impl(bio);
	current->bio_list = bio_list;
	}

	static inline struct bio *
	vdev_bio_alloc(struct block_device *bdev, gfp_t gfp_mask,
	unsigned short nr_vecs)
	{
	struct bio *bio;

	#ifdef HAVE_BIO_ALLOC_4ARG
	bio = bio_alloc(bdev, nr_vecs, 0, gfp_mask);
	#else
	bio = bio_alloc(gfp_mask, nr_vecs);
	if (likely(bio != NULL))
	bio_set_dev(bio, bdev);
	#endif

	return (bio);
	}

	static inline unsigned int
	vdev_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset)
	{
	unsigned long nr_segs = abd_nr_pages_off(zio->io_abd,
	bio_size, abd_offset);

	#ifdef HAVE_BIO_MAX_SEGS
	return (bio_max_segs(nr_segs));
	#else
	return (MIN(nr_segs, BIO_MAX_PAGES));
	#endif
	}

	static int
	__vdev_disk_physio(struct block_device bdev, zio_t zio,
	size_t io_size, uint64_t io_offset, int rw, int flags)
	{
	dio_request_t *dr;
	uint64_t abd_offset;
	uint64_t bio_offset;
	int bio_size;
	int bio_count = 16;
	int error = 0;
	struct blk_plug plug;
	unsigned short nr_vecs;

	/*
	* Accessing outside the block device is never allowed.
	*/
	if (io_offset + io_size > bdev->bd_inode->i_size) {
	vdev_dbgmsg(zio->io_vd,
	"Illegal access %llu size %llu, device size %llu",
	(u_longlong_t)io_offset,
	(u_longlong_t)io_size,
	(u_longlong_t)i_size_read(bdev->bd_inode));
	return (SET_ERROR(EIO));
	}

	retry:
	dr = vdev_disk_dio_alloc(bio_count);

	if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY \| ZIO_FLAG_TRYHARD)))
	bio_set_flags_failfast(bdev, &flags);

	dr->dr_zio = zio;

	/*
	* Since bio's can have up to BIO_MAX_PAGES=256 iovec's, each of which
	* is at least 512 bytes and at most PAGESIZE (typically 4K), one bio
	* can cover at least 128KB and at most 1MB. When the required number
	* of iovec's exceeds this, we are forced to break the IO in multiple
	* bio's and wait for them all to complete. This is likely if the
	* recordsize property is increased beyond 1MB. The default
	* bio_count=16 should typically accommodate the maximum-size zio of
	* 16MB.
	*/

	abd_offset = 0;
	bio_offset = io_offset;
	bio_size = io_size;
	for (int i = 0; i <= dr->dr_bio_count; i++) {

	/* Finished constructing bio's for given buffer */
	if (bio_size <= 0)
	break;

	/*
	* If additional bio's are required, we have to retry, but
	* this should be rare - see the comment above.
	*/
	if (dr->dr_bio_count == i) {
	vdev_disk_dio_free(dr);
	bio_count *= 2;
	goto retry;
	}

	nr_vecs = vdev_bio_max_segs(zio, bio_size, abd_offset);
	dr->dr_bio[i] = vdev_bio_alloc(bdev, GFP_NOIO, nr_vecs);
	if (unlikely(dr->dr_bio[i] == NULL)) {
	vdev_disk_dio_free(dr);
	return (SET_ERROR(ENOMEM));
	}

	/* Matching put called by vdev_disk_physio_completion */
	vdev_disk_dio_get(dr);

	BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
	dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
	dr->dr_bio[i]->bi_private = dr;
	bio_set_op_attrs(dr->dr_bio[i], rw, flags);

	/* Remaining size is returned to become the new size */
	bio_size = abd_bio_map_off(dr->dr_bio[i], zio->io_abd,
	bio_size, abd_offset);

	/* Advance in buffer and construct another bio if needed */
	abd_offset += BIO_BI_SIZE(dr->dr_bio[i]);
	bio_offset += BIO_BI_SIZE(dr->dr_bio[i]);
	}

	/* Extra reference to protect dio_request during vdev_submit_bio */
	vdev_disk_dio_get(dr);

	if (dr->dr_bio_count > 1)
	blk_start_plug(&plug);

	/* Submit all bio's associated with this dio */
	for (int i = 0; i < dr->dr_bio_count; i++) {
	if (dr->dr_bio[i])
	vdev_submit_bio(dr->dr_bio[i]);
	}

	if (dr->dr_bio_count > 1)
	blk_finish_plug(&plug);

	(void) vdev_disk_dio_put(dr);

	return (error);
	}

	BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, error)
	{
	zio_t *zio = bio->bi_private;
	#ifdef HAVE_1ARG_BIO_END_IO_T
	zio->io_error = BIO_END_IO_ERROR(bio);
	#else
	zio->io_error = -error;
	#endif

	if (zio->io_error && (zio->io_error == EOPNOTSUPP))
	zio->io_vd->vdev_nowritecache = B_TRUE;

	bio_put(bio);
	ASSERT3S(zio->io_error, >=, 0);
	if (zio->io_error)
	vdev_disk_error(zio);
	zio_interrupt(zio);
	}

	static int
	vdev_disk_io_flush(struct block_device bdev, zio_t zio)
	{
	struct request_queue *q;
	struct bio *bio;

	q = bdev_get_queue(bdev);
	if (!q)
	return (SET_ERROR(ENXIO));

	bio = vdev_bio_alloc(bdev, GFP_NOIO, 0);
	if (unlikely(bio == NULL))
	return (SET_ERROR(ENOMEM));

	bio->bi_end_io = vdev_disk_io_flush_completion;
	bio->bi_private = zio;
	bio_set_flush(bio);
	vdev_submit_bio(bio);
	invalidate_bdev(bdev);

	return (0);
	}

	static int
	vdev_disk_io_trim(zio_t *zio)
	{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;

	#if defined(HAVE_BLKDEV_ISSUE_SECURE_ERASE)
	if (zio->io_trim_flags & ZIO_TRIM_SECURE) {
	return (-blkdev_issue_secure_erase(vd->vd_bdev,
	zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS));
	} else {
	return (-blkdev_issue_discard(vd->vd_bdev,
	zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS));
	}
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD)
	unsigned long trim_flags = 0;
	#if defined(BLKDEV_DISCARD_SECURE)
	if (zio->io_trim_flags & ZIO_TRIM_SECURE)
	trim_flags \|= BLKDEV_DISCARD_SECURE;
	#endif
	return (-blkdev_issue_discard(vd->vd_bdev,
	zio->io_offset >> 9, zio->io_size >> 9, GFP_NOFS, trim_flags));
	#else
	#error "Unsupported kernel"
	#endif
	}

	static void
	vdev_disk_io_start(zio_t *zio)
	{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;
	int rw, error;

	/*
	* If the vdev is closed, it's likely in the REMOVED or FAULTED state.
	* Nothing to be done here but return failure.
	*/
	if (vd == NULL) {
	zio->io_error = ENXIO;
	zio_interrupt(zio);
	return;
	}

	rw_enter(&vd->vd_lock, RW_READER);

	/*
	* If the vdev is closed, it's likely due to a failed reopen and is
	* in the UNAVAIL state. Nothing to be done here but return failure.
	*/
	if (vd->vd_bdev == NULL) {
	rw_exit(&vd->vd_lock);
	zio->io_error = ENXIO;
	zio_interrupt(zio);
	return;
	}

	switch (zio->io_type) {
	case ZIO_TYPE_IOCTL:

	if (!vdev_readable(v)) {
	rw_exit(&vd->vd_lock);
	zio->io_error = SET_ERROR(ENXIO);
	zio_interrupt(zio);
	return;
	}

	switch (zio->io_cmd) {
	case DKIOCFLUSHWRITECACHE:

	if (zfs_nocacheflush)
	break;

	if (v->vdev_nowritecache) {
	zio->io_error = SET_ERROR(ENOTSUP);
	break;
	}

	error = vdev_disk_io_flush(vd->vd_bdev, zio);
	if (error == 0) {
	rw_exit(&vd->vd_lock);
	return;
	}

	zio->io_error = error;

	break;

	default:
	zio->io_error = SET_ERROR(ENOTSUP);
	}

	rw_exit(&vd->vd_lock);
	zio_execute(zio);
	return;
	case ZIO_TYPE_WRITE:
	rw = WRITE;
	break;

	case ZIO_TYPE_READ:
	rw = READ;
	break;

	case ZIO_TYPE_TRIM:
	zio->io_error = vdev_disk_io_trim(zio);
	rw_exit(&vd->vd_lock);
	zio_interrupt(zio);
	return;

	default:
	rw_exit(&vd->vd_lock);
	zio->io_error = SET_ERROR(ENOTSUP);
	zio_interrupt(zio);
	return;
	}

	zio->io_target_timestamp = zio_handle_io_delay(zio);
	error = __vdev_disk_physio(vd->vd_bdev, zio,
	zio->io_size, zio->io_offset, rw, 0);
	rw_exit(&vd->vd_lock);

	if (error) {
	zio->io_error = error;
	zio_interrupt(zio);
	return;
	}
	}

	static void
	vdev_disk_io_done(zio_t *zio)
	{
	/*
	* If the device returned EIO, we revalidate the media. If it is
	* determined the media has changed this triggers the asynchronous
	* removal of the device from the configuration.
	*/
	if (zio->io_error == EIO) {
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;

	- if (zfs_check_media_change(vd->vd_bdev)) {
	+ if (!zfs_check_disk_status(vd->vd_bdev)) {
	invalidate_bdev(vd->vd_bdev);
	v->vdev_remove_wanted = B_TRUE;
	spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
	}
	}
	}

	static void
	vdev_disk_hold(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* We must have a pathname, and it must be absolute. */
	if (vd->vdev_path == NULL \|\| vd->vdev_path[0] != '/')
	return;

	/*
	* Only prefetch path and devid info if the device has
	* never been opened.
	*/
	if (vd->vdev_tsd != NULL)
	return;

	}

	static void
	vdev_disk_rele(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* XXX: Implement me as a vnode rele for the device */
	}

	vdev_ops_t vdev_disk_ops = {
	.vdev_op_init = NULL,
	.vdev_op_fini = NULL,
	.vdev_op_open = vdev_disk_open,
	.vdev_op_close = vdev_disk_close,
	.vdev_op_asize = vdev_default_asize,
	.vdev_op_min_asize = vdev_default_min_asize,
	.vdev_op_min_alloc = NULL,
	.vdev_op_io_start = vdev_disk_io_start,
	.vdev_op_io_done = vdev_disk_io_done,
	.vdev_op_state_change = NULL,
	.vdev_op_need_resilver = NULL,
	.vdev_op_hold = vdev_disk_hold,
	.vdev_op_rele = vdev_disk_rele,
	.vdev_op_remap = NULL,
	.vdev_op_xlate = vdev_default_xlate,
	.vdev_op_rebuild_asize = NULL,
	.vdev_op_metaslab_init = NULL,
	.vdev_op_config_generate = NULL,
	.vdev_op_nparity = NULL,
	.vdev_op_ndisks = NULL,
	.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
	- .vdev_op_leaf = B_TRUE /* leaf vdev */
	+ .vdev_op_leaf = B_TRUE, /* leaf vdev */
	+ .vdev_op_kobj_evt_post = vdev_disk_kobj_evt_post
	};

	/*
	* The zfs_vdev_scheduler module option has been deprecated. Setting this
	* value no longer has any effect. It has not yet been entirely removed
	* to allow the module to be loaded if this option is specified in the
	* /etc/modprobe.d/zfs.conf file. The following warning will be logged.
	*/
	static int
	param_set_vdev_scheduler(const char val, zfs_kernel_param_t kp)
	{
	int error = param_set_charp(val, kp);
	if (error == 0) {
	printk(KERN_INFO "The 'zfs_vdev_scheduler' module option "
	"is not supported.\n");
	}

	return (error);
	}

	static const char *zfs_vdev_scheduler = "unused";
	module_param_call(zfs_vdev_scheduler, param_set_vdev_scheduler,
	param_get_charp, &zfs_vdev_scheduler, 0644);
	MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler");

	int
	param_set_min_auto_ashift(const char buf, zfs_kernel_param_t kp)
	{
	uint64_t val;
	int error;

	error = kstrtoull(buf, 0, &val);
	if (error < 0)
	return (SET_ERROR(error));

	if (val < ASHIFT_MIN \|\| val > zfs_vdev_max_auto_ashift)
	return (SET_ERROR(-EINVAL));

	error = param_set_ulong(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	return (0);
	}

	int
	param_set_max_auto_ashift(const char buf, zfs_kernel_param_t kp)
	{
	uint64_t val;
	int error;

	error = kstrtoull(buf, 0, &val);
	if (error < 0)
	return (SET_ERROR(error));

	if (val > ASHIFT_MAX \|\| val < zfs_vdev_min_auto_ashift)
	return (SET_ERROR(-EINVAL));

	error = param_set_ulong(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	return (0);
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c
	index 4fd071d3cb20..5935403b49d0 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c
	@@ -1,3019 +1,3019 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013 by Delphix. All rights reserved.
	* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
	*/


	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/sysmacros.h>
	#include <sys/vfs.h>
	#include <sys/vnode.h>
	#include <sys/sid.h>
	#include <sys/file.h>
	#include <sys/stat.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/errno.h>
	#include <sys/fs/zfs.h>
	#include <sys/policy.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_fuid.h>
	#include <sys/zfs_acl.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/dmu.h>
	#include <sys/dnode.h>
	#include <sys/zap.h>
	#include <sys/sa.h>
	#include <sys/trace_acl.h>
	#include <sys/zpl.h>

	#define ALLOW ACE_ACCESS_ALLOWED_ACE_TYPE
	#define DENY ACE_ACCESS_DENIED_ACE_TYPE
	#define MAX_ACE_TYPE ACE_SYSTEM_ALARM_CALLBACK_OBJECT_ACE_TYPE
	#define MIN_ACE_TYPE ALLOW

	#define OWNING_GROUP (ACE_GROUP\|ACE_IDENTIFIER_GROUP)
	#define EVERYONE_ALLOW_MASK (ACE_READ_ACL\|ACE_READ_ATTRIBUTES \| \
	ACE_READ_NAMED_ATTRS\|ACE_SYNCHRONIZE)
	#define EVERYONE_DENY_MASK (ACE_WRITE_ACL\|ACE_WRITE_OWNER \| \
	ACE_WRITE_ATTRIBUTES\|ACE_WRITE_NAMED_ATTRS)
	#define OWNER_ALLOW_MASK (ACE_WRITE_ACL \| ACE_WRITE_OWNER \| \
	ACE_WRITE_ATTRIBUTES\|ACE_WRITE_NAMED_ATTRS)

	#define ZFS_CHECKED_MASKS (ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_READ_DATA\| \
	ACE_READ_NAMED_ATTRS\|ACE_WRITE_DATA\|ACE_WRITE_ATTRIBUTES\| \
	ACE_WRITE_NAMED_ATTRS\|ACE_APPEND_DATA\|ACE_EXECUTE\|ACE_WRITE_OWNER\| \
	ACE_WRITE_ACL\|ACE_DELETE\|ACE_DELETE_CHILD\|ACE_SYNCHRONIZE)

	#define WRITE_MASK_DATA (ACE_WRITE_DATA\|ACE_APPEND_DATA\|ACE_WRITE_NAMED_ATTRS)
	#define WRITE_MASK_ATTRS (ACE_WRITE_ACL\|ACE_WRITE_OWNER\|ACE_WRITE_ATTRIBUTES\| \
	ACE_DELETE\|ACE_DELETE_CHILD)
	#define WRITE_MASK (WRITE_MASK_DATA\|WRITE_MASK_ATTRS)

	#define OGE_CLEAR (ACE_READ_DATA\|ACE_LIST_DIRECTORY\|ACE_WRITE_DATA\| \
	ACE_ADD_FILE\|ACE_APPEND_DATA\|ACE_ADD_SUBDIRECTORY\|ACE_EXECUTE)

	#define OKAY_MASK_BITS (ACE_READ_DATA\|ACE_LIST_DIRECTORY\|ACE_WRITE_DATA\| \
	ACE_ADD_FILE\|ACE_APPEND_DATA\|ACE_ADD_SUBDIRECTORY\|ACE_EXECUTE)

	#define ALL_INHERIT (ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE \| \
	ACE_NO_PROPAGATE_INHERIT_ACE\|ACE_INHERIT_ONLY_ACE\|ACE_INHERITED_ACE)

	#define RESTRICTED_CLEAR (ACE_WRITE_ACL\|ACE_WRITE_OWNER)

	#define V4_ACL_WIDE_FLAGS (ZFS_ACL_AUTO_INHERIT\|ZFS_ACL_DEFAULTED\|\
	ZFS_ACL_PROTECTED)

	#define ZFS_ACL_WIDE_FLAGS (V4_ACL_WIDE_FLAGS\|ZFS_ACL_TRIVIAL\|ZFS_INHERIT_ACE\|\
	ZFS_ACL_OBJ_ACE)

	#define ALL_MODE_EXECS (S_IXUSR \| S_IXGRP \| S_IXOTH)

	#define IDMAP_WK_CREATOR_OWNER_UID 2147483648U

	static uint16_t
	zfs_ace_v0_get_type(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_type);
	}

	static uint16_t
	zfs_ace_v0_get_flags(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_flags);
	}

	static uint32_t
	zfs_ace_v0_get_mask(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_access_mask);
	}

	static uint64_t
	zfs_ace_v0_get_who(void *acep)
	{
	return (((zfs_oldace_t *)acep)->z_fuid);
	}

	static void
	zfs_ace_v0_set_type(void *acep, uint16_t type)
	{
	((zfs_oldace_t *)acep)->z_type = type;
	}

	static void
	zfs_ace_v0_set_flags(void *acep, uint16_t flags)
	{
	((zfs_oldace_t *)acep)->z_flags = flags;
	}

	static void
	zfs_ace_v0_set_mask(void *acep, uint32_t mask)
	{
	((zfs_oldace_t *)acep)->z_access_mask = mask;
	}

	static void
	zfs_ace_v0_set_who(void *acep, uint64_t who)
	{
	((zfs_oldace_t *)acep)->z_fuid = who;
	}

	static size_t
	zfs_ace_v0_size(void *acep)
	{
	(void) acep;
	return (sizeof (zfs_oldace_t));
	}

	static size_t
	zfs_ace_v0_abstract_size(void)
	{
	return (sizeof (zfs_oldace_t));
	}

	static int
	zfs_ace_v0_mask_off(void)
	{
	return (offsetof(zfs_oldace_t, z_access_mask));
	}

	static int
	zfs_ace_v0_data(void acep, void *datap)
	{
	(void) acep;
	*datap = NULL;
	return (0);
	}

	static const acl_ops_t zfs_acl_v0_ops = {
	.ace_mask_get = zfs_ace_v0_get_mask,
	.ace_mask_set = zfs_ace_v0_set_mask,
	.ace_flags_get = zfs_ace_v0_get_flags,
	.ace_flags_set = zfs_ace_v0_set_flags,
	.ace_type_get = zfs_ace_v0_get_type,
	.ace_type_set = zfs_ace_v0_set_type,
	.ace_who_get = zfs_ace_v0_get_who,
	.ace_who_set = zfs_ace_v0_set_who,
	.ace_size = zfs_ace_v0_size,
	.ace_abstract_size = zfs_ace_v0_abstract_size,
	.ace_mask_off = zfs_ace_v0_mask_off,
	.ace_data = zfs_ace_v0_data
	};

	static uint16_t
	zfs_ace_fuid_get_type(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_type);
	}

	static uint16_t
	zfs_ace_fuid_get_flags(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_flags);
	}

	static uint32_t
	zfs_ace_fuid_get_mask(void *acep)
	{
	return (((zfs_ace_hdr_t *)acep)->z_access_mask);
	}

	static uint64_t
	zfs_ace_fuid_get_who(void *args)
	{
	uint16_t entry_type;
	zfs_ace_t *acep = args;

	entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;

	if (entry_type == ACE_OWNER \|\| entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return (-1);
	return (((zfs_ace_t *)acep)->z_fuid);
	}

	static void
	zfs_ace_fuid_set_type(void *acep, uint16_t type)
	{
	((zfs_ace_hdr_t *)acep)->z_type = type;
	}

	static void
	zfs_ace_fuid_set_flags(void *acep, uint16_t flags)
	{
	((zfs_ace_hdr_t *)acep)->z_flags = flags;
	}

	static void
	zfs_ace_fuid_set_mask(void *acep, uint32_t mask)
	{
	((zfs_ace_hdr_t *)acep)->z_access_mask = mask;
	}

	static void
	zfs_ace_fuid_set_who(void *arg, uint64_t who)
	{
	zfs_ace_t *acep = arg;

	uint16_t entry_type = acep->z_hdr.z_flags & ACE_TYPE_FLAGS;

	if (entry_type == ACE_OWNER \|\| entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return;
	acep->z_fuid = who;
	}

	static size_t
	zfs_ace_fuid_size(void *acep)
	{
	zfs_ace_hdr_t *zacep = acep;
	uint16_t entry_type;

	switch (zacep->z_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	return (sizeof (zfs_object_ace_t));
	case ALLOW:
	case DENY:
	entry_type =
	(((zfs_ace_hdr_t *)acep)->z_flags & ACE_TYPE_FLAGS);
	if (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)
	return (sizeof (zfs_ace_hdr_t));
	zfs_fallthrough;
	default:
	return (sizeof (zfs_ace_t));
	}
	}

	static size_t
	zfs_ace_fuid_abstract_size(void)
	{
	return (sizeof (zfs_ace_hdr_t));
	}

	static int
	zfs_ace_fuid_mask_off(void)
	{
	return (offsetof(zfs_ace_hdr_t, z_access_mask));
	}

	static int
	zfs_ace_fuid_data(void acep, void *datap)
	{
	zfs_ace_t *zacep = acep;
	zfs_object_ace_t *zobjp;

	switch (zacep->z_hdr.z_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	zobjp = acep;
	*datap = (caddr_t)zobjp + sizeof (zfs_ace_t);
	return (sizeof (zfs_object_ace_t) - sizeof (zfs_ace_t));
	default:
	*datap = NULL;
	return (0);
	}
	}

	static const acl_ops_t zfs_acl_fuid_ops = {
	.ace_mask_get = zfs_ace_fuid_get_mask,
	.ace_mask_set = zfs_ace_fuid_set_mask,
	.ace_flags_get = zfs_ace_fuid_get_flags,
	.ace_flags_set = zfs_ace_fuid_set_flags,
	.ace_type_get = zfs_ace_fuid_get_type,
	.ace_type_set = zfs_ace_fuid_set_type,
	.ace_who_get = zfs_ace_fuid_get_who,
	.ace_who_set = zfs_ace_fuid_set_who,
	.ace_size = zfs_ace_fuid_size,
	.ace_abstract_size = zfs_ace_fuid_abstract_size,
	.ace_mask_off = zfs_ace_fuid_mask_off,
	.ace_data = zfs_ace_fuid_data
	};

	/*
	* The following three functions are provided for compatibility with
	* older ZPL version in order to determine if the file use to have
	* an external ACL and what version of ACL previously existed on the
	* file. Would really be nice to not need this, sigh.
	*/
	uint64_t
	zfs_external_acl(znode_t *zp)
	{
	zfs_acl_phys_t acl_phys;
	int error;

	if (zp->z_is_sa)
	return (0);

	/*
	* Need to deal with a potential
	* race where zfs_sa_upgrade could cause
	* z_isa_sa to change.
	*
	* If the lookup fails then the state of z_is_sa should have
	* changed.
	*/

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
	&acl_phys, sizeof (acl_phys))) == 0)
	return (acl_phys.z_acl_extern_obj);
	else {
	/*
	* after upgrade the SA_ZPL_ZNODE_ACL should have been
	* removed
	*/
	VERIFY(zp->z_is_sa && error == ENOENT);
	return (0);
	}
	}

	/*
	* Determine size of ACL in bytes
	*
	* This is more complicated than it should be since we have to deal
	* with old external ACLs.
	*/
	static int
	zfs_acl_znode_info(znode_t zp, int aclsize, int *aclcount,
	zfs_acl_phys_t *aclphys)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	uint64_t acl_count;
	int size;
	int error;

	ASSERT(MUTEX_HELD(&zp->z_acl_lock));
	if (zp->z_is_sa) {
	if ((error = sa_size(zp->z_sa_hdl, SA_ZPL_DACL_ACES(zfsvfs),
	&size)) != 0)
	return (error);
	*aclsize = size;
	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_COUNT(zfsvfs),
	&acl_count, sizeof (acl_count))) != 0)
	return (error);
	*aclcount = acl_count;
	} else {
	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
	aclphys, sizeof (*aclphys))) != 0)
	return (error);

	if (aclphys->z_acl_version == ZFS_ACL_VERSION_INITIAL) {
	*aclsize = ZFS_ACL_SIZE(aclphys->z_acl_size);
	*aclcount = aclphys->z_acl_size;
	} else {
	*aclsize = aclphys->z_acl_size;
	*aclcount = aclphys->z_acl_count;
	}
	}
	return (0);
	}

	int
	zfs_znode_acl_version(znode_t *zp)
	{
	zfs_acl_phys_t acl_phys;

	if (zp->z_is_sa)
	return (ZFS_ACL_VERSION_FUID);
	else {
	int error;

	/*
	* Need to deal with a potential
	* race where zfs_sa_upgrade could cause
	* z_isa_sa to change.
	*
	* If the lookup fails then the state of z_is_sa should have
	* changed.
	*/
	if ((error = sa_lookup(zp->z_sa_hdl,
	SA_ZPL_ZNODE_ACL(ZTOZSB(zp)),
	&acl_phys, sizeof (acl_phys))) == 0)
	return (acl_phys.z_acl_version);
	else {
	/*
	* After upgrade SA_ZPL_ZNODE_ACL should have
	* been removed.
	*/
	VERIFY(zp->z_is_sa && error == ENOENT);
	return (ZFS_ACL_VERSION_FUID);
	}
	}
	}

	static int
	zfs_acl_version(int version)
	{
	if (version < ZPL_VERSION_FUID)
	return (ZFS_ACL_VERSION_INITIAL);
	else
	return (ZFS_ACL_VERSION_FUID);
	}

	static int
	zfs_acl_version_zp(znode_t *zp)
	{
	return (zfs_acl_version(ZTOZSB(zp)->z_version));
	}

	zfs_acl_t *
	zfs_acl_alloc(int vers)
	{
	zfs_acl_t *aclp;

	aclp = kmem_zalloc(sizeof (zfs_acl_t), KM_SLEEP);
	list_create(&aclp->z_acl, sizeof (zfs_acl_node_t),
	offsetof(zfs_acl_node_t, z_next));
	aclp->z_version = vers;
	if (vers == ZFS_ACL_VERSION_FUID)
	aclp->z_ops = &zfs_acl_fuid_ops;
	else
	aclp->z_ops = &zfs_acl_v0_ops;
	return (aclp);
	}

	zfs_acl_node_t *
	zfs_acl_node_alloc(size_t bytes)
	{
	zfs_acl_node_t *aclnode;

	aclnode = kmem_zalloc(sizeof (zfs_acl_node_t), KM_SLEEP);
	if (bytes) {
	aclnode->z_acldata = kmem_alloc(bytes, KM_SLEEP);
	aclnode->z_allocdata = aclnode->z_acldata;
	aclnode->z_allocsize = bytes;
	aclnode->z_size = bytes;
	}

	return (aclnode);
	}

	static void
	zfs_acl_node_free(zfs_acl_node_t *aclnode)
	{
	if (aclnode->z_allocsize)
	kmem_free(aclnode->z_allocdata, aclnode->z_allocsize);
	kmem_free(aclnode, sizeof (zfs_acl_node_t));
	}

	static void
	zfs_acl_release_nodes(zfs_acl_t *aclp)
	{
	zfs_acl_node_t *aclnode;

	while ((aclnode = list_head(&aclp->z_acl))) {
	list_remove(&aclp->z_acl, aclnode);
	zfs_acl_node_free(aclnode);
	}
	aclp->z_acl_count = 0;
	aclp->z_acl_bytes = 0;
	}

	void
	zfs_acl_free(zfs_acl_t *aclp)
	{
	zfs_acl_release_nodes(aclp);
	list_destroy(&aclp->z_acl);
	kmem_free(aclp, sizeof (zfs_acl_t));
	}

	static boolean_t
	zfs_acl_valid_ace_type(uint_t type, uint_t flags)
	{
	uint16_t entry_type;

	switch (type) {
	case ALLOW:
	case DENY:
	case ACE_SYSTEM_AUDIT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_ACE_TYPE:
	entry_type = flags & ACE_TYPE_FLAGS;
	return (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE \|\| entry_type == 0 \|\|
	entry_type == ACE_IDENTIFIER_GROUP);
	default:
	- if (type >= MIN_ACE_TYPE && type <= MAX_ACE_TYPE)
	+ if (type <= MAX_ACE_TYPE)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static boolean_t
	zfs_ace_valid(umode_t obj_mode, zfs_acl_t *aclp, uint16_t type, uint16_t iflags)
	{
	/*
	* first check type of entry
	*/

	if (!zfs_acl_valid_ace_type(type, iflags))
	return (B_FALSE);

	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	if (aclp->z_version < ZFS_ACL_VERSION_FUID)
	return (B_FALSE);
	aclp->z_hints \|= ZFS_ACL_OBJ_ACE;
	}

	/*
	* next check inheritance level flags
	*/

	if (S_ISDIR(obj_mode) &&
	(iflags & (ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	if (iflags & (ACE_INHERIT_ONLY_ACE\|ACE_NO_PROPAGATE_INHERIT_ACE)) {
	if ((iflags & (ACE_FILE_INHERIT_ACE\|
	ACE_DIRECTORY_INHERIT_ACE)) == 0) {
	return (B_FALSE);
	}
	}

	return (B_TRUE);
	}

	static void *
	zfs_acl_next_ace(zfs_acl_t aclp, void start, uint64_t *who,
	uint32_t access_mask, uint16_t iflags, uint16_t *type)
	{
	zfs_acl_node_t *aclnode;

	ASSERT(aclp);

	if (start == NULL) {
	aclnode = list_head(&aclp->z_acl);
	if (aclnode == NULL)
	return (NULL);

	aclp->z_next_ace = aclnode->z_acldata;
	aclp->z_curr_node = aclnode;
	aclnode->z_ace_idx = 0;
	}

	aclnode = aclp->z_curr_node;

	if (aclnode == NULL)
	return (NULL);

	if (aclnode->z_ace_idx >= aclnode->z_ace_count) {
	aclnode = list_next(&aclp->z_acl, aclnode);
	if (aclnode == NULL)
	return (NULL);
	else {
	aclp->z_curr_node = aclnode;
	aclnode->z_ace_idx = 0;
	aclp->z_next_ace = aclnode->z_acldata;
	}
	}

	if (aclnode->z_ace_idx < aclnode->z_ace_count) {
	void *acep = aclp->z_next_ace;
	size_t ace_size;

	/*
	* Make sure we don't overstep our bounds
	*/
	ace_size = aclp->z_ops->ace_size(acep);

	if (((caddr_t)acep + ace_size) >
	((caddr_t)aclnode->z_acldata + aclnode->z_size)) {
	return (NULL);
	}

	*iflags = aclp->z_ops->ace_flags_get(acep);
	*type = aclp->z_ops->ace_type_get(acep);
	*access_mask = aclp->z_ops->ace_mask_get(acep);
	*who = aclp->z_ops->ace_who_get(acep);
	aclp->z_next_ace = (caddr_t)aclp->z_next_ace + ace_size;
	aclnode->z_ace_idx++;

	return ((void *)acep);
	}
	return (NULL);
	}

	static uint64_t
	zfs_ace_walk(void *datap, uint64_t cookie, int aclcnt,
	uint16_t flags, uint16_t type, uint32_t *mask)
	{
	(void) aclcnt;
	zfs_acl_t *aclp = datap;
	zfs_ace_hdr_t acep = (zfs_ace_hdr_t )(uintptr_t)cookie;
	uint64_t who;

	acep = zfs_acl_next_ace(aclp, acep, &who, mask,
	flags, type);
	return ((uint64_t)(uintptr_t)acep);
	}

	/*
	* Copy ACE to internal ZFS format.
	* While processing the ACL each ACE will be validated for correctness.
	* ACE FUIDs will be created later.
	*/
	static int
	zfs_copy_ace_2_fuid(zfsvfs_t zfsvfs, umode_t obj_mode, zfs_acl_t aclp,
	void datap, zfs_ace_t z_acl, uint64_t aclcnt, size_t *size,
	zfs_fuid_info_t *fuidp, cred_t cr)
	{
	int i;
	uint16_t entry_type;
	zfs_ace_t *aceptr = z_acl;
	ace_t *acep = datap;
	zfs_object_ace_t *zobjacep;
	ace_object_t *aceobjp;

	for (i = 0; i != aclcnt; i++) {
	aceptr->z_hdr.z_access_mask = acep->a_access_mask;
	aceptr->z_hdr.z_flags = acep->a_flags;
	aceptr->z_hdr.z_type = acep->a_type;
	entry_type = aceptr->z_hdr.z_flags & ACE_TYPE_FLAGS;
	if (entry_type != ACE_OWNER && entry_type != OWNING_GROUP &&
	entry_type != ACE_EVERYONE) {
	aceptr->z_fuid = zfs_fuid_create(zfsvfs, acep->a_who,
	cr, (entry_type == 0) ?
	ZFS_ACE_USER : ZFS_ACE_GROUP, fuidp);
	}

	/*
	* Make sure ACE is valid
	*/
	if (zfs_ace_valid(obj_mode, aclp, aceptr->z_hdr.z_type,
	aceptr->z_hdr.z_flags) != B_TRUE)
	return (SET_ERROR(EINVAL));

	switch (acep->a_type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	zobjacep = (zfs_object_ace_t *)aceptr;
	aceobjp = (ace_object_t *)acep;

	memcpy(zobjacep->z_object_type, aceobjp->a_obj_type,
	sizeof (aceobjp->a_obj_type));
	memcpy(zobjacep->z_inherit_type,
	aceobjp->a_inherit_obj_type,
	sizeof (aceobjp->a_inherit_obj_type));
	acep = (ace_t *)((caddr_t)acep + sizeof (ace_object_t));
	break;
	default:
	acep = (ace_t *)((caddr_t)acep + sizeof (ace_t));
	}

	aceptr = (zfs_ace_t *)((caddr_t)aceptr +
	aclp->z_ops->ace_size(aceptr));
	}

	*size = (caddr_t)aceptr - (caddr_t)z_acl;

	return (0);
	}

	/*
	* Copy ZFS ACEs to fixed size ace_t layout
	*/
	static void
	zfs_copy_fuid_2_ace(zfsvfs_t zfsvfs, zfs_acl_t aclp, cred_t *cr,
	void *datap, int filter)
	{
	uint64_t who;
	uint32_t access_mask;
	uint16_t iflags, type;
	zfs_ace_hdr_t *zacep = NULL;
	ace_t *acep = datap;
	ace_object_t *objacep;
	zfs_object_ace_t *zobjacep;
	size_t ace_size;
	uint16_t entry_type;

	while ((zacep = zfs_acl_next_ace(aclp, zacep,
	&who, &access_mask, &iflags, &type))) {

	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	if (filter) {
	continue;
	}
	zobjacep = (zfs_object_ace_t *)zacep;
	objacep = (ace_object_t *)acep;
	memcpy(objacep->a_obj_type,
	zobjacep->z_object_type,
	sizeof (zobjacep->z_object_type));
	memcpy(objacep->a_inherit_obj_type,
	zobjacep->z_inherit_type,
	sizeof (zobjacep->z_inherit_type));
	ace_size = sizeof (ace_object_t);
	break;
	default:
	ace_size = sizeof (ace_t);
	break;
	}

	entry_type = (iflags & ACE_TYPE_FLAGS);
	if ((entry_type != ACE_OWNER &&
	entry_type != OWNING_GROUP &&
	entry_type != ACE_EVERYONE)) {
	acep->a_who = zfs_fuid_map_id(zfsvfs, who,
	cr, (entry_type & ACE_IDENTIFIER_GROUP) ?
	ZFS_ACE_GROUP : ZFS_ACE_USER);
	} else {
	acep->a_who = (uid_t)(int64_t)who;
	}
	acep->a_access_mask = access_mask;
	acep->a_flags = iflags;
	acep->a_type = type;
	acep = (ace_t *)((caddr_t)acep + ace_size);
	}
	}

	static int
	zfs_copy_ace_2_oldace(umode_t obj_mode, zfs_acl_t aclp, ace_t acep,
	zfs_oldace_t z_acl, int aclcnt, size_t size)
	{
	int i;
	zfs_oldace_t *aceptr = z_acl;

	for (i = 0; i != aclcnt; i++, aceptr++) {
	aceptr->z_access_mask = acep[i].a_access_mask;
	aceptr->z_type = acep[i].a_type;
	aceptr->z_flags = acep[i].a_flags;
	aceptr->z_fuid = acep[i].a_who;
	/*
	* Make sure ACE is valid
	*/
	if (zfs_ace_valid(obj_mode, aclp, aceptr->z_type,
	aceptr->z_flags) != B_TRUE)
	return (SET_ERROR(EINVAL));
	}
	*size = (caddr_t)aceptr - (caddr_t)z_acl;
	return (0);
	}

	/*
	* convert old ACL format to new
	*/
	void
	zfs_acl_xform(znode_t zp, zfs_acl_t aclp, cred_t *cr)
	{
	zfs_oldace_t *oldaclp;
	int i;
	uint16_t type, iflags;
	uint32_t access_mask;
	uint64_t who;
	void *cookie = NULL;
	zfs_acl_node_t *newaclnode;

	ASSERT(aclp->z_version == ZFS_ACL_VERSION_INITIAL);
	/*
	* First create the ACE in a contiguous piece of memory
	* for zfs_copy_ace_2_fuid().
	*
	* We only convert an ACL once, so this won't happen
	* every time.
	*/
	oldaclp = kmem_alloc(sizeof (zfs_oldace_t) * aclp->z_acl_count,
	KM_SLEEP);
	i = 0;
	while ((cookie = zfs_acl_next_ace(aclp, cookie, &who,
	&access_mask, &iflags, &type))) {
	oldaclp[i].z_flags = iflags;
	oldaclp[i].z_type = type;
	oldaclp[i].z_fuid = who;
	oldaclp[i++].z_access_mask = access_mask;
	}

	newaclnode = zfs_acl_node_alloc(aclp->z_acl_count *
	sizeof (zfs_object_ace_t));
	aclp->z_ops = &zfs_acl_fuid_ops;
	VERIFY(zfs_copy_ace_2_fuid(ZTOZSB(zp), ZTOI(zp)->i_mode,
	aclp, oldaclp, newaclnode->z_acldata, aclp->z_acl_count,
	&newaclnode->z_size, NULL, cr) == 0);
	newaclnode->z_ace_count = aclp->z_acl_count;
	aclp->z_version = ZFS_ACL_VERSION;
	kmem_free(oldaclp, aclp->z_acl_count * sizeof (zfs_oldace_t));

	/*
	* Release all previous ACL nodes
	*/

	zfs_acl_release_nodes(aclp);

	list_insert_head(&aclp->z_acl, newaclnode);

	aclp->z_acl_bytes = newaclnode->z_size;
	aclp->z_acl_count = newaclnode->z_ace_count;

	}

	/*
	* Convert unix access mask to v4 access mask
	*/
	static uint32_t
	zfs_unix_to_v4(uint32_t access_mask)
	{
	uint32_t new_mask = 0;

	if (access_mask & S_IXOTH)
	new_mask \|= ACE_EXECUTE;
	if (access_mask & S_IWOTH)
	new_mask \|= ACE_WRITE_DATA;
	if (access_mask & S_IROTH)
	new_mask \|= ACE_READ_DATA;
	return (new_mask);
	}


	static int
	zfs_v4_to_unix(uint32_t access_mask, int *unmapped)
	{
	int new_mask = 0;

	*unmapped = access_mask &
	(ACE_WRITE_OWNER \| ACE_WRITE_ACL \| ACE_DELETE);

	if (access_mask & WRITE_MASK)
	new_mask \|= S_IWOTH;
	if (access_mask & ACE_READ_DATA)
	new_mask \|= S_IROTH;
	if (access_mask & ACE_EXECUTE)
	new_mask \|= S_IXOTH;

	return (new_mask);
	}


	static void
	zfs_set_ace(zfs_acl_t aclp, void acep, uint32_t access_mask,
	uint16_t access_type, uint64_t fuid, uint16_t entry_type)
	{
	uint16_t type = entry_type & ACE_TYPE_FLAGS;

	aclp->z_ops->ace_mask_set(acep, access_mask);
	aclp->z_ops->ace_type_set(acep, access_type);
	aclp->z_ops->ace_flags_set(acep, entry_type);
	if ((type != ACE_OWNER && type != OWNING_GROUP &&
	type != ACE_EVERYONE))
	aclp->z_ops->ace_who_set(acep, fuid);
	}

	/*
	* Determine mode of file based on ACL.
	*/
	uint64_t
	zfs_mode_compute(uint64_t fmode, zfs_acl_t *aclp,
	uint64_t *pflags, uint64_t fuid, uint64_t fgid)
	{
	int entry_type;
	mode_t mode;
	mode_t seen = 0;
	zfs_ace_hdr_t *acep = NULL;
	uint64_t who;
	uint16_t iflags, type;
	uint32_t access_mask;
	boolean_t an_exec_denied = B_FALSE;

	mode = (fmode & (S_IFMT \| S_ISUID \| S_ISGID \| S_ISVTX));

	while ((acep = zfs_acl_next_ace(aclp, acep, &who,
	&access_mask, &iflags, &type))) {

	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	entry_type = (iflags & ACE_TYPE_FLAGS);

	/*
	* Skip over any inherit_only ACEs
	*/
	if (iflags & ACE_INHERIT_ONLY_ACE)
	continue;

	if (entry_type == ACE_OWNER \|\| (entry_type == 0 &&
	who == fuid)) {
	if ((access_mask & ACE_READ_DATA) &&
	(!(seen & S_IRUSR))) {
	seen \|= S_IRUSR;
	if (type == ALLOW) {
	mode \|= S_IRUSR;
	}
	}
	if ((access_mask & ACE_WRITE_DATA) &&
	(!(seen & S_IWUSR))) {
	seen \|= S_IWUSR;
	if (type == ALLOW) {
	mode \|= S_IWUSR;
	}
	}
	if ((access_mask & ACE_EXECUTE) &&
	(!(seen & S_IXUSR))) {
	seen \|= S_IXUSR;
	if (type == ALLOW) {
	mode \|= S_IXUSR;
	}
	}
	} else if (entry_type == OWNING_GROUP \|\|
	(entry_type == ACE_IDENTIFIER_GROUP && who == fgid)) {
	if ((access_mask & ACE_READ_DATA) &&
	(!(seen & S_IRGRP))) {
	seen \|= S_IRGRP;
	if (type == ALLOW) {
	mode \|= S_IRGRP;
	}
	}
	if ((access_mask & ACE_WRITE_DATA) &&
	(!(seen & S_IWGRP))) {
	seen \|= S_IWGRP;
	if (type == ALLOW) {
	mode \|= S_IWGRP;
	}
	}
	if ((access_mask & ACE_EXECUTE) &&
	(!(seen & S_IXGRP))) {
	seen \|= S_IXGRP;
	if (type == ALLOW) {
	mode \|= S_IXGRP;
	}
	}
	} else if (entry_type == ACE_EVERYONE) {
	if ((access_mask & ACE_READ_DATA)) {
	if (!(seen & S_IRUSR)) {
	seen \|= S_IRUSR;
	if (type == ALLOW) {
	mode \|= S_IRUSR;
	}
	}
	if (!(seen & S_IRGRP)) {
	seen \|= S_IRGRP;
	if (type == ALLOW) {
	mode \|= S_IRGRP;
	}
	}
	if (!(seen & S_IROTH)) {
	seen \|= S_IROTH;
	if (type == ALLOW) {
	mode \|= S_IROTH;
	}
	}
	}
	if ((access_mask & ACE_WRITE_DATA)) {
	if (!(seen & S_IWUSR)) {
	seen \|= S_IWUSR;
	if (type == ALLOW) {
	mode \|= S_IWUSR;
	}
	}
	if (!(seen & S_IWGRP)) {
	seen \|= S_IWGRP;
	if (type == ALLOW) {
	mode \|= S_IWGRP;
	}
	}
	if (!(seen & S_IWOTH)) {
	seen \|= S_IWOTH;
	if (type == ALLOW) {
	mode \|= S_IWOTH;
	}
	}
	}
	if ((access_mask & ACE_EXECUTE)) {
	if (!(seen & S_IXUSR)) {
	seen \|= S_IXUSR;
	if (type == ALLOW) {
	mode \|= S_IXUSR;
	}
	}
	if (!(seen & S_IXGRP)) {
	seen \|= S_IXGRP;
	if (type == ALLOW) {
	mode \|= S_IXGRP;
	}
	}
	if (!(seen & S_IXOTH)) {
	seen \|= S_IXOTH;
	if (type == ALLOW) {
	mode \|= S_IXOTH;
	}
	}
	}
	} else {
	/*
	* Only care if this IDENTIFIER_GROUP or
	* USER ACE denies execute access to someone,
	* mode is not affected
	*/
	if ((access_mask & ACE_EXECUTE) && type == DENY)
	an_exec_denied = B_TRUE;
	}
	}

	/*
	* Failure to allow is effectively a deny, so execute permission
	* is denied if it was never mentioned or if we explicitly
	* weren't allowed it.
	*/
	if (!an_exec_denied &&
	((seen & ALL_MODE_EXECS) != ALL_MODE_EXECS \|\|
	(mode & ALL_MODE_EXECS) != ALL_MODE_EXECS))
	an_exec_denied = B_TRUE;

	if (an_exec_denied)
	*pflags &= ~ZFS_NO_EXECS_DENIED;
	else
	*pflags \|= ZFS_NO_EXECS_DENIED;

	return (mode);
	}

	/*
	* Read an external acl object. If the intent is to modify, always
	* create a new acl and leave any cached acl in place.
	*/
	int
	zfs_acl_node_read(struct znode zp, boolean_t have_lock, zfs_acl_t *aclpp,
	boolean_t will_modify)
	{
	zfs_acl_t *aclp;
	int aclsize = 0;
	int acl_count = 0;
	zfs_acl_node_t *aclnode;
	zfs_acl_phys_t znode_acl;
	int version;
	int error;
	boolean_t drop_lock = B_FALSE;

	ASSERT(MUTEX_HELD(&zp->z_acl_lock));

	if (zp->z_acl_cached && !will_modify) {
	*aclpp = zp->z_acl_cached;
	return (0);
	}

	/*
	* close race where znode could be upgrade while trying to
	* read the znode attributes.
	*
	* But this could only happen if the file isn't already an SA
	* znode
	*/
	if (!zp->z_is_sa && !have_lock) {
	mutex_enter(&zp->z_lock);
	drop_lock = B_TRUE;
	}
	version = zfs_znode_acl_version(zp);

	if ((error = zfs_acl_znode_info(zp, &aclsize,
	&acl_count, &znode_acl)) != 0) {
	goto done;
	}

	aclp = zfs_acl_alloc(version);

	aclp->z_acl_count = acl_count;
	aclp->z_acl_bytes = aclsize;

	aclnode = zfs_acl_node_alloc(aclsize);
	aclnode->z_ace_count = aclp->z_acl_count;
	aclnode->z_size = aclsize;

	if (!zp->z_is_sa) {
	if (znode_acl.z_acl_extern_obj) {
	error = dmu_read(ZTOZSB(zp)->z_os,
	znode_acl.z_acl_extern_obj, 0, aclnode->z_size,
	aclnode->z_acldata, DMU_READ_PREFETCH);
	} else {
	memcpy(aclnode->z_acldata, znode_acl.z_ace_data,
	aclnode->z_size);
	}
	} else {
	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DACL_ACES(ZTOZSB(zp)),
	aclnode->z_acldata, aclnode->z_size);
	}

	if (error != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	/* convert checksum errors into IO errors */
	if (error == ECKSUM)
	error = SET_ERROR(EIO);
	goto done;
	}

	list_insert_head(&aclp->z_acl, aclnode);

	*aclpp = aclp;
	if (!will_modify)
	zp->z_acl_cached = aclp;
	done:
	if (drop_lock)
	mutex_exit(&zp->z_lock);
	return (error);
	}

	void
	zfs_acl_data_locator(void *dataptr, uint32_t length, uint32_t buflen,
	boolean_t start, void *userdata)
	{
	(void) buflen;
	zfs_acl_locator_cb_t cb = (zfs_acl_locator_cb_t )userdata;

	if (start) {
	cb->cb_acl_node = list_head(&cb->cb_aclp->z_acl);
	} else {
	cb->cb_acl_node = list_next(&cb->cb_aclp->z_acl,
	cb->cb_acl_node);
	}
	*dataptr = cb->cb_acl_node->z_acldata;
	*length = cb->cb_acl_node->z_size;
	}

	int
	zfs_acl_chown_setattr(znode_t *zp)
	{
	int error;
	zfs_acl_t *aclp;

	if (ZTOZSB(zp)->z_acl_type == ZFS_ACLTYPE_POSIX)
	return (0);

	ASSERT(MUTEX_HELD(&zp->z_lock));
	ASSERT(MUTEX_HELD(&zp->z_acl_lock));

	error = zfs_acl_node_read(zp, B_TRUE, &aclp, B_FALSE);
	if (error == 0 && aclp->z_acl_count > 0)
	zp->z_mode = ZTOI(zp)->i_mode =
	zfs_mode_compute(zp->z_mode, aclp,
	&zp->z_pflags, KUID_TO_SUID(ZTOI(zp)->i_uid),
	KGID_TO_SGID(ZTOI(zp)->i_gid));

	/*
	* Some ZFS implementations (ZEVO) create neither a ZNODE_ACL
	* nor a DACL_ACES SA in which case ENOENT is returned from
	* zfs_acl_node_read() when the SA can't be located.
	* Allow chown/chgrp to succeed in these cases rather than
	* returning an error that makes no sense in the context of
	* the caller.
	*/
	if (error == ENOENT)
	return (0);

	return (error);
	}

	typedef struct trivial_acl {
	uint32_t allow0; /* allow mask for bits only in owner */
	uint32_t deny1; /* deny mask for bits not in owner */
	uint32_t deny2; /* deny mask for bits not in group */
	uint32_t owner; /* allow mask matching mode */
	uint32_t group; /* allow mask matching mode */
	uint32_t everyone; /* allow mask matching mode */
	} trivial_acl_t;

	static void
	acl_trivial_access_masks(mode_t mode, boolean_t isdir, trivial_acl_t *masks)
	{
	uint32_t read_mask = ACE_READ_DATA;
	uint32_t write_mask = ACE_WRITE_DATA\|ACE_APPEND_DATA;
	uint32_t execute_mask = ACE_EXECUTE;

	if (isdir)
	write_mask \|= ACE_DELETE_CHILD;

	masks->deny1 = 0;

	if (!(mode & S_IRUSR) && (mode & (S_IRGRP\|S_IROTH)))
	masks->deny1 \|= read_mask;
	if (!(mode & S_IWUSR) && (mode & (S_IWGRP\|S_IWOTH)))
	masks->deny1 \|= write_mask;
	if (!(mode & S_IXUSR) && (mode & (S_IXGRP\|S_IXOTH)))
	masks->deny1 \|= execute_mask;

	masks->deny2 = 0;
	if (!(mode & S_IRGRP) && (mode & S_IROTH))
	masks->deny2 \|= read_mask;
	if (!(mode & S_IWGRP) && (mode & S_IWOTH))
	masks->deny2 \|= write_mask;
	if (!(mode & S_IXGRP) && (mode & S_IXOTH))
	masks->deny2 \|= execute_mask;

	masks->allow0 = 0;
	if ((mode & S_IRUSR) && (!(mode & S_IRGRP) && (mode & S_IROTH)))
	masks->allow0 \|= read_mask;
	if ((mode & S_IWUSR) && (!(mode & S_IWGRP) && (mode & S_IWOTH)))
	masks->allow0 \|= write_mask;
	if ((mode & S_IXUSR) && (!(mode & S_IXGRP) && (mode & S_IXOTH)))
	masks->allow0 \|= execute_mask;

	masks->owner = ACE_WRITE_ATTRIBUTES\|ACE_WRITE_OWNER\|ACE_WRITE_ACL\|
	ACE_WRITE_NAMED_ATTRS\|ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|
	ACE_READ_NAMED_ATTRS\|ACE_SYNCHRONIZE;
	if (mode & S_IRUSR)
	masks->owner \|= read_mask;
	if (mode & S_IWUSR)
	masks->owner \|= write_mask;
	if (mode & S_IXUSR)
	masks->owner \|= execute_mask;

	masks->group = ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_READ_NAMED_ATTRS\|
	ACE_SYNCHRONIZE;
	if (mode & S_IRGRP)
	masks->group \|= read_mask;
	if (mode & S_IWGRP)
	masks->group \|= write_mask;
	if (mode & S_IXGRP)
	masks->group \|= execute_mask;

	masks->everyone = ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_READ_NAMED_ATTRS\|
	ACE_SYNCHRONIZE;
	if (mode & S_IROTH)
	masks->everyone \|= read_mask;
	if (mode & S_IWOTH)
	masks->everyone \|= write_mask;
	if (mode & S_IXOTH)
	masks->everyone \|= execute_mask;
	}

	/*
	* ace_trivial:
	* determine whether an ace_t acl is trivial
	*
	* Trivialness implies that the acl is composed of only
	* owner, group, everyone entries. ACL can't
	* have read_acl denied, and write_owner/write_acl/write_attributes
	* can only be owner@ entry.
	*/
	static int
	ace_trivial_common(void *acep, int aclcnt,
	uint64_t (walk)(void , uint64_t, int aclcnt,
	uint16_t , uint16_t , uint32_t *))
	{
	uint16_t flags;
	uint32_t mask;
	uint16_t type;
	uint64_t cookie = 0;

	while ((cookie = walk(acep, cookie, aclcnt, &flags, &type, &mask))) {
	switch (flags & ACE_TYPE_FLAGS) {
	case ACE_OWNER:
	case ACE_GROUP\|ACE_IDENTIFIER_GROUP:
	case ACE_EVERYONE:
	break;
	default:
	return (1);
	}

	if (flags & (ACE_FILE_INHERIT_ACE\|
	ACE_DIRECTORY_INHERIT_ACE\|ACE_NO_PROPAGATE_INHERIT_ACE\|
	ACE_INHERIT_ONLY_ACE))
	return (1);

	/*
	* Special check for some special bits
	*
	* Don't allow anybody to deny reading basic
	* attributes or a files ACL.
	*/
	if ((mask & (ACE_READ_ACL\|ACE_READ_ATTRIBUTES)) &&
	(type == ACE_ACCESS_DENIED_ACE_TYPE))
	return (1);

	/*
	* Delete permission is never set by default
	*/
	if (mask & ACE_DELETE)
	return (1);

	/*
	* Child delete permission should be accompanied by write
	*/
	if ((mask & ACE_DELETE_CHILD) && !(mask & ACE_WRITE_DATA))
	return (1);

	/*
	* only allow owner@ to have
	* write_acl/write_owner/write_attributes/write_xattr/
	*/
	if (type == ACE_ACCESS_ALLOWED_ACE_TYPE &&
	(!(flags & ACE_OWNER) && (mask &
	(ACE_WRITE_OWNER\|ACE_WRITE_ACL\| ACE_WRITE_ATTRIBUTES\|
	ACE_WRITE_NAMED_ATTRS))))
	return (1);

	}

	return (0);
	}

	/*
	* common code for setting ACLs.
	*
	* This function is called from zfs_mode_update, zfs_perm_init, and zfs_setacl.
	* zfs_setacl passes a non-NULL inherit pointer (ihp) to indicate that it's
	* already checked the acl and knows whether to inherit.
	*/
	int
	zfs_aclset_common(znode_t zp, zfs_acl_t aclp, cred_t cr, dmu_tx_t tx)
	{
	int error;
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	dmu_object_type_t otype;
	zfs_acl_locator_cb_t locate = { 0 };
	uint64_t mode;
	sa_bulk_attr_t bulk[5];
	uint64_t ctime[2];
	int count = 0;
	zfs_acl_phys_t acl_phys;

	mode = zp->z_mode;

	mode = zfs_mode_compute(mode, aclp, &zp->z_pflags,
	KUID_TO_SUID(ZTOI(zp)->i_uid), KGID_TO_SGID(ZTOI(zp)->i_gid));

	zp->z_mode = ZTOI(zp)->i_mode = mode;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&mode, sizeof (mode));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, sizeof (zp->z_pflags));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, sizeof (ctime));

	if (zp->z_acl_cached) {
	zfs_acl_free(zp->z_acl_cached);
	zp->z_acl_cached = NULL;
	}

	/*
	* Upgrade needed?
	*/
	if (!zfsvfs->z_use_fuids) {
	otype = DMU_OT_OLDACL;
	} else {
	if ((aclp->z_version == ZFS_ACL_VERSION_INITIAL) &&
	(zfsvfs->z_version >= ZPL_VERSION_FUID))
	zfs_acl_xform(zp, aclp, cr);
	ASSERT(aclp->z_version >= ZFS_ACL_VERSION_FUID);
	otype = DMU_OT_ACL;
	}

	/*
	* Arrgh, we have to handle old on disk format
	* as well as newer (preferred) SA format.
	*/

	if (zp->z_is_sa) { /* the easy case, just update the ACL attribute */
	locate.cb_aclp = aclp;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_ACES(zfsvfs),
	zfs_acl_data_locator, &locate, aclp->z_acl_bytes);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DACL_COUNT(zfsvfs),
	NULL, &aclp->z_acl_count, sizeof (uint64_t));
	} else { /* Painful legacy way */
	zfs_acl_node_t *aclnode;
	uint64_t off = 0;
	uint64_t aoid;

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_ZNODE_ACL(zfsvfs),
	&acl_phys, sizeof (acl_phys))) != 0)
	return (error);

	aoid = acl_phys.z_acl_extern_obj;

	if (aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	/*
	* If ACL was previously external and we are now
	* converting to new ACL format then release old
	* ACL object and create a new one.
	*/
	if (aoid &&
	aclp->z_version != acl_phys.z_acl_version) {
	error = dmu_object_free(zfsvfs->z_os, aoid, tx);
	if (error)
	return (error);
	aoid = 0;
	}
	if (aoid == 0) {
	aoid = dmu_object_alloc(zfsvfs->z_os,
	otype, aclp->z_acl_bytes,
	otype == DMU_OT_ACL ?
	DMU_OT_SYSACL : DMU_OT_NONE,
	otype == DMU_OT_ACL ?
	DN_OLD_MAX_BONUSLEN : 0, tx);
	} else {
	(void) dmu_object_set_blocksize(zfsvfs->z_os,
	aoid, aclp->z_acl_bytes, 0, tx);
	}
	acl_phys.z_acl_extern_obj = aoid;
	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	if (aclnode->z_ace_count == 0)
	continue;
	dmu_write(zfsvfs->z_os, aoid, off,
	aclnode->z_size, aclnode->z_acldata, tx);
	off += aclnode->z_size;
	}
	} else {
	void *start = acl_phys.z_ace_data;
	/*
	* Migrating back embedded?
	*/
	if (acl_phys.z_acl_extern_obj) {
	error = dmu_object_free(zfsvfs->z_os,
	acl_phys.z_acl_extern_obj, tx);
	if (error)
	return (error);
	acl_phys.z_acl_extern_obj = 0;
	}

	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	if (aclnode->z_ace_count == 0)
	continue;
	memcpy(start, aclnode->z_acldata,
	aclnode->z_size);
	start = (caddr_t)start + aclnode->z_size;
	}
	}
	/*
	* If Old version then swap count/bytes to match old
	* layout of znode_acl_phys_t.
	*/
	if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
	acl_phys.z_acl_size = aclp->z_acl_count;
	acl_phys.z_acl_count = aclp->z_acl_bytes;
	} else {
	acl_phys.z_acl_size = aclp->z_acl_bytes;
	acl_phys.z_acl_count = aclp->z_acl_count;
	}
	acl_phys.z_acl_version = aclp->z_version;

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
	&acl_phys, sizeof (acl_phys));
	}

	/*
	* Replace ACL wide bits, but first clear them.
	*/
	zp->z_pflags &= ~ZFS_ACL_WIDE_FLAGS;

	zp->z_pflags \|= aclp->z_hints;

	if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
	zp->z_pflags \|= ZFS_ACL_TRIVIAL;

	zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
	return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
	}

	static void
	zfs_acl_chmod(boolean_t isdir, uint64_t mode, boolean_t split, boolean_t trim,
	zfs_acl_t *aclp)
	{
	void *acep = NULL;
	uint64_t who;
	int new_count, new_bytes;
	int ace_size;
	int entry_type;
	uint16_t iflags, type;
	uint32_t access_mask;
	zfs_acl_node_t *newnode;
	size_t abstract_size = aclp->z_ops->ace_abstract_size();
	void *zacep;
	trivial_acl_t masks;

	new_count = new_bytes = 0;

	acl_trivial_access_masks((mode_t)mode, isdir, &masks);

	newnode = zfs_acl_node_alloc((abstract_size * 6) + aclp->z_acl_bytes);

	zacep = newnode->z_acldata;
	if (masks.allow0) {
	zfs_set_ace(aclp, zacep, masks.allow0, ALLOW, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}
	if (masks.deny1) {
	zfs_set_ace(aclp, zacep, masks.deny1, DENY, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}
	if (masks.deny2) {
	zfs_set_ace(aclp, zacep, masks.deny2, DENY, -1, OWNING_GROUP);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	new_count++;
	new_bytes += abstract_size;
	}

	while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
	&iflags, &type))) {
	entry_type = (iflags & ACE_TYPE_FLAGS);
	/*
	* ACEs used to represent the file mode may be divided
	* into an equivalent pair of inherit-only and regular
	* ACEs, if they are inheritable.
	* Skip regular ACEs, which are replaced by the new mode.
	*/
	if (split && (entry_type == ACE_OWNER \|\|
	entry_type == OWNING_GROUP \|\|
	entry_type == ACE_EVERYONE)) {
	if (!isdir \|\| !(iflags &
	(ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	continue;
	/*
	* We preserve owner@, group@, or @everyone
	* permissions, if they are inheritable, by
	* copying them to inherit_only ACEs. This
	* prevents inheritable permissions from being
	* altered along with the file mode.
	*/
	iflags \|= ACE_INHERIT_ONLY_ACE;
	}

	/*
	* If this ACL has any inheritable ACEs, mark that in
	* the hints (which are later masked into the pflags)
	* so create knows to do inheritance.
	*/
	if (isdir && (iflags &
	(ACE_FILE_INHERIT_ACE\|ACE_DIRECTORY_INHERIT_ACE)))
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	if ((type != ALLOW && type != DENY) \|\|
	(iflags & ACE_INHERIT_ONLY_ACE)) {
	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	aclp->z_hints \|= ZFS_ACL_OBJ_ACE;
	break;
	}
	} else {
	/*
	* Limit permissions to be no greater than
	* group permissions.
	* The "aclinherit" and "aclmode" properties
	* affect policy for create and chmod(2),
	* respectively.
	*/
	if ((type == ALLOW) && trim)
	access_mask &= masks.group;
	}
	zfs_set_ace(aclp, zacep, access_mask, type, who, iflags);
	ace_size = aclp->z_ops->ace_size(acep);
	zacep = (void *)((uintptr_t)zacep + ace_size);
	new_count++;
	new_bytes += ace_size;
	}
	zfs_set_ace(aclp, zacep, masks.owner, ALLOW, -1, ACE_OWNER);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	zfs_set_ace(aclp, zacep, masks.group, ALLOW, -1, OWNING_GROUP);
	zacep = (void *)((uintptr_t)zacep + abstract_size);
	zfs_set_ace(aclp, zacep, masks.everyone, ALLOW, -1, ACE_EVERYONE);

	new_count += 3;
	new_bytes += abstract_size * 3;
	zfs_acl_release_nodes(aclp);
	aclp->z_acl_count = new_count;
	aclp->z_acl_bytes = new_bytes;
	newnode->z_ace_count = new_count;
	newnode->z_size = new_bytes;
	list_insert_tail(&aclp->z_acl, newnode);
	}

	int
	zfs_acl_chmod_setattr(znode_t zp, zfs_acl_t *aclp, uint64_t mode)
	{
	int error = 0;

	mutex_enter(&zp->z_acl_lock);
	mutex_enter(&zp->z_lock);
	if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_DISCARD)
	*aclp = zfs_acl_alloc(zfs_acl_version_zp(zp));
	else
	error = zfs_acl_node_read(zp, B_TRUE, aclp, B_TRUE);

	if (error == 0) {
	(*aclp)->z_hints = zp->z_pflags & V4_ACL_WIDE_FLAGS;
	zfs_acl_chmod(S_ISDIR(ZTOI(zp)->i_mode), mode, B_TRUE,
	(ZTOZSB(zp)->z_acl_mode == ZFS_ACL_GROUPMASK), *aclp);
	}
	mutex_exit(&zp->z_lock);
	mutex_exit(&zp->z_acl_lock);

	return (error);
	}

	/*
	* Should ACE be inherited?
	*/
	static int
	zfs_ace_can_use(umode_t obj_mode, uint16_t acep_flags)
	{
	int iflags = (acep_flags & 0xf);

	if (S_ISDIR(obj_mode) && (iflags & ACE_DIRECTORY_INHERIT_ACE))
	return (1);
	else if (iflags & ACE_FILE_INHERIT_ACE)
	return (!(S_ISDIR(obj_mode) &&
	(iflags & ACE_NO_PROPAGATE_INHERIT_ACE)));
	return (0);
	}

	/*
	* inherit inheritable ACEs from parent
	*/
	static zfs_acl_t *
	zfs_acl_inherit(zfsvfs_t zfsvfs, umode_t va_mode, zfs_acl_t paclp,
	uint64_t mode, boolean_t *need_chmod)
	{
	void *pacep = NULL;
	void *acep;
	zfs_acl_node_t *aclnode;
	zfs_acl_t *aclp = NULL;
	uint64_t who;
	uint32_t access_mask;
	uint16_t iflags, newflags, type;
	size_t ace_size;
	void data1, data2;
	size_t data1sz, data2sz;
	uint_t aclinherit;
	boolean_t isdir = S_ISDIR(va_mode);
	boolean_t isreg = S_ISREG(va_mode);

	*need_chmod = B_TRUE;

	aclp = zfs_acl_alloc(paclp->z_version);
	aclinherit = zfsvfs->z_acl_inherit;
	if (aclinherit == ZFS_ACL_DISCARD \|\| S_ISLNK(va_mode))
	return (aclp);

	while ((pacep = zfs_acl_next_ace(paclp, pacep, &who,
	&access_mask, &iflags, &type))) {

	/*
	* don't inherit bogus ACEs
	*/
	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	/*
	* Check if ACE is inheritable by this vnode
	*/
	if ((aclinherit == ZFS_ACL_NOALLOW && type == ALLOW) \|\|
	!zfs_ace_can_use(va_mode, iflags))
	continue;

	/*
	* If owner@, group@, or everyone@ inheritable
	* then zfs_acl_chmod() isn't needed.
	*/
	if ((aclinherit == ZFS_ACL_PASSTHROUGH \|\|
	aclinherit == ZFS_ACL_PASSTHROUGH_X) &&
	((iflags & (ACE_OWNER\|ACE_EVERYONE)) \|\|
	((iflags & OWNING_GROUP) == OWNING_GROUP)) &&
	(isreg \|\| (isdir && (iflags & ACE_DIRECTORY_INHERIT_ACE))))
	*need_chmod = B_FALSE;

	/*
	* Strip inherited execute permission from file if
	* not in mode
	*/
	if (aclinherit == ZFS_ACL_PASSTHROUGH_X && type == ALLOW &&
	!isdir && ((mode & (S_IXUSR\|S_IXGRP\|S_IXOTH)) == 0)) {
	access_mask &= ~ACE_EXECUTE;
	}

	/*
	* Strip write_acl and write_owner from permissions
	* when inheriting an ACE
	*/
	if (aclinherit == ZFS_ACL_RESTRICTED && type == ALLOW) {
	access_mask &= ~RESTRICTED_CLEAR;
	}

	ace_size = aclp->z_ops->ace_size(pacep);
	aclnode = zfs_acl_node_alloc(ace_size);
	list_insert_tail(&aclp->z_acl, aclnode);
	acep = aclnode->z_acldata;

	zfs_set_ace(aclp, acep, access_mask, type,
	who, iflags\|ACE_INHERITED_ACE);

	/*
	* Copy special opaque data if any
	*/
	if ((data1sz = paclp->z_ops->ace_data(pacep, &data1)) != 0) {
	VERIFY((data2sz = aclp->z_ops->ace_data(acep,
	&data2)) == data1sz);
	memcpy(data2, data1, data2sz);
	}

	aclp->z_acl_count++;
	aclnode->z_ace_count++;
	aclp->z_acl_bytes += aclnode->z_size;
	newflags = aclp->z_ops->ace_flags_get(acep);

	/*
	* If ACE is not to be inherited further, or if the vnode is
	* not a directory, remove all inheritance flags
	*/
	if (!isdir \|\| (iflags & ACE_NO_PROPAGATE_INHERIT_ACE)) {
	newflags &= ~ALL_INHERIT;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	continue;
	}

	/*
	* This directory has an inheritable ACE
	*/
	aclp->z_hints \|= ZFS_INHERIT_ACE;

	/*
	* If only FILE_INHERIT is set then turn on
	* inherit_only
	*/
	if ((iflags & (ACE_FILE_INHERIT_ACE \|
	ACE_DIRECTORY_INHERIT_ACE)) == ACE_FILE_INHERIT_ACE) {
	newflags \|= ACE_INHERIT_ONLY_ACE;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	} else {
	newflags &= ~ACE_INHERIT_ONLY_ACE;
	aclp->z_ops->ace_flags_set(acep,
	newflags\|ACE_INHERITED_ACE);
	}
	}
	if (zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
	aclp->z_acl_count != 0) {
	*need_chmod = B_FALSE;
	}

	return (aclp);
	}

	/*
	* Create file system object initial permissions
	* including inheritable ACEs.
	* Also, create FUIDs for owner and group.
	*/
	int
	zfs_acl_ids_create(znode_t dzp, int flag, vattr_t vap, cred_t *cr,
	vsecattr_t vsecp, zfs_acl_ids_t acl_ids)
	{
	int error;
	zfsvfs_t *zfsvfs = ZTOZSB(dzp);
	zfs_acl_t *paclp;
	gid_t gid = vap->va_gid;
	boolean_t need_chmod = B_TRUE;
	boolean_t trim = B_FALSE;
	boolean_t inherited = B_FALSE;

	memset(acl_ids, 0, sizeof (zfs_acl_ids_t));
	acl_ids->z_mode = vap->va_mode;

	if (vsecp)
	if ((error = zfs_vsec_2_aclp(zfsvfs, vap->va_mode, vsecp,
	cr, &acl_ids->z_fuidp, &acl_ids->z_aclp)) != 0)
	return (error);

	acl_ids->z_fuid = vap->va_uid;
	acl_ids->z_fgid = vap->va_gid;
	#ifdef HAVE_KSID
	/*
	* Determine uid and gid.
	*/
	if ((flag & IS_ROOT_NODE) \|\| zfsvfs->z_replay \|\|
	((flag & IS_XATTR) && (S_ISDIR(vap->va_mode)))) {
	acl_ids->z_fuid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid,
	cr, ZFS_OWNER, &acl_ids->z_fuidp);
	acl_ids->z_fgid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
	cr, ZFS_GROUP, &acl_ids->z_fuidp);
	gid = vap->va_gid;
	} else {
	acl_ids->z_fuid = zfs_fuid_create_cred(zfsvfs, ZFS_OWNER,
	cr, &acl_ids->z_fuidp);
	acl_ids->z_fgid = 0;
	if (vap->va_mask & AT_GID) {
	acl_ids->z_fgid = zfs_fuid_create(zfsvfs,
	(uint64_t)vap->va_gid,
	cr, ZFS_GROUP, &acl_ids->z_fuidp);
	gid = vap->va_gid;
	if (acl_ids->z_fgid != KGID_TO_SGID(ZTOI(dzp)->i_gid) &&
	!groupmember(vap->va_gid, cr) &&
	secpolicy_vnode_create_gid(cr) != 0)
	acl_ids->z_fgid = 0;
	}
	if (acl_ids->z_fgid == 0) {
	if (dzp->z_mode & S_ISGID) {
	char *domain;
	uint32_t rid;

	acl_ids->z_fgid = KGID_TO_SGID(
	ZTOI(dzp)->i_gid);
	gid = zfs_fuid_map_id(zfsvfs, acl_ids->z_fgid,
	cr, ZFS_GROUP);

	if (zfsvfs->z_use_fuids &&
	IS_EPHEMERAL(acl_ids->z_fgid)) {
	domain = zfs_fuid_idx_domain(
	&zfsvfs->z_fuid_idx,
	FUID_INDEX(acl_ids->z_fgid));
	rid = FUID_RID(acl_ids->z_fgid);
	zfs_fuid_node_add(&acl_ids->z_fuidp,
	domain, rid,
	FUID_INDEX(acl_ids->z_fgid),
	acl_ids->z_fgid, ZFS_GROUP);
	}
	} else {
	acl_ids->z_fgid = zfs_fuid_create_cred(zfsvfs,
	ZFS_GROUP, cr, &acl_ids->z_fuidp);
	gid = crgetgid(cr);
	}
	}
	}
	#endif /* HAVE_KSID */

	/*
	* If we're creating a directory, and the parent directory has the
	* set-GID bit set, set in on the new directory.
	* Otherwise, if the user is neither privileged nor a member of the
	* file's new group, clear the file's set-GID bit.
	*/

	if (!(flag & IS_ROOT_NODE) && (dzp->z_mode & S_ISGID) &&
	(S_ISDIR(vap->va_mode))) {
	acl_ids->z_mode \|= S_ISGID;
	} else {
	if ((acl_ids->z_mode & S_ISGID) &&
	secpolicy_vnode_setids_setgids(cr, gid) != 0)
	acl_ids->z_mode &= ~S_ISGID;
	}

	if (acl_ids->z_aclp == NULL) {
	mutex_enter(&dzp->z_acl_lock);
	mutex_enter(&dzp->z_lock);
	if (!(flag & IS_ROOT_NODE) &&
	(dzp->z_pflags & ZFS_INHERIT_ACE) &&
	!(dzp->z_pflags & ZFS_XATTR)) {
	VERIFY(0 == zfs_acl_node_read(dzp, B_TRUE,
	&paclp, B_FALSE));
	acl_ids->z_aclp = zfs_acl_inherit(zfsvfs,
	vap->va_mode, paclp, acl_ids->z_mode, &need_chmod);
	inherited = B_TRUE;
	} else {
	acl_ids->z_aclp =
	zfs_acl_alloc(zfs_acl_version_zp(dzp));
	acl_ids->z_aclp->z_hints \|= ZFS_ACL_TRIVIAL;
	}
	mutex_exit(&dzp->z_lock);
	mutex_exit(&dzp->z_acl_lock);

	if (need_chmod) {
	if (S_ISDIR(vap->va_mode))
	acl_ids->z_aclp->z_hints \|=
	ZFS_ACL_AUTO_INHERIT;

	if (zfsvfs->z_acl_mode == ZFS_ACL_GROUPMASK &&
	zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH &&
	zfsvfs->z_acl_inherit != ZFS_ACL_PASSTHROUGH_X)
	trim = B_TRUE;
	zfs_acl_chmod(vap->va_mode, acl_ids->z_mode, B_FALSE,
	trim, acl_ids->z_aclp);
	}
	}

	if (inherited \|\| vsecp) {
	acl_ids->z_mode = zfs_mode_compute(acl_ids->z_mode,
	acl_ids->z_aclp, &acl_ids->z_aclp->z_hints,
	acl_ids->z_fuid, acl_ids->z_fgid);
	if (ace_trivial_common(acl_ids->z_aclp, 0, zfs_ace_walk) == 0)
	acl_ids->z_aclp->z_hints \|= ZFS_ACL_TRIVIAL;
	}

	return (0);
	}

	/*
	* Free ACL and fuid_infop, but not the acl_ids structure
	*/
	void
	zfs_acl_ids_free(zfs_acl_ids_t *acl_ids)
	{
	if (acl_ids->z_aclp)
	zfs_acl_free(acl_ids->z_aclp);
	if (acl_ids->z_fuidp)
	zfs_fuid_info_free(acl_ids->z_fuidp);
	acl_ids->z_aclp = NULL;
	acl_ids->z_fuidp = NULL;
	}

	boolean_t
	zfs_acl_ids_overquota(zfsvfs_t zv, zfs_acl_ids_t acl_ids, uint64_t projid)
	{
	return (zfs_id_overquota(zv, DMU_USERUSED_OBJECT, acl_ids->z_fuid) \|\|
	zfs_id_overquota(zv, DMU_GROUPUSED_OBJECT, acl_ids->z_fgid) \|\|
	(projid != ZFS_DEFAULT_PROJID && projid != ZFS_INVALID_PROJID &&
	zfs_id_overquota(zv, DMU_PROJECTUSED_OBJECT, projid)));
	}

	/*
	* Retrieve a file's ACL
	*/
	int
	zfs_getacl(znode_t zp, vsecattr_t vsecp, boolean_t skipaclchk, cred_t *cr)
	{
	zfs_acl_t *aclp;
	ulong_t mask;
	int error;
	int count = 0;
	int largeace = 0;

	mask = vsecp->vsa_mask & (VSA_ACE \| VSA_ACECNT \|
	VSA_ACE_ACLFLAGS \| VSA_ACE_ALLTYPES);

	if (mask == 0)
	return (SET_ERROR(ENOSYS));

	if ((error = zfs_zaccess(zp, ACE_READ_ACL, 0, skipaclchk, cr)))
	return (error);

	mutex_enter(&zp->z_acl_lock);

	error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
	if (error != 0) {
	mutex_exit(&zp->z_acl_lock);
	return (error);
	}

	/*
	* Scan ACL to determine number of ACEs
	*/
	if ((zp->z_pflags & ZFS_ACL_OBJ_ACE) && !(mask & VSA_ACE_ALLTYPES)) {
	void *zacep = NULL;
	uint64_t who;
	uint32_t access_mask;
	uint16_t type, iflags;

	while ((zacep = zfs_acl_next_ace(aclp, zacep,
	&who, &access_mask, &iflags, &type))) {
	switch (type) {
	case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE:
	case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE:
	case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE:
	largeace++;
	continue;
	default:
	count++;
	}
	}
	vsecp->vsa_aclcnt = count;
	} else
	count = (int)aclp->z_acl_count;

	if (mask & VSA_ACECNT) {
	vsecp->vsa_aclcnt = count;
	}

	if (mask & VSA_ACE) {
	size_t aclsz;

	aclsz = count * sizeof (ace_t) +
	sizeof (ace_object_t) * largeace;

	vsecp->vsa_aclentp = kmem_alloc(aclsz, KM_SLEEP);
	vsecp->vsa_aclentsz = aclsz;

	if (aclp->z_version == ZFS_ACL_VERSION_FUID)
	zfs_copy_fuid_2_ace(ZTOZSB(zp), aclp, cr,
	vsecp->vsa_aclentp, !(mask & VSA_ACE_ALLTYPES));
	else {
	zfs_acl_node_t *aclnode;
	void *start = vsecp->vsa_aclentp;

	for (aclnode = list_head(&aclp->z_acl); aclnode;
	aclnode = list_next(&aclp->z_acl, aclnode)) {
	memcpy(start, aclnode->z_acldata,
	aclnode->z_size);
	start = (caddr_t)start + aclnode->z_size;
	}
	ASSERT((caddr_t)start - (caddr_t)vsecp->vsa_aclentp ==
	aclp->z_acl_bytes);
	}
	}
	if (mask & VSA_ACE_ACLFLAGS) {
	vsecp->vsa_aclflags = 0;
	if (zp->z_pflags & ZFS_ACL_DEFAULTED)
	vsecp->vsa_aclflags \|= ACL_DEFAULTED;
	if (zp->z_pflags & ZFS_ACL_PROTECTED)
	vsecp->vsa_aclflags \|= ACL_PROTECTED;
	if (zp->z_pflags & ZFS_ACL_AUTO_INHERIT)
	vsecp->vsa_aclflags \|= ACL_AUTO_INHERIT;
	}

	mutex_exit(&zp->z_acl_lock);

	return (0);
	}

	int
	zfs_vsec_2_aclp(zfsvfs_t *zfsvfs, umode_t obj_mode,
	vsecattr_t vsecp, cred_t cr, zfs_fuid_info_t fuidp, zfs_acl_t zaclp)
	{
	zfs_acl_t *aclp;
	zfs_acl_node_t *aclnode;
	int aclcnt = vsecp->vsa_aclcnt;
	int error;

	if (vsecp->vsa_aclcnt > MAX_ACL_ENTRIES \|\| vsecp->vsa_aclcnt <= 0)
	return (SET_ERROR(EINVAL));

	aclp = zfs_acl_alloc(zfs_acl_version(zfsvfs->z_version));

	aclp->z_hints = 0;
	aclnode = zfs_acl_node_alloc(aclcnt * sizeof (zfs_object_ace_t));
	if (aclp->z_version == ZFS_ACL_VERSION_INITIAL) {
	if ((error = zfs_copy_ace_2_oldace(obj_mode, aclp,
	(ace_t *)vsecp->vsa_aclentp, aclnode->z_acldata,
	aclcnt, &aclnode->z_size)) != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	return (error);
	}
	} else {
	if ((error = zfs_copy_ace_2_fuid(zfsvfs, obj_mode, aclp,
	vsecp->vsa_aclentp, aclnode->z_acldata, aclcnt,
	&aclnode->z_size, fuidp, cr)) != 0) {
	zfs_acl_free(aclp);
	zfs_acl_node_free(aclnode);
	return (error);
	}
	}
	aclp->z_acl_bytes = aclnode->z_size;
	aclnode->z_ace_count = aclcnt;
	aclp->z_acl_count = aclcnt;
	list_insert_head(&aclp->z_acl, aclnode);

	/*
	* If flags are being set then add them to z_hints
	*/
	if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) {
	if (vsecp->vsa_aclflags & ACL_PROTECTED)
	aclp->z_hints \|= ZFS_ACL_PROTECTED;
	if (vsecp->vsa_aclflags & ACL_DEFAULTED)
	aclp->z_hints \|= ZFS_ACL_DEFAULTED;
	if (vsecp->vsa_aclflags & ACL_AUTO_INHERIT)
	aclp->z_hints \|= ZFS_ACL_AUTO_INHERIT;
	}

	*zaclp = aclp;

	return (0);
	}

	/*
	* Set a file's ACL
	*/
	int
	zfs_setacl(znode_t zp, vsecattr_t vsecp, boolean_t skipaclchk, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	zilog_t *zilog = zfsvfs->z_log;
	ulong_t mask = vsecp->vsa_mask & (VSA_ACE \| VSA_ACECNT);
	dmu_tx_t *tx;
	int error;
	zfs_acl_t *aclp;
	zfs_fuid_info_t *fuidp = NULL;
	boolean_t fuid_dirtied;
	uint64_t acl_obj;

	if (mask == 0)
	return (SET_ERROR(ENOSYS));

	if (zp->z_pflags & ZFS_IMMUTABLE)
	return (SET_ERROR(EPERM));

	if ((error = zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr)))
	return (error);

	error = zfs_vsec_2_aclp(zfsvfs, ZTOI(zp)->i_mode, vsecp, cr, &fuidp,
	&aclp);
	if (error)
	return (error);

	/*
	* If ACL wide flags aren't being set then preserve any
	* existing flags.
	*/
	if (!(vsecp->vsa_mask & VSA_ACE_ACLFLAGS)) {
	aclp->z_hints \|=
	(zp->z_pflags & V4_ACL_WIDE_FLAGS);
	}
	top:
	mutex_enter(&zp->z_acl_lock);
	mutex_enter(&zp->z_lock);

	tx = dmu_tx_create(zfsvfs->z_os);

	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);

	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);

	/*
	* If old version and ACL won't fit in bonus and we aren't
	* upgrading then take out necessary DMU holds
	*/

	if ((acl_obj = zfs_external_acl(zp)) != 0) {
	if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
	zfs_znode_acl_version(zp) <= ZFS_ACL_VERSION_INITIAL) {
	dmu_tx_hold_free(tx, acl_obj, 0,
	DMU_OBJECT_END);
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
	aclp->z_acl_bytes);
	} else {
	dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes);
	}
	} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes);
	}

	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_NOWAIT);
	if (error) {
	mutex_exit(&zp->z_acl_lock);
	mutex_exit(&zp->z_lock);

	if (error == ERESTART) {
	dmu_tx_wait(tx);
	dmu_tx_abort(tx);
	goto top;
	}
	dmu_tx_abort(tx);
	zfs_acl_free(aclp);
	return (error);
	}

	error = zfs_aclset_common(zp, aclp, cr, tx);
	ASSERT(error == 0);
	ASSERT(zp->z_acl_cached == NULL);
	zp->z_acl_cached = aclp;

	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	zfs_log_acl(zilog, tx, zp, vsecp, fuidp);

	if (fuidp)
	zfs_fuid_info_free(fuidp);
	dmu_tx_commit(tx);

	mutex_exit(&zp->z_lock);
	mutex_exit(&zp->z_acl_lock);

	return (error);
	}

	/*
	* Check accesses of interest (AoI) against attributes of the dataset
	* such as read-only. Returns zero if no AoI conflict with dataset
	* attributes, otherwise an appropriate errno is returned.
	*/
	static int
	zfs_zaccess_dataset_check(znode_t *zp, uint32_t v4_mode)
	{
	if ((v4_mode & WRITE_MASK) && (zfs_is_readonly(ZTOZSB(zp))) &&
	(!Z_ISDEV(ZTOI(zp)->i_mode) \|\|
	(Z_ISDEV(ZTOI(zp)->i_mode) && (v4_mode & WRITE_MASK_ATTRS)))) {
	return (SET_ERROR(EROFS));
	}

	/*
	* Intentionally allow ZFS_READONLY through here.
	* See zfs_zaccess_common().
	*/
	if ((v4_mode & WRITE_MASK_DATA) &&
	(zp->z_pflags & ZFS_IMMUTABLE)) {
	return (SET_ERROR(EPERM));
	}

	if ((v4_mode & (ACE_DELETE \| ACE_DELETE_CHILD)) &&
	(zp->z_pflags & ZFS_NOUNLINK)) {
	return (SET_ERROR(EPERM));
	}

	if (((v4_mode & (ACE_READ_DATA\|ACE_EXECUTE)) &&
	(zp->z_pflags & ZFS_AV_QUARANTINED))) {
	return (SET_ERROR(EACCES));
	}

	return (0);
	}

	/*
	* The primary usage of this function is to loop through all of the
	* ACEs in the znode, determining what accesses of interest (AoI) to
	* the caller are allowed or denied. The AoI are expressed as bits in
	* the working_mode parameter. As each ACE is processed, bits covered
	* by that ACE are removed from the working_mode. This removal
	* facilitates two things. The first is that when the working mode is
	* empty (= 0), we know we've looked at all the AoI. The second is
	* that the ACE interpretation rules don't allow a later ACE to undo
	* something granted or denied by an earlier ACE. Removing the
	* discovered access or denial enforces this rule. At the end of
	* processing the ACEs, all AoI that were found to be denied are
	* placed into the working_mode, giving the caller a mask of denied
	* accesses. Returns:
	* 0 if all AoI granted
	* EACCES if the denied mask is non-zero
	* other error if abnormal failure (e.g., IO error)
	*
	* A secondary usage of the function is to determine if any of the
	* AoI are granted. If an ACE grants any access in
	* the working_mode, we immediately short circuit out of the function.
	* This mode is chosen by setting anyaccess to B_TRUE. The
	* working_mode is not a denied access mask upon exit if the function
	* is used in this manner.
	*/
	static int
	zfs_zaccess_aces_check(znode_t zp, uint32_t working_mode,
	boolean_t anyaccess, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	zfs_acl_t *aclp;
	int error;
	uid_t uid = crgetuid(cr);
	uint64_t who;
	uint16_t type, iflags;
	uint16_t entry_type;
	uint32_t access_mask;
	uint32_t deny_mask = 0;
	zfs_ace_hdr_t *acep = NULL;
	boolean_t checkit;
	uid_t gowner;
	uid_t fowner;

	zfs_fuid_map_ids(zp, cr, &fowner, &gowner);

	mutex_enter(&zp->z_acl_lock);

	error = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
	if (error != 0) {
	mutex_exit(&zp->z_acl_lock);
	return (error);
	}

	ASSERT(zp->z_acl_cached);

	while ((acep = zfs_acl_next_ace(aclp, acep, &who, &access_mask,
	&iflags, &type))) {
	uint32_t mask_matched;

	if (!zfs_acl_valid_ace_type(type, iflags))
	continue;

	if (S_ISDIR(ZTOI(zp)->i_mode) &&
	(iflags & ACE_INHERIT_ONLY_ACE))
	continue;

	/* Skip ACE if it does not affect any AoI */
	mask_matched = (access_mask & *working_mode);
	if (!mask_matched)
	continue;

	entry_type = (iflags & ACE_TYPE_FLAGS);

	checkit = B_FALSE;

	switch (entry_type) {
	case ACE_OWNER:
	if (uid == fowner)
	checkit = B_TRUE;
	break;
	case OWNING_GROUP:
	who = gowner;
	zfs_fallthrough;
	case ACE_IDENTIFIER_GROUP:
	checkit = zfs_groupmember(zfsvfs, who, cr);
	break;
	case ACE_EVERYONE:
	checkit = B_TRUE;
	break;

	/* USER Entry */
	default:
	if (entry_type == 0) {
	uid_t newid;

	newid = zfs_fuid_map_id(zfsvfs, who, cr,
	ZFS_ACE_USER);
	if (newid != IDMAP_WK_CREATOR_OWNER_UID &&
	uid == newid)
	checkit = B_TRUE;
	break;
	} else {
	mutex_exit(&zp->z_acl_lock);
	return (SET_ERROR(EIO));
	}
	}

	if (checkit) {
	if (type == DENY) {
	DTRACE_PROBE3(zfs__ace__denies,
	znode_t *, zp,
	zfs_ace_hdr_t *, acep,
	uint32_t, mask_matched);
	deny_mask \|= mask_matched;
	} else {
	DTRACE_PROBE3(zfs__ace__allows,
	znode_t *, zp,
	zfs_ace_hdr_t *, acep,
	uint32_t, mask_matched);
	if (anyaccess) {
	mutex_exit(&zp->z_acl_lock);
	return (0);
	}
	}
	*working_mode &= ~mask_matched;
	}

	/* Are we done? */
	if (*working_mode == 0)
	break;
	}

	mutex_exit(&zp->z_acl_lock);

	/* Put the found 'denies' back on the working mode */
	if (deny_mask) {
	*working_mode \|= deny_mask;
	return (SET_ERROR(EACCES));
	} else if (*working_mode) {
	return (-1);
	}

	return (0);
	}

	/*
	* Return true if any access whatsoever granted, we don't actually
	* care what access is granted.
	*/
	boolean_t
	zfs_has_access(znode_t zp, cred_t cr)
	{
	uint32_t have = ACE_ALL_PERMS;

	if (zfs_zaccess_aces_check(zp, &have, B_TRUE, cr) != 0) {
	uid_t owner;

	owner = zfs_fuid_map_id(ZTOZSB(zp),
	KUID_TO_SUID(ZTOI(zp)->i_uid), cr, ZFS_OWNER);
	return (secpolicy_vnode_any_access(cr, ZTOI(zp), owner) == 0);
	}
	return (B_TRUE);
	}

	/*
	* Simplified access check for case where ACL is known to not contain
	* information beyond what is defined in the mode. In this case, we
	* can pass along to the kernel / vfs generic_permission() check, which
	* evaluates the mode and POSIX ACL.
	*
	* NFSv4 ACLs allow granting permissions that are usually relegated only
	* to the file owner or superuser. Examples are ACE_WRITE_OWNER (chown),
	* ACE_WRITE_ACL(chmod), and ACE_DELETE. ACE_DELETE requests must fail
	* because with conventional posix permissions, right to delete file
	* is determined by write bit on the parent dir.
	*
	* If unmappable perms are requested, then we must return EPERM
	* and include those bits in the working_mode so that the caller of
	* zfs_zaccess_common() can decide whether to perform additional
	* policy / capability checks. EACCES is used in zfs_zaccess_aces_check()
	* to indicate access check failed due to explicit DENY entry, and so
	* we want to avoid that here.
	*/
	static int
	zfs_zaccess_trivial(znode_t zp, uint32_t working_mode, cred_t *cr)
	{
	int err, mask;
	int unmapped = 0;

	ASSERT(zp->z_pflags & ZFS_ACL_TRIVIAL);

	mask = zfs_v4_to_unix(*working_mode, &unmapped);
	if (mask == 0 \|\| unmapped) {
	*working_mode = unmapped;
	return (unmapped ? SET_ERROR(EPERM) : 0);
	}

	#if defined(HAVE_IOPS_PERMISSION_USERNS)
	err = generic_permission(cr->user_ns, ZTOI(zp), mask);
	#else
	err = generic_permission(ZTOI(zp), mask);
	#endif
	if (err != 0) {
	return (SET_ERROR(EPERM));
	}

	*working_mode = unmapped;

	return (0);
	}

	static int
	zfs_zaccess_common(znode_t zp, uint32_t v4_mode, uint32_t working_mode,
	boolean_t check_privs, boolean_t skipaclchk, cred_t cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int err;

	*working_mode = v4_mode;
	*check_privs = B_TRUE;

	/*
	* Short circuit empty requests
	*/
	if (v4_mode == 0 \|\| zfsvfs->z_replay) {
	*working_mode = 0;
	return (0);
	}

	if ((err = zfs_zaccess_dataset_check(zp, v4_mode)) != 0) {
	*check_privs = B_FALSE;
	return (err);
	}

	/*
	* The caller requested that the ACL check be skipped. This
	* would only happen if the caller checked VOP_ACCESS() with a
	* 32 bit ACE mask and already had the appropriate permissions.
	*/
	if (skipaclchk) {
	*working_mode = 0;
	return (0);
	}

	/*
	* Note: ZFS_READONLY represents the "DOS R/O" attribute.
	* When that flag is set, we should behave as if write access
	* were not granted by anything in the ACL. In particular:
	* We _must_ allow writes after opening the file r/w, then
	* setting the DOS R/O attribute, and writing some more.
	* (Similar to how you can write after fchmod(fd, 0444).)
	*
	* Therefore ZFS_READONLY is ignored in the dataset check
	* above, and checked here as if part of the ACL check.
	* Also note: DOS R/O is ignored for directories.
	*/
	if ((v4_mode & WRITE_MASK_DATA) &&
	S_ISDIR(ZTOI(zp)->i_mode) &&
	(zp->z_pflags & ZFS_READONLY)) {
	return (SET_ERROR(EPERM));
	}

	if (zp->z_pflags & ZFS_ACL_TRIVIAL)
	return (zfs_zaccess_trivial(zp, working_mode, cr));

	return (zfs_zaccess_aces_check(zp, working_mode, B_FALSE, cr));
	}

	static int
	zfs_zaccess_append(znode_t zp, uint32_t working_mode, boolean_t *check_privs,
	cred_t *cr)
	{
	if (*working_mode != ACE_WRITE_DATA)
	return (SET_ERROR(EACCES));

	return (zfs_zaccess_common(zp, ACE_APPEND_DATA, working_mode,
	check_privs, B_FALSE, cr));
	}

	int
	zfs_fastaccesschk_execute(znode_t zdp, cred_t cr)
	{
	boolean_t owner = B_FALSE;
	boolean_t groupmbr = B_FALSE;
	boolean_t is_attr;
	uid_t uid = crgetuid(cr);
	int error;

	if (zdp->z_pflags & ZFS_AV_QUARANTINED)
	return (SET_ERROR(EACCES));

	is_attr = ((zdp->z_pflags & ZFS_XATTR) &&
	(S_ISDIR(ZTOI(zdp)->i_mode)));
	if (is_attr)
	goto slow;


	mutex_enter(&zdp->z_acl_lock);

	if (zdp->z_pflags & ZFS_NO_EXECS_DENIED) {
	mutex_exit(&zdp->z_acl_lock);
	return (0);
	}

	if (KUID_TO_SUID(ZTOI(zdp)->i_uid) != 0 \|\|
	KGID_TO_SGID(ZTOI(zdp)->i_gid) != 0) {
	mutex_exit(&zdp->z_acl_lock);
	goto slow;
	}

	if (uid == KUID_TO_SUID(ZTOI(zdp)->i_uid)) {
	owner = B_TRUE;
	if (zdp->z_mode & S_IXUSR) {
	mutex_exit(&zdp->z_acl_lock);
	return (0);
	} else {
	mutex_exit(&zdp->z_acl_lock);
	goto slow;
	}
	}
	if (groupmember(KGID_TO_SGID(ZTOI(zdp)->i_gid), cr)) {
	groupmbr = B_TRUE;
	if (zdp->z_mode & S_IXGRP) {
	mutex_exit(&zdp->z_acl_lock);
	return (0);
	} else {
	mutex_exit(&zdp->z_acl_lock);
	goto slow;
	}
	}
	if (!owner && !groupmbr) {
	if (zdp->z_mode & S_IXOTH) {
	mutex_exit(&zdp->z_acl_lock);
	return (0);
	}
	}

	mutex_exit(&zdp->z_acl_lock);

	slow:
	DTRACE_PROBE(zfs__fastpath__execute__access__miss);
	if ((error = zfs_enter(ZTOZSB(zdp), FTAG)) != 0)
	return (error);
	error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr);
	zfs_exit(ZTOZSB(zdp), FTAG);
	return (error);
	}

	/*
	* Determine whether Access should be granted/denied.
	*
	* The least priv subsystem is always consulted as a basic privilege
	* can define any form of access.
	*/
	int
	zfs_zaccess(znode_t zp, int mode, int flags, boolean_t skipaclchk, cred_t cr)
	{
	uint32_t working_mode;
	int error;
	int is_attr;
	boolean_t check_privs;
	znode_t *xzp;
	znode_t *check_zp = zp;
	mode_t needed_bits;
	uid_t owner;

	is_attr = ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode));

	/*
	* If attribute then validate against base file
	*/
	if (is_attr) {
	if ((error = zfs_zget(ZTOZSB(zp),
	zp->z_xattr_parent, &xzp)) != 0) {
	return (error);
	}

	check_zp = xzp;

	/*
	* fixup mode to map to xattr perms
	*/

	if (mode & (ACE_WRITE_DATA\|ACE_APPEND_DATA)) {
	mode &= ~(ACE_WRITE_DATA\|ACE_APPEND_DATA);
	mode \|= ACE_WRITE_NAMED_ATTRS;
	}

	if (mode & (ACE_READ_DATA\|ACE_EXECUTE)) {
	mode &= ~(ACE_READ_DATA\|ACE_EXECUTE);
	mode \|= ACE_READ_NAMED_ATTRS;
	}
	}

	owner = zfs_fuid_map_id(ZTOZSB(zp), KUID_TO_SUID(ZTOI(zp)->i_uid),
	cr, ZFS_OWNER);
	/*
	* Map the bits required to the standard inode flags
	* S_IRUSR\|S_IWUSR\|S_IXUSR in the needed_bits. Map the bits
	* mapped by working_mode (currently missing) in missing_bits.
	* Call secpolicy_vnode_access2() with (needed_bits & ~checkmode),
	* needed_bits.
	*/
	needed_bits = 0;

	working_mode = mode;
	if ((working_mode & (ACE_READ_ACL\|ACE_READ_ATTRIBUTES)) &&
	owner == crgetuid(cr))
	working_mode &= ~(ACE_READ_ACL\|ACE_READ_ATTRIBUTES);

	if (working_mode & (ACE_READ_DATA\|ACE_READ_NAMED_ATTRS\|
	ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_SYNCHRONIZE))
	needed_bits \|= S_IRUSR;
	if (working_mode & (ACE_WRITE_DATA\|ACE_WRITE_NAMED_ATTRS\|
	ACE_APPEND_DATA\|ACE_WRITE_ATTRIBUTES\|ACE_SYNCHRONIZE))
	needed_bits \|= S_IWUSR;
	if (working_mode & ACE_EXECUTE)
	needed_bits \|= S_IXUSR;

	if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
	&check_privs, skipaclchk, cr)) == 0) {
	if (is_attr)
	zrele(xzp);
	return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
	needed_bits, needed_bits));
	}

	if (error && !check_privs) {
	if (is_attr)
	zrele(xzp);
	return (error);
	}

	if (error && (flags & V_APPEND)) {
	error = zfs_zaccess_append(zp, &working_mode, &check_privs, cr);
	}

	if (error && check_privs) {
	mode_t checkmode = 0;

	/*
	* First check for implicit owner permission on
	* read_acl/read_attributes
	*/

	error = 0;
	ASSERT(working_mode != 0);

	if ((working_mode & (ACE_READ_ACL\|ACE_READ_ATTRIBUTES) &&
	owner == crgetuid(cr)))
	working_mode &= ~(ACE_READ_ACL\|ACE_READ_ATTRIBUTES);

	if (working_mode & (ACE_READ_DATA\|ACE_READ_NAMED_ATTRS\|
	ACE_READ_ACL\|ACE_READ_ATTRIBUTES\|ACE_SYNCHRONIZE))
	checkmode \|= S_IRUSR;
	if (working_mode & (ACE_WRITE_DATA\|ACE_WRITE_NAMED_ATTRS\|
	ACE_APPEND_DATA\|ACE_WRITE_ATTRIBUTES\|ACE_SYNCHRONIZE))
	checkmode \|= S_IWUSR;
	if (working_mode & ACE_EXECUTE)
	checkmode \|= S_IXUSR;

	error = secpolicy_vnode_access2(cr, ZTOI(check_zp), owner,
	needed_bits & ~checkmode, needed_bits);

	if (error == 0 && (working_mode & ACE_WRITE_OWNER))
	error = secpolicy_vnode_chown(cr, owner);
	if (error == 0 && (working_mode & ACE_WRITE_ACL))
	error = secpolicy_vnode_setdac(cr, owner);

	if (error == 0 && (working_mode &
	(ACE_DELETE\|ACE_DELETE_CHILD)))
	error = secpolicy_vnode_remove(cr);

	if (error == 0 && (working_mode & ACE_SYNCHRONIZE)) {
	error = secpolicy_vnode_chown(cr, owner);
	}
	if (error == 0) {
	/*
	* See if any bits other than those already checked
	* for are still present. If so then return EACCES
	*/
	if (working_mode & ~(ZFS_CHECKED_MASKS)) {
	error = SET_ERROR(EACCES);
	}
	}
	} else if (error == 0) {
	error = secpolicy_vnode_access2(cr, ZTOI(zp), owner,
	needed_bits, needed_bits);
	}

	if (is_attr)
	zrele(xzp);

	return (error);
	}

	/*
	* Translate traditional unix S_IRUSR/S_IWUSR/S_IXUSR mode into
	* NFSv4-style ZFS ACL format and call zfs_zaccess()
	*/
	int
	zfs_zaccess_rwx(znode_t zp, mode_t mode, int flags, cred_t cr)
	{
	return (zfs_zaccess(zp, zfs_unix_to_v4(mode >> 6), flags, B_FALSE, cr));
	}

	/*
	* Access function for secpolicy_vnode_setattr
	*/
	int
	zfs_zaccess_unix(znode_t zp, mode_t mode, cred_t cr)
	{
	int v4_mode = zfs_unix_to_v4(mode >> 6);

	return (zfs_zaccess(zp, v4_mode, 0, B_FALSE, cr));
	}

	/* See zfs_zaccess_delete() */
	static const boolean_t zfs_write_implies_delete_child = B_TRUE;

	/*
	* Determine whether delete access should be granted.
	*
	* The following chart outlines how we handle delete permissions which is
	* how recent versions of windows (Windows 2008) handles it. The efficiency
	* comes from not having to check the parent ACL where the object itself grants
	* delete:
	*
	* -------------------------------------------------------
	* \| Parent Dir \| Target Object Permissions \|
	* \| permissions \| \|
	* -------------------------------------------------------
	* \| \| ACL Allows \| ACL Denies\| Delete \|
	* \| \| Delete \| Delete \| unspecified\|
	* -------------------------------------------------------
	* \| ACL Allows \| Permit \| Deny * \| Permit \|
	* \| DELETE_CHILD \| \| \| \|
	* -------------------------------------------------------
	* \| ACL Denies \| Permit \| Deny \| Deny \|
	* \| DELETE_CHILD \| \| \| \|
	* -------------------------------------------------------
	* \| ACL specifies \| \| \| \|
	* \| only allow \| Permit \| Deny * \| Permit \|
	* \| write and \| \| \| \|
	* \| execute \| \| \| \|
	* -------------------------------------------------------
	* \| ACL denies \| \| \| \|
	* \| write and \| Permit \| Deny \| Deny \|
	* \| execute \| \| \| \|
	* -------------------------------------------------------
	* ^
	* \|
	* Re. execute permission on the directory: if that's missing,
	* the vnode lookup of the target will fail before we get here.
	*
	* Re [*] in the table above: NFSv4 would normally Permit delete for
	* these two cells of the matrix.
	* See acl.h for notes on which ACE_... flags should be checked for which
	* operations. Specifically, the NFSv4 committee recommendation is in
	* conflict with the Windows interpretation of DENY ACEs, where DENY ACEs
	* should take precedence ahead of ALLOW ACEs.
	*
	* This implementation always consults the target object's ACL first.
	* If a DENY ACE is present on the target object that specifies ACE_DELETE,
	* delete access is denied. If an ALLOW ACE with ACE_DELETE is present on
	* the target object, access is allowed. If and only if no entries with
	* ACE_DELETE are present in the object's ACL, check the container's ACL
	* for entries with ACE_DELETE_CHILD.
	*
	* A summary of the logic implemented from the table above is as follows:
	*
	* First check for DENY ACEs that apply.
	* If either target or container has a deny, EACCES.
	*
	* Delete access can then be summarized as follows:
	* 1: The object to be deleted grants ACE_DELETE, or
	* 2: The containing directory grants ACE_DELETE_CHILD.
	* In a Windows system, that would be the end of the story.
	* In this system, (2) has some complications...
	* 2a: "sticky" bit on a directory adds restrictions, and
	* 2b: existing ACEs from previous versions of ZFS may
	* not carry ACE_DELETE_CHILD where they should, so we
	* also allow delete when ACE_WRITE_DATA is granted.
	*
	* Note: 2b is technically a work-around for a prior bug,
	* which hopefully can go away some day. For those who
	* no longer need the work around, and for testing, this
	* work-around is made conditional via the tunable:
	* zfs_write_implies_delete_child
	*/
	int
	zfs_zaccess_delete(znode_t dzp, znode_t zp, cred_t *cr)
	{
	uint32_t wanted_dirperms;
	uint32_t dzp_working_mode = 0;
	uint32_t zp_working_mode = 0;
	int dzp_error, zp_error;
	boolean_t dzpcheck_privs;
	boolean_t zpcheck_privs;

	if (zp->z_pflags & (ZFS_IMMUTABLE \| ZFS_NOUNLINK))
	return (SET_ERROR(EPERM));

	/*
	* Case 1:
	* If target object grants ACE_DELETE then we are done. This is
	* indicated by a return value of 0. For this case we don't worry
	* about the sticky bit because sticky only applies to the parent
	* directory and this is the child access result.
	*
	* If we encounter a DENY ACE here, we're also done (EACCES).
	* Note that if we hit a DENY ACE here (on the target) it should
	* take precedence over a DENY ACE on the container, so that when
	* we have more complete auditing support we will be able to
	* report an access failure against the specific target.
	* (This is part of why we're checking the target first.)
	*/
	zp_error = zfs_zaccess_common(zp, ACE_DELETE, &zp_working_mode,
	&zpcheck_privs, B_FALSE, cr);
	if (zp_error == EACCES) {
	/* We hit a DENY ACE. */
	if (!zpcheck_privs)
	return (SET_ERROR(zp_error));
	return (secpolicy_vnode_remove(cr));

	}
	if (zp_error == 0)
	return (0);

	/*
	* Case 2:
	* If the containing directory grants ACE_DELETE_CHILD,
	* or we're in backward compatibility mode and the
	* containing directory has ACE_WRITE_DATA, allow.
	* Case 2b is handled with wanted_dirperms.
	*/
	wanted_dirperms = ACE_DELETE_CHILD;
	if (zfs_write_implies_delete_child)
	wanted_dirperms \|= ACE_WRITE_DATA;
	dzp_error = zfs_zaccess_common(dzp, wanted_dirperms,
	&dzp_working_mode, &dzpcheck_privs, B_FALSE, cr);
	if (dzp_error == EACCES) {
	/* We hit a DENY ACE. */
	if (!dzpcheck_privs)
	return (SET_ERROR(dzp_error));
	return (secpolicy_vnode_remove(cr));
	}

	/*
	* Cases 2a, 2b (continued)
	*
	* Note: dzp_working_mode now contains any permissions
	* that were NOT granted. Therefore, if any of the
	* wanted_dirperms WERE granted, we will have:
	* dzp_working_mode != wanted_dirperms
	* We're really asking if ANY of those permissions
	* were granted, and if so, grant delete access.
	*/
	if (dzp_working_mode != wanted_dirperms)
	dzp_error = 0;

	/*
	* dzp_error is 0 if the container granted us permissions to "modify".
	* If we do not have permission via one or more ACEs, our current
	* privileges may still permit us to modify the container.
	*
	* dzpcheck_privs is false when i.e. the FS is read-only.
	* Otherwise, do privilege checks for the container.
	*/
	if (dzp_error != 0 && dzpcheck_privs) {
	uid_t owner;

	/*
	* The secpolicy call needs the requested access and
	* the current access mode of the container, but it
	* only knows about Unix-style modes (VEXEC, VWRITE),
	* so this must condense the fine-grained ACE bits into
	* Unix modes.
	*
	* The VEXEC flag is easy, because we know that has
	* always been checked before we get here (during the
	* lookup of the target vnode). The container has not
	* granted us permissions to "modify", so we do not set
	* the VWRITE flag in the current access mode.
	*/
	owner = zfs_fuid_map_id(ZTOZSB(dzp),
	KUID_TO_SUID(ZTOI(dzp)->i_uid), cr, ZFS_OWNER);
	dzp_error = secpolicy_vnode_access2(cr, ZTOI(dzp),
	owner, S_IXUSR, S_IWUSR\|S_IXUSR);
	}
	if (dzp_error != 0) {
	/*
	* Note: We may have dzp_error = -1 here (from
	* zfs_zacess_common). Don't return that.
	*/
	return (SET_ERROR(EACCES));
	}


	/*
	* At this point, we know that the directory permissions allow
	* us to modify, but we still need to check for the additional
	* restrictions that apply when the "sticky bit" is set.
	*
	* Yes, zfs_sticky_remove_access() also checks this bit, but
	* checking it here and skipping the call below is nice when
	* you're watching all of this with dtrace.
	*/
	if ((dzp->z_mode & S_ISVTX) == 0)
	return (0);

	/*
	* zfs_sticky_remove_access will succeed if:
	* 1. The sticky bit is absent.
	* 2. We pass the sticky bit restrictions.
	* 3. We have privileges that always allow file removal.
	*/
	return (zfs_sticky_remove_access(dzp, zp, cr));
	}

	int
	zfs_zaccess_rename(znode_t sdzp, znode_t szp, znode_t *tdzp,
	znode_t tzp, cred_t cr)
	{
	int add_perm;
	int error;

	if (szp->z_pflags & ZFS_AV_QUARANTINED)
	return (SET_ERROR(EACCES));

	add_perm = S_ISDIR(ZTOI(szp)->i_mode) ?
	ACE_ADD_SUBDIRECTORY : ACE_ADD_FILE;

	/*
	* Rename permissions are combination of delete permission +
	* add file/subdir permission.
	*/

	/*
	* first make sure we do the delete portion.
	*
	* If that succeeds then check for add_file/add_subdir permissions
	*/

	if ((error = zfs_zaccess_delete(sdzp, szp, cr)))
	return (error);

	/*
	* If we have a tzp, see if we can delete it?
	*/
	if (tzp) {
	if ((error = zfs_zaccess_delete(tdzp, tzp, cr)))
	return (error);
	}

	/*
	* Now check for add permissions
	*/
	error = zfs_zaccess(tdzp, add_perm, 0, B_FALSE, cr);

	return (error);
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	index 2a4e3f378948..819416b68d5f 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	@@ -1,257 +1,259 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/trace_zfs.h>

	typedef struct zfs_dbgmsg {
	procfs_list_node_t zdm_node;
	uint64_t zdm_timestamp;
	- int zdm_size;
	+ uint_t zdm_size;
	char zdm_msg[1]; /* variable length allocation */
	} zfs_dbgmsg_t;

	static procfs_list_t zfs_dbgmsgs;
	-static int zfs_dbgmsg_size = 0;
	-int zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
	+static uint_t zfs_dbgmsg_size = 0;
	+uint_t zfs_dbgmsg_maxsize = 4<<20; /* 4MB */

	/*
	* Internal ZFS debug messages are enabled by default.
	*
	* # Print debug messages
	* cat /proc/spl/kstat/zfs/dbgmsg
	*
	* # Disable the kernel debug message log.
	* echo 0 > /sys/module/zfs/parameters/zfs_dbgmsg_enable
	*
	* # Clear the kernel debug message log.
	* echo 0 >/proc/spl/kstat/zfs/dbgmsg
	*/
	int zfs_dbgmsg_enable = B_TRUE;

	static int
	zfs_dbgmsg_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-12s %-8s\n", "timestamp", "message");
	return (0);
	}

	static int
	zfs_dbgmsg_show(struct seq_file f, void p)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )p;
	seq_printf(f, "%-12llu %-s\n",
	(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);
	return (0);
	}

	static void
	-zfs_dbgmsg_purge(int max_size)
	+zfs_dbgmsg_purge(uint_t max_size)
	{
	while (zfs_dbgmsg_size > max_size) {
	zfs_dbgmsg_t *zdm = list_remove_head(&zfs_dbgmsgs.pl_list);
	if (zdm == NULL)
	return;

	- int size = zdm->zdm_size;
	+ uint_t size = zdm->zdm_size;
	kmem_free(zdm, size);
	zfs_dbgmsg_size -= size;
	}
	}

	static int
	zfs_dbgmsg_clear(procfs_list_t *procfs_list)
	{
	(void) procfs_list;
	mutex_enter(&zfs_dbgmsgs.pl_lock);
	zfs_dbgmsg_purge(0);
	mutex_exit(&zfs_dbgmsgs.pl_lock);
	return (0);
	}

	void
	zfs_dbgmsg_init(void)
	{
	procfs_list_install("zfs",
	NULL,
	"dbgmsg",
	0600,
	&zfs_dbgmsgs,
	zfs_dbgmsg_show,
	zfs_dbgmsg_show_header,
	zfs_dbgmsg_clear,
	offsetof(zfs_dbgmsg_t, zdm_node));
	}

	void
	zfs_dbgmsg_fini(void)
	{
	procfs_list_uninstall(&zfs_dbgmsgs);
	zfs_dbgmsg_purge(0);

	/*
	* TODO - decide how to make this permanent
	*/
	#ifdef _KERNEL
	procfs_list_destroy(&zfs_dbgmsgs);
	#endif
	}

	void
	__set_error(const char file, const char func, int line, int err)
	{
	/*
	* To enable this:
	*
	* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
	*/
	if (zfs_flags & ZFS_DEBUG_SET_ERROR)
	__dprintf(B_FALSE, file, func, line, "error %lu",
	(ulong_t)err);
	}

	void
	__zfs_dbgmsg(char *buf)
	{
	- int size = sizeof (zfs_dbgmsg_t) + strlen(buf);
	+ uint_t size = sizeof (zfs_dbgmsg_t) + strlen(buf);
	zfs_dbgmsg_t *zdm = kmem_zalloc(size, KM_SLEEP);
	zdm->zdm_size = size;
	zdm->zdm_timestamp = gethrestime_sec();
	strcpy(zdm->zdm_msg, buf);

	mutex_enter(&zfs_dbgmsgs.pl_lock);
	procfs_list_add(&zfs_dbgmsgs, zdm);
	zfs_dbgmsg_size += size;
	- zfs_dbgmsg_purge(MAX(zfs_dbgmsg_maxsize, 0));
	+ zfs_dbgmsg_purge(zfs_dbgmsg_maxsize);
	mutex_exit(&zfs_dbgmsgs.pl_lock);
	}

	#ifdef _KERNEL

	void
	__dprintf(boolean_t dprint, const char file, const char func,
	int line, const char *fmt, ...)
	{
	const char *newfile;
	va_list adx;
	size_t size;
	char *buf;
	char *nl;
	int i;
	char *prefix = (dprint) ? "dprintf: " : "";

	size = 1024;
	buf = kmem_alloc(size, KM_SLEEP);

	/*
	* Get rid of annoying prefix to filename.
	*/
	newfile = strrchr(file, '/');
	if (newfile != NULL) {
	newfile = newfile + 1; /* Get rid of leading / */
	} else {
	newfile = file;
	}

	i = snprintf(buf, size, "%s%s:%d:%s(): ", prefix, newfile, line, func);

	if (i < size) {
	va_start(adx, fmt);
	(void) vsnprintf(buf + i, size - i, fmt, adx);
	va_end(adx);
	}

	/*
	* Get rid of trailing newline for dprintf logs.
	*/
	if (dprint && buf[0] != '\0') {
	nl = &buf[strlen(buf) - 1];
	if (*nl == '\n')
	*nl = '\0';
	}

	/*
	* To get this data enable the zfs__dprintf trace point as shown:
	*
	* # Enable zfs__dprintf tracepoint, clear the tracepoint ring buffer
	* $ echo 1 > /sys/kernel/debug/tracing/events/zfs/enable
	* $ echo 0 > /sys/kernel/debug/tracing/trace
	*
	* # Dump the ring buffer.
	* $ cat /sys/kernel/debug/tracing/trace
	*/
	DTRACE_PROBE1(zfs__dprintf, char *, buf);

	/*
	* To get this data:
	*
	* $ cat /proc/spl/kstat/zfs/dbgmsg
	*
	* To clear the buffer:
	* $ echo 0 > /proc/spl/kstat/zfs/dbgmsg
	*/
	__zfs_dbgmsg(buf);

	kmem_free(buf, size);
	}

	#else

	void
	zfs_dbgmsg_print(const char *tag)
	{
	ssize_t ret __attribute__((unused));

	/*
	* We use write() in this function instead of printf()
	* so it is safe to call from a signal handler.
	*/
	ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
	ret = write(STDOUT_FILENO, tag, strlen(tag));
	ret = write(STDOUT_FILENO, ") START:\n", 9);

	mutex_enter(&zfs_dbgmsgs.pl_lock);
	for (zfs_dbgmsg_t *zdm = list_head(&zfs_dbgmsgs.pl_list); zdm != NULL;
	zdm = list_next(&zfs_dbgmsgs.pl_list, zdm)) {
	ret = write(STDOUT_FILENO, zdm->zdm_msg,
	strlen(zdm->zdm_msg));
	ret = write(STDOUT_FILENO, "\n", 1);
	}

	ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
	ret = write(STDOUT_FILENO, tag, strlen(tag));
	ret = write(STDOUT_FILENO, ") END\n", 6);

	mutex_exit(&zfs_dbgmsgs.pl_lock);
	}
	#endif /* _KERNEL */

	#ifdef _KERNEL
	module_param(zfs_dbgmsg_enable, int, 0644);
	MODULE_PARM_DESC(zfs_dbgmsg_enable, "Enable ZFS debug message log");

	-module_param(zfs_dbgmsg_maxsize, int, 0644);
	+/* BEGIN CSTYLED */
	+module_param(zfs_dbgmsg_maxsize, uint, 0644);
	+/* END CSTYLED */
	MODULE_PARM_DESC(zfs_dbgmsg_maxsize, "Maximum ZFS debug log size");
	#endif
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zio_crypt.c b/sys/contrib/openzfs/module/os/linux/zfs/zio_crypt.c
	index dcab02b07894..671300932384 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zio_crypt.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zio_crypt.c
	@@ -1,2048 +1,2048 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2017, Datto, Inc. All rights reserved.
	*/

	#include <sys/zio_crypt.h>
	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dnode.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio.h>
	#include <sys/zil.h>
	#include <sys/sha2.h>
	#include <sys/hkdf.h>
	#include <sys/qat.h>

	/*
	* This file is responsible for handling all of the details of generating
	* encryption parameters and performing encryption and authentication.
	*
	* BLOCK ENCRYPTION PARAMETERS:
	* Encryption /Authentication Algorithm Suite (crypt):
	* The encryption algorithm, mode, and key length we are going to use. We
	* currently support AES in either GCM or CCM modes with 128, 192, and 256 bit
	* keys. All authentication is currently done with SHA512-HMAC.
	*
	* Plaintext:
	* The unencrypted data that we want to encrypt.
	*
	* Initialization Vector (IV):
	* An initialization vector for the encryption algorithms. This is used to
	* "tweak" the encryption algorithms so that two blocks of the same data are
	* encrypted into different ciphertext outputs, thus obfuscating block patterns.
	* The supported encryption modes (AES-GCM and AES-CCM) require that an IV is
	* never reused with the same encryption key. This value is stored unencrypted
	* and must simply be provided to the decryption function. We use a 96 bit IV
	* (as recommended by NIST) for all block encryption. For non-dedup blocks we
	* derive the IV randomly. The first 64 bits of the IV are stored in the second
	* word of DVA[2] and the remaining 32 bits are stored in the upper 32 bits of
	* blk_fill. This is safe because encrypted blocks can't use the upper 32 bits
	* of blk_fill. We only encrypt level 0 blocks, which normally have a fill count
	* of 1. The only exception is for DMU_OT_DNODE objects, where the fill count of
	* level 0 blocks is the number of allocated dnodes in that block. The on-disk
	* format supports at most 2^15 slots per L0 dnode block, because the maximum
	* block size is 16MB (2^24). In either case, for level 0 blocks this number
	* will still be smaller than UINT32_MAX so it is safe to store the IV in the
	* top 32 bits of blk_fill, while leaving the bottom 32 bits of the fill count
	* for the dnode code.
	*
	* Master key:
	* This is the most important secret data of an encrypted dataset. It is used
	* along with the salt to generate that actual encryption keys via HKDF. We
	* do not use the master key to directly encrypt any data because there are
	* theoretical limits on how much data can actually be safely encrypted with
	* any encryption mode. The master key is stored encrypted on disk with the
	* user's wrapping key. Its length is determined by the encryption algorithm.
	* For details on how this is stored see the block comment in dsl_crypt.c
	*
	* Salt:
	* Used as an input to the HKDF function, along with the master key. We use a
	* 64 bit salt, stored unencrypted in the first word of DVA[2]. Any given salt
	* can be used for encrypting many blocks, so we cache the current salt and the
	* associated derived key in zio_crypt_t so we do not need to derive it again
	* needlessly.
	*
	* Encryption Key:
	* A secret binary key, generated from an HKDF function used to encrypt and
	* decrypt data.
	*
	* Message Authentication Code (MAC)
	* The MAC is an output of authenticated encryption modes such as AES-GCM and
	* AES-CCM. Its purpose is to ensure that an attacker cannot modify encrypted
	* data on disk and return garbage to the application. Effectively, it is a
	* checksum that can not be reproduced by an attacker. We store the MAC in the
	* second 128 bits of blk_cksum, leaving the first 128 bits for a truncated
	* regular checksum of the ciphertext which can be used for scrubbing.
	*
	* OBJECT AUTHENTICATION:
	* Some object types, such as DMU_OT_MASTER_NODE cannot be encrypted because
	* they contain some info that always needs to be readable. To prevent this
	* data from being altered, we authenticate this data using SHA512-HMAC. This
	* will produce a MAC (similar to the one produced via encryption) which can
	* be used to verify the object was not modified. HMACs do not require key
	* rotation or IVs, so we can keep up to the full 3 copies of authenticated
	* data.
	*
	* ZIL ENCRYPTION:
	* ZIL blocks have their bp written to disk ahead of the associated data, so we
	* cannot store the MAC there as we normally do. For these blocks the MAC is
	* stored in the embedded checksum within the zil_chain_t header. The salt and
	* IV are generated for the block on bp allocation instead of at encryption
	* time. In addition, ZIL blocks have some pieces that must be left in plaintext
	* for claiming even though all of the sensitive user data still needs to be
	* encrypted. The function zio_crypt_init_uios_zil() handles parsing which
	* pieces of the block need to be encrypted. All data that is not encrypted is
	* authenticated using the AAD mechanisms that the supported encryption modes
	* provide for. In order to preserve the semantics of the ZIL for encrypted
	* datasets, the ZIL is not protected at the objset level as described below.
	*
	* DNODE ENCRYPTION:
	* Similarly to ZIL blocks, the core part of each dnode_phys_t needs to be left
	* in plaintext for scrubbing and claiming, but the bonus buffers might contain
	* sensitive user data. The function zio_crypt_init_uios_dnode() handles parsing
	* which pieces of the block need to be encrypted. For more details about
	* dnode authentication and encryption, see zio_crypt_init_uios_dnode().
	*
	* OBJECT SET AUTHENTICATION:
	* Up to this point, everything we have encrypted and authenticated has been
	* at level 0 (or -2 for the ZIL). If we did not do any further work the
	* on-disk format would be susceptible to attacks that deleted or rearranged
	* the order of level 0 blocks. Ideally, the cleanest solution would be to
	* maintain a tree of authentication MACs going up the bp tree. However, this
	* presents a problem for raw sends. Send files do not send information about
	* indirect blocks so there would be no convenient way to transfer the MACs and
	* they cannot be recalculated on the receive side without the master key which
	* would defeat one of the purposes of raw sends in the first place. Instead,
	* for the indirect levels of the bp tree, we use a regular SHA512 of the MACs
	* from the level below. We also include some portable fields from blk_prop such
	* as the lsize and compression algorithm to prevent the data from being
	* misinterpreted.
	*
	* At the objset level, we maintain 2 separate 256 bit MACs in the
	* objset_phys_t. The first one is "portable" and is the logical root of the
	* MAC tree maintained in the metadnode's bps. The second, is "local" and is
	* used as the root MAC for the user accounting objects, which are also not
	* transferred via "zfs send". The portable MAC is sent in the DRR_BEGIN payload
	* of the send file. The useraccounting code ensures that the useraccounting
	* info is not present upon a receive, so the local MAC can simply be cleared
	* out at that time. For more info about objset_phys_t authentication, see
	* zio_crypt_do_objset_hmacs().
	*
	* CONSIDERATIONS FOR DEDUP:
	* In order for dedup to work, blocks that we want to dedup with one another
	* need to use the same IV and encryption key, so that they will have the same
	* ciphertext. Normally, one should never reuse an IV with the same encryption
	* key or else AES-GCM and AES-CCM can both actually leak the plaintext of both
	* blocks. In this case, however, since we are using the same plaintext as
	* well all that we end up with is a duplicate of the original ciphertext we
	* already had. As a result, an attacker with read access to the raw disk will
	* be able to tell which blocks are the same but this information is given away
	* by dedup anyway. In order to get the same IVs and encryption keys for
	* equivalent blocks of data we use an HMAC of the plaintext. We use an HMAC
	* here so that a reproducible checksum of the plaintext is never available to
	* the attacker. The HMAC key is kept alongside the master key, encrypted on
	* disk. The first 64 bits of the HMAC are used in place of the random salt, and
	* the next 96 bits are used as the IV. As a result of this mechanism, dedup
	* will only work within a clone family since encrypted dedup requires use of
	* the same master and HMAC keys.
	*/

	/*
	* After encrypting many blocks with the same key we may start to run up
	* against the theoretical limits of how much data can securely be encrypted
	* with a single key using the supported encryption modes. The most obvious
	* limitation is that our risk of generating 2 equivalent 96 bit IVs increases
	* the more IVs we generate (which both GCM and CCM modes strictly forbid).
	* This risk actually grows surprisingly quickly over time according to the
	* Birthday Problem. With a total IV space of 2^(96 bits), and assuming we have
	* generated n IVs with a cryptographically secure RNG, the approximate
	* probability p(n) of a collision is given as:
	*
	* p(n) ~= e^(-n(n-1)/(2(2^96)))
	*
	* [http://www.math.cornell.edu/~mec/2008-2009/TianyiZheng/Birthday.html]
	*
	* Assuming that we want to ensure that p(n) never goes over 1 / 1 trillion
	* we must not write more than 398,065,730 blocks with the same encryption key.
	* Therefore, we rotate our keys after 400,000,000 blocks have been written by
	* generating a new random 64 bit salt for our HKDF encryption key generation
	* function.
	*/
	#define ZFS_KEY_MAX_SALT_USES_DEFAULT 400000000
	#define ZFS_CURRENT_MAX_SALT_USES \
	(MIN(zfs_key_max_salt_uses, ZFS_KEY_MAX_SALT_USES_DEFAULT))
	static unsigned long zfs_key_max_salt_uses = ZFS_KEY_MAX_SALT_USES_DEFAULT;

	typedef struct blkptr_auth_buf {
	uint64_t bab_prop; /* blk_prop - portable mask */
	uint8_t bab_mac[ZIO_DATA_MAC_LEN]; /* MAC from blk_cksum */
	uint64_t bab_pad; /* reserved for future use */
	} blkptr_auth_buf_t;

	const zio_crypt_info_t zio_crypt_table[ZIO_CRYPT_FUNCTIONS] = {
	{"", ZC_TYPE_NONE, 0, "inherit"},
	{"", ZC_TYPE_NONE, 0, "on"},
	{"", ZC_TYPE_NONE, 0, "off"},
	{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 16, "aes-128-ccm"},
	{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 24, "aes-192-ccm"},
	{SUN_CKM_AES_CCM, ZC_TYPE_CCM, 32, "aes-256-ccm"},
	{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 16, "aes-128-gcm"},
	{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 24, "aes-192-gcm"},
	{SUN_CKM_AES_GCM, ZC_TYPE_GCM, 32, "aes-256-gcm"}
	};

	void
	zio_crypt_key_destroy(zio_crypt_key_t *key)
	{
	rw_destroy(&key->zk_salt_lock);

	/* free crypto templates */
	crypto_destroy_ctx_template(key->zk_current_tmpl);
	crypto_destroy_ctx_template(key->zk_hmac_tmpl);

	/* zero out sensitive data */
	memset(key, 0, sizeof (zio_crypt_key_t));
	}

	int
	zio_crypt_key_init(uint64_t crypt, zio_crypt_key_t *key)
	{
	int ret;
	crypto_mechanism_t mech;
	uint_t keydata_len;

	ASSERT(key != NULL);
	ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);

	keydata_len = zio_crypt_table[crypt].ci_keylen;
	memset(key, 0, sizeof (zio_crypt_key_t));

	/* fill keydata buffers and salt with random data */
	ret = random_get_bytes((uint8_t *)&key->zk_guid, sizeof (uint64_t));
	if (ret != 0)
	goto error;

	ret = random_get_bytes(key->zk_master_keydata, keydata_len);
	if (ret != 0)
	goto error;

	ret = random_get_bytes(key->zk_hmac_keydata, SHA512_HMAC_KEYLEN);
	if (ret != 0)
	goto error;

	ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
	if (ret != 0)
	goto error;

	/* derive the current key from the master key */
	ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
	key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
	keydata_len);
	if (ret != 0)
	goto error;

	/* initialize keys for the ICP */
	key->zk_current_key.ck_data = key->zk_current_keydata;
	key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);

	key->zk_hmac_key.ck_data = &key->zk_hmac_key;
	key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);

	/*
	* Initialize the crypto templates. It's ok if this fails because
	* this is just an optimization.
	*/
	mech.cm_type = crypto_mech2id(zio_crypt_table[crypt].ci_mechname);
	ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
	&key->zk_current_tmpl);
	if (ret != CRYPTO_SUCCESS)
	key->zk_current_tmpl = NULL;

	mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
	ret = crypto_create_ctx_template(&mech, &key->zk_hmac_key,
	&key->zk_hmac_tmpl);
	if (ret != CRYPTO_SUCCESS)
	key->zk_hmac_tmpl = NULL;

	key->zk_crypt = crypt;
	key->zk_version = ZIO_CRYPT_KEY_CURRENT_VERSION;
	key->zk_salt_count = 0;
	rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);

	return (0);

	error:
	zio_crypt_key_destroy(key);
	return (ret);
	}

	static int
	zio_crypt_key_change_salt(zio_crypt_key_t *key)
	{
	int ret = 0;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	crypto_mechanism_t mech;
	uint_t keydata_len = zio_crypt_table[key->zk_crypt].ci_keylen;

	/* generate a new salt */
	ret = random_get_bytes(salt, ZIO_DATA_SALT_LEN);
	if (ret != 0)
	goto error;

	rw_enter(&key->zk_salt_lock, RW_WRITER);

	/* someone beat us to the salt rotation, just unlock and return */
	if (key->zk_salt_count < ZFS_CURRENT_MAX_SALT_USES)
	goto out_unlock;

	/* derive the current key from the master key and the new salt */
	ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
	salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata, keydata_len);
	if (ret != 0)
	goto out_unlock;

	/* assign the salt and reset the usage count */
	memcpy(key->zk_salt, salt, ZIO_DATA_SALT_LEN);
	key->zk_salt_count = 0;

	/* destroy the old context template and create the new one */
	crypto_destroy_ctx_template(key->zk_current_tmpl);
	ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
	&key->zk_current_tmpl);
	if (ret != CRYPTO_SUCCESS)
	key->zk_current_tmpl = NULL;

	rw_exit(&key->zk_salt_lock);

	return (0);

	out_unlock:
	rw_exit(&key->zk_salt_lock);
	error:
	return (ret);
	}

	/* See comment above zfs_key_max_salt_uses definition for details */
	int
	zio_crypt_key_get_salt(zio_crypt_key_t key, uint8_t salt)
	{
	int ret;
	boolean_t salt_change;

	rw_enter(&key->zk_salt_lock, RW_READER);

	memcpy(salt, key->zk_salt, ZIO_DATA_SALT_LEN);
	salt_change = (atomic_inc_64_nv(&key->zk_salt_count) >=
	ZFS_CURRENT_MAX_SALT_USES);

	rw_exit(&key->zk_salt_lock);

	if (salt_change) {
	ret = zio_crypt_key_change_salt(key);
	if (ret != 0)
	goto error;
	}

	return (0);

	error:
	return (ret);
	}

	/*
	* This function handles all encryption and decryption in zfs. When
	* encrypting it expects puio to reference the plaintext and cuio to
	* reference the ciphertext. cuio must have enough space for the
	* ciphertext + room for a MAC. datalen should be the length of the
	* plaintext / ciphertext alone.
	*/
	static int
	zio_do_crypt_uio(boolean_t encrypt, uint64_t crypt, crypto_key_t *key,
	crypto_ctx_template_t tmpl, uint8_t *ivbuf, uint_t datalen,
	zfs_uio_t puio, zfs_uio_t cuio, uint8_t *authbuf, uint_t auth_len)
	{
	int ret;
	crypto_data_t plaindata, cipherdata;
	CK_AES_CCM_PARAMS ccmp;
	CK_AES_GCM_PARAMS gcmp;
	crypto_mechanism_t mech;
	zio_crypt_info_t crypt_info;
	uint_t plain_full_len, maclen;

	ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);

	/* lookup the encryption info */
	crypt_info = zio_crypt_table[crypt];

	/* the mac will always be the last iovec_t in the cipher uio */
	maclen = cuio->uio_iov[cuio->uio_iovcnt - 1].iov_len;

	ASSERT(maclen <= ZIO_DATA_MAC_LEN);

	/* setup encryption mechanism (same as crypt) */
	mech.cm_type = crypto_mech2id(crypt_info.ci_mechname);

	/*
	* Strangely, the ICP requires that plain_full_len must include
	* the MAC length when decrypting, even though the UIO does not
	* need to have the extra space allocated.
	*/
	if (encrypt) {
	plain_full_len = datalen;
	} else {
	plain_full_len = datalen + maclen;
	}

	/*
	* setup encryption params (currently only AES CCM and AES GCM
	* are supported)
	*/
	if (crypt_info.ci_crypt_type == ZC_TYPE_CCM) {
	ccmp.ulNonceSize = ZIO_DATA_IV_LEN;
	ccmp.ulAuthDataSize = auth_len;
	ccmp.authData = authbuf;
	ccmp.ulMACSize = maclen;
	ccmp.nonce = ivbuf;
	ccmp.ulDataSize = plain_full_len;

	mech.cm_param = (char *)(&ccmp);
	mech.cm_param_len = sizeof (CK_AES_CCM_PARAMS);
	} else {
	gcmp.ulIvLen = ZIO_DATA_IV_LEN;
	gcmp.ulIvBits = CRYPTO_BYTES2BITS(ZIO_DATA_IV_LEN);
	gcmp.ulAADLen = auth_len;
	gcmp.pAAD = authbuf;
	gcmp.ulTagBits = CRYPTO_BYTES2BITS(maclen);
	gcmp.pIv = ivbuf;

	mech.cm_param = (char *)(&gcmp);
	mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
	}

	/* populate the cipher and plain data structs. */
	plaindata.cd_format = CRYPTO_DATA_UIO;
	plaindata.cd_offset = 0;
	plaindata.cd_uio = puio;
	plaindata.cd_length = plain_full_len;

	cipherdata.cd_format = CRYPTO_DATA_UIO;
	cipherdata.cd_offset = 0;
	cipherdata.cd_uio = cuio;
	cipherdata.cd_length = datalen + maclen;

	/* perform the actual encryption */
	if (encrypt) {
	ret = crypto_encrypt(&mech, &plaindata, key, tmpl, &cipherdata);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}
	} else {
	ret = crypto_decrypt(&mech, &cipherdata, key, tmpl, &plaindata);
	if (ret != CRYPTO_SUCCESS) {
	ASSERT3U(ret, ==, CRYPTO_INVALID_MAC);
	ret = SET_ERROR(ECKSUM);
	goto error;
	}
	}

	return (0);

	error:
	return (ret);
	}

	int
	zio_crypt_key_wrap(crypto_key_t cwkey, zio_crypt_key_t key, uint8_t *iv,
	uint8_t mac, uint8_t keydata_out, uint8_t *hmac_keydata_out)
	{
	int ret;
	zfs_uio_t puio, cuio;
	uint64_t aad[3];
	iovec_t plain_iovecs[2], cipher_iovecs[3];
	uint64_t crypt = key->zk_crypt;
	uint_t enc_len, keydata_len, aad_len;

	ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);

	keydata_len = zio_crypt_table[crypt].ci_keylen;

	/* generate iv for wrapping the master and hmac key */
	ret = random_get_pseudo_bytes(iv, WRAPPING_IV_LEN);
	if (ret != 0)
	goto error;

	/* initialize zfs_uio_ts */
	plain_iovecs[0].iov_base = key->zk_master_keydata;
	plain_iovecs[0].iov_len = keydata_len;
	plain_iovecs[1].iov_base = key->zk_hmac_keydata;
	plain_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;

	cipher_iovecs[0].iov_base = keydata_out;
	cipher_iovecs[0].iov_len = keydata_len;
	cipher_iovecs[1].iov_base = hmac_keydata_out;
	cipher_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
	cipher_iovecs[2].iov_base = mac;
	cipher_iovecs[2].iov_len = WRAPPING_MAC_LEN;

	/*
	* Although we don't support writing to the old format, we do
	* support rewrapping the key so that the user can move and
	* quarantine datasets on the old format.
	*/
	if (key->zk_version == 0) {
	aad_len = sizeof (uint64_t);
	aad[0] = LE_64(key->zk_guid);
	} else {
	ASSERT3U(key->zk_version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
	aad_len = sizeof (uint64_t) * 3;
	aad[0] = LE_64(key->zk_guid);
	aad[1] = LE_64(crypt);
	aad[2] = LE_64(key->zk_version);
	}

	enc_len = zio_crypt_table[crypt].ci_keylen + SHA512_HMAC_KEYLEN;
	puio.uio_iov = plain_iovecs;
	puio.uio_iovcnt = 2;
	puio.uio_segflg = UIO_SYSSPACE;
	cuio.uio_iov = cipher_iovecs;
	cuio.uio_iovcnt = 3;
	cuio.uio_segflg = UIO_SYSSPACE;

	/* encrypt the keys and store the resulting ciphertext and mac */
	ret = zio_do_crypt_uio(B_TRUE, crypt, cwkey, NULL, iv, enc_len,
	&puio, &cuio, (uint8_t *)aad, aad_len);
	if (ret != 0)
	goto error;

	return (0);

	error:
	return (ret);
	}

	int
	zio_crypt_key_unwrap(crypto_key_t *cwkey, uint64_t crypt, uint64_t version,
	uint64_t guid, uint8_t keydata, uint8_t hmac_keydata, uint8_t *iv,
	uint8_t mac, zio_crypt_key_t key)
	{
	crypto_mechanism_t mech;
	zfs_uio_t puio, cuio;
	uint64_t aad[3];
	iovec_t plain_iovecs[2], cipher_iovecs[3];
	uint_t enc_len, keydata_len, aad_len;
	int ret;

	ASSERT3U(crypt, <, ZIO_CRYPT_FUNCTIONS);

	rw_init(&key->zk_salt_lock, NULL, RW_DEFAULT, NULL);

	keydata_len = zio_crypt_table[crypt].ci_keylen;

	/* initialize zfs_uio_ts */
	plain_iovecs[0].iov_base = key->zk_master_keydata;
	plain_iovecs[0].iov_len = keydata_len;
	plain_iovecs[1].iov_base = key->zk_hmac_keydata;
	plain_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;

	cipher_iovecs[0].iov_base = keydata;
	cipher_iovecs[0].iov_len = keydata_len;
	cipher_iovecs[1].iov_base = hmac_keydata;
	cipher_iovecs[1].iov_len = SHA512_HMAC_KEYLEN;
	cipher_iovecs[2].iov_base = mac;
	cipher_iovecs[2].iov_len = WRAPPING_MAC_LEN;

	if (version == 0) {
	aad_len = sizeof (uint64_t);
	aad[0] = LE_64(guid);
	} else {
	ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);
	aad_len = sizeof (uint64_t) * 3;
	aad[0] = LE_64(guid);
	aad[1] = LE_64(crypt);
	aad[2] = LE_64(version);
	}

	enc_len = keydata_len + SHA512_HMAC_KEYLEN;
	puio.uio_iov = plain_iovecs;
	puio.uio_segflg = UIO_SYSSPACE;
	puio.uio_iovcnt = 2;
	cuio.uio_iov = cipher_iovecs;
	cuio.uio_iovcnt = 3;
	cuio.uio_segflg = UIO_SYSSPACE;

	/* decrypt the keys and store the result in the output buffers */
	ret = zio_do_crypt_uio(B_FALSE, crypt, cwkey, NULL, iv, enc_len,
	&puio, &cuio, (uint8_t *)aad, aad_len);
	if (ret != 0)
	goto error;

	/* generate a fresh salt */
	ret = random_get_bytes(key->zk_salt, ZIO_DATA_SALT_LEN);
	if (ret != 0)
	goto error;

	/* derive the current key from the master key */
	ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
	key->zk_salt, ZIO_DATA_SALT_LEN, key->zk_current_keydata,
	keydata_len);
	if (ret != 0)
	goto error;

	/* initialize keys for ICP */
	key->zk_current_key.ck_data = key->zk_current_keydata;
	key->zk_current_key.ck_length = CRYPTO_BYTES2BITS(keydata_len);

	key->zk_hmac_key.ck_data = key->zk_hmac_keydata;
	key->zk_hmac_key.ck_length = CRYPTO_BYTES2BITS(SHA512_HMAC_KEYLEN);

	/*
	* Initialize the crypto templates. It's ok if this fails because
	* this is just an optimization.
	*/
	mech.cm_type = crypto_mech2id(zio_crypt_table[crypt].ci_mechname);
	ret = crypto_create_ctx_template(&mech, &key->zk_current_key,
	&key->zk_current_tmpl);
	if (ret != CRYPTO_SUCCESS)
	key->zk_current_tmpl = NULL;

	mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
	ret = crypto_create_ctx_template(&mech, &key->zk_hmac_key,
	&key->zk_hmac_tmpl);
	if (ret != CRYPTO_SUCCESS)
	key->zk_hmac_tmpl = NULL;

	key->zk_crypt = crypt;
	key->zk_version = version;
	key->zk_guid = guid;
	key->zk_salt_count = 0;

	return (0);

	error:
	zio_crypt_key_destroy(key);
	return (ret);
	}

	int
	zio_crypt_generate_iv(uint8_t *ivbuf)
	{
	int ret;

	/* randomly generate the IV */
	ret = random_get_pseudo_bytes(ivbuf, ZIO_DATA_IV_LEN);
	if (ret != 0)
	goto error;

	return (0);

	error:
	memset(ivbuf, 0, ZIO_DATA_IV_LEN);
	return (ret);
	}

	int
	zio_crypt_do_hmac(zio_crypt_key_t key, uint8_t data, uint_t datalen,
	uint8_t *digestbuf, uint_t digestlen)
	{
	int ret;
	crypto_mechanism_t mech;
	crypto_data_t in_data, digest_data;
	uint8_t raw_digestbuf[SHA512_DIGEST_LENGTH];

	ASSERT3U(digestlen, <=, SHA512_DIGEST_LENGTH);

	/* initialize sha512-hmac mechanism and crypto data */
	mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
	mech.cm_param = NULL;
	mech.cm_param_len = 0;

	/* initialize the crypto data */
	in_data.cd_format = CRYPTO_DATA_RAW;
	in_data.cd_offset = 0;
	in_data.cd_length = datalen;
	in_data.cd_raw.iov_base = (char *)data;
	in_data.cd_raw.iov_len = in_data.cd_length;

	digest_data.cd_format = CRYPTO_DATA_RAW;
	digest_data.cd_offset = 0;
	digest_data.cd_length = SHA512_DIGEST_LENGTH;
	digest_data.cd_raw.iov_base = (char *)raw_digestbuf;
	digest_data.cd_raw.iov_len = digest_data.cd_length;

	/* generate the hmac */
	ret = crypto_mac(&mech, &in_data, &key->zk_hmac_key, key->zk_hmac_tmpl,
	&digest_data);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	memcpy(digestbuf, raw_digestbuf, digestlen);

	return (0);

	error:
	memset(digestbuf, 0, digestlen);
	return (ret);
	}

	int
	zio_crypt_generate_iv_salt_dedup(zio_crypt_key_t key, uint8_t data,
	uint_t datalen, uint8_t ivbuf, uint8_t salt)
	{
	int ret;
	uint8_t digestbuf[SHA512_DIGEST_LENGTH];

	ret = zio_crypt_do_hmac(key, data, datalen,
	digestbuf, SHA512_DIGEST_LENGTH);
	if (ret != 0)
	return (ret);

	memcpy(salt, digestbuf, ZIO_DATA_SALT_LEN);
	memcpy(ivbuf, digestbuf + ZIO_DATA_SALT_LEN, ZIO_DATA_IV_LEN);

	return (0);
	}

	/*
	* The following functions are used to encode and decode encryption parameters
	* into blkptr_t and zil_header_t. The ICP wants to use these parameters as
	* byte strings, which normally means that these strings would not need to deal
	* with byteswapping at all. However, both blkptr_t and zil_header_t may be
	* byteswapped by lower layers and so we must "undo" that byteswap here upon
	* decoding and encoding in a non-native byteorder. These functions require
	* that the byteorder bit is correct before being called.
	*/
	void
	zio_crypt_encode_params_bp(blkptr_t bp, uint8_t salt, uint8_t *iv)
	{
	uint64_t val64;
	uint32_t val32;

	ASSERT(BP_IS_ENCRYPTED(bp));

	if (!BP_SHOULD_BYTESWAP(bp)) {
	memcpy(&bp->blk_dva[2].dva_word[0], salt, sizeof (uint64_t));
	memcpy(&bp->blk_dva[2].dva_word[1], iv, sizeof (uint64_t));
	memcpy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
	BP_SET_IV2(bp, val32);
	} else {
	memcpy(&val64, salt, sizeof (uint64_t));
	bp->blk_dva[2].dva_word[0] = BSWAP_64(val64);

	memcpy(&val64, iv, sizeof (uint64_t));
	bp->blk_dva[2].dva_word[1] = BSWAP_64(val64);

	memcpy(&val32, iv + sizeof (uint64_t), sizeof (uint32_t));
	BP_SET_IV2(bp, BSWAP_32(val32));
	}
	}

	void
	zio_crypt_decode_params_bp(const blkptr_t bp, uint8_t salt, uint8_t *iv)
	{
	uint64_t val64;
	uint32_t val32;

	ASSERT(BP_IS_PROTECTED(bp));

	/* for convenience, so callers don't need to check */
	if (BP_IS_AUTHENTICATED(bp)) {
	memset(salt, 0, ZIO_DATA_SALT_LEN);
	memset(iv, 0, ZIO_DATA_IV_LEN);
	return;
	}

	if (!BP_SHOULD_BYTESWAP(bp)) {
	memcpy(salt, &bp->blk_dva[2].dva_word[0], sizeof (uint64_t));
	memcpy(iv, &bp->blk_dva[2].dva_word[1], sizeof (uint64_t));

	val32 = (uint32_t)BP_GET_IV2(bp);
	memcpy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
	} else {
	val64 = BSWAP_64(bp->blk_dva[2].dva_word[0]);
	memcpy(salt, &val64, sizeof (uint64_t));

	val64 = BSWAP_64(bp->blk_dva[2].dva_word[1]);
	memcpy(iv, &val64, sizeof (uint64_t));

	val32 = BSWAP_32((uint32_t)BP_GET_IV2(bp));
	memcpy(iv + sizeof (uint64_t), &val32, sizeof (uint32_t));
	}
	}

	void
	zio_crypt_encode_mac_bp(blkptr_t bp, uint8_t mac)
	{
	uint64_t val64;

	ASSERT(BP_USES_CRYPT(bp));
	ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_OBJSET);

	if (!BP_SHOULD_BYTESWAP(bp)) {
	memcpy(&bp->blk_cksum.zc_word[2], mac, sizeof (uint64_t));
	memcpy(&bp->blk_cksum.zc_word[3], mac + sizeof (uint64_t),
	sizeof (uint64_t));
	} else {
	memcpy(&val64, mac, sizeof (uint64_t));
	bp->blk_cksum.zc_word[2] = BSWAP_64(val64);

	memcpy(&val64, mac + sizeof (uint64_t), sizeof (uint64_t));
	bp->blk_cksum.zc_word[3] = BSWAP_64(val64);
	}
	}

	void
	zio_crypt_decode_mac_bp(const blkptr_t bp, uint8_t mac)
	{
	uint64_t val64;

	ASSERT(BP_USES_CRYPT(bp) \|\| BP_IS_HOLE(bp));

	/* for convenience, so callers don't need to check */
	if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
	memset(mac, 0, ZIO_DATA_MAC_LEN);
	return;
	}

	if (!BP_SHOULD_BYTESWAP(bp)) {
	memcpy(mac, &bp->blk_cksum.zc_word[2], sizeof (uint64_t));
	memcpy(mac + sizeof (uint64_t), &bp->blk_cksum.zc_word[3],
	sizeof (uint64_t));
	} else {
	val64 = BSWAP_64(bp->blk_cksum.zc_word[2]);
	memcpy(mac, &val64, sizeof (uint64_t));

	val64 = BSWAP_64(bp->blk_cksum.zc_word[3]);
	memcpy(mac + sizeof (uint64_t), &val64, sizeof (uint64_t));
	}
	}

	void
	zio_crypt_encode_mac_zil(void data, uint8_t mac)
	{
	zil_chain_t *zilc = data;

	memcpy(&zilc->zc_eck.zec_cksum.zc_word[2], mac, sizeof (uint64_t));
	memcpy(&zilc->zc_eck.zec_cksum.zc_word[3], mac + sizeof (uint64_t),
	sizeof (uint64_t));
	}

	void
	zio_crypt_decode_mac_zil(const void data, uint8_t mac)
	{
	/*
	* The ZIL MAC is embedded in the block it protects, which will
	* not have been byteswapped by the time this function has been called.
	* As a result, we don't need to worry about byteswapping the MAC.
	*/
	const zil_chain_t *zilc = data;

	memcpy(mac, &zilc->zc_eck.zec_cksum.zc_word[2], sizeof (uint64_t));
	memcpy(mac + sizeof (uint64_t), &zilc->zc_eck.zec_cksum.zc_word[3],
	sizeof (uint64_t));
	}

	/*
	* This routine takes a block of dnodes (src_abd) and copies only the bonus
	* buffers to the same offsets in the dst buffer. datalen should be the size
	* of both the src_abd and the dst buffer (not just the length of the bonus
	* buffers).
	*/
	void
	zio_crypt_copy_dnode_bonus(abd_t src_abd, uint8_t dst, uint_t datalen)
	{
	uint_t i, max_dnp = datalen >> DNODE_SHIFT;
	uint8_t *src;
	dnode_phys_t dnp, sdnp, *ddnp;

	src = abd_borrow_buf_copy(src_abd, datalen);

	sdnp = (dnode_phys_t *)src;
	ddnp = (dnode_phys_t *)dst;

	for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
	dnp = &sdnp[i];
	if (dnp->dn_type != DMU_OT_NONE &&
	DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
	dnp->dn_bonuslen != 0) {
	memcpy(DN_BONUS(&ddnp[i]), DN_BONUS(dnp),
	DN_MAX_BONUS_LEN(dnp));
	}
	}

	abd_return_buf(src_abd, src, datalen);
	}

	/*
	* This function decides what fields from blk_prop are included in
	* the on-disk various MAC algorithms.
	*/
	static void
	zio_crypt_bp_zero_nonportable_blkprop(blkptr_t *bp, uint64_t version)
	{
	/*
	* Version 0 did not properly zero out all non-portable fields
	* as it should have done. We maintain this code so that we can
	* do read-only imports of pools on this version.
	*/
	if (version == 0) {
	BP_SET_DEDUP(bp, 0);
	BP_SET_CHECKSUM(bp, 0);
	BP_SET_PSIZE(bp, SPA_MINBLOCKSIZE);
	return;
	}

	ASSERT3U(version, ==, ZIO_CRYPT_KEY_CURRENT_VERSION);

	/*
	* The hole_birth feature might set these fields even if this bp
	* is a hole. We zero them out here to guarantee that raw sends
	* will function with or without the feature.
	*/
	if (BP_IS_HOLE(bp)) {
	bp->blk_prop = 0ULL;
	return;
	}

	/*
	* At L0 we want to verify these fields to ensure that data blocks
	* can not be reinterpreted. For instance, we do not want an attacker
	* to trick us into returning raw lz4 compressed data to the user
	* by modifying the compression bits. At higher levels, we cannot
	* enforce this policy since raw sends do not convey any information
	* about indirect blocks, so these values might be different on the
	* receive side. Fortunately, this does not open any new attack
	* vectors, since any alterations that can be made to a higher level
	* bp must still verify the correct order of the layer below it.
	*/
	if (BP_GET_LEVEL(bp) != 0) {
	BP_SET_BYTEORDER(bp, 0);
	BP_SET_COMPRESS(bp, 0);

	/*
	* psize cannot be set to zero or it will trigger
	* asserts, but the value doesn't really matter as
	* long as it is constant.
	*/
	BP_SET_PSIZE(bp, SPA_MINBLOCKSIZE);
	}

	BP_SET_DEDUP(bp, 0);
	BP_SET_CHECKSUM(bp, 0);
	}

	static void
	zio_crypt_bp_auth_init(uint64_t version, boolean_t should_bswap, blkptr_t *bp,
	blkptr_auth_buf_t bab, uint_t bab_len)
	{
	blkptr_t tmpbp = *bp;

	if (should_bswap)
	byteswap_uint64_array(&tmpbp, sizeof (blkptr_t));

	ASSERT(BP_USES_CRYPT(&tmpbp) \|\| BP_IS_HOLE(&tmpbp));
	ASSERT0(BP_IS_EMBEDDED(&tmpbp));

	zio_crypt_decode_mac_bp(&tmpbp, bab->bab_mac);

	/*
	* We always MAC blk_prop in LE to ensure portability. This
	* must be done after decoding the mac, since the endianness
	* will get zero'd out here.
	*/
	zio_crypt_bp_zero_nonportable_blkprop(&tmpbp, version);
	bab->bab_prop = LE_64(tmpbp.blk_prop);
	bab->bab_pad = 0ULL;

	/* version 0 did not include the padding */
	*bab_len = sizeof (blkptr_auth_buf_t);
	if (version == 0)
	*bab_len -= sizeof (uint64_t);
	}

	static int
	zio_crypt_bp_do_hmac_updates(crypto_context_t ctx, uint64_t version,
	boolean_t should_bswap, blkptr_t *bp)
	{
	int ret;
	uint_t bab_len;
	blkptr_auth_buf_t bab;
	crypto_data_t cd;

	zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
	cd.cd_format = CRYPTO_DATA_RAW;
	cd.cd_offset = 0;
	cd.cd_length = bab_len;
	cd.cd_raw.iov_base = (char *)&bab;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_update(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	return (0);

	error:
	return (ret);
	}

	static void
	zio_crypt_bp_do_indrect_checksum_updates(SHA2_CTX *ctx, uint64_t version,
	boolean_t should_bswap, blkptr_t *bp)
	{
	uint_t bab_len;
	blkptr_auth_buf_t bab;

	zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
	SHA2Update(ctx, &bab, bab_len);
	}

	static void
	zio_crypt_bp_do_aad_updates(uint8_t *aadp, uint_t aad_len, uint64_t version,
	boolean_t should_bswap, blkptr_t *bp)
	{
	uint_t bab_len;
	blkptr_auth_buf_t bab;

	zio_crypt_bp_auth_init(version, should_bswap, bp, &bab, &bab_len);
	memcpy(*aadp, &bab, bab_len);
	*aadp += bab_len;
	*aad_len += bab_len;
	}

	static int
	zio_crypt_do_dnode_hmac_updates(crypto_context_t ctx, uint64_t version,
	boolean_t should_bswap, dnode_phys_t *dnp)
	{
	int ret, i;
	dnode_phys_t *adnp, tmp_dncore;
	size_t dn_core_size = offsetof(dnode_phys_t, dn_blkptr);
	boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
	crypto_data_t cd;

	cd.cd_format = CRYPTO_DATA_RAW;
	cd.cd_offset = 0;

	/*
	* Authenticate the core dnode (masking out non-portable bits).
	* We only copy the first 64 bytes we operate on to avoid the overhead
	* of copying 512-64 unneeded bytes. The compiler seems to be fine
	* with that.
	*/
	memcpy(&tmp_dncore, dnp, dn_core_size);
	adnp = &tmp_dncore;

	if (le_bswap) {
	adnp->dn_datablkszsec = BSWAP_16(adnp->dn_datablkszsec);
	adnp->dn_bonuslen = BSWAP_16(adnp->dn_bonuslen);
	adnp->dn_maxblkid = BSWAP_64(adnp->dn_maxblkid);
	adnp->dn_used = BSWAP_64(adnp->dn_used);
	}
	adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
	adnp->dn_used = 0;

	cd.cd_length = dn_core_size;
	cd.cd_raw.iov_base = (char *)adnp;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_update(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	for (i = 0; i < dnp->dn_nblkptr; i++) {
	ret = zio_crypt_bp_do_hmac_updates(ctx, version,
	should_bswap, &dnp->dn_blkptr[i]);
	if (ret != 0)
	goto error;
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	ret = zio_crypt_bp_do_hmac_updates(ctx, version,
	should_bswap, DN_SPILL_BLKPTR(dnp));
	if (ret != 0)
	goto error;
	}

	return (0);

	error:
	return (ret);
	}

	/*
	* objset_phys_t blocks introduce a number of exceptions to the normal
	* authentication process. objset_phys_t's contain 2 separate HMACS for
	* protecting the integrity of their data. The portable_mac protects the
	* metadnode. This MAC can be sent with a raw send and protects against
	* reordering of data within the metadnode. The local_mac protects the user
	* accounting objects which are not sent from one system to another.
	*
	* In addition, objset blocks are the only blocks that can be modified and
	* written to disk without the key loaded under certain circumstances. During
	* zil_claim() we need to be able to update the zil_header_t to complete
	* claiming log blocks and during raw receives we need to write out the
	* portable_mac from the send file. Both of these actions are possible
	* because these fields are not protected by either MAC so neither one will
	* need to modify the MACs without the key. However, when the modified blocks
	* are written out they will be byteswapped into the host machine's native
	* endianness which will modify fields protected by the MAC. As a result, MAC
	* calculation for objset blocks works slightly differently from other block
	* types. Where other block types MAC the data in whatever endianness is
	* written to disk, objset blocks always MAC little endian version of their
	* values. In the code, should_bswap is the value from BP_SHOULD_BYTESWAP()
	* and le_bswap indicates whether a byteswap is needed to get this block
	* into little endian format.
	*/
	int
	zio_crypt_do_objset_hmacs(zio_crypt_key_t key, void data, uint_t datalen,
	boolean_t should_bswap, uint8_t portable_mac, uint8_t local_mac)
	{
	int ret;
	crypto_mechanism_t mech;
	crypto_context_t ctx;
	crypto_data_t cd;
	objset_phys_t *osp = data;
	uint64_t intval;
	boolean_t le_bswap = (should_bswap == ZFS_HOST_BYTEORDER);
	uint8_t raw_portable_mac[SHA512_DIGEST_LENGTH];
	uint8_t raw_local_mac[SHA512_DIGEST_LENGTH];

	/* initialize HMAC mechanism */
	mech.cm_type = crypto_mech2id(SUN_CKM_SHA512_HMAC);
	mech.cm_param = NULL;
	mech.cm_param_len = 0;

	cd.cd_format = CRYPTO_DATA_RAW;
	cd.cd_offset = 0;

	/* calculate the portable MAC from the portable fields and metadnode */
	ret = crypto_mac_init(&mech, &key->zk_hmac_key, NULL, &ctx);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	/* add in the os_type */
	intval = (le_bswap) ? osp->os_type : BSWAP_64(osp->os_type);
	cd.cd_length = sizeof (uint64_t);
	cd.cd_raw.iov_base = (char *)&intval;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_update(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	/* add in the portable os_flags */
	intval = osp->os_flags;
	if (should_bswap)
	intval = BSWAP_64(intval);
	intval &= OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
	if (!ZFS_HOST_BYTEORDER)
	intval = BSWAP_64(intval);

	cd.cd_length = sizeof (uint64_t);
	cd.cd_raw.iov_base = (char *)&intval;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_update(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	/* add in fields from the metadnode */
	ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
	should_bswap, &osp->os_meta_dnode);
	if (ret)
	goto error;

	/* store the final digest in a temporary buffer and copy what we need */
	cd.cd_length = SHA512_DIGEST_LENGTH;
	cd.cd_raw.iov_base = (char *)raw_portable_mac;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_final(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	memcpy(portable_mac, raw_portable_mac, ZIO_OBJSET_MAC_LEN);

	/*
	* This is necessary here as we check next whether
	* OBJSET_FLAG_USERACCOUNTING_COMPLETE is set in order to
	* decide if the local_mac should be zeroed out. That flag will always
	* be set by dmu_objset_id_quota_upgrade_cb() and
	* dmu_objset_userspace_upgrade_cb() if useraccounting has been
	* completed.
	*/
	intval = osp->os_flags;
	if (should_bswap)
	intval = BSWAP_64(intval);
	boolean_t uacct_incomplete =
	!(intval & OBJSET_FLAG_USERACCOUNTING_COMPLETE);

	/*
	* The local MAC protects the user, group and project accounting.
	* If these objects are not present, the local MAC is zeroed out.
	*/
	if (uacct_incomplete \|\|
	(datalen >= OBJSET_PHYS_SIZE_V3 &&
	osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
	osp->os_groupused_dnode.dn_type == DMU_OT_NONE &&
	osp->os_projectused_dnode.dn_type == DMU_OT_NONE) \|\|
	(datalen >= OBJSET_PHYS_SIZE_V2 &&
	osp->os_userused_dnode.dn_type == DMU_OT_NONE &&
	osp->os_groupused_dnode.dn_type == DMU_OT_NONE) \|\|
	(datalen <= OBJSET_PHYS_SIZE_V1)) {
	memset(local_mac, 0, ZIO_OBJSET_MAC_LEN);
	return (0);
	}

	/* calculate the local MAC from the userused and groupused dnodes */
	ret = crypto_mac_init(&mech, &key->zk_hmac_key, NULL, &ctx);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	/* add in the non-portable os_flags */
	intval = osp->os_flags;
	if (should_bswap)
	intval = BSWAP_64(intval);
	intval &= ~OBJSET_CRYPT_PORTABLE_FLAGS_MASK;
	if (!ZFS_HOST_BYTEORDER)
	intval = BSWAP_64(intval);

	cd.cd_length = sizeof (uint64_t);
	cd.cd_raw.iov_base = (char *)&intval;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_update(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	/* add in fields from the user accounting dnodes */
	if (osp->os_userused_dnode.dn_type != DMU_OT_NONE) {
	ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
	should_bswap, &osp->os_userused_dnode);
	if (ret)
	goto error;
	}

	if (osp->os_groupused_dnode.dn_type != DMU_OT_NONE) {
	ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
	should_bswap, &osp->os_groupused_dnode);
	if (ret)
	goto error;
	}

	if (osp->os_projectused_dnode.dn_type != DMU_OT_NONE &&
	datalen >= OBJSET_PHYS_SIZE_V3) {
	ret = zio_crypt_do_dnode_hmac_updates(ctx, key->zk_version,
	should_bswap, &osp->os_projectused_dnode);
	if (ret)
	goto error;
	}

	/* store the final digest in a temporary buffer and copy what we need */
	cd.cd_length = SHA512_DIGEST_LENGTH;
	cd.cd_raw.iov_base = (char *)raw_local_mac;
	cd.cd_raw.iov_len = cd.cd_length;

	ret = crypto_mac_final(ctx, &cd);
	if (ret != CRYPTO_SUCCESS) {
	ret = SET_ERROR(EIO);
	goto error;
	}

	memcpy(local_mac, raw_local_mac, ZIO_OBJSET_MAC_LEN);

	return (0);

	error:
	memset(portable_mac, 0, ZIO_OBJSET_MAC_LEN);
	memset(local_mac, 0, ZIO_OBJSET_MAC_LEN);
	return (ret);
	}

	static void
	zio_crypt_destroy_uio(zfs_uio_t *uio)
	{
	if (uio->uio_iov)
	kmem_free(uio->uio_iov, uio->uio_iovcnt * sizeof (iovec_t));
	}

	/*
	* This function parses an uncompressed indirect block and returns a checksum
	* of all the portable fields from all of the contained bps. The portable
	* fields are the MAC and all of the fields from blk_prop except for the dedup,
	* checksum, and psize bits. For an explanation of the purpose of this, see
	* the comment block on object set authentication.
	*/
	static int
	zio_crypt_do_indirect_mac_checksum_impl(boolean_t generate, void *buf,
	uint_t datalen, uint64_t version, boolean_t byteswap, uint8_t *cksum)
	{
	blkptr_t *bp;
	int i, epb = datalen >> SPA_BLKPTRSHIFT;
	SHA2_CTX ctx;
	uint8_t digestbuf[SHA512_DIGEST_LENGTH];

	/* checksum all of the MACs from the layer below */
	SHA2Init(SHA512, &ctx);
	for (i = 0, bp = buf; i < epb; i++, bp++) {
	zio_crypt_bp_do_indrect_checksum_updates(&ctx, version,
	byteswap, bp);
	}
	SHA2Final(digestbuf, &ctx);

	if (generate) {
	memcpy(cksum, digestbuf, ZIO_DATA_MAC_LEN);
	return (0);
	}

	if (memcmp(digestbuf, cksum, ZIO_DATA_MAC_LEN) != 0)
	return (SET_ERROR(ECKSUM));

	return (0);
	}

	int
	zio_crypt_do_indirect_mac_checksum(boolean_t generate, void *buf,
	uint_t datalen, boolean_t byteswap, uint8_t *cksum)
	{
	int ret;

	/*
	* Unfortunately, callers of this function will not always have
	* easy access to the on-disk format version. This info is
	* normally found in the DSL Crypto Key, but the checksum-of-MACs
	* is expected to be verifiable even when the key isn't loaded.
	* Here, instead of doing a ZAP lookup for the version for each
	* zio, we simply try both existing formats.
	*/
	ret = zio_crypt_do_indirect_mac_checksum_impl(generate, buf,
	datalen, ZIO_CRYPT_KEY_CURRENT_VERSION, byteswap, cksum);
	if (ret == ECKSUM) {
	ASSERT(!generate);
	ret = zio_crypt_do_indirect_mac_checksum_impl(generate,
	buf, datalen, 0, byteswap, cksum);
	}

	return (ret);
	}

	int
	zio_crypt_do_indirect_mac_checksum_abd(boolean_t generate, abd_t *abd,
	uint_t datalen, boolean_t byteswap, uint8_t *cksum)
	{
	int ret;
	void *buf;

	buf = abd_borrow_buf_copy(abd, datalen);
	ret = zio_crypt_do_indirect_mac_checksum(generate, buf, datalen,
	byteswap, cksum);
	abd_return_buf(abd, buf, datalen);

	return (ret);
	}

	/*
	* Special case handling routine for encrypting / decrypting ZIL blocks.
	* We do not check for the older ZIL chain because the encryption feature
	* was not available before the newer ZIL chain was introduced. The goal
	* here is to encrypt everything except the blkptr_t of a lr_write_t and
	* the zil_chain_t header. Everything that is not encrypted is authenticated.
	*/
	static int
	zio_crypt_init_uios_zil(boolean_t encrypt, uint8_t *plainbuf,
	uint8_t cipherbuf, uint_t datalen, boolean_t byteswap, zfs_uio_t puio,
	zfs_uio_t cuio, uint_t enc_len, uint8_t *authbuf, uint_t auth_len,
	boolean_t *no_crypt)
	{
	int ret;
	uint64_t txtype, lr_len;
	uint_t nr_src, nr_dst, crypt_len;
	uint_t aad_len = 0, nr_iovecs = 0, total_len = 0;
	iovec_t src_iovecs = NULL, dst_iovecs = NULL;
	uint8_t src, dst, slrp, dlrp, blkend, aadp;
	zil_chain_t *zilc;
	lr_t *lr;
	uint8_t *aadbuf = zio_buf_alloc(datalen);

	/* cipherbuf always needs an extra iovec for the MAC */
	if (encrypt) {
	src = plainbuf;
	dst = cipherbuf;
	nr_src = 0;
	nr_dst = 1;
	} else {
	src = cipherbuf;
	dst = plainbuf;
	nr_src = 1;
	nr_dst = 0;
	}
	memset(dst, 0, datalen);

	/* find the start and end record of the log block */
	zilc = (zil_chain_t *)src;
	slrp = src + sizeof (zil_chain_t);
	aadp = aadbuf;
	blkend = src + ((byteswap) ? BSWAP_64(zilc->zc_nused) : zilc->zc_nused);

	/* calculate the number of encrypted iovecs we will need */
	for (; slrp < blkend; slrp += lr_len) {
	lr = (lr_t *)slrp;

	if (!byteswap) {
	txtype = lr->lrc_txtype;
	lr_len = lr->lrc_reclen;
	} else {
	txtype = BSWAP_64(lr->lrc_txtype);
	lr_len = BSWAP_64(lr->lrc_reclen);
	}

	nr_iovecs++;
	if (txtype == TX_WRITE && lr_len != sizeof (lr_write_t))
	nr_iovecs++;
	}

	nr_src += nr_iovecs;
	nr_dst += nr_iovecs;

	/* allocate the iovec arrays */
	if (nr_src != 0) {
	src_iovecs = kmem_alloc(nr_src * sizeof (iovec_t), KM_SLEEP);
	if (src_iovecs == NULL) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}
	}

	if (nr_dst != 0) {
	dst_iovecs = kmem_alloc(nr_dst * sizeof (iovec_t), KM_SLEEP);
	if (dst_iovecs == NULL) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}
	}

	/*
	* Copy the plain zil header over and authenticate everything except
	* the checksum that will store our MAC. If we are writing the data
	* the embedded checksum will not have been calculated yet, so we don't
	* authenticate that.
	*/
	memcpy(dst, src, sizeof (zil_chain_t));
	memcpy(aadp, src, sizeof (zil_chain_t) - sizeof (zio_eck_t));
	aadp += sizeof (zil_chain_t) - sizeof (zio_eck_t);
	aad_len += sizeof (zil_chain_t) - sizeof (zio_eck_t);

	/* loop over records again, filling in iovecs */
	nr_iovecs = 0;
	slrp = src + sizeof (zil_chain_t);
	dlrp = dst + sizeof (zil_chain_t);

	for (; slrp < blkend; slrp += lr_len, dlrp += lr_len) {
	lr = (lr_t *)slrp;

	if (!byteswap) {
	txtype = lr->lrc_txtype;
	lr_len = lr->lrc_reclen;
	} else {
	txtype = BSWAP_64(lr->lrc_txtype);
	lr_len = BSWAP_64(lr->lrc_reclen);
	}

	/* copy the common lr_t */
	memcpy(dlrp, slrp, sizeof (lr_t));
	memcpy(aadp, slrp, sizeof (lr_t));
	aadp += sizeof (lr_t);
	aad_len += sizeof (lr_t);

	ASSERT3P(src_iovecs, !=, NULL);
	ASSERT3P(dst_iovecs, !=, NULL);

	/*
	* If this is a TX_WRITE record we want to encrypt everything
	* except the bp if exists. If the bp does exist we want to
	* authenticate it.
	*/
	if (txtype == TX_WRITE) {
	crypt_len = sizeof (lr_write_t) -
	sizeof (lr_t) - sizeof (blkptr_t);
	src_iovecs[nr_iovecs].iov_base = slrp + sizeof (lr_t);
	src_iovecs[nr_iovecs].iov_len = crypt_len;
	dst_iovecs[nr_iovecs].iov_base = dlrp + sizeof (lr_t);
	dst_iovecs[nr_iovecs].iov_len = crypt_len;

	/* copy the bp now since it will not be encrypted */
	memcpy(dlrp + sizeof (lr_write_t) - sizeof (blkptr_t),
	slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
	sizeof (blkptr_t));
	memcpy(aadp,
	slrp + sizeof (lr_write_t) - sizeof (blkptr_t),
	sizeof (blkptr_t));
	aadp += sizeof (blkptr_t);
	aad_len += sizeof (blkptr_t);
	nr_iovecs++;
	total_len += crypt_len;

	if (lr_len != sizeof (lr_write_t)) {
	crypt_len = lr_len - sizeof (lr_write_t);
	src_iovecs[nr_iovecs].iov_base =
	slrp + sizeof (lr_write_t);
	src_iovecs[nr_iovecs].iov_len = crypt_len;
	dst_iovecs[nr_iovecs].iov_base =
	dlrp + sizeof (lr_write_t);
	dst_iovecs[nr_iovecs].iov_len = crypt_len;
	nr_iovecs++;
	total_len += crypt_len;
	}
	} else {
	crypt_len = lr_len - sizeof (lr_t);
	src_iovecs[nr_iovecs].iov_base = slrp + sizeof (lr_t);
	src_iovecs[nr_iovecs].iov_len = crypt_len;
	dst_iovecs[nr_iovecs].iov_base = dlrp + sizeof (lr_t);
	dst_iovecs[nr_iovecs].iov_len = crypt_len;
	nr_iovecs++;
	total_len += crypt_len;
	}
	}

	*no_crypt = (nr_iovecs == 0);
	*enc_len = total_len;
	*authbuf = aadbuf;
	*auth_len = aad_len;

	if (encrypt) {
	puio->uio_iov = src_iovecs;
	puio->uio_iovcnt = nr_src;
	cuio->uio_iov = dst_iovecs;
	cuio->uio_iovcnt = nr_dst;
	} else {
	puio->uio_iov = dst_iovecs;
	puio->uio_iovcnt = nr_dst;
	cuio->uio_iov = src_iovecs;
	cuio->uio_iovcnt = nr_src;
	}

	return (0);

	error:
	zio_buf_free(aadbuf, datalen);
	if (src_iovecs != NULL)
	kmem_free(src_iovecs, nr_src * sizeof (iovec_t));
	if (dst_iovecs != NULL)
	kmem_free(dst_iovecs, nr_dst * sizeof (iovec_t));

	*enc_len = 0;
	*authbuf = NULL;
	*auth_len = 0;
	*no_crypt = B_FALSE;
	puio->uio_iov = NULL;
	puio->uio_iovcnt = 0;
	cuio->uio_iov = NULL;
	cuio->uio_iovcnt = 0;
	return (ret);
	}

	/*
	* Special case handling routine for encrypting / decrypting dnode blocks.
	*/
	static int
	zio_crypt_init_uios_dnode(boolean_t encrypt, uint64_t version,
	uint8_t plainbuf, uint8_t cipherbuf, uint_t datalen, boolean_t byteswap,
	zfs_uio_t puio, zfs_uio_t cuio, uint_t enc_len, uint8_t *authbuf,
	uint_t auth_len, boolean_t no_crypt)
	{
	int ret;
	uint_t nr_src, nr_dst, crypt_len;
	uint_t aad_len = 0, nr_iovecs = 0, total_len = 0;
	uint_t i, j, max_dnp = datalen >> DNODE_SHIFT;
	iovec_t src_iovecs = NULL, dst_iovecs = NULL;
	uint8_t src, dst, *aadp;
	dnode_phys_t dnp, adnp, sdnp, ddnp;
	uint8_t *aadbuf = zio_buf_alloc(datalen);

	if (encrypt) {
	src = plainbuf;
	dst = cipherbuf;
	nr_src = 0;
	nr_dst = 1;
	} else {
	src = cipherbuf;
	dst = plainbuf;
	nr_src = 1;
	nr_dst = 0;
	}

	sdnp = (dnode_phys_t *)src;
	ddnp = (dnode_phys_t *)dst;
	aadp = aadbuf;

	/*
	* Count the number of iovecs we will need to do the encryption by
	* counting the number of bonus buffers that need to be encrypted.
	*/
	for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
	/*
	* This block may still be byteswapped. However, all of the
	* values we use are either uint8_t's (for which byteswapping
	* is a noop) or a * != 0 check, which will work regardless
	* of whether or not we byteswap.
	*/
	if (sdnp[i].dn_type != DMU_OT_NONE &&
	DMU_OT_IS_ENCRYPTED(sdnp[i].dn_bonustype) &&
	sdnp[i].dn_bonuslen != 0) {
	nr_iovecs++;
	}
	}

	nr_src += nr_iovecs;
	nr_dst += nr_iovecs;

	if (nr_src != 0) {
	src_iovecs = kmem_alloc(nr_src * sizeof (iovec_t), KM_SLEEP);
	if (src_iovecs == NULL) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}
	}

	if (nr_dst != 0) {
	dst_iovecs = kmem_alloc(nr_dst * sizeof (iovec_t), KM_SLEEP);
	if (dst_iovecs == NULL) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}
	}

	nr_iovecs = 0;

	/*
	* Iterate through the dnodes again, this time filling in the uios
	* we allocated earlier. We also concatenate any data we want to
	* authenticate onto aadbuf.
	*/
	for (i = 0; i < max_dnp; i += sdnp[i].dn_extra_slots + 1) {
	dnp = &sdnp[i];

	/* copy over the core fields and blkptrs (kept as plaintext) */
	memcpy(&ddnp[i], dnp,
	(uint8_t )DN_BONUS(dnp) - (uint8_t )dnp);

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	memcpy(DN_SPILL_BLKPTR(&ddnp[i]), DN_SPILL_BLKPTR(dnp),
	sizeof (blkptr_t));
	}

	/*
	* Handle authenticated data. We authenticate everything in
	* the dnode that can be brought over when we do a raw send.
	* This includes all of the core fields as well as the MACs
	* stored in the bp checksums and all of the portable bits
	* from blk_prop. We include the dnode padding here in case it
	* ever gets used in the future. Some dn_flags and dn_used are
	* not portable so we mask those out values out of the
	* authenticated data.
	*/
	crypt_len = offsetof(dnode_phys_t, dn_blkptr);
	memcpy(aadp, dnp, crypt_len);
	adnp = (dnode_phys_t *)aadp;
	adnp->dn_flags &= DNODE_CRYPT_PORTABLE_FLAGS_MASK;
	adnp->dn_used = 0;
	aadp += crypt_len;
	aad_len += crypt_len;

	for (j = 0; j < dnp->dn_nblkptr; j++) {
	zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
	version, byteswap, &dnp->dn_blkptr[j]);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	zio_crypt_bp_do_aad_updates(&aadp, &aad_len,
	version, byteswap, DN_SPILL_BLKPTR(dnp));
	}

	/*
	* If this bonus buffer needs to be encrypted, we prepare an
	* iovec_t. The encryption / decryption functions will fill
	* this in for us with the encrypted or decrypted data.
	* Otherwise we add the bonus buffer to the authenticated
	* data buffer and copy it over to the destination. The
	* encrypted iovec extends to DN_MAX_BONUS_LEN(dnp) so that
	* we can guarantee alignment with the AES block size
	* (128 bits).
	*/
	crypt_len = DN_MAX_BONUS_LEN(dnp);
	if (dnp->dn_type != DMU_OT_NONE &&
	DMU_OT_IS_ENCRYPTED(dnp->dn_bonustype) &&
	dnp->dn_bonuslen != 0) {
	ASSERT3U(nr_iovecs, <, nr_src);
	ASSERT3U(nr_iovecs, <, nr_dst);
	ASSERT3P(src_iovecs, !=, NULL);
	ASSERT3P(dst_iovecs, !=, NULL);
	src_iovecs[nr_iovecs].iov_base = DN_BONUS(dnp);
	src_iovecs[nr_iovecs].iov_len = crypt_len;
	dst_iovecs[nr_iovecs].iov_base = DN_BONUS(&ddnp[i]);
	dst_iovecs[nr_iovecs].iov_len = crypt_len;

	nr_iovecs++;
	total_len += crypt_len;
	} else {
	memcpy(DN_BONUS(&ddnp[i]), DN_BONUS(dnp), crypt_len);
	memcpy(aadp, DN_BONUS(dnp), crypt_len);
	aadp += crypt_len;
	aad_len += crypt_len;
	}
	}

	*no_crypt = (nr_iovecs == 0);
	*enc_len = total_len;
	*authbuf = aadbuf;
	*auth_len = aad_len;

	if (encrypt) {
	puio->uio_iov = src_iovecs;
	puio->uio_iovcnt = nr_src;
	cuio->uio_iov = dst_iovecs;
	cuio->uio_iovcnt = nr_dst;
	} else {
	puio->uio_iov = dst_iovecs;
	puio->uio_iovcnt = nr_dst;
	cuio->uio_iov = src_iovecs;
	cuio->uio_iovcnt = nr_src;
	}

	return (0);

	error:
	zio_buf_free(aadbuf, datalen);
	if (src_iovecs != NULL)
	kmem_free(src_iovecs, nr_src * sizeof (iovec_t));
	if (dst_iovecs != NULL)
	kmem_free(dst_iovecs, nr_dst * sizeof (iovec_t));

	*enc_len = 0;
	*authbuf = NULL;
	*auth_len = 0;
	*no_crypt = B_FALSE;
	puio->uio_iov = NULL;
	puio->uio_iovcnt = 0;
	cuio->uio_iov = NULL;
	cuio->uio_iovcnt = 0;
	return (ret);
	}

	static int
	zio_crypt_init_uios_normal(boolean_t encrypt, uint8_t *plainbuf,
	uint8_t cipherbuf, uint_t datalen, zfs_uio_t puio, zfs_uio_t *cuio,
	uint_t *enc_len)
	{
	(void) encrypt;
	int ret;
	uint_t nr_plain = 1, nr_cipher = 2;
	iovec_t plain_iovecs = NULL, cipher_iovecs = NULL;

	/* allocate the iovecs for the plain and cipher data */
	plain_iovecs = kmem_alloc(nr_plain * sizeof (iovec_t),
	KM_SLEEP);
	if (!plain_iovecs) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}

	cipher_iovecs = kmem_alloc(nr_cipher * sizeof (iovec_t),
	KM_SLEEP);
	if (!cipher_iovecs) {
	ret = SET_ERROR(ENOMEM);
	goto error;
	}

	plain_iovecs[0].iov_base = plainbuf;
	plain_iovecs[0].iov_len = datalen;
	cipher_iovecs[0].iov_base = cipherbuf;
	cipher_iovecs[0].iov_len = datalen;

	*enc_len = datalen;
	puio->uio_iov = plain_iovecs;
	puio->uio_iovcnt = nr_plain;
	cuio->uio_iov = cipher_iovecs;
	cuio->uio_iovcnt = nr_cipher;

	return (0);

	error:
	if (plain_iovecs != NULL)
	kmem_free(plain_iovecs, nr_plain * sizeof (iovec_t));
	if (cipher_iovecs != NULL)
	kmem_free(cipher_iovecs, nr_cipher * sizeof (iovec_t));

	*enc_len = 0;
	puio->uio_iov = NULL;
	puio->uio_iovcnt = 0;
	cuio->uio_iov = NULL;
	cuio->uio_iovcnt = 0;
	return (ret);
	}

	/*
	* This function builds up the plaintext (puio) and ciphertext (cuio) uios so
	* that they can be used for encryption and decryption by zio_do_crypt_uio().
	* Most blocks will use zio_crypt_init_uios_normal(), with ZIL and dnode blocks
	* requiring special handling to parse out pieces that are to be encrypted. The
	* authbuf is used by these special cases to store additional authenticated
	* data (AAD) for the encryption modes.
	*/
	static int
	zio_crypt_init_uios(boolean_t encrypt, uint64_t version, dmu_object_type_t ot,
	uint8_t plainbuf, uint8_t cipherbuf, uint_t datalen, boolean_t byteswap,
	uint8_t mac, zfs_uio_t puio, zfs_uio_t cuio, uint_t enc_len,
	uint8_t *authbuf, uint_t auth_len, boolean_t *no_crypt)
	{
	int ret;
	iovec_t *mac_iov;

	ASSERT(DMU_OT_IS_ENCRYPTED(ot) \|\| ot == DMU_OT_NONE);

	/* route to handler */
	switch (ot) {
	case DMU_OT_INTENT_LOG:
	ret = zio_crypt_init_uios_zil(encrypt, plainbuf, cipherbuf,
	datalen, byteswap, puio, cuio, enc_len, authbuf, auth_len,
	no_crypt);
	break;
	case DMU_OT_DNODE:
	ret = zio_crypt_init_uios_dnode(encrypt, version, plainbuf,
	cipherbuf, datalen, byteswap, puio, cuio, enc_len, authbuf,
	auth_len, no_crypt);
	break;
	default:
	ret = zio_crypt_init_uios_normal(encrypt, plainbuf, cipherbuf,
	datalen, puio, cuio, enc_len);
	*authbuf = NULL;
	*auth_len = 0;
	*no_crypt = B_FALSE;
	break;
	}

	if (ret != 0)
	goto error;

	/* populate the uios */
	puio->uio_segflg = UIO_SYSSPACE;
	cuio->uio_segflg = UIO_SYSSPACE;

	mac_iov = ((iovec_t *)&cuio->uio_iov[cuio->uio_iovcnt - 1]);
	mac_iov->iov_base = mac;
	mac_iov->iov_len = ZIO_DATA_MAC_LEN;

	return (0);

	error:
	return (ret);
	}

	/*
	* Primary encryption / decryption entrypoint for zio data.
	*/
	int
	zio_do_crypt_data(boolean_t encrypt, zio_crypt_key_t *key,
	dmu_object_type_t ot, boolean_t byteswap, uint8_t salt, uint8_t iv,
	uint8_t mac, uint_t datalen, uint8_t plainbuf, uint8_t *cipherbuf,
	boolean_t *no_crypt)
	{
	int ret;
	boolean_t locked = B_FALSE;
	uint64_t crypt = key->zk_crypt;
	uint_t keydata_len = zio_crypt_table[crypt].ci_keylen;
	uint_t enc_len, auth_len;
	zfs_uio_t puio, cuio;
	uint8_t enc_keydata[MASTER_KEY_MAX_LEN];
	crypto_key_t tmp_ckey, *ckey = NULL;
	crypto_ctx_template_t tmpl;
	uint8_t *authbuf = NULL;

	+ memset(&puio, 0, sizeof (puio));
	+ memset(&cuio, 0, sizeof (cuio));
	+
	/*
	* If the needed key is the current one, just use it. Otherwise we
	* need to generate a temporary one from the given salt + master key.
	* If we are encrypting, we must return a copy of the current salt
	* so that it can be stored in the blkptr_t.
	*/
	rw_enter(&key->zk_salt_lock, RW_READER);
	locked = B_TRUE;

	if (memcmp(salt, key->zk_salt, ZIO_DATA_SALT_LEN) == 0) {
	ckey = &key->zk_current_key;
	tmpl = key->zk_current_tmpl;
	} else {
	rw_exit(&key->zk_salt_lock);
	locked = B_FALSE;

	ret = hkdf_sha512(key->zk_master_keydata, keydata_len, NULL, 0,
	salt, ZIO_DATA_SALT_LEN, enc_keydata, keydata_len);
	if (ret != 0)
	goto error;

	tmp_ckey.ck_data = enc_keydata;
	tmp_ckey.ck_length = CRYPTO_BYTES2BITS(keydata_len);

	ckey = &tmp_ckey;
	tmpl = NULL;
	}

	/*
	* Attempt to use QAT acceleration if we can. We currently don't
	* do this for metadnode and ZIL blocks, since they have a much
	* more involved buffer layout and the qat_crypt() function only
	* works in-place.
	*/
	if (qat_crypt_use_accel(datalen) &&
	ot != DMU_OT_INTENT_LOG && ot != DMU_OT_DNODE) {
	uint8_t srcbuf, dstbuf;

	if (encrypt) {
	srcbuf = plainbuf;
	dstbuf = cipherbuf;
	} else {
	srcbuf = cipherbuf;
	dstbuf = plainbuf;
	}

	ret = qat_crypt((encrypt) ? QAT_ENCRYPT : QAT_DECRYPT, srcbuf,
	dstbuf, NULL, 0, iv, mac, ckey, key->zk_crypt, datalen);
	if (ret == CPA_STATUS_SUCCESS) {
	if (locked) {
	rw_exit(&key->zk_salt_lock);
	locked = B_FALSE;
	}

	return (0);
	}
	/* If the hardware implementation fails fall back to software */
	}

	- memset(&puio, 0, sizeof (puio));
	- memset(&cuio, 0, sizeof (cuio));
	-
	/* create uios for encryption */
	ret = zio_crypt_init_uios(encrypt, key->zk_version, ot, plainbuf,
	cipherbuf, datalen, byteswap, mac, &puio, &cuio, &enc_len,
	&authbuf, &auth_len, no_crypt);
	if (ret != 0)
	goto error;

	/* perform the encryption / decryption in software */
	ret = zio_do_crypt_uio(encrypt, key->zk_crypt, ckey, tmpl, iv, enc_len,
	&puio, &cuio, authbuf, auth_len);
	if (ret != 0)
	goto error;

	if (locked) {
	rw_exit(&key->zk_salt_lock);
	locked = B_FALSE;
	}

	if (authbuf != NULL)
	zio_buf_free(authbuf, datalen);
	if (ckey == &tmp_ckey)
	memset(enc_keydata, 0, keydata_len);
	zio_crypt_destroy_uio(&puio);
	zio_crypt_destroy_uio(&cuio);

	return (0);

	error:
	if (locked)
	rw_exit(&key->zk_salt_lock);
	if (authbuf != NULL)
	zio_buf_free(authbuf, datalen);
	if (ckey == &tmp_ckey)
	memset(enc_keydata, 0, keydata_len);
	zio_crypt_destroy_uio(&puio);
	zio_crypt_destroy_uio(&cuio);

	return (ret);
	}

	/*
	* Simple wrapper around zio_do_crypt_data() to work with abd's instead of
	* linear buffers.
	*/
	int
	zio_do_crypt_abd(boolean_t encrypt, zio_crypt_key_t *key, dmu_object_type_t ot,
	boolean_t byteswap, uint8_t salt, uint8_t iv, uint8_t *mac,
	uint_t datalen, abd_t pabd, abd_t cabd, boolean_t *no_crypt)
	{
	int ret;
	void ptmp, ctmp;

	if (encrypt) {
	ptmp = abd_borrow_buf_copy(pabd, datalen);
	ctmp = abd_borrow_buf(cabd, datalen);
	} else {
	ptmp = abd_borrow_buf(pabd, datalen);
	ctmp = abd_borrow_buf_copy(cabd, datalen);
	}

	ret = zio_do_crypt_data(encrypt, key, ot, byteswap, salt, iv, mac,
	datalen, ptmp, ctmp, no_crypt);
	if (ret != 0)
	goto error;

	if (encrypt) {
	abd_return_buf(pabd, ptmp, datalen);
	abd_return_buf_copy(cabd, ctmp, datalen);
	} else {
	abd_return_buf_copy(pabd, ptmp, datalen);
	abd_return_buf(cabd, ctmp, datalen);
	}

	return (0);

	error:
	if (encrypt) {
	abd_return_buf(pabd, ptmp, datalen);
	abd_return_buf_copy(cabd, ctmp, datalen);
	} else {
	abd_return_buf_copy(pabd, ptmp, datalen);
	abd_return_buf(cabd, ctmp, datalen);
	}

	return (ret);
	}

	#if defined(_KERNEL)
	/* CSTYLED */
	module_param(zfs_key_max_salt_uses, ulong, 0644);
	MODULE_PARM_DESC(zfs_key_max_salt_uses, "Max number of times a salt value "
	"can be used for generating encryption keys before it is rotated");
	#endif
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zpl_ctldir.c b/sys/contrib/openzfs/module/os/linux/zfs/zpl_ctldir.c
	index 1a688687ac4b..837629e4a5e0 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zpl_ctldir.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zpl_ctldir.c
	@@ -1,635 +1,635 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (C) 2011 Lawrence Livermore National Security, LLC.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* LLNL-CODE-403049.
	* Rewritten for Linux by:
	* Rohan Puri <rohan.puri15@gmail.com>
	* Brian Behlendorf <behlendorf1@llnl.gov>
	*/

	#include <sys/zfs_znode.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_vnops.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zpl.h>
	#include <sys/dmu.h>
	#include <sys/dsl_dataset.h>
	#include <sys/zap.h>

	/*
	* Common open routine. Disallow any write access.
	*/
	static int
	zpl_common_open(struct inode ip, struct file filp)
	{
	if (filp->f_mode & FMODE_WRITE)
	return (-EACCES);

	return (generic_file_open(ip, filp));
	}

	/*
	* Get root directory contents.
	*/
	static int
	zpl_root_iterate(struct file filp, zpl_dir_context_t ctx)
	{
	zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
	int error = 0;

	if ((error = zpl_enter(zfsvfs, FTAG)) != 0)
	return (error);

	if (!zpl_dir_emit_dots(filp, ctx))
	goto out;

	if (ctx->pos == 2) {
	if (!zpl_dir_emit(ctx, ZFS_SNAPDIR_NAME,
	strlen(ZFS_SNAPDIR_NAME), ZFSCTL_INO_SNAPDIR, DT_DIR))
	goto out;

	ctx->pos++;
	}

	if (ctx->pos == 3) {
	if (!zpl_dir_emit(ctx, ZFS_SHAREDIR_NAME,
	strlen(ZFS_SHAREDIR_NAME), ZFSCTL_INO_SHARES, DT_DIR))
	goto out;

	ctx->pos++;
	}
	out:
	zpl_exit(zfsvfs, FTAG);

	return (error);
	}

	#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
	static int
	zpl_root_readdir(struct file filp, void dirent, filldir_t filldir)
	{
	zpl_dir_context_t ctx =
	ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
	int error;

	error = zpl_root_iterate(filp, &ctx);
	filp->f_pos = ctx.pos;

	return (error);
	}
	#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */

	/*
	* Get root directory attributes.
	*/
	static int
	#ifdef HAVE_USERNS_IOPS_GETATTR
	zpl_root_getattr_impl(struct user_namespace *user_ns,
	const struct path path, struct kstat stat, u32 request_mask,
	unsigned int query_flags)
	#else
	zpl_root_getattr_impl(const struct path path, struct kstat stat,
	u32 request_mask, unsigned int query_flags)
	#endif
	{
	(void) request_mask, (void) query_flags;
	struct inode *ip = path->dentry->d_inode;

	#ifdef HAVE_USERNS_IOPS_GETATTR
	#ifdef HAVE_GENERIC_FILLATTR_USERNS
	generic_fillattr(user_ns, ip, stat);
	#else
	(void) user_ns;
	#endif
	#else
	generic_fillattr(ip, stat);
	#endif
	stat->atime = current_time(ip);

	return (0);
	}
	ZPL_GETATTR_WRAPPER(zpl_root_getattr);

	static struct dentry *
	zpl_root_lookup(struct inode dip, struct dentry dentry, unsigned int flags)
	{
	cred_t *cr = CRED();
	struct inode *ip;
	int error;

	crhold(cr);
	error = -zfsctl_root_lookup(dip, dname(dentry), &ip, 0, cr, NULL, NULL);
	ASSERT3S(error, <=, 0);
	crfree(cr);

	if (error) {
	if (error == -ENOENT)
	return (d_splice_alias(NULL, dentry));
	else
	return (ERR_PTR(error));
	}

	return (d_splice_alias(ip, dentry));
	}

	/*
	* The '.zfs' control directory file and inode operations.
	*/
	const struct file_operations zpl_fops_root = {
	.open = zpl_common_open,
	.llseek = generic_file_llseek,
	.read = generic_read_dir,
	#ifdef HAVE_VFS_ITERATE_SHARED
	.iterate_shared = zpl_root_iterate,
	#elif defined(HAVE_VFS_ITERATE)
	.iterate = zpl_root_iterate,
	#else
	.readdir = zpl_root_readdir,
	#endif
	};

	const struct inode_operations zpl_ops_root = {
	.lookup = zpl_root_lookup,
	.getattr = zpl_root_getattr,
	};

	static struct vfsmount *
	zpl_snapdir_automount(struct path *path)
	{
	int error;

	error = -zfsctl_snapshot_mount(path, 0);
	if (error)
	return (ERR_PTR(error));

	/*
	* Rather than returning the new vfsmount for the snapshot we must
	* return NULL to indicate a mount collision. This is done because
	* the user space mount calls do_add_mount() which adds the vfsmount
	* to the name space. If we returned the new mount here it would be
	* added again to the vfsmount list resulting in list corruption.
	*/
	return (NULL);
	}

	/*
	* Negative dentries must always be revalidated so newly created snapshots
	* can be detected and automounted. Normal dentries should be kept because
	* as of the 3.18 kernel revaliding the mountpoint dentry will result in
	* the snapshot being immediately unmounted.
	*/
	static int
	#ifdef HAVE_D_REVALIDATE_NAMEIDATA
	zpl_snapdir_revalidate(struct dentry dentry, struct nameidata i)
	#else
	zpl_snapdir_revalidate(struct dentry *dentry, unsigned int flags)
	#endif
	{
	return (!!dentry->d_inode);
	}

	-static const dentry_operations_t zpl_dops_snapdirs = {
	+static dentry_operations_t zpl_dops_snapdirs = {
	/*
	* Auto mounting of snapshots is only supported for 2.6.37 and
	* newer kernels. Prior to this kernel the ops->follow_link()
	* callback was used as a hack to trigger the mount. The
	* resulting vfsmount was then explicitly grafted in to the
	* name space. While it might be possible to add compatibility
	* code to accomplish this it would require considerable care.
	*/
	.d_automount = zpl_snapdir_automount,
	.d_revalidate = zpl_snapdir_revalidate,
	};

	static struct dentry *
	zpl_snapdir_lookup(struct inode dip, struct dentry dentry,
	unsigned int flags)
	{
	fstrans_cookie_t cookie;
	cred_t *cr = CRED();
	struct inode *ip = NULL;
	int error;

	crhold(cr);
	cookie = spl_fstrans_mark();
	error = -zfsctl_snapdir_lookup(dip, dname(dentry), &ip,
	0, cr, NULL, NULL);
	ASSERT3S(error, <=, 0);
	spl_fstrans_unmark(cookie);
	crfree(cr);

	if (error && error != -ENOENT)
	return (ERR_PTR(error));

	ASSERT(error == 0 \|\| ip == NULL);
	d_clear_d_op(dentry);
	d_set_d_op(dentry, &zpl_dops_snapdirs);
	dentry->d_flags \|= DCACHE_NEED_AUTOMOUNT;

	return (d_splice_alias(ip, dentry));
	}

	static int
	zpl_snapdir_iterate(struct file filp, zpl_dir_context_t ctx)
	{
	zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
	fstrans_cookie_t cookie;
	char snapname[MAXNAMELEN];
	boolean_t case_conflict;
	uint64_t id, pos;
	int error = 0;

	if ((error = zpl_enter(zfsvfs, FTAG)) != 0)
	return (error);
	cookie = spl_fstrans_mark();

	if (!zpl_dir_emit_dots(filp, ctx))
	goto out;

	/* Start the position at 0 if it already emitted . and .. */
	pos = (ctx->pos == 2 ? 0 : ctx->pos);
	while (error == 0) {
	dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG);
	error = -dmu_snapshot_list_next(zfsvfs->z_os, MAXNAMELEN,
	snapname, &id, &pos, &case_conflict);
	dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG);
	if (error)
	goto out;

	if (!zpl_dir_emit(ctx, snapname, strlen(snapname),
	ZFSCTL_INO_SHARES - id, DT_DIR))
	goto out;

	ctx->pos = pos;
	}
	out:
	spl_fstrans_unmark(cookie);
	zpl_exit(zfsvfs, FTAG);

	if (error == -ENOENT)
	return (0);

	return (error);
	}

	#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
	static int
	zpl_snapdir_readdir(struct file filp, void dirent, filldir_t filldir)
	{
	zpl_dir_context_t ctx =
	ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
	int error;

	error = zpl_snapdir_iterate(filp, &ctx);
	filp->f_pos = ctx.pos;

	return (error);
	}
	#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */

	static int
	#ifdef HAVE_IOPS_RENAME_USERNS
	zpl_snapdir_rename2(struct user_namespace user_ns, struct inode sdip,
	struct dentry sdentry, struct inode tdip, struct dentry *tdentry,
	unsigned int flags)
	#else
	zpl_snapdir_rename2(struct inode sdip, struct dentry sdentry,
	struct inode tdip, struct dentry tdentry, unsigned int flags)
	#endif
	{
	cred_t *cr = CRED();
	int error;

	/* We probably don't want to support renameat2(2) in ctldir */
	if (flags)
	return (-EINVAL);

	crhold(cr);
	error = -zfsctl_snapdir_rename(sdip, dname(sdentry),
	tdip, dname(tdentry), cr, 0);
	ASSERT3S(error, <=, 0);
	crfree(cr);

	return (error);
	}

	#if !defined(HAVE_RENAME_WANTS_FLAGS) && !defined(HAVE_IOPS_RENAME_USERNS)
	static int
	zpl_snapdir_rename(struct inode sdip, struct dentry sdentry,
	struct inode tdip, struct dentry tdentry)
	{
	return (zpl_snapdir_rename2(sdip, sdentry, tdip, tdentry, 0));
	}
	#endif

	static int
	zpl_snapdir_rmdir(struct inode dip, struct dentry dentry)
	{
	cred_t *cr = CRED();
	int error;

	crhold(cr);
	error = -zfsctl_snapdir_remove(dip, dname(dentry), cr, 0);
	ASSERT3S(error, <=, 0);
	crfree(cr);

	return (error);
	}

	static int
	#ifdef HAVE_IOPS_MKDIR_USERNS
	zpl_snapdir_mkdir(struct user_namespace user_ns, struct inode dip,
	struct dentry *dentry, umode_t mode)
	#else
	zpl_snapdir_mkdir(struct inode dip, struct dentry dentry, umode_t mode)
	#endif
	{
	cred_t *cr = CRED();
	vattr_t *vap;
	struct inode *ip;
	int error;

	crhold(cr);
	vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP);
	zpl_vap_init(vap, dip, mode \| S_IFDIR, cr);

	error = -zfsctl_snapdir_mkdir(dip, dname(dentry), vap, &ip, cr, 0);
	if (error == 0) {
	d_clear_d_op(dentry);
	d_set_d_op(dentry, &zpl_dops_snapdirs);
	d_instantiate(dentry, ip);
	}

	kmem_free(vap, sizeof (vattr_t));
	ASSERT3S(error, <=, 0);
	crfree(cr);

	return (error);
	}

	/*
	* Get snapshot directory attributes.
	*/
	static int
	#ifdef HAVE_USERNS_IOPS_GETATTR
	zpl_snapdir_getattr_impl(struct user_namespace *user_ns,
	const struct path path, struct kstat stat, u32 request_mask,
	unsigned int query_flags)
	#else
	zpl_snapdir_getattr_impl(const struct path path, struct kstat stat,
	u32 request_mask, unsigned int query_flags)
	#endif
	{
	(void) request_mask, (void) query_flags;
	struct inode *ip = path->dentry->d_inode;
	zfsvfs_t *zfsvfs = ITOZSB(ip);
	int error;

	if ((error = zpl_enter(zfsvfs, FTAG)) != 0)
	return (error);
	#ifdef HAVE_USERNS_IOPS_GETATTR
	#ifdef HAVE_GENERIC_FILLATTR_USERNS
	generic_fillattr(user_ns, ip, stat);
	#else
	(void) user_ns;
	#endif
	#else
	generic_fillattr(ip, stat);
	#endif

	stat->nlink = stat->size = 2;

	dsl_dataset_t *ds = dmu_objset_ds(zfsvfs->z_os);
	if (dsl_dataset_phys(ds)->ds_snapnames_zapobj != 0) {
	uint64_t snap_count;
	int err = zap_count(
	dmu_objset_pool(ds->ds_objset)->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj, &snap_count);
	if (err != 0) {
	zpl_exit(zfsvfs, FTAG);
	return (-err);
	}
	stat->nlink += snap_count;
	}

	stat->ctime = stat->mtime = dmu_objset_snap_cmtime(zfsvfs->z_os);
	stat->atime = current_time(ip);
	zpl_exit(zfsvfs, FTAG);

	return (0);
	}
	ZPL_GETATTR_WRAPPER(zpl_snapdir_getattr);

	/*
	* The '.zfs/snapshot' directory file operations. These mainly control
	* generating the list of available snapshots when doing an 'ls' in the
	* directory. See zpl_snapdir_readdir().
	*/
	const struct file_operations zpl_fops_snapdir = {
	.open = zpl_common_open,
	.llseek = generic_file_llseek,
	.read = generic_read_dir,
	#ifdef HAVE_VFS_ITERATE_SHARED
	.iterate_shared = zpl_snapdir_iterate,
	#elif defined(HAVE_VFS_ITERATE)
	.iterate = zpl_snapdir_iterate,
	#else
	.readdir = zpl_snapdir_readdir,
	#endif

	};

	/*
	* The '.zfs/snapshot' directory inode operations. These mainly control
	* creating an inode for a snapshot directory and initializing the needed
	* infrastructure to automount the snapshot. See zpl_snapdir_lookup().
	*/
	const struct inode_operations zpl_ops_snapdir = {
	.lookup = zpl_snapdir_lookup,
	.getattr = zpl_snapdir_getattr,
	#if defined(HAVE_RENAME_WANTS_FLAGS) \|\| defined(HAVE_IOPS_RENAME_USERNS)
	.rename = zpl_snapdir_rename2,
	#else
	.rename = zpl_snapdir_rename,
	#endif
	.rmdir = zpl_snapdir_rmdir,
	.mkdir = zpl_snapdir_mkdir,
	};

	static struct dentry *
	zpl_shares_lookup(struct inode dip, struct dentry dentry,
	unsigned int flags)
	{
	fstrans_cookie_t cookie;
	cred_t *cr = CRED();
	struct inode *ip = NULL;
	int error;

	crhold(cr);
	cookie = spl_fstrans_mark();
	error = -zfsctl_shares_lookup(dip, dname(dentry), &ip,
	0, cr, NULL, NULL);
	ASSERT3S(error, <=, 0);
	spl_fstrans_unmark(cookie);
	crfree(cr);

	if (error) {
	if (error == -ENOENT)
	return (d_splice_alias(NULL, dentry));
	else
	return (ERR_PTR(error));
	}

	return (d_splice_alias(ip, dentry));
	}

	static int
	zpl_shares_iterate(struct file filp, zpl_dir_context_t ctx)
	{
	fstrans_cookie_t cookie;
	cred_t *cr = CRED();
	zfsvfs_t *zfsvfs = ITOZSB(file_inode(filp));
	znode_t *dzp;
	int error = 0;

	if ((error = zpl_enter(zfsvfs, FTAG)) != 0)
	return (error);
	cookie = spl_fstrans_mark();

	if (zfsvfs->z_shares_dir == 0) {
	zpl_dir_emit_dots(filp, ctx);
	goto out;
	}

	error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp);
	if (error)
	goto out;

	crhold(cr);
	error = -zfs_readdir(ZTOI(dzp), ctx, cr);
	crfree(cr);

	iput(ZTOI(dzp));
	out:
	spl_fstrans_unmark(cookie);
	zpl_exit(zfsvfs, FTAG);
	ASSERT3S(error, <=, 0);

	return (error);
	}

	#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
	static int
	zpl_shares_readdir(struct file filp, void dirent, filldir_t filldir)
	{
	zpl_dir_context_t ctx =
	ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
	int error;

	error = zpl_shares_iterate(filp, &ctx);
	filp->f_pos = ctx.pos;

	return (error);
	}
	#endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */

	static int
	#ifdef HAVE_USERNS_IOPS_GETATTR
	zpl_shares_getattr_impl(struct user_namespace *user_ns,
	const struct path path, struct kstat stat, u32 request_mask,
	unsigned int query_flags)
	#else
	zpl_shares_getattr_impl(const struct path path, struct kstat stat,
	u32 request_mask, unsigned int query_flags)
	#endif
	{
	(void) request_mask, (void) query_flags;
	struct inode *ip = path->dentry->d_inode;
	zfsvfs_t *zfsvfs = ITOZSB(ip);
	znode_t *dzp;
	int error;

	if ((error = zpl_enter(zfsvfs, FTAG)) != 0)
	return (error);

	if (zfsvfs->z_shares_dir == 0) {
	#ifdef HAVE_USERNS_IOPS_GETATTR
	#ifdef HAVE_GENERIC_FILLATTR_USERNS
	generic_fillattr(user_ns, path->dentry->d_inode, stat);
	#else
	(void) user_ns;
	#endif
	#else
	generic_fillattr(path->dentry->d_inode, stat);
	#endif
	stat->nlink = stat->size = 2;
	stat->atime = current_time(ip);
	zpl_exit(zfsvfs, FTAG);
	return (0);
	}

	error = -zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp);
	if (error == 0) {
	#ifdef HAVE_USERNS_IOPS_GETATTR
	#ifdef HAVE_GENERIC_FILLATTR_USERNS
	error = -zfs_getattr_fast(user_ns, ZTOI(dzp), stat);
	#else
	(void) user_ns;
	#endif
	#else
	error = -zfs_getattr_fast(kcred->user_ns, ZTOI(dzp), stat);
	#endif
	iput(ZTOI(dzp));
	}

	zpl_exit(zfsvfs, FTAG);
	ASSERT3S(error, <=, 0);

	return (error);
	}
	ZPL_GETATTR_WRAPPER(zpl_shares_getattr);

	/*
	* The '.zfs/shares' directory file operations.
	*/
	const struct file_operations zpl_fops_shares = {
	.open = zpl_common_open,
	.llseek = generic_file_llseek,
	.read = generic_read_dir,
	#ifdef HAVE_VFS_ITERATE_SHARED
	.iterate_shared = zpl_shares_iterate,
	#elif defined(HAVE_VFS_ITERATE)
	.iterate = zpl_shares_iterate,
	#else
	.readdir = zpl_shares_readdir,
	#endif

	};

	/*
	* The '.zfs/shares' directory inode operations.
	*/
	const struct inode_operations zpl_ops_shares = {
	.lookup = zpl_shares_lookup,
	.getattr = zpl_shares_getattr,
	};
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	index 8904dc5bb98a..0d4e0dcd5a3d 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	@@ -1,1614 +1,1614 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	*/

	#include <sys/dataset_kstats.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dir.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/zil_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/zio.h>
	#include <sys/zfs_rlock.h>
	#include <sys/spa_impl.h>
	#include <sys/zvol.h>
	#include <sys/zvol_impl.h>

	#include <linux/blkdev_compat.h>
	#include <linux/task_io_accounting_ops.h>

	#ifdef HAVE_BLK_MQ
	#include <linux/blk-mq.h>
	#endif

	static void zvol_request_impl(zvol_state_t zv, struct bio bio,
	struct request *rq, boolean_t force_sync);

	static unsigned int zvol_major = ZVOL_MAJOR;
	static unsigned int zvol_request_sync = 0;
	static unsigned int zvol_prefetch_bytes = (128 * 1024);
	static unsigned long zvol_max_discard_blocks = 16384;

	#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
	static const unsigned int zvol_open_timeout_ms = 1000;
	#endif

	static unsigned int zvol_threads = 0;
	#ifdef HAVE_BLK_MQ
	static unsigned int zvol_blk_mq_threads = 0;
	static unsigned int zvol_blk_mq_actual_threads;
	static boolean_t zvol_use_blk_mq = B_FALSE;

	/*
	* The maximum number of volblocksize blocks to process per thread. Typically,
	* write heavy workloads preform better with higher values here, and read
	* heavy workloads preform better with lower values, but that's not a hard
	* and fast rule. It's basically a knob to tune between "less overhead with
	* less parallelism" and "more overhead, but more parallelism".
	*
	* '8' was chosen as a reasonable, balanced, default based off of sequential
	* read and write tests to a zvol in an NVMe pool (with 16 CPUs).
	*/
	static unsigned int zvol_blk_mq_blocks_per_thread = 8;
	#endif

	#ifndef BLKDEV_DEFAULT_RQ
	/* BLKDEV_MAX_RQ was renamed to BLKDEV_DEFAULT_RQ in the 5.16 kernel */
	#define BLKDEV_DEFAULT_RQ BLKDEV_MAX_RQ
	#endif

	/*
	* Finalize our BIO or request.
	*/
	#ifdef HAVE_BLK_MQ
	#define END_IO(zv, bio, rq, error) do { \
	if (bio) { \
	BIO_END_IO(bio, error); \
	} else { \
	blk_mq_end_request(rq, errno_to_bi_status(error)); \
	} \
	} while (0)
	#else
	#define END_IO(zv, bio, rq, error) BIO_END_IO(bio, error)
	#endif

	#ifdef HAVE_BLK_MQ
	static unsigned int zvol_blk_mq_queue_depth = BLKDEV_DEFAULT_RQ;
	static unsigned int zvol_actual_blk_mq_queue_depth;
	#endif

	struct zvol_state_os {
	struct gendisk zvo_disk; / generic disk */
	struct request_queue zvo_queue; / request queue */
	dev_t zvo_dev; /* device id */

	#ifdef HAVE_BLK_MQ
	struct blk_mq_tag_set tag_set;
	#endif

	/* Set from the global 'zvol_use_blk_mq' at zvol load */
	boolean_t use_blk_mq;
	};

	taskq_t *zvol_taskq;
	static struct ida zvol_ida;

	typedef struct zv_request_stack {
	zvol_state_t *zv;
	struct bio *bio;
	struct request *rq;
	} zv_request_t;

	typedef struct zv_work {
	struct request *rq;
	struct work_struct work;
	} zv_work_t;

	typedef struct zv_request_task {
	zv_request_t zvr;
	taskq_ent_t ent;
	} zv_request_task_t;

	static zv_request_task_t *
	zv_request_task_create(zv_request_t zvr)
	{
	zv_request_task_t *task;
	task = kmem_alloc(sizeof (zv_request_task_t), KM_SLEEP);
	taskq_init_ent(&task->ent);
	task->zvr = zvr;
	return (task);
	}

	static void
	zv_request_task_free(zv_request_task_t *task)
	{
	kmem_free(task, sizeof (*task));
	}

	#ifdef HAVE_BLK_MQ

	/*
	* This is called when a new block multiqueue request comes in. A request
	* contains one or more BIOs.
	*/
	static blk_status_t zvol_mq_queue_rq(struct blk_mq_hw_ctx *hctx,
	const struct blk_mq_queue_data *bd)
	{
	struct request *rq = bd->rq;
	zvol_state_t *zv = rq->q->queuedata;

	/* Tell the kernel that we are starting to process this request */
	blk_mq_start_request(rq);

	if (blk_rq_is_passthrough(rq)) {
	/* Skip non filesystem request */
	blk_mq_end_request(rq, BLK_STS_IOERR);
	return (BLK_STS_IOERR);
	}

	zvol_request_impl(zv, NULL, rq, 0);

	/* Acknowledge to the kernel that we got this request */
	return (BLK_STS_OK);
	}

	static struct blk_mq_ops zvol_blk_mq_queue_ops = {
	.queue_rq = zvol_mq_queue_rq,
	};

	/* Initialize our blk-mq struct */
	static int zvol_blk_mq_alloc_tag_set(zvol_state_t *zv)
	{
	struct zvol_state_os *zso = zv->zv_zso;

	memset(&zso->tag_set, 0, sizeof (zso->tag_set));

	/* Initialize tag set. */
	zso->tag_set.ops = &zvol_blk_mq_queue_ops;
	zso->tag_set.nr_hw_queues = zvol_blk_mq_actual_threads;
	zso->tag_set.queue_depth = zvol_actual_blk_mq_queue_depth;
	zso->tag_set.numa_node = NUMA_NO_NODE;
	zso->tag_set.cmd_size = 0;

	/*
	* We need BLK_MQ_F_BLOCKING here since we do blocking calls in
	* zvol_request_impl()
	*/
	zso->tag_set.flags = BLK_MQ_F_SHOULD_MERGE \| BLK_MQ_F_BLOCKING;
	zso->tag_set.driver_data = zv;

	return (blk_mq_alloc_tag_set(&zso->tag_set));
	}
	#endif /* HAVE_BLK_MQ */

	/*
	* Given a path, return TRUE if path is a ZVOL.
	*/
	boolean_t
	zvol_os_is_zvol(const char *path)
	{
	dev_t dev = 0;

	if (vdev_lookup_bdev(path, &dev) != 0)
	return (B_FALSE);

	if (MAJOR(dev) == zvol_major)
	return (B_TRUE);

	return (B_FALSE);
	}

	static void
	zvol_write(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	int error = 0;
	zfs_uio_t uio;
	zvol_state_t *zv = zvr->zv;
	struct request_queue *q;
	struct gendisk *disk;
	unsigned long start_time = 0;
	boolean_t acct = B_FALSE;

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);
	ASSERT3P(zv->zv_zilog, !=, NULL);

	q = zv->zv_zso->zvo_queue;
	disk = zv->zv_zso->zvo_disk;

	/* bio marked as FLUSH need to flush before write */
	if (io_is_flush(bio, rq))
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	/* Some requests are just for flush and nothing else. */
	if (io_size(bio, rq) == 0) {
	rw_exit(&zv->zv_suspend_lock);
	END_IO(zv, bio, rq, 0);
	return;
	}

	zfs_uio_bvec_init(&uio, bio, rq);

	ssize_t start_resid = uio.uio_resid;

	/*
	* With use_blk_mq, accounting is done by blk_mq_start_request()
	* and blk_mq_end_request(), so we can skip it here.
	*/
	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct) {
	start_time = blk_generic_start_io_acct(q, disk, WRITE,
	bio);
	}
	}

	boolean_t sync =
	io_is_fua(bio, rq) \|\| zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	uio.uio_loffset, uio.uio_resid, RL_WRITER);

	uint64_t volsize = zv->zv_volsize;
	while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
	uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);
	uint64_t off = uio.uio_loffset;
	dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);

	if (bytes > volsize - off) /* don't write past the end */
	bytes = volsize - off;

	dmu_tx_hold_write_by_dnode(tx, zv->zv_dn, off, bytes);

	/* This will only fail for ENOSPC */
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	break;
	}
	error = dmu_write_uio_dnode(zv->zv_dn, &uio, bytes, tx);
	if (error == 0) {
	zvol_log_write(zv, tx, off, bytes, sync);
	}
	dmu_tx_commit(tx);

	if (error)
	break;
	}
	zfs_rangelock_exit(lr);

	int64_t nwritten = start_resid - uio.uio_resid;
	dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten);
	task_io_account_write(nwritten);

	if (sync)
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, WRITE, bio, start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_write_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_write(&task->zvr);
	zv_request_task_free(task);
	}

	static void
	zvol_discard(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	zvol_state_t *zv = zvr->zv;
	uint64_t start = io_offset(bio, rq);
	uint64_t size = io_size(bio, rq);
	uint64_t end = start + size;
	boolean_t sync;
	int error = 0;
	dmu_tx_t *tx;
	struct request_queue *q = zv->zv_zso->zvo_queue;
	struct gendisk *disk = zv->zv_zso->zvo_disk;
	unsigned long start_time = 0;

	boolean_t acct = blk_queue_io_stat(q);

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);
	ASSERT3P(zv->zv_zilog, !=, NULL);

	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct) {
	start_time = blk_generic_start_io_acct(q, disk, WRITE,
	bio);
	}
	}

	sync = io_is_fua(bio, rq) \|\| zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;

	if (end > zv->zv_volsize) {
	error = SET_ERROR(EIO);
	goto unlock;
	}

	/*
	* Align the request to volume block boundaries when a secure erase is
	* not required. This will prevent dnode_free_range() from zeroing out
	* the unaligned parts which is slow (read-modify-write) and useless
	* since we are not freeing any space by doing so.
	*/
	if (!io_is_secure_erase(bio, rq)) {
	start = P2ROUNDUP(start, zv->zv_volblocksize);
	end = P2ALIGN(end, zv->zv_volblocksize);
	size = end - start;
	}

	if (start >= end)
	goto unlock;

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	start, size, RL_WRITER);

	tx = dmu_tx_create(zv->zv_objset);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	} else {
	zvol_log_truncate(zv, tx, start, size, B_TRUE);
	dmu_tx_commit(tx);
	error = dmu_free_long_range(zv->zv_objset,
	ZVOL_OBJ, start, size);
	}
	zfs_rangelock_exit(lr);

	if (error == 0 && sync)
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	unlock:
	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, WRITE, bio,
	start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_discard_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_discard(&task->zvr);
	zv_request_task_free(task);
	}

	static void
	zvol_read(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	int error = 0;
	zfs_uio_t uio;
	boolean_t acct = B_FALSE;
	zvol_state_t *zv = zvr->zv;
	struct request_queue *q;
	struct gendisk *disk;
	unsigned long start_time = 0;

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);

	zfs_uio_bvec_init(&uio, bio, rq);

	q = zv->zv_zso->zvo_queue;
	disk = zv->zv_zso->zvo_disk;

	ssize_t start_resid = uio.uio_resid;

	/*
	* When blk-mq is being used, accounting is done by
	* blk_mq_start_request() and blk_mq_end_request().
	*/
	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct)
	start_time = blk_generic_start_io_acct(q, disk, READ,
	bio);
	}

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	uio.uio_loffset, uio.uio_resid, RL_READER);

	uint64_t volsize = zv->zv_volsize;

	while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
	uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);

	/* don't read past the end */
	if (bytes > volsize - uio.uio_loffset)
	bytes = volsize - uio.uio_loffset;

	error = dmu_read_uio_dnode(zv->zv_dn, &uio, bytes);
	if (error) {
	/* convert checksum errors into IO errors */
	if (error == ECKSUM)
	error = SET_ERROR(EIO);
	break;
	}
	}
	zfs_rangelock_exit(lr);

	int64_t nread = start_resid - uio.uio_resid;
	dataset_kstats_update_read_kstats(&zv->zv_kstat, nread);
	task_io_account_read(nread);

	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, READ, bio, start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_read_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_read(&task->zvr);
	zv_request_task_free(task);
	}


	/*
	* Process a BIO or request
	*
	* Either 'bio' or 'rq' should be set depending on if we are processing a
	* bio or a request (both should not be set).
	*
	* force_sync: Set to 0 to defer processing to a background taskq
	* Set to 1 to process data synchronously
	*/
	static void
	zvol_request_impl(zvol_state_t zv, struct bio bio, struct request *rq,
	boolean_t force_sync)
	{
	fstrans_cookie_t cookie = spl_fstrans_mark();
	uint64_t offset = io_offset(bio, rq);
	uint64_t size = io_size(bio, rq);
	int rw = io_data_dir(bio, rq);

	if (zvol_request_sync)
	force_sync = 1;

	zv_request_t zvr = {
	.zv = zv,
	.bio = bio,
	.rq = rq,
	};

	if (io_has_data(bio, rq) && offset + size > zv->zv_volsize) {
	printk(KERN_INFO "%s: bad access: offset=%llu, size=%lu\n",
	zv->zv_zso->zvo_disk->disk_name,
	(long long unsigned)offset,
	(long unsigned)size);

	END_IO(zv, bio, rq, -SET_ERROR(EIO));
	goto out;
	}

	zv_request_task_t *task;

	if (rw == WRITE) {
	if (unlikely(zv->zv_flags & ZVOL_RDONLY)) {
	END_IO(zv, bio, rq, -SET_ERROR(EROFS));
	goto out;
	}

	/*
	* Prevents the zvol from being suspended, or the ZIL being
	* concurrently opened. Will be released after the i/o
	* completes.
	*/
	rw_enter(&zv->zv_suspend_lock, RW_READER);

	/*
	* Open a ZIL if this is the first time we have written to this
	* zvol. We protect zv->zv_zilog with zv_suspend_lock rather
	* than zv_state_lock so that we don't need to acquire an
	* additional lock in this path.
	*/
	if (zv->zv_zilog == NULL) {
	rw_exit(&zv->zv_suspend_lock);
	rw_enter(&zv->zv_suspend_lock, RW_WRITER);
	if (zv->zv_zilog == NULL) {
	zv->zv_zilog = zil_open(zv->zv_objset,
	zvol_get_data, &zv->zv_kstat.dk_zil_sums);
	zv->zv_flags \|= ZVOL_WRITTEN_TO;
	/* replay / destroy done in zvol_create_minor */
	VERIFY0((zv->zv_zilog->zl_header->zh_flags &
	ZIL_REPLAY_NEEDED));
	}
	rw_downgrade(&zv->zv_suspend_lock);
	}

	/*
	* We don't want this thread to be blocked waiting for i/o to
	* complete, so we instead wait from a taskq callback. The
	* i/o may be a ZIL write (via zil_commit()), or a read of an
	* indirect block, or a read of a data block (if this is a
	* partial-block write). We will indicate that the i/o is
	* complete by calling END_IO() from the taskq callback.
	*
	* This design allows the calling thread to continue and
	* initiate more concurrent operations by calling
	* zvol_request() again. There are typically only a small
	* number of threads available to call zvol_request() (e.g.
	* one per iSCSI target), so keeping the latency of
	* zvol_request() low is important for performance.
	*
	* The zvol_request_sync module parameter allows this
	* behavior to be altered, for performance evaluation
	* purposes. If the callback blocks, setting
	* zvol_request_sync=1 will result in much worse performance.
	*
	* We can have up to zvol_threads concurrent i/o's being
	* processed for all zvols on the system. This is typically
	* a vast improvement over the zvol_request_sync=1 behavior
	* of one i/o at a time per zvol. However, an even better
	* design would be for zvol_request() to initiate the zio
	* directly, and then be notified by the zio_done callback,
	* which would call END_IO(). Unfortunately, the DMU/ZIL
	* interfaces lack this functionality (they block waiting for
	* the i/o to complete).
	*/
	if (io_is_discard(bio, rq) \|\| io_is_secure_erase(bio, rq)) {
	if (force_sync) {
	zvol_discard(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(zvol_taskq,
	zvol_discard_task, task, 0, &task->ent);
	}
	} else {
	if (force_sync) {
	zvol_write(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(zvol_taskq,
	zvol_write_task, task, 0, &task->ent);
	}
	}
	} else {
	/*
	* The SCST driver, and possibly others, may issue READ I/Os
	* with a length of zero bytes. These empty I/Os contain no
	* data and require no additional handling.
	*/
	if (size == 0) {
	END_IO(zv, bio, rq, 0);
	goto out;
	}

	rw_enter(&zv->zv_suspend_lock, RW_READER);

	/* See comment in WRITE case above. */
	if (force_sync) {
	zvol_read(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(zvol_taskq,
	zvol_read_task, task, 0, &task->ent);
	}
	}

	out:
	spl_fstrans_unmark(cookie);
	}

	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	#ifdef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID
	static void
	zvol_submit_bio(struct bio *bio)
	#else
	static blk_qc_t
	zvol_submit_bio(struct bio *bio)
	#endif
	#else
	static MAKE_REQUEST_FN_RET
	zvol_request(struct request_queue q, struct bio bio)
	#endif
	{
	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif
	#endif
	zvol_state_t *zv = q->queuedata;

	zvol_request_impl(zv, bio, NULL, 0);
	#if defined(HAVE_MAKE_REQUEST_FN_RET_QC) \|\| \
	defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
	!defined(HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID)
	return (BLK_QC_T_NONE);
	#endif
	}

	static int
	zvol_open(struct block_device *bdev, fmode_t flag)
	{
	zvol_state_t *zv;
	int error = 0;
	boolean_t drop_suspend = B_FALSE;
	#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
	hrtime_t timeout = MSEC2NSEC(zvol_open_timeout_ms);
	hrtime_t start = gethrtime();

	retry:
	#endif
	rw_enter(&zvol_state_lock, RW_READER);
	/*
	* Obtain a copy of private_data under the zvol_state_lock to make
	* sure that either the result of zvol free code path setting
	* bdev->bd_disk->private_data to NULL is observed, or zvol_os_free()
	* is not called on this zv because of the positive zv_open_count.
	*/
	zv = bdev->bd_disk->private_data;
	if (zv == NULL) {
	rw_exit(&zvol_state_lock);
	return (SET_ERROR(-ENXIO));
	}

	mutex_enter(&zv->zv_state_lock);
	/*
	* Make sure zvol is not suspended during first open
	* (hold zv_suspend_lock) and respect proper lock acquisition
	* ordering - zv_suspend_lock before zv_state_lock
	*/
	if (zv->zv_open_count == 0) {
	if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
	mutex_exit(&zv->zv_state_lock);
	rw_enter(&zv->zv_suspend_lock, RW_READER);
	mutex_enter(&zv->zv_state_lock);
	/* check to see if zv_suspend_lock is needed */
	if (zv->zv_open_count != 0) {
	rw_exit(&zv->zv_suspend_lock);
	} else {
	drop_suspend = B_TRUE;
	}
	} else {
	drop_suspend = B_TRUE;
	}
	}
	rw_exit(&zvol_state_lock);

	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	if (zv->zv_open_count == 0) {
	boolean_t drop_namespace = B_FALSE;

	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));

	/*
	* In all other call paths the spa_namespace_lock is taken
	* before the bdev->bd_mutex lock. However, on open(2)
	* the __blkdev_get() function calls fops->open() with the
	* bdev->bd_mutex lock held. This can result in a deadlock
	* when zvols from one pool are used as vdevs in another.
	*
	* To prevent a lock inversion deadlock we preemptively
	* take the spa_namespace_lock. Normally the lock will not
	* be contended and this is safe because spa_open_common()
	* handles the case where the caller already holds the
	* spa_namespace_lock.
	*
	* When the lock cannot be aquired after multiple retries
	* this must be the vdev on zvol deadlock case and we have
	* no choice but to return an error. For 5.12 and older
	* kernels returning -ERESTARTSYS will result in the
	* bdev->bd_mutex being dropped, then reacquired, and
	* fops->open() being called again. This process can be
	* repeated safely until both locks are acquired. For 5.13
	* and newer the -ERESTARTSYS retry logic was removed from
	* the kernel so the only option is to return the error for
	* the caller to handle it.
	*/
	if (!mutex_owned(&spa_namespace_lock)) {
	if (!mutex_tryenter(&spa_namespace_lock)) {
	mutex_exit(&zv->zv_state_lock);
	rw_exit(&zv->zv_suspend_lock);

	#ifdef HAVE_BLKDEV_GET_ERESTARTSYS
	schedule();
	return (SET_ERROR(-ERESTARTSYS));
	#else
	if ((gethrtime() - start) > timeout)
	return (SET_ERROR(-ERESTARTSYS));

	schedule_timeout(MSEC_TO_TICK(10));
	goto retry;
	#endif
	} else {
	drop_namespace = B_TRUE;
	}
	}

	error = -zvol_first_open(zv, !(flag & FMODE_WRITE));

	if (drop_namespace)
	mutex_exit(&spa_namespace_lock);
	}

	if (error == 0) {
	if ((flag & FMODE_WRITE) && (zv->zv_flags & ZVOL_RDONLY)) {
	if (zv->zv_open_count == 0)
	zvol_last_close(zv);

	error = SET_ERROR(-EROFS);
	} else {
	zv->zv_open_count++;
	}
	}

	mutex_exit(&zv->zv_state_lock);
	if (drop_suspend)
	rw_exit(&zv->zv_suspend_lock);

	if (error == 0)
	zfs_check_media_change(bdev);

	return (error);
	}

	static void
	zvol_release(struct gendisk *disk, fmode_t mode)
	{
	zvol_state_t *zv;
	boolean_t drop_suspend = B_TRUE;

	rw_enter(&zvol_state_lock, RW_READER);
	zv = disk->private_data;

	mutex_enter(&zv->zv_state_lock);
	ASSERT3U(zv->zv_open_count, >, 0);
	/*
	* make sure zvol is not suspended during last close
	* (hold zv_suspend_lock) and respect proper lock acquisition
	* ordering - zv_suspend_lock before zv_state_lock
	*/
	if (zv->zv_open_count == 1) {
	if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
	mutex_exit(&zv->zv_state_lock);
	rw_enter(&zv->zv_suspend_lock, RW_READER);
	mutex_enter(&zv->zv_state_lock);
	/* check to see if zv_suspend_lock is needed */
	if (zv->zv_open_count != 1) {
	rw_exit(&zv->zv_suspend_lock);
	drop_suspend = B_FALSE;
	}
	}
	} else {
	drop_suspend = B_FALSE;
	}
	rw_exit(&zvol_state_lock);

	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	zv->zv_open_count--;
	if (zv->zv_open_count == 0) {
	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
	zvol_last_close(zv);
	}

	mutex_exit(&zv->zv_state_lock);

	if (drop_suspend)
	rw_exit(&zv->zv_suspend_lock);
	}

	static int
	zvol_ioctl(struct block_device *bdev, fmode_t mode,
	unsigned int cmd, unsigned long arg)
	{
	zvol_state_t *zv = bdev->bd_disk->private_data;
	int error = 0;

	ASSERT3U(zv->zv_open_count, >, 0);

	switch (cmd) {
	case BLKFLSBUF:
	fsync_bdev(bdev);
	invalidate_bdev(bdev);
	rw_enter(&zv->zv_suspend_lock, RW_READER);

	if (!(zv->zv_flags & ZVOL_RDONLY))
	txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);

	rw_exit(&zv->zv_suspend_lock);
	break;

	case BLKZNAME:
	mutex_enter(&zv->zv_state_lock);
	error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN);
	mutex_exit(&zv->zv_state_lock);
	break;

	default:
	error = -ENOTTY;
	break;
	}

	return (SET_ERROR(error));
	}

	#ifdef CONFIG_COMPAT
	static int
	zvol_compat_ioctl(struct block_device *bdev, fmode_t mode,
	unsigned cmd, unsigned long arg)
	{
	return (zvol_ioctl(bdev, mode, cmd, arg));
	}
	#else
	#define zvol_compat_ioctl NULL
	#endif

	static unsigned int
	zvol_check_events(struct gendisk *disk, unsigned int clearing)
	{
	unsigned int mask = 0;

	rw_enter(&zvol_state_lock, RW_READER);

	zvol_state_t *zv = disk->private_data;
	if (zv != NULL) {
	mutex_enter(&zv->zv_state_lock);
	mask = zv->zv_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
	zv->zv_changed = 0;
	mutex_exit(&zv->zv_state_lock);
	}

	rw_exit(&zvol_state_lock);

	return (mask);
	}

	static int
	zvol_revalidate_disk(struct gendisk *disk)
	{
	rw_enter(&zvol_state_lock, RW_READER);

	zvol_state_t *zv = disk->private_data;
	if (zv != NULL) {
	mutex_enter(&zv->zv_state_lock);
	set_capacity(zv->zv_zso->zvo_disk,
	zv->zv_volsize >> SECTOR_BITS);
	mutex_exit(&zv->zv_state_lock);
	}

	rw_exit(&zvol_state_lock);

	return (0);
	}

	int
	zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize)
	{
	struct gendisk *disk = zv->zv_zso->zvo_disk;

	#if defined(HAVE_REVALIDATE_DISK_SIZE)
	revalidate_disk_size(disk, zvol_revalidate_disk(disk) == 0);
	#elif defined(HAVE_REVALIDATE_DISK)
	revalidate_disk(disk);
	#else
	zvol_revalidate_disk(disk);
	#endif
	return (0);
	}

	void
	zvol_os_clear_private(zvol_state_t *zv)
	{
	/*
	* Cleared while holding zvol_state_lock as a writer
	* which will prevent zvol_open() from opening it.
	*/
	zv->zv_zso->zvo_disk->private_data = NULL;
	}

	/*
	* Provide a simple virtual geometry for legacy compatibility. For devices
	* smaller than 1 MiB a small head and sector count is used to allow very
	* tiny devices. For devices over 1 Mib a standard head and sector count
	* is used to keep the cylinders count reasonable.
	*/
	static int
	zvol_getgeo(struct block_device bdev, struct hd_geometry geo)
	{
	zvol_state_t *zv = bdev->bd_disk->private_data;
	sector_t sectors;

	ASSERT3U(zv->zv_open_count, >, 0);

	sectors = get_capacity(zv->zv_zso->zvo_disk);

	if (sectors > 2048) {
	geo->heads = 16;
	geo->sectors = 63;
	} else {
	geo->heads = 2;
	geo->sectors = 4;
	}

	geo->start = 0;
	geo->cylinders = sectors / (geo->heads * geo->sectors);

	return (0);
	}

	/*
	* Why have two separate block_device_operations structs?
	*
	* Normally we'd just have one, and assign 'submit_bio' as needed. However,
	* it's possible the user's kernel is built with CONSTIFY_PLUGIN, meaning we
	* can't just change submit_bio dynamically at runtime. So just create two
	* separate structs to get around this.
	*/
	static const struct block_device_operations zvol_ops_blk_mq = {
	.open = zvol_open,
	.release = zvol_release,
	.ioctl = zvol_ioctl,
	.compat_ioctl = zvol_compat_ioctl,
	.check_events = zvol_check_events,
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	.revalidate_disk = zvol_revalidate_disk,
	#endif
	.getgeo = zvol_getgeo,
	.owner = THIS_MODULE,
	};

	static const struct block_device_operations zvol_ops = {
	.open = zvol_open,
	.release = zvol_release,
	.ioctl = zvol_ioctl,
	.compat_ioctl = zvol_compat_ioctl,
	.check_events = zvol_check_events,
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	.revalidate_disk = zvol_revalidate_disk,
	#endif
	.getgeo = zvol_getgeo,
	.owner = THIS_MODULE,
	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	.submit_bio = zvol_submit_bio,
	#endif
	};

	static int
	zvol_alloc_non_blk_mq(struct zvol_state_os *zso)
	{
	#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS)
	#if defined(HAVE_BLK_ALLOC_DISK)
	zso->zvo_disk = blk_alloc_disk(NUMA_NO_NODE);
	if (zso->zvo_disk == NULL)
	return (1);

	zso->zvo_disk->minors = ZVOL_MINORS;
	zso->zvo_queue = zso->zvo_disk->queue;
	#else
	zso->zvo_queue = blk_alloc_queue(NUMA_NO_NODE);
	if (zso->zvo_queue == NULL)
	return (1);

	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	return (1);
	}

	zso->zvo_disk->queue = zso->zvo_queue;
	#endif /* HAVE_BLK_ALLOC_DISK */
	#else
	zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
	if (zso->zvo_queue == NULL)
	return (1);

	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	return (1);
	}

	zso->zvo_disk->queue = zso->zvo_queue;
	#endif /* HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
	return (0);

	}

	static int
	zvol_alloc_blk_mq(zvol_state_t *zv)
	{
	#ifdef HAVE_BLK_MQ
	struct zvol_state_os *zso = zv->zv_zso;

	/* Allocate our blk-mq tag_set */
	if (zvol_blk_mq_alloc_tag_set(zv) != 0)
	return (1);

	#if defined(HAVE_BLK_ALLOC_DISK)
	zso->zvo_disk = blk_mq_alloc_disk(&zso->tag_set, zv);
	if (zso->zvo_disk == NULL) {
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}
	zso->zvo_queue = zso->zvo_disk->queue;
	zso->zvo_disk->minors = ZVOL_MINORS;
	#else
	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}
	/* Allocate queue */
	zso->zvo_queue = blk_mq_init_queue(&zso->tag_set);
	if (IS_ERR(zso->zvo_queue)) {
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}

	/* Our queue is now created, assign it to our disk */
	zso->zvo_disk->queue = zso->zvo_queue;

	#endif
	#endif
	return (0);
	}

	/*
	* Allocate memory for a new zvol_state_t and setup the required
	* request queue and generic disk structures for the block device.
	*/
	static zvol_state_t *
	zvol_alloc(dev_t dev, const char *name)
	{
	zvol_state_t *zv;
	struct zvol_state_os *zso;
	uint64_t volmode;
	int ret;

	if (dsl_prop_get_integer(name, "volmode", &volmode, NULL) != 0)
	return (NULL);

	if (volmode == ZFS_VOLMODE_DEFAULT)
	volmode = zvol_volmode;

	if (volmode == ZFS_VOLMODE_NONE)
	return (NULL);

	zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP);
	zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP);
	zv->zv_zso = zso;
	zv->zv_volmode = volmode;

	list_link_init(&zv->zv_next);
	mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL);

	#ifdef HAVE_BLK_MQ
	zv->zv_zso->use_blk_mq = zvol_use_blk_mq;
	#endif

	/*
	* The block layer has 3 interfaces for getting BIOs:
	*
	* 1. blk-mq request queues (new)
	* 2. submit_bio() (oldest)
	* 3. regular request queues (old).
	*
	* Each of those interfaces has two permutations:
	*
	* a) We have blk_alloc_disk()/blk_mq_alloc_disk(), which allocates
	* both the disk and its queue (5.14 kernel or newer)
	*
	* b) We don't have blk_*alloc_disk(), and have to allocate the
	* disk and the queue separately. (5.13 kernel or older)
	*/
	if (zv->zv_zso->use_blk_mq) {
	ret = zvol_alloc_blk_mq(zv);
	zso->zvo_disk->fops = &zvol_ops_blk_mq;
	} else {
	ret = zvol_alloc_non_blk_mq(zso);
	zso->zvo_disk->fops = &zvol_ops;
	}
	if (ret != 0)
	goto out_kmem;

	blk_queue_set_write_cache(zso->zvo_queue, B_TRUE, B_TRUE);

	/* Limit read-ahead to a single page to prevent over-prefetching. */
	blk_queue_set_read_ahead(zso->zvo_queue, 1);

	if (!zv->zv_zso->use_blk_mq) {
	/* Disable write merging in favor of the ZIO pipeline. */
	blk_queue_flag_set(QUEUE_FLAG_NOMERGES, zso->zvo_queue);
	}

	/* Enable /proc/diskstats */
	blk_queue_flag_set(QUEUE_FLAG_IO_STAT, zso->zvo_queue);

	zso->zvo_queue->queuedata = zv;
	zso->zvo_dev = dev;
	zv->zv_open_count = 0;
	strlcpy(zv->zv_name, name, MAXNAMELEN);

	zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL);
	rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL);

	zso->zvo_disk->major = zvol_major;
	zso->zvo_disk->events = DISK_EVENT_MEDIA_CHANGE;

	/*
	* Setting ZFS_VOLMODE_DEV disables partitioning on ZVOL devices.
	* This is accomplished by limiting the number of minors for the
	* device to one and explicitly disabling partition scanning.
	*/
	if (volmode == ZFS_VOLMODE_DEV) {
	zso->zvo_disk->minors = 1;
	zso->zvo_disk->flags &= ~ZFS_GENHD_FL_EXT_DEVT;
	zso->zvo_disk->flags \|= ZFS_GENHD_FL_NO_PART;
	}

	zso->zvo_disk->first_minor = (dev & MINORMASK);
	zso->zvo_disk->private_data = zv;
	snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d",
	ZVOL_DEV_NAME, (dev & MINORMASK));

	return (zv);

	out_kmem:
	kmem_free(zso, sizeof (struct zvol_state_os));
	kmem_free(zv, sizeof (zvol_state_t));
	return (NULL);
	}

	/*
	* Cleanup then free a zvol_state_t which was created by zvol_alloc().
	* At this time, the structure is not opened by anyone, is taken off
	* the zvol_state_list, and has its private data set to NULL.
	* The zvol_state_lock is dropped.
	*
	* This function may take many milliseconds to complete (e.g. we've seen
	* it take over 256ms), due to the calls to "blk_cleanup_queue" and
	* "del_gendisk". Thus, consumers need to be careful to account for this
	* latency when calling this function.
	*/
	void
	zvol_os_free(zvol_state_t *zv)
	{

	ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
	ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
	ASSERT0(zv->zv_open_count);
	ASSERT3P(zv->zv_zso->zvo_disk->private_data, ==, NULL);

	rw_destroy(&zv->zv_suspend_lock);
	zfs_rangelock_fini(&zv->zv_rangelock);

	del_gendisk(zv->zv_zso->zvo_disk);
	#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
	defined(HAVE_BLK_ALLOC_DISK)
	#if defined(HAVE_BLK_CLEANUP_DISK)
	blk_cleanup_disk(zv->zv_zso->zvo_disk);
	#else
	put_disk(zv->zv_zso->zvo_disk);
	#endif
	#else
	blk_cleanup_queue(zv->zv_zso->zvo_queue);
	put_disk(zv->zv_zso->zvo_disk);
	#endif

	#ifdef HAVE_BLK_MQ
	if (zv->zv_zso->use_blk_mq)
	blk_mq_free_tag_set(&zv->zv_zso->tag_set);
	#endif

	ida_simple_remove(&zvol_ida,
	MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);

	mutex_destroy(&zv->zv_state_lock);
	dataset_kstats_destroy(&zv->zv_kstat);

	kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
	kmem_free(zv, sizeof (zvol_state_t));
	}

	void
	zvol_wait_close(zvol_state_t *zv)
	{
	}

	/*
	* Create a block device minor node and setup the linkage between it
	* and the specified volume. Once this function returns the block
	* device is live and ready for use.
	*/
	int
	zvol_os_create_minor(const char *name)
	{
	zvol_state_t *zv;
	objset_t *os;
	dmu_object_info_t *doi;
	uint64_t volsize;
	uint64_t len;
	unsigned minor = 0;
	int error = 0;
	int idx;
	uint64_t hash = zvol_name_hash(name);

	if (zvol_inhibit_dev)
	return (0);

	idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP));
	if (idx < 0)
	return (SET_ERROR(-idx));
	minor = idx << ZVOL_MINOR_BITS;

	zv = zvol_find_by_name_hash(name, hash, RW_NONE);
	if (zv) {
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));
	mutex_exit(&zv->zv_state_lock);
	ida_simple_remove(&zvol_ida, idx);
	return (SET_ERROR(EEXIST));
	}

	doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);

	error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, B_TRUE, FTAG, &os);
	if (error)
	goto out_doi;

	error = dmu_object_info(os, ZVOL_OBJ, doi);
	if (error)
	goto out_dmu_objset_disown;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
	goto out_dmu_objset_disown;

	zv = zvol_alloc(MKDEV(zvol_major, minor), name);
	if (zv == NULL) {
	error = SET_ERROR(EAGAIN);
	goto out_dmu_objset_disown;
	}
	zv->zv_hash = hash;

	if (dmu_objset_is_snapshot(os))
	zv->zv_flags \|= ZVOL_RDONLY;

	zv->zv_volblocksize = doi->doi_data_block_size;
	zv->zv_volsize = volsize;
	zv->zv_objset = os;

	set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> 9);

	blk_queue_max_hw_sectors(zv->zv_zso->zvo_queue,
	(DMU_MAX_ACCESS / 4) >> 9);

	if (zv->zv_zso->use_blk_mq) {
	/*
	* IO requests can be really big (1MB). When an IO request
	* comes in, it is passed off to zvol_read() or zvol_write()
	* in a new thread, where it is chunked up into 'volblocksize'
	* sized pieces and processed. So for example, if the request
	* is a 1MB write and your volblocksize is 128k, one zvol_write
	* thread will take that request and sequentially do ten 128k
	* IOs. This is due to the fact that the thread needs to lock
	* each volblocksize sized block. So you might be wondering:
	* "instead of passing the whole 1MB request to one thread,
	* why not pass ten individual 128k chunks to ten threads and
	* process the whole write in parallel?" The short answer is
	* that there's a sweet spot number of chunks that balances
	* the greater parallelism with the added overhead of more
	* threads. The sweet spot can be different depending on if you
	* have a read or write heavy workload. Writes typically want
	* high chunk counts while reads typically want lower ones. On
	* a test pool with 6 NVMe drives in a 3x 2-disk mirror
	* configuration, with volblocksize=8k, the sweet spot for good
	* sequential reads and writes was at 8 chunks.
	*/

	/*
	* Below we tell the kernel how big we want our requests
	* to be. You would think that blk_queue_io_opt() would be
	* used to do this since it is used to "set optimal request
	* size for the queue", but that doesn't seem to do
	* anything - the kernel still gives you huge requests
	* with tons of little PAGE_SIZE segments contained within it.
	*
	* Knowing that the kernel will just give you PAGE_SIZE segments
	* no matter what, you can say "ok, I want PAGE_SIZE byte
	* segments, and I want 'N' of them per request", where N is
	* the correct number of segments for the volblocksize and
	* number of chunks you want.
	*/
	#ifdef HAVE_BLK_MQ
	if (zvol_blk_mq_blocks_per_thread != 0) {
	unsigned int chunks;
	chunks = MIN(zvol_blk_mq_blocks_per_thread, UINT16_MAX);

	blk_queue_max_segment_size(zv->zv_zso->zvo_queue,
	PAGE_SIZE);
	blk_queue_max_segments(zv->zv_zso->zvo_queue,
	(zv->zv_volblocksize * chunks) / PAGE_SIZE);
	} else {
	/*
	* Special case: zvol_blk_mq_blocks_per_thread = 0
	* Max everything out.
	*/
	blk_queue_max_segments(zv->zv_zso->zvo_queue,
	UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_zso->zvo_queue,
	UINT_MAX);
	}
	#endif
	} else {
	blk_queue_max_segments(zv->zv_zso->zvo_queue, UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_zso->zvo_queue, UINT_MAX);
	}

	blk_queue_physical_block_size(zv->zv_zso->zvo_queue,
	zv->zv_volblocksize);
	blk_queue_io_opt(zv->zv_zso->zvo_queue, zv->zv_volblocksize);
	blk_queue_max_discard_sectors(zv->zv_zso->zvo_queue,
	(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
	blk_queue_discard_granularity(zv->zv_zso->zvo_queue,
	zv->zv_volblocksize);
	#ifdef QUEUE_FLAG_DISCARD
	blk_queue_flag_set(QUEUE_FLAG_DISCARD, zv->zv_zso->zvo_queue);
	#endif
	#ifdef QUEUE_FLAG_NONROT
	blk_queue_flag_set(QUEUE_FLAG_NONROT, zv->zv_zso->zvo_queue);
	#endif
	#ifdef QUEUE_FLAG_ADD_RANDOM
	blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, zv->zv_zso->zvo_queue);
	#endif
	/* This flag was introduced in kernel version 4.12. */
	#ifdef QUEUE_FLAG_SCSI_PASSTHROUGH
	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, zv->zv_zso->zvo_queue);
	#endif

	ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL);
	error = dataset_kstats_create(&zv->zv_kstat, zv->zv_objset);
	if (error)
	goto out_dmu_objset_disown;
	ASSERT3P(zv->zv_zilog, ==, NULL);
	zv->zv_zilog = zil_open(os, zvol_get_data, &zv->zv_kstat.dk_zil_sums);
	if (spa_writeable(dmu_objset_spa(os))) {
	if (zil_replay_disable)
	zil_destroy(zv->zv_zilog, B_FALSE);
	else
	zil_replay(os, zv, zvol_replay_vector);
	}
	zil_close(zv->zv_zilog);
	zv->zv_zilog = NULL;

	/*
	* When udev detects the addition of the device it will immediately
	* invoke blkid(8) to determine the type of content on the device.
	* Prefetching the blocks commonly scanned by blkid(8) will speed
	* up this process.
	*/
	- len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
	+ len = MIN(zvol_prefetch_bytes, SPA_MAXBLOCKSIZE);
	if (len > 0) {
	dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
	dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
	ZIO_PRIORITY_SYNC_READ);
	}

	zv->zv_objset = NULL;
	out_dmu_objset_disown:
	dmu_objset_disown(os, B_TRUE, FTAG);
	out_doi:
	kmem_free(doi, sizeof (dmu_object_info_t));

	/*
	* Keep in mind that once add_disk() is called, the zvol is
	* announced to the world, and zvol_open()/zvol_release() can
	* be called at any time. Incidentally, add_disk() itself calls
	* zvol_open()->zvol_first_open() and zvol_release()->zvol_last_close()
	* directly as well.
	*/
	if (error == 0) {
	rw_enter(&zvol_state_lock, RW_WRITER);
	zvol_insert(zv);
	rw_exit(&zvol_state_lock);
	#ifdef HAVE_ADD_DISK_RET
	error = add_disk(zv->zv_zso->zvo_disk);
	#else
	add_disk(zv->zv_zso->zvo_disk);
	#endif
	} else {
	ida_simple_remove(&zvol_ida, idx);
	}

	return (error);
	}

	void
	zvol_os_rename_minor(zvol_state_t zv, const char newname)
	{
	int readonly = get_disk_ro(zv->zv_zso->zvo_disk);

	ASSERT(RW_LOCK_HELD(&zvol_state_lock));
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	strlcpy(zv->zv_name, newname, sizeof (zv->zv_name));

	/* move to new hashtable entry */
	zv->zv_hash = zvol_name_hash(zv->zv_name);
	hlist_del(&zv->zv_hlink);
	hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash));

	/*
	* The block device's read-only state is briefly changed causing
	* a KOBJ_CHANGE uevent to be issued. This ensures udev detects
	* the name change and fixes the symlinks. This does not change
	* ZVOL_RDONLY in zv->zv_flags so the actual read-only state never
	* changes. This would normally be done using kobject_uevent() but
	* that is a GPL-only symbol which is why we need this workaround.
	*/
	set_disk_ro(zv->zv_zso->zvo_disk, !readonly);
	set_disk_ro(zv->zv_zso->zvo_disk, readonly);
	}

	void
	zvol_os_set_disk_ro(zvol_state_t *zv, int flags)
	{

	set_disk_ro(zv->zv_zso->zvo_disk, flags);
	}

	void
	zvol_os_set_capacity(zvol_state_t *zv, uint64_t capacity)
	{

	set_capacity(zv->zv_zso->zvo_disk, capacity);
	}

	int
	zvol_init(void)
	{
	int error;

	/*
	* zvol_threads is the module param the user passes in.
	*
	* zvol_actual_threads is what we use internally, since the user can
	* pass zvol_thread = 0 to mean "use all the CPUs" (the default).
	*/
	static unsigned int zvol_actual_threads;

	if (zvol_threads == 0) {
	/*
	* See dde9380a1 for why 32 was chosen here. This should
	* probably be refined to be some multiple of the number
	* of CPUs.
	*/
	zvol_actual_threads = MAX(num_online_cpus(), 32);
	} else {
	zvol_actual_threads = MIN(MAX(zvol_threads, 1), 1024);
	}

	error = register_blkdev(zvol_major, ZVOL_DRIVER);
	if (error) {
	printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error);
	return (error);
	}

	#ifdef HAVE_BLK_MQ
	if (zvol_blk_mq_queue_depth == 0) {
	zvol_actual_blk_mq_queue_depth = BLKDEV_DEFAULT_RQ;
	} else {
	zvol_actual_blk_mq_queue_depth =
	MAX(zvol_blk_mq_queue_depth, BLKDEV_MIN_RQ);
	}

	if (zvol_blk_mq_threads == 0) {
	zvol_blk_mq_actual_threads = num_online_cpus();
	} else {
	zvol_blk_mq_actual_threads = MIN(MAX(zvol_blk_mq_threads, 1),
	1024);
	}
	#endif
	zvol_taskq = taskq_create(ZVOL_DRIVER, zvol_actual_threads, maxclsyspri,
	zvol_actual_threads, INT_MAX, TASKQ_PREPOPULATE \| TASKQ_DYNAMIC);
	if (zvol_taskq == NULL) {
	unregister_blkdev(zvol_major, ZVOL_DRIVER);
	return (-ENOMEM);
	}

	zvol_init_impl();
	ida_init(&zvol_ida);
	return (0);
	}

	void
	zvol_fini(void)
	{
	zvol_fini_impl();
	unregister_blkdev(zvol_major, ZVOL_DRIVER);
	taskq_destroy(zvol_taskq);
	ida_destroy(&zvol_ida);
	}

	/* BEGIN CSTYLED */
	module_param(zvol_inhibit_dev, uint, 0644);
	MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes");

	module_param(zvol_major, uint, 0444);
	MODULE_PARM_DESC(zvol_major, "Major number for zvol device");

	module_param(zvol_threads, uint, 0444);
	MODULE_PARM_DESC(zvol_threads, "Number of threads to handle I/O requests. Set"
	"to 0 to use all active CPUs");

	module_param(zvol_request_sync, uint, 0644);
	MODULE_PARM_DESC(zvol_request_sync, "Synchronously handle bio requests");

	module_param(zvol_max_discard_blocks, ulong, 0444);
	MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard");

	module_param(zvol_prefetch_bytes, uint, 0644);
	MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end");

	module_param(zvol_volmode, uint, 0644);
	MODULE_PARM_DESC(zvol_volmode, "Default volmode property value");

	#ifdef HAVE_BLK_MQ
	module_param(zvol_blk_mq_queue_depth, uint, 0644);
	MODULE_PARM_DESC(zvol_blk_mq_queue_depth, "Default blk-mq queue depth");

	module_param(zvol_use_blk_mq, uint, 0644);
	MODULE_PARM_DESC(zvol_use_blk_mq, "Use the blk-mq API for zvols");

	module_param(zvol_blk_mq_blocks_per_thread, uint, 0644);
	MODULE_PARM_DESC(zvol_blk_mq_blocks_per_thread,
	"Process volblocksize blocks per thread");
	#endif

	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/arc.c b/sys/contrib/openzfs/module/zfs/arc.c
	index b9969bff534e..33865f715b0f 100644
	--- a/sys/contrib/openzfs/module/zfs/arc.c
	+++ b/sys/contrib/openzfs/module/zfs/arc.c
	@@ -1,11200 +1,11199 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2018, Joyent, Inc.
	* Copyright (c) 2011, 2020, Delphix. All rights reserved.
	* Copyright (c) 2014, Saso Kiselkov. All rights reserved.
	* Copyright (c) 2017, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2020, George Amanakis. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2020, The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	*/

	/*
	* DVA-based Adjustable Replacement Cache
	*
	* While much of the theory of operation used here is
	* based on the self-tuning, low overhead replacement cache
	* presented by Megiddo and Modha at FAST 2003, there are some
	* significant differences:
	*
	* 1. The Megiddo and Modha model assumes any page is evictable.
	* Pages in its cache cannot be "locked" into memory. This makes
	* the eviction algorithm simple: evict the last page in the list.
	* This also make the performance characteristics easy to reason
	* about. Our cache is not so simple. At any given moment, some
	* subset of the blocks in the cache are un-evictable because we
	* have handed out a reference to them. Blocks are only evictable
	* when there are no external references active. This makes
	* eviction far more problematic: we choose to evict the evictable
	* blocks that are the "lowest" in the list.
	*
	* There are times when it is not possible to evict the requested
	* space. In these circumstances we are unable to adjust the cache
	* size. To prevent the cache growing unbounded at these times we
	* implement a "cache throttle" that slows the flow of new data
	* into the cache until we can make space available.
	*
	* 2. The Megiddo and Modha model assumes a fixed cache size.
	* Pages are evicted when the cache is full and there is a cache
	* miss. Our model has a variable sized cache. It grows with
	* high use, but also tries to react to memory pressure from the
	* operating system: decreasing its size when system memory is
	* tight.
	*
	* 3. The Megiddo and Modha model assumes a fixed page size. All
	* elements of the cache are therefore exactly the same size. So
	* when adjusting the cache size following a cache miss, its simply
	* a matter of choosing a single page to evict. In our model, we
	* have variable sized cache blocks (ranging from 512 bytes to
	* 128K bytes). We therefore choose a set of blocks to evict to make
	* space for a cache miss that approximates as closely as possible
	* the space used by the new block.
	*
	* See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache"
	* by N. Megiddo & D. Modha, FAST 2003
	*/

	/*
	* The locking model:
	*
	* A new reference to a cache buffer can be obtained in two
	* ways: 1) via a hash table lookup using the DVA as a key,
	* or 2) via one of the ARC lists. The arc_read() interface
	* uses method 1, while the internal ARC algorithms for
	* adjusting the cache use method 2. We therefore provide two
	* types of locks: 1) the hash table lock array, and 2) the
	* ARC list locks.
	*
	* Buffers do not have their own mutexes, rather they rely on the
	* hash table mutexes for the bulk of their protection (i.e. most
	* fields in the arc_buf_hdr_t are protected by these mutexes).
	*
	* buf_hash_find() returns the appropriate mutex (held) when it
	* locates the requested buffer in the hash table. It returns
	* NULL for the mutex if the buffer was not in the table.
	*
	* buf_hash_remove() expects the appropriate hash mutex to be
	* already held before it is invoked.
	*
	* Each ARC state also has a mutex which is used to protect the
	* buffer list associated with the state. When attempting to
	* obtain a hash table lock while holding an ARC list lock you
	* must use: mutex_tryenter() to avoid deadlock. Also note that
	* the active state mutex must be held before the ghost state mutex.
	*
	* It as also possible to register a callback which is run when the
	* arc_meta_limit is reached and no buffers can be safely evicted. In
	* this case the arc user should drop a reference on some arc buffers so
	* they can be reclaimed and the arc_meta_limit honored. For example,
	* when using the ZPL each dentry holds a references on a znode. These
	* dentries must be pruned before the arc buffer holding the znode can
	* be safely evicted.
	*
	* Note that the majority of the performance stats are manipulated
	* with atomic operations.
	*
	* The L2ARC uses the l2ad_mtx on each vdev for the following:
	*
	* - L2ARC buflist creation
	* - L2ARC buflist eviction
	* - L2ARC write completion, which walks L2ARC buflists
	* - ARC header destruction, as it removes from L2ARC buflists
	* - ARC header release, as it removes from L2ARC buflists
	*/

	/*
	* ARC operation:
	*
	* Every block that is in the ARC is tracked by an arc_buf_hdr_t structure.
	* This structure can point either to a block that is still in the cache or to
	* one that is only accessible in an L2 ARC device, or it can provide
	* information about a block that was recently evicted. If a block is
	* only accessible in the L2ARC, then the arc_buf_hdr_t only has enough
	* information to retrieve it from the L2ARC device. This information is
	* stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block
	* that is in this state cannot access the data directly.
	*
	* Blocks that are actively being referenced or have not been evicted
	* are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within
	* the arc_buf_hdr_t that will point to the data block in memory. A block can
	* only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC
	* caches data in two ways -- in a list of ARC buffers (arc_buf_t) and
	* also in the arc_buf_hdr_t's private physical data block pointer (b_pabd).
	*
	* The L1ARC's data pointer may or may not be uncompressed. The ARC has the
	* ability to store the physical data (b_pabd) associated with the DVA of the
	* arc_buf_hdr_t. Since the b_pabd is a copy of the on-disk physical block,
	* it will match its on-disk compression characteristics. This behavior can be
	* disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the
	* compressed ARC functionality is disabled, the b_pabd will point to an
	* uncompressed version of the on-disk data.
	*
	* Data in the L1ARC is not accessed by consumers of the ARC directly. Each
	* arc_buf_hdr_t can have multiple ARC buffers (arc_buf_t) which reference it.
	* Each ARC buffer (arc_buf_t) is being actively accessed by a specific ARC
	* consumer. The ARC will provide references to this data and will keep it
	* cached until it is no longer in use. The ARC caches only the L1ARC's physical
	* data block and will evict any arc_buf_t that is no longer referenced. The
	* amount of memory consumed by the arc_buf_ts' data buffers can be seen via the
	* "overhead_size" kstat.
	*
	* Depending on the consumer, an arc_buf_t can be requested in uncompressed or
	* compressed form. The typical case is that consumers will want uncompressed
	* data, and when that happens a new data buffer is allocated where the data is
	* decompressed for them to use. Currently the only consumer who wants
	* compressed arc_buf_t's is "zfs send", when it streams data exactly as it
	* exists on disk. When this happens, the arc_buf_t's data buffer is shared
	* with the arc_buf_hdr_t.
	*
	* Here is a diagram showing an arc_buf_hdr_t referenced by two arc_buf_t's. The
	* first one is owned by a compressed send consumer (and therefore references
	* the same compressed data buffer as the arc_buf_hdr_t) and the second could be
	* used by any other consumer (and has its own uncompressed copy of the data
	* buffer).
	*
	* arc_buf_hdr_t
	* +-----------+
	* \| fields \|
	* \| common to \|
	* \| L1- and \|
	* \| L2ARC \|
	* +-----------+
	* \| l2arc_buf_hdr_t
	* \| \|
	* +-----------+
	* \| l1arc_buf_hdr_t
	* \| \| arc_buf_t
	* \| b_buf +------------>+-----------+ arc_buf_t
	* \| b_pabd +-+ \|b_next +---->+-----------+
	* +-----------+ \| \|-----------\| \|b_next +-->NULL
	* \| \|b_comp = T \| +-----------+
	* \| \|b_data +-+ \|b_comp = F \|
	* \| +-----------+ \| \|b_data +-+
	* +->+------+ \| +-----------+ \|
	* compressed \| \| \| \|
	* data \| \|<--------------+ \| uncompressed
	* +------+ compressed, \| data
	* shared +-->+------+
	* data \| \|
	* \| \|
	* +------+
	*
	* When a consumer reads a block, the ARC must first look to see if the
	* arc_buf_hdr_t is cached. If the hdr is cached then the ARC allocates a new
	* arc_buf_t and either copies uncompressed data into a new data buffer from an
	* existing uncompressed arc_buf_t, decompresses the hdr's b_pabd buffer into a
	* new data buffer, or shares the hdr's b_pabd buffer, depending on whether the
	* hdr is compressed and the desired compression characteristics of the
	* arc_buf_t consumer. If the arc_buf_t ends up sharing data with the
	* arc_buf_hdr_t and both of them are uncompressed then the arc_buf_t must be
	* the last buffer in the hdr's b_buf list, however a shared compressed buf can
	* be anywhere in the hdr's list.
	*
	* The diagram below shows an example of an uncompressed ARC hdr that is
	* sharing its data with an arc_buf_t (note that the shared uncompressed buf is
	* the last element in the buf list):
	*
	* arc_buf_hdr_t
	* +-----------+
	* \| \|
	* \| \|
	* \| \|
	* +-----------+
	* l2arc_buf_hdr_t\| \|
	* \| \|
	* +-----------+
	* l1arc_buf_hdr_t\| \|
	* \| \| arc_buf_t (shared)
	* \| b_buf +------------>+---------+ arc_buf_t
	* \| \| \|b_next +---->+---------+
	* \| b_pabd +-+ \|---------\| \|b_next +-->NULL
	* +-----------+ \| \| \| +---------+
	* \| \|b_data +-+ \| \|
	* \| +---------+ \| \|b_data +-+
	* +->+------+ \| +---------+ \|
	* \| \| \| \|
	* uncompressed \| \| \| \|
	* data +------+ \| \|
	* ^ +->+------+ \|
	* \| uncompressed \| \| \|
	* \| data \| \| \|
	* \| +------+ \|
	* +---------------------------------+
	*
	* Writing to the ARC requires that the ARC first discard the hdr's b_pabd
	* since the physical block is about to be rewritten. The new data contents
	* will be contained in the arc_buf_t. As the I/O pipeline performs the write,
	* it may compress the data before writing it to disk. The ARC will be called
	* with the transformed data and will memcpy the transformed on-disk block into
	* a newly allocated b_pabd. Writes are always done into buffers which have
	* either been loaned (and hence are new and don't have other readers) or
	* buffers which have been released (and hence have their own hdr, if there
	* were originally other readers of the buf's original hdr). This ensures that
	* the ARC only needs to update a single buf and its hdr after a write occurs.
	*
	* When the L2ARC is in use, it will also take advantage of the b_pabd. The
	* L2ARC will always write the contents of b_pabd to the L2ARC. This means
	* that when compressed ARC is enabled that the L2ARC blocks are identical
	* to the on-disk block in the main data pool. This provides a significant
	* advantage since the ARC can leverage the bp's checksum when reading from the
	* L2ARC to determine if the contents are valid. However, if the compressed
	* ARC is disabled, then the L2ARC's block must be transformed to look
	* like the physical block in the main data pool before comparing the
	* checksum and determining its validity.
	*
	* The L1ARC has a slightly different system for storing encrypted data.
	* Raw (encrypted + possibly compressed) data has a few subtle differences from
	* data that is just compressed. The biggest difference is that it is not
	* possible to decrypt encrypted data (or vice-versa) if the keys aren't loaded.
	* The other difference is that encryption cannot be treated as a suggestion.
	* If a caller would prefer compressed data, but they actually wind up with
	* uncompressed data the worst thing that could happen is there might be a
	* performance hit. If the caller requests encrypted data, however, we must be
	* sure they actually get it or else secret information could be leaked. Raw
	* data is stored in hdr->b_crypt_hdr.b_rabd. An encrypted header, therefore,
	* may have both an encrypted version and a decrypted version of its data at
	* once. When a caller needs a raw arc_buf_t, it is allocated and the data is
	* copied out of this header. To avoid complications with b_pabd, raw buffers
	* cannot be shared.
	*/

	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/zfs_refcount.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/dsl_pool.h>
	#include <sys/multilist.h>
	#include <sys/abd.h>
	#include <sys/zil.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/callb.h>
	#include <sys/kstat.h>
	#include <sys/zthr.h>
	#include <zfs_fletcher.h>
	#include <sys/arc_impl.h>
	#include <sys/trace_zfs.h>
	#include <sys/aggsum.h>
	#include <sys/wmsum.h>
	#include <cityhash.h>
	#include <sys/vdev_trim.h>
	#include <sys/zfs_racct.h>
	#include <sys/zstd/zstd.h>

	#ifndef _KERNEL
	/* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
	boolean_t arc_watch = B_FALSE;
	#endif

	/*
	* This thread's job is to keep enough free memory in the system, by
	* calling arc_kmem_reap_soon() plus arc_reduce_target_size(), which improves
	* arc_available_memory().
	*/
	static zthr_t *arc_reap_zthr;

	/*
	* This thread's job is to keep arc_size under arc_c, by calling
	* arc_evict(), which improves arc_is_overflowing().
	*/
	static zthr_t *arc_evict_zthr;
	static arc_buf_hdr_t **arc_state_evict_markers;
	static int arc_state_evict_marker_count;

	static kmutex_t arc_evict_lock;
	static boolean_t arc_evict_needed = B_FALSE;

	/*
	* Count of bytes evicted since boot.
	*/
	static uint64_t arc_evict_count;

	/*
	* List of arc_evict_waiter_t's, representing threads waiting for the
	* arc_evict_count to reach specific values.
	*/
	static list_t arc_evict_waiters;

	/*
	* When arc_is_overflowing(), arc_get_data_impl() waits for this percent of
	* the requested amount of data to be evicted. For example, by default for
	* every 2KB that's evicted, 1KB of it may be "reused" by a new allocation.
	* Since this is above 100%, it ensures that progress is made towards getting
	* arc_size under arc_c. Since this is finite, it ensures that allocations
	* can still happen, even during the potentially long time that arc_size is
	* more than arc_c.
	*/
	-static int zfs_arc_eviction_pct = 200;
	+static uint_t zfs_arc_eviction_pct = 200;

	/*
	* The number of headers to evict in arc_evict_state_impl() before
	* dropping the sublist lock and evicting from another sublist. A lower
	* value means we're more likely to evict the "correct" header (i.e. the
	* oldest header in the arc state), but comes with higher overhead
	* (i.e. more invocations of arc_evict_state_impl()).
	*/
	-static int zfs_arc_evict_batch_limit = 10;
	+static uint_t zfs_arc_evict_batch_limit = 10;

	/* number of seconds before growing cache again */
	-int arc_grow_retry = 5;
	+uint_t arc_grow_retry = 5;

	/*
	* Minimum time between calls to arc_kmem_reap_soon().
	*/
	static const int arc_kmem_cache_reap_retry_ms = 1000;

	/* shift of arc_c for calculating overflow limit in arc_get_data_impl */
	static int zfs_arc_overflow_shift = 8;

	/* shift of arc_c for calculating both min and max arc_p */
	-static int arc_p_min_shift = 4;
	+static uint_t arc_p_min_shift = 4;

	/* log2(fraction of arc to reclaim) */
	-int arc_shrink_shift = 7;
	+uint_t arc_shrink_shift = 7;

	/* percent of pagecache to reclaim arc to */
	#ifdef _KERNEL
	uint_t zfs_arc_pc_percent = 0;
	#endif

	/*
	* log2(fraction of ARC which must be free to allow growing).
	* I.e. If there is less than arc_c >> arc_no_grow_shift free memory,
	* when reading a new block into the ARC, we will evict an equal-sized block
	* from the ARC.
	*
	* This must be less than arc_shrink_shift, so that when we shrink the ARC,
	* we will still not allow it to grow.
	*/
	-int arc_no_grow_shift = 5;
	+uint_t arc_no_grow_shift = 5;


	/*
	* minimum lifespan of a prefetch block in clock ticks
	* (initialized in arc_init())
	*/
	-static int arc_min_prefetch_ms;
	-static int arc_min_prescient_prefetch_ms;
	+static uint_t arc_min_prefetch_ms;
	+static uint_t arc_min_prescient_prefetch_ms;

	/*
	* If this percent of memory is free, don't throttle.
	*/
	-int arc_lotsfree_percent = 10;
	+uint_t arc_lotsfree_percent = 10;

	/*
	* The arc has filled available memory and has now warmed up.
	*/
	boolean_t arc_warm;

	/*
	* These tunables are for performance analysis.
	*/
	unsigned long zfs_arc_max = 0;
	unsigned long zfs_arc_min = 0;
	unsigned long zfs_arc_meta_limit = 0;
	unsigned long zfs_arc_meta_min = 0;
	static unsigned long zfs_arc_dnode_limit = 0;
	static unsigned long zfs_arc_dnode_reduce_percent = 10;
	-static int zfs_arc_grow_retry = 0;
	-static int zfs_arc_shrink_shift = 0;
	-static int zfs_arc_p_min_shift = 0;
	-int zfs_arc_average_blocksize = 8 * 1024; /* 8KB */
	+static uint_t zfs_arc_grow_retry = 0;
	+static uint_t zfs_arc_shrink_shift = 0;
	+static uint_t zfs_arc_p_min_shift = 0;
	+uint_t zfs_arc_average_blocksize = 8 * 1024; /* 8KB */

	/*
	* ARC dirty data constraints for arc_tempreserve_space() throttle:
	* * total dirty data limit
	* * anon block dirty limit
	* * each pool's anon allowance
	*/
	static const unsigned long zfs_arc_dirty_limit_percent = 50;
	static const unsigned long zfs_arc_anon_limit_percent = 25;
	static const unsigned long zfs_arc_pool_dirty_percent = 20;

	/*
	* Enable or disable compressed arc buffers.
	*/
	int zfs_compressed_arc_enabled = B_TRUE;

	/*
	* ARC will evict meta buffers that exceed arc_meta_limit. This
	* tunable make arc_meta_limit adjustable for different workloads.
	*/
	static unsigned long zfs_arc_meta_limit_percent = 75;

	/*
	* Percentage that can be consumed by dnodes of ARC meta buffers.
	*/
	static unsigned long zfs_arc_dnode_limit_percent = 10;

	/*
	* These tunables are Linux-specific
	*/
	static unsigned long zfs_arc_sys_free = 0;
	-static int zfs_arc_min_prefetch_ms = 0;
	-static int zfs_arc_min_prescient_prefetch_ms = 0;
	+static uint_t zfs_arc_min_prefetch_ms = 0;
	+static uint_t zfs_arc_min_prescient_prefetch_ms = 0;
	static int zfs_arc_p_dampener_disable = 1;
	-static int zfs_arc_meta_prune = 10000;
	-static int zfs_arc_meta_strategy = ARC_STRATEGY_META_BALANCED;
	-static int zfs_arc_meta_adjust_restarts = 4096;
	-static int zfs_arc_lotsfree_percent = 10;
	+static uint_t zfs_arc_meta_prune = 10000;
	+static uint_t zfs_arc_meta_strategy = ARC_STRATEGY_META_BALANCED;
	+static uint_t zfs_arc_meta_adjust_restarts = 4096;
	+static uint_t zfs_arc_lotsfree_percent = 10;

	/*
	* Number of arc_prune threads
	*/
	static int zfs_arc_prune_task_threads = 1;

	/* The 6 states: */
	arc_state_t ARC_anon;
	arc_state_t ARC_mru;
	arc_state_t ARC_mru_ghost;
	arc_state_t ARC_mfu;
	arc_state_t ARC_mfu_ghost;
	arc_state_t ARC_l2c_only;

	arc_stats_t arc_stats = {
	{ "hits", KSTAT_DATA_UINT64 },
	{ "misses", KSTAT_DATA_UINT64 },
	{ "demand_data_hits", KSTAT_DATA_UINT64 },
	{ "demand_data_misses", KSTAT_DATA_UINT64 },
	{ "demand_metadata_hits", KSTAT_DATA_UINT64 },
	{ "demand_metadata_misses", KSTAT_DATA_UINT64 },
	{ "prefetch_data_hits", KSTAT_DATA_UINT64 },
	{ "prefetch_data_misses", KSTAT_DATA_UINT64 },
	{ "prefetch_metadata_hits", KSTAT_DATA_UINT64 },
	{ "prefetch_metadata_misses", KSTAT_DATA_UINT64 },
	{ "mru_hits", KSTAT_DATA_UINT64 },
	{ "mru_ghost_hits", KSTAT_DATA_UINT64 },
	{ "mfu_hits", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_hits", KSTAT_DATA_UINT64 },
	{ "deleted", KSTAT_DATA_UINT64 },
	{ "mutex_miss", KSTAT_DATA_UINT64 },
	{ "access_skip", KSTAT_DATA_UINT64 },
	{ "evict_skip", KSTAT_DATA_UINT64 },
	{ "evict_not_enough", KSTAT_DATA_UINT64 },
	{ "evict_l2_cached", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible_mfu", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible_mru", KSTAT_DATA_UINT64 },
	{ "evict_l2_ineligible", KSTAT_DATA_UINT64 },
	{ "evict_l2_skip", KSTAT_DATA_UINT64 },
	{ "hash_elements", KSTAT_DATA_UINT64 },
	{ "hash_elements_max", KSTAT_DATA_UINT64 },
	{ "hash_collisions", KSTAT_DATA_UINT64 },
	{ "hash_chains", KSTAT_DATA_UINT64 },
	{ "hash_chain_max", KSTAT_DATA_UINT64 },
	{ "p", KSTAT_DATA_UINT64 },
	{ "c", KSTAT_DATA_UINT64 },
	{ "c_min", KSTAT_DATA_UINT64 },
	{ "c_max", KSTAT_DATA_UINT64 },
	{ "size", KSTAT_DATA_UINT64 },
	{ "compressed_size", KSTAT_DATA_UINT64 },
	{ "uncompressed_size", KSTAT_DATA_UINT64 },
	{ "overhead_size", KSTAT_DATA_UINT64 },
	{ "hdr_size", KSTAT_DATA_UINT64 },
	{ "data_size", KSTAT_DATA_UINT64 },
	{ "metadata_size", KSTAT_DATA_UINT64 },
	{ "dbuf_size", KSTAT_DATA_UINT64 },
	{ "dnode_size", KSTAT_DATA_UINT64 },
	{ "bonus_size", KSTAT_DATA_UINT64 },
	#if defined(COMPAT_FREEBSD11)
	{ "other_size", KSTAT_DATA_UINT64 },
	#endif
	{ "anon_size", KSTAT_DATA_UINT64 },
	{ "anon_evictable_data", KSTAT_DATA_UINT64 },
	{ "anon_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mru_size", KSTAT_DATA_UINT64 },
	{ "mru_evictable_data", KSTAT_DATA_UINT64 },
	{ "mru_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mru_ghost_size", KSTAT_DATA_UINT64 },
	{ "mru_ghost_evictable_data", KSTAT_DATA_UINT64 },
	{ "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_size", KSTAT_DATA_UINT64 },
	{ "mfu_evictable_data", KSTAT_DATA_UINT64 },
	{ "mfu_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_size", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "l2_hits", KSTAT_DATA_UINT64 },
	{ "l2_misses", KSTAT_DATA_UINT64 },
	{ "l2_prefetch_asize", KSTAT_DATA_UINT64 },
	{ "l2_mru_asize", KSTAT_DATA_UINT64 },
	{ "l2_mfu_asize", KSTAT_DATA_UINT64 },
	{ "l2_bufc_data_asize", KSTAT_DATA_UINT64 },
	{ "l2_bufc_metadata_asize", KSTAT_DATA_UINT64 },
	{ "l2_feeds", KSTAT_DATA_UINT64 },
	{ "l2_rw_clash", KSTAT_DATA_UINT64 },
	{ "l2_read_bytes", KSTAT_DATA_UINT64 },
	{ "l2_write_bytes", KSTAT_DATA_UINT64 },
	{ "l2_writes_sent", KSTAT_DATA_UINT64 },
	{ "l2_writes_done", KSTAT_DATA_UINT64 },
	{ "l2_writes_error", KSTAT_DATA_UINT64 },
	{ "l2_writes_lock_retry", KSTAT_DATA_UINT64 },
	{ "l2_evict_lock_retry", KSTAT_DATA_UINT64 },
	{ "l2_evict_reading", KSTAT_DATA_UINT64 },
	{ "l2_evict_l1cached", KSTAT_DATA_UINT64 },
	{ "l2_free_on_write", KSTAT_DATA_UINT64 },
	{ "l2_abort_lowmem", KSTAT_DATA_UINT64 },
	{ "l2_cksum_bad", KSTAT_DATA_UINT64 },
	{ "l2_io_error", KSTAT_DATA_UINT64 },
	{ "l2_size", KSTAT_DATA_UINT64 },
	{ "l2_asize", KSTAT_DATA_UINT64 },
	{ "l2_hdr_size", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_writes", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_avg_asize", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_asize", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_count", KSTAT_DATA_UINT64 },
	{ "l2_data_to_meta_ratio", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_success", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_unsupported", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_io_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_dh_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_cksum_lb_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_lowmem", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_size", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_asize", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_bufs", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_bufs_precached", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_log_blks", KSTAT_DATA_UINT64 },
	{ "memory_throttle_count", KSTAT_DATA_UINT64 },
	{ "memory_direct_count", KSTAT_DATA_UINT64 },
	{ "memory_indirect_count", KSTAT_DATA_UINT64 },
	{ "memory_all_bytes", KSTAT_DATA_UINT64 },
	{ "memory_free_bytes", KSTAT_DATA_UINT64 },
	{ "memory_available_bytes", KSTAT_DATA_INT64 },
	{ "arc_no_grow", KSTAT_DATA_UINT64 },
	{ "arc_tempreserve", KSTAT_DATA_UINT64 },
	{ "arc_loaned_bytes", KSTAT_DATA_UINT64 },
	{ "arc_prune", KSTAT_DATA_UINT64 },
	{ "arc_meta_used", KSTAT_DATA_UINT64 },
	{ "arc_meta_limit", KSTAT_DATA_UINT64 },
	{ "arc_dnode_limit", KSTAT_DATA_UINT64 },
	{ "arc_meta_max", KSTAT_DATA_UINT64 },
	{ "arc_meta_min", KSTAT_DATA_UINT64 },
	{ "async_upgrade_sync", KSTAT_DATA_UINT64 },
	{ "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 },
	{ "demand_hit_prescient_prefetch", KSTAT_DATA_UINT64 },
	{ "arc_need_free", KSTAT_DATA_UINT64 },
	{ "arc_sys_free", KSTAT_DATA_UINT64 },
	{ "arc_raw_size", KSTAT_DATA_UINT64 },
	{ "cached_only_in_progress", KSTAT_DATA_UINT64 },
	{ "abd_chunk_waste_size", KSTAT_DATA_UINT64 },
	};

	arc_sums_t arc_sums;

	#define ARCSTAT_MAX(stat, val) { \
	uint64_t m; \
	while ((val) > (m = arc_stats.stat.value.ui64) && \
	(m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \
	continue; \
	}

	/*
	* We define a macro to allow ARC hits/misses to be easily broken down by
	* two separate conditions, giving a total of four different subtypes for
	* each of hits and misses (so eight statistics total).
	*/
	#define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \
	if (cond1) { \
	if (cond2) { \
	ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \
	} else { \
	ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \
	} \
	} else { \
	if (cond2) { \
	ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \
	} else { \
	ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\
	} \
	}

	/*
	* This macro allows us to use kstats as floating averages. Each time we
	* update this kstat, we first factor it and the update value by
	* ARCSTAT_AVG_FACTOR to shrink the new value's contribution to the overall
	* average. This macro assumes that integer loads and stores are atomic, but
	* is not safe for multiple writers updating the kstat in parallel (only the
	* last writer's update will remain).
	*/
	#define ARCSTAT_F_AVG_FACTOR 3
	#define ARCSTAT_F_AVG(stat, value) \
	do { \
	uint64_t x = ARCSTAT(stat); \
	x = x - x / ARCSTAT_F_AVG_FACTOR + \
	(value) / ARCSTAT_F_AVG_FACTOR; \
	ARCSTAT(stat) = x; \
	} while (0)

	static kstat_t *arc_ksp;

	/*
	* There are several ARC variables that are critical to export as kstats --
	* but we don't want to have to grovel around in the kstat whenever we wish to
	* manipulate them. For these variables, we therefore define them to be in
	* terms of the statistic variable. This assures that we are not introducing
	* the possibility of inconsistency by having shadow copies of the variables,
	* while still allowing the code to be readable.
	*/
	#define arc_tempreserve ARCSTAT(arcstat_tempreserve)
	#define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes)
	#define arc_meta_limit ARCSTAT(arcstat_meta_limit) /* max size for metadata */
	/* max size for dnodes */
	#define arc_dnode_size_limit ARCSTAT(arcstat_dnode_limit)
	#define arc_meta_min ARCSTAT(arcstat_meta_min) /* min size for metadata */
	#define arc_need_free ARCSTAT(arcstat_need_free) /* waiting to be evicted */

	hrtime_t arc_growtime;
	list_t arc_prune_list;
	kmutex_t arc_prune_mtx;
	taskq_t *arc_prune_taskq;

	#define GHOST_STATE(state) \
	((state) == arc_mru_ghost \|\| (state) == arc_mfu_ghost \|\| \
	(state) == arc_l2c_only)

	#define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE)
	#define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS)
	#define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR)
	#define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH)
	#define HDR_PRESCIENT_PREFETCH(hdr) \
	((hdr)->b_flags & ARC_FLAG_PRESCIENT_PREFETCH)
	#define HDR_COMPRESSION_ENABLED(hdr) \
	((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC)

	#define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE)
	#define HDR_L2_READING(hdr) \
	(((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \
	((hdr)->b_flags & ARC_FLAG_HAS_L2HDR))
	#define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING)
	#define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED)
	#define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD)
	#define HDR_PROTECTED(hdr) ((hdr)->b_flags & ARC_FLAG_PROTECTED)
	#define HDR_NOAUTH(hdr) ((hdr)->b_flags & ARC_FLAG_NOAUTH)
	#define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA)

	#define HDR_ISTYPE_METADATA(hdr) \
	((hdr)->b_flags & ARC_FLAG_BUFC_METADATA)
	#define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr))

	#define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR)
	#define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)
	#define HDR_HAS_RABD(hdr) \
	(HDR_HAS_L1HDR(hdr) && HDR_PROTECTED(hdr) && \
	(hdr)->b_crypt_hdr.b_rabd != NULL)
	#define HDR_ENCRYPTED(hdr) \
	(HDR_PROTECTED(hdr) && DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))
	#define HDR_AUTHENTICATED(hdr) \
	(HDR_PROTECTED(hdr) && !DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))

	/* For storing compression mode in b_flags */
	#define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1)

	#define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \
	HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS))
	#define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \
	HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp));

	#define ARC_BUF_LAST(buf) ((buf)->b_next == NULL)
	#define ARC_BUF_SHARED(buf) ((buf)->b_flags & ARC_BUF_FLAG_SHARED)
	#define ARC_BUF_COMPRESSED(buf) ((buf)->b_flags & ARC_BUF_FLAG_COMPRESSED)
	#define ARC_BUF_ENCRYPTED(buf) ((buf)->b_flags & ARC_BUF_FLAG_ENCRYPTED)

	/*
	* Other sizes
	*/

	#define HDR_FULL_CRYPT_SIZE ((int64_t)sizeof (arc_buf_hdr_t))
	#define HDR_FULL_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_crypt_hdr))
	#define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr))

	/*
	* Hash table routines
	*/

	#define BUF_LOCKS 2048
	typedef struct buf_hash_table {
	uint64_t ht_mask;
	arc_buf_hdr_t **ht_table;
	kmutex_t ht_locks[BUF_LOCKS] ____cacheline_aligned;
	} buf_hash_table_t;

	static buf_hash_table_t buf_hash_table;

	#define BUF_HASH_INDEX(spa, dva, birth) \
	(buf_hash(spa, dva, birth) & buf_hash_table.ht_mask)
	#define BUF_HASH_LOCK(idx) (&buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)])
	#define HDR_LOCK(hdr) \
	(BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth)))

	uint64_t zfs_crc64_table[256];

	/*
	* Level 2 ARC
	*/

	#define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */
	#define L2ARC_HEADROOM 2 /* num of writes */

	/*
	* If we discover during ARC scan any buffers to be compressed, we boost
	* our headroom for the next scanning cycle by this percentage multiple.
	*/
	#define L2ARC_HEADROOM_BOOST 200
	#define L2ARC_FEED_SECS 1 /* caching interval secs */
	#define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */

	/*
	* We can feed L2ARC from two states of ARC buffers, mru and mfu,
	* and each of the state has two types: data and metadata.
	*/
	#define L2ARC_FEED_TYPES 4

	/* L2ARC Performance Tunables */
	unsigned long l2arc_write_max = L2ARC_WRITE_SIZE; /* def max write size */
	unsigned long l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */
	unsigned long l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */
	unsigned long l2arc_headroom_boost = L2ARC_HEADROOM_BOOST;
	unsigned long l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */
	unsigned long l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */
	int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */
	int l2arc_feed_again = B_TRUE; /* turbo warmup */
	int l2arc_norw = B_FALSE; /* no reads during writes */
	-static int l2arc_meta_percent = 33; /* limit on headers size */
	+static uint_t l2arc_meta_percent = 33; /* limit on headers size */

	/*
	* L2ARC Internals
	*/
	static list_t L2ARC_dev_list; /* device list */
	static list_t l2arc_dev_list; / device list pointer */
	static kmutex_t l2arc_dev_mtx; /* device list mutex */
	static l2arc_dev_t l2arc_dev_last; / last device used */
	static list_t L2ARC_free_on_write; /* free after write buf list */
	static list_t l2arc_free_on_write; / free after write list ptr */
	static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */
	static uint64_t l2arc_ndev; /* number of devices */

	typedef struct l2arc_read_callback {
	arc_buf_hdr_t l2rcb_hdr; / read header */
	blkptr_t l2rcb_bp; /* original blkptr */
	zbookmark_phys_t l2rcb_zb; /* original bookmark */
	int l2rcb_flags; /* original flags */
	abd_t l2rcb_abd; / temporary buffer */
	} l2arc_read_callback_t;

	typedef struct l2arc_data_free {
	/* protected by l2arc_free_on_write_mtx */
	abd_t *l2df_abd;
	size_t l2df_size;
	arc_buf_contents_t l2df_type;
	list_node_t l2df_list_node;
	} l2arc_data_free_t;

	typedef enum arc_fill_flags {
	ARC_FILL_LOCKED = 1 << 0, /* hdr lock is held */
	ARC_FILL_COMPRESSED = 1 << 1, /* fill with compressed data */
	ARC_FILL_ENCRYPTED = 1 << 2, /* fill with encrypted data */
	ARC_FILL_NOAUTH = 1 << 3, /* don't attempt to authenticate */
	ARC_FILL_IN_PLACE = 1 << 4 /* fill in place (special case) */
	} arc_fill_flags_t;

	typedef enum arc_ovf_level {
	ARC_OVF_NONE, /* ARC within target size. */
	ARC_OVF_SOME, /* ARC is slightly overflowed. */
	ARC_OVF_SEVERE /* ARC is severely overflowed. */
	} arc_ovf_level_t;

	static kmutex_t l2arc_feed_thr_lock;
	static kcondvar_t l2arc_feed_thr_cv;
	static uint8_t l2arc_thread_exit;

	static kmutex_t l2arc_rebuild_thr_lock;
	static kcondvar_t l2arc_rebuild_thr_cv;

	enum arc_hdr_alloc_flags {
	ARC_HDR_ALLOC_RDATA = 0x1,
	ARC_HDR_DO_ADAPT = 0x2,
	ARC_HDR_USE_RESERVE = 0x4,
	};


	static abd_t arc_get_data_abd(arc_buf_hdr_t , uint64_t, const void *, int);
	static void arc_get_data_buf(arc_buf_hdr_t , uint64_t, const void *);
	static void arc_get_data_impl(arc_buf_hdr_t , uint64_t, const void , int);
	static void arc_free_data_abd(arc_buf_hdr_t , abd_t , uint64_t, const void *);
	static void arc_free_data_buf(arc_buf_hdr_t , void , uint64_t, const void *);
	static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size,
	const void *tag);
	static void arc_hdr_free_abd(arc_buf_hdr_t *, boolean_t);
	static void arc_hdr_alloc_abd(arc_buf_hdr_t *, int);
	static void arc_access(arc_buf_hdr_t , kmutex_t );
	static void arc_buf_watch(arc_buf_t *);

	static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *);
	static uint32_t arc_bufc_to_flags(arc_buf_contents_t);
	static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
	static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);

	static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *);
	static void l2arc_read_done(zio_t *);
	static void l2arc_do_free_on_write(void);
	static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
	boolean_t state_only);

	#define l2arc_hdr_arcstats_increment(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_TRUE, B_FALSE)
	#define l2arc_hdr_arcstats_decrement(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_FALSE, B_FALSE)
	#define l2arc_hdr_arcstats_increment_state(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_TRUE, B_TRUE)
	#define l2arc_hdr_arcstats_decrement_state(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_FALSE, B_TRUE)

	/*
	* l2arc_exclude_special : A zfs module parameter that controls whether buffers
	* present on special vdevs are eligibile for caching in L2ARC. If
	* set to 1, exclude dbufs on special vdevs from being cached to
	* L2ARC.
	*/
	int l2arc_exclude_special = 0;

	/*
	* l2arc_mfuonly : A ZFS module parameter that controls whether only MFU
	* metadata and data are cached from ARC into L2ARC.
	*/
	static int l2arc_mfuonly = 0;

	/*
	* L2ARC TRIM
	* l2arc_trim_ahead : A ZFS module parameter that controls how much ahead of
	* the current write size (l2arc_write_max) we should TRIM if we
	* have filled the device. It is defined as a percentage of the
	* write size. If set to 100 we trim twice the space required to
	* accommodate upcoming writes. A minimum of 64MB will be trimmed.
	* It also enables TRIM of the whole L2ARC device upon creation or
	* addition to an existing pool or if the header of the device is
	* invalid upon importing a pool or onlining a cache device. The
	* default is 0, which disables TRIM on L2ARC altogether as it can
	* put significant stress on the underlying storage devices. This
	* will vary depending of how well the specific device handles
	* these commands.
	*/
	static unsigned long l2arc_trim_ahead = 0;

	/*
	* Performance tuning of L2ARC persistence:
	*
	* l2arc_rebuild_enabled : A ZFS module parameter that controls whether adding
	* an L2ARC device (either at pool import or later) will attempt
	* to rebuild L2ARC buffer contents.
	* l2arc_rebuild_blocks_min_l2size : A ZFS module parameter that controls
	* whether log blocks are written to the L2ARC device. If the L2ARC
	* device is less than 1GB, the amount of data l2arc_evict()
	* evicts is significant compared to the amount of restored L2ARC
	* data. In this case do not write log blocks in L2ARC in order
	* not to waste space.
	*/
	static int l2arc_rebuild_enabled = B_TRUE;
	static unsigned long l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024;

	/* L2ARC persistence rebuild control routines. */
	void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen);
	static __attribute__((noreturn)) void l2arc_dev_rebuild_thread(void *arg);
	static int l2arc_rebuild(l2arc_dev_t *dev);

	/* L2ARC persistence read I/O routines. */
	static int l2arc_dev_hdr_read(l2arc_dev_t *dev);
	static int l2arc_log_blk_read(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t this_lp, const l2arc_log_blkptr_t next_lp,
	l2arc_log_blk_phys_t this_lb, l2arc_log_blk_phys_t next_lb,
	zio_t this_io, zio_t *next_io);
	static zio_t l2arc_log_blk_fetch(vdev_t vd,
	const l2arc_log_blkptr_t lp, l2arc_log_blk_phys_t lb);
	static void l2arc_log_blk_fetch_abort(zio_t *zio);

	/* L2ARC persistence block restoration routines. */
	static void l2arc_log_blk_restore(l2arc_dev_t *dev,
	const l2arc_log_blk_phys_t *lb, uint64_t lb_asize);
	static void l2arc_hdr_restore(const l2arc_log_ent_phys_t *le,
	l2arc_dev_t *dev);

	/* L2ARC persistence write I/O routines. */
	static void l2arc_log_blk_commit(l2arc_dev_t dev, zio_t pio,
	l2arc_write_callback_t *cb);

	/* L2ARC persistence auxiliary routines. */
	boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t *lbp);
	static boolean_t l2arc_log_blk_insert(l2arc_dev_t *dev,
	const arc_buf_hdr_t *ab);
	boolean_t l2arc_range_check_overlap(uint64_t bottom,
	uint64_t top, uint64_t check);
	static void l2arc_blk_fetch_done(zio_t *zio);
	static inline uint64_t
	l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev);

	/*
	* We use Cityhash for this. It's fast, and has good hash properties without
	* requiring any large static buffers.
	*/
	static uint64_t
	buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth)
	{
	return (cityhash4(spa, dva->dva_word[0], dva->dva_word[1], birth));
	}

	#define HDR_EMPTY(hdr) \
	((hdr)->b_dva.dva_word[0] == 0 && \
	(hdr)->b_dva.dva_word[1] == 0)

	#define HDR_EMPTY_OR_LOCKED(hdr) \
	(HDR_EMPTY(hdr) \|\| MUTEX_HELD(HDR_LOCK(hdr)))

	#define HDR_EQUAL(spa, dva, birth, hdr) \
	((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
	((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
	((hdr)->b_birth == birth) && ((hdr)->b_spa == spa)

	static void
	buf_discard_identity(arc_buf_hdr_t *hdr)
	{
	hdr->b_dva.dva_word[0] = 0;
	hdr->b_dva.dva_word[1] = 0;
	hdr->b_birth = 0;
	}

	static arc_buf_hdr_t *
	buf_hash_find(uint64_t spa, const blkptr_t bp, kmutex_t *lockp)
	{
	const dva_t *dva = BP_IDENTITY(bp);
	uint64_t birth = BP_PHYSICAL_BIRTH(bp);
	uint64_t idx = BUF_HASH_INDEX(spa, dva, birth);
	kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
	arc_buf_hdr_t *hdr;

	mutex_enter(hash_lock);
	for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL;
	hdr = hdr->b_hash_next) {
	if (HDR_EQUAL(spa, dva, birth, hdr)) {
	*lockp = hash_lock;
	return (hdr);
	}
	}
	mutex_exit(hash_lock);
	*lockp = NULL;
	return (NULL);
	}

	/*
	* Insert an entry into the hash table. If there is already an element
	* equal to elem in the hash table, then the already existing element
	* will be returned and the new element will not be inserted.
	* Otherwise returns NULL.
	* If lockp == NULL, the caller is assumed to already hold the hash lock.
	*/
	static arc_buf_hdr_t *
	buf_hash_insert(arc_buf_hdr_t hdr, kmutex_t *lockp)
	{
	uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);
	kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
	arc_buf_hdr_t *fhdr;
	uint32_t i;

	ASSERT(!DVA_IS_EMPTY(&hdr->b_dva));
	ASSERT(hdr->b_birth != 0);
	ASSERT(!HDR_IN_HASH_TABLE(hdr));

	if (lockp != NULL) {
	*lockp = hash_lock;
	mutex_enter(hash_lock);
	} else {
	ASSERT(MUTEX_HELD(hash_lock));
	}

	for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL;
	fhdr = fhdr->b_hash_next, i++) {
	if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr))
	return (fhdr);
	}

	hdr->b_hash_next = buf_hash_table.ht_table[idx];
	buf_hash_table.ht_table[idx] = hdr;
	arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);

	/* collect some hash table performance data */
	if (i > 0) {
	ARCSTAT_BUMP(arcstat_hash_collisions);
	if (i == 1)
	ARCSTAT_BUMP(arcstat_hash_chains);

	ARCSTAT_MAX(arcstat_hash_chain_max, i);
	}
	uint64_t he = atomic_inc_64_nv(
	&arc_stats.arcstat_hash_elements.value.ui64);
	ARCSTAT_MAX(arcstat_hash_elements_max, he);

	return (NULL);
	}

	static void
	buf_hash_remove(arc_buf_hdr_t *hdr)
	{
	arc_buf_hdr_t fhdr, *hdrp;
	uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);

	ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx)));
	ASSERT(HDR_IN_HASH_TABLE(hdr));

	hdrp = &buf_hash_table.ht_table[idx];
	while ((fhdr = *hdrp) != hdr) {
	ASSERT3P(fhdr, !=, NULL);
	hdrp = &fhdr->b_hash_next;
	}
	*hdrp = hdr->b_hash_next;
	hdr->b_hash_next = NULL;
	arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE);

	/* collect some hash table performance data */
	atomic_dec_64(&arc_stats.arcstat_hash_elements.value.ui64);

	if (buf_hash_table.ht_table[idx] &&
	buf_hash_table.ht_table[idx]->b_hash_next == NULL)
	ARCSTAT_BUMPDOWN(arcstat_hash_chains);
	}

	/*
	* Global data structures and functions for the buf kmem cache.
	*/

	static kmem_cache_t *hdr_full_cache;
	static kmem_cache_t *hdr_full_crypt_cache;
	static kmem_cache_t *hdr_l2only_cache;
	static kmem_cache_t *buf_cache;

	static void
	buf_fini(void)
	{
	#if defined(_KERNEL)
	/*
	* Large allocations which do not require contiguous pages
	* should be using vmem_free() in the linux kernel\
	*/
	vmem_free(buf_hash_table.ht_table,
	(buf_hash_table.ht_mask + 1) * sizeof (void *));
	#else
	kmem_free(buf_hash_table.ht_table,
	(buf_hash_table.ht_mask + 1) * sizeof (void *));
	#endif
	for (int i = 0; i < BUF_LOCKS; i++)
	mutex_destroy(BUF_HASH_LOCK(i));
	kmem_cache_destroy(hdr_full_cache);
	kmem_cache_destroy(hdr_full_crypt_cache);
	kmem_cache_destroy(hdr_l2only_cache);
	kmem_cache_destroy(buf_cache);
	}

	/*
	* Constructor callback - called when the cache is empty
	* and a new buf is requested.
	*/
	static int
	hdr_full_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_hdr_t *hdr = vbuf;

	memset(hdr, 0, HDR_FULL_SIZE);
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	cv_init(&hdr->b_l1hdr.b_cv, NULL, CV_DEFAULT, NULL);
	zfs_refcount_create(&hdr->b_l1hdr.b_refcnt);
	mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL);
	list_link_init(&hdr->b_l1hdr.b_arc_node);
	list_link_init(&hdr->b_l2hdr.b_l2node);
	multilist_link_init(&hdr->b_l1hdr.b_arc_node);
	arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS);

	return (0);
	}

	static int
	hdr_full_crypt_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused;
	arc_buf_hdr_t *hdr = vbuf;

	hdr_full_cons(vbuf, unused, kmflag);
	memset(&hdr->b_crypt_hdr, 0, sizeof (hdr->b_crypt_hdr));
	arc_space_consume(sizeof (hdr->b_crypt_hdr), ARC_SPACE_HDRS);

	return (0);
	}

	static int
	hdr_l2only_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_hdr_t *hdr = vbuf;

	memset(hdr, 0, HDR_L2ONLY_SIZE);
	arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);

	return (0);
	}

	static int
	buf_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_t *buf = vbuf;

	memset(buf, 0, sizeof (arc_buf_t));
	mutex_init(&buf->b_evict_lock, NULL, MUTEX_DEFAULT, NULL);
	arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS);

	return (0);
	}

	/*
	* Destructor callback - called when a cached buf is
	* no longer required.
	*/
	static void
	hdr_full_dest(void vbuf, void unused)
	{
	(void) unused;
	arc_buf_hdr_t *hdr = vbuf;

	ASSERT(HDR_EMPTY(hdr));
	cv_destroy(&hdr->b_l1hdr.b_cv);
	zfs_refcount_destroy(&hdr->b_l1hdr.b_refcnt);
	mutex_destroy(&hdr->b_l1hdr.b_freeze_lock);
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS);
	}

	static void
	hdr_full_crypt_dest(void vbuf, void unused)
	{
	(void) vbuf, (void) unused;

	hdr_full_dest(vbuf, unused);
	arc_space_return(sizeof (((arc_buf_hdr_t *)NULL)->b_crypt_hdr),
	ARC_SPACE_HDRS);
	}

	static void
	hdr_l2only_dest(void vbuf, void unused)
	{
	(void) unused;
	arc_buf_hdr_t *hdr = vbuf;

	ASSERT(HDR_EMPTY(hdr));
	arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);
	}

	static void
	buf_dest(void vbuf, void unused)
	{
	(void) unused;
	arc_buf_t *buf = vbuf;

	mutex_destroy(&buf->b_evict_lock);
	arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS);
	}

	static void
	buf_init(void)
	{
	uint64_t *ct = NULL;
	uint64_t hsize = 1ULL << 12;
	int i, j;

	/*
	* The hash table is big enough to fill all of physical memory
	* with an average block size of zfs_arc_average_blocksize (default 8K).
	* By default, the table will take up
	* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
	*/
	while (hsize * zfs_arc_average_blocksize < arc_all_memory())
	hsize <<= 1;
	retry:
	buf_hash_table.ht_mask = hsize - 1;
	#if defined(_KERNEL)
	/*
	* Large allocations which do not require contiguous pages
	* should be using vmem_alloc() in the linux kernel
	*/
	buf_hash_table.ht_table =
	vmem_zalloc(hsize * sizeof (void*), KM_SLEEP);
	#else
	buf_hash_table.ht_table =
	kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP);
	#endif
	if (buf_hash_table.ht_table == NULL) {
	ASSERT(hsize > (1ULL << 8));
	hsize >>= 1;
	goto retry;
	}

	hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE,
	0, hdr_full_cons, hdr_full_dest, NULL, NULL, NULL, 0);
	hdr_full_crypt_cache = kmem_cache_create("arc_buf_hdr_t_full_crypt",
	HDR_FULL_CRYPT_SIZE, 0, hdr_full_crypt_cons, hdr_full_crypt_dest,
	NULL, NULL, NULL, 0);
	hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only",
	HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, NULL,
	NULL, NULL, 0);
	buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t),
	0, buf_cons, buf_dest, NULL, NULL, NULL, 0);

	for (i = 0; i < 256; i++)
	for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--)
	ct = (ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY);

	for (i = 0; i < BUF_LOCKS; i++)
	mutex_init(BUF_HASH_LOCK(i), NULL, MUTEX_DEFAULT, NULL);
	}

	#define ARC_MINTIME (hz>>4) /* 62 ms */

	/*
	* This is the size that the buf occupies in memory. If the buf is compressed,
	* it will correspond to the compressed size. You should use this method of
	* getting the buf size unless you explicitly need the logical size.
	*/
	uint64_t
	arc_buf_size(arc_buf_t *buf)
	{
	return (ARC_BUF_COMPRESSED(buf) ?
	HDR_GET_PSIZE(buf->b_hdr) : HDR_GET_LSIZE(buf->b_hdr));
	}

	uint64_t
	arc_buf_lsize(arc_buf_t *buf)
	{
	return (HDR_GET_LSIZE(buf->b_hdr));
	}

	/*
	* This function will return B_TRUE if the buffer is encrypted in memory.
	* This buffer can be decrypted by calling arc_untransform().
	*/
	boolean_t
	arc_is_encrypted(arc_buf_t *buf)
	{
	return (ARC_BUF_ENCRYPTED(buf) != 0);
	}

	/*
	* Returns B_TRUE if the buffer represents data that has not had its MAC
	* verified yet.
	*/
	boolean_t
	arc_is_unauthenticated(arc_buf_t *buf)
	{
	return (HDR_NOAUTH(buf->b_hdr) != 0);
	}

	void
	arc_get_raw_params(arc_buf_t buf, boolean_t byteorder, uint8_t *salt,
	uint8_t iv, uint8_t mac)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(HDR_PROTECTED(hdr));

	memcpy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN);
	memcpy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN);
	memcpy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN);
	*byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
	ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
	}

	/*
	* Indicates how this buffer is compressed in memory. If it is not compressed
	* the value will be ZIO_COMPRESS_OFF. It can be made normally readable with
	* arc_untransform() as long as it is also unencrypted.
	*/
	enum zio_compress
	arc_get_compression(arc_buf_t *buf)
	{
	return (ARC_BUF_COMPRESSED(buf) ?
	HDR_GET_COMPRESS(buf->b_hdr) : ZIO_COMPRESS_OFF);
	}

	/*
	* Return the compression algorithm used to store this data in the ARC. If ARC
	* compression is enabled or this is an encrypted block, this will be the same
	* as what's used to store it on-disk. Otherwise, this will be ZIO_COMPRESS_OFF.
	*/
	static inline enum zio_compress
	arc_hdr_get_compress(arc_buf_hdr_t *hdr)
	{
	return (HDR_COMPRESSION_ENABLED(hdr) ?
	HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF);
	}

	uint8_t
	arc_get_complevel(arc_buf_t *buf)
	{
	return (buf->b_hdr->b_complevel);
	}

	static inline boolean_t
	arc_buf_is_shared(arc_buf_t *buf)
	{
	boolean_t shared = (buf->b_data != NULL &&
	buf->b_hdr->b_l1hdr.b_pabd != NULL &&
	abd_is_linear(buf->b_hdr->b_l1hdr.b_pabd) &&
	buf->b_data == abd_to_buf(buf->b_hdr->b_l1hdr.b_pabd));
	IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr));
	IMPLY(shared, ARC_BUF_SHARED(buf));
	IMPLY(shared, ARC_BUF_COMPRESSED(buf) \|\| ARC_BUF_LAST(buf));

	/*
	* It would be nice to assert arc_can_share() too, but the "hdr isn't
	* already being shared" requirement prevents us from doing that.
	*/

	return (shared);
	}

	/*
	* Free the checksum associated with this header. If there is no checksum, this
	* is a no-op.
	*/
	static inline void
	arc_cksum_free(arc_buf_hdr_t *hdr)
	{
	ASSERT(HDR_HAS_L1HDR(hdr));

	mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
	if (hdr->b_l1hdr.b_freeze_cksum != NULL) {
	kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t));
	hdr->b_l1hdr.b_freeze_cksum = NULL;
	}
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	}

	/*
	* Return true iff at least one of the bufs on hdr is not compressed.
	* Encrypted buffers count as compressed.
	*/
	static boolean_t
	arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr)
	{
	ASSERT(hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY_OR_LOCKED(hdr));

	for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) {
	if (!ARC_BUF_COMPRESSED(b)) {
	return (B_TRUE);
	}
	}
	return (B_FALSE);
	}


	/*
	* If we've turned on the ZFS_DEBUG_MODIFY flag, verify that the buf's data
	* matches the checksum that is stored in the hdr. If there is no checksum,
	* or if the buf is compressed, this is a no-op.
	*/
	static void
	arc_cksum_verify(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;
	zio_cksum_t zc;

	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(hdr));

	mutex_enter(&hdr->b_l1hdr.b_freeze_lock);

	if (hdr->b_l1hdr.b_freeze_cksum == NULL \|\| HDR_IO_ERROR(hdr)) {
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	return;
	}

	fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, &zc);
	if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc))
	panic("buffer modified while frozen!");
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	}

	/*
	* This function makes the assumption that data stored in the L2ARC
	* will be transformed exactly as it is in the main pool. Because of
	* this we can verify the checksum against the reading process's bp.
	*/
	static boolean_t
	arc_cksum_is_equal(arc_buf_hdr_t hdr, zio_t zio)
	{
	ASSERT(!BP_IS_EMBEDDED(zio->io_bp));
	VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr));

	/*
	* Block pointers always store the checksum for the logical data.
	* If the block pointer has the gang bit set, then the checksum
	* it represents is for the reconstituted data and not for an
	* individual gang member. The zio pipeline, however, must be able to
	* determine the checksum of each of the gang constituents so it
	* treats the checksum comparison differently than what we need
	* for l2arc blocks. This prevents us from using the
	* zio_checksum_error() interface directly. Instead we must call the
	* zio_checksum_error_impl() so that we can ensure the checksum is
	* generated using the correct checksum algorithm and accounts for the
	* logical I/O size and not just a gang fragment.
	*/
	return (zio_checksum_error_impl(zio->io_spa, zio->io_bp,
	BP_GET_CHECKSUM(zio->io_bp), zio->io_abd, zio->io_size,
	zio->io_offset, NULL) == 0);
	}

	/*
	* Given a buf full of data, if ZFS_DEBUG_MODIFY is enabled this computes a
	* checksum and attaches it to the buf's hdr so that we can ensure that the buf
	* isn't modified later on. If buf is compressed or there is already a checksum
	* on the hdr, this is a no-op (we only checksum uncompressed bufs).
	*/
	static void
	arc_cksum_compute(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	ASSERT(HDR_HAS_L1HDR(hdr));

	mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock);
	if (hdr->b_l1hdr.b_freeze_cksum != NULL \|\| ARC_BUF_COMPRESSED(buf)) {
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	return;
	}

	ASSERT(!ARC_BUF_ENCRYPTED(buf));
	ASSERT(!ARC_BUF_COMPRESSED(buf));
	hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t),
	KM_SLEEP);
	fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL,
	hdr->b_l1hdr.b_freeze_cksum);
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	arc_buf_watch(buf);
	}

	#ifndef _KERNEL
	void
	arc_buf_sigsegv(int sig, siginfo_t si, void unused)
	{
	(void) sig, (void) unused;
	panic("Got SIGSEGV at address: 0x%lx\n", (long)si->si_addr);
	}
	#endif

	static void
	arc_buf_unwatch(arc_buf_t *buf)
	{
	#ifndef _KERNEL
	if (arc_watch) {
	ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
	PROT_READ \| PROT_WRITE));
	}
	#else
	(void) buf;
	#endif
	}

	static void
	arc_buf_watch(arc_buf_t *buf)
	{
	#ifndef _KERNEL
	if (arc_watch)
	ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
	PROT_READ));
	#else
	(void) buf;
	#endif
	}

	static arc_buf_contents_t
	arc_buf_type(arc_buf_hdr_t *hdr)
	{
	arc_buf_contents_t type;
	if (HDR_ISTYPE_METADATA(hdr)) {
	type = ARC_BUFC_METADATA;
	} else {
	type = ARC_BUFC_DATA;
	}
	VERIFY3U(hdr->b_type, ==, type);
	return (type);
	}

	boolean_t
	arc_is_metadata(arc_buf_t *buf)
	{
	return (HDR_ISTYPE_METADATA(buf->b_hdr) != 0);
	}

	static uint32_t
	arc_bufc_to_flags(arc_buf_contents_t type)
	{
	switch (type) {
	case ARC_BUFC_DATA:
	/* metadata field is 0 if buffer contains normal data */
	return (0);
	case ARC_BUFC_METADATA:
	return (ARC_FLAG_BUFC_METADATA);
	default:
	break;
	}
	panic("undefined ARC buffer type!");
	return ((uint32_t)-1);
	}

	void
	arc_buf_thaw(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));

	arc_cksum_verify(buf);

	/*
	* Compressed buffers do not manipulate the b_freeze_cksum.
	*/
	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(hdr));
	arc_cksum_free(hdr);
	arc_buf_unwatch(buf);
	}

	void
	arc_buf_freeze(arc_buf_t *buf)
	{
	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(buf->b_hdr));
	arc_cksum_compute(buf);
	}

	/*
	* The arc_buf_hdr_t's b_flags should never be modified directly. Instead,
	* the following functions should be used to ensure that the flags are
	* updated in a thread-safe way. When manipulating the flags either
	* the hash_lock must be held or the hdr must be undiscoverable. This
	* ensures that we're not racing with any other threads when updating
	* the flags.
	*/
	static inline void
	arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	hdr->b_flags \|= flags;
	}

	static inline void
	arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	hdr->b_flags &= ~flags;
	}

	/*
	* Setting the compression bits in the arc_buf_hdr_t's b_flags is
	* done in a special way since we have to clear and set bits
	* at the same time. Consumers that wish to set the compression bits
	* must use this function to ensure that the flags are updated in
	* thread-safe manner.
	*/
	static void
	arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Holes and embedded blocks will always have a psize = 0 so
	* we ignore the compression of the blkptr and set the
	* want to uncompress them. Mark them as uncompressed.
	*/
	if (!zfs_compressed_arc_enabled \|\| HDR_GET_PSIZE(hdr) == 0) {
	arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
	ASSERT(!HDR_COMPRESSION_ENABLED(hdr));
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
	ASSERT(HDR_COMPRESSION_ENABLED(hdr));
	}

	HDR_SET_COMPRESS(hdr, cmp);
	ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp);
	}

	/*
	* Looks for another buf on the same hdr which has the data decompressed, copies
	* from it, and returns true. If no such buf exists, returns false.
	*/
	static boolean_t
	arc_buf_try_copy_decompressed_data(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;
	boolean_t copied = B_FALSE;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(!ARC_BUF_COMPRESSED(buf));

	for (arc_buf_t *from = hdr->b_l1hdr.b_buf; from != NULL;
	from = from->b_next) {
	/* can't use our own data buffer */
	if (from == buf) {
	continue;
	}

	if (!ARC_BUF_COMPRESSED(from)) {
	memcpy(buf->b_data, from->b_data, arc_buf_size(buf));
	copied = B_TRUE;
	break;
	}
	}

	/*
	* There were no decompressed bufs, so there should not be a
	* checksum on the hdr either.
	*/
	if (zfs_flags & ZFS_DEBUG_MODIFY)
	EQUIV(!copied, hdr->b_l1hdr.b_freeze_cksum == NULL);

	return (copied);
	}

	/*
	* Allocates an ARC buf header that's in an evicted & L2-cached state.
	* This is used during l2arc reconstruction to make empty ARC buffers
	* which circumvent the regular disk->arc->l2arc path and instead come
	* into being in the reverse order, i.e. l2arc->arc.
	*/
	static arc_buf_hdr_t *
	arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev,
	dva_t dva, uint64_t daddr, int32_t psize, uint64_t birth,
	enum zio_compress compress, uint8_t complevel, boolean_t protected,
	boolean_t prefetch, arc_state_type_t arcs_state)
	{
	arc_buf_hdr_t *hdr;

	ASSERT(size != 0);
	hdr = kmem_cache_alloc(hdr_l2only_cache, KM_SLEEP);
	hdr->b_birth = birth;
	hdr->b_type = type;
	hdr->b_flags = 0;
	arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) \| ARC_FLAG_HAS_L2HDR);
	HDR_SET_LSIZE(hdr, size);
	HDR_SET_PSIZE(hdr, psize);
	arc_hdr_set_compress(hdr, compress);
	hdr->b_complevel = complevel;
	if (protected)
	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
	if (prefetch)
	arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
	hdr->b_spa = spa_load_guid(dev->l2ad_vdev->vdev_spa);

	hdr->b_dva = dva;

	hdr->b_l2hdr.b_dev = dev;
	hdr->b_l2hdr.b_daddr = daddr;
	hdr->b_l2hdr.b_arcs_state = arcs_state;

	return (hdr);
	}

	/*
	* Return the size of the block, b_pabd, that is stored in the arc_buf_hdr_t.
	*/
	static uint64_t
	arc_hdr_size(arc_buf_hdr_t *hdr)
	{
	uint64_t size;

	if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
	HDR_GET_PSIZE(hdr) > 0) {
	size = HDR_GET_PSIZE(hdr);
	} else {
	ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0);
	size = HDR_GET_LSIZE(hdr);
	}
	return (size);
	}

	static int
	arc_hdr_authenticate(arc_buf_hdr_t hdr, spa_t spa, uint64_t dsobj)
	{
	int ret;
	uint64_t csize;
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	void *tmpbuf = NULL;
	abd_t *abd = hdr->b_l1hdr.b_pabd;

	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT(HDR_AUTHENTICATED(hdr));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	/*
	* The MAC is calculated on the compressed data that is stored on disk.
	* However, if compressed arc is disabled we will only have the
	* decompressed data available to us now. Compress it into a temporary
	* abd so we can verify the MAC. The performance overhead of this will
	* be relatively low, since most objects in an encrypted objset will
	* be encrypted (instead of authenticated) anyway.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {
	tmpbuf = zio_buf_alloc(lsize);
	abd = abd_get_from_buf(tmpbuf, lsize);
	abd_take_ownership_of_buf(abd, B_TRUE);
	csize = zio_compress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, tmpbuf, lsize, hdr->b_complevel);
	ASSERT3U(csize, <=, psize);
	abd_zero_off(abd, csize, psize - csize);
	}

	/*
	* Authentication is best effort. We authenticate whenever the key is
	* available. If we succeed we clear ARC_FLAG_NOAUTH.
	*/
	if (hdr->b_crypt_hdr.b_ot == DMU_OT_OBJSET) {
	ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
	ASSERT3U(lsize, ==, psize);
	ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, abd,
	psize, hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
	} else {
	ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, abd, psize,
	hdr->b_crypt_hdr.b_mac);
	}

	if (ret == 0)
	arc_hdr_clear_flags(hdr, ARC_FLAG_NOAUTH);
	else if (ret != ENOENT)
	goto error;

	if (tmpbuf != NULL)
	abd_free(abd);

	return (0);

	error:
	if (tmpbuf != NULL)
	abd_free(abd);

	return (ret);
	}

	/*
	* This function will take a header that only has raw encrypted data in
	* b_crypt_hdr.b_rabd and decrypt it into a new buffer which is stored in
	* b_l1hdr.b_pabd. If designated in the header flags, this function will
	* also decompress the data.
	*/
	static int
	arc_hdr_decrypt(arc_buf_hdr_t hdr, spa_t spa, const zbookmark_phys_t *zb)
	{
	int ret;
	abd_t *cabd = NULL;
	void *tmp = NULL;
	boolean_t no_crypt = B_FALSE;
	boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);

	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT(HDR_ENCRYPTED(hdr));

	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT);

	ret = spa_do_crypt_abd(B_FALSE, spa, zb, hdr->b_crypt_hdr.b_ot,
	B_FALSE, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv,
	hdr->b_crypt_hdr.b_mac, HDR_GET_PSIZE(hdr), hdr->b_l1hdr.b_pabd,
	hdr->b_crypt_hdr.b_rabd, &no_crypt);
	if (ret != 0)
	goto error;

	if (no_crypt) {
	abd_copy(hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd,
	HDR_GET_PSIZE(hdr));
	}

	/*
	* If this header has disabled arc compression but the b_pabd is
	* compressed after decrypting it, we need to decompress the newly
	* decrypted data.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {
	/*
	* We want to make sure that we are correctly honoring the
	* zfs_abd_scatter_enabled setting, so we allocate an abd here
	* and then loan a buffer from it, rather than allocating a
	* linear buffer and wrapping it in an abd later.
	*/
	cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
	ARC_HDR_DO_ADAPT);
	tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));

	ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
	HDR_GET_LSIZE(hdr), &hdr->b_complevel);
	if (ret != 0) {
	abd_return_buf(cabd, tmp, arc_hdr_size(hdr));
	goto error;
	}

	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = cabd;
	}

	return (0);

	error:
	arc_hdr_free_abd(hdr, B_FALSE);
	if (cabd != NULL)
	arc_free_data_buf(hdr, cabd, arc_hdr_size(hdr), hdr);

	return (ret);
	}

	/*
	* This function is called during arc_buf_fill() to prepare the header's
	* abd plaintext pointer for use. This involves authenticated protected
	* data and decrypting encrypted data into the plaintext abd.
	*/
	static int
	arc_fill_hdr_crypt(arc_buf_hdr_t hdr, kmutex_t hash_lock, spa_t *spa,
	const zbookmark_phys_t *zb, boolean_t noauth)
	{
	int ret;

	ASSERT(HDR_PROTECTED(hdr));

	if (hash_lock != NULL)
	mutex_enter(hash_lock);

	if (HDR_NOAUTH(hdr) && !noauth) {
	/*
	* The caller requested authenticated data but our data has
	* not been authenticated yet. Verify the MAC now if we can.
	*/
	ret = arc_hdr_authenticate(hdr, spa, zb->zb_objset);
	if (ret != 0)
	goto error;
	} else if (HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd == NULL) {
	/*
	* If we only have the encrypted version of the data, but the
	* unencrypted version was requested we take this opportunity
	* to store the decrypted version in the header for future use.
	*/
	ret = arc_hdr_decrypt(hdr, spa, zb);
	if (ret != 0)
	goto error;
	}

	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	return (0);

	error:
	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	return (ret);
	}

	/*
	* This function is used by the dbuf code to decrypt bonus buffers in place.
	* The dbuf code itself doesn't have any locking for decrypting a shared dnode
	* block, so we use the hash lock here to protect against concurrent calls to
	* arc_buf_fill().
	*/
	static void
	arc_buf_untransform_in_place(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(HDR_ENCRYPTED(hdr));
	ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data,
	arc_buf_size(buf));
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	hdr->b_crypt_hdr.b_ebufcnt -= 1;
	}

	/*
	* Given a buf that has a data buffer attached to it, this function will
	* efficiently fill the buf with data of the specified compression setting from
	* the hdr and update the hdr's b_freeze_cksum if necessary. If the buf and hdr
	* are already sharing a data buf, no copy is performed.
	*
	* If the buf is marked as compressed but uncompressed data was requested, this
	* will allocate a new data buffer for the buf, remove that flag, and fill the
	* buf with uncompressed data. You can't request a compressed buf on a hdr with
	* uncompressed data, and (since we haven't added support for it yet) if you
	* want compressed data your buf must already be marked as compressed and have
	* the correct-sized data buffer.
	*/
	static int
	arc_buf_fill(arc_buf_t buf, spa_t spa, const zbookmark_phys_t *zb,
	arc_fill_flags_t flags)
	{
	int error = 0;
	arc_buf_hdr_t *hdr = buf->b_hdr;
	boolean_t hdr_compressed =
	(arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	boolean_t compressed = (flags & ARC_FILL_COMPRESSED) != 0;
	boolean_t encrypted = (flags & ARC_FILL_ENCRYPTED) != 0;
	dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap;
	kmutex_t *hash_lock = (flags & ARC_FILL_LOCKED) ? NULL : HDR_LOCK(hdr);

	ASSERT3P(buf->b_data, !=, NULL);
	IMPLY(compressed, hdr_compressed \|\| ARC_BUF_ENCRYPTED(buf));
	IMPLY(compressed, ARC_BUF_COMPRESSED(buf));
	IMPLY(encrypted, HDR_ENCRYPTED(hdr));
	IMPLY(encrypted, ARC_BUF_ENCRYPTED(buf));
	IMPLY(encrypted, ARC_BUF_COMPRESSED(buf));
	IMPLY(encrypted, !ARC_BUF_SHARED(buf));

	/*
	* If the caller wanted encrypted data we just need to copy it from
	* b_rabd and potentially byteswap it. We won't be able to do any
	* further transforms on it.
	*/
	if (encrypted) {
	ASSERT(HDR_HAS_RABD(hdr));
	abd_copy_to_buf(buf->b_data, hdr->b_crypt_hdr.b_rabd,
	HDR_GET_PSIZE(hdr));
	goto byteswap;
	}

	/*
	* Adjust encrypted and authenticated headers to accommodate
	* the request if needed. Dnode blocks (ARC_FILL_IN_PLACE) are
	* allowed to fail decryption due to keys not being loaded
	* without being marked as an IO error.
	*/
	if (HDR_PROTECTED(hdr)) {
	error = arc_fill_hdr_crypt(hdr, hash_lock, spa,
	zb, !!(flags & ARC_FILL_NOAUTH));
	if (error == EACCES && (flags & ARC_FILL_IN_PLACE) != 0) {
	return (error);
	} else if (error != 0) {
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	return (error);
	}
	}

	/*
	* There is a special case here for dnode blocks which are
	* decrypting their bonus buffers. These blocks may request to
	* be decrypted in-place. This is necessary because there may
	* be many dnodes pointing into this buffer and there is
	* currently no method to synchronize replacing the backing
	* b_data buffer and updating all of the pointers. Here we use
	* the hash lock to ensure there are no races. If the need
	* arises for other types to be decrypted in-place, they must
	* add handling here as well.
	*/
	if ((flags & ARC_FILL_IN_PLACE) != 0) {
	ASSERT(!hdr_compressed);
	ASSERT(!compressed);
	ASSERT(!encrypted);

	if (HDR_ENCRYPTED(hdr) && ARC_BUF_ENCRYPTED(buf)) {
	ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);

	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_buf_untransform_in_place(buf);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	/* Compute the hdr's checksum if necessary */
	arc_cksum_compute(buf);
	}

	return (0);
	}

	if (hdr_compressed == compressed) {
	if (!arc_buf_is_shared(buf)) {
	abd_copy_to_buf(buf->b_data, hdr->b_l1hdr.b_pabd,
	arc_buf_size(buf));
	}
	} else {
	ASSERT(hdr_compressed);
	ASSERT(!compressed);

	/*
	* If the buf is sharing its data with the hdr, unlink it and
	* allocate a new data buffer for the buf.
	*/
	if (arc_buf_is_shared(buf)) {
	ASSERT(ARC_BUF_COMPRESSED(buf));

	/* We need to give the buf its own b_data */
	buf->b_flags &= ~ARC_BUF_FLAG_SHARED;
	buf->b_data =
	arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);

	/* Previously overhead was 0; just add new overhead */
	ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
	} else if (ARC_BUF_COMPRESSED(buf)) {
	/* We need to reallocate the buf's b_data */
	arc_free_data_buf(hdr, buf->b_data, HDR_GET_PSIZE(hdr),
	buf);
	buf->b_data =
	arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);

	/* We increased the size of b_data; update overhead */
	ARCSTAT_INCR(arcstat_overhead_size,
	HDR_GET_LSIZE(hdr) - HDR_GET_PSIZE(hdr));
	}

	/*
	* Regardless of the buf's previous compression settings, it
	* should not be compressed at the end of this function.
	*/
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;

	/*
	* Try copying the data from another buf which already has a
	* decompressed version. If that's not possible, it's time to
	* bite the bullet and decompress the data from the hdr.
	*/
	if (arc_buf_try_copy_decompressed_data(buf)) {
	/* Skip byteswapping and checksumming (already done) */
	return (0);
	} else {
	error = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, buf->b_data,
	HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr),
	&hdr->b_complevel);

	/*
	* Absent hardware errors or software bugs, this should
	* be impossible, but log it anyway so we can debug it.
	*/
	if (error != 0) {
	zfs_dbgmsg(
	"hdr %px, compress %d, psize %d, lsize %d",
	hdr, arc_hdr_get_compress(hdr),
	HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr));
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	return (SET_ERROR(EIO));
	}
	}
	}

	byteswap:
	/* Byteswap the buf's data if necessary */
	if (bswap != DMU_BSWAP_NUMFUNCS) {
	ASSERT(!HDR_SHARED_DATA(hdr));
	ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS);
	dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr));
	}

	/* Compute the hdr's checksum if necessary */
	arc_cksum_compute(buf);

	return (0);
	}

	/*
	* If this function is being called to decrypt an encrypted buffer or verify an
	* authenticated one, the key must be loaded and a mapping must be made
	* available in the keystore via spa_keystore_create_mapping() or one of its
	* callers.
	*/
	int
	arc_untransform(arc_buf_t buf, spa_t spa, const zbookmark_phys_t *zb,
	boolean_t in_place)
	{
	int ret;
	arc_fill_flags_t flags = 0;

	if (in_place)
	flags \|= ARC_FILL_IN_PLACE;

	ret = arc_buf_fill(buf, spa, zb, flags);
	if (ret == ECKSUM) {
	/*
	* Convert authentication and decryption errors to EIO
	* (and generate an ereport) before leaving the ARC.
	*/
	ret = SET_ERROR(EIO);
	spa_log_error(spa, zb);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, zb, NULL, 0);
	}

	return (ret);
	}

	/*
	* Increment the amount of evictable space in the arc_state_t's refcount.
	* We account for the space used by the hdr and the arc buf individually
	* so that we can add and remove them from the refcount individually.
	*/
	static void
	arc_evictable_space_increment(arc_buf_hdr_t hdr, arc_state_t state)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));

	if (GHOST_STATE(state)) {
	ASSERT0(hdr->b_l1hdr.b_bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	HDR_GET_LSIZE(hdr), hdr);
	return;
	}

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	arc_hdr_size(hdr), hdr);
	}
	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	HDR_GET_PSIZE(hdr), hdr);
	}

	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	if (arc_buf_is_shared(buf))
	continue;
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}
	}

	/*
	* Decrement the amount of evictable space in the arc_state_t's refcount.
	* We account for the space used by the hdr and the arc buf individually
	* so that we can add and remove them from the refcount individually.
	*/
	static void
	arc_evictable_space_decrement(arc_buf_hdr_t hdr, arc_state_t state)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));

	if (GHOST_STATE(state)) {
	ASSERT0(hdr->b_l1hdr.b_bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	HDR_GET_LSIZE(hdr), hdr);
	return;
	}

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	arc_hdr_size(hdr), hdr);
	}
	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	HDR_GET_PSIZE(hdr), hdr);
	}

	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	if (arc_buf_is_shared(buf))
	continue;
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}
	}

	/*
	* Add a reference to this hdr indicating that someone is actively
	* referencing that memory. When the refcount transitions from 0 to 1,
	* we remove it from the respective arc_state_t list to indicate that
	* it is not evictable.
	*/
	static void
	add_reference(arc_buf_hdr_t hdr, const void tag)
	{
	arc_state_t *state;

	ASSERT(HDR_HAS_L1HDR(hdr));
	if (!HDR_EMPTY(hdr) && !MUTEX_HELD(HDR_LOCK(hdr))) {
	ASSERT(hdr->b_l1hdr.b_state == arc_anon);
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	}

	state = hdr->b_l1hdr.b_state;

	if ((zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) &&
	(state != arc_anon)) {
	/* We don't use the L2-only state list. */
	if (state != arc_l2c_only) {
	multilist_remove(&state->arcs_list[arc_buf_type(hdr)],
	hdr);
	arc_evictable_space_decrement(hdr, state);
	}
	/* remove the prefetch flag if we get a reference */
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH);
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	}

	/*
	* Remove a reference from this hdr. When the reference transitions from
	* 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's
	* list making it eligible for eviction.
	*/
	static int
	remove_reference(arc_buf_hdr_t hdr, kmutex_t hash_lock, const void *tag)
	{
	int cnt;
	arc_state_t *state = hdr->b_l1hdr.b_state;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(state == arc_anon \|\| MUTEX_HELD(hash_lock));
	ASSERT(!GHOST_STATE(state));

	/*
	* arc_l2c_only counts as a ghost state so we don't need to explicitly
	* check to prevent usage of the arc_l2c_only list.
	*/
	if (((cnt = zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) == 0) &&
	(state != arc_anon)) {
	multilist_insert(&state->arcs_list[arc_buf_type(hdr)], hdr);
	ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
	arc_evictable_space_increment(hdr, state);
	}
	return (cnt);
	}

	/*
	* Returns detailed information about a specific arc buffer. When the
	* state_index argument is set the function will calculate the arc header
	* list position for its arc state. Since this requires a linear traversal
	* callers are strongly encourage not to do this. However, it can be helpful
	* for targeted analysis so the functionality is provided.
	*/
	void
	arc_buf_info(arc_buf_t ab, arc_buf_info_t abi, int state_index)
	{
	(void) state_index;
	arc_buf_hdr_t *hdr = ab->b_hdr;
	l1arc_buf_hdr_t *l1hdr = NULL;
	l2arc_buf_hdr_t *l2hdr = NULL;
	arc_state_t *state = NULL;

	memset(abi, 0, sizeof (arc_buf_info_t));

	if (hdr == NULL)
	return;

	abi->abi_flags = hdr->b_flags;

	if (HDR_HAS_L1HDR(hdr)) {
	l1hdr = &hdr->b_l1hdr;
	state = l1hdr->b_state;
	}
	if (HDR_HAS_L2HDR(hdr))
	l2hdr = &hdr->b_l2hdr;

	if (l1hdr) {
	abi->abi_bufcnt = l1hdr->b_bufcnt;
	abi->abi_access = l1hdr->b_arc_access;
	abi->abi_mru_hits = l1hdr->b_mru_hits;
	abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits;
	abi->abi_mfu_hits = l1hdr->b_mfu_hits;
	abi->abi_mfu_ghost_hits = l1hdr->b_mfu_ghost_hits;
	abi->abi_holds = zfs_refcount_count(&l1hdr->b_refcnt);
	}

	if (l2hdr) {
	abi->abi_l2arc_dattr = l2hdr->b_daddr;
	abi->abi_l2arc_hits = l2hdr->b_hits;
	}

	abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON;
	abi->abi_state_contents = arc_buf_type(hdr);
	abi->abi_size = arc_hdr_size(hdr);
	}

	/*
	* Move the supplied buffer to the indicated state. The hash lock
	* for the buffer must be held by the caller.
	*/
	static void
	arc_change_state(arc_state_t new_state, arc_buf_hdr_t hdr,
	kmutex_t *hash_lock)
	{
	arc_state_t *old_state;
	int64_t refcnt;
	uint32_t bufcnt;
	boolean_t update_old, update_new;
	arc_buf_contents_t buftype = arc_buf_type(hdr);

	/*
	* We almost always have an L1 hdr here, since we call arc_hdr_realloc()
	* in arc_read() when bringing a buffer out of the L2ARC. However, the
	* L1 hdr doesn't always exist when we change state to arc_anon before
	* destroying a header, in which case reallocating to add the L1 hdr is
	* pointless.
	*/
	if (HDR_HAS_L1HDR(hdr)) {
	old_state = hdr->b_l1hdr.b_state;
	refcnt = zfs_refcount_count(&hdr->b_l1hdr.b_refcnt);
	bufcnt = hdr->b_l1hdr.b_bufcnt;
	update_old = (bufcnt > 0 \|\| hdr->b_l1hdr.b_pabd != NULL \|\|
	HDR_HAS_RABD(hdr));
	} else {
	old_state = arc_l2c_only;
	refcnt = 0;
	bufcnt = 0;
	update_old = B_FALSE;
	}
	update_new = update_old;

	ASSERT(MUTEX_HELD(hash_lock));
	ASSERT3P(new_state, !=, old_state);
	ASSERT(!GHOST_STATE(new_state) \|\| bufcnt == 0);
	ASSERT(old_state != arc_anon \|\| bufcnt <= 1);

	/*
	* If this buffer is evictable, transfer it from the
	* old state list to the new state list.
	*/
	if (refcnt == 0) {
	if (old_state != arc_anon && old_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	multilist_remove(&old_state->arcs_list[buftype], hdr);

	if (GHOST_STATE(old_state)) {
	ASSERT0(bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	update_old = B_TRUE;
	}
	arc_evictable_space_decrement(hdr, old_state);
	}
	if (new_state != arc_anon && new_state != arc_l2c_only) {
	/*
	* An L1 header always exists here, since if we're
	* moving to some L1-cached state (i.e. not l2c_only or
	* anonymous), we realloc the header to add an L1hdr
	* beforehand.
	*/
	ASSERT(HDR_HAS_L1HDR(hdr));
	multilist_insert(&new_state->arcs_list[buftype], hdr);

	if (GHOST_STATE(new_state)) {
	ASSERT0(bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	update_new = B_TRUE;
	}
	arc_evictable_space_increment(hdr, new_state);
	}
	}

	ASSERT(!HDR_EMPTY(hdr));
	if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr))
	buf_hash_remove(hdr);

	/* adjust state sizes (ignore arc_l2c_only) */

	if (update_new && new_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	if (GHOST_STATE(new_state)) {
	ASSERT0(bufcnt);

	/*
	* When moving a header to a ghost state, we first
	* remove all arc buffers. Thus, we'll have a
	* bufcnt of zero, and no arc buffer to use for
	* the reference. As a result, we use the arc
	* header pointer for the reference.
	*/
	(void) zfs_refcount_add_many(&new_state->arcs_size,
	HDR_GET_LSIZE(hdr), hdr);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	} else {
	uint32_t buffers = 0;

	/*
	* Each individual buffer holds a unique reference,
	* thus we must remove each of these references one
	* at a time.
	*/
	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	ASSERT3U(bufcnt, !=, 0);
	buffers++;

	/*
	* When the arc_buf_t is sharing the data
	* block with the hdr, the owner of the
	* reference belongs to the hdr. Only
	* add to the refcount if the arc_buf_t is
	* not shared.
	*/
	if (arc_buf_is_shared(buf))
	continue;

	(void) zfs_refcount_add_many(
	&new_state->arcs_size,
	arc_buf_size(buf), buf);
	}
	ASSERT3U(bufcnt, ==, buffers);

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_add_many(
	&new_state->arcs_size,
	arc_hdr_size(hdr), hdr);
	}

	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_add_many(
	&new_state->arcs_size,
	HDR_GET_PSIZE(hdr), hdr);
	}
	}
	}

	if (update_old && old_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	if (GHOST_STATE(old_state)) {
	ASSERT0(bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));

	/*
	* When moving a header off of a ghost state,
	* the header will not contain any arc buffers.
	* We use the arc header pointer for the reference
	* which is exactly what we did when we put the
	* header on the ghost state.
	*/

	(void) zfs_refcount_remove_many(&old_state->arcs_size,
	HDR_GET_LSIZE(hdr), hdr);
	} else {
	uint32_t buffers = 0;

	/*
	* Each individual buffer holds a unique reference,
	* thus we must remove each of these references one
	* at a time.
	*/
	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	ASSERT3U(bufcnt, !=, 0);
	buffers++;

	/*
	* When the arc_buf_t is sharing the data
	* block with the hdr, the owner of the
	* reference belongs to the hdr. Only
	* add to the refcount if the arc_buf_t is
	* not shared.
	*/
	if (arc_buf_is_shared(buf))
	continue;

	(void) zfs_refcount_remove_many(
	&old_state->arcs_size, arc_buf_size(buf),
	buf);
	}
	ASSERT3U(bufcnt, ==, buffers);
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\|
	HDR_HAS_RABD(hdr));

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_remove_many(
	&old_state->arcs_size, arc_hdr_size(hdr),
	hdr);
	}

	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_remove_many(
	&old_state->arcs_size, HDR_GET_PSIZE(hdr),
	hdr);
	}
	}
	}

	if (HDR_HAS_L1HDR(hdr)) {
	hdr->b_l1hdr.b_state = new_state;

	if (HDR_HAS_L2HDR(hdr) && new_state != arc_l2c_only) {
	l2arc_hdr_arcstats_decrement_state(hdr);
	hdr->b_l2hdr.b_arcs_state = new_state->arcs_state;
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	}
	}

	void
	arc_space_consume(uint64_t space, arc_space_type_t type)
	{
	ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);

	switch (type) {
	default:
	break;
	case ARC_SPACE_DATA:
	ARCSTAT_INCR(arcstat_data_size, space);
	break;
	case ARC_SPACE_META:
	ARCSTAT_INCR(arcstat_metadata_size, space);
	break;
	case ARC_SPACE_BONUS:
	ARCSTAT_INCR(arcstat_bonus_size, space);
	break;
	case ARC_SPACE_DNODE:
	aggsum_add(&arc_sums.arcstat_dnode_size, space);
	break;
	case ARC_SPACE_DBUF:
	ARCSTAT_INCR(arcstat_dbuf_size, space);
	break;
	case ARC_SPACE_HDRS:
	ARCSTAT_INCR(arcstat_hdr_size, space);
	break;
	case ARC_SPACE_L2HDRS:
	aggsum_add(&arc_sums.arcstat_l2_hdr_size, space);
	break;
	case ARC_SPACE_ABD_CHUNK_WASTE:
	/*
	* Note: this includes space wasted by all scatter ABD's, not
	* just those allocated by the ARC. But the vast majority of
	* scatter ABD's come from the ARC, because other users are
	* very short-lived.
	*/
	ARCSTAT_INCR(arcstat_abd_chunk_waste_size, space);
	break;
	}

	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE)
	aggsum_add(&arc_sums.arcstat_meta_used, space);

	aggsum_add(&arc_sums.arcstat_size, space);
	}

	void
	arc_space_return(uint64_t space, arc_space_type_t type)
	{
	ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);

	switch (type) {
	default:
	break;
	case ARC_SPACE_DATA:
	ARCSTAT_INCR(arcstat_data_size, -space);
	break;
	case ARC_SPACE_META:
	ARCSTAT_INCR(arcstat_metadata_size, -space);
	break;
	case ARC_SPACE_BONUS:
	ARCSTAT_INCR(arcstat_bonus_size, -space);
	break;
	case ARC_SPACE_DNODE:
	aggsum_add(&arc_sums.arcstat_dnode_size, -space);
	break;
	case ARC_SPACE_DBUF:
	ARCSTAT_INCR(arcstat_dbuf_size, -space);
	break;
	case ARC_SPACE_HDRS:
	ARCSTAT_INCR(arcstat_hdr_size, -space);
	break;
	case ARC_SPACE_L2HDRS:
	aggsum_add(&arc_sums.arcstat_l2_hdr_size, -space);
	break;
	case ARC_SPACE_ABD_CHUNK_WASTE:
	ARCSTAT_INCR(arcstat_abd_chunk_waste_size, -space);
	break;
	}

	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE) {
	ASSERT(aggsum_compare(&arc_sums.arcstat_meta_used,
	space) >= 0);
	ARCSTAT_MAX(arcstat_meta_max,
	aggsum_upper_bound(&arc_sums.arcstat_meta_used));
	aggsum_add(&arc_sums.arcstat_meta_used, -space);
	}

	ASSERT(aggsum_compare(&arc_sums.arcstat_size, space) >= 0);
	aggsum_add(&arc_sums.arcstat_size, -space);
	}

	/*
	* Given a hdr and a buf, returns whether that buf can share its b_data buffer
	* with the hdr's b_pabd.
	*/
	static boolean_t
	arc_can_share(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	/*
	* The criteria for sharing a hdr's data are:
	* 1. the buffer is not encrypted
	* 2. the hdr's compression matches the buf's compression
	* 3. the hdr doesn't need to be byteswapped
	* 4. the hdr isn't already being shared
	* 5. the buf is either compressed or it is the last buf in the hdr list
	*
	* Criterion #5 maintains the invariant that shared uncompressed
	* bufs must be the final buf in the hdr's b_buf list. Reading this, you
	* might ask, "if a compressed buf is allocated first, won't that be the
	* last thing in the list?", but in that case it's impossible to create
	* a shared uncompressed buf anyway (because the hdr must be compressed
	* to have the compressed buf). You might also think that #3 is
	* sufficient to make this guarantee, however it's possible
	* (specifically in the rare L2ARC write race mentioned in
	* arc_buf_alloc_impl()) there will be an existing uncompressed buf that
	* is shareable, but wasn't at the time of its allocation. Rather than
	* allow a new shared uncompressed buf to be created and then shuffle
	* the list around to make it the last element, this simply disallows
	* sharing if the new buf isn't the first to be added.
	*/
	ASSERT3P(buf->b_hdr, ==, hdr);
	boolean_t hdr_compressed =
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF;
	boolean_t buf_compressed = ARC_BUF_COMPRESSED(buf) != 0;
	return (!ARC_BUF_ENCRYPTED(buf) &&
	buf_compressed == hdr_compressed &&
	hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS &&
	!HDR_SHARED_DATA(hdr) &&
	(ARC_BUF_LAST(buf) \|\| ARC_BUF_COMPRESSED(buf)));
	}

	/*
	* Allocate a buf for this hdr. If you care about the data that's in the hdr,
	* or if you want a compressed buffer, pass those flags in. Returns 0 if the
	* copy was made successfully, or an error code otherwise.
	*/
	static int
	arc_buf_alloc_impl(arc_buf_hdr_t hdr, spa_t spa, const zbookmark_phys_t *zb,
	const void *tag, boolean_t encrypted, boolean_t compressed,
	boolean_t noauth, boolean_t fill, arc_buf_t **ret)
	{
	arc_buf_t *buf;
	arc_fill_flags_t flags = ARC_FILL_LOCKED;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
	VERIFY(hdr->b_type == ARC_BUFC_DATA \|\|
	hdr->b_type == ARC_BUFC_METADATA);
	ASSERT3P(ret, !=, NULL);
	ASSERT3P(*ret, ==, NULL);
	IMPLY(encrypted, compressed);

	buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
	buf->b_hdr = hdr;
	buf->b_data = NULL;
	buf->b_next = hdr->b_l1hdr.b_buf;
	buf->b_flags = 0;

	add_reference(hdr, tag);

	/*
	* We're about to change the hdr's b_flags. We must either
	* hold the hash_lock or be undiscoverable.
	*/
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Only honor requests for compressed bufs if the hdr is actually
	* compressed. This must be overridden if the buffer is encrypted since
	* encrypted buffers cannot be decompressed.
	*/
	if (encrypted) {
	buf->b_flags \|= ARC_BUF_FLAG_COMPRESSED;
	buf->b_flags \|= ARC_BUF_FLAG_ENCRYPTED;
	flags \|= ARC_FILL_COMPRESSED \| ARC_FILL_ENCRYPTED;
	} else if (compressed &&
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
	buf->b_flags \|= ARC_BUF_FLAG_COMPRESSED;
	flags \|= ARC_FILL_COMPRESSED;
	}

	if (noauth) {
	ASSERT0(encrypted);
	flags \|= ARC_FILL_NOAUTH;
	}

	/*
	* If the hdr's data can be shared then we share the data buffer and
	* set the appropriate bit in the hdr's b_flags to indicate the hdr is
	* sharing it's b_pabd with the arc_buf_t. Otherwise, we allocate a new
	* buffer to store the buf's data.
	*
	* There are two additional restrictions here because we're sharing
	* hdr -> buf instead of the usual buf -> hdr. First, the hdr can't be
	* actively involved in an L2ARC write, because if this buf is used by
	* an arc_write() then the hdr's data buffer will be released when the
	* write completes, even though the L2ARC write might still be using it.
	* Second, the hdr's ABD must be linear so that the buf's user doesn't
	* need to be ABD-aware. It must be allocated via
	* zio_[data_]buf_alloc(), not as a page, because we need to be able
	* to abd_release_ownership_of_buf(), which isn't allowed on "linear
	* page" buffers because the ABD code needs to handle freeing them
	* specially.
	*/
	boolean_t can_share = arc_can_share(hdr, buf) &&
	!HDR_L2_WRITING(hdr) &&
	hdr->b_l1hdr.b_pabd != NULL &&
	abd_is_linear(hdr->b_l1hdr.b_pabd) &&
	!abd_is_linear_page(hdr->b_l1hdr.b_pabd);

	/* Set up b_data and sharing */
	if (can_share) {
	buf->b_data = abd_to_buf(hdr->b_l1hdr.b_pabd);
	buf->b_flags \|= ARC_BUF_FLAG_SHARED;
	arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
	} else {
	buf->b_data =
	arc_get_data_buf(hdr, arc_buf_size(buf), buf);
	ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
	}
	VERIFY3P(buf->b_data, !=, NULL);

	hdr->b_l1hdr.b_buf = buf;
	hdr->b_l1hdr.b_bufcnt += 1;
	if (encrypted)
	hdr->b_crypt_hdr.b_ebufcnt += 1;

	/*
	* If the user wants the data from the hdr, we need to either copy or
	* decompress the data.
	*/
	if (fill) {
	ASSERT3P(zb, !=, NULL);
	return (arc_buf_fill(buf, spa, zb, flags));
	}

	return (0);
	}

	static const char *arc_onloan_tag = "onloan";

	static inline void
	arc_loaned_bytes_update(int64_t delta)
	{
	atomic_add_64(&arc_loaned_bytes, delta);

	/* assert that it did not wrap around */
	ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);
	}

	/*
	* Loan out an anonymous arc buffer. Loaned buffers are not counted as in
	* flight data by arc_tempreserve_space() until they are "returned". Loaned
	* buffers must be returned to the arc before they can be used by the DMU or
	* freed.
	*/
	arc_buf_t *
	arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size)
	{
	arc_buf_t *buf = arc_alloc_buf(spa, arc_onloan_tag,
	is_metadata ? ARC_BUFC_METADATA : ARC_BUFC_DATA, size);

	arc_loaned_bytes_update(arc_buf_size(buf));

	return (buf);
	}

	arc_buf_t *
	arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag,
	psize, lsize, compression_type, complevel);

	arc_loaned_bytes_update(arc_buf_size(buf));

	return (buf);
	}

	arc_buf_t *
	arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t *mac,
	dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj,
	byteorder, salt, iv, mac, ot, psize, lsize, compression_type,
	complevel);

	atomic_add_64(&arc_loaned_bytes, psize);
	return (buf);
	}


	/*
	* Return a loaned arc buffer to the arc.
	*/
	void
	arc_return_buf(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(HDR_HAS_L1HDR(hdr));
	(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag);
	(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);

	arc_loaned_bytes_update(-arc_buf_size(buf));
	}

	/* Detach an arc_buf from a dbuf (tag) */
	void
	arc_loan_inuse_buf(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(HDR_HAS_L1HDR(hdr));
	(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
	(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag);

	arc_loaned_bytes_update(arc_buf_size(buf));
	}

	static void
	l2arc_free_abd_on_write(abd_t *abd, size_t size, arc_buf_contents_t type)
	{
	l2arc_data_free_t df = kmem_alloc(sizeof (df), KM_SLEEP);

	df->l2df_abd = abd;
	df->l2df_size = size;
	df->l2df_type = type;
	mutex_enter(&l2arc_free_on_write_mtx);
	list_insert_head(l2arc_free_on_write, df);
	mutex_exit(&l2arc_free_on_write_mtx);
	}

	static void
	arc_hdr_free_on_write(arc_buf_hdr_t *hdr, boolean_t free_rdata)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;
	arc_buf_contents_t type = arc_buf_type(hdr);
	uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);

	/* protected by hash lock, if in the hash table */
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT(state != arc_anon && state != arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	size, hdr);
	}
	(void) zfs_refcount_remove_many(&state->arcs_size, size, hdr);
	if (type == ARC_BUFC_METADATA) {
	arc_space_return(size, ARC_SPACE_META);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	arc_space_return(size, ARC_SPACE_DATA);
	}

	if (free_rdata) {
	l2arc_free_abd_on_write(hdr->b_crypt_hdr.b_rabd, size, type);
	} else {
	l2arc_free_abd_on_write(hdr->b_l1hdr.b_pabd, size, type);
	}
	}

	/*
	* Share the arc_buf_t's data with the hdr. Whenever we are sharing the
	* data buffer, we transfer the refcount ownership to the hdr and update
	* the appropriate kstats.
	*/
	static void
	arc_share_buf(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(arc_can_share(hdr, buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!ARC_BUF_ENCRYPTED(buf));
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Start sharing the data buffer. We transfer the
	* refcount ownership to the hdr since it always owns
	* the refcount whenever an arc_buf_t is shared.
	*/
	zfs_refcount_transfer_ownership_many(&hdr->b_l1hdr.b_state->arcs_size,
	arc_hdr_size(hdr), buf, hdr);
	hdr->b_l1hdr.b_pabd = abd_get_from_buf(buf->b_data, arc_buf_size(buf));
	abd_take_ownership_of_buf(hdr->b_l1hdr.b_pabd,
	HDR_ISTYPE_METADATA(hdr));
	arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
	buf->b_flags \|= ARC_BUF_FLAG_SHARED;

	/*
	* Since we've transferred ownership to the hdr we need
	* to increment its compressed and uncompressed kstats and
	* decrement the overhead size.
	*/
	ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr));
	ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
	ARCSTAT_INCR(arcstat_overhead_size, -arc_buf_size(buf));
	}

	static void
	arc_unshare_buf(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(arc_buf_is_shared(buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* We are no longer sharing this buffer so we need
	* to transfer its ownership to the rightful owner.
	*/
	zfs_refcount_transfer_ownership_many(&hdr->b_l1hdr.b_state->arcs_size,
	arc_hdr_size(hdr), hdr, buf);
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
	abd_release_ownership_of_buf(hdr->b_l1hdr.b_pabd);
	abd_free(hdr->b_l1hdr.b_pabd);
	hdr->b_l1hdr.b_pabd = NULL;
	buf->b_flags &= ~ARC_BUF_FLAG_SHARED;

	/*
	* Since the buffer is no longer shared between
	* the arc buf and the hdr, count it as overhead.
	*/
	ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr));
	ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
	ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
	}

	/*
	* Remove an arc_buf_t from the hdr's buf list and return the last
	* arc_buf_t on the list. If no buffers remain on the list then return
	* NULL.
	*/
	static arc_buf_t *
	arc_buf_remove(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	arc_buf_t **bufp = &hdr->b_l1hdr.b_buf;
	arc_buf_t *lastbuf = NULL;

	/*
	* Remove the buf from the hdr list and locate the last
	* remaining buffer on the list.
	*/
	while (*bufp != NULL) {
	if (*bufp == buf)
	*bufp = buf->b_next;

	/*
	* If we've removed a buffer in the middle of
	* the list then update the lastbuf and update
	* bufp.
	*/
	if (*bufp != NULL) {
	lastbuf = *bufp;
	bufp = &(*bufp)->b_next;
	}
	}
	buf->b_next = NULL;
	ASSERT3P(lastbuf, !=, buf);
	IMPLY(hdr->b_l1hdr.b_bufcnt > 0, lastbuf != NULL);
	IMPLY(hdr->b_l1hdr.b_bufcnt > 0, hdr->b_l1hdr.b_buf != NULL);
	IMPLY(lastbuf != NULL, ARC_BUF_LAST(lastbuf));

	return (lastbuf);
	}

	/*
	* Free up buf->b_data and pull the arc_buf_t off of the arc_buf_hdr_t's
	* list and free it.
	*/
	static void
	arc_buf_destroy_impl(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* Free up the data associated with the buf but only if we're not
	* sharing this with the hdr. If we are sharing it with the hdr, the
	* hdr is responsible for doing the free.
	*/
	if (buf->b_data != NULL) {
	/*
	* We're about to change the hdr's b_flags. We must either
	* hold the hash_lock or be undiscoverable.
	*/
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	arc_cksum_verify(buf);
	arc_buf_unwatch(buf);

	if (arc_buf_is_shared(buf)) {
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
	} else {
	uint64_t size = arc_buf_size(buf);
	arc_free_data_buf(hdr, buf->b_data, size, buf);
	ARCSTAT_INCR(arcstat_overhead_size, -size);
	}
	buf->b_data = NULL;

	ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
	hdr->b_l1hdr.b_bufcnt -= 1;

	if (ARC_BUF_ENCRYPTED(buf)) {
	hdr->b_crypt_hdr.b_ebufcnt -= 1;

	/*
	* If we have no more encrypted buffers and we've
	* already gotten a copy of the decrypted data we can
	* free b_rabd to save some space.
	*/
	if (hdr->b_crypt_hdr.b_ebufcnt == 0 &&
	HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd != NULL &&
	!HDR_IO_IN_PROGRESS(hdr)) {
	arc_hdr_free_abd(hdr, B_TRUE);
	}
	}
	}

	arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);

	if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) {
	/*
	* If the current arc_buf_t is sharing its data buffer with the
	* hdr, then reassign the hdr's b_pabd to share it with the new
	* buffer at the end of the list. The shared buffer is always
	* the last one on the hdr's buffer list.
	*
	* There is an equivalent case for compressed bufs, but since
	* they aren't guaranteed to be the last buf in the list and
	* that is an exceedingly rare case, we just allow that space be
	* wasted temporarily. We must also be careful not to share
	* encrypted buffers, since they cannot be shared.
	*/
	if (lastbuf != NULL && !ARC_BUF_ENCRYPTED(lastbuf)) {
	/* Only one buf can be shared at once */
	VERIFY(!arc_buf_is_shared(lastbuf));
	/* hdr is uncompressed so can't have compressed buf */
	VERIFY(!ARC_BUF_COMPRESSED(lastbuf));

	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	arc_hdr_free_abd(hdr, B_FALSE);

	/*
	* We must setup a new shared block between the
	* last buffer and the hdr. The data would have
	* been allocated by the arc buf so we need to transfer
	* ownership to the hdr since it's now being shared.
	*/
	arc_share_buf(hdr, lastbuf);
	}
	} else if (HDR_SHARED_DATA(hdr)) {
	/*
	* Uncompressed shared buffers are always at the end
	* of the list. Compressed buffers don't have the
	* same requirements. This makes it hard to
	* simply assert that the lastbuf is shared so
	* we rely on the hdr's compression flags to determine
	* if we have a compressed, shared buffer.
	*/
	ASSERT3P(lastbuf, !=, NULL);
	ASSERT(arc_buf_is_shared(lastbuf) \|\|
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	}

	/*
	* Free the checksum if we're removing the last uncompressed buf from
	* this hdr.
	*/
	if (!arc_hdr_has_uncompressed_buf(hdr)) {
	arc_cksum_free(hdr);
	}

	/* clean up the buf */
	buf->b_hdr = NULL;
	kmem_cache_free(buf_cache, buf);
	}

	static void
	arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, int alloc_flags)
	{
	uint64_t size;
	boolean_t alloc_rdata = ((alloc_flags & ARC_HDR_ALLOC_RDATA) != 0);

	ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!HDR_SHARED_DATA(hdr) \|\| alloc_rdata);
	IMPLY(alloc_rdata, HDR_PROTECTED(hdr));

	if (alloc_rdata) {
	size = HDR_GET_PSIZE(hdr);
	ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL);
	hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr,
	alloc_flags);
	ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL);
	ARCSTAT_INCR(arcstat_raw_size, size);
	} else {
	size = arc_hdr_size(hdr);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr,
	alloc_flags);
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	}

	ARCSTAT_INCR(arcstat_compressed_size, size);
	ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
	}

	static void
	arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata)
	{
	uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));
	IMPLY(free_rdata, HDR_HAS_RABD(hdr));

	/*
	* If the hdr is currently being written to the l2arc then
	* we defer freeing the data by adding it to the l2arc_free_on_write
	* list. The l2arc will free the data once it's finished
	* writing it to the l2arc device.
	*/
	if (HDR_L2_WRITING(hdr)) {
	arc_hdr_free_on_write(hdr, free_rdata);
	ARCSTAT_BUMP(arcstat_l2_free_on_write);
	} else if (free_rdata) {
	arc_free_data_abd(hdr, hdr->b_crypt_hdr.b_rabd, size, hdr);
	} else {
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, size, hdr);
	}

	if (free_rdata) {
	hdr->b_crypt_hdr.b_rabd = NULL;
	ARCSTAT_INCR(arcstat_raw_size, -size);
	} else {
	hdr->b_l1hdr.b_pabd = NULL;
	}

	if (hdr->b_l1hdr.b_pabd == NULL && !HDR_HAS_RABD(hdr))
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;

	ARCSTAT_INCR(arcstat_compressed_size, -size);
	ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
	}

	/*
	* Allocate empty anonymous ARC header. The header will get its identity
	* assigned and buffers attached later as part of read or write operations.
	*
	* In case of read arc_read() assigns header its identify (b_dva + b_birth),
	* inserts it into ARC hash to become globally visible and allocates physical
	* (b_pabd) or raw (b_rabd) ABD buffer to read into from disk. On disk read
	* completion arc_read_done() allocates ARC buffer(s) as needed, potentially
	* sharing one of them with the physical ABD buffer.
	*
	* In case of write arc_alloc_buf() allocates ARC buffer to be filled with
	* data. Then after compression and/or encryption arc_write_ready() allocates
	* and fills (or potentially shares) physical (b_pabd) or raw (b_rabd) ABD
	* buffer. On disk write completion arc_write_done() assigns the header its
	* new identity (b_dva + b_birth) and inserts into ARC hash.
	*
	* In case of partial overwrite the old data is read first as described. Then
	* arc_release() either allocates new anonymous ARC header and moves the ARC
	* buffer to it, or reuses the old ARC header by discarding its identity and
	* removing it from ARC hash. After buffer modification normal write process
	* follows as described.
	*/
	static arc_buf_hdr_t *
	arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
	boolean_t protected, enum zio_compress compression_type, uint8_t complevel,
	arc_buf_contents_t type)
	{
	arc_buf_hdr_t *hdr;

	VERIFY(type == ARC_BUFC_DATA \|\| type == ARC_BUFC_METADATA);
	if (protected) {
	hdr = kmem_cache_alloc(hdr_full_crypt_cache, KM_PUSHPAGE);
	} else {
	hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);
	}

	ASSERT(HDR_EMPTY(hdr));
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	HDR_SET_PSIZE(hdr, psize);
	HDR_SET_LSIZE(hdr, lsize);
	hdr->b_spa = spa;
	hdr->b_type = type;
	hdr->b_flags = 0;
	arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) \| ARC_FLAG_HAS_L1HDR);
	arc_hdr_set_compress(hdr, compression_type);
	hdr->b_complevel = complevel;
	if (protected)
	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);

	hdr->b_l1hdr.b_state = arc_anon;
	hdr->b_l1hdr.b_arc_access = 0;
	hdr->b_l1hdr.b_mru_hits = 0;
	hdr->b_l1hdr.b_mru_ghost_hits = 0;
	hdr->b_l1hdr.b_mfu_hits = 0;
	hdr->b_l1hdr.b_mfu_ghost_hits = 0;
	hdr->b_l1hdr.b_bufcnt = 0;
	hdr->b_l1hdr.b_buf = NULL;

	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));

	return (hdr);
	}

	/*
	* Transition between the two allocation states for the arc_buf_hdr struct.
	* The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without
	* (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller
	* version is used when a cache buffer is only in the L2ARC in order to reduce
	* memory usage.
	*/
	static arc_buf_hdr_t *
	arc_hdr_realloc(arc_buf_hdr_t hdr, kmem_cache_t old, kmem_cache_t *new)
	{
	ASSERT(HDR_HAS_L2HDR(hdr));

	arc_buf_hdr_t *nhdr;
	l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;

	ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) \|\|
	(old == hdr_l2only_cache && new == hdr_full_cache));

	/*
	* if the caller wanted a new full header and the header is to be
	* encrypted we will actually allocate the header from the full crypt
	* cache instead. The same applies to freeing from the old cache.
	*/
	if (HDR_PROTECTED(hdr) && new == hdr_full_cache)
	new = hdr_full_crypt_cache;
	if (HDR_PROTECTED(hdr) && old == hdr_full_cache)
	old = hdr_full_crypt_cache;

	nhdr = kmem_cache_alloc(new, KM_PUSHPAGE);

	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	buf_hash_remove(hdr);

	memcpy(nhdr, hdr, HDR_L2ONLY_SIZE);

	if (new == hdr_full_cache \|\| new == hdr_full_crypt_cache) {
	arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR);
	/*
	* arc_access and arc_change_state need to be aware that a
	* header has just come out of L2ARC, so we set its state to
	* l2c_only even though it's about to change.
	*/
	nhdr->b_l1hdr.b_state = arc_l2c_only;

	/* Verify previous threads set to NULL before freeing */
	ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	} else {
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT0(hdr->b_l1hdr.b_bufcnt);
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);

	/*
	* If we've reached here, We must have been called from
	* arc_evict_hdr(), as such we should have already been
	* removed from any ghost list we were previously on
	* (which protects us from racing with arc_evict_state),
	* thus no locking is needed during this check.
	*/
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));

	/*
	* A buffer must not be moved into the arc_l2c_only
	* state if it's not finished being written out to the
	* l2arc device. Otherwise, the b_l1hdr.b_pabd field
	* might try to be accessed, even though it was removed.
	*/
	VERIFY(!HDR_L2_WRITING(hdr));
	VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));

	arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR);
	}
	/*
	* The header has been reallocated so we need to re-insert it into any
	* lists it was on.
	*/
	(void) buf_hash_insert(nhdr, NULL);

	ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node));

	mutex_enter(&dev->l2ad_mtx);

	/*
	* We must place the realloc'ed header back into the list at
	* the same spot. Otherwise, if it's placed earlier in the list,
	* l2arc_write_buffers() could find it during the function's
	* write phase, and try to write it out to the l2arc.
	*/
	list_insert_after(&dev->l2ad_buflist, hdr, nhdr);
	list_remove(&dev->l2ad_buflist, hdr);

	mutex_exit(&dev->l2ad_mtx);

	/*
	* Since we're using the pointer address as the tag when
	* incrementing and decrementing the l2ad_alloc refcount, we
	* must remove the old pointer (that we're about to destroy) and
	* add the new pointer to the refcount. Otherwise we'd remove
	* the wrong pointer address when calling arc_hdr_destroy() later.
	*/

	(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(nhdr), nhdr);

	buf_discard_identity(hdr);
	kmem_cache_free(old, hdr);

	return (nhdr);
	}

	/*
	* This function allows an L1 header to be reallocated as a crypt
	* header and vice versa. If we are going to a crypt header, the
	* new fields will be zeroed out.
	*/
	static arc_buf_hdr_t *
	arc_hdr_realloc_crypt(arc_buf_hdr_t *hdr, boolean_t need_crypt)
	{
	arc_buf_hdr_t *nhdr;
	arc_buf_t *buf;
	kmem_cache_t ncache, ocache;

	/*
	* This function requires that hdr is in the arc_anon state.
	* Therefore it won't have any L2ARC data for us to worry
	* about copying.
	*/
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!HDR_HAS_L2HDR(hdr));
	ASSERT3U(!!HDR_PROTECTED(hdr), !=, need_crypt);
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	ASSERT(!list_link_active(&hdr->b_l2hdr.b_l2node));
	ASSERT3P(hdr->b_hash_next, ==, NULL);

	if (need_crypt) {
	ncache = hdr_full_crypt_cache;
	ocache = hdr_full_cache;
	} else {
	ncache = hdr_full_cache;
	ocache = hdr_full_crypt_cache;
	}

	nhdr = kmem_cache_alloc(ncache, KM_PUSHPAGE);

	/*
	* Copy all members that aren't locks or condvars to the new header.
	* No lists are pointing to us (as we asserted above), so we don't
	* need to worry about the list nodes.
	*/
	nhdr->b_dva = hdr->b_dva;
	nhdr->b_birth = hdr->b_birth;
	nhdr->b_type = hdr->b_type;
	nhdr->b_flags = hdr->b_flags;
	nhdr->b_psize = hdr->b_psize;
	nhdr->b_lsize = hdr->b_lsize;
	nhdr->b_spa = hdr->b_spa;
	nhdr->b_l1hdr.b_freeze_cksum = hdr->b_l1hdr.b_freeze_cksum;
	nhdr->b_l1hdr.b_bufcnt = hdr->b_l1hdr.b_bufcnt;
	nhdr->b_l1hdr.b_byteswap = hdr->b_l1hdr.b_byteswap;
	nhdr->b_l1hdr.b_state = hdr->b_l1hdr.b_state;
	nhdr->b_l1hdr.b_arc_access = hdr->b_l1hdr.b_arc_access;
	nhdr->b_l1hdr.b_mru_hits = hdr->b_l1hdr.b_mru_hits;
	nhdr->b_l1hdr.b_mru_ghost_hits = hdr->b_l1hdr.b_mru_ghost_hits;
	nhdr->b_l1hdr.b_mfu_hits = hdr->b_l1hdr.b_mfu_hits;
	nhdr->b_l1hdr.b_mfu_ghost_hits = hdr->b_l1hdr.b_mfu_ghost_hits;
	nhdr->b_l1hdr.b_acb = hdr->b_l1hdr.b_acb;
	nhdr->b_l1hdr.b_pabd = hdr->b_l1hdr.b_pabd;

	/*
	* This zfs_refcount_add() exists only to ensure that the individual
	* arc buffers always point to a header that is referenced, avoiding
	* a small race condition that could trigger ASSERTs.
	*/
	(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, FTAG);
	nhdr->b_l1hdr.b_buf = hdr->b_l1hdr.b_buf;
	for (buf = nhdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) {
	mutex_enter(&buf->b_evict_lock);
	buf->b_hdr = nhdr;
	mutex_exit(&buf->b_evict_lock);
	}

	zfs_refcount_transfer(&nhdr->b_l1hdr.b_refcnt, &hdr->b_l1hdr.b_refcnt);
	(void) zfs_refcount_remove(&nhdr->b_l1hdr.b_refcnt, FTAG);
	ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));

	if (need_crypt) {
	arc_hdr_set_flags(nhdr, ARC_FLAG_PROTECTED);
	} else {
	arc_hdr_clear_flags(nhdr, ARC_FLAG_PROTECTED);
	}

	/* unset all members of the original hdr */
	memset(&hdr->b_dva, 0, sizeof (dva_t));
	hdr->b_birth = 0;
	hdr->b_type = ARC_BUFC_INVALID;
	hdr->b_flags = 0;
	hdr->b_psize = 0;
	hdr->b_lsize = 0;
	hdr->b_spa = 0;
	hdr->b_l1hdr.b_freeze_cksum = NULL;
	hdr->b_l1hdr.b_buf = NULL;
	hdr->b_l1hdr.b_bufcnt = 0;
	hdr->b_l1hdr.b_byteswap = 0;
	hdr->b_l1hdr.b_state = NULL;
	hdr->b_l1hdr.b_arc_access = 0;
	hdr->b_l1hdr.b_mru_hits = 0;
	hdr->b_l1hdr.b_mru_ghost_hits = 0;
	hdr->b_l1hdr.b_mfu_hits = 0;
	hdr->b_l1hdr.b_mfu_ghost_hits = 0;
	hdr->b_l1hdr.b_acb = NULL;
	hdr->b_l1hdr.b_pabd = NULL;

	if (ocache == hdr_full_crypt_cache) {
	ASSERT(!HDR_HAS_RABD(hdr));
	hdr->b_crypt_hdr.b_ot = DMU_OT_NONE;
	hdr->b_crypt_hdr.b_ebufcnt = 0;
	hdr->b_crypt_hdr.b_dsobj = 0;
	memset(hdr->b_crypt_hdr.b_salt, 0, ZIO_DATA_SALT_LEN);
	memset(hdr->b_crypt_hdr.b_iv, 0, ZIO_DATA_IV_LEN);
	memset(hdr->b_crypt_hdr.b_mac, 0, ZIO_DATA_MAC_LEN);
	}

	buf_discard_identity(hdr);
	kmem_cache_free(ocache, hdr);

	return (nhdr);
	}

	/*
	* This function is used by the send / receive code to convert a newly
	* allocated arc_buf_t to one that is suitable for a raw encrypted write. It
	* is also used to allow the root objset block to be updated without altering
	* its embedded MACs. Both block types will always be uncompressed so we do not
	* have to worry about compression type or psize.
	*/
	void
	arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder,
	dmu_object_type_t ot, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(ot == DMU_OT_DNODE \|\| ot == DMU_OT_OBJSET);
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);

	buf->b_flags \|= (ARC_BUF_FLAG_COMPRESSED \| ARC_BUF_FLAG_ENCRYPTED);
	if (!HDR_PROTECTED(hdr))
	hdr = arc_hdr_realloc_crypt(hdr, B_TRUE);
	hdr->b_crypt_hdr.b_dsobj = dsobj;
	hdr->b_crypt_hdr.b_ot = ot;
	hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
	DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
	if (!arc_hdr_has_uncompressed_buf(hdr))
	arc_cksum_free(hdr);

	if (salt != NULL)
	memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN);
	if (iv != NULL)
	memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN);
	if (mac != NULL)
	memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN);
	}

	/*
	* Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller.
	* The buf is returned thawed since we expect the consumer to modify it.
	*/
	arc_buf_t *
	arc_alloc_buf(spa_t spa, const void tag, arc_buf_contents_t type,
	int32_t size)
	{
	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size,
	B_FALSE, ZIO_COMPRESS_OFF, 0, type);

	arc_buf_t *buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE,
	B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);

	return (buf);
	}

	/*
	* Allocate a compressed buf in the same manner as arc_alloc_buf. Don't use this
	* for bufs containing metadata.
	*/
	arc_buf_t *
	arc_alloc_compressed_buf(spa_t spa, const void tag, uint64_t psize,
	uint64_t lsize, enum zio_compress compression_type, uint8_t complevel)
	{
	ASSERT3U(lsize, >, 0);
	ASSERT3U(lsize, >=, psize);
	ASSERT3U(compression_type, >, ZIO_COMPRESS_OFF);
	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);

	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
	B_FALSE, compression_type, complevel, ARC_BUFC_DATA);

	arc_buf_t *buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE,
	B_TRUE, B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);

	/*
	* To ensure that the hdr has the correct data in it if we call
	* arc_untransform() on this buf before it's been written to disk,
	* it's easiest if we just set up sharing between the buf and the hdr.
	*/
	arc_share_buf(hdr, buf);

	return (buf);
	}

	arc_buf_t *
	arc_alloc_raw_buf(spa_t spa, const void tag, uint64_t dsobj,
	boolean_t byteorder, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_hdr_t *hdr;
	arc_buf_t *buf;
	arc_buf_contents_t type = DMU_OT_IS_METADATA(ot) ?
	ARC_BUFC_METADATA : ARC_BUFC_DATA;

	ASSERT3U(lsize, >, 0);
	ASSERT3U(lsize, >=, psize);
	ASSERT3U(compression_type, >=, ZIO_COMPRESS_OFF);
	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);

	hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE,
	compression_type, complevel, type);

	hdr->b_crypt_hdr.b_dsobj = dsobj;
	hdr->b_crypt_hdr.b_ot = ot;
	hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
	DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
	memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN);
	memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN);
	memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN);

	/*
	* This buffer will be considered encrypted even if the ot is not an
	* encrypted type. It will become authenticated instead in
	* arc_write_ready().
	*/
	buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_TRUE, B_TRUE,
	B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);

	return (buf);
	}

	static void
	l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
	boolean_t state_only)
	{
	l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
	l2arc_dev_t *dev = l2hdr->b_dev;
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
	arc_buf_contents_t type = hdr->b_type;
	int64_t lsize_s;
	int64_t psize_s;
	int64_t asize_s;

	if (incr) {
	lsize_s = lsize;
	psize_s = psize;
	asize_s = asize;
	} else {
	lsize_s = -lsize;
	psize_s = -psize;
	asize_s = -asize;
	}

	/* If the buffer is a prefetch, count it as such. */
	if (HDR_PREFETCH(hdr)) {
	ARCSTAT_INCR(arcstat_l2_prefetch_asize, asize_s);
	} else {
	/*
	* We use the value stored in the L2 header upon initial
	* caching in L2ARC. This value will be updated in case
	* an MRU/MRU_ghost buffer transitions to MFU but the L2ARC
	* metadata (log entry) cannot currently be updated. Having
	* the ARC state in the L2 header solves the problem of a
	* possibly absent L1 header (apparent in buffers restored
	* from persistent L2ARC).
	*/
	switch (hdr->b_l2hdr.b_arcs_state) {
	case ARC_STATE_MRU_GHOST:
	case ARC_STATE_MRU:
	ARCSTAT_INCR(arcstat_l2_mru_asize, asize_s);
	break;
	case ARC_STATE_MFU_GHOST:
	case ARC_STATE_MFU:
	ARCSTAT_INCR(arcstat_l2_mfu_asize, asize_s);
	break;
	default:
	break;
	}
	}

	if (state_only)
	return;

	ARCSTAT_INCR(arcstat_l2_psize, psize_s);
	ARCSTAT_INCR(arcstat_l2_lsize, lsize_s);

	switch (type) {
	case ARC_BUFC_DATA:
	ARCSTAT_INCR(arcstat_l2_bufc_data_asize, asize_s);
	break;
	case ARC_BUFC_METADATA:
	ARCSTAT_INCR(arcstat_l2_bufc_metadata_asize, asize_s);
	break;
	default:
	break;
	}
	}


	static void
	arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr)
	{
	l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
	l2arc_dev_t *dev = l2hdr->b_dev;
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);

	ASSERT(MUTEX_HELD(&dev->l2ad_mtx));
	ASSERT(HDR_HAS_L2HDR(hdr));

	list_remove(&dev->l2ad_buflist, hdr);

	l2arc_hdr_arcstats_decrement(hdr);
	vdev_space_update(dev->l2ad_vdev, -asize, 0, 0);

	(void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr),
	hdr);
	arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
	}

	static void
	arc_hdr_destroy(arc_buf_hdr_t *hdr)
	{
	if (HDR_HAS_L1HDR(hdr)) {
	ASSERT(hdr->b_l1hdr.b_buf == NULL \|\|
	hdr->b_l1hdr.b_bufcnt > 0);
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	}
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT(!HDR_IN_HASH_TABLE(hdr));

	if (HDR_HAS_L2HDR(hdr)) {
	l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
	boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx);

	if (!buflist_held)
	mutex_enter(&dev->l2ad_mtx);

	/*
	* Even though we checked this conditional above, we
	* need to check this again now that we have the
	* l2ad_mtx. This is because we could be racing with
	* another thread calling l2arc_evict() which might have
	* destroyed this header's L2 portion as we were waiting
	* to acquire the l2ad_mtx. If that happens, we don't
	* want to re-destroy the header's L2 portion.
	*/
	if (HDR_HAS_L2HDR(hdr)) {

	if (!HDR_EMPTY(hdr))
	buf_discard_identity(hdr);

	arc_hdr_l2hdr_destroy(hdr);
	}

	if (!buflist_held)
	mutex_exit(&dev->l2ad_mtx);
	}

	/*
	* The header's identify can only be safely discarded once it is no
	* longer discoverable. This requires removing it from the hash table
	* and the l2arc header list. After this point the hash lock can not
	* be used to protect the header.
	*/
	if (!HDR_EMPTY(hdr))
	buf_discard_identity(hdr);

	if (HDR_HAS_L1HDR(hdr)) {
	arc_cksum_free(hdr);

	while (hdr->b_l1hdr.b_buf != NULL)
	arc_buf_destroy_impl(hdr->b_l1hdr.b_buf);

	if (hdr->b_l1hdr.b_pabd != NULL)
	arc_hdr_free_abd(hdr, B_FALSE);

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);
	}

	ASSERT3P(hdr->b_hash_next, ==, NULL);
	if (HDR_HAS_L1HDR(hdr)) {
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);

	if (!HDR_PROTECTED(hdr)) {
	kmem_cache_free(hdr_full_cache, hdr);
	} else {
	kmem_cache_free(hdr_full_crypt_cache, hdr);
	}
	} else {
	kmem_cache_free(hdr_l2only_cache, hdr);
	}
	}

	void
	arc_buf_destroy(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	if (hdr->b_l1hdr.b_state == arc_anon) {
	ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	VERIFY0(remove_reference(hdr, NULL, tag));
	arc_hdr_destroy(hdr);
	return;
	}

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	ASSERT3P(hdr, ==, buf->b_hdr);
	ASSERT(hdr->b_l1hdr.b_bufcnt > 0);
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
	ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon);
	ASSERT3P(buf->b_data, !=, NULL);

	(void) remove_reference(hdr, hash_lock, tag);
	arc_buf_destroy_impl(buf);
	mutex_exit(hash_lock);
	}

	/*
	* Evict the arc_buf_hdr that is provided as a parameter. The resultant
	* state of the header is dependent on its state prior to entering this
	* function. The following transitions are possible:
	*
	* - arc_mru -> arc_mru_ghost
	* - arc_mfu -> arc_mfu_ghost
	* - arc_mru_ghost -> arc_l2c_only
	* - arc_mru_ghost -> deleted
	* - arc_mfu_ghost -> arc_l2c_only
	* - arc_mfu_ghost -> deleted
	*
	* Return total size of evicted data buffers for eviction progress tracking.
	* When evicting from ghost states return logical buffer size to make eviction
	* progress at the same (or at least comparable) rate as from non-ghost states.
	*
	* Return *real_evicted for actual ARC size reduction to wake up threads
	* waiting for it. For non-ghost states it includes size of evicted data
	* buffers (the headers are not freed there). For ghost states it includes
	* only the evicted headers size.
	*/
	static int64_t
	arc_evict_hdr(arc_buf_hdr_t hdr, kmutex_t hash_lock, uint64_t *real_evicted)
	{
	arc_state_t evicted_state, state;
	int64_t bytes_evicted = 0;
	- int min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ?
	+ uint_t min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ?
	arc_min_prescient_prefetch_ms : arc_min_prefetch_ms;

	ASSERT(MUTEX_HELD(hash_lock));
	ASSERT(HDR_HAS_L1HDR(hdr));

	*real_evicted = 0;
	state = hdr->b_l1hdr.b_state;
	if (GHOST_STATE(state)) {
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);

	/*
	* l2arc_write_buffers() relies on a header's L1 portion
	* (i.e. its b_pabd field) during it's write phase.
	* Thus, we cannot push a header onto the arc_l2c_only
	* state (removing its L1 piece) until the header is
	* done being written to the l2arc.
	*/
	if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) {
	ARCSTAT_BUMP(arcstat_evict_l2_skip);
	return (bytes_evicted);
	}

	ARCSTAT_BUMP(arcstat_deleted);
	bytes_evicted += HDR_GET_LSIZE(hdr);

	DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr);

	if (HDR_HAS_L2HDR(hdr)) {
	ASSERT(hdr->b_l1hdr.b_pabd == NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	/*
	* This buffer is cached on the 2nd Level ARC;
	* don't destroy the header.
	*/
	arc_change_state(arc_l2c_only, hdr, hash_lock);
	/*
	* dropping from L1+L2 cached to L2-only,
	* realloc to remove the L1 header.
	*/
	hdr = arc_hdr_realloc(hdr, hdr_full_cache,
	hdr_l2only_cache);
	*real_evicted += HDR_FULL_SIZE - HDR_L2ONLY_SIZE;
	} else {
	arc_change_state(arc_anon, hdr, hash_lock);
	arc_hdr_destroy(hdr);
	*real_evicted += HDR_FULL_SIZE;
	}
	return (bytes_evicted);
	}

	ASSERT(state == arc_mru \|\| state == arc_mfu);
	evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost;

	/* prefetch buffers have a minimum lifespan */
	if (HDR_IO_IN_PROGRESS(hdr) \|\|
	((hdr->b_flags & (ARC_FLAG_PREFETCH \| ARC_FLAG_INDIRECT)) &&
	ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access <
	MSEC_TO_TICK(min_lifetime))) {
	ARCSTAT_BUMP(arcstat_evict_skip);
	return (bytes_evicted);
	}

	ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));
	while (hdr->b_l1hdr.b_buf) {
	arc_buf_t *buf = hdr->b_l1hdr.b_buf;
	if (!mutex_tryenter(&buf->b_evict_lock)) {
	ARCSTAT_BUMP(arcstat_mutex_miss);
	break;
	}
	if (buf->b_data != NULL) {
	bytes_evicted += HDR_GET_LSIZE(hdr);
	*real_evicted += HDR_GET_LSIZE(hdr);
	}
	mutex_exit(&buf->b_evict_lock);
	arc_buf_destroy_impl(buf);
	}

	if (HDR_HAS_L2HDR(hdr)) {
	ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr));
	} else {
	if (l2arc_write_eligible(hdr->b_spa, hdr)) {
	ARCSTAT_INCR(arcstat_evict_l2_eligible,
	HDR_GET_LSIZE(hdr));

	switch (state->arcs_state) {
	case ARC_STATE_MRU:
	ARCSTAT_INCR(
	arcstat_evict_l2_eligible_mru,
	HDR_GET_LSIZE(hdr));
	break;
	case ARC_STATE_MFU:
	ARCSTAT_INCR(
	arcstat_evict_l2_eligible_mfu,
	HDR_GET_LSIZE(hdr));
	break;
	default:
	break;
	}
	} else {
	ARCSTAT_INCR(arcstat_evict_l2_ineligible,
	HDR_GET_LSIZE(hdr));
	}
	}

	if (hdr->b_l1hdr.b_bufcnt == 0) {
	arc_cksum_free(hdr);

	bytes_evicted += arc_hdr_size(hdr);
	*real_evicted += arc_hdr_size(hdr);

	/*
	* If this hdr is being evicted and has a compressed
	* buffer then we discard it here before we change states.
	* This ensures that the accounting is updated correctly
	* in arc_free_data_impl().
	*/
	if (hdr->b_l1hdr.b_pabd != NULL)
	arc_hdr_free_abd(hdr, B_FALSE);

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);

	arc_change_state(evicted_state, hdr, hash_lock);
	ASSERT(HDR_IN_HASH_TABLE(hdr));
	arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);
	DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
	}

	return (bytes_evicted);
	}

	static void
	arc_set_need_free(void)
	{
	ASSERT(MUTEX_HELD(&arc_evict_lock));
	int64_t remaining = arc_free_memory() - arc_sys_free / 2;
	arc_evict_waiter_t *aw = list_tail(&arc_evict_waiters);
	if (aw == NULL) {
	arc_need_free = MAX(-remaining, 0);
	} else {
	arc_need_free =
	MAX(-remaining, (int64_t)(aw->aew_count - arc_evict_count));
	}
	}

	static uint64_t
	arc_evict_state_impl(multilist_t ml, int idx, arc_buf_hdr_t marker,
	uint64_t spa, uint64_t bytes)
	{
	multilist_sublist_t *mls;
	uint64_t bytes_evicted = 0, real_evicted = 0;
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	- int evict_count = zfs_arc_evict_batch_limit;
	+ uint_t evict_count = zfs_arc_evict_batch_limit;

	ASSERT3P(marker, !=, NULL);

	mls = multilist_sublist_lock(ml, idx);

	for (hdr = multilist_sublist_prev(mls, marker); likely(hdr != NULL);
	hdr = multilist_sublist_prev(mls, marker)) {
	- if ((evict_count <= 0) \|\| (bytes_evicted >= bytes))
	+ if ((evict_count == 0) \|\| (bytes_evicted >= bytes))
	break;

	/*
	* To keep our iteration location, move the marker
	* forward. Since we're not holding hdr's hash lock, we
	* must be very careful and not remove 'hdr' from the
	* sublist. Otherwise, other consumers might mistake the
	* 'hdr' as not being on a sublist when they call the
	* multilist_link_active() function (they all rely on
	* the hash lock protecting concurrent insertions and
	* removals). multilist_sublist_move_forward() was
	* specifically implemented to ensure this is the case
	* (only 'marker' will be removed and re-inserted).
	*/
	multilist_sublist_move_forward(mls, marker);

	/*
	* The only case where the b_spa field should ever be
	* zero, is the marker headers inserted by
	* arc_evict_state(). It's possible for multiple threads
	* to be calling arc_evict_state() concurrently (e.g.
	* dsl_pool_close() and zio_inject_fault()), so we must
	* skip any markers we see from these other threads.
	*/
	if (hdr->b_spa == 0)
	continue;

	/* we're only interested in evicting buffers of a certain spa */
	if (spa != 0 && hdr->b_spa != spa) {
	ARCSTAT_BUMP(arcstat_evict_skip);
	continue;
	}

	hash_lock = HDR_LOCK(hdr);

	/*
	* We aren't calling this function from any code path
	* that would already be holding a hash lock, so we're
	* asserting on this assumption to be defensive in case
	* this ever changes. Without this check, it would be
	* possible to incorrectly increment arcstat_mutex_miss
	* below (e.g. if the code changed such that we called
	* this function with a hash lock held).
	*/
	ASSERT(!MUTEX_HELD(hash_lock));

	if (mutex_tryenter(hash_lock)) {
	uint64_t revicted;
	uint64_t evicted = arc_evict_hdr(hdr, hash_lock,
	&revicted);
	mutex_exit(hash_lock);

	bytes_evicted += evicted;
	real_evicted += revicted;

	/*
	* If evicted is zero, arc_evict_hdr() must have
	* decided to skip this header, don't increment
	* evict_count in this case.
	*/
	if (evicted != 0)
	evict_count--;

	} else {
	ARCSTAT_BUMP(arcstat_mutex_miss);
	}
	}

	multilist_sublist_unlock(mls);

	/*
	* Increment the count of evicted bytes, and wake up any threads that
	* are waiting for the count to reach this value. Since the list is
	* ordered by ascending aew_count, we pop off the beginning of the
	* list until we reach the end, or a waiter that's past the current
	* "count". Doing this outside the loop reduces the number of times
	* we need to acquire the global arc_evict_lock.
	*
	* Only wake when there's sufficient free memory in the system
	* (specifically, arc_sys_free/2, which by default is a bit more than
	* 1/64th of RAM). See the comments in arc_wait_for_eviction().
	*/
	mutex_enter(&arc_evict_lock);
	arc_evict_count += real_evicted;

	if (arc_free_memory() > arc_sys_free / 2) {
	arc_evict_waiter_t *aw;
	while ((aw = list_head(&arc_evict_waiters)) != NULL &&
	aw->aew_count <= arc_evict_count) {
	list_remove(&arc_evict_waiters, aw);
	cv_broadcast(&aw->aew_cv);
	}
	}
	arc_set_need_free();
	mutex_exit(&arc_evict_lock);

	/*
	* If the ARC size is reduced from arc_c_max to arc_c_min (especially
	* if the average cached block is small), eviction can be on-CPU for
	* many seconds. To ensure that other threads that may be bound to
	* this CPU are able to make progress, make a voluntary preemption
	* call here.
	*/
	kpreempt(KPREEMPT_SYNC);

	return (bytes_evicted);
	}

	/*
	* Allocate an array of buffer headers used as placeholders during arc state
	* eviction.
	*/
	static arc_buf_hdr_t **
	arc_state_alloc_markers(int count)
	{
	arc_buf_hdr_t **markers;

	markers = kmem_zalloc(sizeof (markers) count, KM_SLEEP);
	for (int i = 0; i < count; i++) {
	markers[i] = kmem_cache_alloc(hdr_full_cache, KM_SLEEP);

	/*
	* A b_spa of 0 is used to indicate that this header is
	* a marker. This fact is used in arc_evict_type() and
	* arc_evict_state_impl().
	*/
	markers[i]->b_spa = 0;

	}
	return (markers);
	}

	static void
	arc_state_free_markers(arc_buf_hdr_t **markers, int count)
	{
	for (int i = 0; i < count; i++)
	kmem_cache_free(hdr_full_cache, markers[i]);
	kmem_free(markers, sizeof (markers) count);
	}

	/*
	* Evict buffers from the given arc state, until we've removed the
	* specified number of bytes. Move the removed buffers to the
	* appropriate evict state.
	*
	* This function makes a "best effort". It skips over any buffers
	* it can't get a hash_lock on, and so, may not catch all candidates.
	* It may also return without evicting as much space as requested.
	*
	* If bytes is specified using the special value ARC_EVICT_ALL, this
	* will evict all available (i.e. unlocked and evictable) buffers from
	* the given arc state; which is used by arc_flush().
	*/
	static uint64_t
	arc_evict_state(arc_state_t *state, uint64_t spa, uint64_t bytes,
	arc_buf_contents_t type)
	{
	uint64_t total_evicted = 0;
	multilist_t *ml = &state->arcs_list[type];
	int num_sublists;
	arc_buf_hdr_t **markers;

	num_sublists = multilist_get_num_sublists(ml);

	/*
	* If we've tried to evict from each sublist, made some
	* progress, but still have not hit the target number of bytes
	* to evict, we want to keep trying. The markers allow us to
	* pick up where we left off for each individual sublist, rather
	* than starting from the tail each time.
	*/
	if (zthr_iscurthread(arc_evict_zthr)) {
	markers = arc_state_evict_markers;
	ASSERT3S(num_sublists, <=, arc_state_evict_marker_count);
	} else {
	markers = arc_state_alloc_markers(num_sublists);
	}
	for (int i = 0; i < num_sublists; i++) {
	multilist_sublist_t *mls;

	mls = multilist_sublist_lock(ml, i);
	multilist_sublist_insert_tail(mls, markers[i]);
	multilist_sublist_unlock(mls);
	}

	/*
	* While we haven't hit our target number of bytes to evict, or
	* we're evicting all available buffers.
	*/
	while (total_evicted < bytes) {
	int sublist_idx = multilist_get_random_index(ml);
	uint64_t scan_evicted = 0;

	/*
	* Try to reduce pinned dnodes with a floor of arc_dnode_limit.
	* Request that 10% of the LRUs be scanned by the superblock
	* shrinker.
	*/
	if (type == ARC_BUFC_DATA && aggsum_compare(
	&arc_sums.arcstat_dnode_size, arc_dnode_size_limit) > 0) {
	arc_prune_async((aggsum_upper_bound(
	&arc_sums.arcstat_dnode_size) -
	arc_dnode_size_limit) / sizeof (dnode_t) /
	zfs_arc_dnode_reduce_percent);
	}

	/*
	* Start eviction using a randomly selected sublist,
	* this is to try and evenly balance eviction across all
	* sublists. Always starting at the same sublist
	* (e.g. index 0) would cause evictions to favor certain
	* sublists over others.
	*/
	for (int i = 0; i < num_sublists; i++) {
	uint64_t bytes_remaining;
	uint64_t bytes_evicted;

	if (total_evicted < bytes)
	bytes_remaining = bytes - total_evicted;
	else
	break;

	bytes_evicted = arc_evict_state_impl(ml, sublist_idx,
	markers[sublist_idx], spa, bytes_remaining);

	scan_evicted += bytes_evicted;
	total_evicted += bytes_evicted;

	/* we've reached the end, wrap to the beginning */
	if (++sublist_idx >= num_sublists)
	sublist_idx = 0;
	}

	/*
	* If we didn't evict anything during this scan, we have
	* no reason to believe we'll evict more during another
	* scan, so break the loop.
	*/
	if (scan_evicted == 0) {
	/* This isn't possible, let's make that obvious */
	ASSERT3S(bytes, !=, 0);

	/*
	* When bytes is ARC_EVICT_ALL, the only way to
	* break the loop is when scan_evicted is zero.
	* In that case, we actually have evicted enough,
	* so we don't want to increment the kstat.
	*/
	if (bytes != ARC_EVICT_ALL) {
	ASSERT3S(total_evicted, <, bytes);
	ARCSTAT_BUMP(arcstat_evict_not_enough);
	}

	break;
	}
	}

	for (int i = 0; i < num_sublists; i++) {
	multilist_sublist_t *mls = multilist_sublist_lock(ml, i);
	multilist_sublist_remove(mls, markers[i]);
	multilist_sublist_unlock(mls);
	}
	if (markers != arc_state_evict_markers)
	arc_state_free_markers(markers, num_sublists);

	return (total_evicted);
	}

	/*
	* Flush all "evictable" data of the given type from the arc state
	* specified. This will not evict any "active" buffers (i.e. referenced).
	*
	* When 'retry' is set to B_FALSE, the function will make a single pass
	* over the state and evict any buffers that it can. Since it doesn't
	* continually retry the eviction, it might end up leaving some buffers
	* in the ARC due to lock misses.
	*
	* When 'retry' is set to B_TRUE, the function will continually retry the
	* eviction until all evictable buffers have been removed from the
	* state. As a result, if concurrent insertions into the state are
	* allowed (e.g. if the ARC isn't shutting down), this function might
	* wind up in an infinite loop, continually trying to evict buffers.
	*/
	static uint64_t
	arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type,
	boolean_t retry)
	{
	uint64_t evicted = 0;

	while (zfs_refcount_count(&state->arcs_esize[type]) != 0) {
	evicted += arc_evict_state(state, spa, ARC_EVICT_ALL, type);

	if (!retry)
	break;
	}

	return (evicted);
	}

	/*
	* Evict the specified number of bytes from the state specified,
	* restricting eviction to the spa and type given. This function
	* prevents us from trying to evict more from a state's list than
	* is "evictable", and to skip evicting altogether when passed a
	* negative value for "bytes". In contrast, arc_evict_state() will
	* evict everything it can, when passed a negative value for "bytes".
	*/
	static uint64_t
	arc_evict_impl(arc_state_t *state, uint64_t spa, int64_t bytes,
	arc_buf_contents_t type)
	{
	uint64_t delta;

	if (bytes > 0 && zfs_refcount_count(&state->arcs_esize[type]) > 0) {
	delta = MIN(zfs_refcount_count(&state->arcs_esize[type]),
	bytes);
	return (arc_evict_state(state, spa, delta, type));
	}

	return (0);
	}

	/*
	* The goal of this function is to evict enough meta data buffers from the
	* ARC in order to enforce the arc_meta_limit. Achieving this is slightly
	* more complicated than it appears because it is common for data buffers
	* to have holds on meta data buffers. In addition, dnode meta data buffers
	* will be held by the dnodes in the block preventing them from being freed.
	* This means we can't simply traverse the ARC and expect to always find
	* enough unheld meta data buffer to release.
	*
	* Therefore, this function has been updated to make alternating passes
	* over the ARC releasing data buffers and then newly unheld meta data
	* buffers. This ensures forward progress is maintained and meta_used
	* will decrease. Normally this is sufficient, but if required the ARC
	* will call the registered prune callbacks causing dentry and inodes to
	* be dropped from the VFS cache. This will make dnode meta data buffers
	* available for reclaim.
	*/
	static uint64_t
	arc_evict_meta_balanced(uint64_t meta_used)
	{
	- int64_t delta, prune = 0, adjustmnt;
	- uint64_t total_evicted = 0;
	+ int64_t delta, adjustmnt;
	+ uint64_t total_evicted = 0, prune = 0;
	arc_buf_contents_t type = ARC_BUFC_DATA;
	- int restarts = MAX(zfs_arc_meta_adjust_restarts, 0);
	+ uint_t restarts = zfs_arc_meta_adjust_restarts;

	restart:
	/*
	* This slightly differs than the way we evict from the mru in
	* arc_evict because we don't have a "target" value (i.e. no
	* "meta" arc_p). As a result, I think we can completely
	* cannibalize the metadata in the MRU before we evict the
	* metadata from the MFU. I think we probably need to implement a
	* "metadata arc_p" value to do this properly.
	*/
	adjustmnt = meta_used - arc_meta_limit;

	if (adjustmnt > 0 &&
	zfs_refcount_count(&arc_mru->arcs_esize[type]) > 0) {
	delta = MIN(zfs_refcount_count(&arc_mru->arcs_esize[type]),
	adjustmnt);
	total_evicted += arc_evict_impl(arc_mru, 0, delta, type);
	adjustmnt -= delta;
	}

	/*
	* We can't afford to recalculate adjustmnt here. If we do,
	* new metadata buffers can sneak into the MRU or ANON lists,
	* thus penalize the MFU metadata. Although the fudge factor is
	* small, it has been empirically shown to be significant for
	* certain workloads (e.g. creating many empty directories). As
	* such, we use the original calculation for adjustmnt, and
	* simply decrement the amount of data evicted from the MRU.
	*/

	if (adjustmnt > 0 &&
	zfs_refcount_count(&arc_mfu->arcs_esize[type]) > 0) {
	delta = MIN(zfs_refcount_count(&arc_mfu->arcs_esize[type]),
	adjustmnt);
	total_evicted += arc_evict_impl(arc_mfu, 0, delta, type);
	}

	adjustmnt = meta_used - arc_meta_limit;

	if (adjustmnt > 0 &&
	zfs_refcount_count(&arc_mru_ghost->arcs_esize[type]) > 0) {
	delta = MIN(adjustmnt,
	zfs_refcount_count(&arc_mru_ghost->arcs_esize[type]));
	total_evicted += arc_evict_impl(arc_mru_ghost, 0, delta, type);
	adjustmnt -= delta;
	}

	if (adjustmnt > 0 &&
	zfs_refcount_count(&arc_mfu_ghost->arcs_esize[type]) > 0) {
	delta = MIN(adjustmnt,
	zfs_refcount_count(&arc_mfu_ghost->arcs_esize[type]));
	total_evicted += arc_evict_impl(arc_mfu_ghost, 0, delta, type);
	}

	/*
	* If after attempting to make the requested adjustment to the ARC
	* the meta limit is still being exceeded then request that the
	* higher layers drop some cached objects which have holds on ARC
	* meta buffers. Requests to the upper layers will be made with
	* increasingly large scan sizes until the ARC is below the limit.
	*/
	if (meta_used > arc_meta_limit) {
	if (type == ARC_BUFC_DATA) {
	type = ARC_BUFC_METADATA;
	} else {
	type = ARC_BUFC_DATA;

	if (zfs_arc_meta_prune) {
	prune += zfs_arc_meta_prune;
	arc_prune_async(prune);
	}
	}

	if (restarts > 0) {
	restarts--;
	goto restart;
	}
	}
	return (total_evicted);
	}

	/*
	* Evict metadata buffers from the cache, such that arcstat_meta_used is
	* capped by the arc_meta_limit tunable.
	*/
	static uint64_t
	arc_evict_meta_only(uint64_t meta_used)
	{
	uint64_t total_evicted = 0;
	int64_t target;

	/*
	* If we're over the meta limit, we want to evict enough
	* metadata to get back under the meta limit. We don't want to
	* evict so much that we drop the MRU below arc_p, though. If
	* we're over the meta limit more than we're over arc_p, we
	* evict some from the MRU here, and some from the MFU below.
	*/
	target = MIN((int64_t)(meta_used - arc_meta_limit),
	(int64_t)(zfs_refcount_count(&arc_anon->arcs_size) +
	zfs_refcount_count(&arc_mru->arcs_size) - arc_p));

	total_evicted += arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);

	/*
	* Similar to the above, we want to evict enough bytes to get us
	* below the meta limit, but not so much as to drop us below the
	* space allotted to the MFU (which is defined as arc_c - arc_p).
	*/
	target = MIN((int64_t)(meta_used - arc_meta_limit),
	(int64_t)(zfs_refcount_count(&arc_mfu->arcs_size) -
	(arc_c - arc_p)));

	total_evicted += arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);

	return (total_evicted);
	}

	static uint64_t
	arc_evict_meta(uint64_t meta_used)
	{
	if (zfs_arc_meta_strategy == ARC_STRATEGY_META_ONLY)
	return (arc_evict_meta_only(meta_used));
	else
	return (arc_evict_meta_balanced(meta_used));
	}

	/*
	* Return the type of the oldest buffer in the given arc state
	*
	* This function will select a random sublist of type ARC_BUFC_DATA and
	* a random sublist of type ARC_BUFC_METADATA. The tail of each sublist
	* is compared, and the type which contains the "older" buffer will be
	* returned.
	*/
	static arc_buf_contents_t
	arc_evict_type(arc_state_t *state)
	{
	multilist_t *data_ml = &state->arcs_list[ARC_BUFC_DATA];
	multilist_t *meta_ml = &state->arcs_list[ARC_BUFC_METADATA];
	int data_idx = multilist_get_random_index(data_ml);
	int meta_idx = multilist_get_random_index(meta_ml);
	multilist_sublist_t *data_mls;
	multilist_sublist_t *meta_mls;
	arc_buf_contents_t type;
	arc_buf_hdr_t *data_hdr;
	arc_buf_hdr_t *meta_hdr;

	/*
	* We keep the sublist lock until we're finished, to prevent
	* the headers from being destroyed via arc_evict_state().
	*/
	data_mls = multilist_sublist_lock(data_ml, data_idx);
	meta_mls = multilist_sublist_lock(meta_ml, meta_idx);

	/*
	* These two loops are to ensure we skip any markers that
	* might be at the tail of the lists due to arc_evict_state().
	*/

	for (data_hdr = multilist_sublist_tail(data_mls); data_hdr != NULL;
	data_hdr = multilist_sublist_prev(data_mls, data_hdr)) {
	if (data_hdr->b_spa != 0)
	break;
	}

	for (meta_hdr = multilist_sublist_tail(meta_mls); meta_hdr != NULL;
	meta_hdr = multilist_sublist_prev(meta_mls, meta_hdr)) {
	if (meta_hdr->b_spa != 0)
	break;
	}

	if (data_hdr == NULL && meta_hdr == NULL) {
	type = ARC_BUFC_DATA;
	} else if (data_hdr == NULL) {
	ASSERT3P(meta_hdr, !=, NULL);
	type = ARC_BUFC_METADATA;
	} else if (meta_hdr == NULL) {
	ASSERT3P(data_hdr, !=, NULL);
	type = ARC_BUFC_DATA;
	} else {
	ASSERT3P(data_hdr, !=, NULL);
	ASSERT3P(meta_hdr, !=, NULL);

	/* The headers can't be on the sublist without an L1 header */
	ASSERT(HDR_HAS_L1HDR(data_hdr));
	ASSERT(HDR_HAS_L1HDR(meta_hdr));

	if (data_hdr->b_l1hdr.b_arc_access <
	meta_hdr->b_l1hdr.b_arc_access) {
	type = ARC_BUFC_DATA;
	} else {
	type = ARC_BUFC_METADATA;
	}
	}

	multilist_sublist_unlock(meta_mls);
	multilist_sublist_unlock(data_mls);

	return (type);
	}

	/*
	* Evict buffers from the cache, such that arcstat_size is capped by arc_c.
	*/
	static uint64_t
	arc_evict(void)
	{
	uint64_t total_evicted = 0;
	uint64_t bytes;
	int64_t target;
	uint64_t asize = aggsum_value(&arc_sums.arcstat_size);
	uint64_t ameta = aggsum_value(&arc_sums.arcstat_meta_used);

	/*
	* If we're over arc_meta_limit, we want to correct that before
	* potentially evicting data buffers below.
	*/
	total_evicted += arc_evict_meta(ameta);

	/*
	* Adjust MRU size
	*
	* If we're over the target cache size, we want to evict enough
	* from the list to get back to our target size. We don't want
	* to evict too much from the MRU, such that it drops below
	* arc_p. So, if we're over our target cache size more than
	* the MRU is over arc_p, we'll evict enough to get back to
	* arc_p here, and then evict more from the MFU below.
	*/
	target = MIN((int64_t)(asize - arc_c),
	(int64_t)(zfs_refcount_count(&arc_anon->arcs_size) +
	zfs_refcount_count(&arc_mru->arcs_size) + ameta - arc_p));

	/*
	* If we're below arc_meta_min, always prefer to evict data.
	* Otherwise, try to satisfy the requested number of bytes to
	* evict from the type which contains older buffers; in an
	* effort to keep newer buffers in the cache regardless of their
	* type. If we cannot satisfy the number of bytes from this
	* type, spill over into the next type.
	*/
	if (arc_evict_type(arc_mru) == ARC_BUFC_METADATA &&
	ameta > arc_meta_min) {
	bytes = arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);
	total_evicted += bytes;

	/*
	* If we couldn't evict our target number of bytes from
	* metadata, we try to get the rest from data.
	*/
	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mru, 0, target, ARC_BUFC_DATA);
	} else {
	bytes = arc_evict_impl(arc_mru, 0, target, ARC_BUFC_DATA);
	total_evicted += bytes;

	/*
	* If we couldn't evict our target number of bytes from
	* data, we try to get the rest from metadata.
	*/
	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mru, 0, target, ARC_BUFC_METADATA);
	}

	/*
	* Re-sum ARC stats after the first round of evictions.
	*/
	asize = aggsum_value(&arc_sums.arcstat_size);
	ameta = aggsum_value(&arc_sums.arcstat_meta_used);


	/*
	* Adjust MFU size
	*
	* Now that we've tried to evict enough from the MRU to get its
	* size back to arc_p, if we're still above the target cache
	* size, we evict the rest from the MFU.
	*/
	target = asize - arc_c;

	if (arc_evict_type(arc_mfu) == ARC_BUFC_METADATA &&
	ameta > arc_meta_min) {
	bytes = arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);
	total_evicted += bytes;

	/*
	* If we couldn't evict our target number of bytes from
	* metadata, we try to get the rest from data.
	*/
	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_DATA);
	} else {
	bytes = arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_DATA);
	total_evicted += bytes;

	/*
	* If we couldn't evict our target number of bytes from
	* data, we try to get the rest from data.
	*/
	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mfu, 0, target, ARC_BUFC_METADATA);
	}

	/*
	* Adjust ghost lists
	*
	* In addition to the above, the ARC also defines target values
	* for the ghost lists. The sum of the mru list and mru ghost
	* list should never exceed the target size of the cache, and
	* the sum of the mru list, mfu list, mru ghost list, and mfu
	* ghost list should never exceed twice the target size of the
	* cache. The following logic enforces these limits on the ghost
	* caches, and evicts from them as needed.
	*/
	target = zfs_refcount_count(&arc_mru->arcs_size) +
	zfs_refcount_count(&arc_mru_ghost->arcs_size) - arc_c;

	bytes = arc_evict_impl(arc_mru_ghost, 0, target, ARC_BUFC_DATA);
	total_evicted += bytes;

	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mru_ghost, 0, target, ARC_BUFC_METADATA);

	/*
	* We assume the sum of the mru list and mfu list is less than
	* or equal to arc_c (we enforced this above), which means we
	* can use the simpler of the two equations below:
	*
	* mru + mfu + mru ghost + mfu ghost <= 2 * arc_c
	* mru ghost + mfu ghost <= arc_c
	*/
	target = zfs_refcount_count(&arc_mru_ghost->arcs_size) +
	zfs_refcount_count(&arc_mfu_ghost->arcs_size) - arc_c;

	bytes = arc_evict_impl(arc_mfu_ghost, 0, target, ARC_BUFC_DATA);
	total_evicted += bytes;

	target -= bytes;

	total_evicted +=
	arc_evict_impl(arc_mfu_ghost, 0, target, ARC_BUFC_METADATA);

	return (total_evicted);
	}

	void
	arc_flush(spa_t *spa, boolean_t retry)
	{
	uint64_t guid = 0;

	/*
	* If retry is B_TRUE, a spa must not be specified since we have
	* no good way to determine if all of a spa's buffers have been
	* evicted from an arc state.
	*/
	ASSERT(!retry \|\| spa == 0);

	if (spa != NULL)
	guid = spa_load_guid(spa);

	(void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry);
	}

	void
	arc_reduce_target_size(int64_t to_free)
	{
	uint64_t asize = aggsum_value(&arc_sums.arcstat_size);

	/*
	* All callers want the ARC to actually evict (at least) this much
	* memory. Therefore we reduce from the lower of the current size and
	* the target size. This way, even if arc_c is much higher than
	* arc_size (as can be the case after many calls to arc_freed(), we will
	* immediately have arc_c < arc_size and therefore the arc_evict_zthr
	* will evict.
	*/
	uint64_t c = MIN(arc_c, asize);

	if (c > to_free && c - to_free > arc_c_min) {
	arc_c = c - to_free;
	atomic_add_64(&arc_p, -(arc_p >> arc_shrink_shift));
	if (arc_p > arc_c)
	arc_p = (arc_c >> 1);
	ASSERT(arc_c >= arc_c_min);
	ASSERT((int64_t)arc_p >= 0);
	} else {
	arc_c = arc_c_min;
	}

	if (asize > arc_c) {
	/* See comment in arc_evict_cb_check() on why lock+flag */
	mutex_enter(&arc_evict_lock);
	arc_evict_needed = B_TRUE;
	mutex_exit(&arc_evict_lock);
	zthr_wakeup(arc_evict_zthr);
	}
	}

	/*
	* Determine if the system is under memory pressure and is asking
	* to reclaim memory. A return value of B_TRUE indicates that the system
	* is under memory pressure and that the arc should adjust accordingly.
	*/
	boolean_t
	arc_reclaim_needed(void)
	{
	return (arc_available_memory() < 0);
	}

	void
	arc_kmem_reap_soon(void)
	{
	size_t i;
	kmem_cache_t *prev_cache = NULL;
	kmem_cache_t *prev_data_cache = NULL;

	#ifdef _KERNEL
	if ((aggsum_compare(&arc_sums.arcstat_meta_used,
	arc_meta_limit) >= 0) && zfs_arc_meta_prune) {
	/*
	* We are exceeding our meta-data cache limit.
	* Prune some entries to release holds on meta-data.
	*/
	arc_prune_async(zfs_arc_meta_prune);
	}
	#if defined(_ILP32)
	/*
	* Reclaim unused memory from all kmem caches.
	*/
	kmem_reap();
	#endif
	#endif

	for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) {
	#if defined(_ILP32)
	/* reach upper limit of cache size on 32-bit */
	if (zio_buf_cache[i] == NULL)
	break;
	#endif
	if (zio_buf_cache[i] != prev_cache) {
	prev_cache = zio_buf_cache[i];
	kmem_cache_reap_now(zio_buf_cache[i]);
	}
	if (zio_data_buf_cache[i] != prev_data_cache) {
	prev_data_cache = zio_data_buf_cache[i];
	kmem_cache_reap_now(zio_data_buf_cache[i]);
	}
	}
	kmem_cache_reap_now(buf_cache);
	kmem_cache_reap_now(hdr_full_cache);
	kmem_cache_reap_now(hdr_l2only_cache);
	kmem_cache_reap_now(zfs_btree_leaf_cache);
	abd_cache_reap_now();
	}

	static boolean_t
	arc_evict_cb_check(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	#ifdef ZFS_DEBUG
	/*
	* This is necessary in order to keep the kstat information
	* up to date for tools that display kstat data such as the
	* mdb ::arc dcmd and the Linux crash utility. These tools
	* typically do not call kstat's update function, but simply
	* dump out stats from the most recent update. Without
	* this call, these commands may show stale stats for the
	* anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even
	* with this call, the data might be out of date if the
	* evict thread hasn't been woken recently; but that should
	* suffice. The arc_state_t structures can be queried
	* directly if more accurate information is needed.
	*/
	if (arc_ksp != NULL)
	arc_ksp->ks_update(arc_ksp, KSTAT_READ);
	#endif

	/*
	* We have to rely on arc_wait_for_eviction() to tell us when to
	* evict, rather than checking if we are overflowing here, so that we
	* are sure to not leave arc_wait_for_eviction() waiting on aew_cv.
	* If we have become "not overflowing" since arc_wait_for_eviction()
	* checked, we need to wake it up. We could broadcast the CV here,
	* but arc_wait_for_eviction() may have not yet gone to sleep. We
	* would need to use a mutex to ensure that this function doesn't
	* broadcast until arc_wait_for_eviction() has gone to sleep (e.g.
	* the arc_evict_lock). However, the lock ordering of such a lock
	* would necessarily be incorrect with respect to the zthr_lock,
	* which is held before this function is called, and is held by
	* arc_wait_for_eviction() when it calls zthr_wakeup().
	*/
	return (arc_evict_needed);
	}

	/*
	* Keep arc_size under arc_c by running arc_evict which evicts data
	* from the ARC.
	*/
	static void
	arc_evict_cb(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	uint64_t evicted = 0;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	/* Evict from cache */
	evicted = arc_evict();

	/*
	* If evicted is zero, we couldn't evict anything
	* via arc_evict(). This could be due to hash lock
	* collisions, but more likely due to the majority of
	* arc buffers being unevictable. Therefore, even if
	* arc_size is above arc_c, another pass is unlikely to
	* be helpful and could potentially cause us to enter an
	* infinite loop. Additionally, zthr_iscancelled() is
	* checked here so that if the arc is shutting down, the
	* broadcast will wake any remaining arc evict waiters.
	*/
	mutex_enter(&arc_evict_lock);
	arc_evict_needed = !zthr_iscancelled(arc_evict_zthr) &&
	evicted > 0 && aggsum_compare(&arc_sums.arcstat_size, arc_c) > 0;
	if (!arc_evict_needed) {
	/*
	* We're either no longer overflowing, or we
	* can't evict anything more, so we should wake
	* arc_get_data_impl() sooner.
	*/
	arc_evict_waiter_t *aw;
	while ((aw = list_remove_head(&arc_evict_waiters)) != NULL) {
	cv_broadcast(&aw->aew_cv);
	}
	arc_set_need_free();
	}
	mutex_exit(&arc_evict_lock);
	spl_fstrans_unmark(cookie);
	}

	static boolean_t
	arc_reap_cb_check(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	int64_t free_memory = arc_available_memory();
	static int reap_cb_check_counter = 0;

	/*
	* If a kmem reap is already active, don't schedule more. We must
	* check for this because kmem_cache_reap_soon() won't actually
	* block on the cache being reaped (this is to prevent callers from
	* becoming implicitly blocked by a system-wide kmem reap -- which,
	* on a system with many, many full magazines, can take minutes).
	*/
	if (!kmem_cache_reap_active() && free_memory < 0) {

	arc_no_grow = B_TRUE;
	arc_warm = B_TRUE;
	/*
	* Wait at least zfs_grow_retry (default 5) seconds
	* before considering growing.
	*/
	arc_growtime = gethrtime() + SEC2NSEC(arc_grow_retry);
	return (B_TRUE);
	} else if (free_memory < arc_c >> arc_no_grow_shift) {
	arc_no_grow = B_TRUE;
	} else if (gethrtime() >= arc_growtime) {
	arc_no_grow = B_FALSE;
	}

	/*
	* Called unconditionally every 60 seconds to reclaim unused
	* zstd compression and decompression context. This is done
	* here to avoid the need for an independent thread.
	*/
	if (!((reap_cb_check_counter++) % 60))
	zfs_zstd_cache_reap_now();

	return (B_FALSE);
	}

	/*
	* Keep enough free memory in the system by reaping the ARC's kmem
	* caches. To cause more slabs to be reapable, we may reduce the
	* target size of the cache (arc_c), causing the arc_evict_cb()
	* to free more buffers.
	*/
	static void
	arc_reap_cb(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	int64_t free_memory;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	/*
	* Kick off asynchronous kmem_reap()'s of all our caches.
	*/
	arc_kmem_reap_soon();

	/*
	* Wait at least arc_kmem_cache_reap_retry_ms between
	* arc_kmem_reap_soon() calls. Without this check it is possible to
	* end up in a situation where we spend lots of time reaping
	* caches, while we're near arc_c_min. Waiting here also gives the
	* subsequent free memory check a chance of finding that the
	* asynchronous reap has already freed enough memory, and we don't
	* need to call arc_reduce_target_size().
	*/
	delay((hz * arc_kmem_cache_reap_retry_ms + 999) / 1000);

	/*
	* Reduce the target size as needed to maintain the amount of free
	* memory in the system at a fraction of the arc_size (1/128th by
	* default). If oversubscribed (free_memory < 0) then reduce the
	* target arc_size by the deficit amount plus the fractional
	* amount. If free memory is positive but less than the fractional
	* amount, reduce by what is needed to hit the fractional amount.
	*/
	free_memory = arc_available_memory();

	int64_t can_free = arc_c - arc_c_min;
	if (can_free > 0) {
	int64_t to_free = (can_free >> arc_shrink_shift) - free_memory;
	if (to_free > 0)
	arc_reduce_target_size(to_free);
	}
	spl_fstrans_unmark(cookie);
	}

	#ifdef _KERNEL
	/*
	* Determine the amount of memory eligible for eviction contained in the
	* ARC. All clean data reported by the ghost lists can always be safely
	* evicted. Due to arc_c_min, the same does not hold for all clean data
	* contained by the regular mru and mfu lists.
	*
	* In the case of the regular mru and mfu lists, we need to report as
	* much clean data as possible, such that evicting that same reported
	* data will not bring arc_size below arc_c_min. Thus, in certain
	* circumstances, the total amount of clean data in the mru and mfu
	* lists might not actually be evictable.
	*
	* The following two distinct cases are accounted for:
	*
	* 1. The sum of the amount of dirty data contained by both the mru and
	* mfu lists, plus the ARC's other accounting (e.g. the anon list),
	* is greater than or equal to arc_c_min.
	* (i.e. amount of dirty data >= arc_c_min)
	*
	* This is the easy case; all clean data contained by the mru and mfu
	* lists is evictable. Evicting all clean data can only drop arc_size
	* to the amount of dirty data, which is greater than arc_c_min.
	*
	* 2. The sum of the amount of dirty data contained by both the mru and
	* mfu lists, plus the ARC's other accounting (e.g. the anon list),
	* is less than arc_c_min.
	* (i.e. arc_c_min > amount of dirty data)
	*
	* 2.1. arc_size is greater than or equal arc_c_min.
	* (i.e. arc_size >= arc_c_min > amount of dirty data)
	*
	* In this case, not all clean data from the regular mru and mfu
	* lists is actually evictable; we must leave enough clean data
	* to keep arc_size above arc_c_min. Thus, the maximum amount of
	* evictable data from the two lists combined, is exactly the
	* difference between arc_size and arc_c_min.
	*
	* 2.2. arc_size is less than arc_c_min
	* (i.e. arc_c_min > arc_size > amount of dirty data)
	*
	* In this case, none of the data contained in the mru and mfu
	* lists is evictable, even if it's clean. Since arc_size is
	* already below arc_c_min, evicting any more would only
	* increase this negative difference.
	*/

	#endif /* _KERNEL */

	/*
	* Adapt arc info given the number of bytes we are trying to add and
	* the state that we are coming from. This function is only called
	* when we are adding new content to the cache.
	*/
	static void
	arc_adapt(int bytes, arc_state_t *state)
	{
	int mult;
	uint64_t arc_p_min = (arc_c >> arc_p_min_shift);
	int64_t mrug_size = zfs_refcount_count(&arc_mru_ghost->arcs_size);
	int64_t mfug_size = zfs_refcount_count(&arc_mfu_ghost->arcs_size);

	ASSERT(bytes > 0);
	/*
	* Adapt the target size of the MRU list:
	* - if we just hit in the MRU ghost list, then increase
	* the target size of the MRU list.
	* - if we just hit in the MFU ghost list, then increase
	* the target size of the MFU list by decreasing the
	* target size of the MRU list.
	*/
	if (state == arc_mru_ghost) {
	mult = (mrug_size >= mfug_size) ? 1 : (mfug_size / mrug_size);
	if (!zfs_arc_p_dampener_disable)
	mult = MIN(mult, 10); /* avoid wild arc_p adjustment */

	arc_p = MIN(arc_c - arc_p_min, arc_p + bytes * mult);
	} else if (state == arc_mfu_ghost) {
	uint64_t delta;

	mult = (mfug_size >= mrug_size) ? 1 : (mrug_size / mfug_size);
	if (!zfs_arc_p_dampener_disable)
	mult = MIN(mult, 10);

	delta = MIN(bytes * mult, arc_p);
	arc_p = MAX(arc_p_min, arc_p - delta);
	}
	ASSERT((int64_t)arc_p >= 0);

	/*
	* Wake reap thread if we do not have any available memory
	*/
	if (arc_reclaim_needed()) {
	zthr_wakeup(arc_reap_zthr);
	return;
	}

	if (arc_no_grow)
	return;

	if (arc_c >= arc_c_max)
	return;

	/*
	* If we're within (2 * maxblocksize) bytes of the target
	* cache size, increment the target cache size
	*/
	ASSERT3U(arc_c, >=, 2ULL << SPA_MAXBLOCKSHIFT);
	if (aggsum_upper_bound(&arc_sums.arcstat_size) >=
	arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) {
	atomic_add_64(&arc_c, (int64_t)bytes);
	if (arc_c > arc_c_max)
	arc_c = arc_c_max;
	else if (state == arc_anon)
	atomic_add_64(&arc_p, (int64_t)bytes);
	if (arc_p > arc_c)
	arc_p = arc_c;
	}
	ASSERT((int64_t)arc_p >= 0);
	}

	/*
	* Check if arc_size has grown past our upper threshold, determined by
	* zfs_arc_overflow_shift.
	*/
	static arc_ovf_level_t
	arc_is_overflowing(boolean_t use_reserve)
	{
	/* Always allow at least one block of overflow */
	int64_t overflow = MAX(SPA_MAXBLOCKSIZE,
	arc_c >> zfs_arc_overflow_shift);

	/*
	* We just compare the lower bound here for performance reasons. Our
	* primary goals are to make sure that the arc never grows without
	* bound, and that it can reach its maximum size. This check
	* accomplishes both goals. The maximum amount we could run over by is
	* 2 * aggsum_borrow_multiplier * NUM_CPUS * the average size of a block
	* in the ARC. In practice, that's in the tens of MB, which is low
	* enough to be safe.
	*/
	int64_t over = aggsum_lower_bound(&arc_sums.arcstat_size) -
	arc_c - overflow / 2;
	if (!use_reserve)
	overflow /= 2;
	return (over < 0 ? ARC_OVF_NONE :
	over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE);
	}

	static abd_t *
	arc_get_data_abd(arc_buf_hdr_t hdr, uint64_t size, const void tag,
	int alloc_flags)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_get_data_impl(hdr, size, tag, alloc_flags);
	if (type == ARC_BUFC_METADATA) {
	return (abd_alloc(size, B_TRUE));
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	return (abd_alloc(size, B_FALSE));
	}
	}

	static void *
	arc_get_data_buf(arc_buf_hdr_t hdr, uint64_t size, const void tag)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_get_data_impl(hdr, size, tag, ARC_HDR_DO_ADAPT);
	if (type == ARC_BUFC_METADATA) {
	return (zio_buf_alloc(size));
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	return (zio_data_buf_alloc(size));
	}
	}

	/*
	* Wait for the specified amount of data (in bytes) to be evicted from the
	* ARC, and for there to be sufficient free memory in the system. Waiting for
	* eviction ensures that the memory used by the ARC decreases. Waiting for
	* free memory ensures that the system won't run out of free pages, regardless
	* of ARC behavior and settings. See arc_lowmem_init().
	*/
	void
	arc_wait_for_eviction(uint64_t amount, boolean_t use_reserve)
	{
	switch (arc_is_overflowing(use_reserve)) {
	case ARC_OVF_NONE:
	return;
	case ARC_OVF_SOME:
	/*
	* This is a bit racy without taking arc_evict_lock, but the
	* worst that can happen is we either call zthr_wakeup() extra
	* time due to race with other thread here, or the set flag
	* get cleared by arc_evict_cb(), which is unlikely due to
	* big hysteresis, but also not important since at this level
	* of overflow the eviction is purely advisory. Same time
	* taking the global lock here every time without waiting for
	* the actual eviction creates a significant lock contention.
	*/
	if (!arc_evict_needed) {
	arc_evict_needed = B_TRUE;
	zthr_wakeup(arc_evict_zthr);
	}
	return;
	case ARC_OVF_SEVERE:
	default:
	{
	arc_evict_waiter_t aw;
	list_link_init(&aw.aew_node);
	cv_init(&aw.aew_cv, NULL, CV_DEFAULT, NULL);

	uint64_t last_count = 0;
	mutex_enter(&arc_evict_lock);
	if (!list_is_empty(&arc_evict_waiters)) {
	arc_evict_waiter_t *last =
	list_tail(&arc_evict_waiters);
	last_count = last->aew_count;
	} else if (!arc_evict_needed) {
	arc_evict_needed = B_TRUE;
	zthr_wakeup(arc_evict_zthr);
	}
	/*
	* Note, the last waiter's count may be less than
	* arc_evict_count if we are low on memory in which
	* case arc_evict_state_impl() may have deferred
	* wakeups (but still incremented arc_evict_count).
	*/
	aw.aew_count = MAX(last_count, arc_evict_count) + amount;

	list_insert_tail(&arc_evict_waiters, &aw);

	arc_set_need_free();

	DTRACE_PROBE3(arc__wait__for__eviction,
	uint64_t, amount,
	uint64_t, arc_evict_count,
	uint64_t, aw.aew_count);

	/*
	* We will be woken up either when arc_evict_count reaches
	* aew_count, or when the ARC is no longer overflowing and
	* eviction completes.
	* In case of "false" wakeup, we will still be on the list.
	*/
	do {
	cv_wait(&aw.aew_cv, &arc_evict_lock);
	} while (list_link_active(&aw.aew_node));
	mutex_exit(&arc_evict_lock);

	cv_destroy(&aw.aew_cv);
	}
	}
	}

	/*
	* Allocate a block and return it to the caller. If we are hitting the
	* hard limit for the cache size, we must sleep, waiting for the eviction
	* thread to catch up. If we're past the target size but below the hard
	* limit, we'll only signal the reclaim thread and continue on.
	*/
	static void
	arc_get_data_impl(arc_buf_hdr_t hdr, uint64_t size, const void tag,
	int alloc_flags)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;
	arc_buf_contents_t type = arc_buf_type(hdr);

	if (alloc_flags & ARC_HDR_DO_ADAPT)
	arc_adapt(size, state);

	/*
	* If arc_size is currently overflowing, we must be adding data
	* faster than we are evicting. To ensure we don't compound the
	* problem by adding more data and forcing arc_size to grow even
	* further past it's target size, we wait for the eviction thread to
	* make some progress. We also wait for there to be sufficient free
	* memory in the system, as measured by arc_free_memory().
	*
	* Specifically, we wait for zfs_arc_eviction_pct percent of the
	* requested size to be evicted. This should be more than 100%, to
	* ensure that that progress is also made towards getting arc_size
	* under arc_c. See the comment above zfs_arc_eviction_pct.
	*/
	arc_wait_for_eviction(size * zfs_arc_eviction_pct / 100,
	alloc_flags & ARC_HDR_USE_RESERVE);

	VERIFY3U(hdr->b_type, ==, type);
	if (type == ARC_BUFC_METADATA) {
	arc_space_consume(size, ARC_SPACE_META);
	} else {
	arc_space_consume(size, ARC_SPACE_DATA);
	}

	/*
	* Update the state size. Note that ghost states have a
	* "ghost size" and so don't need to be updated.
	*/
	if (!GHOST_STATE(state)) {

	(void) zfs_refcount_add_many(&state->arcs_size, size, tag);

	/*
	* If this is reached via arc_read, the link is
	* protected by the hash lock. If reached via
	* arc_buf_alloc, the header should not be accessed by
	* any other thread. And, if reached via arc_read_done,
	* the hash lock will protect it if it's found in the
	* hash table; otherwise no other thread should be
	* trying to [add\|remove]_reference it.
	*/
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	size, tag);
	}

	/*
	* If we are growing the cache, and we are adding anonymous
	* data, and we have outgrown arc_p, update arc_p
	*/
	if (aggsum_upper_bound(&arc_sums.arcstat_size) < arc_c &&
	hdr->b_l1hdr.b_state == arc_anon &&
	(zfs_refcount_count(&arc_anon->arcs_size) +
	zfs_refcount_count(&arc_mru->arcs_size) > arc_p))
	arc_p = MIN(arc_c, arc_p + size);
	}
	}

	static void
	arc_free_data_abd(arc_buf_hdr_t hdr, abd_t abd, uint64_t size,
	const void *tag)
	{
	arc_free_data_impl(hdr, size, tag);
	abd_free(abd);
	}

	static void
	arc_free_data_buf(arc_buf_hdr_t hdr, void buf, uint64_t size, const void *tag)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_free_data_impl(hdr, size, tag);
	if (type == ARC_BUFC_METADATA) {
	zio_buf_free(buf, size);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	zio_data_buf_free(buf, size);
	}
	}

	/*
	* Free the arc data buffer.
	*/
	static void
	arc_free_data_impl(arc_buf_hdr_t hdr, uint64_t size, const void tag)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;
	arc_buf_contents_t type = arc_buf_type(hdr);

	/* protected by hash lock, if in the hash table */
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT(state != arc_anon && state != arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	size, tag);
	}
	(void) zfs_refcount_remove_many(&state->arcs_size, size, tag);

	VERIFY3U(hdr->b_type, ==, type);
	if (type == ARC_BUFC_METADATA) {
	arc_space_return(size, ARC_SPACE_META);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	arc_space_return(size, ARC_SPACE_DATA);
	}
	}

	/*
	* This routine is called whenever a buffer is accessed.
	* NOTE: the hash lock is dropped in this function.
	*/
	static void
	arc_access(arc_buf_hdr_t hdr, kmutex_t hash_lock)
	{
	clock_t now;

	ASSERT(MUTEX_HELD(hash_lock));
	ASSERT(HDR_HAS_L1HDR(hdr));

	if (hdr->b_l1hdr.b_state == arc_anon) {
	/*
	* This buffer is not in the cache, and does not
	* appear in our "ghost" list. Add the new buffer
	* to the MRU state.
	*/

	ASSERT0(hdr->b_l1hdr.b_arc_access);
	hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
	DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mru, hdr, hash_lock);

	} else if (hdr->b_l1hdr.b_state == arc_mru) {
	now = ddi_get_lbolt();

	/*
	* If this buffer is here because of a prefetch, then either:
	* - clear the flag if this is a "referencing" read
	* (any subsequent access will bump this into the MFU state).
	* or
	* - move the buffer to the head of the list if this is
	* another prefetch (to make it less likely to be evicted).
	*/
	if (HDR_PREFETCH(hdr) \|\| HDR_PRESCIENT_PREFETCH(hdr)) {
	if (zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) {
	/* link protected by hash lock */
	ASSERT(multilist_link_active(
	&hdr->b_l1hdr.b_arc_node));
	} else {
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PREFETCH \|
	ARC_FLAG_PRESCIENT_PREFETCH);
	hdr->b_l1hdr.b_mru_hits++;
	ARCSTAT_BUMP(arcstat_mru_hits);
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	hdr->b_l1hdr.b_arc_access = now;
	return;
	}

	/*
	* This buffer has been "accessed" only once so far,
	* but it is still in the cache. Move it to the MFU
	* state.
	*/
	if (ddi_time_after(now, hdr->b_l1hdr.b_arc_access +
	ARC_MINTIME)) {
	/*
	* More than 125ms have passed since we
	* instantiated this buffer. Move it to the
	* most frequently used state.
	*/
	hdr->b_l1hdr.b_arc_access = now;
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mfu, hdr, hash_lock);
	}
	hdr->b_l1hdr.b_mru_hits++;
	ARCSTAT_BUMP(arcstat_mru_hits);
	} else if (hdr->b_l1hdr.b_state == arc_mru_ghost) {
	arc_state_t *new_state;
	/*
	* This buffer has been "accessed" recently, but
	* was evicted from the cache. Move it to the
	* MFU state.
	*/
	if (HDR_PREFETCH(hdr) \|\| HDR_PRESCIENT_PREFETCH(hdr)) {
	new_state = arc_mru;
	if (zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) > 0) {
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PREFETCH \|
	ARC_FLAG_PRESCIENT_PREFETCH);
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
	} else {
	new_state = arc_mfu;
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	}

	hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
	arc_change_state(new_state, hdr, hash_lock);

	hdr->b_l1hdr.b_mru_ghost_hits++;
	ARCSTAT_BUMP(arcstat_mru_ghost_hits);
	} else if (hdr->b_l1hdr.b_state == arc_mfu) {
	/*
	* This buffer has been accessed more than once and is
	* still in the cache. Keep it in the MFU state.
	*
	* NOTE: an add_reference() that occurred when we did
	* the arc_read() will have kicked this off the list.
	* If it was a prefetch, we will explicitly move it to
	* the head of the list now.
	*/

	hdr->b_l1hdr.b_mfu_hits++;
	ARCSTAT_BUMP(arcstat_mfu_hits);
	hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
	} else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) {
	arc_state_t *new_state = arc_mfu;
	/*
	* This buffer has been accessed more than once but has
	* been evicted from the cache. Move it back to the
	* MFU state.
	*/

	if (HDR_PREFETCH(hdr) \|\| HDR_PRESCIENT_PREFETCH(hdr)) {
	/*
	* This is a prefetch access...
	* move this block back to the MRU state.
	*/
	new_state = arc_mru;
	}

	hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	arc_change_state(new_state, hdr, hash_lock);

	hdr->b_l1hdr.b_mfu_ghost_hits++;
	ARCSTAT_BUMP(arcstat_mfu_ghost_hits);
	} else if (hdr->b_l1hdr.b_state == arc_l2c_only) {
	/*
	* This buffer is on the 2nd Level ARC.
	*/

	hdr->b_l1hdr.b_arc_access = ddi_get_lbolt();
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mfu, hdr, hash_lock);
	} else {
	cmn_err(CE_PANIC, "invalid arc state 0x%p",
	hdr->b_l1hdr.b_state);
	}
	}

	/*
	* This routine is called by dbuf_hold() to update the arc_access() state
	* which otherwise would be skipped for entries in the dbuf cache.
	*/
	void
	arc_buf_access(arc_buf_t *buf)
	{
	mutex_enter(&buf->b_evict_lock);
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* Avoid taking the hash_lock when possible as an optimization.
	* The header must be checked again under the hash_lock in order
	* to handle the case where it is concurrently being released.
	*/
	if (hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY(hdr)) {
	mutex_exit(&buf->b_evict_lock);
	return;
	}

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	if (hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY(hdr)) {
	mutex_exit(hash_lock);
	mutex_exit(&buf->b_evict_lock);
	ARCSTAT_BUMP(arcstat_access_skip);
	return;
	}

	mutex_exit(&buf->b_evict_lock);

	ASSERT(hdr->b_l1hdr.b_state == arc_mru \|\|
	hdr->b_l1hdr.b_state == arc_mfu);

	DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
	arc_access(hdr, hash_lock);
	mutex_exit(hash_lock);

	ARCSTAT_BUMP(arcstat_hits);
	ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr) && !HDR_PRESCIENT_PREFETCH(hdr),
	demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits);
	}

	/* a generic arc_read_done_func_t which you can use */
	void
	arc_bcopy_func(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void arg)
	{
	(void) zio, (void) zb, (void) bp;

	if (buf == NULL)
	return;

	memcpy(arg, buf->b_data, arc_buf_size(buf));
	arc_buf_destroy(buf, arg);
	}

	/* a generic arc_read_done_func_t */
	void
	arc_getbuf_func(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void arg)
	{
	(void) zb, (void) bp;
	arc_buf_t **bufp = arg;

	if (buf == NULL) {
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	*bufp = NULL;
	} else {
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	*bufp = buf;
	ASSERT(buf->b_data != NULL);
	}
	}

	static void
	arc_hdr_verify(arc_buf_hdr_t hdr, blkptr_t bp)
	{
	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp)) {
	ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0);
	ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF);
	} else {
	if (HDR_COMPRESSION_ENABLED(hdr)) {
	ASSERT3U(arc_hdr_get_compress(hdr), ==,
	BP_GET_COMPRESS(bp));
	}
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
	ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp));
	ASSERT3U(!!HDR_PROTECTED(hdr), ==, BP_IS_PROTECTED(bp));
	}
	}

	static void
	arc_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	arc_buf_hdr_t *hdr = zio->io_private;
	kmutex_t *hash_lock = NULL;
	arc_callback_t *callback_list;
	arc_callback_t *acb;
	boolean_t freeable = B_FALSE;

	/*
	* The hdr was inserted into hash-table and removed from lists
	* prior to starting I/O. We should find this header, since
	* it's in the hash table, and it should be legit since it's
	* not possible to evict it during the I/O. The only possible
	* reason for it not to be found is if we were freed during the
	* read.
	*/
	if (HDR_IN_HASH_TABLE(hdr)) {
	arc_buf_hdr_t *found;

	ASSERT3U(hdr->b_birth, ==, BP_PHYSICAL_BIRTH(zio->io_bp));
	ASSERT3U(hdr->b_dva.dva_word[0], ==,
	BP_IDENTITY(zio->io_bp)->dva_word[0]);
	ASSERT3U(hdr->b_dva.dva_word[1], ==,
	BP_IDENTITY(zio->io_bp)->dva_word[1]);

	found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock);

	ASSERT((found == hdr &&
	DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) \|\|
	(found == hdr && HDR_L2_READING(hdr)));
	ASSERT3P(hash_lock, !=, NULL);
	}

	if (BP_IS_PROTECTED(bp)) {
	hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
	hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
	zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv);

	if (zio->io_error == 0) {
	if (BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG) {
	void *tmpbuf;

	tmpbuf = abd_borrow_buf_copy(zio->io_abd,
	sizeof (zil_chain_t));
	zio_crypt_decode_mac_zil(tmpbuf,
	hdr->b_crypt_hdr.b_mac);
	abd_return_buf(zio->io_abd, tmpbuf,
	sizeof (zil_chain_t));
	} else {
	zio_crypt_decode_mac_bp(bp,
	hdr->b_crypt_hdr.b_mac);
	}
	}
	}

	if (zio->io_error == 0) {
	/* byteswap if necessary */
	if (BP_SHOULD_BYTESWAP(zio->io_bp)) {
	if (BP_GET_LEVEL(zio->io_bp) > 0) {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
	} else {
	hdr->b_l1hdr.b_byteswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp));
	}
	} else {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	}
	if (!HDR_L2_READING(hdr)) {
	hdr->b_complevel = zio->io_prop.zp_complevel;
	}
	}

	arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED);
	if (l2arc_noprefetch && HDR_PREFETCH(hdr))
	arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE);

	callback_list = hdr->b_l1hdr.b_acb;
	ASSERT3P(callback_list, !=, NULL);

	if (hash_lock && zio->io_error == 0 &&
	hdr->b_l1hdr.b_state == arc_anon) {
	/*
	* Only call arc_access on anonymous buffers. This is because
	* if we've issued an I/O for an evicted buffer, we've already
	* called arc_access (to prevent any simultaneous readers from
	* getting confused).
	*/
	arc_access(hdr, hash_lock);
	}

	/*
	* If a read request has a callback (i.e. acb_done is not NULL), then we
	* make a buf containing the data according to the parameters which were
	* passed in. The implementation of arc_buf_alloc_impl() ensures that we
	* aren't needlessly decompressing the data multiple times.
	*/
	int callback_cnt = 0;
	for (acb = callback_list; acb != NULL; acb = acb->acb_next) {
	if (!acb->acb_done \|\| acb->acb_nobuf)
	continue;

	callback_cnt++;

	if (zio->io_error != 0)
	continue;

	int error = arc_buf_alloc_impl(hdr, zio->io_spa,
	&acb->acb_zb, acb->acb_private, acb->acb_encrypted,
	acb->acb_compressed, acb->acb_noauth, B_TRUE,
	&acb->acb_buf);

	/*
	* Assert non-speculative zios didn't fail because an
	* encryption key wasn't loaded
	*/
	ASSERT((zio->io_flags & ZIO_FLAG_SPECULATIVE) \|\|
	error != EACCES);

	/*
	* If we failed to decrypt, report an error now (as the zio
	* layer would have done if it had done the transforms).
	*/
	if (error == ECKSUM) {
	ASSERT(BP_IS_PROTECTED(bp));
	error = SET_ERROR(EIO);
	if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(zio->io_spa, &acb->acb_zb);
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_AUTHENTICATION,
	zio->io_spa, NULL, &acb->acb_zb, zio, 0);
	}
	}

	if (error != 0) {
	/*
	* Decompression or decryption failed. Set
	* io_error so that when we call acb_done
	* (below), we will indicate that the read
	* failed. Note that in the unusual case
	* where one callback is compressed and another
	* uncompressed, we will mark all of them
	* as failed, even though the uncompressed
	* one can't actually fail. In this case,
	* the hdr will not be anonymous, because
	* if there are multiple callbacks, it's
	* because multiple threads found the same
	* arc buf in the hash table.
	*/
	zio->io_error = error;
	}
	}

	/*
	* If there are multiple callbacks, we must have the hash lock,
	* because the only way for multiple threads to find this hdr is
	* in the hash table. This ensures that if there are multiple
	* callbacks, the hdr is not anonymous. If it were anonymous,
	* we couldn't use arc_buf_destroy() in the error case below.
	*/
	ASSERT(callback_cnt < 2 \|\| hash_lock != NULL);

	hdr->b_l1hdr.b_acb = NULL;
	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	if (callback_cnt == 0)
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));

	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt) \|\|
	callback_list != NULL);

	if (zio->io_error == 0) {
	arc_hdr_verify(hdr, zio->io_bp);
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hdr->b_l1hdr.b_state != arc_anon)
	arc_change_state(arc_anon, hdr, hash_lock);
	if (HDR_IN_HASH_TABLE(hdr))
	buf_hash_remove(hdr);
	freeable = zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt);
	}

	/*
	* Broadcast before we drop the hash_lock to avoid the possibility
	* that the hdr (and hence the cv) might be freed before we get to
	* the cv_broadcast().
	*/
	cv_broadcast(&hdr->b_l1hdr.b_cv);

	if (hash_lock != NULL) {
	mutex_exit(hash_lock);
	} else {
	/*
	* This block was freed while we waited for the read to
	* complete. It has been removed from the hash table and
	* moved to the anonymous state (so that it won't show up
	* in the cache).
	*/
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	freeable = zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt);
	}

	/* execute each callback and free its structure */
	while ((acb = callback_list) != NULL) {
	if (acb->acb_done != NULL) {
	if (zio->io_error != 0 && acb->acb_buf != NULL) {
	/*
	* If arc_buf_alloc_impl() fails during
	* decompression, the buf will still be
	* allocated, and needs to be freed here.
	*/
	arc_buf_destroy(acb->acb_buf,
	acb->acb_private);
	acb->acb_buf = NULL;
	}
	acb->acb_done(zio, &zio->io_bookmark, zio->io_bp,
	acb->acb_buf, acb->acb_private);
	}

	if (acb->acb_zio_dummy != NULL) {
	acb->acb_zio_dummy->io_error = zio->io_error;
	zio_nowait(acb->acb_zio_dummy);
	}

	callback_list = acb->acb_next;
	kmem_free(acb, sizeof (arc_callback_t));
	}

	if (freeable)
	arc_hdr_destroy(hdr);
	}

	/*
	* "Read" the block at the specified DVA (in bp) via the
	* cache. If the block is found in the cache, invoke the provided
	* callback immediately and return. Note that the `zio' parameter
	* in the callback will be NULL in this case, since no IO was
	* required. If the block is not in the cache pass the read request
	* on to the spa with a substitute callback function, so that the
	* requested block will be added to the cache.
	*
	* If a read request arrives for a block that has a read in-progress,
	* either wait for the in-progress read to complete (and return the
	* results); or, if this is a read with a "done" func, add a record
	* to the read to invoke the "done" func when the read completes,
	* and return; or just return.
	*
	* arc_read_done() will invoke all the requested "done" functions
	* for readers of this block.
	*/
	int
	arc_read(zio_t pio, spa_t spa, const blkptr_t *bp,
	arc_read_done_func_t done, void private, zio_priority_t priority,
	int zio_flags, arc_flags_t arc_flags, const zbookmark_phys_t zb)
	{
	arc_buf_hdr_t *hdr = NULL;
	kmutex_t *hash_lock = NULL;
	zio_t *rzio;
	uint64_t guid = spa_load_guid(spa);
	boolean_t compressed_read = (zio_flags & ZIO_FLAG_RAW_COMPRESS) != 0;
	boolean_t encrypted_read = BP_IS_ENCRYPTED(bp) &&
	(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
	boolean_t noauth_read = BP_IS_AUTHENTICATED(bp) &&
	(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
	boolean_t embedded_bp = !!BP_IS_EMBEDDED(bp);
	boolean_t no_buf = *arc_flags & ARC_FLAG_NO_BUF;
	int rc = 0;

	ASSERT(!embedded_bp \|\|
	BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_REDACTED(bp));

	/*
	* Normally SPL_FSTRANS will already be set since kernel threads which
	* expect to call the DMU interfaces will set it when created. System
	* calls are similarly handled by setting/cleaning the bit in the
	* registered callback (module/os/.../zfs/zpl_*).
	*
	* External consumers such as Lustre which call the exported DMU
	* interfaces may not have set SPL_FSTRANS. To avoid a deadlock
	* on the hash_lock always set and clear the bit.
	*/
	fstrans_cookie_t cookie = spl_fstrans_mark();
	top:
	/*
	* Verify the block pointer contents are reasonable. This should
	* always be the case since the blkptr is protected by a checksum.
	* However, if there is damage it's desirable to detect this early
	* and treat it as a checksum error. This allows an alternate blkptr
	* to be tried when one is available (e.g. ditto blocks).
	*/
	if (!zfs_blkptr_verify(spa, bp, zio_flags & ZIO_FLAG_CONFIG_WRITER,
	BLK_VERIFY_LOG)) {
	rc = SET_ERROR(ECKSUM);
	goto out;
	}

	if (!embedded_bp) {
	/*
	* Embedded BP's have no DVA and require no I/O to "read".
	* Create an anonymous arc buf to back it.
	*/
	hdr = buf_hash_find(guid, bp, &hash_lock);
	}

	/*
	* Determine if we have an L1 cache hit or a cache miss. For simplicity
	* we maintain encrypted data separately from compressed / uncompressed
	* data. If the user is requesting raw encrypted data and we don't have
	* that in the header we will read from disk to guarantee that we can
	* get it even if the encryption keys aren't loaded.
	*/
	if (hdr != NULL && HDR_HAS_L1HDR(hdr) && (HDR_HAS_RABD(hdr) \|\|
	(hdr->b_l1hdr.b_pabd != NULL && !encrypted_read))) {
	arc_buf_t *buf = NULL;
	*arc_flags \|= ARC_FLAG_CACHED;

	if (HDR_IO_IN_PROGRESS(hdr)) {
	zio_t *head_zio = hdr->b_l1hdr.b_acb->acb_zio_head;

	if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
	mutex_exit(hash_lock);
	ARCSTAT_BUMP(arcstat_cached_only_in_progress);
	rc = SET_ERROR(ENOENT);
	goto out;
	}

	ASSERT3P(head_zio, !=, NULL);
	if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) &&
	priority == ZIO_PRIORITY_SYNC_READ) {
	/*
	* This is a sync read that needs to wait for
	* an in-flight async read. Request that the
	* zio have its priority upgraded.
	*/
	zio_change_priority(head_zio, priority);
	DTRACE_PROBE1(arc__async__upgrade__sync,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_async_upgrade_sync);
	}
	if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) {
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PREDICTIVE_PREFETCH);
	}

	/*
	* If there are multiple threads reading the same block
	* and that block is not yet in the ARC, then only one
	* thread will do the physical I/O and all other
	* threads will wait until that I/O completes.
	* Synchronous reads use the b_cv whereas nowait reads
	* register a callback. Both are signalled/called in
	* arc_read_done.
	*
	* Errors of the physical I/O may need to be propagated
	* to the pio. For synchronous reads, we simply restart
	* this function and it will reassess. Nowait reads
	* attach the acb_zio_dummy zio to pio and
	* arc_read_done propagates the physical I/O's io_error
	* to acb_zio_dummy, and thereby to pio.
	*/

	if (*arc_flags & ARC_FLAG_WAIT) {
	cv_wait(&hdr->b_l1hdr.b_cv, hash_lock);
	mutex_exit(hash_lock);
	goto top;
	}
	ASSERT(*arc_flags & ARC_FLAG_NOWAIT);

	if (done) {
	arc_callback_t *acb = NULL;

	acb = kmem_zalloc(sizeof (arc_callback_t),
	KM_SLEEP);
	acb->acb_done = done;
	acb->acb_private = private;
	acb->acb_compressed = compressed_read;
	acb->acb_encrypted = encrypted_read;
	acb->acb_noauth = noauth_read;
	acb->acb_nobuf = no_buf;
	acb->acb_zb = *zb;
	if (pio != NULL)
	acb->acb_zio_dummy = zio_null(pio,
	spa, NULL, NULL, NULL, zio_flags);

	ASSERT3P(acb->acb_done, !=, NULL);
	acb->acb_zio_head = head_zio;
	acb->acb_next = hdr->b_l1hdr.b_acb;
	hdr->b_l1hdr.b_acb = acb;
	}
	mutex_exit(hash_lock);
	goto out;
	}

	ASSERT(hdr->b_l1hdr.b_state == arc_mru \|\|
	hdr->b_l1hdr.b_state == arc_mfu);

	if (done && !no_buf) {
	if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) {
	/*
	* This is a demand read which does not have to
	* wait for i/o because we did a predictive
	* prefetch i/o for it, which has completed.
	*/
	DTRACE_PROBE1(
	arc__demand__hit__predictive__prefetch,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(
	arcstat_demand_hit_predictive_prefetch);
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PREDICTIVE_PREFETCH);
	}

	if (hdr->b_flags & ARC_FLAG_PRESCIENT_PREFETCH) {
	ARCSTAT_BUMP(
	arcstat_demand_hit_prescient_prefetch);
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PRESCIENT_PREFETCH);
	}

	ASSERT(!embedded_bp \|\| !BP_IS_HOLE(bp));

	/* Get a buf with the desired data in it. */
	rc = arc_buf_alloc_impl(hdr, spa, zb, private,
	encrypted_read, compressed_read, noauth_read,
	B_TRUE, &buf);
	if (rc == ECKSUM) {
	/*
	* Convert authentication and decryption errors
	* to EIO (and generate an ereport if needed)
	* before leaving the ARC.
	*/
	rc = SET_ERROR(EIO);
	if ((zio_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(spa, zb);
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, zb, NULL, 0);
	}
	}
	if (rc != 0) {
	(void) remove_reference(hdr, hash_lock,
	private);
	arc_buf_destroy_impl(buf);
	buf = NULL;
	}

	/* assert any errors weren't due to unloaded keys */
	ASSERT((zio_flags & ZIO_FLAG_SPECULATIVE) \|\|
	rc != EACCES);
	} else if (*arc_flags & ARC_FLAG_PREFETCH &&
	zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
	arc_access(hdr, hash_lock);
	if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH)
	arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH);
	if (*arc_flags & ARC_FLAG_L2CACHE)
	arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
	mutex_exit(hash_lock);
	ARCSTAT_BUMP(arcstat_hits);
	ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
	demand, prefetch, !HDR_ISTYPE_METADATA(hdr),
	data, metadata, hits);

	if (done)
	done(NULL, zb, bp, buf, private);
	} else {
	uint64_t lsize = BP_GET_LSIZE(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	arc_callback_t *acb;
	vdev_t *vd = NULL;
	uint64_t addr = 0;
	boolean_t devw = B_FALSE;
	uint64_t size;
	abd_t *hdr_abd;
	int alloc_flags = encrypted_read ? ARC_HDR_ALLOC_RDATA : 0;

	if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
	rc = SET_ERROR(ENOENT);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	goto out;
	}

	if (hdr == NULL) {
	/*
	* This block is not in the cache or it has
	* embedded data.
	*/
	arc_buf_hdr_t *exists = NULL;
	arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);
	hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
	BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), 0, type);

	if (!embedded_bp) {
	hdr->b_dva = *BP_IDENTITY(bp);
	hdr->b_birth = BP_PHYSICAL_BIRTH(bp);
	exists = buf_hash_insert(hdr, &hash_lock);
	}
	if (exists != NULL) {
	/* somebody beat us to the hash insert */
	mutex_exit(hash_lock);
	buf_discard_identity(hdr);
	arc_hdr_destroy(hdr);
	goto top; /* restart the IO request */
	}
	alloc_flags \|= ARC_HDR_DO_ADAPT;
	} else {
	/*
	* This block is in the ghost cache or encrypted data
	* was requested and we didn't have it. If it was
	* L2-only (and thus didn't have an L1 hdr),
	* we realloc the header to add an L1 hdr.
	*/
	if (!HDR_HAS_L1HDR(hdr)) {
	hdr = arc_hdr_realloc(hdr, hdr_l2only_cache,
	hdr_full_cache);
	}

	if (GHOST_STATE(hdr->b_l1hdr.b_state)) {
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT0(zfs_refcount_count(
	&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	} else if (HDR_IO_IN_PROGRESS(hdr)) {
	/*
	* If this header already had an IO in progress
	* and we are performing another IO to fetch
	* encrypted data we must wait until the first
	* IO completes so as not to confuse
	* arc_read_done(). This should be very rare
	* and so the performance impact shouldn't
	* matter.
	*/
	cv_wait(&hdr->b_l1hdr.b_cv, hash_lock);
	mutex_exit(hash_lock);
	goto top;
	}

	/*
	* This is a delicate dance that we play here.
	* This hdr might be in the ghost list so we access
	* it to move it out of the ghost list before we
	* initiate the read. If it's a prefetch then
	* it won't have a callback so we'll remove the
	* reference that arc_buf_alloc_impl() created. We
	* do this after we've called arc_access() to
	* avoid hitting an assert in remove_reference().
	*/
	arc_adapt(arc_hdr_size(hdr), hdr->b_l1hdr.b_state);
	arc_access(hdr, hash_lock);
	}

	arc_hdr_alloc_abd(hdr, alloc_flags);
	if (encrypted_read) {
	ASSERT(HDR_HAS_RABD(hdr));
	size = HDR_GET_PSIZE(hdr);
	hdr_abd = hdr->b_crypt_hdr.b_rabd;
	zio_flags \|= ZIO_FLAG_RAW;
	} else {
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	size = arc_hdr_size(hdr);
	hdr_abd = hdr->b_l1hdr.b_pabd;

	if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	}

	/*
	* For authenticated bp's, we do not ask the ZIO layer
	* to authenticate them since this will cause the entire
	* IO to fail if the key isn't loaded. Instead, we
	* defer authentication until arc_buf_fill(), which will
	* verify the data when the key is available.
	*/
	if (BP_IS_AUTHENTICATED(bp))
	zio_flags \|= ZIO_FLAG_RAW_ENCRYPT;
	}

	if (*arc_flags & ARC_FLAG_PREFETCH &&
	zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH)
	arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH);
	if (*arc_flags & ARC_FLAG_L2CACHE)
	arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);
	if (BP_IS_AUTHENTICATED(bp))
	arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
	if (BP_GET_LEVEL(bp) > 0)
	arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT);
	if (*arc_flags & ARC_FLAG_PREDICTIVE_PREFETCH)
	arc_hdr_set_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH);
	ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state));

	acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP);
	acb->acb_done = done;
	acb->acb_private = private;
	acb->acb_compressed = compressed_read;
	acb->acb_encrypted = encrypted_read;
	acb->acb_noauth = noauth_read;
	acb->acb_zb = *zb;

	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
	hdr->b_l1hdr.b_acb = acb;
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);

	if (HDR_HAS_L2HDR(hdr) &&
	(vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) {
	devw = hdr->b_l2hdr.b_dev->l2ad_writing;
	addr = hdr->b_l2hdr.b_daddr;
	/*
	* Lock out L2ARC device removal.
	*/
	if (vdev_is_dead(vd) \|\|
	!spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER))
	vd = NULL;
	}

	/*
	* We count both async reads and scrub IOs as asynchronous so
	* that both can be upgraded in the event of a cache hit while
	* the read IO is still in-flight.
	*/
	if (priority == ZIO_PRIORITY_ASYNC_READ \|\|
	priority == ZIO_PRIORITY_SCRUB)
	arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
	else
	arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);

	/*
	* At this point, we have a level 1 cache miss or a blkptr
	* with embedded data. Try again in L2ARC if possible.
	*/
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize);

	/*
	* Skip ARC stat bump for block pointers with embedded
	* data. The data are read from the blkptr itself via
	* decode_embedded_bp_compressed().
	*/
	if (!embedded_bp) {
	DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr,
	blkptr_t *, bp, uint64_t, lsize,
	zbookmark_phys_t *, zb);
	ARCSTAT_BUMP(arcstat_misses);
	ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr),
	demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data,
	metadata, misses);
	zfs_racct_read(size, 1);
	}

	/* Check if the spa even has l2 configured */
	const boolean_t spa_has_l2 = l2arc_ndev != 0 &&
	spa->spa_l2cache.sav_count > 0;

	if (vd != NULL && spa_has_l2 && !(l2arc_norw && devw)) {
	/*
	* Read from the L2ARC if the following are true:
	* 1. The L2ARC vdev was previously cached.
	* 2. This buffer still has L2ARC metadata.
	* 3. This buffer isn't currently writing to the L2ARC.
	* 4. The L2ARC entry wasn't evicted, which may
	* also have invalidated the vdev.
	* 5. This isn't prefetch or l2arc_noprefetch is 0.
	*/
	if (HDR_HAS_L2HDR(hdr) &&
	!HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr) &&
	!(l2arc_noprefetch && HDR_PREFETCH(hdr))) {
	l2arc_read_callback_t *cb;
	abd_t *abd;
	uint64_t asize;

	DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_hits);
	hdr->b_l2hdr.b_hits++;

	cb = kmem_zalloc(sizeof (l2arc_read_callback_t),
	KM_SLEEP);
	cb->l2rcb_hdr = hdr;
	cb->l2rcb_bp = *bp;
	cb->l2rcb_zb = *zb;
	cb->l2rcb_flags = zio_flags;

	/*
	* When Compressed ARC is disabled, but the
	* L2ARC block is compressed, arc_hdr_size()
	* will have returned LSIZE rather than PSIZE.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr) &&
	HDR_GET_PSIZE(hdr) != 0) {
	size = HDR_GET_PSIZE(hdr);
	}

	asize = vdev_psize_to_asize(vd, size);
	if (asize != size) {
	abd = abd_alloc_for_io(asize,
	HDR_ISTYPE_METADATA(hdr));
	cb->l2rcb_abd = abd;
	} else {
	abd = hdr_abd;
	}

	ASSERT(addr >= VDEV_LABEL_START_SIZE &&
	addr + asize <= vd->vdev_psize -
	VDEV_LABEL_END_SIZE);

	/*
	* l2arc read. The SCL_L2ARC lock will be
	* released by l2arc_read_done().
	* Issue a null zio if the underlying buffer
	* was squashed to zero size by compression.
	*/
	ASSERT3U(arc_hdr_get_compress(hdr), !=,
	ZIO_COMPRESS_EMPTY);
	rzio = zio_read_phys(pio, vd, addr,
	asize, abd,
	ZIO_CHECKSUM_OFF,
	l2arc_read_done, cb, priority,
	zio_flags \| ZIO_FLAG_DONT_CACHE \|
	ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY, B_FALSE);
	acb->acb_zio_head = rzio;

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	DTRACE_PROBE2(l2arc__read, vdev_t *, vd,
	zio_t *, rzio);
	ARCSTAT_INCR(arcstat_l2_read_bytes,
	HDR_GET_PSIZE(hdr));

	if (*arc_flags & ARC_FLAG_NOWAIT) {
	zio_nowait(rzio);
	goto out;
	}

	ASSERT(*arc_flags & ARC_FLAG_WAIT);
	if (zio_wait(rzio) == 0)
	goto out;

	/* l2arc read error; goto zio_read() */
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	} else {
	DTRACE_PROBE1(l2arc__miss,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_misses);
	if (HDR_L2_WRITING(hdr))
	ARCSTAT_BUMP(arcstat_l2_rw_clash);
	spa_config_exit(spa, SCL_L2ARC, vd);
	}
	} else {
	if (vd != NULL)
	spa_config_exit(spa, SCL_L2ARC, vd);

	/*
	* Only a spa with l2 should contribute to l2
	* miss stats. (Including the case of having a
	* faulted cache device - that's also a miss.)
	*/
	if (spa_has_l2) {
	/*
	* Skip ARC stat bump for block pointers with
	* embedded data. The data are read from the
	* blkptr itself via
	* decode_embedded_bp_compressed().
	*/
	if (!embedded_bp) {
	DTRACE_PROBE1(l2arc__miss,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_misses);
	}
	}
	}

	rzio = zio_read(pio, spa, bp, hdr_abd, size,
	arc_read_done, hdr, priority, zio_flags, zb);
	acb->acb_zio_head = rzio;

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	if (*arc_flags & ARC_FLAG_WAIT) {
	rc = zio_wait(rzio);
	goto out;
	}

	ASSERT(*arc_flags & ARC_FLAG_NOWAIT);
	zio_nowait(rzio);
	}

	out:
	/* embedded bps don't actually go to disk */
	if (!embedded_bp)
	spa_read_history_add(spa, zb, *arc_flags);
	spl_fstrans_unmark(cookie);
	return (rc);
	}

	arc_prune_t *
	arc_add_prune_callback(arc_prune_func_t func, void private)
	{
	arc_prune_t *p;

	p = kmem_alloc(sizeof (*p), KM_SLEEP);
	p->p_pfunc = func;
	p->p_private = private;
	list_link_init(&p->p_node);
	zfs_refcount_create(&p->p_refcnt);

	mutex_enter(&arc_prune_mtx);
	zfs_refcount_add(&p->p_refcnt, &arc_prune_list);
	list_insert_head(&arc_prune_list, p);
	mutex_exit(&arc_prune_mtx);

	return (p);
	}

	void
	arc_remove_prune_callback(arc_prune_t *p)
	{
	boolean_t wait = B_FALSE;
	mutex_enter(&arc_prune_mtx);
	list_remove(&arc_prune_list, p);
	if (zfs_refcount_remove(&p->p_refcnt, &arc_prune_list) > 0)
	wait = B_TRUE;
	mutex_exit(&arc_prune_mtx);

	/* wait for arc_prune_task to finish */
	if (wait)
	taskq_wait_outstanding(arc_prune_taskq, 0);
	ASSERT0(zfs_refcount_count(&p->p_refcnt));
	zfs_refcount_destroy(&p->p_refcnt);
	kmem_free(p, sizeof (*p));
	}

	/*
	* Notify the arc that a block was freed, and thus will never be used again.
	*/
	void
	arc_freed(spa_t spa, const blkptr_t bp)
	{
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	uint64_t guid = spa_load_guid(spa);

	ASSERT(!BP_IS_EMBEDDED(bp));

	hdr = buf_hash_find(guid, bp, &hash_lock);
	if (hdr == NULL)
	return;

	/*
	* We might be trying to free a block that is still doing I/O
	* (i.e. prefetch) or has a reference (i.e. a dedup-ed,
	* dmu_sync-ed block). If this block is being prefetched, then it
	* would still have the ARC_FLAG_IO_IN_PROGRESS flag set on the hdr
	* until the I/O completes. A block may also have a reference if it is
	* part of a dedup-ed, dmu_synced write. The dmu_sync() function would
	* have written the new block to its final resting place on disk but
	* without the dedup flag set. This would have left the hdr in the MRU
	* state and discoverable. When the txg finally syncs it detects that
	* the block was overridden in open context and issues an override I/O.
	* Since this is a dedup block, the override I/O will determine if the
	* block is already in the DDT. If so, then it will replace the io_bp
	* with the bp from the DDT and allow the I/O to finish. When the I/O
	* reaches the done callback, dbuf_write_override_done, it will
	* check to see if the io_bp and io_bp_override are identical.
	* If they are not, then it indicates that the bp was replaced with
	* the bp in the DDT and the override bp is freed. This allows
	* us to arrive here with a reference on a block that is being
	* freed. So if we have an I/O in progress, or a reference to
	* this hdr, then we don't destroy the hdr.
	*/
	if (!HDR_HAS_L1HDR(hdr) \|\| (!HDR_IO_IN_PROGRESS(hdr) &&
	zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt))) {
	arc_change_state(arc_anon, hdr, hash_lock);
	arc_hdr_destroy(hdr);
	mutex_exit(hash_lock);
	} else {
	mutex_exit(hash_lock);
	}

	}

	/*
	* Release this buffer from the cache, making it an anonymous buffer. This
	* must be done after a read and prior to modifying the buffer contents.
	* If the buffer has more than one reference, we must make
	* a new hdr for the buffer.
	*/
	void
	arc_release(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* It would be nice to assert that if its DMU metadata (level >
	* 0 \|\| it's the dnode file), then it must be syncing context.
	* But we don't know that information at this level.
	*/

	mutex_enter(&buf->b_evict_lock);

	ASSERT(HDR_HAS_L1HDR(hdr));

	/*
	* We don't grab the hash lock prior to this check, because if
	* the buffer's header is in the arc_anon state, it won't be
	* linked into the hash table.
	*/
	if (hdr->b_l1hdr.b_state == arc_anon) {
	mutex_exit(&buf->b_evict_lock);
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT(!HDR_IN_HASH_TABLE(hdr));
	ASSERT(!HDR_HAS_L2HDR(hdr));

	ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);
	ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1);
	ASSERT(!list_link_active(&hdr->b_l1hdr.b_arc_node));

	hdr->b_l1hdr.b_arc_access = 0;

	/*
	* If the buf is being overridden then it may already
	* have a hdr that is not empty.
	*/
	buf_discard_identity(hdr);
	arc_buf_thaw(buf);

	return;
	}

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	/*
	* This assignment is only valid as long as the hash_lock is
	* held, we must be careful not to reference state or the
	* b_state field after dropping the lock.
	*/
	arc_state_t *state = hdr->b_l1hdr.b_state;
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
	ASSERT3P(state, !=, arc_anon);

	/* this buffer is not on any list */
	ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), >, 0);

	if (HDR_HAS_L2HDR(hdr)) {
	mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx);

	/*
	* We have to recheck this conditional again now that
	* we're holding the l2ad_mtx to prevent a race with
	* another thread which might be concurrently calling
	* l2arc_evict(). In that case, l2arc_evict() might have
	* destroyed the header's L2 portion as we were waiting
	* to acquire the l2ad_mtx.
	*/
	if (HDR_HAS_L2HDR(hdr))
	arc_hdr_l2hdr_destroy(hdr);

	mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx);
	}

	/*
	* Do we have more than one buf?
	*/
	if (hdr->b_l1hdr.b_bufcnt > 1) {
	arc_buf_hdr_t *nhdr;
	uint64_t spa = hdr->b_spa;
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	boolean_t protected = HDR_PROTECTED(hdr);
	enum zio_compress compress = arc_hdr_get_compress(hdr);
	arc_buf_contents_t type = arc_buf_type(hdr);
	VERIFY3U(hdr->b_type, ==, type);

	ASSERT(hdr->b_l1hdr.b_buf != buf \|\| buf->b_next != NULL);
	(void) remove_reference(hdr, hash_lock, tag);

	if (arc_buf_is_shared(buf) && !ARC_BUF_COMPRESSED(buf)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
	ASSERT(ARC_BUF_LAST(buf));
	}

	/*
	* Pull the data off of this hdr and attach it to
	* a new anonymous hdr. Also find the last buffer
	* in the hdr's buffer list.
	*/
	arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);
	ASSERT3P(lastbuf, !=, NULL);

	/*
	* If the current arc_buf_t and the hdr are sharing their data
	* buffer, then we must stop sharing that block.
	*/
	if (arc_buf_is_shared(buf)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
	VERIFY(!arc_buf_is_shared(lastbuf));

	/*
	* First, sever the block sharing relationship between
	* buf and the arc_buf_hdr_t.
	*/
	arc_unshare_buf(hdr, buf);

	/*
	* Now we need to recreate the hdr's b_pabd. Since we
	* have lastbuf handy, we try to share with it, but if
	* we can't then we allocate a new b_pabd and copy the
	* data from buf into it.
	*/
	if (arc_can_share(hdr, lastbuf)) {
	arc_share_buf(hdr, lastbuf);
	} else {
	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT);
	abd_copy_from_buf(hdr->b_l1hdr.b_pabd,
	buf->b_data, psize);
	}
	VERIFY3P(lastbuf->b_data, !=, NULL);
	} else if (HDR_SHARED_DATA(hdr)) {
	/*
	* Uncompressed shared buffers are always at the end
	* of the list. Compressed buffers don't have the
	* same requirements. This makes it hard to
	* simply assert that the lastbuf is shared so
	* we rely on the hdr's compression flags to determine
	* if we have a compressed, shared buffer.
	*/
	ASSERT(arc_buf_is_shared(lastbuf) \|\|
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	ASSERT(!ARC_BUF_SHARED(buf));
	}

	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));
	ASSERT3P(state, !=, arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_size,
	arc_buf_size(buf), buf);

	if (zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
	ASSERT3P(state, !=, arc_l2c_only);
	(void) zfs_refcount_remove_many(
	&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}

	hdr->b_l1hdr.b_bufcnt -= 1;
	if (ARC_BUF_ENCRYPTED(buf))
	hdr->b_crypt_hdr.b_ebufcnt -= 1;

	arc_cksum_verify(buf);
	arc_buf_unwatch(buf);

	/* if this is the last uncompressed buf free the checksum */
	if (!arc_hdr_has_uncompressed_buf(hdr))
	arc_cksum_free(hdr);

	mutex_exit(hash_lock);

	/*
	* Allocate a new hdr. The new hdr will contain a b_pabd
	* buffer which will be freed in arc_write().
	*/
	nhdr = arc_hdr_alloc(spa, psize, lsize, protected,
	compress, hdr->b_complevel, type);
	ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT0(nhdr->b_l1hdr.b_bufcnt);
	ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt));
	VERIFY3U(nhdr->b_type, ==, type);
	ASSERT(!HDR_SHARED_DATA(nhdr));

	nhdr->b_l1hdr.b_buf = buf;
	nhdr->b_l1hdr.b_bufcnt = 1;
	if (ARC_BUF_ENCRYPTED(buf))
	nhdr->b_crypt_hdr.b_ebufcnt = 1;
	(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, tag);
	buf->b_hdr = nhdr;

	mutex_exit(&buf->b_evict_lock);
	(void) zfs_refcount_add_many(&arc_anon->arcs_size,
	arc_buf_size(buf), buf);
	} else {
	mutex_exit(&buf->b_evict_lock);
	ASSERT(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 1);
	/* protected by hash lock, or hdr is on arc_anon */
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	hdr->b_l1hdr.b_mru_hits = 0;
	hdr->b_l1hdr.b_mru_ghost_hits = 0;
	hdr->b_l1hdr.b_mfu_hits = 0;
	hdr->b_l1hdr.b_mfu_ghost_hits = 0;
	arc_change_state(arc_anon, hdr, hash_lock);
	hdr->b_l1hdr.b_arc_access = 0;

	mutex_exit(hash_lock);
	buf_discard_identity(hdr);
	arc_buf_thaw(buf);
	}
	}

	int
	arc_released(arc_buf_t *buf)
	{
	int released;

	mutex_enter(&buf->b_evict_lock);
	released = (buf->b_data != NULL &&
	buf->b_hdr->b_l1hdr.b_state == arc_anon);
	mutex_exit(&buf->b_evict_lock);
	return (released);
	}

	#ifdef ZFS_DEBUG
	int
	arc_referenced(arc_buf_t *buf)
	{
	int referenced;

	mutex_enter(&buf->b_evict_lock);
	referenced = (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt));
	mutex_exit(&buf->b_evict_lock);
	return (referenced);
	}
	#endif

	static void
	arc_write_ready(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;
	arc_buf_hdr_t *hdr = buf->b_hdr;
	blkptr_t *bp = zio->io_bp;
	uint64_t psize = BP_IS_HOLE(bp) ? 0 : BP_GET_PSIZE(bp);
	fstrans_cookie_t cookie = spl_fstrans_mark();

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!zfs_refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt));
	ASSERT(hdr->b_l1hdr.b_bufcnt > 0);

	/*
	* If we're reexecuting this zio because the pool suspended, then
	* cleanup any state that was previously set the first time the
	* callback was invoked.
	*/
	if (zio->io_flags & ZIO_FLAG_REEXECUTED) {
	arc_cksum_free(hdr);
	arc_buf_unwatch(buf);
	if (hdr->b_l1hdr.b_pabd != NULL) {
	if (arc_buf_is_shared(buf)) {
	arc_unshare_buf(hdr, buf);
	} else {
	arc_hdr_free_abd(hdr, B_FALSE);
	}
	}

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);
	}
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	ASSERT(!HDR_SHARED_DATA(hdr));
	ASSERT(!arc_buf_is_shared(buf));

	callback->awcb_ready(zio, buf, callback->awcb_private);

	if (HDR_IO_IN_PROGRESS(hdr))
	ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED);

	arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);

	if (BP_IS_PROTECTED(bp) != !!HDR_PROTECTED(hdr))
	hdr = arc_hdr_realloc_crypt(hdr, BP_IS_PROTECTED(bp));

	if (BP_IS_PROTECTED(bp)) {
	/* ZIL blocks are written through zio_rewrite */
	ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);
	ASSERT(HDR_PROTECTED(hdr));

	if (BP_SHOULD_BYTESWAP(bp)) {
	if (BP_GET_LEVEL(bp) > 0) {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
	} else {
	hdr->b_l1hdr.b_byteswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
	}
	} else {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	}

	hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
	hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
	zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv);
	zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac);
	}

	/*
	* If this block was written for raw encryption but the zio layer
	* ended up only authenticating it, adjust the buffer flags now.
	*/
	if (BP_IS_AUTHENTICATED(bp) && ARC_BUF_ENCRYPTED(buf)) {
	arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	if (BP_GET_COMPRESS(bp) == ZIO_COMPRESS_OFF)
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	} else if (BP_IS_HOLE(bp) && ARC_BUF_ENCRYPTED(buf)) {
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	}

	/* this must be done after the buffer flags are adjusted */
	arc_cksum_compute(buf);

	enum zio_compress compress;
	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp)) {
	compress = ZIO_COMPRESS_OFF;
	} else {
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
	compress = BP_GET_COMPRESS(bp);
	}
	HDR_SET_PSIZE(hdr, psize);
	arc_hdr_set_compress(hdr, compress);
	hdr->b_complevel = zio->io_prop.zp_complevel;

	if (zio->io_error != 0 \|\| psize == 0)
	goto out;

	/*
	* Fill the hdr with data. If the buffer is encrypted we have no choice
	* but to copy the data into b_radb. If the hdr is compressed, the data
	* we want is available from the zio, otherwise we can take it from
	* the buf.
	*
	* We might be able to share the buf's data with the hdr here. However,
	* doing so would cause the ARC to be full of linear ABDs if we write a
	* lot of shareable data. As a compromise, we check whether scattered
	* ABDs are allowed, and assume that if they are then the user wants
	* the ARC to be primarily filled with them regardless of the data being
	* written. Therefore, if they're allowed then we allocate one and copy
	* the data into it; otherwise, we share the data directly if we can.
	*/
	if (ARC_BUF_ENCRYPTED(buf)) {
	ASSERT3U(psize, >, 0);
	ASSERT(ARC_BUF_COMPRESSED(buf));
	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT \| ARC_HDR_ALLOC_RDATA \|
	ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
	} else if (!abd_size_alloc_linear(arc_buf_size(buf)) \|\|
	!arc_can_share(hdr, buf)) {
	/*
	* Ideally, we would always copy the io_abd into b_pabd, but the
	* user may have disabled compressed ARC, thus we must check the
	* hdr's compression setting rather than the io_bp's.
	*/
	if (BP_IS_ENCRYPTED(bp)) {
	ASSERT3U(psize, >, 0);
	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT \|
	ARC_HDR_ALLOC_RDATA \| ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
	} else if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
	!ARC_BUF_COMPRESSED(buf)) {
	ASSERT3U(psize, >, 0);
	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT \|
	ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_l1hdr.b_pabd, zio->io_abd, psize);
	} else {
	ASSERT3U(zio->io_orig_size, ==, arc_hdr_size(hdr));
	arc_hdr_alloc_abd(hdr, ARC_HDR_DO_ADAPT \|
	ARC_HDR_USE_RESERVE);
	abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data,
	arc_buf_size(buf));
	}
	} else {
	ASSERT3P(buf->b_data, ==, abd_to_buf(zio->io_orig_abd));
	ASSERT3U(zio->io_orig_size, ==, arc_buf_size(buf));
	ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1);

	arc_share_buf(hdr, buf);
	}

	out:
	arc_hdr_verify(hdr, bp);
	spl_fstrans_unmark(cookie);
	}

	static void
	arc_write_children_ready(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;

	callback->awcb_children_ready(zio, buf, callback->awcb_private);
	}

	/*
	* The SPA calls this callback for each physical write that happens on behalf
	* of a logical write. See the comment in dbuf_write_physdone() for details.
	*/
	static void
	arc_write_physdone(zio_t *zio)
	{
	arc_write_callback_t *cb = zio->io_private;
	if (cb->awcb_physdone != NULL)
	cb->awcb_physdone(zio, cb->awcb_buf, cb->awcb_private);
	}

	static void
	arc_write_done(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);

	if (zio->io_error == 0) {
	arc_hdr_verify(hdr, zio->io_bp);

	if (BP_IS_HOLE(zio->io_bp) \|\| BP_IS_EMBEDDED(zio->io_bp)) {
	buf_discard_identity(hdr);
	} else {
	hdr->b_dva = *BP_IDENTITY(zio->io_bp);
	hdr->b_birth = BP_PHYSICAL_BIRTH(zio->io_bp);
	}
	} else {
	ASSERT(HDR_EMPTY(hdr));
	}

	/*
	* If the block to be written was all-zero or compressed enough to be
	* embedded in the BP, no write was performed so there will be no
	* dva/birth/checksum. The buffer must therefore remain anonymous
	* (and uncached).
	*/
	if (!HDR_EMPTY(hdr)) {
	arc_buf_hdr_t *exists;
	kmutex_t *hash_lock;

	ASSERT3U(zio->io_error, ==, 0);

	arc_cksum_verify(buf);

	exists = buf_hash_insert(hdr, &hash_lock);
	if (exists != NULL) {
	/*
	* This can only happen if we overwrite for
	* sync-to-convergence, because we remove
	* buffers from the hash table when we arc_free().
	*/
	if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
	if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
	panic("bad overwrite, hdr=%p exists=%p",
	(void )hdr, (void )exists);
	ASSERT(zfs_refcount_is_zero(
	&exists->b_l1hdr.b_refcnt));
	arc_change_state(arc_anon, exists, hash_lock);
	arc_hdr_destroy(exists);
	mutex_exit(hash_lock);
	exists = buf_hash_insert(hdr, &hash_lock);
	ASSERT3P(exists, ==, NULL);
	} else if (zio->io_flags & ZIO_FLAG_NOPWRITE) {
	/* nopwrite */
	ASSERT(zio->io_prop.zp_nopwrite);
	if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
	panic("bad nopwrite, hdr=%p exists=%p",
	(void )hdr, (void )exists);
	} else {
	/* Dedup */
	ASSERT(hdr->b_l1hdr.b_bufcnt == 1);
	ASSERT(hdr->b_l1hdr.b_state == arc_anon);
	ASSERT(BP_GET_DEDUP(zio->io_bp));
	ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
	}
	}
	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	/* if it's not anon, we are doing a scrub */
	if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon)
	arc_access(hdr, hash_lock);
	mutex_exit(hash_lock);
	} else {
	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	}

	ASSERT(!zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	callback->awcb_done(zio, buf, callback->awcb_private);

	abd_free(zio->io_abd);
	kmem_free(callback, sizeof (arc_write_callback_t));
	}

	zio_t *
	arc_write(zio_t pio, spa_t spa, uint64_t txg,
	blkptr_t bp, arc_buf_t buf, boolean_t l2arc,
	const zio_prop_t zp, arc_write_done_func_t ready,
	arc_write_done_func_t children_ready, arc_write_done_func_t physdone,
	arc_write_done_func_t done, void private, zio_priority_t priority,
	int zio_flags, const zbookmark_phys_t *zb)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;
	arc_write_callback_t *callback;
	zio_t *zio;
	zio_prop_t localprop = *zp;

	ASSERT3P(ready, !=, NULL);
	ASSERT3P(done, !=, NULL);
	ASSERT(!HDR_IO_ERROR(hdr));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
	ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0);
	if (l2arc)
	arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);

	if (ARC_BUF_ENCRYPTED(buf)) {
	ASSERT(ARC_BUF_COMPRESSED(buf));
	localprop.zp_encrypt = B_TRUE;
	localprop.zp_compress = HDR_GET_COMPRESS(hdr);
	localprop.zp_complevel = hdr->b_complevel;
	localprop.zp_byteorder =
	(hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
	ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
	memcpy(localprop.zp_salt, hdr->b_crypt_hdr.b_salt,
	ZIO_DATA_SALT_LEN);
	memcpy(localprop.zp_iv, hdr->b_crypt_hdr.b_iv,
	ZIO_DATA_IV_LEN);
	memcpy(localprop.zp_mac, hdr->b_crypt_hdr.b_mac,
	ZIO_DATA_MAC_LEN);
	if (DMU_OT_IS_ENCRYPTED(localprop.zp_type)) {
	localprop.zp_nopwrite = B_FALSE;
	localprop.zp_copies =
	MIN(localprop.zp_copies, SPA_DVAS_PER_BP - 1);
	}
	zio_flags \|= ZIO_FLAG_RAW;
	} else if (ARC_BUF_COMPRESSED(buf)) {
	ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf));
	localprop.zp_compress = HDR_GET_COMPRESS(hdr);
	localprop.zp_complevel = hdr->b_complevel;
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	}
	callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
	callback->awcb_ready = ready;
	callback->awcb_children_ready = children_ready;
	callback->awcb_physdone = physdone;
	callback->awcb_done = done;
	callback->awcb_private = private;
	callback->awcb_buf = buf;

	/*
	* The hdr's b_pabd is now stale, free it now. A new data block
	* will be allocated when the zio pipeline calls arc_write_ready().
	*/
	if (hdr->b_l1hdr.b_pabd != NULL) {
	/*
	* If the buf is currently sharing the data block with
	* the hdr then we need to break that relationship here.
	* The hdr will remain with a NULL data pointer and the
	* buf will take sole ownership of the block.
	*/
	if (arc_buf_is_shared(buf)) {
	arc_unshare_buf(hdr, buf);
	} else {
	arc_hdr_free_abd(hdr, B_FALSE);
	}
	VERIFY3P(buf->b_data, !=, NULL);
	}

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);

	if (!(zio_flags & ZIO_FLAG_RAW))
	arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF);

	ASSERT(!arc_buf_is_shared(buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);

	zio = zio_write(pio, spa, txg, bp,
	abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)),
	HDR_GET_LSIZE(hdr), arc_buf_size(buf), &localprop, arc_write_ready,
	(children_ready != NULL) ? arc_write_children_ready : NULL,
	arc_write_physdone, arc_write_done, callback,
	priority, zio_flags, zb);

	return (zio);
	}

	void
	arc_tempreserve_clear(uint64_t reserve)
	{
	atomic_add_64(&arc_tempreserve, -reserve);
	ASSERT((int64_t)arc_tempreserve >= 0);
	}

	int
	arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg)
	{
	int error;
	uint64_t anon_size;

	if (!arc_no_grow &&
	reserve > arc_c/4 &&
	reserve * 4 > (2ULL << SPA_MAXBLOCKSHIFT))
	arc_c = MIN(arc_c_max, reserve * 4);

	/*
	* Throttle when the calculated memory footprint for the TXG
	* exceeds the target ARC size.
	*/
	if (reserve > arc_c) {
	DMU_TX_STAT_BUMP(dmu_tx_memory_reserve);
	return (SET_ERROR(ERESTART));
	}

	/*
	* Don't count loaned bufs as in flight dirty data to prevent long
	* network delays from blocking transactions that are ready to be
	* assigned to a txg.
	*/

	/* assert that it has not wrapped around */
	ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);

	anon_size = MAX((int64_t)(zfs_refcount_count(&arc_anon->arcs_size) -
	arc_loaned_bytes), 0);

	/*
	* Writes will, almost always, require additional memory allocations
	* in order to compress/encrypt/etc the data. We therefore need to
	* make sure that there is sufficient available memory for this.
	*/
	error = arc_memory_throttle(spa, reserve, txg);
	if (error != 0)
	return (error);

	/*
	* Throttle writes when the amount of dirty data in the cache
	* gets too large. We try to keep the cache less than half full
	* of dirty blocks so that our sync times don't grow too large.
	*
	* In the case of one pool being built on another pool, we want
	* to make sure we don't end up throttling the lower (backing)
	* pool when the upper pool is the majority contributor to dirty
	* data. To insure we make forward progress during throttling, we
	* also check the current pool's net dirty data and only throttle
	* if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty
	* data in the cache.
	*
	* Note: if two requests come in concurrently, we might let them
	* both succeed, when one of them should fail. Not a huge deal.
	*/
	uint64_t total_dirty = reserve + arc_tempreserve + anon_size;
	uint64_t spa_dirty_anon = spa_dirty_data(spa);
	uint64_t rarc_c = arc_warm ? arc_c : arc_c_max;
	if (total_dirty > rarc_c * zfs_arc_dirty_limit_percent / 100 &&
	anon_size > rarc_c * zfs_arc_anon_limit_percent / 100 &&
	spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) {
	#ifdef ZFS_DEBUG
	uint64_t meta_esize = zfs_refcount_count(
	&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	uint64_t data_esize =
	zfs_refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
	"anon_data=%lluK tempreserve=%lluK rarc_c=%lluK\n",
	(u_longlong_t)arc_tempreserve >> 10,
	(u_longlong_t)meta_esize >> 10,
	(u_longlong_t)data_esize >> 10,
	(u_longlong_t)reserve >> 10,
	(u_longlong_t)rarc_c >> 10);
	#endif
	DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle);
	return (SET_ERROR(ERESTART));
	}
	atomic_add_64(&arc_tempreserve, reserve);
	return (0);
	}

	static void
	arc_kstat_update_state(arc_state_t state, kstat_named_t size,
	kstat_named_t evict_data, kstat_named_t evict_metadata)
	{
	size->value.ui64 = zfs_refcount_count(&state->arcs_size);
	evict_data->value.ui64 =
	zfs_refcount_count(&state->arcs_esize[ARC_BUFC_DATA]);
	evict_metadata->value.ui64 =
	zfs_refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]);
	}

	static int
	arc_kstat_update(kstat_t *ksp, int rw)
	{
	arc_stats_t *as = ksp->ks_data;

	if (rw == KSTAT_WRITE)
	return (SET_ERROR(EACCES));

	as->arcstat_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hits);
	as->arcstat_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_misses);
	as->arcstat_demand_data_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_data_hits);
	as->arcstat_demand_data_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_data_misses);
	as->arcstat_demand_metadata_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_metadata_hits);
	as->arcstat_demand_metadata_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_metadata_misses);
	as->arcstat_prefetch_data_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_data_hits);
	as->arcstat_prefetch_data_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_data_misses);
	as->arcstat_prefetch_metadata_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_metadata_hits);
	as->arcstat_prefetch_metadata_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_metadata_misses);
	as->arcstat_mru_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mru_hits);
	as->arcstat_mru_ghost_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mru_ghost_hits);
	as->arcstat_mfu_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mfu_hits);
	as->arcstat_mfu_ghost_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mfu_ghost_hits);
	as->arcstat_deleted.value.ui64 =
	wmsum_value(&arc_sums.arcstat_deleted);
	as->arcstat_mutex_miss.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mutex_miss);
	as->arcstat_access_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_access_skip);
	as->arcstat_evict_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_skip);
	as->arcstat_evict_not_enough.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_not_enough);
	as->arcstat_evict_l2_cached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_cached);
	as->arcstat_evict_l2_eligible.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible);
	as->arcstat_evict_l2_eligible_mfu.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mfu);
	as->arcstat_evict_l2_eligible_mru.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mru);
	as->arcstat_evict_l2_ineligible.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_ineligible);
	as->arcstat_evict_l2_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_skip);
	as->arcstat_hash_collisions.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hash_collisions);
	as->arcstat_hash_chains.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hash_chains);
	as->arcstat_size.value.ui64 =
	aggsum_value(&arc_sums.arcstat_size);
	as->arcstat_compressed_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_compressed_size);
	as->arcstat_uncompressed_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_uncompressed_size);
	as->arcstat_overhead_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_overhead_size);
	as->arcstat_hdr_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hdr_size);
	as->arcstat_data_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_data_size);
	as->arcstat_metadata_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_metadata_size);
	as->arcstat_dbuf_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_dbuf_size);
	#if defined(COMPAT_FREEBSD11)
	as->arcstat_other_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_bonus_size) +
	aggsum_value(&arc_sums.arcstat_dnode_size) +
	wmsum_value(&arc_sums.arcstat_dbuf_size);
	#endif

	arc_kstat_update_state(arc_anon,
	&as->arcstat_anon_size,
	&as->arcstat_anon_evictable_data,
	&as->arcstat_anon_evictable_metadata);
	arc_kstat_update_state(arc_mru,
	&as->arcstat_mru_size,
	&as->arcstat_mru_evictable_data,
	&as->arcstat_mru_evictable_metadata);
	arc_kstat_update_state(arc_mru_ghost,
	&as->arcstat_mru_ghost_size,
	&as->arcstat_mru_ghost_evictable_data,
	&as->arcstat_mru_ghost_evictable_metadata);
	arc_kstat_update_state(arc_mfu,
	&as->arcstat_mfu_size,
	&as->arcstat_mfu_evictable_data,
	&as->arcstat_mfu_evictable_metadata);
	arc_kstat_update_state(arc_mfu_ghost,
	&as->arcstat_mfu_ghost_size,
	&as->arcstat_mfu_ghost_evictable_data,
	&as->arcstat_mfu_ghost_evictable_metadata);

	as->arcstat_dnode_size.value.ui64 =
	aggsum_value(&arc_sums.arcstat_dnode_size);
	as->arcstat_bonus_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_bonus_size);
	as->arcstat_l2_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_hits);
	as->arcstat_l2_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_misses);
	as->arcstat_l2_prefetch_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_prefetch_asize);
	as->arcstat_l2_mru_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_mru_asize);
	as->arcstat_l2_mfu_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_mfu_asize);
	as->arcstat_l2_bufc_data_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_bufc_data_asize);
	as->arcstat_l2_bufc_metadata_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_bufc_metadata_asize);
	as->arcstat_l2_feeds.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_feeds);
	as->arcstat_l2_rw_clash.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rw_clash);
	as->arcstat_l2_read_bytes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_read_bytes);
	as->arcstat_l2_write_bytes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_write_bytes);
	as->arcstat_l2_writes_sent.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_sent);
	as->arcstat_l2_writes_done.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_done);
	as->arcstat_l2_writes_error.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_error);
	as->arcstat_l2_writes_lock_retry.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_lock_retry);
	as->arcstat_l2_evict_lock_retry.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_lock_retry);
	as->arcstat_l2_evict_reading.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_reading);
	as->arcstat_l2_evict_l1cached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_l1cached);
	as->arcstat_l2_free_on_write.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_free_on_write);
	as->arcstat_l2_abort_lowmem.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_abort_lowmem);
	as->arcstat_l2_cksum_bad.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_cksum_bad);
	as->arcstat_l2_io_error.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_io_error);
	as->arcstat_l2_lsize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_lsize);
	as->arcstat_l2_psize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_psize);
	as->arcstat_l2_hdr_size.value.ui64 =
	aggsum_value(&arc_sums.arcstat_l2_hdr_size);
	as->arcstat_l2_log_blk_writes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_writes);
	as->arcstat_l2_log_blk_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_asize);
	as->arcstat_l2_log_blk_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_count);
	as->arcstat_l2_rebuild_success.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_success);
	as->arcstat_l2_rebuild_abort_unsupported.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
	as->arcstat_l2_rebuild_abort_io_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
	as->arcstat_l2_rebuild_abort_dh_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
	as->arcstat_l2_rebuild_abort_cksum_lb_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
	as->arcstat_l2_rebuild_abort_lowmem.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
	as->arcstat_l2_rebuild_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_size);
	as->arcstat_l2_rebuild_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_asize);
	as->arcstat_l2_rebuild_bufs.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs);
	as->arcstat_l2_rebuild_bufs_precached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs_precached);
	as->arcstat_l2_rebuild_log_blks.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_log_blks);
	as->arcstat_memory_throttle_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_throttle_count);
	as->arcstat_memory_direct_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_direct_count);
	as->arcstat_memory_indirect_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_indirect_count);

	as->arcstat_memory_all_bytes.value.ui64 =
	arc_all_memory();
	as->arcstat_memory_free_bytes.value.ui64 =
	arc_free_memory();
	as->arcstat_memory_available_bytes.value.i64 =
	arc_available_memory();

	as->arcstat_prune.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prune);
	as->arcstat_meta_used.value.ui64 =
	aggsum_value(&arc_sums.arcstat_meta_used);
	as->arcstat_async_upgrade_sync.value.ui64 =
	wmsum_value(&arc_sums.arcstat_async_upgrade_sync);
	as->arcstat_demand_hit_predictive_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_hit_predictive_prefetch);
	as->arcstat_demand_hit_prescient_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_hit_prescient_prefetch);
	as->arcstat_raw_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_raw_size);
	as->arcstat_cached_only_in_progress.value.ui64 =
	wmsum_value(&arc_sums.arcstat_cached_only_in_progress);
	as->arcstat_abd_chunk_waste_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_abd_chunk_waste_size);

	return (0);
	}

	/*
	* This function must return indices evenly distributed between all
	* sublists of the multilist. This is needed due to how the ARC eviction
	* code is laid out; arc_evict_state() assumes ARC buffers are evenly
	* distributed between all sublists and uses this assumption when
	* deciding which sublist to evict from and how much to evict from it.
	*/
	static unsigned int
	arc_state_multilist_index_func(multilist_t ml, void obj)
	{
	arc_buf_hdr_t *hdr = obj;

	/*
	* We rely on b_dva to generate evenly distributed index
	* numbers using buf_hash below. So, as an added precaution,
	* let's make sure we never add empty buffers to the arc lists.
	*/
	ASSERT(!HDR_EMPTY(hdr));

	/*
	* The assumption here, is the hash value for a given
	* arc_buf_hdr_t will remain constant throughout its lifetime
	* (i.e. its b_spa, b_dva, and b_birth fields don't change).
	* Thus, we don't need to store the header's sublist index
	* on insertion, as this index can be recalculated on removal.
	*
	* Also, the low order bits of the hash value are thought to be
	* distributed evenly. Otherwise, in the case that the multilist
	* has a power of two number of sublists, each sublists' usage
	* would not be evenly distributed. In this context full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) %
	multilist_get_num_sublists(ml));
	}

	static unsigned int
	arc_state_l2c_multilist_index_func(multilist_t ml, void obj)
	{
	panic("Header %p insert into arc_l2c_only %p", obj, ml);
	}

	#define WARN_IF_TUNING_IGNORED(tuning, value, do_warn) do { \
	if ((do_warn) && (tuning) && ((tuning) != (value))) { \
	cmn_err(CE_WARN, \
	"ignoring tunable %s (using %llu instead)", \
	(#tuning), (u_longlong_t)(value)); \
	} \
	} while (0)

	/*
	* Called during module initialization and periodically thereafter to
	* apply reasonable changes to the exposed performance tunings. Can also be
	* called explicitly by param_set_arc_*() functions when ARC tunables are
	* updated manually. Non-zero zfs_* values which differ from the currently set
	* values will be applied.
	*/
	void
	arc_tuning_update(boolean_t verbose)
	{
	uint64_t allmem = arc_all_memory();
	unsigned long limit;

	/* Valid range: 32M - <arc_c_max> */
	if ((zfs_arc_min) && (zfs_arc_min != arc_c_min) &&
	(zfs_arc_min >= 2ULL << SPA_MAXBLOCKSHIFT) &&
	(zfs_arc_min <= arc_c_max)) {
	arc_c_min = zfs_arc_min;
	arc_c = MAX(arc_c, arc_c_min);
	}
	WARN_IF_TUNING_IGNORED(zfs_arc_min, arc_c_min, verbose);

	/* Valid range: 64M - <all physical memory> */
	if ((zfs_arc_max) && (zfs_arc_max != arc_c_max) &&
	(zfs_arc_max >= MIN_ARC_MAX) && (zfs_arc_max < allmem) &&
	(zfs_arc_max > arc_c_min)) {
	arc_c_max = zfs_arc_max;
	arc_c = MIN(arc_c, arc_c_max);
	arc_p = (arc_c >> 1);
	if (arc_meta_limit > arc_c_max)
	arc_meta_limit = arc_c_max;
	if (arc_dnode_size_limit > arc_meta_limit)
	arc_dnode_size_limit = arc_meta_limit;
	}
	WARN_IF_TUNING_IGNORED(zfs_arc_max, arc_c_max, verbose);

	/* Valid range: 16M - <arc_c_max> */
	if ((zfs_arc_meta_min) && (zfs_arc_meta_min != arc_meta_min) &&
	(zfs_arc_meta_min >= 1ULL << SPA_MAXBLOCKSHIFT) &&
	(zfs_arc_meta_min <= arc_c_max)) {
	arc_meta_min = zfs_arc_meta_min;
	if (arc_meta_limit < arc_meta_min)
	arc_meta_limit = arc_meta_min;
	if (arc_dnode_size_limit < arc_meta_min)
	arc_dnode_size_limit = arc_meta_min;
	}
	WARN_IF_TUNING_IGNORED(zfs_arc_meta_min, arc_meta_min, verbose);

	/* Valid range: <arc_meta_min> - <arc_c_max> */
	limit = zfs_arc_meta_limit ? zfs_arc_meta_limit :
	MIN(zfs_arc_meta_limit_percent, 100) * arc_c_max / 100;
	if ((limit != arc_meta_limit) &&
	(limit >= arc_meta_min) &&
	(limit <= arc_c_max))
	arc_meta_limit = limit;
	WARN_IF_TUNING_IGNORED(zfs_arc_meta_limit, arc_meta_limit, verbose);

	/* Valid range: <arc_meta_min> - <arc_meta_limit> */
	limit = zfs_arc_dnode_limit ? zfs_arc_dnode_limit :
	MIN(zfs_arc_dnode_limit_percent, 100) * arc_meta_limit / 100;
	if ((limit != arc_dnode_size_limit) &&
	(limit >= arc_meta_min) &&
	(limit <= arc_meta_limit))
	arc_dnode_size_limit = limit;
	WARN_IF_TUNING_IGNORED(zfs_arc_dnode_limit, arc_dnode_size_limit,
	verbose);

	/* Valid range: 1 - N */
	if (zfs_arc_grow_retry)
	arc_grow_retry = zfs_arc_grow_retry;

	/* Valid range: 1 - N */
	if (zfs_arc_shrink_shift) {
	arc_shrink_shift = zfs_arc_shrink_shift;
	arc_no_grow_shift = MIN(arc_no_grow_shift, arc_shrink_shift -1);
	}

	/* Valid range: 1 - N */
	if (zfs_arc_p_min_shift)
	arc_p_min_shift = zfs_arc_p_min_shift;

	/* Valid range: 1 - N ms */
	if (zfs_arc_min_prefetch_ms)
	arc_min_prefetch_ms = zfs_arc_min_prefetch_ms;

	/* Valid range: 1 - N ms */
	if (zfs_arc_min_prescient_prefetch_ms) {
	arc_min_prescient_prefetch_ms =
	zfs_arc_min_prescient_prefetch_ms;
	}

	/* Valid range: 0 - 100 */
	- if ((zfs_arc_lotsfree_percent >= 0) &&
	- (zfs_arc_lotsfree_percent <= 100))
	+ if (zfs_arc_lotsfree_percent <= 100)
	arc_lotsfree_percent = zfs_arc_lotsfree_percent;
	WARN_IF_TUNING_IGNORED(zfs_arc_lotsfree_percent, arc_lotsfree_percent,
	verbose);

	/* Valid range: 0 - <all physical memory> */
	if ((zfs_arc_sys_free) && (zfs_arc_sys_free != arc_sys_free))
	- arc_sys_free = MIN(MAX(zfs_arc_sys_free, 0), allmem);
	+ arc_sys_free = MIN(zfs_arc_sys_free, allmem);
	WARN_IF_TUNING_IGNORED(zfs_arc_sys_free, arc_sys_free, verbose);
	}

	static void
	arc_state_multilist_init(multilist_t *ml,
	multilist_sublist_index_func_t index_func, int maxcountp)
	{
	multilist_create(ml, sizeof (arc_buf_hdr_t),
	offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), index_func);
	maxcountp = MAX(maxcountp, multilist_get_num_sublists(ml));
	}

	static void
	arc_state_init(void)
	{
	int num_sublists = 0;

	arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);

	/*
	* L2 headers should never be on the L2 state list since they don't
	* have L1 headers allocated. Special index function asserts that.
	*/
	arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
	arc_state_l2c_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
	arc_state_l2c_multilist_index_func, &num_sublists);

	/*
	* Keep track of the number of markers needed to reclaim buffers from
	* any ARC state. The markers will be pre-allocated so as to minimize
	* the number of memory allocations performed by the eviction thread.
	*/
	arc_state_evict_marker_count = num_sublists;

	zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);

	zfs_refcount_create(&arc_anon->arcs_size);
	zfs_refcount_create(&arc_mru->arcs_size);
	zfs_refcount_create(&arc_mru_ghost->arcs_size);
	zfs_refcount_create(&arc_mfu->arcs_size);
	zfs_refcount_create(&arc_mfu_ghost->arcs_size);
	zfs_refcount_create(&arc_l2c_only->arcs_size);

	wmsum_init(&arc_sums.arcstat_hits, 0);
	wmsum_init(&arc_sums.arcstat_misses, 0);
	wmsum_init(&arc_sums.arcstat_demand_data_hits, 0);
	wmsum_init(&arc_sums.arcstat_demand_data_misses, 0);
	wmsum_init(&arc_sums.arcstat_demand_metadata_hits, 0);
	wmsum_init(&arc_sums.arcstat_demand_metadata_misses, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_data_hits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_data_misses, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_metadata_hits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_metadata_misses, 0);
	wmsum_init(&arc_sums.arcstat_mru_hits, 0);
	wmsum_init(&arc_sums.arcstat_mru_ghost_hits, 0);
	wmsum_init(&arc_sums.arcstat_mfu_hits, 0);
	wmsum_init(&arc_sums.arcstat_mfu_ghost_hits, 0);
	wmsum_init(&arc_sums.arcstat_deleted, 0);
	wmsum_init(&arc_sums.arcstat_mutex_miss, 0);
	wmsum_init(&arc_sums.arcstat_access_skip, 0);
	wmsum_init(&arc_sums.arcstat_evict_skip, 0);
	wmsum_init(&arc_sums.arcstat_evict_not_enough, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_cached, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mfu, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mru, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_ineligible, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_skip, 0);
	wmsum_init(&arc_sums.arcstat_hash_collisions, 0);
	wmsum_init(&arc_sums.arcstat_hash_chains, 0);
	aggsum_init(&arc_sums.arcstat_size, 0);
	wmsum_init(&arc_sums.arcstat_compressed_size, 0);
	wmsum_init(&arc_sums.arcstat_uncompressed_size, 0);
	wmsum_init(&arc_sums.arcstat_overhead_size, 0);
	wmsum_init(&arc_sums.arcstat_hdr_size, 0);
	wmsum_init(&arc_sums.arcstat_data_size, 0);
	wmsum_init(&arc_sums.arcstat_metadata_size, 0);
	wmsum_init(&arc_sums.arcstat_dbuf_size, 0);
	aggsum_init(&arc_sums.arcstat_dnode_size, 0);
	wmsum_init(&arc_sums.arcstat_bonus_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_hits, 0);
	wmsum_init(&arc_sums.arcstat_l2_misses, 0);
	wmsum_init(&arc_sums.arcstat_l2_prefetch_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_mru_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_mfu_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_bufc_data_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_bufc_metadata_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_feeds, 0);
	wmsum_init(&arc_sums.arcstat_l2_rw_clash, 0);
	wmsum_init(&arc_sums.arcstat_l2_read_bytes, 0);
	wmsum_init(&arc_sums.arcstat_l2_write_bytes, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_sent, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_done, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_error, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_lock_retry, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_lock_retry, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_reading, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_l1cached, 0);
	wmsum_init(&arc_sums.arcstat_l2_free_on_write, 0);
	wmsum_init(&arc_sums.arcstat_l2_abort_lowmem, 0);
	wmsum_init(&arc_sums.arcstat_l2_cksum_bad, 0);
	wmsum_init(&arc_sums.arcstat_l2_io_error, 0);
	wmsum_init(&arc_sums.arcstat_l2_lsize, 0);
	wmsum_init(&arc_sums.arcstat_l2_psize, 0);
	aggsum_init(&arc_sums.arcstat_l2_hdr_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_writes, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_count, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_success, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_unsupported, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_io_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_dh_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_lowmem, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs_precached, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_log_blks, 0);
	wmsum_init(&arc_sums.arcstat_memory_throttle_count, 0);
	wmsum_init(&arc_sums.arcstat_memory_direct_count, 0);
	wmsum_init(&arc_sums.arcstat_memory_indirect_count, 0);
	wmsum_init(&arc_sums.arcstat_prune, 0);
	aggsum_init(&arc_sums.arcstat_meta_used, 0);
	wmsum_init(&arc_sums.arcstat_async_upgrade_sync, 0);
	wmsum_init(&arc_sums.arcstat_demand_hit_predictive_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_demand_hit_prescient_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_raw_size, 0);
	wmsum_init(&arc_sums.arcstat_cached_only_in_progress, 0);
	wmsum_init(&arc_sums.arcstat_abd_chunk_waste_size, 0);

	arc_anon->arcs_state = ARC_STATE_ANON;
	arc_mru->arcs_state = ARC_STATE_MRU;
	arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST;
	arc_mfu->arcs_state = ARC_STATE_MFU;
	arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST;
	arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY;
	}

	static void
	arc_state_fini(void)
	{
	zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);

	zfs_refcount_destroy(&arc_anon->arcs_size);
	zfs_refcount_destroy(&arc_mru->arcs_size);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_size);
	zfs_refcount_destroy(&arc_mfu->arcs_size);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_size);
	zfs_refcount_destroy(&arc_l2c_only->arcs_size);

	multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]);

	wmsum_fini(&arc_sums.arcstat_hits);
	wmsum_fini(&arc_sums.arcstat_misses);
	wmsum_fini(&arc_sums.arcstat_demand_data_hits);
	wmsum_fini(&arc_sums.arcstat_demand_data_misses);
	wmsum_fini(&arc_sums.arcstat_demand_metadata_hits);
	wmsum_fini(&arc_sums.arcstat_demand_metadata_misses);
	wmsum_fini(&arc_sums.arcstat_prefetch_data_hits);
	wmsum_fini(&arc_sums.arcstat_prefetch_data_misses);
	wmsum_fini(&arc_sums.arcstat_prefetch_metadata_hits);
	wmsum_fini(&arc_sums.arcstat_prefetch_metadata_misses);
	wmsum_fini(&arc_sums.arcstat_mru_hits);
	wmsum_fini(&arc_sums.arcstat_mru_ghost_hits);
	wmsum_fini(&arc_sums.arcstat_mfu_hits);
	wmsum_fini(&arc_sums.arcstat_mfu_ghost_hits);
	wmsum_fini(&arc_sums.arcstat_deleted);
	wmsum_fini(&arc_sums.arcstat_mutex_miss);
	wmsum_fini(&arc_sums.arcstat_access_skip);
	wmsum_fini(&arc_sums.arcstat_evict_skip);
	wmsum_fini(&arc_sums.arcstat_evict_not_enough);
	wmsum_fini(&arc_sums.arcstat_evict_l2_cached);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mfu);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mru);
	wmsum_fini(&arc_sums.arcstat_evict_l2_ineligible);
	wmsum_fini(&arc_sums.arcstat_evict_l2_skip);
	wmsum_fini(&arc_sums.arcstat_hash_collisions);
	wmsum_fini(&arc_sums.arcstat_hash_chains);
	aggsum_fini(&arc_sums.arcstat_size);
	wmsum_fini(&arc_sums.arcstat_compressed_size);
	wmsum_fini(&arc_sums.arcstat_uncompressed_size);
	wmsum_fini(&arc_sums.arcstat_overhead_size);
	wmsum_fini(&arc_sums.arcstat_hdr_size);
	wmsum_fini(&arc_sums.arcstat_data_size);
	wmsum_fini(&arc_sums.arcstat_metadata_size);
	wmsum_fini(&arc_sums.arcstat_dbuf_size);
	aggsum_fini(&arc_sums.arcstat_dnode_size);
	wmsum_fini(&arc_sums.arcstat_bonus_size);
	wmsum_fini(&arc_sums.arcstat_l2_hits);
	wmsum_fini(&arc_sums.arcstat_l2_misses);
	wmsum_fini(&arc_sums.arcstat_l2_prefetch_asize);
	wmsum_fini(&arc_sums.arcstat_l2_mru_asize);
	wmsum_fini(&arc_sums.arcstat_l2_mfu_asize);
	wmsum_fini(&arc_sums.arcstat_l2_bufc_data_asize);
	wmsum_fini(&arc_sums.arcstat_l2_bufc_metadata_asize);
	wmsum_fini(&arc_sums.arcstat_l2_feeds);
	wmsum_fini(&arc_sums.arcstat_l2_rw_clash);
	wmsum_fini(&arc_sums.arcstat_l2_read_bytes);
	wmsum_fini(&arc_sums.arcstat_l2_write_bytes);
	wmsum_fini(&arc_sums.arcstat_l2_writes_sent);
	wmsum_fini(&arc_sums.arcstat_l2_writes_done);
	wmsum_fini(&arc_sums.arcstat_l2_writes_error);
	wmsum_fini(&arc_sums.arcstat_l2_writes_lock_retry);
	wmsum_fini(&arc_sums.arcstat_l2_evict_lock_retry);
	wmsum_fini(&arc_sums.arcstat_l2_evict_reading);
	wmsum_fini(&arc_sums.arcstat_l2_evict_l1cached);
	wmsum_fini(&arc_sums.arcstat_l2_free_on_write);
	wmsum_fini(&arc_sums.arcstat_l2_abort_lowmem);
	wmsum_fini(&arc_sums.arcstat_l2_cksum_bad);
	wmsum_fini(&arc_sums.arcstat_l2_io_error);
	wmsum_fini(&arc_sums.arcstat_l2_lsize);
	wmsum_fini(&arc_sums.arcstat_l2_psize);
	aggsum_fini(&arc_sums.arcstat_l2_hdr_size);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_writes);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_asize);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_count);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_success);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_size);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_asize);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs_precached);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_log_blks);
	wmsum_fini(&arc_sums.arcstat_memory_throttle_count);
	wmsum_fini(&arc_sums.arcstat_memory_direct_count);
	wmsum_fini(&arc_sums.arcstat_memory_indirect_count);
	wmsum_fini(&arc_sums.arcstat_prune);
	aggsum_fini(&arc_sums.arcstat_meta_used);
	wmsum_fini(&arc_sums.arcstat_async_upgrade_sync);
	wmsum_fini(&arc_sums.arcstat_demand_hit_predictive_prefetch);
	wmsum_fini(&arc_sums.arcstat_demand_hit_prescient_prefetch);
	wmsum_fini(&arc_sums.arcstat_raw_size);
	wmsum_fini(&arc_sums.arcstat_cached_only_in_progress);
	wmsum_fini(&arc_sums.arcstat_abd_chunk_waste_size);
	}

	uint64_t
	arc_target_bytes(void)
	{
	return (arc_c);
	}

	void
	arc_set_limits(uint64_t allmem)
	{
	/* Set min cache to 1/32 of all memory, or 32MB, whichever is more. */
	arc_c_min = MAX(allmem / 32, 2ULL << SPA_MAXBLOCKSHIFT);

	/* How to set default max varies by platform. */
	arc_c_max = arc_default_max(arc_c_min, allmem);
	}
	void
	arc_init(void)
	{
	uint64_t percent, allmem = arc_all_memory();
	mutex_init(&arc_evict_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&arc_evict_waiters, sizeof (arc_evict_waiter_t),
	offsetof(arc_evict_waiter_t, aew_node));

	arc_min_prefetch_ms = 1000;
	arc_min_prescient_prefetch_ms = 6000;

	#if defined(_KERNEL)
	arc_lowmem_init();
	#endif

	arc_set_limits(allmem);

	#ifdef _KERNEL
	/*
	* If zfs_arc_max is non-zero at init, meaning it was set in the kernel
	* environment before the module was loaded, don't block setting the
	* maximum because it is less than arc_c_min, instead, reset arc_c_min
	* to a lower value.
	* zfs_arc_min will be handled by arc_tuning_update().
	*/
	if (zfs_arc_max != 0 && zfs_arc_max >= MIN_ARC_MAX &&
	zfs_arc_max < allmem) {
	arc_c_max = zfs_arc_max;
	if (arc_c_min >= arc_c_max) {
	arc_c_min = MAX(zfs_arc_max / 2,
	2ULL << SPA_MAXBLOCKSHIFT);
	}
	}
	#else
	/*
	* In userland, there's only the memory pressure that we artificially
	* create (see arc_available_memory()). Don't let arc_c get too
	* small, because it can cause transactions to be larger than
	* arc_c, causing arc_tempreserve_space() to fail.
	*/
	arc_c_min = MAX(arc_c_max / 2, 2ULL << SPA_MAXBLOCKSHIFT);
	#endif

	arc_c = arc_c_min;
	arc_p = (arc_c >> 1);

	/* Set min to 1/2 of arc_c_min */
	arc_meta_min = 1ULL << SPA_MAXBLOCKSHIFT;
	/*
	* Set arc_meta_limit to a percent of arc_c_max with a floor of
	* arc_meta_min, and a ceiling of arc_c_max.
	*/
	percent = MIN(zfs_arc_meta_limit_percent, 100);
	arc_meta_limit = MAX(arc_meta_min, (percent * arc_c_max) / 100);
	percent = MIN(zfs_arc_dnode_limit_percent, 100);
	arc_dnode_size_limit = (percent * arc_meta_limit) / 100;

	/* Apply user specified tunings */
	arc_tuning_update(B_TRUE);

	/* if kmem_flags are set, lets try to use less memory */
	if (kmem_debugging())
	arc_c = arc_c / 2;
	if (arc_c < arc_c_min)
	arc_c = arc_c_min;

	arc_register_hotplug();

	arc_state_init();

	buf_init();

	list_create(&arc_prune_list, sizeof (arc_prune_t),
	offsetof(arc_prune_t, p_node));
	mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);

	arc_prune_taskq = taskq_create("arc_prune", zfs_arc_prune_task_threads,
	defclsyspri, 100, INT_MAX, TASKQ_PREPOPULATE \| TASKQ_DYNAMIC);

	arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
	sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);

	if (arc_ksp != NULL) {
	arc_ksp->ks_data = &arc_stats;
	arc_ksp->ks_update = arc_kstat_update;
	kstat_install(arc_ksp);
	}

	arc_state_evict_markers =
	arc_state_alloc_markers(arc_state_evict_marker_count);
	arc_evict_zthr = zthr_create("arc_evict",
	arc_evict_cb_check, arc_evict_cb, NULL, defclsyspri);
	arc_reap_zthr = zthr_create_timer("arc_reap",
	arc_reap_cb_check, arc_reap_cb, NULL, SEC2NSEC(1), minclsyspri);

	arc_warm = B_FALSE;

	/*
	* Calculate maximum amount of dirty data per pool.
	*
	* If it has been set by a module parameter, take that.
	* Otherwise, use a percentage of physical memory defined by
	* zfs_dirty_data_max_percent (default 10%) with a cap at
	* zfs_dirty_data_max_max (default 4G or 25% of physical memory).
	*/
	#ifdef __LP64__
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#else
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(1ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#endif

	if (zfs_dirty_data_max == 0) {
	zfs_dirty_data_max = allmem *
	zfs_dirty_data_max_percent / 100;
	zfs_dirty_data_max = MIN(zfs_dirty_data_max,
	zfs_dirty_data_max_max);
	}

	if (zfs_wrlog_data_max == 0) {

	/*
	* dp_wrlog_total is reduced for each txg at the end of
	* spa_sync(). However, dp_dirty_total is reduced every time
	* a block is written out. Thus under normal operation,
	* dp_wrlog_total could grow 2 times as big as
	* zfs_dirty_data_max.
	*/
	zfs_wrlog_data_max = zfs_dirty_data_max * 2;
	}
	}

	void
	arc_fini(void)
	{
	arc_prune_t *p;

	#ifdef _KERNEL
	arc_lowmem_fini();
	#endif /* _KERNEL */

	/* Use B_TRUE to ensure all buffers are evicted */
	arc_flush(NULL, B_TRUE);

	if (arc_ksp != NULL) {
	kstat_delete(arc_ksp);
	arc_ksp = NULL;
	}

	taskq_wait(arc_prune_taskq);
	taskq_destroy(arc_prune_taskq);

	mutex_enter(&arc_prune_mtx);
	while ((p = list_head(&arc_prune_list)) != NULL) {
	list_remove(&arc_prune_list, p);
	zfs_refcount_remove(&p->p_refcnt, &arc_prune_list);
	zfs_refcount_destroy(&p->p_refcnt);
	kmem_free(p, sizeof (*p));
	}
	mutex_exit(&arc_prune_mtx);

	list_destroy(&arc_prune_list);
	mutex_destroy(&arc_prune_mtx);

	(void) zthr_cancel(arc_evict_zthr);
	(void) zthr_cancel(arc_reap_zthr);
	arc_state_free_markers(arc_state_evict_markers,
	arc_state_evict_marker_count);

	mutex_destroy(&arc_evict_lock);
	list_destroy(&arc_evict_waiters);

	/*
	* Free any buffers that were tagged for destruction. This needs
	* to occur before arc_state_fini() runs and destroys the aggsum
	* values which are updated when freeing scatter ABDs.
	*/
	l2arc_do_free_on_write();

	/*
	* buf_fini() must proceed arc_state_fini() because buf_fin() may
	* trigger the release of kmem magazines, which can callback to
	* arc_space_return() which accesses aggsums freed in act_state_fini().
	*/
	buf_fini();
	arc_state_fini();

	arc_unregister_hotplug();

	/*
	* We destroy the zthrs after all the ARC state has been
	* torn down to avoid the case of them receiving any
	* wakeup() signals after they are destroyed.
	*/
	zthr_destroy(arc_evict_zthr);
	zthr_destroy(arc_reap_zthr);

	ASSERT0(arc_loaned_bytes);
	}

	/*
	* Level 2 ARC
	*
	* The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk.
	* It uses dedicated storage devices to hold cached data, which are populated
	* using large infrequent writes. The main role of this cache is to boost
	* the performance of random read workloads. The intended L2ARC devices
	* include short-stroked disks, solid state disks, and other media with
	* substantially faster read latency than disk.
	*
	* +-----------------------+
	* \| ARC \|
	* +-----------------------+
	* \| ^ ^
	* \| \| \|
	* l2arc_feed_thread() arc_read()
	* \| \| \|
	* \| l2arc read \|
	* V \| \|
	* +---------------+ \|
	* \| L2ARC \| \|
	* +---------------+ \|
	* \| ^ \|
	* l2arc_write() \| \|
	* \| \| \|
	* V \| \|
	* +-------+ +-------+
	* \| vdev \| \| vdev \|
	* \| cache \| \| cache \|
	* +-------+ +-------+
	* +=========+ .-----.
	* : L2ARC : \|-_____-\|
	* : devices : \| Disks \|
	* +=========+ `-_____-'
	*
	* Read requests are satisfied from the following sources, in order:
	*
	* 1) ARC
	* 2) vdev cache of L2ARC devices
	* 3) L2ARC devices
	* 4) vdev cache of disks
	* 5) disks
	*
	* Some L2ARC device types exhibit extremely slow write performance.
	* To accommodate for this there are some significant differences between
	* the L2ARC and traditional cache design:
	*
	* 1. There is no eviction path from the ARC to the L2ARC. Evictions from
	* the ARC behave as usual, freeing buffers and placing headers on ghost
	* lists. The ARC does not send buffers to the L2ARC during eviction as
	* this would add inflated write latencies for all ARC memory pressure.
	*
	* 2. The L2ARC attempts to cache data from the ARC before it is evicted.
	* It does this by periodically scanning buffers from the eviction-end of
	* the MFU and MRU ARC lists, copying them to the L2ARC devices if they are
	* not already there. It scans until a headroom of buffers is satisfied,
	* which itself is a buffer for ARC eviction. If a compressible buffer is
	* found during scanning and selected for writing to an L2ARC device, we
	* temporarily boost scanning headroom during the next scan cycle to make
	* sure we adapt to compression effects (which might significantly reduce
	* the data volume we write to L2ARC). The thread that does this is
	* l2arc_feed_thread(), illustrated below; example sizes are included to
	* provide a better sense of ratio than this diagram:
	*
	* head --> tail
	* +---------------------+----------+
	* ARC_mfu \|:::::#:::::::::::::::\|o#o###o###\|-->. # already on L2ARC
	* +---------------------+----------+ \| o L2ARC eligible
	* ARC_mru \|:#:::::::::::::::::::\|#o#ooo####\|-->\| : ARC buffer
	* +---------------------+----------+ \|
	* 15.9 Gbytes ^ 32 Mbytes \|
	* headroom \|
	* l2arc_feed_thread()
	* \|
	* l2arc write hand <--[oooo]--'
	* \| 8 Mbyte
	* \| write max
	* V
	* +==============================+
	* L2ARC dev \|####\|#\|###\|###\| \|####\| ... \|
	* +==============================+
	* 32 Gbytes
	*
	* 3. If an ARC buffer is copied to the L2ARC but then hit instead of
	* evicted, then the L2ARC has cached a buffer much sooner than it probably
	* needed to, potentially wasting L2ARC device bandwidth and storage. It is
	* safe to say that this is an uncommon case, since buffers at the end of
	* the ARC lists have moved there due to inactivity.
	*
	* 4. If the ARC evicts faster than the L2ARC can maintain a headroom,
	* then the L2ARC simply misses copying some buffers. This serves as a
	* pressure valve to prevent heavy read workloads from both stalling the ARC
	* with waits and clogging the L2ARC with writes. This also helps prevent
	* the potential for the L2ARC to churn if it attempts to cache content too
	* quickly, such as during backups of the entire pool.
	*
	* 5. After system boot and before the ARC has filled main memory, there are
	* no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru
	* lists can remain mostly static. Instead of searching from tail of these
	* lists as pictured, the l2arc_feed_thread() will search from the list heads
	* for eligible buffers, greatly increasing its chance of finding them.
	*
	* The L2ARC device write speed is also boosted during this time so that
	* the L2ARC warms up faster. Since there have been no ARC evictions yet,
	* there are no L2ARC reads, and no fear of degrading read performance
	* through increased writes.
	*
	* 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that
	* the vdev queue can aggregate them into larger and fewer writes. Each
	* device is written to in a rotor fashion, sweeping writes through
	* available space then repeating.
	*
	* 7. The L2ARC does not store dirty content. It never needs to flush
	* write buffers back to disk based storage.
	*
	* 8. If an ARC buffer is written (and dirtied) which also exists in the
	* L2ARC, the now stale L2ARC buffer is immediately dropped.
	*
	* The performance of the L2ARC can be tweaked by a number of tunables, which
	* may be necessary for different workloads:
	*
	* l2arc_write_max max write bytes per interval
	* l2arc_write_boost extra write bytes during device warmup
	* l2arc_noprefetch skip caching prefetched buffers
	* l2arc_headroom number of max device writes to precache
	* l2arc_headroom_boost when we find compressed buffers during ARC
	* scanning, we multiply headroom by this
	* percentage factor for the next scan cycle,
	* since more compressed buffers are likely to
	* be present
	* l2arc_feed_secs seconds between L2ARC writing
	*
	* Tunables may be removed or added as future performance improvements are
	* integrated, and also may become zpool properties.
	*
	* There are three key functions that control how the L2ARC warms up:
	*
	* l2arc_write_eligible() check if a buffer is eligible to cache
	* l2arc_write_size() calculate how much to write
	* l2arc_write_interval() calculate sleep delay between writes
	*
	* These three functions determine what to write, how much, and how quickly
	* to send writes.
	*
	* L2ARC persistence:
	*
	* When writing buffers to L2ARC, we periodically add some metadata to
	* make sure we can pick them up after reboot, thus dramatically reducing
	* the impact that any downtime has on the performance of storage systems
	* with large caches.
	*
	* The implementation works fairly simply by integrating the following two
	* modifications:
	*
	* *) When writing to the L2ARC, we occasionally write a "l2arc log block",
	* which is an additional piece of metadata which describes what's been
	* written. This allows us to rebuild the arc_buf_hdr_t structures of the
	* main ARC buffers. There are 2 linked-lists of log blocks headed by
	* dh_start_lbps[2]. We alternate which chain we append to, so they are
	* time-wise and offset-wise interleaved, but that is an optimization rather
	* than for correctness. The log block also includes a pointer to the
	* previous block in its chain.
	*
	* *) We reserve SPA_MINBLOCKSIZE of space at the start of each L2ARC device
	* for our header bookkeeping purposes. This contains a device header,
	* which contains our top-level reference structures. We update it each
	* time we write a new log block, so that we're able to locate it in the
	* L2ARC device. If this write results in an inconsistent device header
	* (e.g. due to power failure), we detect this by verifying the header's
	* checksum and simply fail to reconstruct the L2ARC after reboot.
	*
	* Implementation diagram:
	*
	* +=== L2ARC device (not to scale) ======================================+
	* \| ___two newest log block pointers__.__________ \|
	* \| / \dh_start_lbps[1] \|
	* \| / \ \dh_start_lbps[0]\|
	* \|.___/__. V V \|
	* \|\|L2 dev\|....\|lb \|bufs \|lb \|bufs \|lb \|bufs \|lb \|bufs \|lb \|---(empty)---\|
	* \|\| hdr\| ^ /^ /^ / / \|
	* \|+------+ ...--\-------/ \-----/--\------/ / \|
	* \| \--------------/ \--------------/ \|
	* +======================================================================+
	*
	* As can be seen on the diagram, rather than using a simple linked list,
	* we use a pair of linked lists with alternating elements. This is a
	* performance enhancement due to the fact that we only find out the
	* address of the next log block access once the current block has been
	* completely read in. Obviously, this hurts performance, because we'd be
	* keeping the device's I/O queue at only a 1 operation deep, thus
	* incurring a large amount of I/O round-trip latency. Having two lists
	* allows us to fetch two log blocks ahead of where we are currently
	* rebuilding L2ARC buffers.
	*
	* On-device data structures:
	*
	* L2ARC device header: l2arc_dev_hdr_phys_t
	* L2ARC log block: l2arc_log_blk_phys_t
	*
	* L2ARC reconstruction:
	*
	* When writing data, we simply write in the standard rotary fashion,
	* evicting buffers as we go and simply writing new data over them (writing
	* a new log block every now and then). This obviously means that once we
	* loop around the end of the device, we will start cutting into an already
	* committed log block (and its referenced data buffers), like so:
	*
	* current write head__ __old tail
	* \ /
	* V V
	* <--\|bufs \|lb \|bufs \|lb \| \|bufs \|lb \|bufs \|lb \|-->
	* ^ ^^^^^^^^^___________________________________
	* \| \
	* <<nextwrite>> may overwrite this blk and/or its bufs --'
	*
	* When importing the pool, we detect this situation and use it to stop
	* our scanning process (see l2arc_rebuild).
	*
	* There is one significant caveat to consider when rebuilding ARC contents
	* from an L2ARC device: what about invalidated buffers? Given the above
	* construction, we cannot update blocks which we've already written to amend
	* them to remove buffers which were invalidated. Thus, during reconstruction,
	* we might be populating the cache with buffers for data that's not on the
	* main pool anymore, or may have been overwritten!
	*
	* As it turns out, this isn't a problem. Every arc_read request includes
	* both the DVA and, crucially, the birth TXG of the BP the caller is
	* looking for. So even if the cache were populated by completely rotten
	* blocks for data that had been long deleted and/or overwritten, we'll
	* never actually return bad data from the cache, since the DVA with the
	* birth TXG uniquely identify a block in space and time - once created,
	* a block is immutable on disk. The worst thing we have done is wasted
	* some time and memory at l2arc rebuild to reconstruct outdated ARC
	* entries that will get dropped from the l2arc as it is being updated
	* with new blocks.
	*
	* L2ARC buffers that have been evicted by l2arc_evict() ahead of the write
	* hand are not restored. This is done by saving the offset (in bytes)
	* l2arc_evict() has evicted to in the L2ARC device header and taking it
	* into account when restoring buffers.
	*/

	static boolean_t
	l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr)
	{
	/*
	* A buffer is not eligible for the L2ARC if it:
	* 1. belongs to a different spa.
	* 2. is already cached on the L2ARC.
	* 3. has an I/O in progress (it may be an incomplete read).
	* 4. is flagged not eligible (zfs property).
	*/
	if (hdr->b_spa != spa_guid \|\| HDR_HAS_L2HDR(hdr) \|\|
	HDR_IO_IN_PROGRESS(hdr) \|\| !HDR_L2CACHE(hdr))
	return (B_FALSE);

	return (B_TRUE);
	}

	static uint64_t
	l2arc_write_size(l2arc_dev_t *dev)
	{
	uint64_t size, dev_size, tsize;

	/*
	* Make sure our globals have meaningful values in case the user
	* altered them.
	*/
	size = l2arc_write_max;
	if (size == 0) {
	cmn_err(CE_NOTE, "Bad value for l2arc_write_max, value must "
	"be greater than zero, resetting it to the default (%d)",
	L2ARC_WRITE_SIZE);
	size = l2arc_write_max = L2ARC_WRITE_SIZE;
	}

	if (arc_warm == B_FALSE)
	size += l2arc_write_boost;

	/*
	* Make sure the write size does not exceed the size of the cache
	* device. This is important in l2arc_evict(), otherwise infinite
	* iteration can occur.
	*/
	dev_size = dev->l2ad_end - dev->l2ad_start;
	tsize = size + l2arc_log_blk_overhead(size, dev);
	if (dev->l2ad_vdev->vdev_has_trim && l2arc_trim_ahead > 0)
	tsize += MAX(64 * 1024 * 1024,
	(tsize * l2arc_trim_ahead) / 100);

	if (tsize >= dev_size) {
	cmn_err(CE_NOTE, "l2arc_write_max or l2arc_write_boost "
	"plus the overhead of log blocks (persistent L2ARC, "
	"%llu bytes) exceeds the size of the cache device "
	"(guid %llu), resetting them to the default (%d)",
	(u_longlong_t)l2arc_log_blk_overhead(size, dev),
	(u_longlong_t)dev->l2ad_vdev->vdev_guid, L2ARC_WRITE_SIZE);
	size = l2arc_write_max = l2arc_write_boost = L2ARC_WRITE_SIZE;

	if (arc_warm == B_FALSE)
	size += l2arc_write_boost;
	}

	return (size);

	}

	static clock_t
	l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote)
	{
	clock_t interval, next, now;

	/*
	* If the ARC lists are busy, increase our write rate; if the
	* lists are stale, idle back. This is achieved by checking
	* how much we previously wrote - if it was more than half of
	* what we wanted, schedule the next write much sooner.
	*/
	if (l2arc_feed_again && wrote > (wanted / 2))
	interval = (hz * l2arc_feed_min_ms) / 1000;
	else
	interval = hz * l2arc_feed_secs;

	now = ddi_get_lbolt();
	next = MAX(now, MIN(now + interval, began + interval));

	return (next);
	}

	/*
	* Cycle through L2ARC devices. This is how L2ARC load balances.
	* If a device is returned, this also returns holding the spa config lock.
	*/
	static l2arc_dev_t *
	l2arc_dev_get_next(void)
	{
	l2arc_dev_t first, next = NULL;

	/*
	* Lock out the removal of spas (spa_namespace_lock), then removal
	* of cache devices (l2arc_dev_mtx). Once a device has been selected,
	* both locks will be dropped and a spa config lock held instead.
	*/
	mutex_enter(&spa_namespace_lock);
	mutex_enter(&l2arc_dev_mtx);

	/* if there are no vdevs, there is nothing to do */
	if (l2arc_ndev == 0)
	goto out;

	first = NULL;
	next = l2arc_dev_last;
	do {
	/* loop around the list looking for a non-faulted vdev */
	if (next == NULL) {
	next = list_head(l2arc_dev_list);
	} else {
	next = list_next(l2arc_dev_list, next);
	if (next == NULL)
	next = list_head(l2arc_dev_list);
	}

	/* if we have come back to the start, bail out */
	if (first == NULL)
	first = next;
	else if (next == first)
	break;

	} while (vdev_is_dead(next->l2ad_vdev) \|\| next->l2ad_rebuild \|\|
	next->l2ad_trim_all);

	/* if we were unable to find any usable vdevs, return NULL */
	if (vdev_is_dead(next->l2ad_vdev) \|\| next->l2ad_rebuild \|\|
	next->l2ad_trim_all)
	next = NULL;

	l2arc_dev_last = next;

	out:
	mutex_exit(&l2arc_dev_mtx);

	/*
	* Grab the config lock to prevent the 'next' device from being
	* removed while we are writing to it.
	*/
	if (next != NULL)
	spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER);
	mutex_exit(&spa_namespace_lock);

	return (next);
	}

	/*
	* Free buffers that were tagged for destruction.
	*/
	static void
	l2arc_do_free_on_write(void)
	{
	list_t *buflist;
	l2arc_data_free_t df, df_prev;

	mutex_enter(&l2arc_free_on_write_mtx);
	buflist = l2arc_free_on_write;

	for (df = list_tail(buflist); df; df = df_prev) {
	df_prev = list_prev(buflist, df);
	ASSERT3P(df->l2df_abd, !=, NULL);
	abd_free(df->l2df_abd);
	list_remove(buflist, df);
	kmem_free(df, sizeof (l2arc_data_free_t));
	}

	mutex_exit(&l2arc_free_on_write_mtx);
	}

	/*
	* A write to a cache device has completed. Update all headers to allow
	* reads from these buffers to begin.
	*/
	static void
	l2arc_write_done(zio_t *zio)
	{
	l2arc_write_callback_t *cb;
	l2arc_lb_abd_buf_t *abd_buf;
	l2arc_lb_ptr_buf_t *lb_ptr_buf;
	l2arc_dev_t *dev;
	l2arc_dev_hdr_phys_t *l2dhdr;
	list_t *buflist;
	arc_buf_hdr_t head, hdr, *hdr_prev;
	kmutex_t *hash_lock;
	int64_t bytes_dropped = 0;

	cb = zio->io_private;
	ASSERT3P(cb, !=, NULL);
	dev = cb->l2wcb_dev;
	l2dhdr = dev->l2ad_dev_hdr;
	ASSERT3P(dev, !=, NULL);
	head = cb->l2wcb_head;
	ASSERT3P(head, !=, NULL);
	buflist = &dev->l2ad_buflist;
	ASSERT3P(buflist, !=, NULL);
	DTRACE_PROBE2(l2arc__iodone, zio_t *, zio,
	l2arc_write_callback_t *, cb);

	/*
	* All writes completed, or an error was hit.
	*/
	top:
	mutex_enter(&dev->l2ad_mtx);
	for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) {
	hdr_prev = list_prev(buflist, hdr);

	hash_lock = HDR_LOCK(hdr);

	/*
	* We cannot use mutex_enter or else we can deadlock
	* with l2arc_write_buffers (due to swapping the order
	* the hash lock and l2ad_mtx are taken).
	*/
	if (!mutex_tryenter(hash_lock)) {
	/*
	* Missed the hash lock. We must retry so we
	* don't leave the ARC_FLAG_L2_WRITING bit set.
	*/
	ARCSTAT_BUMP(arcstat_l2_writes_lock_retry);

	/*
	* We don't want to rescan the headers we've
	* already marked as having been written out, so
	* we reinsert the head node so we can pick up
	* where we left off.
	*/
	list_remove(buflist, head);
	list_insert_after(buflist, hdr, head);

	mutex_exit(&dev->l2ad_mtx);

	/*
	* We wait for the hash lock to become available
	* to try and prevent busy waiting, and increase
	* the chance we'll be able to acquire the lock
	* the next time around.
	*/
	mutex_enter(hash_lock);
	mutex_exit(hash_lock);
	goto top;
	}

	/*
	* We could not have been moved into the arc_l2c_only
	* state while in-flight due to our ARC_FLAG_L2_WRITING
	* bit being set. Let's just ensure that's being enforced.
	*/
	ASSERT(HDR_HAS_L1HDR(hdr));

	/*
	* Skipped - drop L2ARC entry and mark the header as no
	* longer L2 eligibile.
	*/
	if (zio->io_error != 0) {
	/*
	* Error - drop L2ARC entry.
	*/
	list_remove(buflist, hdr);
	arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);

	uint64_t psize = HDR_GET_PSIZE(hdr);
	l2arc_hdr_arcstats_decrement(hdr);

	bytes_dropped +=
	vdev_psize_to_asize(dev->l2ad_vdev, psize);
	(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);
	}

	/*
	* Allow ARC to begin reads and ghost list evictions to
	* this L2ARC entry.
	*/
	arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING);

	mutex_exit(hash_lock);
	}

	/*
	* Free the allocated abd buffers for writing the log blocks.
	* If the zio failed reclaim the allocated space and remove the
	* pointers to these log blocks from the log block pointer list
	* of the L2ARC device.
	*/
	while ((abd_buf = list_remove_tail(&cb->l2wcb_abd_list)) != NULL) {
	abd_free(abd_buf->abd);
	zio_buf_free(abd_buf, sizeof (*abd_buf));
	if (zio->io_error != 0) {
	lb_ptr_buf = list_remove_head(&dev->l2ad_lbptr_list);
	/*
	* L2BLK_GET_PSIZE returns aligned size for log
	* blocks.
	*/
	uint64_t asize =
	L2BLK_GET_PSIZE((lb_ptr_buf->lb_ptr)->lbp_prop);
	bytes_dropped += asize;
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
	ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
	zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
	lb_ptr_buf);
	zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf);
	kmem_free(lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
	}
	}
	list_destroy(&cb->l2wcb_abd_list);

	if (zio->io_error != 0) {
	ARCSTAT_BUMP(arcstat_l2_writes_error);

	/*
	* Restore the lbps array in the header to its previous state.
	* If the list of log block pointers is empty, zero out the
	* log block pointers in the device header.
	*/
	lb_ptr_buf = list_head(&dev->l2ad_lbptr_list);
	for (int i = 0; i < 2; i++) {
	if (lb_ptr_buf == NULL) {
	/*
	* If the list is empty zero out the device
	* header. Otherwise zero out the second log
	* block pointer in the header.
	*/
	if (i == 0) {
	memset(l2dhdr, 0,
	dev->l2ad_dev_hdr_asize);
	} else {
	memset(&l2dhdr->dh_start_lbps[i], 0,
	sizeof (l2arc_log_blkptr_t));
	}
	break;
	}
	memcpy(&l2dhdr->dh_start_lbps[i], lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	lb_ptr_buf = list_next(&dev->l2ad_lbptr_list,
	lb_ptr_buf);
	}
	}

	ARCSTAT_BUMP(arcstat_l2_writes_done);
	list_remove(buflist, head);
	ASSERT(!HDR_HAS_L1HDR(head));
	kmem_cache_free(hdr_l2only_cache, head);
	mutex_exit(&dev->l2ad_mtx);

	ASSERT(dev->l2ad_vdev != NULL);
	vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0);

	l2arc_do_free_on_write();

	kmem_free(cb, sizeof (l2arc_write_callback_t));
	}

	static int
	l2arc_untransform(zio_t zio, l2arc_read_callback_t cb)
	{
	int ret;
	spa_t *spa = zio->io_spa;
	arc_buf_hdr_t *hdr = cb->l2rcb_hdr;
	blkptr_t *bp = zio->io_bp;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	/*
	* ZIL data is never be written to the L2ARC, so we don't need
	* special handling for its unique MAC storage.
	*/
	ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);
	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	/*
	* If the data was encrypted, decrypt it now. Note that
	* we must check the bp here and not the hdr, since the
	* hdr does not have its encryption parameters updated
	* until arc_read_done().
	*/
	if (BP_IS_ENCRYPTED(bp)) {
	abd_t *eabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
	ARC_HDR_DO_ADAPT \| ARC_HDR_USE_RESERVE);

	zio_crypt_decode_params_bp(bp, salt, iv);
	zio_crypt_decode_mac_bp(bp, mac);

	ret = spa_do_crypt_abd(B_FALSE, spa, &cb->l2rcb_zb,
	BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
	salt, iv, mac, HDR_GET_PSIZE(hdr), eabd,
	hdr->b_l1hdr.b_pabd, &no_crypt);
	if (ret != 0) {
	arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
	goto error;
	}

	/*
	* If we actually performed decryption, replace b_pabd
	* with the decrypted data. Otherwise we can just throw
	* our decryption buffer away.
	*/
	if (!no_crypt) {
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = eabd;
	zio->io_abd = eabd;
	} else {
	arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
	}
	}

	/*
	* If the L2ARC block was compressed, but ARC compression
	* is disabled we decompress the data into a new buffer and
	* replace the existing data.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {
	abd_t *cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
	ARC_HDR_DO_ADAPT \| ARC_HDR_USE_RESERVE);
	void *tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));

	ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
	HDR_GET_LSIZE(hdr), &hdr->b_complevel);
	if (ret != 0) {
	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr);
	goto error;
	}

	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = cabd;
	zio->io_abd = cabd;
	zio->io_size = HDR_GET_LSIZE(hdr);
	}

	return (0);

	error:
	return (ret);
	}


	/*
	* A read to a cache device completed. Validate buffer contents before
	* handing over to the regular ARC routines.
	*/
	static void
	l2arc_read_done(zio_t *zio)
	{
	int tfm_error = 0;
	l2arc_read_callback_t *cb = zio->io_private;
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	boolean_t valid_cksum;
	boolean_t using_rdata = (BP_IS_ENCRYPTED(&cb->l2rcb_bp) &&
	(cb->l2rcb_flags & ZIO_FLAG_RAW_ENCRYPT));

	ASSERT3P(zio->io_vd, !=, NULL);
	ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE);

	spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd);

	ASSERT3P(cb, !=, NULL);
	hdr = cb->l2rcb_hdr;
	ASSERT3P(hdr, !=, NULL);

	hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));

	/*
	* If the data was read into a temporary buffer,
	* move it and free the buffer.
	*/
	if (cb->l2rcb_abd != NULL) {
	ASSERT3U(arc_hdr_size(hdr), <, zio->io_size);
	if (zio->io_error == 0) {
	if (using_rdata) {
	abd_copy(hdr->b_crypt_hdr.b_rabd,
	cb->l2rcb_abd, arc_hdr_size(hdr));
	} else {
	abd_copy(hdr->b_l1hdr.b_pabd,
	cb->l2rcb_abd, arc_hdr_size(hdr));
	}
	}

	/*
	* The following must be done regardless of whether
	* there was an error:
	* - free the temporary buffer
	* - point zio to the real ARC buffer
	* - set zio size accordingly
	* These are required because zio is either re-used for
	* an I/O of the block in the case of the error
	* or the zio is passed to arc_read_done() and it
	* needs real data.
	*/
	abd_free(cb->l2rcb_abd);
	zio->io_size = zio->io_orig_size = arc_hdr_size(hdr);

	if (using_rdata) {
	ASSERT(HDR_HAS_RABD(hdr));
	zio->io_abd = zio->io_orig_abd =
	hdr->b_crypt_hdr.b_rabd;
	} else {
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	zio->io_abd = zio->io_orig_abd = hdr->b_l1hdr.b_pabd;
	}
	}

	ASSERT3P(zio->io_abd, !=, NULL);

	/*
	* Check this survived the L2ARC journey.
	*/
	ASSERT(zio->io_abd == hdr->b_l1hdr.b_pabd \|\|
	(HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd));
	zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
	zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
	zio->io_prop.zp_complevel = hdr->b_complevel;

	valid_cksum = arc_cksum_is_equal(hdr, zio);

	/*
	* b_rabd will always match the data as it exists on disk if it is
	* being used. Therefore if we are reading into b_rabd we do not
	* attempt to untransform the data.
	*/
	if (valid_cksum && !using_rdata)
	tfm_error = l2arc_untransform(zio, cb);

	if (valid_cksum && tfm_error == 0 && zio->io_error == 0 &&
	!HDR_L2_EVICTED(hdr)) {
	mutex_exit(hash_lock);
	zio->io_private = hdr;
	arc_read_done(zio);
	} else {
	/*
	* Buffer didn't survive caching. Increment stats and
	* reissue to the original storage device.
	*/
	if (zio->io_error != 0) {
	ARCSTAT_BUMP(arcstat_l2_io_error);
	} else {
	zio->io_error = SET_ERROR(EIO);
	}
	if (!valid_cksum \|\| tfm_error != 0)
	ARCSTAT_BUMP(arcstat_l2_cksum_bad);

	/*
	* If there's no waiter, issue an async i/o to the primary
	* storage now. If there is a waiter, the caller must
	* issue the i/o in a context where it's OK to block.
	*/
	if (zio->io_waiter == NULL) {
	zio_t *pio = zio_unique_parent(zio);
	void *abd = (using_rdata) ?
	hdr->b_crypt_hdr.b_rabd : hdr->b_l1hdr.b_pabd;

	ASSERT(!pio \|\| pio->io_child_type == ZIO_CHILD_LOGICAL);

	zio = zio_read(pio, zio->io_spa, zio->io_bp,
	abd, zio->io_size, arc_read_done,
	hdr, zio->io_priority, cb->l2rcb_flags,
	&cb->l2rcb_zb);

	/*
	* Original ZIO will be freed, so we need to update
	* ARC header with the new ZIO pointer to be used
	* by zio_change_priority() in arc_read().
	*/
	for (struct arc_callback *acb = hdr->b_l1hdr.b_acb;
	acb != NULL; acb = acb->acb_next)
	acb->acb_zio_head = zio;

	mutex_exit(hash_lock);
	zio_nowait(zio);
	} else {
	mutex_exit(hash_lock);
	}
	}

	kmem_free(cb, sizeof (l2arc_read_callback_t));
	}

	/*
	* This is the list priority from which the L2ARC will search for pages to
	* cache. This is used within loops (0..3) to cycle through lists in the
	* desired order. This order can have a significant effect on cache
	* performance.
	*
	* Currently the metadata lists are hit first, MFU then MRU, followed by
	* the data lists. This function returns a locked list, and also returns
	* the lock pointer.
	*/
	static multilist_sublist_t *
	l2arc_sublist_lock(int list_num)
	{
	multilist_t *ml = NULL;
	unsigned int idx;

	ASSERT(list_num >= 0 && list_num < L2ARC_FEED_TYPES);

	switch (list_num) {
	case 0:
	ml = &arc_mfu->arcs_list[ARC_BUFC_METADATA];
	break;
	case 1:
	ml = &arc_mru->arcs_list[ARC_BUFC_METADATA];
	break;
	case 2:
	ml = &arc_mfu->arcs_list[ARC_BUFC_DATA];
	break;
	case 3:
	ml = &arc_mru->arcs_list[ARC_BUFC_DATA];
	break;
	default:
	return (NULL);
	}

	/*
	* Return a randomly-selected sublist. This is acceptable
	* because the caller feeds only a little bit of data for each
	* call (8MB). Subsequent calls will result in different
	* sublists being selected.
	*/
	idx = multilist_get_random_index(ml);
	return (multilist_sublist_lock(ml, idx));
	}

	/*
	* Calculates the maximum overhead of L2ARC metadata log blocks for a given
	* L2ARC write size. l2arc_evict and l2arc_write_size need to include this
	* overhead in processing to make sure there is enough headroom available
	* when writing buffers.
	*/
	static inline uint64_t
	l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev)
	{
	if (dev->l2ad_log_entries == 0) {
	return (0);
	} else {
	uint64_t log_entries = write_sz >> SPA_MINBLOCKSHIFT;

	uint64_t log_blocks = (log_entries +
	dev->l2ad_log_entries - 1) /
	dev->l2ad_log_entries;

	return (vdev_psize_to_asize(dev->l2ad_vdev,
	sizeof (l2arc_log_blk_phys_t)) * log_blocks);
	}
	}

	/*
	* Evict buffers from the device write hand to the distance specified in
	* bytes. This distance may span populated buffers, it may span nothing.
	* This is clearing a region on the L2ARC device ready for writing.
	* If the 'all' boolean is set, every buffer is evicted.
	*/
	static void
	l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all)
	{
	list_t *buflist;
	arc_buf_hdr_t hdr, hdr_prev;
	kmutex_t *hash_lock;
	uint64_t taddr;
	l2arc_lb_ptr_buf_t lb_ptr_buf, lb_ptr_buf_prev;
	vdev_t *vd = dev->l2ad_vdev;
	boolean_t rerun;

	buflist = &dev->l2ad_buflist;

	/*
	* We need to add in the worst case scenario of log block overhead.
	*/
	distance += l2arc_log_blk_overhead(distance, dev);
	if (vd->vdev_has_trim && l2arc_trim_ahead > 0) {
	/*
	* Trim ahead of the write size 64MB or (l2arc_trim_ahead/100)
	* times the write size, whichever is greater.
	*/
	distance += MAX(64 * 1024 * 1024,
	(distance * l2arc_trim_ahead) / 100);
	}

	top:
	rerun = B_FALSE;
	if (dev->l2ad_hand >= (dev->l2ad_end - distance)) {
	/*
	* When there is no space to accommodate upcoming writes,
	* evict to the end. Then bump the write and evict hands
	* to the start and iterate. This iteration does not
	* happen indefinitely as we make sure in
	* l2arc_write_size() that when the write hand is reset,
	* the write size does not exceed the end of the device.
	*/
	rerun = B_TRUE;
	taddr = dev->l2ad_end;
	} else {
	taddr = dev->l2ad_hand + distance;
	}
	DTRACE_PROBE4(l2arc__evict, l2arc_dev_t , dev, list_t , buflist,
	uint64_t, taddr, boolean_t, all);

	if (!all) {
	/*
	* This check has to be placed after deciding whether to
	* iterate (rerun).
	*/
	if (dev->l2ad_first) {
	/*
	* This is the first sweep through the device. There is
	* nothing to evict. We have already trimmmed the
	* whole device.
	*/
	goto out;
	} else {
	/*
	* Trim the space to be evicted.
	*/
	if (vd->vdev_has_trim && dev->l2ad_evict < taddr &&
	l2arc_trim_ahead > 0) {
	/*
	* We have to drop the spa_config lock because
	* vdev_trim_range() will acquire it.
	* l2ad_evict already accounts for the label
	* size. To prevent vdev_trim_ranges() from
	* adding it again, we subtract it from
	* l2ad_evict.
	*/
	spa_config_exit(dev->l2ad_spa, SCL_L2ARC, dev);
	vdev_trim_simple(vd,
	dev->l2ad_evict - VDEV_LABEL_START_SIZE,
	taddr - dev->l2ad_evict);
	spa_config_enter(dev->l2ad_spa, SCL_L2ARC, dev,
	RW_READER);
	}

	/*
	* When rebuilding L2ARC we retrieve the evict hand
	* from the header of the device. Of note, l2arc_evict()
	* does not actually delete buffers from the cache
	* device, but trimming may do so depending on the
	* hardware implementation. Thus keeping track of the
	* evict hand is useful.
	*/
	dev->l2ad_evict = MAX(dev->l2ad_evict, taddr);
	}
	}

	retry:
	mutex_enter(&dev->l2ad_mtx);
	/*
	* We have to account for evicted log blocks. Run vdev_space_update()
	* on log blocks whose offset (in bytes) is before the evicted offset
	* (in bytes) by searching in the list of pointers to log blocks
	* present in the L2ARC device.
	*/
	for (lb_ptr_buf = list_tail(&dev->l2ad_lbptr_list); lb_ptr_buf;
	lb_ptr_buf = lb_ptr_buf_prev) {

	lb_ptr_buf_prev = list_prev(&dev->l2ad_lbptr_list, lb_ptr_buf);

	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	uint64_t asize = L2BLK_GET_PSIZE(
	(lb_ptr_buf->lb_ptr)->lbp_prop);

	/*
	* We don't worry about log blocks left behind (ie
	* lbp_payload_start < l2ad_hand) because l2arc_write_buffers()
	* will never write more than l2arc_evict() evicts.
	*/
	if (!all && l2arc_log_blkptr_valid(dev, lb_ptr_buf->lb_ptr)) {
	break;
	} else {
	vdev_space_update(vd, -asize, 0, 0);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
	ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
	zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
	lb_ptr_buf);
	zfs_refcount_remove(&dev->l2ad_lb_count, lb_ptr_buf);
	list_remove(&dev->l2ad_lbptr_list, lb_ptr_buf);
	kmem_free(lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
	}
	}

	for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) {
	hdr_prev = list_prev(buflist, hdr);

	ASSERT(!HDR_EMPTY(hdr));
	hash_lock = HDR_LOCK(hdr);

	/*
	* We cannot use mutex_enter or else we can deadlock
	* with l2arc_write_buffers (due to swapping the order
	* the hash lock and l2ad_mtx are taken).
	*/
	if (!mutex_tryenter(hash_lock)) {
	/*
	* Missed the hash lock. Retry.
	*/
	ARCSTAT_BUMP(arcstat_l2_evict_lock_retry);
	mutex_exit(&dev->l2ad_mtx);
	mutex_enter(hash_lock);
	mutex_exit(hash_lock);
	goto retry;
	}

	/*
	* A header can't be on this list if it doesn't have L2 header.
	*/
	ASSERT(HDR_HAS_L2HDR(hdr));

	/* Ensure this header has finished being written. */
	ASSERT(!HDR_L2_WRITING(hdr));
	ASSERT(!HDR_L2_WRITE_HEAD(hdr));

	if (!all && (hdr->b_l2hdr.b_daddr >= dev->l2ad_evict \|\|
	hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) {
	/*
	* We've evicted to the target address,
	* or the end of the device.
	*/
	mutex_exit(hash_lock);
	break;
	}

	if (!HDR_HAS_L1HDR(hdr)) {
	ASSERT(!HDR_L2_READING(hdr));
	/*
	* This doesn't exist in the ARC. Destroy.
	* arc_hdr_destroy() will call list_remove()
	* and decrement arcstat_l2_lsize.
	*/
	arc_change_state(arc_anon, hdr, hash_lock);
	arc_hdr_destroy(hdr);
	} else {
	ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only);
	ARCSTAT_BUMP(arcstat_l2_evict_l1cached);
	/*
	* Invalidate issued or about to be issued
	* reads, since we may be about to write
	* over this location.
	*/
	if (HDR_L2_READING(hdr)) {
	ARCSTAT_BUMP(arcstat_l2_evict_reading);
	arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED);
	}

	arc_hdr_l2hdr_destroy(hdr);
	}
	mutex_exit(hash_lock);
	}
	mutex_exit(&dev->l2ad_mtx);

	out:
	/*
	* We need to check if we evict all buffers, otherwise we may iterate
	* unnecessarily.
	*/
	if (!all && rerun) {
	/*
	* Bump device hand to the device start if it is approaching the
	* end. l2arc_evict() has already evicted ahead for this case.
	*/
	dev->l2ad_hand = dev->l2ad_start;
	dev->l2ad_evict = dev->l2ad_start;
	dev->l2ad_first = B_FALSE;
	goto top;
	}

	if (!all) {
	/*
	* In case of cache device removal (all) the following
	* assertions may be violated without functional consequences
	* as the device is about to be removed.
	*/
	ASSERT3U(dev->l2ad_hand + distance, <, dev->l2ad_end);
	if (!dev->l2ad_first)
	ASSERT3U(dev->l2ad_hand, <, dev->l2ad_evict);
	}
	}

	/*
	* Handle any abd transforms that might be required for writing to the L2ARC.
	* If successful, this function will always return an abd with the data
	* transformed as it is on disk in a new abd of asize bytes.
	*/
	static int
	l2arc_apply_transforms(spa_t spa, arc_buf_hdr_t hdr, uint64_t asize,
	abd_t **abd_out)
	{
	int ret;
	void *tmp = NULL;
	abd_t cabd = NULL, eabd = NULL, *to_write = hdr->b_l1hdr.b_pabd;
	enum zio_compress compress = HDR_GET_COMPRESS(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t size = arc_hdr_size(hdr);
	boolean_t ismd = HDR_ISTYPE_METADATA(hdr);
	boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
	dsl_crypto_key_t *dck = NULL;
	uint8_t mac[ZIO_DATA_MAC_LEN] = { 0 };
	boolean_t no_crypt = B_FALSE;

	ASSERT((HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) \|\|
	HDR_ENCRYPTED(hdr) \|\| HDR_SHARED_DATA(hdr) \|\| psize != asize);
	ASSERT3U(psize, <=, asize);

	/*
	* If this data simply needs its own buffer, we simply allocate it
	* and copy the data. This may be done to eliminate a dependency on a
	* shared buffer or to reallocate the buffer to match asize.
	*/
	if (HDR_HAS_RABD(hdr) && asize != psize) {
	ASSERT3U(asize, >=, psize);
	to_write = abd_alloc_for_io(asize, ismd);
	abd_copy(to_write, hdr->b_crypt_hdr.b_rabd, psize);
	if (psize != asize)
	abd_zero_off(to_write, psize, asize - psize);
	goto out;
	}

	if ((compress == ZIO_COMPRESS_OFF \|\| HDR_COMPRESSION_ENABLED(hdr)) &&
	!HDR_ENCRYPTED(hdr)) {
	ASSERT3U(size, ==, psize);
	to_write = abd_alloc_for_io(asize, ismd);
	abd_copy(to_write, hdr->b_l1hdr.b_pabd, size);
	if (size != asize)
	abd_zero_off(to_write, size, asize - size);
	goto out;
	}

	if (compress != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) {
	/*
	* In some cases, we can wind up with size > asize, so
	* we need to opt for the larger allocation option here.
	*
	* (We also need abd_return_buf_copy in all cases because
	* it's an ASSERT() to modify the buffer before returning it
	* with arc_return_buf(), and all the compressors
	* write things before deciding to fail compression in nearly
	* every case.)
	*/
	cabd = abd_alloc_for_io(size, ismd);
	tmp = abd_borrow_buf(cabd, size);

	psize = zio_compress_data(compress, to_write, tmp, size,
	hdr->b_complevel);

	if (psize >= asize) {
	psize = HDR_GET_PSIZE(hdr);
	abd_return_buf_copy(cabd, tmp, size);
	HDR_SET_COMPRESS(hdr, ZIO_COMPRESS_OFF);
	to_write = cabd;
	abd_copy(to_write, hdr->b_l1hdr.b_pabd, psize);
	if (psize != asize)
	abd_zero_off(to_write, psize, asize - psize);
	goto encrypt;
	}
	ASSERT3U(psize, <=, HDR_GET_PSIZE(hdr));
	if (psize < asize)
	memset((char *)tmp + psize, 0, asize - psize);
	psize = HDR_GET_PSIZE(hdr);
	abd_return_buf_copy(cabd, tmp, size);
	to_write = cabd;
	}

	encrypt:
	if (HDR_ENCRYPTED(hdr)) {
	eabd = abd_alloc_for_io(asize, ismd);

	/*
	* If the dataset was disowned before the buffer
	* made it to this point, the key to re-encrypt
	* it won't be available. In this case we simply
	* won't write the buffer to the L2ARC.
	*/
	ret = spa_keystore_lookup_key(spa, hdr->b_crypt_hdr.b_dsobj,
	FTAG, &dck);
	if (ret != 0)
	goto error;

	ret = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
	hdr->b_crypt_hdr.b_ot, bswap, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv, mac, psize, to_write, eabd,
	&no_crypt);
	if (ret != 0)
	goto error;

	if (no_crypt)
	abd_copy(eabd, to_write, psize);

	if (psize != asize)
	abd_zero_off(eabd, psize, asize - psize);

	/* assert that the MAC we got here matches the one we saved */
	ASSERT0(memcmp(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN));
	spa_keystore_dsl_key_rele(spa, dck, FTAG);

	if (to_write == cabd)
	abd_free(cabd);

	to_write = eabd;
	}

	out:
	ASSERT3P(to_write, !=, hdr->b_l1hdr.b_pabd);
	*abd_out = to_write;
	return (0);

	error:
	if (dck != NULL)
	spa_keystore_dsl_key_rele(spa, dck, FTAG);
	if (cabd != NULL)
	abd_free(cabd);
	if (eabd != NULL)
	abd_free(eabd);

	*abd_out = NULL;
	return (ret);
	}

	static void
	l2arc_blk_fetch_done(zio_t *zio)
	{
	l2arc_read_callback_t *cb;

	cb = zio->io_private;
	if (cb->l2rcb_abd != NULL)
	abd_free(cb->l2rcb_abd);
	kmem_free(cb, sizeof (l2arc_read_callback_t));
	}

	/*
	* Find and write ARC buffers to the L2ARC device.
	*
	* An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid
	* for reading until they have completed writing.
	* The headroom_boost is an in-out parameter used to maintain headroom boost
	* state between calls to this function.
	*
	* Returns the number of bytes actually written (which may be smaller than
	* the delta by which the device hand has changed due to alignment and the
	* writing of log blocks).
	*/
	static uint64_t
	l2arc_write_buffers(spa_t spa, l2arc_dev_t dev, uint64_t target_sz)
	{
	arc_buf_hdr_t hdr, hdr_prev, *head;
	uint64_t write_asize, write_psize, write_lsize, headroom;
	boolean_t full;
	l2arc_write_callback_t *cb = NULL;
	zio_t pio, wzio;
	uint64_t guid = spa_load_guid(spa);
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;

	ASSERT3P(dev->l2ad_vdev, !=, NULL);

	pio = NULL;
	write_lsize = write_asize = write_psize = 0;
	full = B_FALSE;
	head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE);
	arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD \| ARC_FLAG_HAS_L2HDR);

	/*
	* Copy buffers for L2ARC writing.
	*/
	for (int pass = 0; pass < L2ARC_FEED_TYPES; pass++) {
	/*
	* If pass == 1 or 3, we cache MRU metadata and data
	* respectively.
	*/
	if (l2arc_mfuonly) {
	if (pass == 1 \|\| pass == 3)
	continue;
	}

	multilist_sublist_t *mls = l2arc_sublist_lock(pass);
	uint64_t passed_sz = 0;

	VERIFY3P(mls, !=, NULL);

	/*
	* L2ARC fast warmup.
	*
	* Until the ARC is warm and starts to evict, read from the
	* head of the ARC lists rather than the tail.
	*/
	if (arc_warm == B_FALSE)
	hdr = multilist_sublist_head(mls);
	else
	hdr = multilist_sublist_tail(mls);

	headroom = target_sz * l2arc_headroom;
	if (zfs_compressed_arc_enabled)
	headroom = (headroom * l2arc_headroom_boost) / 100;

	for (; hdr; hdr = hdr_prev) {
	kmutex_t *hash_lock;
	abd_t *to_write = NULL;

	if (arc_warm == B_FALSE)
	hdr_prev = multilist_sublist_next(mls, hdr);
	else
	hdr_prev = multilist_sublist_prev(mls, hdr);

	hash_lock = HDR_LOCK(hdr);
	if (!mutex_tryenter(hash_lock)) {
	/*
	* Skip this buffer rather than waiting.
	*/
	continue;
	}

	passed_sz += HDR_GET_LSIZE(hdr);
	if (l2arc_headroom != 0 && passed_sz > headroom) {
	/*
	* Searched too far.
	*/
	mutex_exit(hash_lock);
	break;
	}

	if (!l2arc_write_eligible(guid, hdr)) {
	mutex_exit(hash_lock);
	continue;
	}

	ASSERT(HDR_HAS_L1HDR(hdr));

	ASSERT3U(HDR_GET_PSIZE(hdr), >, 0);
	ASSERT3U(arc_hdr_size(hdr), >, 0);
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\|
	HDR_HAS_RABD(hdr));
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev,
	psize);

	if ((write_asize + asize) > target_sz) {
	full = B_TRUE;
	mutex_exit(hash_lock);
	break;
	}

	/*
	* We rely on the L1 portion of the header below, so
	* it's invalid for this header to have been evicted out
	* of the ghost cache, prior to being written out. The
	* ARC_FLAG_L2_WRITING bit ensures this won't happen.
	*/
	arc_hdr_set_flags(hdr, ARC_FLAG_L2_WRITING);

	/*
	* If this header has b_rabd, we can use this since it
	* must always match the data exactly as it exists on
	* disk. Otherwise, the L2ARC can normally use the
	* hdr's data, but if we're sharing data between the
	* hdr and one of its bufs, L2ARC needs its own copy of
	* the data so that the ZIO below can't race with the
	* buf consumer. To ensure that this copy will be
	* available for the lifetime of the ZIO and be cleaned
	* up afterwards, we add it to the l2arc_free_on_write
	* queue. If we need to apply any transforms to the
	* data (compression, encryption) we will also need the
	* extra buffer.
	*/
	if (HDR_HAS_RABD(hdr) && psize == asize) {
	to_write = hdr->b_crypt_hdr.b_rabd;
	} else if ((HDR_COMPRESSION_ENABLED(hdr) \|\|
	HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) &&
	!HDR_ENCRYPTED(hdr) && !HDR_SHARED_DATA(hdr) &&
	psize == asize) {
	to_write = hdr->b_l1hdr.b_pabd;
	} else {
	int ret;
	arc_buf_contents_t type = arc_buf_type(hdr);

	ret = l2arc_apply_transforms(spa, hdr, asize,
	&to_write);
	if (ret != 0) {
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_L2_WRITING);
	mutex_exit(hash_lock);
	continue;
	}

	l2arc_free_abd_on_write(to_write, asize, type);
	}

	if (pio == NULL) {
	/*
	* Insert a dummy header on the buflist so
	* l2arc_write_done() can find where the
	* write buffers begin without searching.
	*/
	mutex_enter(&dev->l2ad_mtx);
	list_insert_head(&dev->l2ad_buflist, head);
	mutex_exit(&dev->l2ad_mtx);

	cb = kmem_alloc(
	sizeof (l2arc_write_callback_t), KM_SLEEP);
	cb->l2wcb_dev = dev;
	cb->l2wcb_head = head;
	/*
	* Create a list to save allocated abd buffers
	* for l2arc_log_blk_commit().
	*/
	list_create(&cb->l2wcb_abd_list,
	sizeof (l2arc_lb_abd_buf_t),
	offsetof(l2arc_lb_abd_buf_t, node));
	pio = zio_root(spa, l2arc_write_done, cb,
	ZIO_FLAG_CANFAIL);
	}

	hdr->b_l2hdr.b_dev = dev;
	hdr->b_l2hdr.b_hits = 0;

	hdr->b_l2hdr.b_daddr = dev->l2ad_hand;
	hdr->b_l2hdr.b_arcs_state =
	hdr->b_l1hdr.b_state->arcs_state;
	arc_hdr_set_flags(hdr, ARC_FLAG_HAS_L2HDR);

	mutex_enter(&dev->l2ad_mtx);
	list_insert_head(&dev->l2ad_buflist, hdr);
	mutex_exit(&dev->l2ad_mtx);

	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);

	wzio = zio_write_phys(pio, dev->l2ad_vdev,
	hdr->b_l2hdr.b_daddr, asize, to_write,
	ZIO_CHECKSUM_OFF, NULL, hdr,
	ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_CANFAIL, B_FALSE);

	write_lsize += HDR_GET_LSIZE(hdr);
	DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev,
	zio_t *, wzio);

	write_psize += psize;
	write_asize += asize;
	dev->l2ad_hand += asize;
	l2arc_hdr_arcstats_increment(hdr);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	mutex_exit(hash_lock);

	/*
	* Append buf info to current log and commit if full.
	* arcstat_l2_{size,asize} kstats are updated
	* internally.
	*/
	if (l2arc_log_blk_insert(dev, hdr))
	l2arc_log_blk_commit(dev, pio, cb);

	zio_nowait(wzio);
	}

	multilist_sublist_unlock(mls);

	if (full == B_TRUE)
	break;
	}

	/* No buffers selected for writing? */
	if (pio == NULL) {
	ASSERT0(write_lsize);
	ASSERT(!HDR_HAS_L1HDR(head));
	kmem_cache_free(hdr_l2only_cache, head);

	/*
	* Although we did not write any buffers l2ad_evict may
	* have advanced.
	*/
	if (dev->l2ad_evict != l2dhdr->dh_evict)
	l2arc_dev_hdr_update(dev);

	return (0);
	}

	if (!dev->l2ad_first)
	ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict);

	ASSERT3U(write_asize, <=, target_sz);
	ARCSTAT_BUMP(arcstat_l2_writes_sent);
	ARCSTAT_INCR(arcstat_l2_write_bytes, write_psize);

	dev->l2ad_writing = B_TRUE;
	(void) zio_wait(pio);
	dev->l2ad_writing = B_FALSE;

	/*
	* Update the device header after the zio completes as
	* l2arc_write_done() may have updated the memory holding the log block
	* pointers in the device header.
	*/
	l2arc_dev_hdr_update(dev);

	return (write_asize);
	}

	static boolean_t
	l2arc_hdr_limit_reached(void)
	{
	int64_t s = aggsum_upper_bound(&arc_sums.arcstat_l2_hdr_size);

	return (arc_reclaim_needed() \|\| (s > arc_meta_limit * 3 / 4) \|\|
	(s > (arc_warm ? arc_c : arc_c_max) * l2arc_meta_percent / 100));
	}

	/*
	* This thread feeds the L2ARC at regular intervals. This is the beating
	* heart of the L2ARC.
	*/
	static __attribute__((noreturn)) void
	l2arc_feed_thread(void *unused)
	{
	(void) unused;
	callb_cpr_t cpr;
	l2arc_dev_t *dev;
	spa_t *spa;
	uint64_t size, wrote;
	clock_t begin, next = ddi_get_lbolt();
	fstrans_cookie_t cookie;

	CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG);

	mutex_enter(&l2arc_feed_thr_lock);

	cookie = spl_fstrans_mark();
	while (l2arc_thread_exit == 0) {
	CALLB_CPR_SAFE_BEGIN(&cpr);
	(void) cv_timedwait_idle(&l2arc_feed_thr_cv,
	&l2arc_feed_thr_lock, next);
	CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock);
	next = ddi_get_lbolt() + hz;

	/*
	* Quick check for L2ARC devices.
	*/
	mutex_enter(&l2arc_dev_mtx);
	if (l2arc_ndev == 0) {
	mutex_exit(&l2arc_dev_mtx);
	continue;
	}
	mutex_exit(&l2arc_dev_mtx);
	begin = ddi_get_lbolt();

	/*
	* This selects the next l2arc device to write to, and in
	* doing so the next spa to feed from: dev->l2ad_spa. This
	* will return NULL if there are now no l2arc devices or if
	* they are all faulted.
	*
	* If a device is returned, its spa's config lock is also
	* held to prevent device removal. l2arc_dev_get_next()
	* will grab and release l2arc_dev_mtx.
	*/
	if ((dev = l2arc_dev_get_next()) == NULL)
	continue;

	spa = dev->l2ad_spa;
	ASSERT3P(spa, !=, NULL);

	/*
	* If the pool is read-only then force the feed thread to
	* sleep a little longer.
	*/
	if (!spa_writeable(spa)) {
	next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz;
	spa_config_exit(spa, SCL_L2ARC, dev);
	continue;
	}

	/*
	* Avoid contributing to memory pressure.
	*/
	if (l2arc_hdr_limit_reached()) {
	ARCSTAT_BUMP(arcstat_l2_abort_lowmem);
	spa_config_exit(spa, SCL_L2ARC, dev);
	continue;
	}

	ARCSTAT_BUMP(arcstat_l2_feeds);

	size = l2arc_write_size(dev);

	/*
	* Evict L2ARC buffers that will be overwritten.
	*/
	l2arc_evict(dev, size, B_FALSE);

	/*
	* Write ARC buffers.
	*/
	wrote = l2arc_write_buffers(spa, dev, size);

	/*
	* Calculate interval between writes.
	*/
	next = l2arc_write_interval(begin, size, wrote);
	spa_config_exit(spa, SCL_L2ARC, dev);
	}
	spl_fstrans_unmark(cookie);

	l2arc_thread_exit = 0;
	cv_broadcast(&l2arc_feed_thr_cv);
	CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */
	thread_exit();
	}

	boolean_t
	l2arc_vdev_present(vdev_t *vd)
	{
	return (l2arc_vdev_get(vd) != NULL);
	}

	/*
	* Returns the l2arc_dev_t associated with a particular vdev_t or NULL if
	* the vdev_t isn't an L2ARC device.
	*/
	l2arc_dev_t *
	l2arc_vdev_get(vdev_t *vd)
	{
	l2arc_dev_t *dev;

	mutex_enter(&l2arc_dev_mtx);
	for (dev = list_head(l2arc_dev_list); dev != NULL;
	dev = list_next(l2arc_dev_list, dev)) {
	if (dev->l2ad_vdev == vd)
	break;
	}
	mutex_exit(&l2arc_dev_mtx);

	return (dev);
	}

	static void
	l2arc_rebuild_dev(l2arc_dev_t *dev, boolean_t reopen)
	{
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	spa_t *spa = dev->l2ad_spa;

	/*
	* The L2ARC has to hold at least the payload of one log block for
	* them to be restored (persistent L2ARC). The payload of a log block
	* depends on the amount of its log entries. We always write log blocks
	* with 1022 entries. How many of them are committed or restored depends
	* on the size of the L2ARC device. Thus the maximum payload of
	* one log block is 1022 * SPA_MAXBLOCKSIZE = 16GB. If the L2ARC device
	* is less than that, we reduce the amount of committed and restored
	* log entries per block so as to enable persistence.
	*/
	if (dev->l2ad_end < l2arc_rebuild_blocks_min_l2size) {
	dev->l2ad_log_entries = 0;
	} else {
	dev->l2ad_log_entries = MIN((dev->l2ad_end -
	dev->l2ad_start) >> SPA_MAXBLOCKSHIFT,
	L2ARC_LOG_BLK_MAX_ENTRIES);
	}

	/*
	* Read the device header, if an error is returned do not rebuild L2ARC.
	*/
	if (l2arc_dev_hdr_read(dev) == 0 && dev->l2ad_log_entries > 0) {
	/*
	* If we are onlining a cache device (vdev_reopen) that was
	* still present (l2arc_vdev_present()) and rebuild is enabled,
	* we should evict all ARC buffers and pointers to log blocks
	* and reclaim their space before restoring its contents to
	* L2ARC.
	*/
	if (reopen) {
	if (!l2arc_rebuild_enabled) {
	return;
	} else {
	l2arc_evict(dev, 0, B_TRUE);
	/* start a new log block */
	dev->l2ad_log_ent_idx = 0;
	dev->l2ad_log_blk_payload_asize = 0;
	dev->l2ad_log_blk_payload_start = 0;
	}
	}
	/*
	* Just mark the device as pending for a rebuild. We won't
	* be starting a rebuild in line here as it would block pool
	* import. Instead spa_load_impl will hand that off to an
	* async task which will call l2arc_spa_rebuild_start.
	*/
	dev->l2ad_rebuild = B_TRUE;
	} else if (spa_writeable(spa)) {
	/*
	* In this case TRIM the whole device if l2arc_trim_ahead > 0,
	* otherwise create a new header. We zero out the memory holding
	* the header to reset dh_start_lbps. If we TRIM the whole
	* device the new header will be written by
	* vdev_trim_l2arc_thread() at the end of the TRIM to update the
	* trim_state in the header too. When reading the header, if
	* trim_state is not VDEV_TRIM_COMPLETE and l2arc_trim_ahead > 0
	* we opt to TRIM the whole device again.
	*/
	if (l2arc_trim_ahead > 0) {
	dev->l2ad_trim_all = B_TRUE;
	} else {
	memset(l2dhdr, 0, l2dhdr_asize);
	l2arc_dev_hdr_update(dev);
	}
	}
	}

	/*
	* Add a vdev for use by the L2ARC. By this point the spa has already
	* validated the vdev and opened it.
	*/
	void
	l2arc_add_vdev(spa_t spa, vdev_t vd)
	{
	l2arc_dev_t *adddev;
	uint64_t l2dhdr_asize;

	ASSERT(!l2arc_vdev_present(vd));

	/*
	* Create a new l2arc device entry.
	*/
	adddev = vmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP);
	adddev->l2ad_spa = spa;
	adddev->l2ad_vdev = vd;
	/* leave extra size for an l2arc device header */
	l2dhdr_asize = adddev->l2ad_dev_hdr_asize =
	MAX(sizeof (*adddev->l2ad_dev_hdr), 1 << vd->vdev_ashift);
	adddev->l2ad_start = VDEV_LABEL_START_SIZE + l2dhdr_asize;
	adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd);
	ASSERT3U(adddev->l2ad_start, <, adddev->l2ad_end);
	adddev->l2ad_hand = adddev->l2ad_start;
	adddev->l2ad_evict = adddev->l2ad_start;
	adddev->l2ad_first = B_TRUE;
	adddev->l2ad_writing = B_FALSE;
	adddev->l2ad_trim_all = B_FALSE;
	list_link_init(&adddev->l2ad_node);
	adddev->l2ad_dev_hdr = kmem_zalloc(l2dhdr_asize, KM_SLEEP);

	mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL);
	/*
	* This is a list of all ARC buffers that are still valid on the
	* device.
	*/
	list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t),
	offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node));

	/*
	* This is a list of pointers to log blocks that are still present
	* on the device.
	*/
	list_create(&adddev->l2ad_lbptr_list, sizeof (l2arc_lb_ptr_buf_t),
	offsetof(l2arc_lb_ptr_buf_t, node));

	vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand);
	zfs_refcount_create(&adddev->l2ad_alloc);
	zfs_refcount_create(&adddev->l2ad_lb_asize);
	zfs_refcount_create(&adddev->l2ad_lb_count);

	/*
	* Decide if dev is eligible for L2ARC rebuild or whole device
	* trimming. This has to happen before the device is added in the
	* cache device list and l2arc_dev_mtx is released. Otherwise
	* l2arc_feed_thread() might already start writing on the
	* device.
	*/
	l2arc_rebuild_dev(adddev, B_FALSE);

	/*
	* Add device to global list
	*/
	mutex_enter(&l2arc_dev_mtx);
	list_insert_head(l2arc_dev_list, adddev);
	atomic_inc_64(&l2arc_ndev);
	mutex_exit(&l2arc_dev_mtx);
	}

	/*
	* Decide if a vdev is eligible for L2ARC rebuild, called from vdev_reopen()
	* in case of onlining a cache device.
	*/
	void
	l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen)
	{
	l2arc_dev_t *dev = NULL;

	dev = l2arc_vdev_get(vd);
	ASSERT3P(dev, !=, NULL);

	/*
	* In contrast to l2arc_add_vdev() we do not have to worry about
	* l2arc_feed_thread() invalidating previous content when onlining a
	* cache device. The device parameters (l2ad*) are not cleared when
	* offlining the device and writing new buffers will not invalidate
	* all previous content. In worst case only buffers that have not had
	* their log block written to the device will be lost.
	* When onlining the cache device (ie offline->online without exporting
	* the pool in between) this happens:
	* vdev_reopen() -> vdev_open() -> l2arc_rebuild_vdev()
	* \| \|
	* vdev_is_dead() = B_FALSE l2ad_rebuild = B_TRUE
	* During the time where vdev_is_dead = B_FALSE and until l2ad_rebuild
	* is set to B_TRUE we might write additional buffers to the device.
	*/
	l2arc_rebuild_dev(dev, reopen);
	}

	/*
	* Remove a vdev from the L2ARC.
	*/
	void
	l2arc_remove_vdev(vdev_t *vd)
	{
	l2arc_dev_t *remdev = NULL;

	/*
	* Find the device by vdev
	*/
	remdev = l2arc_vdev_get(vd);
	ASSERT3P(remdev, !=, NULL);

	/*
	* Cancel any ongoing or scheduled rebuild.
	*/
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (remdev->l2ad_rebuild_began == B_TRUE) {
	remdev->l2ad_rebuild_cancel = B_TRUE;
	while (remdev->l2ad_rebuild == B_TRUE)
	cv_wait(&l2arc_rebuild_thr_cv, &l2arc_rebuild_thr_lock);
	}
	mutex_exit(&l2arc_rebuild_thr_lock);

	/*
	* Remove device from global list
	*/
	mutex_enter(&l2arc_dev_mtx);
	list_remove(l2arc_dev_list, remdev);
	l2arc_dev_last = NULL; /* may have been invalidated */
	atomic_dec_64(&l2arc_ndev);
	mutex_exit(&l2arc_dev_mtx);

	/*
	* Clear all buflists and ARC references. L2ARC device flush.
	*/
	l2arc_evict(remdev, 0, B_TRUE);
	list_destroy(&remdev->l2ad_buflist);
	ASSERT(list_is_empty(&remdev->l2ad_lbptr_list));
	list_destroy(&remdev->l2ad_lbptr_list);
	mutex_destroy(&remdev->l2ad_mtx);
	zfs_refcount_destroy(&remdev->l2ad_alloc);
	zfs_refcount_destroy(&remdev->l2ad_lb_asize);
	zfs_refcount_destroy(&remdev->l2ad_lb_count);
	kmem_free(remdev->l2ad_dev_hdr, remdev->l2ad_dev_hdr_asize);
	vmem_free(remdev, sizeof (l2arc_dev_t));
	}

	void
	l2arc_init(void)
	{
	l2arc_thread_exit = 0;
	l2arc_ndev = 0;

	mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&l2arc_rebuild_thr_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&l2arc_rebuild_thr_cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL);

	l2arc_dev_list = &L2ARC_dev_list;
	l2arc_free_on_write = &L2ARC_free_on_write;
	list_create(l2arc_dev_list, sizeof (l2arc_dev_t),
	offsetof(l2arc_dev_t, l2ad_node));
	list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t),
	offsetof(l2arc_data_free_t, l2df_list_node));
	}

	void
	l2arc_fini(void)
	{
	mutex_destroy(&l2arc_feed_thr_lock);
	cv_destroy(&l2arc_feed_thr_cv);
	mutex_destroy(&l2arc_rebuild_thr_lock);
	cv_destroy(&l2arc_rebuild_thr_cv);
	mutex_destroy(&l2arc_dev_mtx);
	mutex_destroy(&l2arc_free_on_write_mtx);

	list_destroy(l2arc_dev_list);
	list_destroy(l2arc_free_on_write);
	}

	void
	l2arc_start(void)
	{
	if (!(spa_mode_global & SPA_MODE_WRITE))
	return;

	(void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0,
	TS_RUN, defclsyspri);
	}

	void
	l2arc_stop(void)
	{
	if (!(spa_mode_global & SPA_MODE_WRITE))
	return;

	mutex_enter(&l2arc_feed_thr_lock);
	cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */
	l2arc_thread_exit = 1;
	while (l2arc_thread_exit != 0)
	cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock);
	mutex_exit(&l2arc_feed_thr_lock);
	}

	/*
	* Punches out rebuild threads for the L2ARC devices in a spa. This should
	* be called after pool import from the spa async thread, since starting
	* these threads directly from spa_import() will make them part of the
	* "zpool import" context and delay process exit (and thus pool import).
	*/
	void
	l2arc_spa_rebuild_start(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	/*
	* Locate the spa's l2arc devices and kick off rebuild threads.
	*/
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	l2arc_dev_t *dev =
	l2arc_vdev_get(spa->spa_l2cache.sav_vdevs[i]);
	if (dev == NULL) {
	/* Don't attempt a rebuild if the vdev is UNAVAIL */
	continue;
	}
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (dev->l2ad_rebuild && !dev->l2ad_rebuild_cancel) {
	dev->l2ad_rebuild_began = B_TRUE;
	(void) thread_create(NULL, 0, l2arc_dev_rebuild_thread,
	dev, 0, &p0, TS_RUN, minclsyspri);
	}
	mutex_exit(&l2arc_rebuild_thr_lock);
	}
	}

	/*
	* Main entry point for L2ARC rebuilding.
	*/
	static __attribute__((noreturn)) void
	l2arc_dev_rebuild_thread(void *arg)
	{
	l2arc_dev_t *dev = arg;

	VERIFY(!dev->l2ad_rebuild_cancel);
	VERIFY(dev->l2ad_rebuild);
	(void) l2arc_rebuild(dev);
	mutex_enter(&l2arc_rebuild_thr_lock);
	dev->l2ad_rebuild_began = B_FALSE;
	dev->l2ad_rebuild = B_FALSE;
	mutex_exit(&l2arc_rebuild_thr_lock);

	thread_exit();
	}

	/*
	* This function implements the actual L2ARC metadata rebuild. It:
	* starts reading the log block chain and restores each block's contents
	* to memory (reconstructing arc_buf_hdr_t's).
	*
	* Operation stops under any of the following conditions:
	*
	* 1) We reach the end of the log block chain.
	* 2) We encounter any error condition (cksum errors, io errors)
	*/
	static int
	l2arc_rebuild(l2arc_dev_t *dev)
	{
	vdev_t *vd = dev->l2ad_vdev;
	spa_t *spa = vd->vdev_spa;
	int err = 0;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	l2arc_log_blk_phys_t this_lb, next_lb;
	zio_t this_io = NULL, next_io = NULL;
	l2arc_log_blkptr_t lbps[2];
	l2arc_lb_ptr_buf_t *lb_ptr_buf;
	boolean_t lock_held;

	this_lb = vmem_zalloc(sizeof (*this_lb), KM_SLEEP);
	next_lb = vmem_zalloc(sizeof (*next_lb), KM_SLEEP);

	/*
	* We prevent device removal while issuing reads to the device,
	* then during the rebuilding phases we drop this lock again so
	* that a spa_unload or device remove can be initiated - this is
	* safe, because the spa will signal us to stop before removing
	* our device and wait for us to stop.
	*/
	spa_config_enter(spa, SCL_L2ARC, vd, RW_READER);
	lock_held = B_TRUE;

	/*
	* Retrieve the persistent L2ARC device state.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	dev->l2ad_evict = MAX(l2dhdr->dh_evict, dev->l2ad_start);
	dev->l2ad_hand = MAX(l2dhdr->dh_start_lbps[0].lbp_daddr +
	L2BLK_GET_PSIZE((&l2dhdr->dh_start_lbps[0])->lbp_prop),
	dev->l2ad_start);
	dev->l2ad_first = !!(l2dhdr->dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);

	vd->vdev_trim_action_time = l2dhdr->dh_trim_action_time;
	vd->vdev_trim_state = l2dhdr->dh_trim_state;

	/*
	* In case the zfs module parameter l2arc_rebuild_enabled is false
	* we do not start the rebuild process.
	*/
	if (!l2arc_rebuild_enabled)
	goto out;

	/* Prepare the rebuild process */
	memcpy(lbps, l2dhdr->dh_start_lbps, sizeof (lbps));

	/* Start the rebuild process */
	for (;;) {
	if (!l2arc_log_blkptr_valid(dev, &lbps[0]))
	break;

	if ((err = l2arc_log_blk_read(dev, &lbps[0], &lbps[1],
	this_lb, next_lb, this_io, &next_io)) != 0)
	goto out;

	/*
	* Our memory pressure valve. If the system is running low
	* on memory, rather than swamping memory with new ARC buf
	* hdrs, we opt not to rebuild the L2ARC. At this point,
	* however, we have already set up our L2ARC dev to chain in
	* new metadata log blocks, so the user may choose to offline/
	* online the L2ARC dev at a later time (or re-import the pool)
	* to reconstruct it (when there's less memory pressure).
	*/
	if (l2arc_hdr_limit_reached()) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_lowmem);
	cmn_err(CE_NOTE, "System running low on memory, "
	"aborting L2ARC rebuild.");
	err = SET_ERROR(ENOMEM);
	goto out;
	}

	spa_config_exit(spa, SCL_L2ARC, vd);
	lock_held = B_FALSE;

	/*
	* Now that we know that the next_lb checks out alright, we
	* can start reconstruction from this log block.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	uint64_t asize = L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
	l2arc_log_blk_restore(dev, this_lb, asize);

	/*
	* log block restored, include its pointer in the list of
	* pointers to log blocks present in the L2ARC device.
	*/
	lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
	lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t),
	KM_SLEEP);
	memcpy(lb_ptr_buf->lb_ptr, &lbps[0],
	sizeof (l2arc_log_blkptr_t));
	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_lbptr_list, lb_ptr_buf);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_count);
	zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
	zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
	mutex_exit(&dev->l2ad_mtx);
	vdev_space_update(vd, asize, 0, 0);

	/*
	* Protection against loops of log blocks:
	*
	* l2ad_hand l2ad_evict
	* V V
	* l2ad_start \|=======================================\| l2ad_end
	* -----\|\|\|----\|\|\|---\|\|\|----\|\|\|
	* (3) (2) (1) (0)
	* ---\|\|\|---\|\|\|----\|\|\|---\|\|\|
	* (7) (6) (5) (4)
	*
	* In this situation the pointer of log block (4) passes
	* l2arc_log_blkptr_valid() but the log block should not be
	* restored as it is overwritten by the payload of log block
	* (0). Only log blocks (0)-(3) should be restored. We check
	* whether l2ad_evict lies in between the payload starting
	* offset of the next log block (lbps[1].lbp_payload_start)
	* and the payload starting offset of the present log block
	* (lbps[0].lbp_payload_start). If true and this isn't the
	* first pass, we are looping from the beginning and we should
	* stop.
	*/
	if (l2arc_range_check_overlap(lbps[1].lbp_payload_start,
	lbps[0].lbp_payload_start, dev->l2ad_evict) &&
	!dev->l2ad_first)
	goto out;

	kpreempt(KPREEMPT_SYNC);
	for (;;) {
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (dev->l2ad_rebuild_cancel) {
	dev->l2ad_rebuild = B_FALSE;
	cv_signal(&l2arc_rebuild_thr_cv);
	mutex_exit(&l2arc_rebuild_thr_lock);
	err = SET_ERROR(ECANCELED);
	goto out;
	}
	mutex_exit(&l2arc_rebuild_thr_lock);
	if (spa_config_tryenter(spa, SCL_L2ARC, vd,
	RW_READER)) {
	lock_held = B_TRUE;
	break;
	}
	/*
	* L2ARC config lock held by somebody in writer,
	* possibly due to them trying to remove us. They'll
	* likely to want us to shut down, so after a little
	* delay, we check l2ad_rebuild_cancel and retry
	* the lock again.
	*/
	delay(1);
	}

	/*
	* Continue with the next log block.
	*/
	lbps[0] = lbps[1];
	lbps[1] = this_lb->lb_prev_lbp;
	PTR_SWAP(this_lb, next_lb);
	this_io = next_io;
	next_io = NULL;
	}

	if (this_io != NULL)
	l2arc_log_blk_fetch_abort(this_io);
	out:
	if (next_io != NULL)
	l2arc_log_blk_fetch_abort(next_io);
	vmem_free(this_lb, sizeof (*this_lb));
	vmem_free(next_lb, sizeof (*next_lb));

	if (!l2arc_rebuild_enabled) {
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"disabled");
	} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) > 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_success);
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"successful, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) == 0) {
	/*
	* No error but also nothing restored, meaning the lbps array
	* in the device header points to invalid/non-present log
	* blocks. Reset the header.
	*/
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"no valid log blocks");
	memset(l2dhdr, 0, dev->l2ad_dev_hdr_asize);
	l2arc_dev_hdr_update(dev);
	} else if (err == ECANCELED) {
	/*
	* In case the rebuild was canceled do not log to spa history
	* log as the pool may be in the process of being removed.
	*/
	zfs_dbgmsg("L2ARC rebuild aborted, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	} else if (err != 0) {
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"aborted, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	}

	if (lock_held)
	spa_config_exit(spa, SCL_L2ARC, vd);

	return (err);
	}

	/*
	* Attempts to read the device header on the provided L2ARC device and writes
	* it to `hdr'. On success, this function returns 0, otherwise the appropriate
	* error code is returned.
	*/
	static int
	l2arc_dev_hdr_read(l2arc_dev_t *dev)
	{
	int err;
	uint64_t guid;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	abd_t *abd;

	guid = spa_guid(dev->l2ad_vdev->vdev_spa);

	abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);

	err = zio_wait(zio_read_phys(NULL, dev->l2ad_vdev,
	VDEV_LABEL_START_SIZE, l2dhdr_asize, abd,
	ZIO_CHECKSUM_LABEL, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_SPECULATIVE, B_FALSE));

	abd_free(abd);

	if (err != 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_dh_errors);
	zfs_dbgmsg("L2ARC IO error (%d) while reading device header, "
	"vdev guid: %llu", err,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	return (err);
	}

	if (l2dhdr->dh_magic == BSWAP_64(L2ARC_DEV_HDR_MAGIC))
	byteswap_uint64_array(l2dhdr, sizeof (*l2dhdr));

	if (l2dhdr->dh_magic != L2ARC_DEV_HDR_MAGIC \|\|
	l2dhdr->dh_spa_guid != guid \|\|
	l2dhdr->dh_vdev_guid != dev->l2ad_vdev->vdev_guid \|\|
	l2dhdr->dh_version != L2ARC_PERSISTENT_VERSION \|\|
	l2dhdr->dh_log_entries != dev->l2ad_log_entries \|\|
	l2dhdr->dh_end != dev->l2ad_end \|\|
	!l2arc_range_check_overlap(dev->l2ad_start, dev->l2ad_end,
	l2dhdr->dh_evict) \|\|
	(l2dhdr->dh_trim_state != VDEV_TRIM_COMPLETE &&
	l2arc_trim_ahead > 0)) {
	/*
	* Attempt to rebuild a device containing no actual dev hdr
	* or containing a header from some other pool or from another
	* version of persistent L2ARC.
	*/
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_unsupported);
	return (SET_ERROR(ENOTSUP));
	}

	return (0);
	}

	/*
	* Reads L2ARC log blocks from storage and validates their contents.
	*
	* This function implements a simple fetcher to make sure that while
	* we're processing one buffer the L2ARC is already fetching the next
	* one in the chain.
	*
	* The arguments this_lp and next_lp point to the current and next log block
	* address in the block chain. Similarly, this_lb and next_lb hold the
	* l2arc_log_blk_phys_t's of the current and next L2ARC blk.
	*
	* The `this_io' and `next_io' arguments are used for block fetching.
	* When issuing the first blk IO during rebuild, you should pass NULL for
	* `this_io'. This function will then issue a sync IO to read the block and
	* also issue an async IO to fetch the next block in the block chain. The
	* fetched IO is returned in `next_io'. On subsequent calls to this
	* function, pass the value returned in `next_io' from the previous call
	* as `this_io' and a fresh `next_io' pointer to hold the next fetch IO.
	* Prior to the call, you should initialize your `next_io' pointer to be
	* NULL. If no fetch IO was issued, the pointer is left set at NULL.
	*
	* On success, this function returns 0, otherwise it returns an appropriate
	* error code. On error the fetching IO is aborted and cleared before
	* returning from this function. Therefore, if we return `success', the
	* caller can assume that we have taken care of cleanup of fetch IOs.
	*/
	static int
	l2arc_log_blk_read(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t this_lbp, const l2arc_log_blkptr_t next_lbp,
	l2arc_log_blk_phys_t this_lb, l2arc_log_blk_phys_t next_lb,
	zio_t this_io, zio_t *next_io)
	{
	int err = 0;
	zio_cksum_t cksum;
	abd_t *abd = NULL;
	uint64_t asize;

	ASSERT(this_lbp != NULL && next_lbp != NULL);
	ASSERT(this_lb != NULL && next_lb != NULL);
	ASSERT(next_io != NULL && *next_io == NULL);
	ASSERT(l2arc_log_blkptr_valid(dev, this_lbp));

	/*
	* Check to see if we have issued the IO for this log block in a
	* previous run. If not, this is the first call, so issue it now.
	*/
	if (this_io == NULL) {
	this_io = l2arc_log_blk_fetch(dev->l2ad_vdev, this_lbp,
	this_lb);
	}

	/*
	* Peek to see if we can start issuing the next IO immediately.
	*/
	if (l2arc_log_blkptr_valid(dev, next_lbp)) {
	/*
	* Start issuing IO for the next log block early - this
	* should help keep the L2ARC device busy while we
	* decompress and restore this log block.
	*/
	*next_io = l2arc_log_blk_fetch(dev->l2ad_vdev, next_lbp,
	next_lb);
	}

	/* Wait for the IO to read this log block to complete */
	if ((err = zio_wait(this_io)) != 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_io_errors);
	zfs_dbgmsg("L2ARC IO error (%d) while reading log block, "
	"offset: %llu, vdev guid: %llu", err,
	(u_longlong_t)this_lbp->lbp_daddr,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	goto cleanup;
	}

	/*
	* Make sure the buffer checks out.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	asize = L2BLK_GET_PSIZE((this_lbp)->lbp_prop);
	fletcher_4_native(this_lb, asize, NULL, &cksum);
	if (!ZIO_CHECKSUM_EQUAL(cksum, this_lbp->lbp_cksum)) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_cksum_lb_errors);
	zfs_dbgmsg("L2ARC log block cksum failed, offset: %llu, "
	"vdev guid: %llu, l2ad_hand: %llu, l2ad_evict: %llu",
	(u_longlong_t)this_lbp->lbp_daddr,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid,
	(u_longlong_t)dev->l2ad_hand,
	(u_longlong_t)dev->l2ad_evict);
	err = SET_ERROR(ECKSUM);
	goto cleanup;
	}

	/* Now we can take our time decoding this buffer */
	switch (L2BLK_GET_COMPRESS((this_lbp)->lbp_prop)) {
	case ZIO_COMPRESS_OFF:
	break;
	case ZIO_COMPRESS_LZ4:
	abd = abd_alloc_for_io(asize, B_TRUE);
	abd_copy_from_buf_off(abd, this_lb, 0, asize);
	if ((err = zio_decompress_data(
	L2BLK_GET_COMPRESS((this_lbp)->lbp_prop),
	abd, this_lb, asize, sizeof (*this_lb), NULL)) != 0) {
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	break;
	default:
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	if (this_lb->lb_magic == BSWAP_64(L2ARC_LOG_BLK_MAGIC))
	byteswap_uint64_array(this_lb, sizeof (*this_lb));
	if (this_lb->lb_magic != L2ARC_LOG_BLK_MAGIC) {
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	cleanup:
	/* Abort an in-flight fetch I/O in case of error */
	if (err != 0 && *next_io != NULL) {
	l2arc_log_blk_fetch_abort(*next_io);
	*next_io = NULL;
	}
	if (abd != NULL)
	abd_free(abd);
	return (err);
	}

	/*
	* Restores the payload of a log block to ARC. This creates empty ARC hdr
	* entries which only contain an l2arc hdr, essentially restoring the
	* buffers to their L2ARC evicted state. This function also updates space
	* usage on the L2ARC vdev to make sure it tracks restored buffers.
	*/
	static void
	l2arc_log_blk_restore(l2arc_dev_t dev, const l2arc_log_blk_phys_t lb,
	uint64_t lb_asize)
	{
	uint64_t size = 0, asize = 0;
	uint64_t log_entries = dev->l2ad_log_entries;

	/*
	* Usually arc_adapt() is called only for data, not headers, but
	* since we may allocate significant amount of memory here, let ARC
	* grow its arc_c.
	*/
	arc_adapt(log_entries * HDR_L2ONLY_SIZE, arc_l2c_only);

	for (int i = log_entries - 1; i >= 0; i--) {
	/*
	* Restore goes in the reverse temporal direction to preserve
	* correct temporal ordering of buffers in the l2ad_buflist.
	* l2arc_hdr_restore also does a list_insert_tail instead of
	* list_insert_head on the l2ad_buflist:
	*
	* LIST l2ad_buflist LIST
	* HEAD <------ (time) ------ TAIL
	* direction +-----+-----+-----+-----+-----+ direction
	* of l2arc <== \| buf \| buf \| buf \| buf \| buf \| ===> of rebuild
	* fill +-----+-----+-----+-----+-----+
	* ^ ^
	* \| \|
	* \| \|
	* l2arc_feed_thread l2arc_rebuild
	* will place new bufs here restores bufs here
	*
	* During l2arc_rebuild() the device is not used by
	* l2arc_feed_thread() as dev->l2ad_rebuild is set to true.
	*/
	size += L2BLK_GET_LSIZE((&lb->lb_entries[i])->le_prop);
	asize += vdev_psize_to_asize(dev->l2ad_vdev,
	L2BLK_GET_PSIZE((&lb->lb_entries[i])->le_prop));
	l2arc_hdr_restore(&lb->lb_entries[i], dev);
	}

	/*
	* Record rebuild stats:
	* size Logical size of restored buffers in the L2ARC
	* asize Aligned size of restored buffers in the L2ARC
	*/
	ARCSTAT_INCR(arcstat_l2_rebuild_size, size);
	ARCSTAT_INCR(arcstat_l2_rebuild_asize, asize);
	ARCSTAT_INCR(arcstat_l2_rebuild_bufs, log_entries);
	ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, lb_asize);
	ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio, asize / lb_asize);
	ARCSTAT_BUMP(arcstat_l2_rebuild_log_blks);
	}

	/*
	* Restores a single ARC buf hdr from a log entry. The ARC buffer is put
	* into a state indicating that it has been evicted to L2ARC.
	*/
	static void
	l2arc_hdr_restore(const l2arc_log_ent_phys_t le, l2arc_dev_t dev)
	{
	arc_buf_hdr_t hdr, exists;
	kmutex_t *hash_lock;
	arc_buf_contents_t type = L2BLK_GET_TYPE((le)->le_prop);
	uint64_t asize;

	/*
	* Do all the allocation before grabbing any locks, this lets us
	* sleep if memory is full and we don't have to deal with failed
	* allocations.
	*/
	hdr = arc_buf_alloc_l2only(L2BLK_GET_LSIZE((le)->le_prop), type,
	dev, le->le_dva, le->le_daddr,
	L2BLK_GET_PSIZE((le)->le_prop), le->le_birth,
	L2BLK_GET_COMPRESS((le)->le_prop), le->le_complevel,
	L2BLK_GET_PROTECTED((le)->le_prop),
	L2BLK_GET_PREFETCH((le)->le_prop),
	L2BLK_GET_STATE((le)->le_prop));
	asize = vdev_psize_to_asize(dev->l2ad_vdev,
	L2BLK_GET_PSIZE((le)->le_prop));

	/*
	* vdev_space_update() has to be called before arc_hdr_destroy() to
	* avoid underflow since the latter also calls vdev_space_update().
	*/
	l2arc_hdr_arcstats_increment(hdr);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_buflist, hdr);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr);
	mutex_exit(&dev->l2ad_mtx);

	exists = buf_hash_insert(hdr, &hash_lock);
	if (exists) {
	/* Buffer was already cached, no need to restore it. */
	arc_hdr_destroy(hdr);
	/*
	* If the buffer is already cached, check whether it has
	* L2ARC metadata. If not, enter them and update the flag.
	* This is important is case of onlining a cache device, since
	* we previously evicted all L2ARC metadata from ARC.
	*/
	if (!HDR_HAS_L2HDR(exists)) {
	arc_hdr_set_flags(exists, ARC_FLAG_HAS_L2HDR);
	exists->b_l2hdr.b_dev = dev;
	exists->b_l2hdr.b_daddr = le->le_daddr;
	exists->b_l2hdr.b_arcs_state =
	L2BLK_GET_STATE((le)->le_prop);
	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_buflist, exists);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(exists), exists);
	mutex_exit(&dev->l2ad_mtx);
	l2arc_hdr_arcstats_increment(exists);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
	}
	ARCSTAT_BUMP(arcstat_l2_rebuild_bufs_precached);
	}

	mutex_exit(hash_lock);
	}

	/*
	* Starts an asynchronous read IO to read a log block. This is used in log
	* block reconstruction to start reading the next block before we are done
	* decoding and reconstructing the current block, to keep the l2arc device
	* nice and hot with read IO to process.
	* The returned zio will contain a newly allocated memory buffers for the IO
	* data which should then be freed by the caller once the zio is no longer
	* needed (i.e. due to it having completed). If you wish to abort this
	* zio, you should do so using l2arc_log_blk_fetch_abort, which takes
	* care of disposing of the allocated buffers correctly.
	*/
	static zio_t *
	l2arc_log_blk_fetch(vdev_t vd, const l2arc_log_blkptr_t lbp,
	l2arc_log_blk_phys_t *lb)
	{
	uint32_t asize;
	zio_t *pio;
	l2arc_read_callback_t *cb;

	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
	ASSERT(asize <= sizeof (l2arc_log_blk_phys_t));

	cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP);
	cb->l2rcb_abd = abd_get_from_buf(lb, asize);
	pio = zio_root(vd->vdev_spa, l2arc_blk_fetch_done, cb,
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY);
	(void) zio_nowait(zio_read_phys(pio, vd, lbp->lbp_daddr, asize,
	cb->l2rcb_abd, ZIO_CHECKSUM_OFF, NULL, NULL,
	ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY, B_FALSE));

	return (pio);
	}

	/*
	* Aborts a zio returned from l2arc_log_blk_fetch and frees the data
	* buffers allocated for it.
	*/
	static void
	l2arc_log_blk_fetch_abort(zio_t *zio)
	{
	(void) zio_wait(zio);
	}

	/*
	* Creates a zio to update the device header on an l2arc device.
	*/
	void
	l2arc_dev_hdr_update(l2arc_dev_t *dev)
	{
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	abd_t *abd;
	int err;

	VERIFY(spa_config_held(dev->l2ad_spa, SCL_STATE_ALL, RW_READER));

	l2dhdr->dh_magic = L2ARC_DEV_HDR_MAGIC;
	l2dhdr->dh_version = L2ARC_PERSISTENT_VERSION;
	l2dhdr->dh_spa_guid = spa_guid(dev->l2ad_vdev->vdev_spa);
	l2dhdr->dh_vdev_guid = dev->l2ad_vdev->vdev_guid;
	l2dhdr->dh_log_entries = dev->l2ad_log_entries;
	l2dhdr->dh_evict = dev->l2ad_evict;
	l2dhdr->dh_start = dev->l2ad_start;
	l2dhdr->dh_end = dev->l2ad_end;
	l2dhdr->dh_lb_asize = zfs_refcount_count(&dev->l2ad_lb_asize);
	l2dhdr->dh_lb_count = zfs_refcount_count(&dev->l2ad_lb_count);
	l2dhdr->dh_flags = 0;
	l2dhdr->dh_trim_action_time = dev->l2ad_vdev->vdev_trim_action_time;
	l2dhdr->dh_trim_state = dev->l2ad_vdev->vdev_trim_state;
	if (dev->l2ad_first)
	l2dhdr->dh_flags \|= L2ARC_DEV_HDR_EVICT_FIRST;

	abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);

	err = zio_wait(zio_write_phys(NULL, dev->l2ad_vdev,
	VDEV_LABEL_START_SIZE, l2dhdr_asize, abd, ZIO_CHECKSUM_LABEL, NULL,
	NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE));

	abd_free(abd);

	if (err != 0) {
	zfs_dbgmsg("L2ARC IO error (%d) while writing device header, "
	"vdev guid: %llu", err,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	}
	}

	/*
	* Commits a log block to the L2ARC device. This routine is invoked from
	* l2arc_write_buffers when the log block fills up.
	* This function allocates some memory to temporarily hold the serialized
	* buffer to be written. This is then released in l2arc_write_done.
	*/
	static void
	l2arc_log_blk_commit(l2arc_dev_t dev, zio_t pio, l2arc_write_callback_t *cb)
	{
	l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	uint64_t psize, asize;
	zio_t *wzio;
	l2arc_lb_abd_buf_t *abd_buf;
	uint8_t *tmpbuf;
	l2arc_lb_ptr_buf_t *lb_ptr_buf;

	VERIFY3S(dev->l2ad_log_ent_idx, ==, dev->l2ad_log_entries);

	tmpbuf = zio_buf_alloc(sizeof (*lb));
	abd_buf = zio_buf_alloc(sizeof (*abd_buf));
	abd_buf->abd = abd_get_from_buf(lb, sizeof (*lb));
	lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
	lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t), KM_SLEEP);

	/* link the buffer into the block chain */
	lb->lb_prev_lbp = l2dhdr->dh_start_lbps[1];
	lb->lb_magic = L2ARC_LOG_BLK_MAGIC;

	/*
	* l2arc_log_blk_commit() may be called multiple times during a single
	* l2arc_write_buffers() call. Save the allocated abd buffers in a list
	* so we can free them in l2arc_write_done() later on.
	*/
	list_insert_tail(&cb->l2wcb_abd_list, abd_buf);

	/* try to compress the buffer */
	psize = zio_compress_data(ZIO_COMPRESS_LZ4,
	abd_buf->abd, tmpbuf, sizeof (*lb), 0);

	/* a log block is never entirely zero */
	ASSERT(psize != 0);
	asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
	ASSERT(asize <= sizeof (*lb));

	/*
	* Update the start log block pointer in the device header to point
	* to the log block we're about to write.
	*/
	l2dhdr->dh_start_lbps[1] = l2dhdr->dh_start_lbps[0];
	l2dhdr->dh_start_lbps[0].lbp_daddr = dev->l2ad_hand;
	l2dhdr->dh_start_lbps[0].lbp_payload_asize =
	dev->l2ad_log_blk_payload_asize;
	l2dhdr->dh_start_lbps[0].lbp_payload_start =
	dev->l2ad_log_blk_payload_start;
	L2BLK_SET_LSIZE(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop, sizeof (*lb));
	L2BLK_SET_PSIZE(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop, asize);
	L2BLK_SET_CHECKSUM(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_CHECKSUM_FLETCHER_4);
	if (asize < sizeof (*lb)) {
	/* compression succeeded */
	memset(tmpbuf + psize, 0, asize - psize);
	L2BLK_SET_COMPRESS(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_COMPRESS_LZ4);
	} else {
	/* compression failed */
	memcpy(tmpbuf, lb, sizeof (*lb));
	L2BLK_SET_COMPRESS(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_COMPRESS_OFF);
	}

	/* checksum what we're about to write */
	fletcher_4_native(tmpbuf, asize, NULL,
	&l2dhdr->dh_start_lbps[0].lbp_cksum);

	abd_free(abd_buf->abd);

	/* perform the write itself */
	abd_buf->abd = abd_get_from_buf(tmpbuf, sizeof (*lb));
	abd_take_ownership_of_buf(abd_buf->abd, B_TRUE);
	wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand,
	asize, abd_buf->abd, ZIO_CHECKSUM_OFF, NULL, NULL,
	ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE);
	DTRACE_PROBE2(l2arc__write, vdev_t , dev->l2ad_vdev, zio_t , wzio);
	(void) zio_nowait(wzio);

	dev->l2ad_hand += asize;
	/*
	* Include the committed log block's pointer in the list of pointers
	* to log blocks present in the L2ARC device.
	*/
	memcpy(lb_ptr_buf->lb_ptr, &l2dhdr->dh_start_lbps[0],
	sizeof (l2arc_log_blkptr_t));
	mutex_enter(&dev->l2ad_mtx);
	list_insert_head(&dev->l2ad_lbptr_list, lb_ptr_buf);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_count);
	zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
	zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
	mutex_exit(&dev->l2ad_mtx);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	/* bump the kstats */
	ARCSTAT_INCR(arcstat_l2_write_bytes, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_writes);
	ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, asize);
	ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio,
	dev->l2ad_log_blk_payload_asize / asize);

	/* start a new log block */
	dev->l2ad_log_ent_idx = 0;
	dev->l2ad_log_blk_payload_asize = 0;
	dev->l2ad_log_blk_payload_start = 0;
	}

	/*
	* Validates an L2ARC log block address to make sure that it can be read
	* from the provided L2ARC device.
	*/
	boolean_t
	l2arc_log_blkptr_valid(l2arc_dev_t dev, const l2arc_log_blkptr_t lbp)
	{
	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	uint64_t asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
	uint64_t end = lbp->lbp_daddr + asize - 1;
	uint64_t start = lbp->lbp_payload_start;
	boolean_t evicted = B_FALSE;

	/*
	* A log block is valid if all of the following conditions are true:
	* - it fits entirely (including its payload) between l2ad_start and
	* l2ad_end
	* - it has a valid size
	* - neither the log block itself nor part of its payload was evicted
	* by l2arc_evict():
	*
	* l2ad_hand l2ad_evict
	* \| \| lbp_daddr
	* \| start \| \| end
	* \| \| \| \| \|
	* V V V V V
	* l2ad_start ============================================ l2ad_end
	* --------------------------\|\|\|\|
	* ^ ^
	* \| log block
	* payload
	*/

	evicted =
	l2arc_range_check_overlap(start, end, dev->l2ad_hand) \|\|
	l2arc_range_check_overlap(start, end, dev->l2ad_evict) \|\|
	l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, start) \|\|
	l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, end);

	return (start >= dev->l2ad_start && end <= dev->l2ad_end &&
	asize > 0 && asize <= sizeof (l2arc_log_blk_phys_t) &&
	(!evicted \|\| dev->l2ad_first));
	}

	/*
	* Inserts ARC buffer header `hdr' into the current L2ARC log block on
	* the device. The buffer being inserted must be present in L2ARC.
	* Returns B_TRUE if the L2ARC log block is full and needs to be committed
	* to L2ARC, or B_FALSE if it still has room for more ARC buffers.
	*/
	static boolean_t
	l2arc_log_blk_insert(l2arc_dev_t dev, const arc_buf_hdr_t hdr)
	{
	l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
	l2arc_log_ent_phys_t *le;

	if (dev->l2ad_log_entries == 0)
	return (B_FALSE);

	int index = dev->l2ad_log_ent_idx++;

	ASSERT3S(index, <, dev->l2ad_log_entries);
	ASSERT(HDR_HAS_L2HDR(hdr));

	le = &lb->lb_entries[index];
	memset(le, 0, sizeof (*le));
	le->le_dva = hdr->b_dva;
	le->le_birth = hdr->b_birth;
	le->le_daddr = hdr->b_l2hdr.b_daddr;
	if (index == 0)
	dev->l2ad_log_blk_payload_start = le->le_daddr;
	L2BLK_SET_LSIZE((le)->le_prop, HDR_GET_LSIZE(hdr));
	L2BLK_SET_PSIZE((le)->le_prop, HDR_GET_PSIZE(hdr));
	L2BLK_SET_COMPRESS((le)->le_prop, HDR_GET_COMPRESS(hdr));
	le->le_complevel = hdr->b_complevel;
	L2BLK_SET_TYPE((le)->le_prop, hdr->b_type);
	L2BLK_SET_PROTECTED((le)->le_prop, !!(HDR_PROTECTED(hdr)));
	L2BLK_SET_PREFETCH((le)->le_prop, !!(HDR_PREFETCH(hdr)));
	L2BLK_SET_STATE((le)->le_prop, hdr->b_l1hdr.b_state->arcs_state);

	dev->l2ad_log_blk_payload_asize += vdev_psize_to_asize(dev->l2ad_vdev,
	HDR_GET_PSIZE(hdr));

	return (dev->l2ad_log_ent_idx == dev->l2ad_log_entries);
	}

	/*
	* Checks whether a given L2ARC device address sits in a time-sequential
	* range. The trick here is that the L2ARC is a rotary buffer, so we can't
	* just do a range comparison, we need to handle the situation in which the
	* range wraps around the end of the L2ARC device. Arguments:
	* bottom -- Lower end of the range to check (written to earlier).
	* top -- Upper end of the range to check (written to later).
	* check -- The address for which we want to determine if it sits in
	* between the top and bottom.
	*
	* The 3-way conditional below represents the following cases:
	*
	* bottom < top : Sequentially ordered case:
	* <check>--------+-------------------+
	* \| (overlap here?) \|
	* L2ARC dev V V
	* \|---------------<bottom>============<top>--------------\|
	*
	* bottom > top: Looped-around case:
	* <check>--------+------------------+
	* \| (overlap here?) \|
	* L2ARC dev V V
	* \|===============<top>---------------<bottom>===========\|
	* ^ ^
	* \| (or here?) \|
	* +---------------+---------<check>
	*
	* top == bottom : Just a single address comparison.
	*/
	boolean_t
	l2arc_range_check_overlap(uint64_t bottom, uint64_t top, uint64_t check)
	{
	if (bottom < top)
	return (bottom <= check && check <= top);
	else if (bottom > top)
	return (check <= top \|\| bottom <= check);
	else
	return (check == top);
	}

	EXPORT_SYMBOL(arc_buf_size);
	EXPORT_SYMBOL(arc_write);
	EXPORT_SYMBOL(arc_read);
	EXPORT_SYMBOL(arc_buf_info);
	EXPORT_SYMBOL(arc_getbuf_func);
	EXPORT_SYMBOL(arc_add_prune_callback);
	EXPORT_SYMBOL(arc_remove_prune_callback);

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min, param_set_arc_min,
	- param_get_long, ZMOD_RW, "Minimum ARC size in bytes");
	+ param_get_ulong, ZMOD_RW, "Minimum ARC size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, max, param_set_arc_max,
	- param_get_long, ZMOD_RW, "Maximum ARC size in bytes");
	+ param_get_ulong, ZMOD_RW, "Maximum ARC size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_limit, param_set_arc_long,
	- param_get_long, ZMOD_RW, "Metadata limit for ARC size in bytes");
	+ param_get_ulong, ZMOD_RW, "Metadata limit for ARC size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_limit_percent,
	- param_set_arc_long, param_get_long, ZMOD_RW,
	+ param_set_arc_long, param_get_ulong, ZMOD_RW,
	"Percent of ARC size for ARC meta limit");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, meta_min, param_set_arc_long,
	- param_get_long, ZMOD_RW, "Minimum ARC metadata size in bytes");
	+ param_get_ulong, ZMOD_RW, "Minimum ARC metadata size in bytes");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_prune, INT, ZMOD_RW,
	"Meta objects to scan for prune");

	-ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_adjust_restarts, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_adjust_restarts, UINT, ZMOD_RW,
	"Limit number of restarts in arc_evict_meta");

	-ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_strategy, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_strategy, UINT, ZMOD_RW,
	"Meta reclaim strategy");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, grow_retry, param_set_arc_int,
	- param_get_int, ZMOD_RW, "Seconds before growing ARC size");
	+ param_get_uint, ZMOD_RW, "Seconds before growing ARC size");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, p_dampener_disable, INT, ZMOD_RW,
	"Disable arc_p adapt dampener");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, shrink_shift, param_set_arc_int,
	- param_get_int, ZMOD_RW, "log2(fraction of ARC to reclaim)");
	+ param_get_uint, ZMOD_RW, "log2(fraction of ARC to reclaim)");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, pc_percent, UINT, ZMOD_RW,
	"Percent of pagecache to reclaim ARC to");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, p_min_shift, param_set_arc_int,
	- param_get_int, ZMOD_RW, "arc_c shift to calc min/max arc_p");
	+ param_get_uint, ZMOD_RW, "arc_c shift to calc min/max arc_p");

	-ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, average_blocksize, INT, ZMOD_RD,
	+ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, average_blocksize, UINT, ZMOD_RD,
	"Target average block size");

	ZFS_MODULE_PARAM(zfs, zfs_, compressed_arc_enabled, INT, ZMOD_RW,
	"Disable compressed ARC buffers");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prefetch_ms, param_set_arc_int,
	- param_get_int, ZMOD_RW, "Min life of prefetch block in ms");
	+ param_get_uint, ZMOD_RW, "Min life of prefetch block in ms");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prescient_prefetch_ms,
	- param_set_arc_int, param_get_int, ZMOD_RW,
	+ param_set_arc_int, param_get_uint, ZMOD_RW,
	"Min life of prescient prefetched block in ms");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_max, ULONG, ZMOD_RW,
	"Max write bytes per interval");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_boost, ULONG, ZMOD_RW,
	"Extra write bytes during device warmup");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom, ULONG, ZMOD_RW,
	"Number of max device writes to precache");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom_boost, ULONG, ZMOD_RW,
	"Compressed l2arc_headroom multiplier");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, trim_ahead, ULONG, ZMOD_RW,
	"TRIM ahead L2ARC write size multiplier");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_secs, ULONG, ZMOD_RW,
	"Seconds between L2ARC writing");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_min_ms, ULONG, ZMOD_RW,
	"Min feed interval in milliseconds");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, noprefetch, INT, ZMOD_RW,
	"Skip caching prefetched buffers");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_again, INT, ZMOD_RW,
	"Turbo L2ARC warmup");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, norw, INT, ZMOD_RW,
	"No reads during writes");

	-ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, meta_percent, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, meta_percent, UINT, ZMOD_RW,
	"Percent of ARC size allowed for L2ARC-only headers");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_enabled, INT, ZMOD_RW,
	"Rebuild the L2ARC when importing a pool");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_blocks_min_l2size, ULONG, ZMOD_RW,
	"Min size in bytes to write rebuild log blocks in L2ARC");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, mfuonly, INT, ZMOD_RW,
	"Cache only MFU data from ARC into L2ARC");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, exclude_special, INT, ZMOD_RW,
	"Exclude dbufs on special vdevs from being cached to L2ARC if set.");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, lotsfree_percent, param_set_arc_int,
	- param_get_int, ZMOD_RW, "System free memory I/O throttle in bytes");
	+ param_get_uint, ZMOD_RW, "System free memory I/O throttle in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, sys_free, param_set_arc_long,
	- param_get_long, ZMOD_RW, "System free memory target size in bytes");
	+ param_get_ulong, ZMOD_RW, "System free memory target size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit, param_set_arc_long,
	- param_get_long, ZMOD_RW, "Minimum bytes of dnodes in ARC");
	+ param_get_ulong, ZMOD_RW, "Minimum bytes of dnodes in ARC");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit_percent,
	- param_set_arc_long, param_get_long, ZMOD_RW,
	+ param_set_arc_long, param_get_ulong, ZMOD_RW,
	"Percent of ARC meta buffers for dnodes");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, dnode_reduce_percent, ULONG, ZMOD_RW,
	"Percentage of excess dnodes to try to unpin");

	-ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, eviction_pct, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, eviction_pct, UINT, ZMOD_RW,
	"When full, ARC allocation waits for eviction of this % of alloc size");

	-ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, evict_batch_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, evict_batch_limit, UINT, ZMOD_RW,
	"The number of headers to evict per sublist before moving to the next");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, prune_task_threads, INT, ZMOD_RW,
	"Number of arc_prune threads");
	diff --git a/sys/contrib/openzfs/module/zfs/dbuf.c b/sys/contrib/openzfs/module/zfs/dbuf.c
	index 80cab8177bc6..db1123d37d98 100644
	--- a/sys/contrib/openzfs/module/zfs/dbuf.c
	+++ b/sys/contrib/openzfs/module/zfs/dbuf.c
	@@ -1,5119 +1,5142 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/dmu.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_impl.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dmu_tx.h>
	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/dmu_zfetch.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/zfeature.h>
	#include <sys/blkptr.h>
	#include <sys/range_tree.h>
	#include <sys/trace_zfs.h>
	#include <sys/callb.h>
	#include <sys/abd.h>
	#include <sys/vdev.h>
	#include <cityhash.h>
	#include <sys/spa_impl.h>
	#include <sys/wmsum.h>
	#include <sys/vdev_impl.h>

	static kstat_t *dbuf_ksp;

	typedef struct dbuf_stats {
	/*
	* Various statistics about the size of the dbuf cache.
	*/
	kstat_named_t cache_count;
	kstat_named_t cache_size_bytes;
	kstat_named_t cache_size_bytes_max;
	/*
	* Statistics regarding the bounds on the dbuf cache size.
	*/
	kstat_named_t cache_target_bytes;
	kstat_named_t cache_lowater_bytes;
	kstat_named_t cache_hiwater_bytes;
	/*
	* Total number of dbuf cache evictions that have occurred.
	*/
	kstat_named_t cache_total_evicts;
	/*
	* The distribution of dbuf levels in the dbuf cache and
	* the total size of all dbufs at each level.
	*/
	kstat_named_t cache_levels[DN_MAX_LEVELS];
	kstat_named_t cache_levels_bytes[DN_MAX_LEVELS];
	/*
	* Statistics about the dbuf hash table.
	*/
	kstat_named_t hash_hits;
	kstat_named_t hash_misses;
	kstat_named_t hash_collisions;
	kstat_named_t hash_elements;
	kstat_named_t hash_elements_max;
	/*
	* Number of sublists containing more than one dbuf in the dbuf
	* hash table. Keep track of the longest hash chain.
	*/
	kstat_named_t hash_chains;
	kstat_named_t hash_chain_max;
	/*
	* Number of times a dbuf_create() discovers that a dbuf was
	* already created and in the dbuf hash table.
	*/
	kstat_named_t hash_insert_race;
	+ /*
	+ * Number of entries in the hash table dbuf and mutex arrays.
	+ */
	+ kstat_named_t hash_table_count;
	+ kstat_named_t hash_mutex_count;
	/*
	* Statistics about the size of the metadata dbuf cache.
	*/
	kstat_named_t metadata_cache_count;
	kstat_named_t metadata_cache_size_bytes;
	kstat_named_t metadata_cache_size_bytes_max;
	/*
	* For diagnostic purposes, this is incremented whenever we can't add
	* something to the metadata cache because it's full, and instead put
	* the data in the regular dbuf cache.
	*/
	kstat_named_t metadata_cache_overflow;
	} dbuf_stats_t;

	dbuf_stats_t dbuf_stats = {
	{ "cache_count", KSTAT_DATA_UINT64 },
	{ "cache_size_bytes", KSTAT_DATA_UINT64 },
	{ "cache_size_bytes_max", KSTAT_DATA_UINT64 },
	{ "cache_target_bytes", KSTAT_DATA_UINT64 },
	{ "cache_lowater_bytes", KSTAT_DATA_UINT64 },
	{ "cache_hiwater_bytes", KSTAT_DATA_UINT64 },
	{ "cache_total_evicts", KSTAT_DATA_UINT64 },
	{ { "cache_levels_N", KSTAT_DATA_UINT64 } },
	{ { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } },
	{ "hash_hits", KSTAT_DATA_UINT64 },
	{ "hash_misses", KSTAT_DATA_UINT64 },
	{ "hash_collisions", KSTAT_DATA_UINT64 },
	{ "hash_elements", KSTAT_DATA_UINT64 },
	{ "hash_elements_max", KSTAT_DATA_UINT64 },
	{ "hash_chains", KSTAT_DATA_UINT64 },
	{ "hash_chain_max", KSTAT_DATA_UINT64 },
	{ "hash_insert_race", KSTAT_DATA_UINT64 },
	+ { "hash_table_count", KSTAT_DATA_UINT64 },
	+ { "hash_mutex_count", KSTAT_DATA_UINT64 },
	{ "metadata_cache_count", KSTAT_DATA_UINT64 },
	{ "metadata_cache_size_bytes", KSTAT_DATA_UINT64 },
	{ "metadata_cache_size_bytes_max", KSTAT_DATA_UINT64 },
	{ "metadata_cache_overflow", KSTAT_DATA_UINT64 }
	};

	struct {
	wmsum_t cache_count;
	wmsum_t cache_total_evicts;
	wmsum_t cache_levels[DN_MAX_LEVELS];
	wmsum_t cache_levels_bytes[DN_MAX_LEVELS];
	wmsum_t hash_hits;
	wmsum_t hash_misses;
	wmsum_t hash_collisions;
	wmsum_t hash_chains;
	wmsum_t hash_insert_race;
	wmsum_t metadata_cache_count;
	wmsum_t metadata_cache_overflow;
	} dbuf_sums;

	#define DBUF_STAT_INCR(stat, val) \
	wmsum_add(&dbuf_sums.stat, val);
	#define DBUF_STAT_DECR(stat, val) \
	DBUF_STAT_INCR(stat, -(val));
	#define DBUF_STAT_BUMP(stat) \
	DBUF_STAT_INCR(stat, 1);
	#define DBUF_STAT_BUMPDOWN(stat) \
	DBUF_STAT_INCR(stat, -1);
	#define DBUF_STAT_MAX(stat, v) { \
	uint64_t _m; \
	while ((v) > (_m = dbuf_stats.stat.value.ui64) && \
	(_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\
	continue; \
	}

	static boolean_t dbuf_undirty(dmu_buf_impl_t db, dmu_tx_t tx);
	static void dbuf_write(dbuf_dirty_record_t dr, arc_buf_t data, dmu_tx_t *tx);
	static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr);
	static int dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags);

	/*
	* Global data structures and functions for the dbuf cache.
	*/
	static kmem_cache_t *dbuf_kmem_cache;
	static taskq_t *dbu_evict_taskq;

	static kthread_t *dbuf_cache_evict_thread;
	static kmutex_t dbuf_evict_lock;
	static kcondvar_t dbuf_evict_cv;
	static boolean_t dbuf_evict_thread_exit;

	/*
	* There are two dbuf caches; each dbuf can only be in one of them at a time.
	*
	* 1. Cache of metadata dbufs, to help make read-heavy administrative commands
	* from /sbin/zfs run faster. The "metadata cache" specifically stores dbufs
	* that represent the metadata that describes filesystems/snapshots/
	* bookmarks/properties/etc. We only evict from this cache when we export a
	* pool, to short-circuit as much I/O as possible for all administrative
	* commands that need the metadata. There is no eviction policy for this
	* cache, because we try to only include types in it which would occupy a
	* very small amount of space per object but create a large impact on the
	* performance of these commands. Instead, after it reaches a maximum size
	* (which should only happen on very small memory systems with a very large
	* number of filesystem objects), we stop taking new dbufs into the
	* metadata cache, instead putting them in the normal dbuf cache.
	*
	* 2. LRU cache of dbufs. The dbuf cache maintains a list of dbufs that
	* are not currently held but have been recently released. These dbufs
	* are not eligible for arc eviction until they are aged out of the cache.
	* Dbufs that are aged out of the cache will be immediately destroyed and
	* become eligible for arc eviction.
	*
	* Dbufs are added to these caches once the last hold is released. If a dbuf is
	* later accessed and still exists in the dbuf cache, then it will be removed
	* from the cache and later re-added to the head of the cache.
	*
	* If a given dbuf meets the requirements for the metadata cache, it will go
	* there, otherwise it will be considered for the generic LRU dbuf cache. The
	* caches and the refcounts tracking their sizes are stored in an array indexed
	* by those caches' matching enum values (from dbuf_cached_state_t).
	*/
	typedef struct dbuf_cache {
	multilist_t cache;
	zfs_refcount_t size ____cacheline_aligned;
	} dbuf_cache_t;
	dbuf_cache_t dbuf_caches[DB_CACHE_MAX];

	/* Size limits for the caches */
	static unsigned long dbuf_cache_max_bytes = ULONG_MAX;
	static unsigned long dbuf_metadata_cache_max_bytes = ULONG_MAX;

	/* Set the default sizes of the caches to log2 fraction of arc size */
	-static int dbuf_cache_shift = 5;
	-static int dbuf_metadata_cache_shift = 6;
	+static uint_t dbuf_cache_shift = 5;
	+static uint_t dbuf_metadata_cache_shift = 6;
	+
	+/* Set the dbuf hash mutex count as log2 shift (dynamic by default) */
	+static uint_t dbuf_mutex_cache_shift = 0;

	static unsigned long dbuf_cache_target_bytes(void);
	static unsigned long dbuf_metadata_cache_target_bytes(void);

	/*
	* The LRU dbuf cache uses a three-stage eviction policy:
	* - A low water marker designates when the dbuf eviction thread
	* should stop evicting from the dbuf cache.
	* - When we reach the maximum size (aka mid water mark), we
	* signal the eviction thread to run.
	* - The high water mark indicates when the eviction thread
	* is unable to keep up with the incoming load and eviction must
	* happen in the context of the calling thread.
	*
	* The dbuf cache:
	* (max size)
	* low water mid water hi water
	* +----------------------------------------+----------+----------+
	* \| \| \| \|
	* \| \| \| \|
	* \| \| \| \|
	* \| \| \| \|
	* +----------------------------------------+----------+----------+
	* stop signal evict
	* evicting eviction directly
	* thread
	*
	* The high and low water marks indicate the operating range for the eviction
	* thread. The low water mark is, by default, 90% of the total size of the
	* cache and the high water mark is at 110% (both of these percentages can be
	* changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct,
	* respectively). The eviction thread will try to ensure that the cache remains
	* within this range by waking up every second and checking if the cache is
	* above the low water mark. The thread can also be woken up by callers adding
	* elements into the cache if the cache is larger than the mid water (i.e max
	* cache size). Once the eviction thread is woken up and eviction is required,
	* it will continue evicting buffers until it's able to reduce the cache size
	* to the low water mark. If the cache size continues to grow and hits the high
	* water mark, then callers adding elements to the cache will begin to evict
	* directly from the cache until the cache is no longer above the high water
	* mark.
	*/

	/*
	* The percentage above and below the maximum cache size.
	*/
	static uint_t dbuf_cache_hiwater_pct = 10;
	static uint_t dbuf_cache_lowater_pct = 10;

	static int
	dbuf_cons(void vdb, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	dmu_buf_impl_t *db = vdb;
	memset(db, 0, sizeof (dmu_buf_impl_t));

	mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL);
	rw_init(&db->db_rwlock, NULL, RW_DEFAULT, NULL);
	cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL);
	multilist_link_init(&db->db_cache_link);
	zfs_refcount_create(&db->db_holds);

	return (0);
	}

	static void
	dbuf_dest(void vdb, void unused)
	{
	(void) unused;
	dmu_buf_impl_t *db = vdb;
	mutex_destroy(&db->db_mtx);
	rw_destroy(&db->db_rwlock);
	cv_destroy(&db->db_changed);
	ASSERT(!multilist_link_active(&db->db_cache_link));
	zfs_refcount_destroy(&db->db_holds);
	}

	/*
	* dbuf hash table routines
	*/
	static dbuf_hash_table_t dbuf_hash_table;

	/*
	* We use Cityhash for this. It's fast, and has good hash properties without
	* requiring any large static buffers.
	*/
	static uint64_t
	dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid)
	{
	return (cityhash4((uintptr_t)os, obj, (uint64_t)lvl, blkid));
	}

	#define DTRACE_SET_STATE(db, why) \
	DTRACE_PROBE2(dbuf__state_change, dmu_buf_impl_t *, db, \
	const char *, why)

	#define DBUF_EQUAL(dbuf, os, obj, level, blkid) \
	((dbuf)->db.db_object == (obj) && \
	(dbuf)->db_objset == (os) && \
	(dbuf)->db_level == (level) && \
	(dbuf)->db_blkid == (blkid))

	dmu_buf_impl_t *
	dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	uint64_t hv;
	uint64_t idx;
	dmu_buf_impl_t *db;

	hv = dbuf_hash(os, obj, level, blkid);
	idx = hv & h->hash_table_mask;

	- rw_enter(DBUF_HASH_RWLOCK(h, idx), RW_READER);
	+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
	for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) {
	if (DBUF_EQUAL(db, os, obj, level, blkid)) {
	mutex_enter(&db->db_mtx);
	if (db->db_state != DB_EVICTING) {
	- rw_exit(DBUF_HASH_RWLOCK(h, idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
	return (db);
	}
	mutex_exit(&db->db_mtx);
	}
	}
	- rw_exit(DBUF_HASH_RWLOCK(h, idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
	return (NULL);
	}

	static dmu_buf_impl_t *
	dbuf_find_bonus(objset_t *os, uint64_t object)
	{
	dnode_t *dn;
	dmu_buf_impl_t *db = NULL;

	if (dnode_hold(os, object, FTAG, &dn) == 0) {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_bonus != NULL) {
	db = dn->dn_bonus;
	mutex_enter(&db->db_mtx);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	}
	return (db);
	}

	/*
	* Insert an entry into the hash table. If there is already an element
	* equal to elem in the hash table, then the already existing element
	* will be returned and the new element will not be inserted.
	* Otherwise returns NULL.
	*/
	static dmu_buf_impl_t *
	dbuf_hash_insert(dmu_buf_impl_t *db)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	objset_t *os = db->db_objset;
	uint64_t obj = db->db.db_object;
	int level = db->db_level;
	uint64_t blkid, hv, idx;
	dmu_buf_impl_t *dbf;
	uint32_t i;

	blkid = db->db_blkid;
	hv = dbuf_hash(os, obj, level, blkid);
	idx = hv & h->hash_table_mask;

	- rw_enter(DBUF_HASH_RWLOCK(h, idx), RW_WRITER);
	+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
	for (dbf = h->hash_table[idx], i = 0; dbf != NULL;
	dbf = dbf->db_hash_next, i++) {
	if (DBUF_EQUAL(dbf, os, obj, level, blkid)) {
	mutex_enter(&dbf->db_mtx);
	if (dbf->db_state != DB_EVICTING) {
	- rw_exit(DBUF_HASH_RWLOCK(h, idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
	return (dbf);
	}
	mutex_exit(&dbf->db_mtx);
	}
	}

	if (i > 0) {
	DBUF_STAT_BUMP(hash_collisions);
	if (i == 1)
	DBUF_STAT_BUMP(hash_chains);

	DBUF_STAT_MAX(hash_chain_max, i);
	}

	mutex_enter(&db->db_mtx);
	db->db_hash_next = h->hash_table[idx];
	h->hash_table[idx] = db;
	- rw_exit(DBUF_HASH_RWLOCK(h, idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
	uint64_t he = atomic_inc_64_nv(&dbuf_stats.hash_elements.value.ui64);
	DBUF_STAT_MAX(hash_elements_max, he);

	return (NULL);
	}

	/*
	* This returns whether this dbuf should be stored in the metadata cache, which
	* is based on whether it's from one of the dnode types that store data related
	* to traversing dataset hierarchies.
	*/
	static boolean_t
	dbuf_include_in_metadata_cache(dmu_buf_impl_t *db)
	{
	DB_DNODE_ENTER(db);
	dmu_object_type_t type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	/* Check if this dbuf is one of the types we care about */
	if (DMU_OT_IS_METADATA_CACHED(type)) {
	/* If we hit this, then we set something up wrong in dmu_ot */
	ASSERT(DMU_OT_IS_METADATA(type));

	/*
	* Sanity check for small-memory systems: don't allocate too
	* much memory for this purpose.
	*/
	if (zfs_refcount_count(
	&dbuf_caches[DB_DBUF_METADATA_CACHE].size) >
	dbuf_metadata_cache_target_bytes()) {
	DBUF_STAT_BUMP(metadata_cache_overflow);
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Remove an entry from the hash table. It must be in the EVICTING state.
	*/
	static void
	dbuf_hash_remove(dmu_buf_impl_t *db)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	uint64_t hv, idx;
	dmu_buf_impl_t dbf, *dbp;

	hv = dbuf_hash(db->db_objset, db->db.db_object,
	db->db_level, db->db_blkid);
	idx = hv & h->hash_table_mask;

	/*
	* We mustn't hold db_mtx to maintain lock ordering:
	- * DBUF_HASH_RWLOCK > db_mtx.
	+ * DBUF_HASH_MUTEX > db_mtx.
	*/
	ASSERT(zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_state == DB_EVICTING);
	ASSERT(!MUTEX_HELD(&db->db_mtx));

	- rw_enter(DBUF_HASH_RWLOCK(h, idx), RW_WRITER);
	+ mutex_enter(DBUF_HASH_MUTEX(h, idx));
	dbp = &h->hash_table[idx];
	while ((dbf = *dbp) != db) {
	dbp = &dbf->db_hash_next;
	ASSERT(dbf != NULL);
	}
	*dbp = db->db_hash_next;
	db->db_hash_next = NULL;
	if (h->hash_table[idx] &&
	h->hash_table[idx]->db_hash_next == NULL)
	DBUF_STAT_BUMPDOWN(hash_chains);
	- rw_exit(DBUF_HASH_RWLOCK(h, idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, idx));
	atomic_dec_64(&dbuf_stats.hash_elements.value.ui64);
	}

	typedef enum {
	DBVU_EVICTING,
	DBVU_NOT_EVICTING
	} dbvu_verify_type_t;

	static void
	dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type)
	{
	#ifdef ZFS_DEBUG
	int64_t holds;

	if (db->db_user == NULL)
	return;

	/* Only data blocks support the attachment of user data. */
	ASSERT(db->db_level == 0);

	/* Clients must resolve a dbuf before attaching user data. */
	ASSERT(db->db.db_data != NULL);
	ASSERT3U(db->db_state, ==, DB_CACHED);

	holds = zfs_refcount_count(&db->db_holds);
	if (verify_type == DBVU_EVICTING) {
	/*
	* Immediate eviction occurs when holds == dirtycnt.
	* For normal eviction buffers, holds is zero on
	* eviction, except when dbuf_fix_old_data() calls
	* dbuf_clear_data(). However, the hold count can grow
	* during eviction even though db_mtx is held (see
	* dmu_bonus_hold() for an example), so we can only
	* test the generic invariant that holds >= dirtycnt.
	*/
	ASSERT3U(holds, >=, db->db_dirtycnt);
	} else {
	if (db->db_user_immediate_evict == TRUE)
	ASSERT3U(holds, >=, db->db_dirtycnt);
	else
	ASSERT3U(holds, >, 0);
	}
	#endif
	}

	static void
	dbuf_evict_user(dmu_buf_impl_t *db)
	{
	dmu_buf_user_t *dbu = db->db_user;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (dbu == NULL)
	return;

	dbuf_verify_user(db, DBVU_EVICTING);
	db->db_user = NULL;

	#ifdef ZFS_DEBUG
	if (dbu->dbu_clear_on_evict_dbufp != NULL)
	*dbu->dbu_clear_on_evict_dbufp = NULL;
	#endif

	/*
	* There are two eviction callbacks - one that we call synchronously
	* and one that we invoke via a taskq. The async one is useful for
	* avoiding lock order reversals and limiting stack depth.
	*
	* Note that if we have a sync callback but no async callback,
	* it's likely that the sync callback will free the structure
	* containing the dbu. In that case we need to take care to not
	* dereference dbu after calling the sync evict func.
	*/
	boolean_t has_async = (dbu->dbu_evict_func_async != NULL);

	if (dbu->dbu_evict_func_sync != NULL)
	dbu->dbu_evict_func_sync(dbu);

	if (has_async) {
	taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async,
	dbu, 0, &dbu->dbu_tqent);
	}
	}

	boolean_t
	dbuf_is_metadata(dmu_buf_impl_t *db)
	{
	/*
	* Consider indirect blocks and spill blocks to be meta data.
	*/
	if (db->db_level > 0 \|\| db->db_blkid == DMU_SPILL_BLKID) {
	return (B_TRUE);
	} else {
	boolean_t is_metadata;

	DB_DNODE_ENTER(db);
	is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type);
	DB_DNODE_EXIT(db);

	return (is_metadata);
	}
	}

	/*
	* We want to exclude buffers that are on a special allocation class from
	* L2ARC.
	*/
	boolean_t
	dbuf_is_l2cacheable(dmu_buf_impl_t *db)
	{
	vdev_t *vd = NULL;
	zfs_cache_type_t cache = db->db_objset->os_secondary_cache;
	blkptr_t *bp = db->db_blkptr;

	if (bp != NULL && !BP_IS_HOLE(bp)) {
	uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
	vdev_t *rvd = db->db_objset->os_spa->spa_root_vdev;

	if (vdev < rvd->vdev_children)
	vd = rvd->vdev_child[vdev];

	if (cache == ZFS_CACHE_ALL \|\|
	(dbuf_is_metadata(db) && cache == ZFS_CACHE_METADATA)) {
	if (vd == NULL)
	return (B_TRUE);

	if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
	vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) \|\|
	l2arc_exclude_special == 0)
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}

	static inline boolean_t
	dnode_level_is_l2cacheable(blkptr_t bp, dnode_t dn, int64_t level)
	{
	vdev_t *vd = NULL;
	zfs_cache_type_t cache = dn->dn_objset->os_secondary_cache;

	if (bp != NULL && !BP_IS_HOLE(bp)) {
	uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
	vdev_t *rvd = dn->dn_objset->os_spa->spa_root_vdev;

	if (vdev < rvd->vdev_children)
	vd = rvd->vdev_child[vdev];

	if (cache == ZFS_CACHE_ALL \|\| ((level > 0 \|\|
	DMU_OT_IS_METADATA(dn->dn_handle->dnh_dnode->dn_type)) &&
	cache == ZFS_CACHE_METADATA)) {
	if (vd == NULL)
	return (B_TRUE);

	if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
	vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) \|\|
	l2arc_exclude_special == 0)
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}


	/*
	* This function must return indices evenly distributed between all
	* sublists of the multilist. This is needed due to how the dbuf eviction
	* code is laid out; dbuf_evict_thread() assumes dbufs are evenly
	* distributed between all sublists and uses this assumption when
	* deciding which sublist to evict from and how much to evict from it.
	*/
	static unsigned int
	dbuf_cache_multilist_index_func(multilist_t ml, void obj)
	{
	dmu_buf_impl_t *db = obj;

	/*
	* The assumption here, is the hash value for a given
	* dmu_buf_impl_t will remain constant throughout it's lifetime
	* (i.e. it's objset, object, level and blkid fields don't change).
	* Thus, we don't need to store the dbuf's sublist index
	* on insertion, as this index can be recalculated on removal.
	*
	* Also, the low order bits of the hash value are thought to be
	* distributed evenly. Otherwise, in the case that the multilist
	* has a power of two number of sublists, each sublists' usage
	* would not be evenly distributed. In this context full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)dbuf_hash(db->db_objset, db->db.db_object,
	db->db_level, db->db_blkid) %
	multilist_get_num_sublists(ml));
	}

	/*
	* The target size of the dbuf cache can grow with the ARC target,
	* unless limited by the tunable dbuf_cache_max_bytes.
	*/
	static inline unsigned long
	dbuf_cache_target_bytes(void)
	{
	return (MIN(dbuf_cache_max_bytes,
	arc_target_bytes() >> dbuf_cache_shift));
	}

	/*
	* The target size of the dbuf metadata cache can grow with the ARC target,
	* unless limited by the tunable dbuf_metadata_cache_max_bytes.
	*/
	static inline unsigned long
	dbuf_metadata_cache_target_bytes(void)
	{
	return (MIN(dbuf_metadata_cache_max_bytes,
	arc_target_bytes() >> dbuf_metadata_cache_shift));
	}

	static inline uint64_t
	dbuf_cache_hiwater_bytes(void)
	{
	uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
	return (dbuf_cache_target +
	(dbuf_cache_target * dbuf_cache_hiwater_pct) / 100);
	}

	static inline uint64_t
	dbuf_cache_lowater_bytes(void)
	{
	uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
	return (dbuf_cache_target -
	(dbuf_cache_target * dbuf_cache_lowater_pct) / 100);
	}

	static inline boolean_t
	dbuf_cache_above_lowater(void)
	{
	return (zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) >
	dbuf_cache_lowater_bytes());
	}

	/*
	* Evict the oldest eligible dbuf from the dbuf cache.
	*/
	static void
	dbuf_evict_one(void)
	{
	int idx = multilist_get_random_index(&dbuf_caches[DB_DBUF_CACHE].cache);
	multilist_sublist_t *mls = multilist_sublist_lock(
	&dbuf_caches[DB_DBUF_CACHE].cache, idx);

	ASSERT(!MUTEX_HELD(&dbuf_evict_lock));

	dmu_buf_impl_t *db = multilist_sublist_tail(mls);
	while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) {
	db = multilist_sublist_prev(mls, db);
	}

	DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db,
	multilist_sublist_t *, mls);

	if (db != NULL) {
	multilist_sublist_remove(mls, db);
	multilist_sublist_unlock(mls);
	(void) zfs_refcount_remove_many(
	&dbuf_caches[DB_DBUF_CACHE].size, db->db.db_size, db);
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
	db->db.db_size);
	ASSERT3U(db->db_caching_status, ==, DB_DBUF_CACHE);
	db->db_caching_status = DB_NO_CACHE;
	dbuf_destroy(db);
	DBUF_STAT_BUMP(cache_total_evicts);
	} else {
	multilist_sublist_unlock(mls);
	}
	}

	/*
	* The dbuf evict thread is responsible for aging out dbufs from the
	* cache. Once the cache has reached it's maximum size, dbufs are removed
	* and destroyed. The eviction thread will continue running until the size
	* of the dbuf cache is at or below the maximum size. Once the dbuf is aged
	* out of the cache it is destroyed and becomes eligible for arc eviction.
	*/
	static __attribute__((noreturn)) void
	dbuf_evict_thread(void *unused)
	{
	(void) unused;
	callb_cpr_t cpr;

	CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG);

	mutex_enter(&dbuf_evict_lock);
	while (!dbuf_evict_thread_exit) {
	while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
	CALLB_CPR_SAFE_BEGIN(&cpr);
	(void) cv_timedwait_idle_hires(&dbuf_evict_cv,
	&dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0);
	CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock);
	}
	mutex_exit(&dbuf_evict_lock);

	/*
	* Keep evicting as long as we're above the low water mark
	* for the cache. We do this without holding the locks to
	* minimize lock contention.
	*/
	while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
	dbuf_evict_one();
	}

	mutex_enter(&dbuf_evict_lock);
	}

	dbuf_evict_thread_exit = B_FALSE;
	cv_broadcast(&dbuf_evict_cv);
	CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */
	thread_exit();
	}

	/*
	* Wake up the dbuf eviction thread if the dbuf cache is at its max size.
	* If the dbuf cache is at its high water mark, then evict a dbuf from the
	* dbuf cache using the caller's context.
	*/
	static void
	dbuf_evict_notify(uint64_t size)
	{
	/*
	* We check if we should evict without holding the dbuf_evict_lock,
	* because it's OK to occasionally make the wrong decision here,
	* and grabbing the lock results in massive lock contention.
	*/
	if (size > dbuf_cache_target_bytes()) {
	if (size > dbuf_cache_hiwater_bytes())
	dbuf_evict_one();
	cv_signal(&dbuf_evict_cv);
	}
	}

	static int
	dbuf_kstat_update(kstat_t *ksp, int rw)
	{
	dbuf_stats_t *ds = ksp->ks_data;
	+ dbuf_hash_table_t *h = &dbuf_hash_table;

	if (rw == KSTAT_WRITE)
	return (SET_ERROR(EACCES));

	ds->cache_count.value.ui64 =
	wmsum_value(&dbuf_sums.cache_count);
	ds->cache_size_bytes.value.ui64 =
	zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size);
	ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes();
	ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes();
	ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes();
	ds->cache_total_evicts.value.ui64 =
	wmsum_value(&dbuf_sums.cache_total_evicts);
	for (int i = 0; i < DN_MAX_LEVELS; i++) {
	ds->cache_levels[i].value.ui64 =
	wmsum_value(&dbuf_sums.cache_levels[i]);
	ds->cache_levels_bytes[i].value.ui64 =
	wmsum_value(&dbuf_sums.cache_levels_bytes[i]);
	}
	ds->hash_hits.value.ui64 =
	wmsum_value(&dbuf_sums.hash_hits);
	ds->hash_misses.value.ui64 =
	wmsum_value(&dbuf_sums.hash_misses);
	ds->hash_collisions.value.ui64 =
	wmsum_value(&dbuf_sums.hash_collisions);
	ds->hash_chains.value.ui64 =
	wmsum_value(&dbuf_sums.hash_chains);
	ds->hash_insert_race.value.ui64 =
	wmsum_value(&dbuf_sums.hash_insert_race);
	+ ds->hash_table_count.value.ui64 = h->hash_table_mask + 1;
	+ ds->hash_mutex_count.value.ui64 = h->hash_mutex_mask + 1;
	ds->metadata_cache_count.value.ui64 =
	wmsum_value(&dbuf_sums.metadata_cache_count);
	ds->metadata_cache_size_bytes.value.ui64 = zfs_refcount_count(
	&dbuf_caches[DB_DBUF_METADATA_CACHE].size);
	ds->metadata_cache_overflow.value.ui64 =
	wmsum_value(&dbuf_sums.metadata_cache_overflow);
	return (0);
	}

	void
	dbuf_init(void)
	{
	- uint64_t hsize = 1ULL << 16;
	+ uint64_t hmsize, hsize = 1ULL << 16;
	dbuf_hash_table_t *h = &dbuf_hash_table;
	- int i;

	/*
	* The hash table is big enough to fill one eighth of physical memory
	* with an average block size of zfs_arc_average_blocksize (default 8K).
	* By default, the table will take up
	* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
	*/
	while (hsize * zfs_arc_average_blocksize < arc_all_memory() / 8)
	hsize <<= 1;

	-retry:
	- h->hash_table_mask = hsize - 1;
	-#if defined(_KERNEL)
	+ h->hash_table = NULL;
	+ while (h->hash_table == NULL) {
	+ h->hash_table_mask = hsize - 1;
	+
	+ h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP);
	+ if (h->hash_table == NULL)
	+ hsize >>= 1;
	+
	+ ASSERT3U(hsize, >=, 1ULL << 10);
	+ }
	+
	/*
	- * Large allocations which do not require contiguous pages
	- * should be using vmem_alloc() in the linux kernel
	+ * The hash table buckets are protected by an array of mutexes where
	+ * each mutex is reponsible for protecting 128 buckets. A minimum
	+ * array size of 8192 is targeted to avoid contention.
	*/
	- h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP);
	-#else
	- h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP);
	-#endif
	- if (h->hash_table == NULL) {
	- /* XXX - we should really return an error instead of assert */
	- ASSERT(hsize > (1ULL << 10));
	- hsize >>= 1;
	- goto retry;
	+ if (dbuf_mutex_cache_shift == 0)
	+ hmsize = MAX(hsize >> 7, 1ULL << 13);
	+ else
	+ hmsize = 1ULL << MIN(dbuf_mutex_cache_shift, 24);
	+
	+ h->hash_mutexes = NULL;
	+ while (h->hash_mutexes == NULL) {
	+ h->hash_mutex_mask = hmsize - 1;
	+
	+ h->hash_mutexes = vmem_zalloc(hmsize * sizeof (kmutex_t),
	+ KM_SLEEP);
	+ if (h->hash_mutexes == NULL)
	+ hmsize >>= 1;
	}

	dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t",
	sizeof (dmu_buf_impl_t),
	0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0);

	- for (i = 0; i < DBUF_RWLOCKS; i++)
	- rw_init(&h->hash_rwlocks[i], NULL, RW_DEFAULT, NULL);
	+ for (int i = 0; i < hmsize; i++)
	+ mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL);

	dbuf_stats_init(h);

	/*
	* All entries are queued via taskq_dispatch_ent(), so min/maxalloc
	* configuration is not required.
	*/
	dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0);

	for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
	multilist_create(&dbuf_caches[dcs].cache,
	sizeof (dmu_buf_impl_t),
	offsetof(dmu_buf_impl_t, db_cache_link),
	dbuf_cache_multilist_index_func);
	zfs_refcount_create(&dbuf_caches[dcs].size);
	}

	dbuf_evict_thread_exit = B_FALSE;
	mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL);
	dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread,
	NULL, 0, &p0, TS_RUN, minclsyspri);

	wmsum_init(&dbuf_sums.cache_count, 0);
	wmsum_init(&dbuf_sums.cache_total_evicts, 0);
	- for (i = 0; i < DN_MAX_LEVELS; i++) {
	+ for (int i = 0; i < DN_MAX_LEVELS; i++) {
	wmsum_init(&dbuf_sums.cache_levels[i], 0);
	wmsum_init(&dbuf_sums.cache_levels_bytes[i], 0);
	}
	wmsum_init(&dbuf_sums.hash_hits, 0);
	wmsum_init(&dbuf_sums.hash_misses, 0);
	wmsum_init(&dbuf_sums.hash_collisions, 0);
	wmsum_init(&dbuf_sums.hash_chains, 0);
	wmsum_init(&dbuf_sums.hash_insert_race, 0);
	wmsum_init(&dbuf_sums.metadata_cache_count, 0);
	wmsum_init(&dbuf_sums.metadata_cache_overflow, 0);

	dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc",
	KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);
	if (dbuf_ksp != NULL) {
	- for (i = 0; i < DN_MAX_LEVELS; i++) {
	+ for (int i = 0; i < DN_MAX_LEVELS; i++) {
	snprintf(dbuf_stats.cache_levels[i].name,
	KSTAT_STRLEN, "cache_level_%d", i);
	dbuf_stats.cache_levels[i].data_type =
	KSTAT_DATA_UINT64;
	snprintf(dbuf_stats.cache_levels_bytes[i].name,
	KSTAT_STRLEN, "cache_level_%d_bytes", i);
	dbuf_stats.cache_levels_bytes[i].data_type =
	KSTAT_DATA_UINT64;
	}
	dbuf_ksp->ks_data = &dbuf_stats;
	dbuf_ksp->ks_update = dbuf_kstat_update;
	kstat_install(dbuf_ksp);
	}
	}

	void
	dbuf_fini(void)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	- int i;

	dbuf_stats_destroy();

	- for (i = 0; i < DBUF_RWLOCKS; i++)
	- rw_destroy(&h->hash_rwlocks[i]);
	-#if defined(_KERNEL)
	- /*
	- * Large allocations which do not require contiguous pages
	- * should be using vmem_free() in the linux kernel
	- */
	+ for (int i = 0; i < (h->hash_mutex_mask + 1); i++)
	+ mutex_destroy(&h->hash_mutexes[i]);
	+
	vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
	-#else
	- kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
	-#endif
	+ vmem_free(h->hash_mutexes, (h->hash_mutex_mask + 1) *
	+ sizeof (kmutex_t));
	+
	kmem_cache_destroy(dbuf_kmem_cache);
	taskq_destroy(dbu_evict_taskq);

	mutex_enter(&dbuf_evict_lock);
	dbuf_evict_thread_exit = B_TRUE;
	while (dbuf_evict_thread_exit) {
	cv_signal(&dbuf_evict_cv);
	cv_wait(&dbuf_evict_cv, &dbuf_evict_lock);
	}
	mutex_exit(&dbuf_evict_lock);

	mutex_destroy(&dbuf_evict_lock);
	cv_destroy(&dbuf_evict_cv);

	for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
	zfs_refcount_destroy(&dbuf_caches[dcs].size);
	multilist_destroy(&dbuf_caches[dcs].cache);
	}

	if (dbuf_ksp != NULL) {
	kstat_delete(dbuf_ksp);
	dbuf_ksp = NULL;
	}

	wmsum_fini(&dbuf_sums.cache_count);
	wmsum_fini(&dbuf_sums.cache_total_evicts);
	- for (i = 0; i < DN_MAX_LEVELS; i++) {
	+ for (int i = 0; i < DN_MAX_LEVELS; i++) {
	wmsum_fini(&dbuf_sums.cache_levels[i]);
	wmsum_fini(&dbuf_sums.cache_levels_bytes[i]);
	}
	wmsum_fini(&dbuf_sums.hash_hits);
	wmsum_fini(&dbuf_sums.hash_misses);
	wmsum_fini(&dbuf_sums.hash_collisions);
	wmsum_fini(&dbuf_sums.hash_chains);
	wmsum_fini(&dbuf_sums.hash_insert_race);
	wmsum_fini(&dbuf_sums.metadata_cache_count);
	wmsum_fini(&dbuf_sums.metadata_cache_overflow);
	}

	/*
	* Other stuff.
	*/

	#ifdef ZFS_DEBUG
	static void
	dbuf_verify(dmu_buf_impl_t *db)
	{
	dnode_t *dn;
	dbuf_dirty_record_t *dr;
	uint32_t txg_prev;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY))
	return;

	ASSERT(db->db_objset != NULL);
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	if (dn == NULL) {
	ASSERT(db->db_parent == NULL);
	ASSERT(db->db_blkptr == NULL);
	} else {
	ASSERT3U(db->db.db_object, ==, dn->dn_object);
	ASSERT3P(db->db_objset, ==, dn->dn_objset);
	ASSERT3U(db->db_level, <, dn->dn_nlevels);
	ASSERT(db->db_blkid == DMU_BONUS_BLKID \|\|
	db->db_blkid == DMU_SPILL_BLKID \|\|
	!avl_is_empty(&dn->dn_dbufs));
	}
	if (db->db_blkid == DMU_BONUS_BLKID) {
	ASSERT(dn != NULL);
	ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
	ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID);
	} else if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn != NULL);
	ASSERT0(db->db.db_offset);
	} else {
	ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size);
	}

	if ((dr = list_head(&db->db_dirty_records)) != NULL) {
	ASSERT(dr->dr_dbuf == db);
	txg_prev = dr->dr_txg;
	for (dr = list_next(&db->db_dirty_records, dr); dr != NULL;
	dr = list_next(&db->db_dirty_records, dr)) {
	ASSERT(dr->dr_dbuf == db);
	ASSERT(txg_prev > dr->dr_txg);
	txg_prev = dr->dr_txg;
	}
	}

	/*
	* We can't assert that db_size matches dn_datablksz because it
	* can be momentarily different when another thread is doing
	* dnode_set_blksz().
	*/
	if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) {
	dr = db->db_data_pending;
	/*
	* It should only be modified in syncing context, so
	* make sure we only have one copy of the data.
	*/
	ASSERT(dr == NULL \|\| dr->dt.dl.dr_data == db->db_buf);
	}

	/* verify db->db_blkptr */
	if (db->db_blkptr) {
	if (db->db_parent == dn->dn_dbuf) {
	/* db is pointed to by the dnode */
	/* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */
	if (DMU_OBJECT_IS_SPECIAL(db->db.db_object))
	ASSERT(db->db_parent == NULL);
	else
	ASSERT(db->db_parent != NULL);
	if (db->db_blkid != DMU_SPILL_BLKID)
	ASSERT3P(db->db_blkptr, ==,
	&dn->dn_phys->dn_blkptr[db->db_blkid]);
	} else {
	/* db is pointed to by an indirect block */
	int epb __maybe_unused = db->db_parent->db.db_size >>
	SPA_BLKPTRSHIFT;
	ASSERT3U(db->db_parent->db_level, ==, db->db_level+1);
	ASSERT3U(db->db_parent->db.db_object, ==,
	db->db.db_object);
	/*
	* dnode_grow_indblksz() can make this fail if we don't
	* have the parent's rwlock. XXX indblksz no longer
	* grows. safe to do this now?
	*/
	if (RW_LOCK_HELD(&db->db_parent->db_rwlock)) {
	ASSERT3P(db->db_blkptr, ==,
	((blkptr_t *)db->db_parent->db.db_data +
	db->db_blkid % epb));
	}
	}
	}
	if ((db->db_blkptr == NULL \|\| BP_IS_HOLE(db->db_blkptr)) &&
	(db->db_buf == NULL \|\| db->db_buf->b_data) &&
	db->db.db_data && db->db_blkid != DMU_BONUS_BLKID &&
	db->db_state != DB_FILL && !dn->dn_free_txg) {
	/*
	* If the blkptr isn't set but they have nonzero data,
	* it had better be dirty, otherwise we'll lose that
	* data when we evict this buffer.
	*
	* There is an exception to this rule for indirect blocks; in
	* this case, if the indirect block is a hole, we fill in a few
	* fields on each of the child blocks (importantly, birth time)
	* to prevent hole birth times from being lost when you
	* partially fill in a hole.
	*/
	if (db->db_dirtycnt == 0) {
	if (db->db_level == 0) {
	uint64_t *buf = db->db.db_data;
	int i;

	for (i = 0; i < db->db.db_size >> 3; i++) {
	ASSERT(buf[i] == 0);
	}
	} else {
	blkptr_t *bps = db->db.db_data;
	ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==,
	db->db.db_size);
	/*
	* We want to verify that all the blkptrs in the
	* indirect block are holes, but we may have
	* automatically set up a few fields for them.
	* We iterate through each blkptr and verify
	* they only have those fields set.
	*/
	for (int i = 0;
	i < db->db.db_size / sizeof (blkptr_t);
	i++) {
	blkptr_t *bp = &bps[i];
	ASSERT(ZIO_CHECKSUM_IS_ZERO(
	&bp->blk_cksum));
	ASSERT(
	DVA_IS_EMPTY(&bp->blk_dva[0]) &&
	DVA_IS_EMPTY(&bp->blk_dva[1]) &&
	DVA_IS_EMPTY(&bp->blk_dva[2]));
	ASSERT0(bp->blk_fill);
	ASSERT0(bp->blk_pad[0]);
	ASSERT0(bp->blk_pad[1]);
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(BP_IS_HOLE(bp));
	ASSERT0(bp->blk_phys_birth);
	}
	}
	}
	}
	DB_DNODE_EXIT(db);
	}
	#endif

	static void
	dbuf_clear_data(dmu_buf_impl_t *db)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));
	dbuf_evict_user(db);
	ASSERT3P(db->db_buf, ==, NULL);
	db->db.db_data = NULL;
	if (db->db_state != DB_NOFILL) {
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "clear data");
	}
	}

	static void
	dbuf_set_data(dmu_buf_impl_t db, arc_buf_t buf)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(buf != NULL);

	db->db_buf = buf;
	ASSERT(buf->b_data != NULL);
	db->db.db_data = buf->b_data;
	}

	static arc_buf_t *
	dbuf_alloc_arcbuf(dmu_buf_impl_t *db)
	{
	spa_t *spa = db->db_objset->os_spa;

	return (arc_alloc_buf(spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size));
	}

	/*
	* Loan out an arc_buf for read. Return the loaned arc_buf.
	*/
	arc_buf_t *
	dbuf_loan_arcbuf(dmu_buf_impl_t *db)
	{
	arc_buf_t *abuf;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	mutex_enter(&db->db_mtx);
	if (arc_released(db->db_buf) \|\| zfs_refcount_count(&db->db_holds) > 1) {
	int blksz = db->db.db_size;
	spa_t *spa = db->db_objset->os_spa;

	mutex_exit(&db->db_mtx);
	abuf = arc_loan_buf(spa, B_FALSE, blksz);
	memcpy(abuf->b_data, db->db.db_data, blksz);
	} else {
	abuf = db->db_buf;
	arc_loan_inuse_buf(abuf, db);
	db->db_buf = NULL;
	dbuf_clear_data(db);
	mutex_exit(&db->db_mtx);
	}
	return (abuf);
	}

	/*
	* Calculate which level n block references the data at the level 0 offset
	* provided.
	*/
	uint64_t
	dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset)
	{
	if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) {
	/*
	* The level n blkid is equal to the level 0 blkid divided by
	* the number of level 0s in a level n block.
	*
	* The level 0 blkid is offset >> datablkshift =
	* offset / 2^datablkshift.
	*
	* The number of level 0s in a level n is the number of block
	* pointers in an indirect block, raised to the power of level.
	* This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level =
	* 2^(level*(indblkshift - SPA_BLKPTRSHIFT)).
	*
	* Thus, the level n blkid is: offset /
	* ((2^datablkshift)(2^(level(indblkshift-SPA_BLKPTRSHIFT))))
	* = offset / 2^(datablkshift + level *
	* (indblkshift - SPA_BLKPTRSHIFT))
	* = offset >> (datablkshift + level *
	* (indblkshift - SPA_BLKPTRSHIFT))
	*/

	const unsigned exp = dn->dn_datablkshift +
	level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT);

	if (exp >= 8 * sizeof (offset)) {
	/* This only happens on the highest indirection level */
	ASSERT3U(level, ==, dn->dn_nlevels - 1);
	return (0);
	}

	ASSERT3U(exp, <, 8 * sizeof (offset));

	return (offset >> exp);
	} else {
	ASSERT3U(offset, <, dn->dn_datablksz);
	return (0);
	}
	}

	/*
	* This function is used to lock the parent of the provided dbuf. This should be
	* used when modifying or reading db_blkptr.
	*/
	db_lock_type_t
	dmu_buf_lock_parent(dmu_buf_impl_t db, krw_t rw, const void tag)
	{
	enum db_lock_type ret = DLT_NONE;
	if (db->db_parent != NULL) {
	rw_enter(&db->db_parent->db_rwlock, rw);
	ret = DLT_PARENT;
	} else if (dmu_objset_ds(db->db_objset) != NULL) {
	rrw_enter(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, rw,
	tag);
	ret = DLT_OBJSET;
	}
	/*
	* We only return a DLT_NONE lock when it's the top-most indirect block
	* of the meta-dnode of the MOS.
	*/
	return (ret);
	}

	/*
	* We need to pass the lock type in because it's possible that the block will
	* move from being the topmost indirect block in a dnode (and thus, have no
	* parent) to not the top-most via an indirection increase. This would cause a
	* panic if we didn't pass the lock type in.
	*/
	void
	dmu_buf_unlock_parent(dmu_buf_impl_t db, db_lock_type_t type, const void tag)
	{
	if (type == DLT_PARENT)
	rw_exit(&db->db_parent->db_rwlock);
	else if (type == DLT_OBJSET)
	rrw_exit(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, tag);
	}

	static void
	dbuf_read_done(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void vdb)
	{
	(void) zb, (void) bp;
	dmu_buf_impl_t *db = vdb;

	mutex_enter(&db->db_mtx);
	ASSERT3U(db->db_state, ==, DB_READ);
	/*
	* All reads are synchronous, so we must have a hold on the dbuf
	*/
	ASSERT(zfs_refcount_count(&db->db_holds) > 0);
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	if (buf == NULL) {
	/* i/o error */
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT3P(db->db_buf, ==, NULL);
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "i/o error");
	} else if (db->db_level == 0 && db->db_freed_in_flight) {
	/* freed in flight */
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	arc_release(buf, db);
	memset(buf->b_data, 0, db->db.db_size);
	arc_buf_freeze(buf);
	db->db_freed_in_flight = FALSE;
	dbuf_set_data(db, buf);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "freed in flight");
	} else {
	/* success */
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	dbuf_set_data(db, buf);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "successful read");
	}
	cv_broadcast(&db->db_changed);
	dbuf_rele_and_unlock(db, NULL, B_FALSE);
	}

	/*
	* Shortcut for performing reads on bonus dbufs. Returns
	* an error if we fail to verify the dnode associated with
	* a decrypted block. Otherwise success.
	*/
	static int
	dbuf_read_bonus(dmu_buf_impl_t db, dnode_t dn, uint32_t flags)
	{
	int bonuslen, max_bonuslen, err;

	err = dbuf_read_verify_dnode_crypt(db, flags);
	if (err)
	return (err);

	bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen);
	max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(DB_DNODE_HELD(db));
	ASSERT3U(bonuslen, <=, db->db.db_size);
	db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP);
	arc_space_consume(max_bonuslen, ARC_SPACE_BONUS);
	if (bonuslen < max_bonuslen)
	memset(db->db.db_data, 0, max_bonuslen);
	if (bonuslen)
	memcpy(db->db.db_data, DN_BONUS(dn->dn_phys), bonuslen);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "bonus buffer filled");
	return (0);
	}

	static void
	dbuf_handle_indirect_hole(dmu_buf_impl_t db, dnode_t dn)
	{
	blkptr_t *bps = db->db.db_data;
	uint32_t indbs = 1ULL << dn->dn_indblkshift;
	int n_bps = indbs >> SPA_BLKPTRSHIFT;

	for (int i = 0; i < n_bps; i++) {
	blkptr_t *bp = &bps[i];

	ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, indbs);
	BP_SET_LSIZE(bp, BP_GET_LEVEL(db->db_blkptr) == 1 ?
	dn->dn_datablksz : BP_GET_LSIZE(db->db_blkptr));
	BP_SET_TYPE(bp, BP_GET_TYPE(db->db_blkptr));
	BP_SET_LEVEL(bp, BP_GET_LEVEL(db->db_blkptr) - 1);
	BP_SET_BIRTH(bp, db->db_blkptr->blk_birth, 0);
	}
	}

	/*
	* Handle reads on dbufs that are holes, if necessary. This function
	* requires that the dbuf's mutex is held. Returns success (0) if action
	* was taken, ENOENT if no action was taken.
	*/
	static int
	dbuf_read_hole(dmu_buf_impl_t db, dnode_t dn)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));

	int is_hole = db->db_blkptr == NULL \|\| BP_IS_HOLE(db->db_blkptr);
	/*
	* For level 0 blocks only, if the above check fails:
	* Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync()
	* processes the delete record and clears the bp while we are waiting
	* for the dn_mtx (resulting in a "no" from block_freed).
	*/
	if (!is_hole && db->db_level == 0) {
	is_hole = dnode_block_freed(dn, db->db_blkid) \|\|
	BP_IS_HOLE(db->db_blkptr);
	}

	if (is_hole) {
	dbuf_set_data(db, dbuf_alloc_arcbuf(db));
	memset(db->db.db_data, 0, db->db.db_size);

	if (db->db_blkptr != NULL && db->db_level > 0 &&
	BP_IS_HOLE(db->db_blkptr) &&
	db->db_blkptr->blk_birth != 0) {
	dbuf_handle_indirect_hole(db, dn);
	}
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "hole read satisfied");
	return (0);
	}
	return (ENOENT);
	}

	/*
	* This function ensures that, when doing a decrypting read of a block,
	* we make sure we have decrypted the dnode associated with it. We must do
	* this so that we ensure we are fully authenticating the checksum-of-MACs
	* tree from the root of the objset down to this block. Indirect blocks are
	* always verified against their secure checksum-of-MACs assuming that the
	* dnode containing them is correct. Now that we are doing a decrypting read,
	* we can be sure that the key is loaded and verify that assumption. This is
	* especially important considering that we always read encrypted dnode
	* blocks as raw data (without verifying their MACs) to start, and
	* decrypt / authenticate them when we need to read an encrypted bonus buffer.
	*/
	static int
	dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags)
	{
	int err = 0;
	objset_t *os = db->db_objset;
	arc_buf_t *dnode_abuf;
	dnode_t *dn;
	zbookmark_phys_t zb;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (!os->os_encrypted \|\| os->os_raw_receive \|\|
	(flags & DB_RF_NO_DECRYPT) != 0)
	return (0);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	dnode_abuf = (dn->dn_dbuf != NULL) ? dn->dn_dbuf->db_buf : NULL;

	if (dnode_abuf == NULL \|\| !arc_is_encrypted(dnode_abuf)) {
	DB_DNODE_EXIT(db);
	return (0);
	}

	SET_BOOKMARK(&zb, dmu_objset_id(os),
	DMU_META_DNODE_OBJECT, 0, dn->dn_dbuf->db_blkid);
	err = arc_untransform(dnode_abuf, os->os_spa, &zb, B_TRUE);

	/*
	* An error code of EACCES tells us that the key is still not
	* available. This is ok if we are only reading authenticated
	* (and therefore non-encrypted) blocks.
	*/
	if (err == EACCES && ((db->db_blkid != DMU_BONUS_BLKID &&
	!DMU_OT_IS_ENCRYPTED(dn->dn_type)) \|\|
	(db->db_blkid == DMU_BONUS_BLKID &&
	!DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))))
	err = 0;

	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Drops db_mtx and the parent lock specified by dblt and tag before
	* returning.
	*/
	static int
	dbuf_read_impl(dmu_buf_impl_t db, zio_t zio, uint32_t flags,
	db_lock_type_t dblt, const void *tag)
	{
	dnode_t *dn;
	zbookmark_phys_t zb;
	uint32_t aflags = ARC_FLAG_NOWAIT;
	int err, zio_flags;

	err = zio_flags = 0;
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db_state == DB_UNCACHED);
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db_parent == NULL \|\|
	RW_LOCK_HELD(&db->db_parent->db_rwlock));

	if (db->db_blkid == DMU_BONUS_BLKID) {
	err = dbuf_read_bonus(db, dn, flags);
	goto early_unlock;
	}

	err = dbuf_read_hole(db, dn);
	if (err == 0)
	goto early_unlock;

	/*
	* Any attempt to read a redacted block should result in an error. This
	* will never happen under normal conditions, but can be useful for
	* debugging purposes.
	*/
	if (BP_IS_REDACTED(db->db_blkptr)) {
	ASSERT(dsl_dataset_feature_is_active(
	db->db_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));
	err = SET_ERROR(EIO);
	goto early_unlock;
	}

	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);

	/*
	* All bps of an encrypted os should have the encryption bit set.
	* If this is not true it indicates tampering and we report an error.
	*/
	if (db->db_objset->os_encrypted && !BP_USES_CRYPT(db->db_blkptr)) {
	spa_log_error(db->db_objset->os_spa, &zb);
	zfs_panic_recover("unencrypted block in encrypted "
	"object set %llu", dmu_objset_id(db->db_objset));
	err = SET_ERROR(EIO);
	goto early_unlock;
	}

	err = dbuf_read_verify_dnode_crypt(db, flags);
	if (err != 0)
	goto early_unlock;

	DB_DNODE_EXIT(db);

	db->db_state = DB_READ;
	DTRACE_SET_STATE(db, "read issued");
	mutex_exit(&db->db_mtx);

	if (dbuf_is_l2cacheable(db))
	aflags \|= ARC_FLAG_L2CACHE;

	dbuf_add_ref(db, NULL);

	zio_flags = (flags & DB_RF_CANFAIL) ?
	ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED;

	if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr))
	zio_flags \|= ZIO_FLAG_RAW;
	/*
	* The zio layer will copy the provided blkptr later, but we need to
	* do this now so that we can release the parent's rwlock. We have to
	* do that now so that if dbuf_read_done is called synchronously (on
	* an l1 cache hit) we don't acquire the db_mtx while holding the
	* parent's rwlock, which would be a lock ordering violation.
	*/
	blkptr_t bp = *db->db_blkptr;
	dmu_buf_unlock_parent(db, dblt, tag);
	(void) arc_read(zio, db->db_objset->os_spa, &bp,
	dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags,
	&aflags, &zb);
	return (err);
	early_unlock:
	DB_DNODE_EXIT(db);
	mutex_exit(&db->db_mtx);
	dmu_buf_unlock_parent(db, dblt, tag);
	return (err);
	}

	/*
	* This is our just-in-time copy function. It makes a copy of buffers that
	* have been modified in a previous transaction group before we access them in
	* the current active group.
	*
	* This function is used in three places: when we are dirtying a buffer for the
	* first time in a txg, when we are freeing a range in a dnode that includes
	* this buffer, and when we are accessing a buffer which was received compressed
	* and later referenced in a WRITE_BYREF record.
	*
	* Note that when we are called from dbuf_free_range() we do not put a hold on
	* the buffer, we just traverse the active dbuf list for the dnode.
	*/
	static void
	dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg)
	{
	dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db.db_data != NULL);
	ASSERT(db->db_level == 0);
	ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT);

	if (dr == NULL \|\|
	(dr->dt.dl.dr_data !=
	((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf)))
	return;

	/*
	* If the last dirty record for this dbuf has not yet synced
	* and its referencing the dbuf data, either:
	* reset the reference to point to a new copy,
	* or (if there a no active holders)
	* just null out the current db_data pointer.
	*/
	ASSERT3U(dr->dr_txg, >=, txg - 2);
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dnode_t *dn = DB_DNODE(db);
	int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP);
	arc_space_consume(bonuslen, ARC_SPACE_BONUS);
	memcpy(dr->dt.dl.dr_data, db->db.db_data, bonuslen);
	} else if (zfs_refcount_count(&db->db_holds) > db->db_dirtycnt) {
	dnode_t *dn = DB_DNODE(db);
	int size = arc_buf_size(db->db_buf);
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	spa_t *spa = db->db_objset->os_spa;
	enum zio_compress compress_type =
	arc_get_compression(db->db_buf);
	uint8_t complevel = arc_get_complevel(db->db_buf);

	if (arc_is_encrypted(db->db_buf)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(db->db_buf, &byteorder, salt,
	iv, mac);
	dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db,
	dmu_objset_id(dn->dn_objset), byteorder, salt, iv,
	mac, dn->dn_type, size, arc_buf_lsize(db->db_buf),
	compress_type, complevel);
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	ASSERT3U(type, ==, ARC_BUFC_DATA);
	dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db,
	size, arc_buf_lsize(db->db_buf), compress_type,
	complevel);
	} else {
	dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size);
	}
	memcpy(dr->dt.dl.dr_data->b_data, db->db.db_data, size);
	} else {
	db->db_buf = NULL;
	dbuf_clear_data(db);
	}
	}

	int
	dbuf_read(dmu_buf_impl_t db, zio_t zio, uint32_t flags)
	{
	int err = 0;
	boolean_t prefetch;
	dnode_t *dn;

	/*
	* We don't have to hold the mutex to check db_state because it
	* can't be freed while we have a hold on the buffer.
	*/
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	if (db->db_state == DB_NOFILL)
	return (SET_ERROR(EIO));

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	(flags & DB_RF_NOPREFETCH) == 0 && dn != NULL &&
	DBUF_IS_CACHEABLE(db);

	mutex_enter(&db->db_mtx);
	if (db->db_state == DB_CACHED) {
	spa_t *spa = dn->dn_objset->os_spa;

	/*
	* Ensure that this block's dnode has been decrypted if
	* the caller has requested decrypted data.
	*/
	err = dbuf_read_verify_dnode_crypt(db, flags);

	/*
	* If the arc buf is compressed or encrypted and the caller
	* requested uncompressed data, we need to untransform it
	* before returning. We also call arc_untransform() on any
	* unauthenticated blocks, which will verify their MAC if
	* the key is now available.
	*/
	if (err == 0 && db->db_buf != NULL &&
	(flags & DB_RF_NO_DECRYPT) == 0 &&
	(arc_is_encrypted(db->db_buf) \|\|
	arc_is_unauthenticated(db->db_buf) \|\|
	arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) {
	zbookmark_phys_t zb;

	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);
	dbuf_fix_old_data(db, spa_syncing_txg(spa));
	err = arc_untransform(db->db_buf, spa, &zb, B_FALSE);
	dbuf_set_data(db, db->db_buf);
	}
	mutex_exit(&db->db_mtx);
	if (err == 0 && prefetch) {
	dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
	B_FALSE, flags & DB_RF_HAVESTRUCT);
	}
	DB_DNODE_EXIT(db);
	DBUF_STAT_BUMP(hash_hits);
	} else if (db->db_state == DB_UNCACHED) {
	spa_t *spa = dn->dn_objset->os_spa;
	boolean_t need_wait = B_FALSE;

	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);

	if (zio == NULL &&
	db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)) {
	zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	need_wait = B_TRUE;
	}
	err = dbuf_read_impl(db, zio, flags, dblt, FTAG);
	/*
	* dbuf_read_impl has dropped db_mtx and our parent's rwlock
	* for us
	*/
	if (!err && prefetch) {
	dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
	db->db_state != DB_CACHED,
	flags & DB_RF_HAVESTRUCT);
	}

	DB_DNODE_EXIT(db);
	DBUF_STAT_BUMP(hash_misses);

	/*
	* If we created a zio_root we must execute it to avoid
	* leaking it, even if it isn't attached to any work due
	* to an error in dbuf_read_impl().
	*/
	if (need_wait) {
	if (err == 0)
	err = zio_wait(zio);
	else
	VERIFY0(zio_wait(zio));
	}
	} else {
	/*
	* Another reader came in while the dbuf was in flight
	* between UNCACHED and CACHED. Either a writer will finish
	* writing the buffer (sending the dbuf to CACHED) or the
	* first reader's request will reach the read_done callback
	* and send the dbuf to CACHED. Otherwise, a failure
	* occurred and the dbuf went to UNCACHED.
	*/
	mutex_exit(&db->db_mtx);
	if (prefetch) {
	dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE,
	B_TRUE, flags & DB_RF_HAVESTRUCT);
	}
	DB_DNODE_EXIT(db);
	DBUF_STAT_BUMP(hash_misses);

	/* Skip the wait per the caller's request. */
	if ((flags & DB_RF_NEVERWAIT) == 0) {
	mutex_enter(&db->db_mtx);
	while (db->db_state == DB_READ \|\|
	db->db_state == DB_FILL) {
	ASSERT(db->db_state == DB_READ \|\|
	(flags & DB_RF_HAVESTRUCT) == 0);
	DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *,
	db, zio_t *, zio);
	cv_wait(&db->db_changed, &db->db_mtx);
	}
	if (db->db_state == DB_UNCACHED)
	err = SET_ERROR(EIO);
	mutex_exit(&db->db_mtx);
	}
	}

	return (err);
	}

	static void
	dbuf_noread(dmu_buf_impl_t *db)
	{
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	mutex_enter(&db->db_mtx);
	while (db->db_state == DB_READ \|\| db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);
	if (db->db_state == DB_UNCACHED) {
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	dbuf_set_data(db, dbuf_alloc_arcbuf(db));
	db->db_state = DB_FILL;
	DTRACE_SET_STATE(db, "assigning filled buffer");
	} else if (db->db_state == DB_NOFILL) {
	dbuf_clear_data(db);
	} else {
	ASSERT3U(db->db_state, ==, DB_CACHED);
	}
	mutex_exit(&db->db_mtx);
	}

	void
	dbuf_unoverride(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	blkptr_t *bp = &dr->dt.dl.dr_overridden_by;
	uint64_t txg = dr->dr_txg;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	/*
	* This assert is valid because dmu_sync() expects to be called by
	* a zilog's get_data while holding a range lock. This call only
	* comes from dbuf_dirty() callers who must also hold a range lock.
	*/
	ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC);
	ASSERT(db->db_level == 0);

	if (db->db_blkid == DMU_BONUS_BLKID \|\|
	dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN)
	return;

	ASSERT(db->db_data_pending != dr);

	/* free this block */
	if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite)
	zio_free(db->db_objset->os_spa, txg, bp);

	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	dr->dt.dl.dr_nopwrite = B_FALSE;
	dr->dt.dl.dr_has_raw_params = B_FALSE;

	/*
	* Release the already-written buffer, so we leave it in
	* a consistent dirty state. Note that all callers are
	* modifying the buffer, so they will immediately do
	* another (redundant) arc_release(). Therefore, leave
	* the buf thawed to save the effort of freezing &
	* immediately re-thawing it.
	*/
	arc_release(dr->dt.dl.dr_data, db);
	}

	/*
	* Evict (if its unreferenced) or clear (if its referenced) any level-0
	* data blocks in the free range, so that any future readers will find
	* empty blocks.
	*/
	void
	dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db_search;
	dmu_buf_impl_t db, db_next;
	uint64_t txg = tx->tx_txg;
	avl_index_t where;
	dbuf_dirty_record_t *dr;

	if (end_blkid > dn->dn_maxblkid &&
	!(start_blkid == DMU_SPILL_BLKID \|\| end_blkid == DMU_SPILL_BLKID))
	end_blkid = dn->dn_maxblkid;
	dprintf_dnode(dn, "start=%llu end=%llu\n", (u_longlong_t)start_blkid,
	(u_longlong_t)end_blkid);

	db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP);
	db_search->db_level = 0;
	db_search->db_blkid = start_blkid;
	db_search->db_state = DB_SEARCH;

	mutex_enter(&dn->dn_dbufs_mtx);
	db = avl_find(&dn->dn_dbufs, db_search, &where);
	ASSERT3P(db, ==, NULL);

	db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER);

	for (; db != NULL; db = db_next) {
	db_next = AVL_NEXT(&dn->dn_dbufs, db);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	if (db->db_level != 0 \|\| db->db_blkid > end_blkid) {
	break;
	}
	ASSERT3U(db->db_blkid, >=, start_blkid);

	/* found a level 0 buffer in the range */
	mutex_enter(&db->db_mtx);
	if (dbuf_undirty(db, tx)) {
	/* mutex has been dropped and dbuf destroyed */
	continue;
	}

	if (db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL \|\|
	db->db_state == DB_EVICTING) {
	ASSERT(db->db.db_data == NULL);
	mutex_exit(&db->db_mtx);
	continue;
	}
	if (db->db_state == DB_READ \|\| db->db_state == DB_FILL) {
	/* will be handled in dbuf_read_done or dbuf_rele */
	db->db_freed_in_flight = TRUE;
	mutex_exit(&db->db_mtx);
	continue;
	}
	if (zfs_refcount_count(&db->db_holds) == 0) {
	ASSERT(db->db_buf);
	dbuf_destroy(db);
	continue;
	}
	/* The dbuf is referenced */

	dr = list_head(&db->db_dirty_records);
	if (dr != NULL) {
	if (dr->dr_txg == txg) {
	/*
	* This buffer is "in-use", re-adjust the file
	* size to reflect that this buffer may
	* contain new data when we sync.
	*/
	if (db->db_blkid != DMU_SPILL_BLKID &&
	db->db_blkid > dn->dn_maxblkid)
	dn->dn_maxblkid = db->db_blkid;
	dbuf_unoverride(dr);
	} else {
	/*
	* This dbuf is not dirty in the open context.
	* Either uncache it (if its not referenced in
	* the open context) or reset its contents to
	* empty.
	*/
	dbuf_fix_old_data(db, txg);
	}
	}
	/* clear the contents if its cached */
	if (db->db_state == DB_CACHED) {
	ASSERT(db->db.db_data != NULL);
	arc_release(db->db_buf, db);
	rw_enter(&db->db_rwlock, RW_WRITER);
	memset(db->db.db_data, 0, db->db.db_size);
	rw_exit(&db->db_rwlock);
	arc_buf_freeze(db->db_buf);
	}

	mutex_exit(&db->db_mtx);
	}

	mutex_exit(&dn->dn_dbufs_mtx);
	kmem_free(db_search, sizeof (dmu_buf_impl_t));
	}

	void
	dbuf_new_size(dmu_buf_impl_t db, int size, dmu_tx_t tx)
	{
	arc_buf_t buf, old_buf;
	dbuf_dirty_record_t *dr;
	int osize = db->db.db_size;
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	dnode_t *dn;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	/*
	* XXX we should be doing a dbuf_read, checking the return
	* value and returning that up to our callers
	*/
	dmu_buf_will_dirty(&db->db, tx);

	/* create the data buffer for the new block */
	buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size);

	/* copy old block data to the new block */
	old_buf = db->db_buf;
	memcpy(buf->b_data, old_buf->b_data, MIN(osize, size));
	/* zero the remainder */
	if (size > osize)
	memset((uint8_t *)buf->b_data + osize, 0, size - osize);

	mutex_enter(&db->db_mtx);
	dbuf_set_data(db, buf);
	arc_buf_destroy(old_buf, db);
	db->db.db_size = size;

	dr = list_head(&db->db_dirty_records);
	/* dirty record added by dmu_buf_will_dirty() */
	VERIFY(dr != NULL);
	if (db->db_level == 0)
	dr->dt.dl.dr_data = buf;
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	ASSERT3U(dr->dr_accounted, ==, osize);
	dr->dr_accounted = size;
	mutex_exit(&db->db_mtx);

	dmu_objset_willuse_space(dn->dn_objset, size - osize, tx);
	DB_DNODE_EXIT(db);
	}

	void
	dbuf_release_bp(dmu_buf_impl_t *db)
	{
	objset_t *os __maybe_unused = db->db_objset;

	ASSERT(dsl_pool_sync_context(dmu_objset_pool(os)));
	ASSERT(arc_released(os->os_phys_buf) \|\|
	list_link_active(&os->os_dsl_dataset->ds_synced_link));
	ASSERT(db->db_parent == NULL \|\| arc_released(db->db_parent->db_buf));

	(void) arc_release(db->db_buf, db);
	}

	/*
	* We already have a dirty record for this TXG, and we are being
	* dirtied again.
	*/
	static void
	dbuf_redirty(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) {
	/*
	* If this buffer has already been written out,
	* we now need to reset its state.
	*/
	dbuf_unoverride(dr);
	if (db->db.db_object != DMU_META_DNODE_OBJECT &&
	db->db_state != DB_NOFILL) {
	/* Already released on initial dirty, so just thaw. */
	ASSERT(arc_released(db->db_buf));
	arc_buf_thaw(db->db_buf);
	}
	}
	}

	dbuf_dirty_record_t *
	dbuf_dirty_lightweight(dnode_t dn, uint64_t blkid, dmu_tx_t tx)
	{
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	IMPLY(dn->dn_objset->os_raw_receive, dn->dn_maxblkid >= blkid);
	dnode_new_blkid(dn, blkid, tx, B_TRUE, B_FALSE);
	ASSERT(dn->dn_maxblkid >= blkid);

	dbuf_dirty_record_t dr = kmem_zalloc(sizeof (dr), KM_SLEEP);
	list_link_init(&dr->dr_dirty_node);
	list_link_init(&dr->dr_dbuf_node);
	dr->dr_dnode = dn;
	dr->dr_txg = tx->tx_txg;
	dr->dt.dll.dr_blkid = blkid;
	dr->dr_accounted = dn->dn_datablksz;

	/*
	* There should not be any dbuf for the block that we're dirtying.
	* Otherwise the buffer contents could be inconsistent between the
	* dbuf and the lightweight dirty record.
	*/
	ASSERT3P(NULL, ==, dbuf_find(dn->dn_objset, dn->dn_object, 0, blkid));

	mutex_enter(&dn->dn_mtx);
	int txgoff = tx->tx_txg & TXG_MASK;
	if (dn->dn_free_ranges[txgoff] != NULL) {
	range_tree_clear(dn->dn_free_ranges[txgoff], blkid, 1);
	}

	if (dn->dn_nlevels == 1) {
	ASSERT3U(blkid, <, dn->dn_nblkptr);
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_setdirty(dn, tx);
	} else {
	mutex_exit(&dn->dn_mtx);

	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	dmu_buf_impl_t *parent_db = dbuf_hold_level(dn,
	1, blkid >> epbs, FTAG);
	rw_exit(&dn->dn_struct_rwlock);
	if (parent_db == NULL) {
	kmem_free(dr, sizeof (*dr));
	return (NULL);
	}
	int err = dbuf_read(parent_db, NULL,
	(DB_RF_NOPREFETCH \| DB_RF_CANFAIL));
	if (err != 0) {
	dbuf_rele(parent_db, FTAG);
	kmem_free(dr, sizeof (*dr));
	return (NULL);
	}

	dbuf_dirty_record_t *parent_dr = dbuf_dirty(parent_db, tx);
	dbuf_rele(parent_db, FTAG);
	mutex_enter(&parent_dr->dt.di.dr_mtx);
	ASSERT3U(parent_dr->dr_txg, ==, tx->tx_txg);
	list_insert_tail(&parent_dr->dt.di.dr_children, dr);
	mutex_exit(&parent_dr->dt.di.dr_mtx);
	dr->dr_parent = parent_dr;
	}

	dmu_objset_willuse_space(dn->dn_objset, dr->dr_accounted, tx);

	return (dr);
	}

	dbuf_dirty_record_t *
	dbuf_dirty(dmu_buf_impl_t db, dmu_tx_t tx)
	{
	dnode_t *dn;
	objset_t *os;
	dbuf_dirty_record_t dr, dr_next, *dr_head;
	int txgoff = tx->tx_txg & TXG_MASK;
	boolean_t drop_struct_rwlock = B_FALSE;

	ASSERT(tx->tx_txg != 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	DMU_TX_DIRTY_BUF(tx, db);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	/*
	* Shouldn't dirty a regular buffer in syncing context. Private
	* objects may be dirtied in syncing context, but only if they
	* were already pre-dirtied in open context.
	*/
	#ifdef ZFS_DEBUG
	if (dn->dn_objset->os_dsl_dataset != NULL) {
	rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock,
	RW_READER, FTAG);
	}
	ASSERT(!dmu_tx_is_syncing(tx) \|\|
	BP_IS_HOLE(dn->dn_objset->os_rootbp) \|\|
	DMU_OBJECT_IS_SPECIAL(dn->dn_object) \|\|
	dn->dn_objset->os_dsl_dataset == NULL);
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG);
	#endif
	/*
	* We make this assert for private objects as well, but after we
	* check if we're already dirty. They are allowed to re-dirty
	* in syncing context.
	*/
	ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT \|\|
	dn->dn_dirtyctx == DN_UNDIRTIED \|\| dn->dn_dirtyctx ==
	(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));

	mutex_enter(&db->db_mtx);
	/*
	* XXX make this true for indirects too? The problem is that
	* transactions created with dmu_tx_create_assigned() from
	* syncing context don't bother holding ahead.
	*/
	ASSERT(db->db_level != 0 \|\|
	db->db_state == DB_CACHED \|\| db->db_state == DB_FILL \|\|
	db->db_state == DB_NOFILL);

	mutex_enter(&dn->dn_mtx);
	dnode_set_dirtyctx(dn, tx, db);
	if (tx->tx_txg > dn->dn_dirty_txg)
	dn->dn_dirty_txg = tx->tx_txg;
	mutex_exit(&dn->dn_mtx);

	if (db->db_blkid == DMU_SPILL_BLKID)
	dn->dn_have_spill = B_TRUE;

	/*
	* If this buffer is already dirty, we're done.
	*/
	dr_head = list_head(&db->db_dirty_records);
	ASSERT(dr_head == NULL \|\| dr_head->dr_txg <= tx->tx_txg \|\|
	db->db.db_object == DMU_META_DNODE_OBJECT);
	dr_next = dbuf_find_dirty_lte(db, tx->tx_txg);
	if (dr_next && dr_next->dr_txg == tx->tx_txg) {
	DB_DNODE_EXIT(db);

	dbuf_redirty(dr_next);
	mutex_exit(&db->db_mtx);
	return (dr_next);
	}

	/*
	* Only valid if not already dirty.
	*/
	ASSERT(dn->dn_object == 0 \|\|
	dn->dn_dirtyctx == DN_UNDIRTIED \|\| dn->dn_dirtyctx ==
	(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));

	ASSERT3U(dn->dn_nlevels, >, db->db_level);

	/*
	* We should only be dirtying in syncing context if it's the
	* mos or we're initializing the os or it's a special object.
	* However, we are allowed to dirty in syncing context provided
	* we already dirtied it in open context. Hence we must make
	* this assertion only if we're not already dirty.
	*/
	os = dn->dn_objset;
	VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa));
	#ifdef ZFS_DEBUG
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG);
	ASSERT(!dmu_tx_is_syncing(tx) \|\| DMU_OBJECT_IS_SPECIAL(dn->dn_object) \|\|
	os->os_dsl_dataset == NULL \|\| BP_IS_HOLE(os->os_rootbp));
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG);
	#endif
	ASSERT(db->db.db_size != 0);

	dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);

	if (db->db_blkid != DMU_BONUS_BLKID) {
	dmu_objset_willuse_space(os, db->db.db_size, tx);
	}

	/*
	* If this buffer is dirty in an old transaction group we need
	* to make a copy of it so that the changes we make in this
	* transaction group won't leak out when we sync the older txg.
	*/
	dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP);
	list_link_init(&dr->dr_dirty_node);
	list_link_init(&dr->dr_dbuf_node);
	dr->dr_dnode = dn;
	if (db->db_level == 0) {
	void *data_old = db->db_buf;

	if (db->db_state != DB_NOFILL) {
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dbuf_fix_old_data(db, tx->tx_txg);
	data_old = db->db.db_data;
	} else if (db->db.db_object != DMU_META_DNODE_OBJECT) {
	/*
	* Release the data buffer from the cache so
	* that we can modify it without impacting
	* possible other users of this cached data
	* block. Note that indirect blocks and
	* private objects are not released until the
	* syncing state (since they are only modified
	* then).
	*/
	arc_release(db->db_buf, db);
	dbuf_fix_old_data(db, tx->tx_txg);
	data_old = db->db_buf;
	}
	ASSERT(data_old != NULL);
	}
	dr->dt.dl.dr_data = data_old;
	} else {
	mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL);
	list_create(&dr->dt.di.dr_children,
	sizeof (dbuf_dirty_record_t),
	offsetof(dbuf_dirty_record_t, dr_dirty_node));
	}
	if (db->db_blkid != DMU_BONUS_BLKID)
	dr->dr_accounted = db->db.db_size;
	dr->dr_dbuf = db;
	dr->dr_txg = tx->tx_txg;
	list_insert_before(&db->db_dirty_records, dr_next, dr);

	/*
	* We could have been freed_in_flight between the dbuf_noread
	* and dbuf_dirty. We win, as though the dbuf_noread() had
	* happened after the free.
	*/
	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	db->db_blkid != DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (dn->dn_free_ranges[txgoff] != NULL) {
	range_tree_clear(dn->dn_free_ranges[txgoff],
	db->db_blkid, 1);
	}
	mutex_exit(&dn->dn_mtx);
	db->db_freed_in_flight = FALSE;
	}

	/*
	* This buffer is now part of this txg
	*/
	dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg);
	db->db_dirtycnt += 1;
	ASSERT3U(db->db_dirtycnt, <=, 3);

	mutex_exit(&db->db_mtx);

	if (db->db_blkid == DMU_BONUS_BLKID \|\|
	db->db_blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	dnode_setdirty(dn, tx);
	DB_DNODE_EXIT(db);
	return (dr);
	}

	if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	drop_struct_rwlock = B_TRUE;
	}

	/*
	* If we are overwriting a dedup BP, then unless it is snapshotted,
	* when we get to syncing context we will need to decrement its
	* refcount in the DDT. Prefetch the relevant DDT block so that
	* syncing context won't have to wait for the i/o.
	*/
	if (db->db_blkptr != NULL) {
	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);
	ddt_prefetch(os->os_spa, db->db_blkptr);
	dmu_buf_unlock_parent(db, dblt, FTAG);
	}

	/*
	* We need to hold the dn_struct_rwlock to make this assertion,
	* because it protects dn_phys / dn_next_nlevels from changing.
	*/
	ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) \|\|
	dn->dn_phys->dn_nlevels > db->db_level \|\|
	dn->dn_next_nlevels[txgoff] > db->db_level \|\|
	dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level \|\|
	dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level);


	if (db->db_level == 0) {
	ASSERT(!db->db_objset->os_raw_receive \|\|
	dn->dn_maxblkid >= db->db_blkid);
	dnode_new_blkid(dn, db->db_blkid, tx,
	drop_struct_rwlock, B_FALSE);
	ASSERT(dn->dn_maxblkid >= db->db_blkid);
	}

	if (db->db_level+1 < dn->dn_nlevels) {
	dmu_buf_impl_t *parent = db->db_parent;
	dbuf_dirty_record_t *di;
	int parent_held = FALSE;

	if (db->db_parent == NULL \|\| db->db_parent == dn->dn_dbuf) {
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	parent = dbuf_hold_level(dn, db->db_level + 1,
	db->db_blkid >> epbs, FTAG);
	ASSERT(parent != NULL);
	parent_held = TRUE;
	}
	if (drop_struct_rwlock)
	rw_exit(&dn->dn_struct_rwlock);
	ASSERT3U(db->db_level + 1, ==, parent->db_level);
	di = dbuf_dirty(parent, tx);
	if (parent_held)
	dbuf_rele(parent, FTAG);

	mutex_enter(&db->db_mtx);
	/*
	* Since we've dropped the mutex, it's possible that
	* dbuf_undirty() might have changed this out from under us.
	*/
	if (list_head(&db->db_dirty_records) == dr \|\|
	dn->dn_object == DMU_META_DNODE_OBJECT) {
	mutex_enter(&di->dt.di.dr_mtx);
	ASSERT3U(di->dr_txg, ==, tx->tx_txg);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&di->dt.di.dr_children, dr);
	mutex_exit(&di->dt.di.dr_mtx);
	dr->dr_parent = di;
	}
	mutex_exit(&db->db_mtx);
	} else {
	ASSERT(db->db_level + 1 == dn->dn_nlevels);
	ASSERT(db->db_blkid < dn->dn_nblkptr);
	ASSERT(db->db_parent == NULL \|\| db->db_parent == dn->dn_dbuf);
	mutex_enter(&dn->dn_mtx);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	if (drop_struct_rwlock)
	rw_exit(&dn->dn_struct_rwlock);
	}

	dnode_setdirty(dn, tx);
	DB_DNODE_EXIT(db);
	return (dr);
	}

	static void
	dbuf_undirty_bonus(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;

	if (dr->dt.dl.dr_data != db->db.db_data) {
	struct dnode *dn = dr->dr_dnode;
	int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);

	kmem_free(dr->dt.dl.dr_data, max_bonuslen);
	arc_space_return(max_bonuslen, ARC_SPACE_BONUS);
	}
	db->db_data_pending = NULL;
	ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
	list_remove(&db->db_dirty_records, dr);
	if (dr->dr_dbuf->db_level != 0) {
	mutex_destroy(&dr->dt.di.dr_mtx);
	list_destroy(&dr->dt.di.dr_children);
	}
	kmem_free(dr, sizeof (dbuf_dirty_record_t));
	ASSERT3U(db->db_dirtycnt, >, 0);
	db->db_dirtycnt -= 1;
	}

	/*
	* Undirty a buffer in the transaction group referenced by the given
	* transaction. Return whether this evicted the dbuf.
	*/
	static boolean_t
	dbuf_undirty(dmu_buf_impl_t db, dmu_tx_t tx)
	{
	uint64_t txg = tx->tx_txg;

	ASSERT(txg != 0);

	/*
	* Due to our use of dn_nlevels below, this can only be called
	* in open context, unless we are operating on the MOS.
	* From syncing context, dn_nlevels may be different from the
	* dn_nlevels used when dbuf was dirtied.
	*/
	ASSERT(db->db_objset ==
	dmu_objset_pool(db->db_objset)->dp_meta_objset \|\|
	txg != spa_syncing_txg(dmu_objset_spa(db->db_objset)));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT0(db->db_level);
	ASSERT(MUTEX_HELD(&db->db_mtx));

	/*
	* If this buffer is not dirty, we're done.
	*/
	dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, txg);
	if (dr == NULL)
	return (B_FALSE);
	ASSERT(dr->dr_dbuf == db);

	dnode_t *dn = dr->dr_dnode;

	dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);

	ASSERT(db->db.db_size != 0);

	dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset),
	dr->dr_accounted, txg);

	list_remove(&db->db_dirty_records, dr);

	/*
	* Note that there are three places in dbuf_dirty()
	* where this dirty record may be put on a list.
	* Make sure to do a list_remove corresponding to
	* every one of those list_insert calls.
	*/
	if (dr->dr_parent) {
	mutex_enter(&dr->dr_parent->dt.di.dr_mtx);
	list_remove(&dr->dr_parent->dt.di.dr_children, dr);
	mutex_exit(&dr->dr_parent->dt.di.dr_mtx);
	} else if (db->db_blkid == DMU_SPILL_BLKID \|\|
	db->db_level + 1 == dn->dn_nlevels) {
	ASSERT(db->db_blkptr == NULL \|\| db->db_parent == dn->dn_dbuf);
	mutex_enter(&dn->dn_mtx);
	list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr);
	mutex_exit(&dn->dn_mtx);
	}

	if (db->db_state != DB_NOFILL) {
	dbuf_unoverride(dr);

	ASSERT(db->db_buf != NULL);
	ASSERT(dr->dt.dl.dr_data != NULL);
	if (dr->dt.dl.dr_data != db->db_buf)
	arc_buf_destroy(dr->dt.dl.dr_data, db);
	}

	kmem_free(dr, sizeof (dbuf_dirty_record_t));

	ASSERT(db->db_dirtycnt > 0);
	db->db_dirtycnt -= 1;

	if (zfs_refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) {
	ASSERT(db->db_state == DB_NOFILL \|\| arc_released(db->db_buf));
	dbuf_destroy(db);
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static void
	dmu_buf_will_dirty_impl(dmu_buf_t db_fake, int flags, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	ASSERT(tx->tx_txg != 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	/*
	* Quick check for dirtiness. For already dirty blocks, this
	* reduces runtime of this function by >90%, and overall performance
	* by 50% for some workloads (e.g. file deletion with indirect blocks
	* cached).
	*/
	mutex_enter(&db->db_mtx);

	if (db->db_state == DB_CACHED) {
	dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg);
	/*
	* It's possible that it is already dirty but not cached,
	* because there are some calls to dbuf_dirty() that don't
	* go through dmu_buf_will_dirty().
	*/
	if (dr != NULL) {
	/* This dbuf is already dirty and cached. */
	dbuf_redirty(dr);
	mutex_exit(&db->db_mtx);
	return;
	}
	}
	mutex_exit(&db->db_mtx);

	DB_DNODE_ENTER(db);
	if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock))
	flags \|= DB_RF_HAVESTRUCT;
	DB_DNODE_EXIT(db);
	(void) dbuf_read(db, NULL, flags);
	(void) dbuf_dirty(db, tx);
	}

	void
	dmu_buf_will_dirty(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_will_dirty_impl(db_fake,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH, tx);
	}

	boolean_t
	dmu_buf_is_dirty(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dbuf_dirty_record_t *dr;

	mutex_enter(&db->db_mtx);
	dr = dbuf_find_dirty_eq(db, tx->tx_txg);
	mutex_exit(&db->db_mtx);
	return (dr != NULL);
	}

	void
	dmu_buf_will_not_fill(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	db->db_state = DB_NOFILL;
	DTRACE_SET_STATE(db, "allocating NOFILL buffer");
	dmu_buf_will_fill(db_fake, tx);
	}

	void
	dmu_buf_will_fill(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(tx->tx_txg != 0);
	ASSERT(db->db_level == 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT \|\|
	dmu_tx_private_ok(tx));

	dbuf_noread(db);
	(void) dbuf_dirty(db, tx);
	}

	/*
	* This function is effectively the same as dmu_buf_will_dirty(), but
	* indicates the caller expects raw encrypted data in the db, and provides
	* the crypt params (byteorder, salt, iv, mac) which should be stored in the
	* blkptr_t when this dbuf is written. This is only used for blocks of
	* dnodes, during raw receive.
	*/
	void
	dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t mac, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dbuf_dirty_record_t *dr;

	/*
	* dr_has_raw_params is only processed for blocks of dnodes
	* (see dbuf_sync_dnode_leaf_crypt()).
	*/
	ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
	ASSERT3U(db->db_level, ==, 0);
	ASSERT(db->db_objset->os_raw_receive);

	dmu_buf_will_dirty_impl(db_fake,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH \| DB_RF_NO_DECRYPT, tx);

	dr = dbuf_find_dirty_eq(db, tx->tx_txg);

	ASSERT3P(dr, !=, NULL);

	dr->dt.dl.dr_has_raw_params = B_TRUE;
	dr->dt.dl.dr_byteorder = byteorder;
	memcpy(dr->dt.dl.dr_salt, salt, ZIO_DATA_SALT_LEN);
	memcpy(dr->dt.dl.dr_iv, iv, ZIO_DATA_IV_LEN);
	memcpy(dr->dt.dl.dr_mac, mac, ZIO_DATA_MAC_LEN);
	}

	static void
	dbuf_override_impl(dmu_buf_impl_t db, const blkptr_t bp, dmu_tx_t *tx)
	{
	struct dirty_leaf *dl;
	dbuf_dirty_record_t *dr;

	dr = list_head(&db->db_dirty_records);
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	dl = &dr->dt.dl;
	dl->dr_overridden_by = *bp;
	dl->dr_override_state = DR_OVERRIDDEN;
	dl->dr_overridden_by.blk_birth = dr->dr_txg;
	}

	void
	dmu_buf_fill_done(dmu_buf_t dbuf, dmu_tx_t tx)
	{
	(void) tx;
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	dbuf_states_t old_state;
	mutex_enter(&db->db_mtx);
	DBUF_VERIFY(db);

	old_state = db->db_state;
	db->db_state = DB_CACHED;
	if (old_state == DB_FILL) {
	if (db->db_level == 0 && db->db_freed_in_flight) {
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	/* we were freed while filling */
	/* XXX dbuf_undirty? */
	memset(db->db.db_data, 0, db->db.db_size);
	db->db_freed_in_flight = FALSE;
	DTRACE_SET_STATE(db,
	"fill done handling freed in flight");
	} else {
	DTRACE_SET_STATE(db, "fill done");
	}
	cv_broadcast(&db->db_changed);
	}
	mutex_exit(&db->db_mtx);
	}

	void
	dmu_buf_write_embedded(dmu_buf_t dbuf, void data,
	bp_embedded_type_t etype, enum zio_compress comp,
	int uncompressed_size, int compressed_size, int byteorder,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	struct dirty_leaf *dl;
	dmu_object_type_t type;
	dbuf_dirty_record_t *dr;

	if (etype == BP_EMBEDDED_TYPE_DATA) {
	ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset),
	SPA_FEATURE_EMBEDDED_DATA));
	}

	DB_DNODE_ENTER(db);
	type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	ASSERT0(db->db_level);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	dmu_buf_will_not_fill(dbuf, tx);

	dr = list_head(&db->db_dirty_records);
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	dl = &dr->dt.dl;
	encode_embedded_bp_compressed(&dl->dr_overridden_by,
	data, comp, uncompressed_size, compressed_size);
	BPE_SET_ETYPE(&dl->dr_overridden_by, etype);
	BP_SET_TYPE(&dl->dr_overridden_by, type);
	BP_SET_LEVEL(&dl->dr_overridden_by, 0);
	BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder);

	dl->dr_override_state = DR_OVERRIDDEN;
	dl->dr_overridden_by.blk_birth = dr->dr_txg;
	}

	void
	dmu_buf_redact(dmu_buf_t dbuf, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	dmu_object_type_t type;
	ASSERT(dsl_dataset_feature_is_active(db->db_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));

	DB_DNODE_ENTER(db);
	type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	ASSERT0(db->db_level);
	dmu_buf_will_not_fill(dbuf, tx);

	blkptr_t bp = { { { {0} } } };
	BP_SET_TYPE(&bp, type);
	BP_SET_LEVEL(&bp, 0);
	BP_SET_BIRTH(&bp, tx->tx_txg, 0);
	BP_SET_REDACTED(&bp);
	BPE_SET_LSIZE(&bp, dbuf->db_size);

	dbuf_override_impl(db, &bp, tx);
	}

	/*
	* Directly assign a provided arc buf to a given dbuf if it's not referenced
	* by anybody except our caller. Otherwise copy arcbuf's contents to dbuf.
	*/
	void
	dbuf_assign_arcbuf(dmu_buf_impl_t db, arc_buf_t buf, dmu_tx_t *tx)
	{
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(db->db_level == 0);
	ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf));
	ASSERT(buf != NULL);
	ASSERT3U(arc_buf_lsize(buf), ==, db->db.db_size);
	ASSERT(tx->tx_txg != 0);

	arc_return_buf(buf, db);
	ASSERT(arc_released(buf));

	mutex_enter(&db->db_mtx);

	while (db->db_state == DB_READ \|\| db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);

	ASSERT(db->db_state == DB_CACHED \|\| db->db_state == DB_UNCACHED);

	if (db->db_state == DB_CACHED &&
	zfs_refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) {
	/*
	* In practice, we will never have a case where we have an
	* encrypted arc buffer while additional holds exist on the
	* dbuf. We don't handle this here so we simply assert that
	* fact instead.
	*/
	ASSERT(!arc_is_encrypted(buf));
	mutex_exit(&db->db_mtx);
	(void) dbuf_dirty(db, tx);
	memcpy(db->db.db_data, buf->b_data, db->db.db_size);
	arc_buf_destroy(buf, db);
	return;
	}

	if (db->db_state == DB_CACHED) {
	dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);

	ASSERT(db->db_buf != NULL);
	if (dr != NULL && dr->dr_txg == tx->tx_txg) {
	ASSERT(dr->dt.dl.dr_data == db->db_buf);

	if (!arc_released(db->db_buf)) {
	ASSERT(dr->dt.dl.dr_override_state ==
	DR_OVERRIDDEN);
	arc_release(db->db_buf, db);
	}
	dr->dt.dl.dr_data = buf;
	arc_buf_destroy(db->db_buf, db);
	} else if (dr == NULL \|\| dr->dt.dl.dr_data != db->db_buf) {
	arc_release(db->db_buf, db);
	arc_buf_destroy(db->db_buf, db);
	}
	db->db_buf = NULL;
	}
	ASSERT(db->db_buf == NULL);
	dbuf_set_data(db, buf);
	db->db_state = DB_FILL;
	DTRACE_SET_STATE(db, "filling assigned arcbuf");
	mutex_exit(&db->db_mtx);
	(void) dbuf_dirty(db, tx);
	dmu_buf_fill_done(&db->db, tx);
	}

	void
	dbuf_destroy(dmu_buf_impl_t *db)
	{
	dnode_t *dn;
	dmu_buf_impl_t *parent = db->db_parent;
	dmu_buf_impl_t *dndb;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(zfs_refcount_is_zero(&db->db_holds));

	if (db->db_buf != NULL) {
	arc_buf_destroy(db->db_buf, db);
	db->db_buf = NULL;
	}

	if (db->db_blkid == DMU_BONUS_BLKID) {
	int slots = DB_DNODE(db)->dn_num_slots;
	int bonuslen = DN_SLOTS_TO_BONUSLEN(slots);
	if (db->db.db_data != NULL) {
	kmem_free(db->db.db_data, bonuslen);
	arc_space_return(bonuslen, ARC_SPACE_BONUS);
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "buffer cleared");
	}
	}

	dbuf_clear_data(db);

	if (multilist_link_active(&db->db_cache_link)) {
	ASSERT(db->db_caching_status == DB_DBUF_CACHE \|\|
	db->db_caching_status == DB_DBUF_METADATA_CACHE);

	multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);
	(void) zfs_refcount_remove_many(
	&dbuf_caches[db->db_caching_status].size,
	db->db.db_size, db);

	if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMPDOWN(metadata_cache_count);
	} else {
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
	db->db.db_size);
	}
	db->db_caching_status = DB_NO_CACHE;
	}

	ASSERT(db->db_state == DB_UNCACHED \|\| db->db_state == DB_NOFILL);
	ASSERT(db->db_data_pending == NULL);
	ASSERT(list_is_empty(&db->db_dirty_records));

	db->db_state = DB_EVICTING;
	DTRACE_SET_STATE(db, "buffer eviction started");
	db->db_blkptr = NULL;

	/*
	* Now that db_state is DB_EVICTING, nobody else can find this via
	* the hash table. We can now drop db_mtx, which allows us to
	* acquire the dn_dbufs_mtx.
	*/
	mutex_exit(&db->db_mtx);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	dndb = dn->dn_dbuf;
	if (db->db_blkid != DMU_BONUS_BLKID) {
	boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx);
	if (needlock)
	mutex_enter_nested(&dn->dn_dbufs_mtx,
	NESTED_SINGLE);
	avl_remove(&dn->dn_dbufs, db);
	membar_producer();
	DB_DNODE_EXIT(db);
	if (needlock)
	mutex_exit(&dn->dn_dbufs_mtx);
	/*
	* Decrementing the dbuf count means that the hold corresponding
	* to the removed dbuf is no longer discounted in dnode_move(),
	* so the dnode cannot be moved until after we release the hold.
	* The membar_producer() ensures visibility of the decremented
	* value in dnode_move(), since DB_DNODE_EXIT doesn't actually
	* release any lock.
	*/
	mutex_enter(&dn->dn_mtx);
	dnode_rele_and_unlock(dn, db, B_TRUE);
	db->db_dnode_handle = NULL;

	dbuf_hash_remove(db);
	} else {
	DB_DNODE_EXIT(db);
	}

	ASSERT(zfs_refcount_is_zero(&db->db_holds));

	db->db_parent = NULL;

	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	ASSERT(db->db_hash_next == NULL);
	ASSERT(db->db_blkptr == NULL);
	ASSERT(db->db_data_pending == NULL);
	ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);
	ASSERT(!multilist_link_active(&db->db_cache_link));

	/*
	* If this dbuf is referenced from an indirect dbuf,
	* decrement the ref count on the indirect dbuf.
	*/
	if (parent && parent != dndb) {
	mutex_enter(&parent->db_mtx);
	dbuf_rele_and_unlock(parent, db, B_TRUE);
	}

	kmem_cache_free(dbuf_kmem_cache, db);
	arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
	}

	/*
	* Note: While bpp will always be updated if the function returns success,
	* parentp will not be updated if the dnode does not have dn_dbuf filled in;
	* this happens when the dnode is the meta-dnode, or {user\|group\|project}used
	* object.
	*/
	__attribute__((always_inline))
	static inline int
	dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse,
	dmu_buf_impl_t parentp, blkptr_t bpp)
	{
	*parentp = NULL;
	*bpp = NULL;

	ASSERT(blkid != DMU_BONUS_BLKID);

	if (blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (dn->dn_have_spill &&
	(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
	*bpp = DN_SPILL_BLKPTR(dn->dn_phys);
	else
	*bpp = NULL;
	dbuf_add_ref(dn->dn_dbuf, NULL);
	*parentp = dn->dn_dbuf;
	mutex_exit(&dn->dn_mtx);
	return (0);
	}

	int nlevels =
	(dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels;
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;

	ASSERT3U(level * epbs, <, 64);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	/*
	* This assertion shouldn't trip as long as the max indirect block size
	* is less than 1M. The reason for this is that up to that point,
	* the number of levels required to address an entire object with blocks
	* of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In
	* other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55
	* (i.e. we can address the entire object), objects will all use at most
	* N-1 levels and the assertion won't overflow. However, once epbs is
	* 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be
	* enough to address an entire object, so objects will have 5 levels,
	* but then this assertion will overflow.
	*
	* All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we
	* need to redo this logic to handle overflows.
	*/
	ASSERT(level >= nlevels \|\|
	((nlevels - level - 1) * epbs) +
	highbit64(dn->dn_phys->dn_nblkptr) <= 64);
	if (level >= nlevels \|\|
	blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr <<
	((nlevels - level - 1) * epbs)) \|\|
	(fail_sparse &&
	blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) {
	/* the buffer has no parent yet */
	return (SET_ERROR(ENOENT));
	} else if (level < nlevels-1) {
	/* this block is referenced from an indirect block */
	int err;

	err = dbuf_hold_impl(dn, level + 1,
	blkid >> epbs, fail_sparse, FALSE, NULL, parentp);

	if (err)
	return (err);
	err = dbuf_read(*parentp, NULL,
	(DB_RF_HAVESTRUCT \| DB_RF_NOPREFETCH \| DB_RF_CANFAIL));
	if (err) {
	dbuf_rele(*parentp, NULL);
	*parentp = NULL;
	return (err);
	}
	rw_enter(&(*parentp)->db_rwlock, RW_READER);
	bpp = ((blkptr_t )(*parentp)->db.db_data) +
	(blkid & ((1ULL << epbs) - 1));
	if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))
	ASSERT(BP_IS_HOLE(*bpp));
	rw_exit(&(*parentp)->db_rwlock);
	return (0);
	} else {
	/* the block is referenced from the dnode */
	ASSERT3U(level, ==, nlevels-1);
	ASSERT(dn->dn_phys->dn_nblkptr == 0 \|\|
	blkid < dn->dn_phys->dn_nblkptr);
	if (dn->dn_dbuf) {
	dbuf_add_ref(dn->dn_dbuf, NULL);
	*parentp = dn->dn_dbuf;
	}
	*bpp = &dn->dn_phys->dn_blkptr[blkid];
	return (0);
	}
	}

	static dmu_buf_impl_t *
	dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid,
	dmu_buf_impl_t parent, blkptr_t blkptr)
	{
	objset_t *os = dn->dn_objset;
	dmu_buf_impl_t db, odb;

	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	ASSERT(dn->dn_type != DMU_OT_NONE);

	db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP);

	list_create(&db->db_dirty_records, sizeof (dbuf_dirty_record_t),
	offsetof(dbuf_dirty_record_t, dr_dbuf_node));

	db->db_objset = os;
	db->db.db_object = dn->dn_object;
	db->db_level = level;
	db->db_blkid = blkid;
	db->db_dirtycnt = 0;
	db->db_dnode_handle = dn->dn_handle;
	db->db_parent = parent;
	db->db_blkptr = blkptr;

	db->db_user = NULL;
	db->db_user_immediate_evict = FALSE;
	db->db_freed_in_flight = FALSE;
	db->db_pending_evict = FALSE;

	if (blkid == DMU_BONUS_BLKID) {
	ASSERT3P(parent, ==, dn->dn_dbuf);
	db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) -
	(dn->dn_nblkptr-1) * sizeof (blkptr_t);
	ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
	db->db.db_offset = DMU_BONUS_BLKID;
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "bonus buffer created");
	db->db_caching_status = DB_NO_CACHE;
	/* the bonus dbuf is not placed in the hash table */
	arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
	return (db);
	} else if (blkid == DMU_SPILL_BLKID) {
	db->db.db_size = (blkptr != NULL) ?
	BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE;
	db->db.db_offset = 0;
	} else {
	int blocksize =
	db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz;
	db->db.db_size = blocksize;
	db->db.db_offset = db->db_blkid * blocksize;
	}

	/*
	* Hold the dn_dbufs_mtx while we get the new dbuf
	* in the hash table and added to the dbufs list.
	* This prevents a possible deadlock with someone
	* trying to look up this dbuf before it's added to the
	* dn_dbufs list.
	*/
	mutex_enter(&dn->dn_dbufs_mtx);
	db->db_state = DB_EVICTING; /* not worth logging this state change */
	if ((odb = dbuf_hash_insert(db)) != NULL) {
	/* someone else inserted it first */
	mutex_exit(&dn->dn_dbufs_mtx);
	kmem_cache_free(dbuf_kmem_cache, db);
	DBUF_STAT_BUMP(hash_insert_race);
	return (odb);
	}
	avl_add(&dn->dn_dbufs, db);

	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "regular buffer created");
	db->db_caching_status = DB_NO_CACHE;
	mutex_exit(&dn->dn_dbufs_mtx);
	arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);

	if (parent && parent != dn->dn_dbuf)
	dbuf_add_ref(parent, db);

	ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT \|\|
	zfs_refcount_count(&dn->dn_holds) > 0);
	(void) zfs_refcount_add(&dn->dn_holds, db);

	dprintf_dbuf(db, "db=%p\n", db);

	return (db);
	}

	/*
	* This function returns a block pointer and information about the object,
	* given a dnode and a block. This is a publicly accessible version of
	* dbuf_findbp that only returns some information, rather than the
	* dbuf. Note that the dnode passed in must be held, and the dn_struct_rwlock
	* should be locked as (at least) a reader.
	*/
	int
	dbuf_dnode_findbp(dnode_t *dn, uint64_t level, uint64_t blkid,
	blkptr_t bp, uint16_t datablkszsec, uint8_t *indblkshift)
	{
	dmu_buf_impl_t *dbp = NULL;
	blkptr_t *bp2;
	int err = 0;
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));

	err = dbuf_findbp(dn, level, blkid, B_FALSE, &dbp, &bp2);
	if (err == 0) {
	bp = bp2;
	if (dbp != NULL)
	dbuf_rele(dbp, NULL);
	if (datablkszsec != NULL)
	*datablkszsec = dn->dn_phys->dn_datablkszsec;
	if (indblkshift != NULL)
	*indblkshift = dn->dn_phys->dn_indblkshift;
	}

	return (err);
	}

	typedef struct dbuf_prefetch_arg {
	spa_t dpa_spa; / The spa to issue the prefetch in. */
	zbookmark_phys_t dpa_zb; /* The target block to prefetch. */
	int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */
	int dpa_curlevel; /* The current level that we're reading */
	dnode_t dpa_dnode; / The dnode associated with the prefetch */
	zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */
	zio_t dpa_zio; / The parent zio_t for all prefetches. */
	arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */
	dbuf_prefetch_fn dpa_cb; /* prefetch completion callback */
	void dpa_arg; / prefetch completion arg */
	} dbuf_prefetch_arg_t;

	static void
	dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done)
	{
	if (dpa->dpa_cb != NULL) {
	dpa->dpa_cb(dpa->dpa_arg, dpa->dpa_zb.zb_level,
	dpa->dpa_zb.zb_blkid, io_done);
	}
	kmem_free(dpa, sizeof (*dpa));
	}

	static void
	dbuf_issue_final_prefetch_done(zio_t zio, const zbookmark_phys_t zb,
	const blkptr_t iobp, arc_buf_t abuf, void *private)
	{
	(void) zio, (void) zb, (void) iobp;
	dbuf_prefetch_arg_t *dpa = private;

	if (abuf != NULL)
	arc_buf_destroy(abuf, private);

	dbuf_prefetch_fini(dpa, B_TRUE);
	}

	/*
	* Actually issue the prefetch read for the block given.
	*/
	static void
	dbuf_issue_final_prefetch(dbuf_prefetch_arg_t dpa, blkptr_t bp)
	{
	ASSERT(!BP_IS_REDACTED(bp) \|\|
	dsl_dataset_feature_is_active(
	dpa->dpa_dnode->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));

	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp) \|\| BP_IS_REDACTED(bp))
	return (dbuf_prefetch_fini(dpa, B_FALSE));

	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE;
	arc_flags_t aflags =
	dpa->dpa_aflags \| ARC_FLAG_NOWAIT \| ARC_FLAG_PREFETCH \|
	ARC_FLAG_NO_BUF;

	/* dnodes are always read as raw and then converted later */
	if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) &&
	dpa->dpa_curlevel == 0)
	zio_flags \|= ZIO_FLAG_RAW;

	ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));
	ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level);
	ASSERT(dpa->dpa_zio != NULL);
	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp,
	dbuf_issue_final_prefetch_done, dpa,
	dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb);
	}

	/*
	* Called when an indirect block above our prefetch target is read in. This
	* will either read in the next indirect block down the tree or issue the actual
	* prefetch if the next block down is our target.
	*/
	static void
	dbuf_prefetch_indirect_done(zio_t zio, const zbookmark_phys_t zb,
	const blkptr_t iobp, arc_buf_t abuf, void *private)
	{
	(void) zb, (void) iobp;
	dbuf_prefetch_arg_t *dpa = private;

	ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel);
	ASSERT3S(dpa->dpa_curlevel, >, 0);

	if (abuf == NULL) {
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	}
	ASSERT(zio == NULL \|\| zio->io_error == 0);

	/*
	* The dpa_dnode is only valid if we are called with a NULL
	* zio. This indicates that the arc_read() returned without
	* first calling zio_read() to issue a physical read. Once
	* a physical read is made the dpa_dnode must be invalidated
	* as the locks guarding it may have been dropped. If the
	* dpa_dnode is still valid, then we want to add it to the dbuf
	* cache. To do so, we must hold the dbuf associated with the block
	* we just prefetched, read its contents so that we associate it
	* with an arc_buf_t, and then release it.
	*/
	if (zio != NULL) {
	ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel);
	if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) {
	ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size);
	} else {
	ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size);
	}
	ASSERT3P(zio->io_spa, ==, dpa->dpa_spa);

	dpa->dpa_dnode = NULL;
	} else if (dpa->dpa_dnode != NULL) {
	uint64_t curblkid = dpa->dpa_zb.zb_blkid >>
	(dpa->dpa_epbs * (dpa->dpa_curlevel -
	dpa->dpa_zb.zb_level));
	dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode,
	dpa->dpa_curlevel, curblkid, FTAG);
	if (db == NULL) {
	arc_buf_destroy(abuf, private);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	}
	(void) dbuf_read(db, NULL,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH \| DB_RF_HAVESTRUCT);
	dbuf_rele(db, FTAG);
	}

	dpa->dpa_curlevel--;
	uint64_t nextblkid = dpa->dpa_zb.zb_blkid >>
	(dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level));
	blkptr_t bp = ((blkptr_t )abuf->b_data) +
	P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs);

	ASSERT(!BP_IS_REDACTED(bp) \|\|
	dsl_dataset_feature_is_active(
	dpa->dpa_dnode->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));
	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp)) {
	arc_buf_destroy(abuf, private);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	} else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) {
	ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid);
	dbuf_issue_final_prefetch(dpa, bp);
	} else {
	arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
	zbookmark_phys_t zb;

	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
	if (dpa->dpa_aflags & ARC_FLAG_L2CACHE)
	iter_aflags \|= ARC_FLAG_L2CACHE;

	ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));

	SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset,
	dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid);

	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
	bp, dbuf_prefetch_indirect_done, dpa,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&iter_aflags, &zb);
	}

	arc_buf_destroy(abuf, private);
	}

	/*
	* Issue prefetch reads for the given block on the given level. If the indirect
	* blocks above that block are not in memory, we will read them in
	* asynchronously. As a result, this call never blocks waiting for a read to
	* complete. Note that the prefetch might fail if the dataset is encrypted and
	* the encryption key is unmapped before the IO completes.
	*/
	int
	dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
	zio_priority_t prio, arc_flags_t aflags, dbuf_prefetch_fn cb,
	void *arg)
	{
	blkptr_t bp;
	int epbs, nlevels, curlevel;
	uint64_t curblkid;

	ASSERT(blkid != DMU_BONUS_BLKID);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));

	if (blkid > dn->dn_maxblkid)
	goto no_issue;

	if (level == 0 && dnode_block_freed(dn, blkid))
	goto no_issue;

	/*
	* This dnode hasn't been written to disk yet, so there's nothing to
	* prefetch.
	*/
	nlevels = dn->dn_phys->dn_nlevels;
	if (level >= nlevels \|\| dn->dn_phys->dn_nblkptr == 0)
	goto no_issue;

	epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
	if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level))
	goto no_issue;

	dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object,
	level, blkid);
	if (db != NULL) {
	mutex_exit(&db->db_mtx);
	/*
	* This dbuf already exists. It is either CACHED, or
	* (we assume) about to be read or filled.
	*/
	goto no_issue;
	}

	/*
	* Find the closest ancestor (indirect block) of the target block
	* that is present in the cache. In this indirect block, we will
	* find the bp that is at curlevel, curblkid.
	*/
	curlevel = level;
	curblkid = blkid;
	while (curlevel < nlevels - 1) {
	int parent_level = curlevel + 1;
	uint64_t parent_blkid = curblkid >> epbs;
	dmu_buf_impl_t *db;

	if (dbuf_hold_impl(dn, parent_level, parent_blkid,
	FALSE, TRUE, FTAG, &db) == 0) {
	blkptr_t *bpp = db->db_buf->b_data;
	bp = bpp[P2PHASE(curblkid, 1 << epbs)];
	dbuf_rele(db, FTAG);
	break;
	}

	curlevel = parent_level;
	curblkid = parent_blkid;
	}

	if (curlevel == nlevels - 1) {
	/* No cached indirect blocks found. */
	ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr);
	bp = dn->dn_phys->dn_blkptr[curblkid];
	}
	ASSERT(!BP_IS_REDACTED(&bp) \|\|
	dsl_dataset_feature_is_active(dn->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));
	if (BP_IS_HOLE(&bp) \|\| BP_IS_REDACTED(&bp))
	goto no_issue;

	ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp));

	zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL,
	ZIO_FLAG_CANFAIL);

	dbuf_prefetch_arg_t dpa = kmem_zalloc(sizeof (dpa), KM_SLEEP);
	dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
	SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
	dn->dn_object, level, blkid);
	dpa->dpa_curlevel = curlevel;
	dpa->dpa_prio = prio;
	dpa->dpa_aflags = aflags;
	dpa->dpa_spa = dn->dn_objset->os_spa;
	dpa->dpa_dnode = dn;
	dpa->dpa_epbs = epbs;
	dpa->dpa_zio = pio;
	dpa->dpa_cb = cb;
	dpa->dpa_arg = arg;

	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
	if (dnode_level_is_l2cacheable(&bp, dn, level))
	dpa->dpa_aflags \|= ARC_FLAG_L2CACHE;

	/*
	* If we have the indirect just above us, no need to do the asynchronous
	* prefetch chain; we'll just run the last step ourselves. If we're at
	* a higher level, though, we want to issue the prefetches for all the
	* indirect blocks asynchronously, so we can go on with whatever we were
	* doing.
	*/
	if (curlevel == level) {
	ASSERT3U(curblkid, ==, blkid);
	dbuf_issue_final_prefetch(dpa, &bp);
	} else {
	arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
	zbookmark_phys_t zb;

	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
	if (dnode_level_is_l2cacheable(&bp, dn, level))
	iter_aflags \|= ARC_FLAG_L2CACHE;

	SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
	dn->dn_object, curlevel, curblkid);
	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
	&bp, dbuf_prefetch_indirect_done, dpa,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&iter_aflags, &zb);
	}
	/*
	* We use pio here instead of dpa_zio since it's possible that
	* dpa may have already been freed.
	*/
	zio_nowait(pio);
	return (1);
	no_issue:
	if (cb != NULL)
	cb(arg, level, blkid, B_FALSE);
	return (0);
	}

	int
	dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio,
	arc_flags_t aflags)
	{

	return (dbuf_prefetch_impl(dn, level, blkid, prio, aflags, NULL, NULL));
	}

	/*
	* Helper function for dbuf_hold_impl() to copy a buffer. Handles
	* the case of encrypted, compressed and uncompressed buffers by
	* allocating the new buffer, respectively, with arc_alloc_raw_buf(),
	* arc_alloc_compressed_buf() or arc_alloc_buf().*
	*
	* NOTE: Declared noinline to avoid stack bloat in dbuf_hold_impl().
	*/
	noinline static void
	dbuf_hold_copy(dnode_t dn, dmu_buf_impl_t db)
	{
	dbuf_dirty_record_t *dr = db->db_data_pending;
	arc_buf_t *data = dr->dt.dl.dr_data;
	enum zio_compress compress_type = arc_get_compression(data);
	uint8_t complevel = arc_get_complevel(data);

	if (arc_is_encrypted(data)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(data, &byteorder, salt, iv, mac);
	dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db,
	dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac,
	dn->dn_type, arc_buf_size(data), arc_buf_lsize(data),
	compress_type, complevel));
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	dbuf_set_data(db, arc_alloc_compressed_buf(
	dn->dn_objset->os_spa, db, arc_buf_size(data),
	arc_buf_lsize(data), compress_type, complevel));
	} else {
	dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db,
	DBUF_GET_BUFC_TYPE(db), db->db.db_size));
	}

	rw_enter(&db->db_rwlock, RW_WRITER);
	memcpy(db->db.db_data, data->b_data, arc_buf_size(data));
	rw_exit(&db->db_rwlock);
	}

	/*
	* Returns with db_holds incremented, and db_mtx not held.
	* Note: dn_struct_rwlock must be held.
	*/
	int
	dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid,
	boolean_t fail_sparse, boolean_t fail_uncached,
	const void tag, dmu_buf_impl_t *dbp)
	{
	dmu_buf_impl_t db, parent = NULL;

	/* If the pool has been created, verify the tx_sync_lock is not held */
	spa_t *spa = dn->dn_objset->os_spa;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	if (dp != NULL) {
	ASSERT(!MUTEX_HELD(&dp->dp_tx.tx_sync_lock));
	}

	ASSERT(blkid != DMU_BONUS_BLKID);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	ASSERT3U(dn->dn_nlevels, >, level);

	*dbp = NULL;

	/* dbuf_find() returns with db_mtx held */
	db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid);

	if (db == NULL) {
	blkptr_t *bp = NULL;
	int err;

	if (fail_uncached)
	return (SET_ERROR(ENOENT));

	ASSERT3P(parent, ==, NULL);
	err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp);
	if (fail_sparse) {
	if (err == 0 && bp && BP_IS_HOLE(bp))
	err = SET_ERROR(ENOENT);
	if (err) {
	if (parent)
	dbuf_rele(parent, NULL);
	return (err);
	}
	}
	if (err && err != ENOENT)
	return (err);
	db = dbuf_create(dn, level, blkid, parent, bp);
	}

	if (fail_uncached && db->db_state != DB_CACHED) {
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(ENOENT));
	}

	if (db->db_buf != NULL) {
	arc_buf_access(db->db_buf);
	ASSERT3P(db->db.db_data, ==, db->db_buf->b_data);
	}

	ASSERT(db->db_buf == NULL \|\| arc_referenced(db->db_buf));

	/*
	* If this buffer is currently syncing out, and we are
	* still referencing it from db_data, we need to make a copy
	* of it in case we decide we want to dirty it again in this txg.
	*/
	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	dn->dn_object != DMU_META_DNODE_OBJECT &&
	db->db_state == DB_CACHED && db->db_data_pending) {
	dbuf_dirty_record_t *dr = db->db_data_pending;
	if (dr->dt.dl.dr_data == db->db_buf)
	dbuf_hold_copy(dn, db);
	}

	if (multilist_link_active(&db->db_cache_link)) {
	ASSERT(zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_caching_status == DB_DBUF_CACHE \|\|
	db->db_caching_status == DB_DBUF_METADATA_CACHE);

	multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);
	(void) zfs_refcount_remove_many(
	&dbuf_caches[db->db_caching_status].size,
	db->db.db_size, db);

	if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMPDOWN(metadata_cache_count);
	} else {
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
	db->db.db_size);
	}
	db->db_caching_status = DB_NO_CACHE;
	}
	(void) zfs_refcount_add(&db->db_holds, tag);
	DBUF_VERIFY(db);
	mutex_exit(&db->db_mtx);

	/* NOTE: we can't rele the parent until after we drop the db_mtx */
	if (parent)
	dbuf_rele(parent, NULL);

	ASSERT3P(DB_DNODE(db), ==, dn);
	ASSERT3U(db->db_blkid, ==, blkid);
	ASSERT3U(db->db_level, ==, level);
	*dbp = db;

	return (0);
	}

	dmu_buf_impl_t *
	dbuf_hold(dnode_t dn, uint64_t blkid, const void tag)
	{
	return (dbuf_hold_level(dn, 0, blkid, tag));
	}

	dmu_buf_impl_t *
	dbuf_hold_level(dnode_t dn, int level, uint64_t blkid, const void tag)
	{
	dmu_buf_impl_t *db;
	int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db);
	return (err ? NULL : db);
	}

	void
	dbuf_create_bonus(dnode_t *dn)
	{
	ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock));

	ASSERT(dn->dn_bonus == NULL);
	dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL);
	}

	int
	dbuf_spill_set_blksz(dmu_buf_t db_fake, uint64_t blksz, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	if (db->db_blkid != DMU_SPILL_BLKID)
	return (SET_ERROR(ENOTSUP));
	if (blksz == 0)
	blksz = SPA_MINBLOCKSIZE;
	ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset)));
	blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE);

	dbuf_new_size(db, blksz, tx);

	return (0);
	}

	void
	dbuf_rm_spill(dnode_t dn, dmu_tx_t tx)
	{
	dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx);
	}

	#pragma weak dmu_buf_add_ref = dbuf_add_ref
	void
	dbuf_add_ref(dmu_buf_impl_t db, const void tag)
	{
	int64_t holds = zfs_refcount_add(&db->db_holds, tag);
	VERIFY3S(holds, >, 1);
	}

	#pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref
	boolean_t
	dbuf_try_add_ref(dmu_buf_t db_fake, objset_t os, uint64_t obj, uint64_t blkid,
	const void *tag)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dmu_buf_impl_t *found_db;
	boolean_t result = B_FALSE;

	if (blkid == DMU_BONUS_BLKID)
	found_db = dbuf_find_bonus(os, obj);
	else
	found_db = dbuf_find(os, obj, 0, blkid);

	if (found_db != NULL) {
	if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) {
	(void) zfs_refcount_add(&db->db_holds, tag);
	result = B_TRUE;
	}
	mutex_exit(&found_db->db_mtx);
	}
	return (result);
	}

	/*
	* If you call dbuf_rele() you had better not be referencing the dnode handle
	* unless you have some other direct or indirect hold on the dnode. (An indirect
	* hold is a hold on one of the dnode's dbufs, including the bonus buffer.)
	* Without that, the dbuf_rele() could lead to a dnode_rele() followed by the
	* dnode's parent dbuf evicting its dnode handles.
	*/
	void
	dbuf_rele(dmu_buf_impl_t db, const void tag)
	{
	mutex_enter(&db->db_mtx);
	dbuf_rele_and_unlock(db, tag, B_FALSE);
	}

	void
	dmu_buf_rele(dmu_buf_t db, const void tag)
	{
	dbuf_rele((dmu_buf_impl_t *)db, tag);
	}

	/*
	* dbuf_rele() for an already-locked dbuf. This is necessary to allow
	* db_dirtycnt and db_holds to be updated atomically. The 'evicting'
	* argument should be set if we are already in the dbuf-evicting code
	* path, in which case we don't want to recursively evict. This allows us to
	* avoid deeply nested stacks that would have a call flow similar to this:
	*
	* dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify()
	* ^ \|
	* \| \|
	* +-----dbuf_destroy()<--dbuf_evict_one()<--------+
	*
	*/
	void
	dbuf_rele_and_unlock(dmu_buf_impl_t db, const void tag, boolean_t evicting)
	{
	int64_t holds;
	uint64_t size;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	DBUF_VERIFY(db);

	/*
	* Remove the reference to the dbuf before removing its hold on the
	* dnode so we can guarantee in dnode_move() that a referenced bonus
	* buffer has a corresponding dnode hold.
	*/
	holds = zfs_refcount_remove(&db->db_holds, tag);
	ASSERT(holds >= 0);

	/*
	* We can't freeze indirects if there is a possibility that they
	* may be modified in the current syncing context.
	*/
	if (db->db_buf != NULL &&
	holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) {
	arc_buf_freeze(db->db_buf);
	}

	if (holds == db->db_dirtycnt &&
	db->db_level == 0 && db->db_user_immediate_evict)
	dbuf_evict_user(db);

	if (holds == 0) {
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dnode_t *dn;
	boolean_t evict_dbuf = db->db_pending_evict;

	/*
	* If the dnode moves here, we cannot cross this
	* barrier until the move completes.
	*/
	DB_DNODE_ENTER(db);

	dn = DB_DNODE(db);
	atomic_dec_32(&dn->dn_dbufs_count);

	/*
	* Decrementing the dbuf count means that the bonus
	* buffer's dnode hold is no longer discounted in
	* dnode_move(). The dnode cannot move until after
	* the dnode_rele() below.
	*/
	DB_DNODE_EXIT(db);

	/*
	* Do not reference db after its lock is dropped.
	* Another thread may evict it.
	*/
	mutex_exit(&db->db_mtx);

	if (evict_dbuf)
	dnode_evict_bonus(dn);

	dnode_rele(dn, db);
	} else if (db->db_buf == NULL) {
	/*
	* This is a special case: we never associated this
	* dbuf with any data allocated from the ARC.
	*/
	ASSERT(db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL);
	dbuf_destroy(db);
	} else if (arc_released(db->db_buf)) {
	/*
	* This dbuf has anonymous data associated with it.
	*/
	dbuf_destroy(db);
	} else {
	boolean_t do_arc_evict = B_FALSE;
	blkptr_t bp;
	spa_t *spa = dmu_objset_spa(db->db_objset);

	if (!DBUF_IS_CACHEABLE(db) &&
	db->db_blkptr != NULL &&
	!BP_IS_HOLE(db->db_blkptr) &&
	!BP_IS_EMBEDDED(db->db_blkptr)) {
	do_arc_evict = B_TRUE;
	bp = *db->db_blkptr;
	}

	if (!DBUF_IS_CACHEABLE(db) \|\|
	db->db_pending_evict) {
	dbuf_destroy(db);
	} else if (!multilist_link_active(&db->db_cache_link)) {
	ASSERT3U(db->db_caching_status, ==,
	DB_NO_CACHE);

	dbuf_cached_state_t dcs =
	dbuf_include_in_metadata_cache(db) ?
	DB_DBUF_METADATA_CACHE : DB_DBUF_CACHE;
	db->db_caching_status = dcs;

	multilist_insert(&dbuf_caches[dcs].cache, db);
	uint64_t db_size = db->db.db_size;
	size = zfs_refcount_add_many(
	&dbuf_caches[dcs].size, db_size, db);
	uint8_t db_level = db->db_level;
	mutex_exit(&db->db_mtx);

	if (dcs == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMP(metadata_cache_count);
	DBUF_STAT_MAX(
	metadata_cache_size_bytes_max,
	size);
	} else {
	DBUF_STAT_BUMP(cache_count);
	DBUF_STAT_MAX(cache_size_bytes_max,
	size);
	DBUF_STAT_BUMP(cache_levels[db_level]);
	DBUF_STAT_INCR(
	cache_levels_bytes[db_level],
	db_size);
	}

	if (dcs == DB_DBUF_CACHE && !evicting)
	dbuf_evict_notify(size);
	}

	if (do_arc_evict)
	arc_freed(spa, &bp);
	}
	} else {
	mutex_exit(&db->db_mtx);
	}

	}

	#pragma weak dmu_buf_refcount = dbuf_refcount
	uint64_t
	dbuf_refcount(dmu_buf_impl_t *db)
	{
	return (zfs_refcount_count(&db->db_holds));
	}

	uint64_t
	dmu_buf_user_refcount(dmu_buf_t *db_fake)
	{
	uint64_t holds;
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	mutex_enter(&db->db_mtx);
	ASSERT3U(zfs_refcount_count(&db->db_holds), >=, db->db_dirtycnt);
	holds = zfs_refcount_count(&db->db_holds) - db->db_dirtycnt;
	mutex_exit(&db->db_mtx);

	return (holds);
	}

	void *
	dmu_buf_replace_user(dmu_buf_t db_fake, dmu_buf_user_t old_user,
	dmu_buf_user_t *new_user)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	mutex_enter(&db->db_mtx);
	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	if (db->db_user == old_user)
	db->db_user = new_user;
	else
	old_user = db->db_user;
	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	mutex_exit(&db->db_mtx);

	return (old_user);
	}

	void *
	dmu_buf_set_user(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	return (dmu_buf_replace_user(db_fake, NULL, user));
	}

	void *
	dmu_buf_set_user_ie(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	db->db_user_immediate_evict = TRUE;
	return (dmu_buf_set_user(db_fake, user));
	}

	void *
	dmu_buf_remove_user(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	return (dmu_buf_replace_user(db_fake, user, NULL));
	}

	void *
	dmu_buf_get_user(dmu_buf_t *db_fake)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	return (db->db_user);
	}

	void
	dmu_buf_user_evict_wait(void)
	{
	taskq_wait(dbu_evict_taskq);
	}

	blkptr_t *
	dmu_buf_get_blkptr(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	return (dbi->db_blkptr);
	}

	objset_t *
	dmu_buf_get_objset(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	return (dbi->db_objset);
	}

	dnode_t *
	dmu_buf_dnode_enter(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	DB_DNODE_ENTER(dbi);
	return (DB_DNODE(dbi));
	}

	void
	dmu_buf_dnode_exit(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	DB_DNODE_EXIT(dbi);
	}

	static void
	dbuf_check_blkptr(dnode_t dn, dmu_buf_impl_t db)
	{
	/* ASSERT(dmu_tx_is_syncing(tx) */
	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (db->db_blkptr != NULL)
	return;

	if (db->db_blkid == DMU_SPILL_BLKID) {
	db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys);
	BP_ZERO(db->db_blkptr);
	return;
	}
	if (db->db_level == dn->dn_phys->dn_nlevels-1) {
	/*
	* This buffer was allocated at a time when there was
	* no available blkptrs from the dnode, or it was
	* inappropriate to hook it in (i.e., nlevels mismatch).
	*/
	ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr);
	ASSERT(db->db_parent == NULL);
	db->db_parent = dn->dn_dbuf;
	db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid];
	DBUF_VERIFY(db);
	} else {
	dmu_buf_impl_t *parent = db->db_parent;
	int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;

	ASSERT(dn->dn_phys->dn_nlevels > 1);
	if (parent == NULL) {
	mutex_exit(&db->db_mtx);
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	parent = dbuf_hold_level(dn, db->db_level + 1,
	db->db_blkid >> epbs, db);
	rw_exit(&dn->dn_struct_rwlock);
	mutex_enter(&db->db_mtx);
	db->db_parent = parent;
	}
	db->db_blkptr = (blkptr_t *)parent->db.db_data +
	(db->db_blkid & ((1ULL << epbs) - 1));
	DBUF_VERIFY(db);
	}
	}

	static void
	dbuf_sync_bonus(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	void *data = dr->dt.dl.dr_data;

	ASSERT0(db->db_level);
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db_blkid == DMU_BONUS_BLKID);
	ASSERT(data != NULL);

	dnode_t *dn = dr->dr_dnode;
	ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=,
	DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1));
	memcpy(DN_BONUS(dn->dn_phys), data, DN_MAX_BONUS_LEN(dn->dn_phys));

	dbuf_sync_leaf_verify_bonus_dnode(dr);

	dbuf_undirty_bonus(dr);
	dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);
	}

	/*
	* When syncing out a blocks of dnodes, adjust the block to deal with
	* encryption. Normally, we make sure the block is decrypted before writing
	* it. If we have crypt params, then we are writing a raw (encrypted) block,
	* from a raw receive. In this case, set the ARC buf's crypt params so
	* that the BP will be filled with the correct byteorder, salt, iv, and mac.
	*/
	static void
	dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr)
	{
	int err;
	dmu_buf_impl_t *db = dr->dr_dbuf;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
	ASSERT3U(db->db_level, ==, 0);

	if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) {
	zbookmark_phys_t zb;

	/*
	* Unfortunately, there is currently no mechanism for
	* syncing context to handle decryption errors. An error
	* here is only possible if an attacker maliciously
	* changed a dnode block and updated the associated
	* checksums going up the block tree.
	*/
	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);
	err = arc_untransform(db->db_buf, db->db_objset->os_spa,
	&zb, B_TRUE);
	if (err)
	panic("Invalid dnode block MAC");
	} else if (dr->dt.dl.dr_has_raw_params) {
	(void) arc_release(dr->dt.dl.dr_data, db);
	arc_convert_to_raw(dr->dt.dl.dr_data,
	dmu_objset_id(db->db_objset),
	dr->dt.dl.dr_byteorder, DMU_OT_DNODE,
	dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac);
	}
	}

	/*
	* dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it
	* is critical the we not allow the compiler to inline this function in to
	* dbuf_sync_list() thereby drastically bloating the stack usage.
	*/
	noinline static void
	dbuf_sync_indirect(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;

	ASSERT(dmu_tx_is_syncing(tx));

	dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);

	mutex_enter(&db->db_mtx);

	ASSERT(db->db_level > 0);
	DBUF_VERIFY(db);

	/* Read the block if it hasn't been read yet. */
	if (db->db_buf == NULL) {
	mutex_exit(&db->db_mtx);
	(void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED);
	mutex_enter(&db->db_mtx);
	}
	ASSERT3U(db->db_state, ==, DB_CACHED);
	ASSERT(db->db_buf != NULL);

	/* Indirect block size must match what the dnode thinks it is. */
	ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
	dbuf_check_blkptr(dn, db);

	/* Provide the pending dirty record to child dbufs */
	db->db_data_pending = dr;

	mutex_exit(&db->db_mtx);

	dbuf_write(dr, db->db_buf, tx);

	zio_t *zio = dr->dr_zio;
	mutex_enter(&dr->dt.di.dr_mtx);
	dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx);
	ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
	mutex_exit(&dr->dt.di.dr_mtx);
	zio_nowait(zio);
	}

	/*
	* Verify that the size of the data in our bonus buffer does not exceed
	* its recorded size.
	*
	* The purpose of this verification is to catch any cases in development
	* where the size of a phys structure (i.e space_map_phys_t) grows and,
	* due to incorrect feature management, older pools expect to read more
	* data even though they didn't actually write it to begin with.
	*
	* For a example, this would catch an error in the feature logic where we
	* open an older pool and we expect to write the space map histogram of
	* a space map with size SPACE_MAP_SIZE_V0.
	*/
	static void
	dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr)
	{
	#ifdef ZFS_DEBUG
	dnode_t *dn = dr->dr_dnode;

	/*
	* Encrypted bonus buffers can have data past their bonuslen.
	* Skip the verification of these blocks.
	*/
	if (DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))
	return;

	uint16_t bonuslen = dn->dn_phys->dn_bonuslen;
	uint16_t maxbonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	ASSERT3U(bonuslen, <=, maxbonuslen);

	arc_buf_t *datap = dr->dt.dl.dr_data;
	char datap_end = ((char )datap) + bonuslen;
	char datap_max = ((char )datap) + maxbonuslen;

	/* ensure that everything is zero after our data */
	for (; datap_end < datap_max; datap_end++)
	ASSERT(*datap_end == 0);
	#endif
	}

	static blkptr_t *
	dbuf_lightweight_bp(dbuf_dirty_record_t *dr)
	{
	/* This must be a lightweight dirty record. */
	ASSERT3P(dr->dr_dbuf, ==, NULL);
	dnode_t *dn = dr->dr_dnode;

	if (dn->dn_phys->dn_nlevels == 1) {
	VERIFY3U(dr->dt.dll.dr_blkid, <, dn->dn_phys->dn_nblkptr);
	return (&dn->dn_phys->dn_blkptr[dr->dt.dll.dr_blkid]);
	} else {
	dmu_buf_impl_t *parent_db = dr->dr_parent->dr_dbuf;
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	VERIFY3U(parent_db->db_level, ==, 1);
	VERIFY3P(parent_db->db_dnode_handle->dnh_dnode, ==, dn);
	VERIFY3U(dr->dt.dll.dr_blkid >> epbs, ==, parent_db->db_blkid);
	blkptr_t *bp = parent_db->db.db_data;
	return (&bp[dr->dt.dll.dr_blkid & ((1 << epbs) - 1)]);
	}
	}

	static void
	dbuf_lightweight_ready(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	blkptr_t *bp = zio->io_bp;

	if (zio->io_error != 0)
	return;

	dnode_t *dn = dr->dr_dnode;

	blkptr_t *bp_orig = dbuf_lightweight_bp(dr);
	spa_t *spa = dmu_objset_spa(dn->dn_objset);
	int64_t delta = bp_get_dsize_sync(spa, bp) -
	bp_get_dsize_sync(spa, bp_orig);
	dnode_diduse_space(dn, delta);

	uint64_t blkid = dr->dt.dll.dr_blkid;
	mutex_enter(&dn->dn_mtx);
	if (blkid > dn->dn_phys->dn_maxblkid) {
	ASSERT0(dn->dn_objset->os_raw_receive);
	dn->dn_phys->dn_maxblkid = blkid;
	}
	mutex_exit(&dn->dn_mtx);

	if (!BP_IS_EMBEDDED(bp)) {
	uint64_t fill = BP_IS_HOLE(bp) ? 0 : 1;
	BP_SET_FILL(bp, fill);
	}

	dmu_buf_impl_t *parent_db;
	EQUIV(dr->dr_parent == NULL, dn->dn_phys->dn_nlevels == 1);
	if (dr->dr_parent == NULL) {
	parent_db = dn->dn_dbuf;
	} else {
	parent_db = dr->dr_parent->dr_dbuf;
	}
	rw_enter(&parent_db->db_rwlock, RW_WRITER);
	bp_orig = bp;
	rw_exit(&parent_db->db_rwlock);
	}

	static void
	dbuf_lightweight_physdone(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	dsl_pool_t *dp = spa_get_dsl(zio->io_spa);
	ASSERT3U(dr->dr_txg, ==, zio->io_txg);

	/*
	* The callback will be called io_phys_children times. Retire one
	* portion of our dirty space each time we are called. Any rounding
	* error will be cleaned up by dbuf_lightweight_done().
	*/
	int delta = dr->dr_accounted / zio->io_phys_children;
	dsl_pool_undirty_space(dp, delta, zio->io_txg);
	}

	static void
	dbuf_lightweight_done(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;

	VERIFY0(zio->io_error);

	objset_t *os = dr->dr_dnode->dn_objset;
	dmu_tx_t *tx = os->os_synctx;

	if (zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)) {
	ASSERT(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
	} else {
	dsl_dataset_t *ds = os->os_dsl_dataset;
	(void) dsl_dataset_block_kill(ds, &zio->io_bp_orig, tx, B_TRUE);
	dsl_dataset_block_born(ds, zio->io_bp, tx);
	}

	/*
	* See comment in dbuf_write_done().
	*/
	if (zio->io_phys_children == 0) {
	dsl_pool_undirty_space(dmu_objset_pool(os),
	dr->dr_accounted, zio->io_txg);
	} else {
	dsl_pool_undirty_space(dmu_objset_pool(os),
	dr->dr_accounted % zio->io_phys_children, zio->io_txg);
	}

	abd_free(dr->dt.dll.dr_abd);
	kmem_free(dr, sizeof (*dr));
	}

	noinline static void
	dbuf_sync_lightweight(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dnode_t *dn = dr->dr_dnode;
	zio_t *pio;
	if (dn->dn_phys->dn_nlevels == 1) {
	pio = dn->dn_zio;
	} else {
	pio = dr->dr_parent->dr_zio;
	}

	zbookmark_phys_t zb = {
	.zb_objset = dmu_objset_id(dn->dn_objset),
	.zb_object = dn->dn_object,
	.zb_level = 0,
	.zb_blkid = dr->dt.dll.dr_blkid,
	};

	/*
	* See comment in dbuf_write(). This is so that zio->io_bp_orig
	* will have the old BP in dbuf_lightweight_done().
	*/
	dr->dr_bp_copy = *dbuf_lightweight_bp(dr);

	dr->dr_zio = zio_write(pio, dmu_objset_spa(dn->dn_objset),
	dmu_tx_get_txg(tx), &dr->dr_bp_copy, dr->dt.dll.dr_abd,
	dn->dn_datablksz, abd_get_size(dr->dt.dll.dr_abd),
	&dr->dt.dll.dr_props, dbuf_lightweight_ready, NULL,
	dbuf_lightweight_physdone, dbuf_lightweight_done, dr,
	ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_MUSTSUCCEED \| dr->dt.dll.dr_flags, &zb);

	zio_nowait(dr->dr_zio);
	}

	/*
	* dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is
	* critical the we not allow the compiler to inline this function in to
	* dbuf_sync_list() thereby drastically bloating the stack usage.
	*/
	noinline static void
	dbuf_sync_leaf(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	arc_buf_t **datap = &dr->dt.dl.dr_data;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;
	objset_t *os;
	uint64_t txg = tx->tx_txg;

	ASSERT(dmu_tx_is_syncing(tx));

	dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);

	mutex_enter(&db->db_mtx);
	/*
	* To be synced, we must be dirtied. But we
	* might have been freed after the dirty.
	*/
	if (db->db_state == DB_UNCACHED) {
	/* This buffer has been freed since it was dirtied */
	ASSERT(db->db.db_data == NULL);
	} else if (db->db_state == DB_FILL) {
	/* This buffer was freed and is now being re-filled */
	ASSERT(db->db.db_data != dr->dt.dl.dr_data);
	} else {
	ASSERT(db->db_state == DB_CACHED \|\| db->db_state == DB_NOFILL);
	}
	DBUF_VERIFY(db);

	if (db->db_blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
	/*
	* In the previous transaction group, the bonus buffer
	* was entirely used to store the attributes for the
	* dnode which overrode the dn_spill field. However,
	* when adding more attributes to the file a spill
	* block was required to hold the extra attributes.
	*
	* Make sure to clear the garbage left in the dn_spill
	* field from the previous attributes in the bonus
	* buffer. Otherwise, after writing out the spill
	* block to the new allocated dva, it will free
	* the old block pointed to by the invalid dn_spill.
	*/
	db->db_blkptr = NULL;
	}
	dn->dn_phys->dn_flags \|= DNODE_FLAG_SPILL_BLKPTR;
	mutex_exit(&dn->dn_mtx);
	}

	/*
	* If this is a bonus buffer, simply copy the bonus data into the
	* dnode. It will be written out when the dnode is synced (and it
	* will be synced, since it must have been dirty for dbuf_sync to
	* be called).
	*/
	if (db->db_blkid == DMU_BONUS_BLKID) {
	ASSERT(dr->dr_dbuf == db);
	dbuf_sync_bonus(dr, tx);
	return;
	}

	os = dn->dn_objset;

	/*
	* This function may have dropped the db_mtx lock allowing a dmu_sync
	* operation to sneak in. As a result, we need to ensure that we
	* don't check the dr_override_state until we have returned from
	* dbuf_check_blkptr.
	*/
	dbuf_check_blkptr(dn, db);

	/*
	* If this buffer is in the middle of an immediate write,
	* wait for the synchronous IO to complete.
	*/
	while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
	ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
	cv_wait(&db->db_changed, &db->db_mtx);
	ASSERT(dr->dt.dl.dr_override_state != DR_NOT_OVERRIDDEN);
	}

	/*
	* If this is a dnode block, ensure it is appropriately encrypted
	* or decrypted, depending on what we are writing to it this txg.
	*/
	if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT)
	dbuf_prepare_encrypted_dnode_leaf(dr);

	if (db->db_state != DB_NOFILL &&
	dn->dn_object != DMU_META_DNODE_OBJECT &&
	zfs_refcount_count(&db->db_holds) > 1 &&
	dr->dt.dl.dr_override_state != DR_OVERRIDDEN &&
	*datap == db->db_buf) {
	/*
	* If this buffer is currently "in use" (i.e., there
	* are active holds and db_data still references it),
	* then make a copy before we start the write so that
	* any modifications from the open txg will not leak
	* into this write.
	*
	* NOTE: this copy does not need to be made for
	* objects only modified in the syncing context (e.g.
	* DNONE_DNODE blocks).
	*/
	int psize = arc_buf_size(*datap);
	int lsize = arc_buf_lsize(*datap);
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	enum zio_compress compress_type = arc_get_compression(*datap);
	uint8_t complevel = arc_get_complevel(*datap);

	if (arc_is_encrypted(*datap)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(*datap, &byteorder, salt, iv, mac);
	*datap = arc_alloc_raw_buf(os->os_spa, db,
	dmu_objset_id(os), byteorder, salt, iv, mac,
	dn->dn_type, psize, lsize, compress_type,
	complevel);
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	ASSERT3U(type, ==, ARC_BUFC_DATA);
	*datap = arc_alloc_compressed_buf(os->os_spa, db,
	psize, lsize, compress_type, complevel);
	} else {
	*datap = arc_alloc_buf(os->os_spa, db, type, psize);
	}
	memcpy((*datap)->b_data, db->db.db_data, psize);
	}
	db->db_data_pending = dr;

	mutex_exit(&db->db_mtx);

	dbuf_write(dr, *datap, tx);

	ASSERT(!list_link_active(&dr->dr_dirty_node));
	if (dn->dn_object == DMU_META_DNODE_OBJECT) {
	list_insert_tail(&dn->dn_dirty_records[txg & TXG_MASK], dr);
	} else {
	zio_nowait(dr->dr_zio);
	}
	}

	void
	dbuf_sync_list(list_t list, int level, dmu_tx_t tx)
	{
	dbuf_dirty_record_t *dr;

	while ((dr = list_head(list))) {
	if (dr->dr_zio != NULL) {
	/*
	* If we find an already initialized zio then we
	* are processing the meta-dnode, and we have finished.
	* The dbufs for all dnodes are put back on the list
	* during processing, so that we can zio_wait()
	* these IOs after initiating all child IOs.
	*/
	ASSERT3U(dr->dr_dbuf->db.db_object, ==,
	DMU_META_DNODE_OBJECT);
	break;
	}
	list_remove(list, dr);
	if (dr->dr_dbuf == NULL) {
	dbuf_sync_lightweight(dr, tx);
	} else {
	if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID &&
	dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) {
	VERIFY3U(dr->dr_dbuf->db_level, ==, level);
	}
	if (dr->dr_dbuf->db_level > 0)
	dbuf_sync_indirect(dr, tx);
	else
	dbuf_sync_leaf(dr, tx);
	}
	}
	}

	static void
	dbuf_write_ready(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) buf;
	dmu_buf_impl_t *db = vdb;
	dnode_t *dn;
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	spa_t *spa = zio->io_spa;
	int64_t delta;
	uint64_t fill = 0;
	int i;

	ASSERT3P(db->db_blkptr, !=, NULL);
	ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig);
	dnode_diduse_space(dn, delta - zio->io_prev_space_delta);
	zio->io_prev_space_delta = delta;

	if (bp->blk_birth != 0) {
	ASSERT((db->db_blkid != DMU_SPILL_BLKID &&
	BP_GET_TYPE(bp) == dn->dn_type) \|\|
	(db->db_blkid == DMU_SPILL_BLKID &&
	BP_GET_TYPE(bp) == dn->dn_bonustype) \|\|
	BP_IS_EMBEDDED(bp));
	ASSERT(BP_GET_LEVEL(bp) == db->db_level);
	}

	mutex_enter(&db->db_mtx);

	#ifdef ZFS_DEBUG
	if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
	ASSERT(!(BP_IS_HOLE(bp)) &&
	db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
	}
	#endif

	if (db->db_level == 0) {
	mutex_enter(&dn->dn_mtx);
	if (db->db_blkid > dn->dn_phys->dn_maxblkid &&
	db->db_blkid != DMU_SPILL_BLKID) {
	ASSERT0(db->db_objset->os_raw_receive);
	dn->dn_phys->dn_maxblkid = db->db_blkid;
	}
	mutex_exit(&dn->dn_mtx);

	if (dn->dn_type == DMU_OT_DNODE) {
	i = 0;
	while (i < db->db.db_size) {
	dnode_phys_t *dnp =
	(void )(((char )db->db.db_data) + i);

	i += DNODE_MIN_SIZE;
	if (dnp->dn_type != DMU_OT_NONE) {
	fill++;
	i += dnp->dn_extra_slots *
	DNODE_MIN_SIZE;
	}
	}
	} else {
	if (BP_IS_HOLE(bp)) {
	fill = 0;
	} else {
	fill = 1;
	}
	}
	} else {
	blkptr_t *ibp = db->db.db_data;
	ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
	for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) {
	if (BP_IS_HOLE(ibp))
	continue;
	fill += BP_GET_FILL(ibp);
	}
	}
	DB_DNODE_EXIT(db);

	if (!BP_IS_EMBEDDED(bp))
	BP_SET_FILL(bp, fill);

	mutex_exit(&db->db_mtx);

	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_WRITER, FTAG);
	db->db_blkptr = bp;
	dmu_buf_unlock_parent(db, dblt, FTAG);
	}

	/*
	* This function gets called just prior to running through the compression
	* stage of the zio pipeline. If we're an indirect block comprised of only
	* holes, then we want this indirect to be compressed away to a hole. In
	* order to do that we must zero out any information about the holes that
	* this indirect points to prior to before we try to compress it.
	*/
	static void
	dbuf_write_children_ready(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) zio, (void) buf;
	dmu_buf_impl_t *db = vdb;
	dnode_t *dn;
	blkptr_t *bp;
	unsigned int epbs, i;

	ASSERT3U(db->db_level, >, 0);
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
	ASSERT3U(epbs, <, 31);

	/* Determine if all our children are holes */
	for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) {
	if (!BP_IS_HOLE(bp))
	break;
	}

	/*
	* If all the children are holes, then zero them all out so that
	* we may get compressed away.
	*/
	if (i == 1ULL << epbs) {
	/*
	* We only found holes. Grab the rwlock to prevent
	* anybody from reading the blocks we're about to
	* zero out.
	*/
	rw_enter(&db->db_rwlock, RW_WRITER);
	memset(db->db.db_data, 0, db->db.db_size);
	rw_exit(&db->db_rwlock);
	}
	DB_DNODE_EXIT(db);
	}

	/*
	* The SPA will call this callback several times for each zio - once
	* for every physical child i/o (zio->io_phys_children times). This
	* allows the DMU to monitor the progress of each logical i/o. For example,
	* there may be 2 copies of an indirect block, or many fragments of a RAID-Z
	* block. There may be a long delay before all copies/fragments are completed,
	* so this callback allows us to retire dirty space gradually, as the physical
	* i/os complete.
	*/
	static void
	dbuf_write_physdone(zio_t zio, arc_buf_t buf, void *arg)
	{
	(void) buf;
	dmu_buf_impl_t *db = arg;
	objset_t *os = db->db_objset;
	dsl_pool_t *dp = dmu_objset_pool(os);
	dbuf_dirty_record_t *dr;
	int delta = 0;

	dr = db->db_data_pending;
	ASSERT3U(dr->dr_txg, ==, zio->io_txg);

	/*
	* The callback will be called io_phys_children times. Retire one
	* portion of our dirty space each time we are called. Any rounding
	* error will be cleaned up by dbuf_write_done().
	*/
	delta = dr->dr_accounted / zio->io_phys_children;
	dsl_pool_undirty_space(dp, delta, zio->io_txg);
	}

	static void
	dbuf_write_done(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) buf;
	dmu_buf_impl_t *db = vdb;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	blkptr_t *bp = db->db_blkptr;
	objset_t *os = db->db_objset;
	dmu_tx_t *tx = os->os_synctx;

	ASSERT0(zio->io_error);
	ASSERT(db->db_blkptr == bp);

	/*
	* For nopwrites and rewrites we ensure that the bp matches our
	* original and bypass all the accounting.
	*/
	if (zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)) {
	ASSERT(BP_EQUAL(bp, bp_orig));
	} else {
	dsl_dataset_t *ds = os->os_dsl_dataset;
	(void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE);
	dsl_dataset_block_born(ds, bp, tx);
	}

	mutex_enter(&db->db_mtx);

	DBUF_VERIFY(db);

	dbuf_dirty_record_t *dr = db->db_data_pending;
	dnode_t *dn = dr->dr_dnode;
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	ASSERT(dr->dr_dbuf == db);
	ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
	list_remove(&db->db_dirty_records, dr);

	#ifdef ZFS_DEBUG
	if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
	ASSERT(!(BP_IS_HOLE(db->db_blkptr)) &&
	db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
	}
	#endif

	if (db->db_level == 0) {
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
	if (db->db_state != DB_NOFILL) {
	if (dr->dt.dl.dr_data != db->db_buf)
	arc_buf_destroy(dr->dt.dl.dr_data, db);
	}
	} else {
	ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
	ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift);
	if (!BP_IS_HOLE(db->db_blkptr)) {
	int epbs __maybe_unused = dn->dn_phys->dn_indblkshift -
	SPA_BLKPTRSHIFT;
	ASSERT3U(db->db_blkid, <=,
	dn->dn_phys->dn_maxblkid >> (db->db_level * epbs));
	ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==,
	db->db.db_size);
	}
	mutex_destroy(&dr->dt.di.dr_mtx);
	list_destroy(&dr->dt.di.dr_children);
	}

	cv_broadcast(&db->db_changed);
	ASSERT(db->db_dirtycnt > 0);
	db->db_dirtycnt -= 1;
	db->db_data_pending = NULL;
	dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);

	/*
	* If we didn't do a physical write in this ZIO and we
	* still ended up here, it means that the space of the
	* dbuf that we just released (and undirtied) above hasn't
	* been marked as undirtied in the pool's accounting.
	*
	* Thus, we undirty that space in the pool's view of the
	* world here. For physical writes this type of update
	* happens in dbuf_write_physdone().
	*
	* If we did a physical write, cleanup any rounding errors
	* that came up due to writing multiple copies of a block
	* on disk [see dbuf_write_physdone()].
	*/
	if (zio->io_phys_children == 0) {
	dsl_pool_undirty_space(dmu_objset_pool(os),
	dr->dr_accounted, zio->io_txg);
	} else {
	dsl_pool_undirty_space(dmu_objset_pool(os),
	dr->dr_accounted % zio->io_phys_children, zio->io_txg);
	}

	kmem_free(dr, sizeof (dbuf_dirty_record_t));
	}

	static void
	dbuf_write_nofill_ready(zio_t *zio)
	{
	dbuf_write_ready(zio, NULL, zio->io_private);
	}

	static void
	dbuf_write_nofill_done(zio_t *zio)
	{
	dbuf_write_done(zio, NULL, zio->io_private);
	}

	static void
	dbuf_write_override_ready(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	dmu_buf_impl_t *db = dr->dr_dbuf;

	dbuf_write_ready(zio, NULL, db);
	}

	static void
	dbuf_write_override_done(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	blkptr_t *obp = &dr->dt.dl.dr_overridden_by;

	mutex_enter(&db->db_mtx);
	if (!BP_EQUAL(zio->io_bp, obp)) {
	if (!BP_IS_HOLE(obp))
	dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp);
	arc_release(dr->dt.dl.dr_data, db);
	}
	mutex_exit(&db->db_mtx);

	dbuf_write_done(zio, NULL, db);

	if (zio->io_abd != NULL)
	abd_free(zio->io_abd);
	}

	typedef struct dbuf_remap_impl_callback_arg {
	objset_t *drica_os;
	uint64_t drica_blk_birth;
	dmu_tx_t *drica_tx;
	} dbuf_remap_impl_callback_arg_t;

	static void
	dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size,
	void *arg)
	{
	dbuf_remap_impl_callback_arg_t *drica = arg;
	objset_t *os = drica->drica_os;
	spa_t *spa = dmu_objset_spa(os);
	dmu_tx_t *tx = drica->drica_tx;

	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	if (os == spa_meta_objset(spa)) {
	spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx);
	} else {
	dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset,
	size, drica->drica_blk_birth, tx);
	}
	}

	static void
	dbuf_remap_impl(dnode_t dn, blkptr_t bp, krwlock_t rw, dmu_tx_t tx)
	{
	blkptr_t bp_copy = *bp;
	spa_t *spa = dmu_objset_spa(dn->dn_objset);
	dbuf_remap_impl_callback_arg_t drica;

	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	drica.drica_os = dn->dn_objset;
	drica.drica_blk_birth = bp->blk_birth;
	drica.drica_tx = tx;
	if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback,
	&drica)) {
	/*
	* If the blkptr being remapped is tracked by a livelist,
	* then we need to make sure the livelist reflects the update.
	* First, cancel out the old blkptr by appending a 'FREE'
	* entry. Next, add an 'ALLOC' to track the new version. This
	* way we avoid trying to free an inaccurate blkptr at delete.
	* Note that embedded blkptrs are not tracked in livelists.
	*/
	if (dn->dn_objset != spa_meta_objset(spa)) {
	dsl_dataset_t *ds = dmu_objset_ds(dn->dn_objset);
	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
	bp->blk_birth > ds->ds_dir->dd_origin_txg) {
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(dsl_dir_is_clone(ds->ds_dir));
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LIVELIST));
	bplist_append(&ds->ds_dir->dd_pending_frees,
	bp);
	bplist_append(&ds->ds_dir->dd_pending_allocs,
	&bp_copy);
	}
	}

	/*
	* The db_rwlock prevents dbuf_read_impl() from
	* dereferencing the BP while we are changing it. To
	* avoid lock contention, only grab it when we are actually
	* changing the BP.
	*/
	if (rw != NULL)
	rw_enter(rw, RW_WRITER);
	*bp = bp_copy;
	if (rw != NULL)
	rw_exit(rw);
	}
	}

	/*
	* Remap any existing BP's to concrete vdevs, if possible.
	*/
	static void
	dbuf_remap(dnode_t dn, dmu_buf_impl_t db, dmu_tx_t *tx)
	{
	spa_t *spa = dmu_objset_spa(db->db_objset);
	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL))
	return;

	if (db->db_level > 0) {
	blkptr_t *bp = db->db.db_data;
	for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) {
	dbuf_remap_impl(dn, &bp[i], &db->db_rwlock, tx);
	}
	} else if (db->db.db_object == DMU_META_DNODE_OBJECT) {
	dnode_phys_t *dnp = db->db.db_data;
	ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==,
	DMU_OT_DNODE);
	for (int i = 0; i < db->db.db_size >> DNODE_SHIFT;
	i += dnp[i].dn_extra_slots + 1) {
	for (int j = 0; j < dnp[i].dn_nblkptr; j++) {
	krwlock_t *lock = (dn->dn_dbuf == NULL ? NULL :
	&dn->dn_dbuf->db_rwlock);
	dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], lock,
	tx);
	}
	}
	}
	}


	/* Issue I/O to commit a dirty buffer to disk. */
	static void
	dbuf_write(dbuf_dirty_record_t dr, arc_buf_t data, dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;
	objset_t *os;
	dmu_buf_impl_t *parent = db->db_parent;
	uint64_t txg = tx->tx_txg;
	zbookmark_phys_t zb;
	zio_prop_t zp;
	zio_t pio; / parent I/O */
	int wp_flag = 0;

	ASSERT(dmu_tx_is_syncing(tx));

	os = dn->dn_objset;

	if (db->db_state != DB_NOFILL) {
	if (db->db_level > 0 \|\| dn->dn_type == DMU_OT_DNODE) {
	/*
	* Private object buffers are released here rather
	* than in dbuf_dirty() since they are only modified
	* in the syncing context and we don't want the
	* overhead of making multiple copies of the data.
	*/
	if (BP_IS_HOLE(db->db_blkptr)) {
	arc_buf_thaw(data);
	} else {
	dbuf_release_bp(db);
	}
	dbuf_remap(dn, db, tx);
	}
	}

	if (parent != dn->dn_dbuf) {
	/* Our parent is an indirect block. */
	/* We have a dirty parent that has been scheduled for write. */
	ASSERT(parent && parent->db_data_pending);
	/* Our parent's buffer is one level closer to the dnode. */
	ASSERT(db->db_level == parent->db_level-1);
	/*
	* We're about to modify our parent's db_data by modifying
	* our block pointer, so the parent must be released.
	*/
	ASSERT(arc_released(parent->db_buf));
	pio = parent->db_data_pending->dr_zio;
	} else {
	/* Our parent is the dnode itself. */
	ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 &&
	db->db_blkid != DMU_SPILL_BLKID) \|\|
	(db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0));
	if (db->db_blkid != DMU_SPILL_BLKID)
	ASSERT3P(db->db_blkptr, ==,
	&dn->dn_phys->dn_blkptr[db->db_blkid]);
	pio = dn->dn_zio;
	}

	ASSERT(db->db_level == 0 \|\| data == db->db_buf);
	ASSERT3U(db->db_blkptr->blk_birth, <=, txg);
	ASSERT(pio);

	SET_BOOKMARK(&zb, os->os_dsl_dataset ?
	os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
	db->db.db_object, db->db_level, db->db_blkid);

	if (db->db_blkid == DMU_SPILL_BLKID)
	wp_flag = WP_SPILL;
	wp_flag \|= (db->db_state == DB_NOFILL) ? WP_NOFILL : 0;

	dmu_write_policy(os, dn, db->db_level, wp_flag, &zp);

	/*
	* We copy the blkptr now (rather than when we instantiate the dirty
	* record), because its value can change between open context and
	* syncing context. We do not need to hold dn_struct_rwlock to read
	* db_blkptr because we are in syncing context.
	*/
	dr->dr_bp_copy = *db->db_blkptr;

	if (db->db_level == 0 &&
	dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
	/*
	* The BP for this block has been provided by open context
	* (by dmu_sync() or dmu_buf_write_embedded()).
	*/
	abd_t *contents = (data != NULL) ?
	abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL;

	dr->dr_zio = zio_write(pio, os->os_spa, txg, &dr->dr_bp_copy,
	contents, db->db.db_size, db->db.db_size, &zp,
	dbuf_write_override_ready, NULL, NULL,
	dbuf_write_override_done,
	dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
	mutex_enter(&db->db_mtx);
	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by,
	dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite);
	mutex_exit(&db->db_mtx);
	} else if (db->db_state == DB_NOFILL) {
	ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF \|\|
	zp.zp_checksum == ZIO_CHECKSUM_NOPARITY);
	dr->dr_zio = zio_write(pio, os->os_spa, txg,
	&dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp,
	dbuf_write_nofill_ready, NULL, NULL,
	dbuf_write_nofill_done, db,
	ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_MUSTSUCCEED \| ZIO_FLAG_NODATA, &zb);
	} else {
	ASSERT(arc_released(data));

	/*
	* For indirect blocks, we want to setup the children
	* ready callback so that we can properly handle an indirect
	* block that only contains holes.
	*/
	arc_write_done_func_t *children_ready_cb = NULL;
	if (db->db_level != 0)
	children_ready_cb = dbuf_write_children_ready;

	dr->dr_zio = arc_write(pio, os->os_spa, txg,
	&dr->dr_bp_copy, data, dbuf_is_l2cacheable(db),
	&zp, dbuf_write_ready,
	children_ready_cb, dbuf_write_physdone,
	dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_MUSTSUCCEED, &zb);
	}
	}

	EXPORT_SYMBOL(dbuf_find);
	EXPORT_SYMBOL(dbuf_is_metadata);
	EXPORT_SYMBOL(dbuf_destroy);
	EXPORT_SYMBOL(dbuf_loan_arcbuf);
	EXPORT_SYMBOL(dbuf_whichblock);
	EXPORT_SYMBOL(dbuf_read);
	EXPORT_SYMBOL(dbuf_unoverride);
	EXPORT_SYMBOL(dbuf_free_range);
	EXPORT_SYMBOL(dbuf_new_size);
	EXPORT_SYMBOL(dbuf_release_bp);
	EXPORT_SYMBOL(dbuf_dirty);
	EXPORT_SYMBOL(dmu_buf_set_crypt_params);
	EXPORT_SYMBOL(dmu_buf_will_dirty);
	EXPORT_SYMBOL(dmu_buf_is_dirty);
	EXPORT_SYMBOL(dmu_buf_will_not_fill);
	EXPORT_SYMBOL(dmu_buf_will_fill);
	EXPORT_SYMBOL(dmu_buf_fill_done);
	EXPORT_SYMBOL(dmu_buf_rele);
	EXPORT_SYMBOL(dbuf_assign_arcbuf);
	EXPORT_SYMBOL(dbuf_prefetch);
	EXPORT_SYMBOL(dbuf_hold_impl);
	EXPORT_SYMBOL(dbuf_hold);
	EXPORT_SYMBOL(dbuf_hold_level);
	EXPORT_SYMBOL(dbuf_create_bonus);
	EXPORT_SYMBOL(dbuf_spill_set_blksz);
	EXPORT_SYMBOL(dbuf_rm_spill);
	EXPORT_SYMBOL(dbuf_add_ref);
	EXPORT_SYMBOL(dbuf_rele);
	EXPORT_SYMBOL(dbuf_rele_and_unlock);
	EXPORT_SYMBOL(dbuf_refcount);
	EXPORT_SYMBOL(dbuf_sync_list);
	EXPORT_SYMBOL(dmu_buf_set_user);
	EXPORT_SYMBOL(dmu_buf_set_user_ie);
	EXPORT_SYMBOL(dmu_buf_get_user);
	EXPORT_SYMBOL(dmu_buf_get_blkptr);

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, max_bytes, ULONG, ZMOD_RW,
	"Maximum size in bytes of the dbuf cache.");

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, hiwater_pct, UINT, ZMOD_RW,
	"Percentage over dbuf_cache_max_bytes for direct dbuf eviction.");

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, lowater_pct, UINT, ZMOD_RW,
	"Percentage below dbuf_cache_max_bytes when dbuf eviction stops.");

	ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_max_bytes, ULONG, ZMOD_RW,
	"Maximum size in bytes of dbuf metadata cache.");

	-ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, cache_shift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, cache_shift, UINT, ZMOD_RW,
	"Set size of dbuf cache to log2 fraction of arc size.");

	-ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_shift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_shift, UINT, ZMOD_RW,
	"Set size of dbuf metadata cache to log2 fraction of arc size.");
	+
	+ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, mutex_cache_shift, UINT, ZMOD_RD,
	+ "Set size of dbuf cache mutex array as log2 shift.");
	diff --git a/sys/contrib/openzfs/module/zfs/dbuf_stats.c b/sys/contrib/openzfs/module/zfs/dbuf_stats.c
	index 747fc337db48..e5dc2df30774 100644
	--- a/sys/contrib/openzfs/module/zfs/dbuf_stats.c
	+++ b/sys/contrib/openzfs/module/zfs/dbuf_stats.c
	@@ -1,230 +1,230 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	#include <sys/zfs_context.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_objset.h>

	/*
	* Calculate the index of the arc header for the state, disabled by default.
	*/
	int zfs_dbuf_state_index = 0;

	/*
	* ==========================================================================
	* Dbuf Hash Read Routines
	* ==========================================================================
	*/
	typedef struct dbuf_stats_t {
	kmutex_t lock;
	kstat_t *kstat;
	dbuf_hash_table_t *hash;
	int idx;
	} dbuf_stats_t;

	static dbuf_stats_t dbuf_stats_hash_table;

	static int
	dbuf_stats_hash_table_headers(char *buf, size_t size)
	{
	(void) snprintf(buf, size,
	"%-96s \| %-119s \| %s\n"
	"%-16s %-8s %-8s %-8s %-8s %-10s %-8s %-5s %-5s %-7s %3s \| "
	"%-5s %-5s %-9s %-6s %-8s %-12s "
	"%-6s %-6s %-6s %-6s %-6s %-8s %-8s %-8s %-6s \| "
	"%-6s %-6s %-8s %-8s %-6s %-6s %-6s %-8s %-8s\n",
	"dbuf", "arcbuf", "dnode", "pool", "objset", "object", "level",
	"blkid", "offset", "dbsize", "meta", "state", "dbholds", "dbc",
	"list", "atype", "flags", "count", "asize", "access",
	"mru", "gmru", "mfu", "gmfu", "l2", "l2_dattr", "l2_asize",
	"l2_comp", "aholds", "dtype", "btype", "data_bs", "meta_bs",
	"bsize", "lvls", "dholds", "blocks", "dsize");

	return (0);
	}

	static int
	__dbuf_stats_hash_table_data(char buf, size_t size, dmu_buf_impl_t db)
	{
	arc_buf_info_t abi = { 0 };
	dmu_object_info_t doi = { 0 };
	dnode_t *dn = DB_DNODE(db);
	size_t nwritten;

	if (db->db_buf)
	arc_buf_info(db->db_buf, &abi, zfs_dbuf_state_index);

	__dmu_object_info_from_dnode(dn, &doi);

	nwritten = snprintf(buf, size,
	"%-16s %-8llu %-8lld %-8lld %-8lld %-10llu %-8llu %-5d %-5d "
	"%-7lu %-3d \| %-5d %-5d 0x%-7x %-6lu %-8llu %-12llu "
	"%-6lu %-6lu %-6lu %-6lu %-6lu %-8llu %-8llu %-8d %-6lu \| "
	"%-6d %-6d %-8lu %-8lu %-6llu %-6lu %-6lu %-8llu %-8llu\n",
	/* dmu_buf_impl_t */
	spa_name(dn->dn_objset->os_spa),
	(u_longlong_t)dmu_objset_id(db->db_objset),
	(longlong_t)db->db.db_object,
	(longlong_t)db->db_level,
	(longlong_t)db->db_blkid,
	(u_longlong_t)db->db.db_offset,
	(u_longlong_t)db->db.db_size,
	!!dbuf_is_metadata(db),
	db->db_state,
	(ulong_t)zfs_refcount_count(&db->db_holds),
	multilist_link_active(&db->db_cache_link),
	/* arc_buf_info_t */
	abi.abi_state_type,
	abi.abi_state_contents,
	abi.abi_flags,
	(ulong_t)abi.abi_bufcnt,
	(u_longlong_t)abi.abi_size,
	(u_longlong_t)abi.abi_access,
	(ulong_t)abi.abi_mru_hits,
	(ulong_t)abi.abi_mru_ghost_hits,
	(ulong_t)abi.abi_mfu_hits,
	(ulong_t)abi.abi_mfu_ghost_hits,
	(ulong_t)abi.abi_l2arc_hits,
	(u_longlong_t)abi.abi_l2arc_dattr,
	(u_longlong_t)abi.abi_l2arc_asize,
	abi.abi_l2arc_compress,
	(ulong_t)abi.abi_holds,
	/* dmu_object_info_t */
	doi.doi_type,
	doi.doi_bonus_type,
	(ulong_t)doi.doi_data_block_size,
	(ulong_t)doi.doi_metadata_block_size,
	(u_longlong_t)doi.doi_bonus_size,
	(ulong_t)doi.doi_indirection,
	(ulong_t)zfs_refcount_count(&dn->dn_holds),
	(u_longlong_t)doi.doi_fill_count,
	(u_longlong_t)doi.doi_max_offset);

	if (nwritten >= size)
	return (size);

	return (nwritten + 1);
	}

	static int
	dbuf_stats_hash_table_data(char buf, size_t size, void data)
	{
	dbuf_stats_t dsh = (dbuf_stats_t )data;
	dbuf_hash_table_t *h = dsh->hash;
	dmu_buf_impl_t *db;
	int length, error = 0;

	ASSERT3S(dsh->idx, >=, 0);
	ASSERT3S(dsh->idx, <=, h->hash_table_mask);
	if (size)
	buf[0] = 0;

	- rw_enter(DBUF_HASH_RWLOCK(h, dsh->idx), RW_READER);
	+ mutex_enter(DBUF_HASH_MUTEX(h, dsh->idx));
	for (db = h->hash_table[dsh->idx]; db != NULL; db = db->db_hash_next) {
	/*
	* Returning ENOMEM will cause the data and header functions
	* to be called with a larger scratch buffers.
	*/
	if (size < 512) {
	error = SET_ERROR(ENOMEM);
	break;
	}

	mutex_enter(&db->db_mtx);

	if (db->db_state != DB_EVICTING) {
	length = __dbuf_stats_hash_table_data(buf, size, db);
	buf += length;
	size -= length;
	}

	mutex_exit(&db->db_mtx);
	}
	- rw_exit(DBUF_HASH_RWLOCK(h, dsh->idx));
	+ mutex_exit(DBUF_HASH_MUTEX(h, dsh->idx));

	return (error);
	}

	static void *
	dbuf_stats_hash_table_addr(kstat_t *ksp, loff_t n)
	{
	dbuf_stats_t *dsh = ksp->ks_private;

	ASSERT(MUTEX_HELD(&dsh->lock));

	if (n <= dsh->hash->hash_table_mask) {
	dsh->idx = n;
	return (dsh);
	}

	return (NULL);
	}

	static void
	dbuf_stats_hash_table_init(dbuf_hash_table_t *hash)
	{
	dbuf_stats_t *dsh = &dbuf_stats_hash_table;
	kstat_t *ksp;

	mutex_init(&dsh->lock, NULL, MUTEX_DEFAULT, NULL);
	dsh->hash = hash;

	ksp = kstat_create("zfs", 0, "dbufs", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
	dsh->kstat = ksp;

	if (ksp) {
	ksp->ks_lock = &dsh->lock;
	ksp->ks_ndata = UINT32_MAX;
	ksp->ks_private = dsh;
	kstat_set_raw_ops(ksp, dbuf_stats_hash_table_headers,
	dbuf_stats_hash_table_data, dbuf_stats_hash_table_addr);
	kstat_install(ksp);
	}
	}

	static void
	dbuf_stats_hash_table_destroy(void)
	{
	dbuf_stats_t *dsh = &dbuf_stats_hash_table;
	kstat_t *ksp;

	ksp = dsh->kstat;
	if (ksp)
	kstat_delete(ksp);

	mutex_destroy(&dsh->lock);
	}

	void
	dbuf_stats_init(dbuf_hash_table_t *hash)
	{
	dbuf_stats_hash_table_init(hash);
	}

	void
	dbuf_stats_destroy(void)
	{
	dbuf_stats_hash_table_destroy();
	}

	ZFS_MODULE_PARAM(zfs, zfs_, dbuf_state_index, INT, ZMOD_RW,
	"Calculate arc header index");
	diff --git a/sys/contrib/openzfs/module/zfs/dmu.c b/sys/contrib/openzfs/module/zfs/dmu.c
	index dc6e87cd19fb..9e67eb51f415 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu.c
	@@ -1,2366 +1,2366 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013, Joyent, Inc. All rights reserved.
	* Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_prop.h>
	#include <sys/dmu_zfetch.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/sa.h>
	#include <sys/zfeature.h>
	#include <sys/abd.h>
	#include <sys/trace_zfs.h>
	#include <sys/zfs_racct.h>
	#include <sys/zfs_rlock.h>
	#ifdef _KERNEL
	#include <sys/vmsystm.h>
	#include <sys/zfs_znode.h>
	#endif

	/*
	* Enable/disable nopwrite feature.
	*/
	static int zfs_nopwrite_enabled = 1;

	/*
	* Tunable to control percentage of dirtied L1 blocks from frees allowed into
	* one TXG. After this threshold is crossed, additional dirty blocks from frees
	* will wait until the next TXG.
	* A value of zero will disable this throttle.
	*/
	static unsigned long zfs_per_txg_dirty_frees_percent = 30;

	/*
	* Enable/disable forcing txg sync when dirty checking for holes with lseek().
	* By default this is enabled to ensure accurate hole reporting, it can result
	* in a significant performance penalty for lseek(SEEK_HOLE) heavy workloads.
	* Disabling this option will result in holes never being reported in dirty
	* files which is always safe.
	*/
	static int zfs_dmu_offset_next_sync = 1;

	/*
	* Limit the amount we can prefetch with one call to this amount. This
	* helps to limit the amount of memory that can be used by prefetching.
	* Larger objects should be prefetched a bit at a time.
	*/
	-int dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;
	+uint_t dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;

	const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "unallocated" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "object directory" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "object array" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "packed nvlist" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "packed nvlist size" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj header" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map header" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, TRUE, "ZIL intent log" },
	{DMU_BSWAP_DNODE, TRUE, FALSE, TRUE, "DMU dnode" },
	{DMU_BSWAP_OBJSET, TRUE, TRUE, FALSE, "DMU objset" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL directory" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL directory child map"},
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset snap map" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL props" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL dataset" },
	{DMU_BSWAP_ZNODE, TRUE, FALSE, FALSE, "ZFS znode" },
	{DMU_BSWAP_OLDACL, TRUE, FALSE, TRUE, "ZFS V0 ACL" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "ZFS plain file" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS directory" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "ZFS master node" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS delete queue" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "zvol object" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "zvol prop" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "other uint8[]" },
	{DMU_BSWAP_UINT64, FALSE, FALSE, TRUE, "other uint64[]" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "other ZAP" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "persistent error log" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "SPA history" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA history offsets" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "Pool properties" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL permissions" },
	{DMU_BSWAP_ACL, TRUE, FALSE, TRUE, "ZFS ACL" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "ZFS SYSACL" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "FUID table" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "FUID table size" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset next clones"},
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan work queue" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project used" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project quota"},
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "snapshot refcount tags"},
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT ZAP algorithm" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT statistics" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "System attributes" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA master node" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr registration" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr layouts" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan translations" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "deduplicated block" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL deadlist map" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL deadlist map hdr" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dir clones" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj subobj" }
	};

	-const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
	+dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
	{ byteswap_uint8_array, "uint8" },
	{ byteswap_uint16_array, "uint16" },
	{ byteswap_uint32_array, "uint32" },
	{ byteswap_uint64_array, "uint64" },
	{ zap_byteswap, "zap" },
	{ dnode_buf_byteswap, "dnode" },
	{ dmu_objset_byteswap, "objset" },
	{ zfs_znode_byteswap, "znode" },
	{ zfs_oldacl_byteswap, "oldacl" },
	{ zfs_acl_byteswap, "acl" }
	};

	static int
	dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
	const void tag, dmu_buf_t *dbp)
	{
	uint64_t blkid;
	dmu_buf_impl_t *db;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, tag);
	rw_exit(&dn->dn_struct_rwlock);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}

	*dbp = &db->db;
	return (0);
	}
	int
	dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
	const void tag, dmu_buf_t *dbp)
	{
	dnode_t *dn;
	uint64_t blkid;
	dmu_buf_impl_t *db;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, tag);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}

	*dbp = &db->db;
	return (err);
	}

	int
	dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
	const void tag, dmu_buf_t *dbp, int flags)
	{
	int err;
	int db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	err = dmu_buf_hold_noread_by_dnode(dn, offset, tag, dbp);
	if (err == 0) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )(*dbp);
	err = dbuf_read(db, NULL, db_flags);
	if (err != 0) {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	}

	return (err);
	}

	int
	dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
	const void tag, dmu_buf_t *dbp, int flags)
	{
	int err;
	int db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	err = dmu_buf_hold_noread(os, object, offset, tag, dbp);
	if (err == 0) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )(*dbp);
	err = dbuf_read(db, NULL, db_flags);
	if (err != 0) {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	}

	return (err);
	}

	int
	dmu_bonus_max(void)
	{
	return (DN_OLD_MAX_BONUSLEN);
	}

	int
	dmu_set_bonus(dmu_buf_t db_fake, int newsize, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int error;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (dn->dn_bonus != db) {
	error = SET_ERROR(EINVAL);
	} else if (newsize < 0 \|\| newsize > db_fake->db_size) {
	error = SET_ERROR(EINVAL);
	} else {
	dnode_setbonuslen(dn, newsize, tx);
	error = 0;
	}

	DB_DNODE_EXIT(db);
	return (error);
	}

	int
	dmu_set_bonustype(dmu_buf_t db_fake, dmu_object_type_t type, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int error;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (!DMU_OT_IS_VALID(type)) {
	error = SET_ERROR(EINVAL);
	} else if (dn->dn_bonus != db) {
	error = SET_ERROR(EINVAL);
	} else {
	dnode_setbonus_type(dn, type, tx);
	error = 0;
	}

	DB_DNODE_EXIT(db);
	return (error);
	}

	dmu_object_type_t
	dmu_get_bonustype(dmu_buf_t *db_fake)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	dmu_object_type_t type;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	type = dn->dn_bonustype;
	DB_DNODE_EXIT(db);

	return (type);
	}

	int
	dmu_rm_spill(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int error;

	error = dnode_hold(os, object, FTAG, &dn);
	dbuf_rm_spill(dn, tx);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	dnode_rm_spill(dn, tx);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	return (error);
	}

	/*
	* Lookup and hold the bonus buffer for the provided dnode. If the dnode
	* has not yet been allocated a new bonus dbuf a will be allocated.
	* Returns ENOENT, EIO, or 0.
	*/
	int dmu_bonus_hold_by_dnode(dnode_t dn, const void tag, dmu_buf_t **dbp,
	uint32_t flags)
	{
	dmu_buf_impl_t *db;
	int error;
	uint32_t db_flags = DB_RF_MUST_SUCCEED;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_bonus == NULL) {
	rw_exit(&dn->dn_struct_rwlock);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	if (dn->dn_bonus == NULL)
	dbuf_create_bonus(dn);
	}
	db = dn->dn_bonus;

	/* as long as the bonus buf is held, the dnode will be held */
	if (zfs_refcount_add(&db->db_holds, tag) == 1) {
	VERIFY(dnode_add_ref(dn, db));
	atomic_inc_32(&dn->dn_dbufs_count);
	}

	/*
	* Wait to drop dn_struct_rwlock until after adding the bonus dbuf's
	* hold and incrementing the dbuf count to ensure that dnode_move() sees
	* a dnode hold for every dbuf.
	*/
	rw_exit(&dn->dn_struct_rwlock);

	error = dbuf_read(db, NULL, db_flags);
	if (error) {
	dnode_evict_bonus(dn);
	dbuf_rele(db, tag);
	*dbp = NULL;
	return (error);
	}

	*dbp = &db->db;
	return (0);
	}

	int
	dmu_bonus_hold(objset_t os, uint64_t object, const void tag, dmu_buf_t **dbp)
	{
	dnode_t *dn;
	int error;

	error = dnode_hold(os, object, FTAG, &dn);
	if (error)
	return (error);

	error = dmu_bonus_hold_by_dnode(dn, tag, dbp, DMU_READ_NO_PREFETCH);
	dnode_rele(dn, FTAG);

	return (error);
	}

	/*
	* returns ENOENT, EIO, or 0.
	*
	* This interface will allocate a blank spill dbuf when a spill blk
	* doesn't already exist on the dnode.
	*
	* if you only want to find an already existing spill db, then
	* dmu_spill_hold_existing() should be used.
	*/
	int
	dmu_spill_hold_by_dnode(dnode_t dn, uint32_t flags, const void tag,
	dmu_buf_t **dbp)
	{
	dmu_buf_impl_t *db = NULL;
	int err;

	if ((flags & DB_RF_HAVESTRUCT) == 0)
	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	db = dbuf_hold(dn, DMU_SPILL_BLKID, tag);

	if ((flags & DB_RF_HAVESTRUCT) == 0)
	rw_exit(&dn->dn_struct_rwlock);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}
	err = dbuf_read(db, NULL, flags);
	if (err == 0)
	*dbp = &db->db;
	else {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	return (err);
	}

	int
	dmu_spill_hold_existing(dmu_buf_t bonus, const void tag, dmu_buf_t **dbp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )bonus;
	dnode_t *dn;
	int err;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) {
	err = SET_ERROR(EINVAL);
	} else {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	if (!dn->dn_have_spill) {
	err = SET_ERROR(ENOENT);
	} else {
	err = dmu_spill_hold_by_dnode(dn,
	DB_RF_HAVESTRUCT \| DB_RF_CANFAIL, tag, dbp);
	}

	rw_exit(&dn->dn_struct_rwlock);
	}

	DB_DNODE_EXIT(db);
	return (err);
	}

	int
	dmu_spill_hold_by_bonus(dmu_buf_t bonus, uint32_t flags, const void tag,
	dmu_buf_t **dbp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )bonus;
	dnode_t *dn;
	int err;
	uint32_t db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_spill_hold_by_dnode(dn, db_flags, tag, dbp);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Note: longer-term, we should modify all of the dmu_buf_*() interfaces
	* to take a held dnode rather than <os, object> -- the lookup is wasteful,
	* and can induce severe lock contention when writing to several files
	* whose dnodes are in the same block.
	*/
	int
	dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
	boolean_t read, const void tag, int numbufsp, dmu_buf_t ***dbpp,
	uint32_t flags)
	{
	dmu_buf_t **dbp;
	zstream_t *zs = NULL;
	uint64_t blkid, nblks, i;
	uint32_t dbuf_flags;
	int err;
	zio_t *zio = NULL;
	boolean_t missed = B_FALSE;

	ASSERT(length <= DMU_MAX_ACCESS);

	/*
	* Note: We directly notify the prefetch code of this read, so that
	* we can tell it about the multi-block read. dbuf_read() only knows
	* about the one block it is accessing.
	*/
	dbuf_flags = DB_RF_CANFAIL \| DB_RF_NEVERWAIT \| DB_RF_HAVESTRUCT \|
	DB_RF_NOPREFETCH;

	if ((flags & DMU_READ_NO_DECRYPT) != 0)
	dbuf_flags \|= DB_RF_NO_DECRYPT;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_datablkshift) {
	int blkshift = dn->dn_datablkshift;
	nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) -
	P2ALIGN(offset, 1ULL << blkshift)) >> blkshift;
	} else {
	if (offset + length > dn->dn_datablksz) {
	zfs_panic_recover("zfs: accessing past end of object "
	"%llx/%llx (size=%u access=%llu+%llu)",
	(longlong_t)dn->dn_objset->
	os_dsl_dataset->ds_object,
	(longlong_t)dn->dn_object, dn->dn_datablksz,
	(longlong_t)offset, (longlong_t)length);
	rw_exit(&dn->dn_struct_rwlock);
	return (SET_ERROR(EIO));
	}
	nblks = 1;
	}
	dbp = kmem_zalloc(sizeof (dmu_buf_t ) nblks, KM_SLEEP);

	if (read)
	zio = zio_root(dn->dn_objset->os_spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);
	blkid = dbuf_whichblock(dn, 0, offset);
	if ((flags & DMU_READ_NO_PREFETCH) == 0 &&
	DNODE_META_IS_CACHEABLE(dn) && length <= zfetch_array_rd_sz) {
	/*
	* Prepare the zfetch before initiating the demand reads, so
	* that if multiple threads block on same indirect block, we
	* base predictions on the original less racy request order.
	*/
	zs = dmu_zfetch_prepare(&dn->dn_zfetch, blkid, nblks,
	read && DNODE_IS_CACHEABLE(dn), B_TRUE);
	}
	for (i = 0; i < nblks; i++) {
	dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag);
	if (db == NULL) {
	if (zs)
	dmu_zfetch_run(zs, missed, B_TRUE);
	rw_exit(&dn->dn_struct_rwlock);
	dmu_buf_rele_array(dbp, nblks, tag);
	if (read)
	zio_nowait(zio);
	return (SET_ERROR(EIO));
	}

	/*
	* Initiate async demand data read.
	* We check the db_state after calling dbuf_read() because
	* (1) dbuf_read() may change the state to CACHED due to a
	* hit in the ARC, and (2) on a cache miss, a child will
	* have been added to "zio" but not yet completed, so the
	* state will not yet be CACHED.
	*/
	if (read) {
	(void) dbuf_read(db, zio, dbuf_flags);
	if (db->db_state != DB_CACHED)
	missed = B_TRUE;
	}
	dbp[i] = &db->db;
	}

	if (!read)
	zfs_racct_write(length, nblks);

	if (zs)
	dmu_zfetch_run(zs, missed, B_TRUE);
	rw_exit(&dn->dn_struct_rwlock);

	if (read) {
	/* wait for async read i/o */
	err = zio_wait(zio);
	if (err) {
	dmu_buf_rele_array(dbp, nblks, tag);
	return (err);
	}

	/* wait for other io to complete */
	for (i = 0; i < nblks; i++) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbp[i];
	mutex_enter(&db->db_mtx);
	while (db->db_state == DB_READ \|\|
	db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);
	if (db->db_state == DB_UNCACHED)
	err = SET_ERROR(EIO);
	mutex_exit(&db->db_mtx);
	if (err) {
	dmu_buf_rele_array(dbp, nblks, tag);
	return (err);
	}
	}
	}

	*numbufsp = nblks;
	*dbpp = dbp;
	return (0);
	}

	int
	dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t length, int read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
	numbufsp, dbpp, DMU_READ_PREFETCH);

	dnode_rele(dn, FTAG);

	return (err);
	}

	int
	dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset,
	uint64_t length, boolean_t read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int err;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
	numbufsp, dbpp, DMU_READ_PREFETCH);
	DB_DNODE_EXIT(db);

	return (err);
	}

	void
	dmu_buf_rele_array(dmu_buf_t *dbp_fake, int numbufs, const void tag)
	{
	int i;
	dmu_buf_impl_t dbp = (dmu_buf_impl_t )dbp_fake;

	if (numbufs == 0)
	return;

	for (i = 0; i < numbufs; i++) {
	if (dbp[i])
	dbuf_rele(dbp[i], tag);
	}

	kmem_free(dbp, sizeof (dmu_buf_t ) numbufs);
	}

	/*
	* Issue prefetch i/os for the given blocks. If level is greater than 0, the
	* indirect blocks prefetched will be those that point to the blocks containing
	* the data starting at offset, and continuing to offset + len.
	*
	* Note that if the indirect blocks above the blocks being prefetched are not
	* in cache, they will be asynchronously read in.
	*/
	void
	dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
	uint64_t len, zio_priority_t pri)
	{
	dnode_t *dn;
	uint64_t blkid;
	int nblks, err;

	if (len == 0) { /* they're interested in the bonus buffer */
	dn = DMU_META_DNODE(os);

	if (object == 0 \|\| object >= DN_MAX_OBJECT)
	return;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, level,
	object * sizeof (dnode_phys_t));
	dbuf_prefetch(dn, level, blkid, pri, 0);
	rw_exit(&dn->dn_struct_rwlock);
	return;
	}

	/*
	* See comment before the definition of dmu_prefetch_max.
	*/
	len = MIN(len, dmu_prefetch_max);

	/*
	* XXX - Note, if the dnode for the requested object is not
	* already cached, we will do a synchronous read in the
	* dnode_hold() call. The same is true for any indirects.
	*/
	err = dnode_hold(os, object, FTAG, &dn);
	if (err != 0)
	return;

	/*
	* offset + len - 1 is the last byte we want to prefetch for, and offset
	* is the first. Then dbuf_whichblk(dn, level, off + len - 1) is the
	* last block we want to prefetch, and dbuf_whichblock(dn, level,
	* offset) is the first. Then the number we need to prefetch is the
	* last - first + 1.
	*/
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (level > 0 \|\| dn->dn_datablkshift != 0) {
	nblks = dbuf_whichblock(dn, level, offset + len - 1) -
	dbuf_whichblock(dn, level, offset) + 1;
	} else {
	nblks = (offset < dn->dn_datablksz);
	}

	if (nblks != 0) {
	blkid = dbuf_whichblock(dn, level, offset);
	for (int i = 0; i < nblks; i++)
	dbuf_prefetch(dn, level, blkid + i, pri, 0);
	}
	rw_exit(&dn->dn_struct_rwlock);

	dnode_rele(dn, FTAG);
	}

	/*
	* Get the next "chunk" of file data to free. We traverse the file from
	* the end so that the file gets shorter over time (if we crashes in the
	* middle, this will leave us in a better state). We find allocated file
	* data by simply searching the allocated level 1 indirects.
	*
	* On input, *start should be the first offset that does not need to be
	* freed (e.g. "offset + length"). On return, *start will be the first
	* offset that should be freed and l1blks is set to the number of level 1
	* indirect blocks found within the chunk.
	*/
	static int
	get_next_chunk(dnode_t dn, uint64_t start, uint64_t minimum, uint64_t *l1blks)
	{
	uint64_t blks;
	uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1);
	/* bytes of data covered by a level-1 indirect block */
	uint64_t iblkrange = (uint64_t)dn->dn_datablksz *
	EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);

	ASSERT3U(minimum, <=, *start);

	/*
	* Check if we can free the entire range assuming that all of the
	* L1 blocks in this range have data. If we can, we use this
	* worst case value as an estimate so we can avoid having to look
	* at the object's actual data.
	*/
	uint64_t total_l1blks =
	(roundup(start, iblkrange) - (minimum / iblkrange iblkrange)) /
	iblkrange;
	if (total_l1blks <= maxblks) {
	*l1blks = total_l1blks;
	*start = minimum;
	return (0);
	}
	ASSERT(ISP2(iblkrange));

	for (blks = 0; *start > minimum && blks < maxblks; blks++) {
	int err;

	/*
	* dnode_next_offset(BACKWARDS) will find an allocated L1
	* indirect block at or before the input offset. We must
	* decrement *start so that it is at the end of the region
	* to search.
	*/
	(*start)--;

	err = dnode_next_offset(dn,
	DNODE_FIND_BACKWARDS, start, 2, 1, 0);

	/* if there are no indirect blocks before start, we are done */
	if (err == ESRCH) {
	*start = minimum;
	break;
	} else if (err != 0) {
	*l1blks = blks;
	return (err);
	}

	/* set start to the beginning of this L1 indirect */
	start = P2ALIGN(start, iblkrange);
	}
	if (*start < minimum)
	*start = minimum;
	*l1blks = blks;

	return (0);
	}

	/*
	* If this objset is of type OST_ZFS return true if vfs's unmounted flag is set,
	* otherwise return false.
	* Used below in dmu_free_long_range_impl() to enable abort when unmounting
	*/
	static boolean_t
	dmu_objset_zfs_unmounting(objset_t *os)
	{
	#ifdef _KERNEL
	if (dmu_objset_type(os) == DMU_OST_ZFS)
	return (zfs_get_vfs_flag_unmounted(os));
	#else
	(void) os;
	#endif
	return (B_FALSE);
	}

	static int
	dmu_free_long_range_impl(objset_t os, dnode_t dn, uint64_t offset,
	uint64_t length)
	{
	uint64_t object_size;
	int err;
	uint64_t dirty_frees_threshold;
	dsl_pool_t *dp = dmu_objset_pool(os);

	if (dn == NULL)
	return (SET_ERROR(EINVAL));

	object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
	if (offset >= object_size)
	return (0);

	if (zfs_per_txg_dirty_frees_percent <= 100)
	dirty_frees_threshold =
	zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100;
	else
	dirty_frees_threshold = zfs_dirty_data_max / 20;

	if (length == DMU_OBJECT_END \|\| offset + length > object_size)
	length = object_size - offset;

	while (length != 0) {
	uint64_t chunk_end, chunk_begin, chunk_len;
	uint64_t l1blks;
	dmu_tx_t *tx;

	if (dmu_objset_zfs_unmounting(dn->dn_objset))
	return (SET_ERROR(EINTR));

	chunk_end = chunk_begin = offset + length;

	/* move chunk_begin backwards to the beginning of this chunk */
	err = get_next_chunk(dn, &chunk_begin, offset, &l1blks);
	if (err)
	return (err);
	ASSERT3U(chunk_begin, >=, offset);
	ASSERT3U(chunk_begin, <=, chunk_end);

	chunk_len = chunk_end - chunk_begin;

	tx = dmu_tx_create(os);
	dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len);

	/*
	* Mark this transaction as typically resulting in a net
	* reduction in space used.
	*/
	dmu_tx_mark_netfree(tx);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err) {
	dmu_tx_abort(tx);
	return (err);
	}

	uint64_t txg = dmu_tx_get_txg(tx);

	mutex_enter(&dp->dp_lock);
	uint64_t long_free_dirty =
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK];
	mutex_exit(&dp->dp_lock);

	/*
	* To avoid filling up a TXG with just frees, wait for
	* the next TXG to open before freeing more chunks if
	* we have reached the threshold of frees.
	*/
	if (dirty_frees_threshold != 0 &&
	long_free_dirty >= dirty_frees_threshold) {
	DMU_TX_STAT_BUMP(dmu_tx_dirty_frees_delay);
	dmu_tx_commit(tx);
	txg_wait_open(dp, 0, B_TRUE);
	continue;
	}

	/*
	* In order to prevent unnecessary write throttling, for each
	* TXG, we track the cumulative size of L1 blocks being dirtied
	* in dnode_free_range() below. We compare this number to a
	* tunable threshold, past which we prevent new L1 dirty freeing
	* blocks from being added into the open TXG. See
	* dmu_free_long_range_impl() for details. The threshold
	* prevents write throttle activation due to dirty freeing L1
	* blocks taking up a large percentage of zfs_dirty_data_max.
	*/
	mutex_enter(&dp->dp_lock);
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] +=
	l1blks << dn->dn_indblkshift;
	mutex_exit(&dp->dp_lock);
	DTRACE_PROBE3(free__long__range,
	uint64_t, long_free_dirty, uint64_t, chunk_len,
	uint64_t, txg);
	dnode_free_range(dn, chunk_begin, chunk_len, tx);

	dmu_tx_commit(tx);

	length -= chunk_len;
	}
	return (0);
	}

	int
	dmu_free_long_range(objset_t *os, uint64_t object,
	uint64_t offset, uint64_t length)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err != 0)
	return (err);
	err = dmu_free_long_range_impl(os, dn, offset, length);

	/*
	* It is important to zero out the maxblkid when freeing the entire
	* file, so that (a) subsequent calls to dmu_free_long_range_impl()
	* will take the fast path, and (b) dnode_reallocate() can verify
	* that the entire file has been freed.
	*/
	if (err == 0 && offset == 0 && length == DMU_OBJECT_END)
	dn->dn_maxblkid = 0;

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_free_long_object(objset_t *os, uint64_t object)
	{
	dmu_tx_t *tx;
	int err;

	err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END);
	if (err != 0)
	return (err);

	tx = dmu_tx_create(os);
	dmu_tx_hold_bonus(tx, object);
	dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
	dmu_tx_mark_netfree(tx);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err == 0) {
	err = dmu_object_free(os, object, tx);
	dmu_tx_commit(tx);
	} else {
	dmu_tx_abort(tx);
	}

	return (err);
	}

	int
	dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t size, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	ASSERT(offset < UINT64_MAX);
	ASSERT(size == DMU_OBJECT_END \|\| size <= UINT64_MAX - offset);
	dnode_free_range(dn, offset, size, tx);
	dnode_rele(dn, FTAG);
	return (0);
	}

	static int
	dmu_read_impl(dnode_t *dn, uint64_t offset, uint64_t size,
	void *buf, uint32_t flags)
	{
	dmu_buf_t **dbp;
	int numbufs, err = 0;

	/*
	* Deal with odd block sizes, where there can't be data past the first
	* block. If we ever do the tail block optimization, we will need to
	* handle that here as well.
	*/
	if (dn->dn_maxblkid == 0) {
	uint64_t newsz = offset > dn->dn_datablksz ? 0 :
	MIN(size, dn->dn_datablksz - offset);
	memset((char *)buf + newsz, 0, size - newsz);
	size = newsz;
	}

	while (size > 0) {
	uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2);
	int i;

	/*
	* NB: we could do this block-at-a-time, but it's nice
	* to be reading in parallel.
	*/
	err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
	TRUE, FTAG, &numbufs, &dbp, flags);
	if (err)
	break;

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = offset - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	(void) memcpy(buf, (char *)db->db_data + bufoff, tocpy);

	offset += tocpy;
	size -= tocpy;
	buf = (char *)buf + tocpy;
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}
	return (err);
	}

	int
	dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	void *buf, uint32_t flags)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err != 0)
	return (err);

	err = dmu_read_impl(dn, offset, size, buf, flags);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_read_by_dnode(dnode_t dn, uint64_t offset, uint64_t size, void buf,
	uint32_t flags)
	{
	return (dmu_read_impl(dn, offset, size, buf, flags));
	}

	static void
	dmu_write_impl(dmu_buf_t **dbp, int numbufs, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	int i;

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = offset - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	ASSERT(i == 0 \|\| i == numbufs-1 \|\| tocpy == db->db_size);

	if (tocpy == db->db_size)
	dmu_buf_will_fill(db, tx);
	else
	dmu_buf_will_dirty(db, tx);

	(void) memcpy((char *)db->db_data + bufoff, buf, tocpy);

	if (tocpy == db->db_size)
	dmu_buf_fill_done(db, tx);

	offset += tocpy;
	size -= tocpy;
	buf = (char *)buf + tocpy;
	}
	}

	void
	dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	dmu_buf_t **dbp;
	int numbufs;

	if (size == 0)
	return;

	VERIFY0(dmu_buf_hold_array(os, object, offset, size,
	FALSE, FTAG, &numbufs, &dbp));
	dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	/*
	* Note: Lustre is an external consumer of this interface.
	*/
	void
	dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	dmu_buf_t **dbp;
	int numbufs;

	if (size == 0)
	return;

	VERIFY0(dmu_buf_hold_array_by_dnode(dn, offset, size,
	FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH));
	dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	void
	dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx)
	{
	dmu_buf_t **dbp;
	int numbufs, i;

	if (size == 0)
	return;

	VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
	FALSE, FTAG, &numbufs, &dbp));

	for (i = 0; i < numbufs; i++) {
	dmu_buf_t *db = dbp[i];

	dmu_buf_will_not_fill(db, tx);
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	void
	dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
	void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
	int compressed_size, int byteorder, dmu_tx_t *tx)
	{
	dmu_buf_t *db;

	ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES);
	ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS);
	VERIFY0(dmu_buf_hold_noread(os, object, offset,
	FTAG, &db));

	dmu_buf_write_embedded(db,
	data, (bp_embedded_type_t)etype, (enum zio_compress)comp,
	uncompressed_size, compressed_size, byteorder, tx);

	dmu_buf_rele(db, FTAG);
	}

	void
	dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx)
	{
	int numbufs, i;
	dmu_buf_t **dbp;

	VERIFY0(dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG,
	&numbufs, &dbp));
	for (i = 0; i < numbufs; i++)
	dmu_buf_redact(dbp[i], tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	#ifdef _KERNEL
	int
	dmu_read_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size)
	{
	dmu_buf_t **dbp;
	int numbufs, i, err;

	/*
	* NB: we could do this block-at-a-time, but it's nice
	* to be reading in parallel.
	*/
	err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
	TRUE, FTAG, &numbufs, &dbp, 0);
	if (err)
	return (err);

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = zfs_uio_offset(uio) - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	err = zfs_uio_fault_move((char *)db->db_data + bufoff, tocpy,
	UIO_READ, uio);

	if (err)
	break;

	size -= tocpy;
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);

	return (err);
	}

	/*
	* Read 'size' bytes into the uio buffer.
	* From object zdb->db_object.
	* Starting at zfs_uio_offset(uio).
	*
	* If the caller already has a dbuf in the target object
	* (e.g. its bonus buffer), this routine is faster than dmu_read_uio(),
	* because we don't have to find the dnode_t for the object.
	*/
	int
	dmu_read_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zdb;
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_read_uio_dnode(dn, uio, size);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Read 'size' bytes into the uio buffer.
	* From the specified object
	* Starting at offset zfs_uio_offset(uio).
	*/
	int
	dmu_read_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size)
	{
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_read_uio_dnode(dn, uio, size);

	dnode_rele(dn, FTAG);

	return (err);
	}

	int
	dmu_write_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size, dmu_tx_t *tx)
	{
	dmu_buf_t **dbp;
	int numbufs;
	int err = 0;
	int i;

	err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
	FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH);
	if (err)
	return (err);

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = zfs_uio_offset(uio) - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	ASSERT(i == 0 \|\| i == numbufs-1 \|\| tocpy == db->db_size);

	if (tocpy == db->db_size)
	dmu_buf_will_fill(db, tx);
	else
	dmu_buf_will_dirty(db, tx);

	/*
	* XXX zfs_uiomove could block forever (eg.nfs-backed
	* pages). There needs to be a uiolockdown() function
	* to lock the pages in memory, so that zfs_uiomove won't
	* block.
	*/
	err = zfs_uio_fault_move((char *)db->db_data + bufoff,
	tocpy, UIO_WRITE, uio);

	if (tocpy == db->db_size)
	dmu_buf_fill_done(db, tx);

	if (err)
	break;

	size -= tocpy;
	}

	dmu_buf_rele_array(dbp, numbufs, FTAG);
	return (err);
	}

	/*
	* Write 'size' bytes from the uio buffer.
	* To object zdb->db_object.
	* Starting at offset zfs_uio_offset(uio).
	*
	* If the caller already has a dbuf in the target object
	* (e.g. its bonus buffer), this routine is faster than dmu_write_uio(),
	* because we don't have to find the dnode_t for the object.
	*/
	int
	dmu_write_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zdb;
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_write_uio_dnode(dn, uio, size, tx);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Write 'size' bytes from the uio buffer.
	* To the specified object.
	* Starting at offset zfs_uio_offset(uio).
	*/
	int
	dmu_write_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_write_uio_dnode(dn, uio, size, tx);

	dnode_rele(dn, FTAG);

	return (err);
	}
	#endif /* _KERNEL */

	/*
	* Allocate a loaned anonymous arc buffer.
	*/
	arc_buf_t *
	dmu_request_arcbuf(dmu_buf_t *handle, int size)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )handle;

	return (arc_loan_buf(db->db_objset->os_spa, B_FALSE, size));
	}

	/*
	* Free a loaned arc buffer.
	*/
	void
	dmu_return_arcbuf(arc_buf_t *buf)
	{
	arc_return_buf(buf, FTAG);
	arc_buf_destroy(buf, FTAG);
	}

	/*
	* A "lightweight" write is faster than a regular write (e.g.
	* dmu_write_by_dnode() or dmu_assign_arcbuf_by_dnode()), because it avoids the
	* CPU cost of creating a dmu_buf_impl_t and arc_buf_[hdr_]_t. However, the
	* data can not be read or overwritten until the transaction's txg has been
	* synced. This makes it appropriate for workloads that are known to be
	* (temporarily) write-only, like "zfs receive".
	*
	* A single block is written, starting at the specified offset in bytes. If
	* the call is successful, it returns 0 and the provided abd has been
	* consumed (the caller should not free it).
	*/
	int
	dmu_lightweight_write_by_dnode(dnode_t dn, uint64_t offset, abd_t abd,
	const zio_prop_t zp, enum zio_flag flags, dmu_tx_t tx)
	{
	dbuf_dirty_record_t *dr =
	dbuf_dirty_lightweight(dn, dbuf_whichblock(dn, 0, offset), tx);
	if (dr == NULL)
	return (SET_ERROR(EIO));
	dr->dt.dll.dr_abd = abd;
	dr->dt.dll.dr_props = *zp;
	dr->dt.dll.dr_flags = flags;
	return (0);
	}

	/*
	* When possible directly assign passed loaned arc buffer to a dbuf.
	* If this is not possible copy the contents of passed arc buf via
	* dmu_write().
	*/
	int
	dmu_assign_arcbuf_by_dnode(dnode_t dn, uint64_t offset, arc_buf_t buf,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db;
	objset_t *os = dn->dn_objset;
	uint64_t object = dn->dn_object;
	uint32_t blksz = (uint32_t)arc_buf_lsize(buf);
	uint64_t blkid;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, FTAG);
	if (db == NULL)
	return (SET_ERROR(EIO));
	rw_exit(&dn->dn_struct_rwlock);

	/*
	* We can only assign if the offset is aligned and the arc buf is the
	* same size as the dbuf.
	*/
	if (offset == db->db.db_offset && blksz == db->db.db_size) {
	zfs_racct_write(blksz, 1);
	dbuf_assign_arcbuf(db, buf, tx);
	dbuf_rele(db, FTAG);
	} else {
	/* compressed bufs must always be assignable to their dbuf */
	ASSERT3U(arc_get_compression(buf), ==, ZIO_COMPRESS_OFF);
	ASSERT(!(buf->b_flags & ARC_BUF_FLAG_COMPRESSED));

	dbuf_rele(db, FTAG);
	dmu_write(os, object, offset, blksz, buf->b_data, tx);
	dmu_return_arcbuf(buf);
	}

	return (0);
	}

	int
	dmu_assign_arcbuf_by_dbuf(dmu_buf_t handle, uint64_t offset, arc_buf_t buf,
	dmu_tx_t *tx)
	{
	int err;
	dmu_buf_impl_t dbuf = (dmu_buf_impl_t )handle;

	DB_DNODE_ENTER(dbuf);
	err = dmu_assign_arcbuf_by_dnode(DB_DNODE(dbuf), offset, buf, tx);
	DB_DNODE_EXIT(dbuf);

	return (err);
	}

	typedef struct {
	dbuf_dirty_record_t *dsa_dr;
	dmu_sync_cb_t *dsa_done;
	zgd_t *dsa_zgd;
	dmu_tx_t *dsa_tx;
	} dmu_sync_arg_t;

	static void
	dmu_sync_ready(zio_t zio, arc_buf_t buf, void *varg)
	{
	(void) buf;
	dmu_sync_arg_t *dsa = varg;
	dmu_buf_t *db = dsa->dsa_zgd->zgd_db;
	blkptr_t *bp = zio->io_bp;

	if (zio->io_error == 0) {
	if (BP_IS_HOLE(bp)) {
	/*
	* A block of zeros may compress to a hole, but the
	* block size still needs to be known for replay.
	*/
	BP_SET_LSIZE(bp, db->db_size);
	} else if (!BP_IS_EMBEDDED(bp)) {
	ASSERT(BP_GET_LEVEL(bp) == 0);
	BP_SET_FILL(bp, 1);
	}
	}
	}

	static void
	dmu_sync_late_arrival_ready(zio_t *zio)
	{
	dmu_sync_ready(zio, NULL, zio->io_private);
	}

	static void
	dmu_sync_done(zio_t zio, arc_buf_t buf, void *varg)
	{
	(void) buf;
	dmu_sync_arg_t *dsa = varg;
	dbuf_dirty_record_t *dr = dsa->dsa_dr;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	zgd_t *zgd = dsa->dsa_zgd;

	/*
	* Record the vdev(s) backing this blkptr so they can be flushed after
	* the writes for the lwb have completed.
	*/
	if (zio->io_error == 0) {
	zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);
	}

	mutex_enter(&db->db_mtx);
	ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
	if (zio->io_error == 0) {
	dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE);
	if (dr->dt.dl.dr_nopwrite) {
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	uint8_t chksum = BP_GET_CHECKSUM(bp_orig);

	ASSERT(BP_EQUAL(bp, bp_orig));
	VERIFY(BP_EQUAL(bp, db->db_blkptr));
	ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF);
	VERIFY(zio_checksum_table[chksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE);
	}
	dr->dt.dl.dr_overridden_by = *zio->io_bp;
	dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
	dr->dt.dl.dr_copies = zio->io_prop.zp_copies;

	/*
	* Old style holes are filled with all zeros, whereas
	* new-style holes maintain their lsize, type, level,
	* and birth time (see zio_write_compress). While we
	* need to reset the BP_SET_LSIZE() call that happened
	* in dmu_sync_ready for old style holes, we do not
	* want to wipe out the information contained in new
	* style holes. Thus, only zero out the block pointer if
	* it's an old style hole.
	*/
	if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) &&
	dr->dt.dl.dr_overridden_by.blk_birth == 0)
	BP_ZERO(&dr->dt.dl.dr_overridden_by);
	} else {
	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	}
	cv_broadcast(&db->db_changed);
	mutex_exit(&db->db_mtx);

	dsa->dsa_done(dsa->dsa_zgd, zio->io_error);

	kmem_free(dsa, sizeof (*dsa));
	}

	static void
	dmu_sync_late_arrival_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	dmu_sync_arg_t *dsa = zio->io_private;
	zgd_t *zgd = dsa->dsa_zgd;

	if (zio->io_error == 0) {
	/*
	* Record the vdev(s) backing this blkptr so they can be
	* flushed after the writes for the lwb have completed.
	*/
	zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);

	if (!BP_IS_HOLE(bp)) {
	blkptr_t *bp_orig __maybe_unused = &zio->io_bp_orig;
	ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE));
	ASSERT(BP_IS_HOLE(bp_orig) \|\| !BP_EQUAL(bp, bp_orig));
	ASSERT(zio->io_bp->blk_birth == zio->io_txg);
	ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa));
	zio_free(zio->io_spa, zio->io_txg, zio->io_bp);
	}
	}

	dmu_tx_commit(dsa->dsa_tx);

	dsa->dsa_done(dsa->dsa_zgd, zio->io_error);

	abd_free(zio->io_abd);
	kmem_free(dsa, sizeof (*dsa));
	}

	static int
	dmu_sync_late_arrival(zio_t pio, objset_t os, dmu_sync_cb_t done, zgd_t zgd,
	zio_prop_t zp, zbookmark_phys_t zb)
	{
	dmu_sync_arg_t *dsa;
	dmu_tx_t *tx;

	tx = dmu_tx_create(os);
	dmu_tx_hold_space(tx, zgd->zgd_db->db_size);
	if (dmu_tx_assign(tx, TXG_WAIT) != 0) {
	dmu_tx_abort(tx);
	/* Make zl_get_data do txg_waited_synced() */
	return (SET_ERROR(EIO));
	}

	/*
	* In order to prevent the zgd's lwb from being free'd prior to
	* dmu_sync_late_arrival_done() being called, we have to ensure
	* the lwb's "max txg" takes this tx's txg into account.
	*/
	zil_lwb_add_txg(zgd->zgd_lwb, dmu_tx_get_txg(tx));

	dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
	dsa->dsa_dr = NULL;
	dsa->dsa_done = done;
	dsa->dsa_zgd = zgd;
	dsa->dsa_tx = tx;

	/*
	* Since we are currently syncing this txg, it's nontrivial to
	* determine what BP to nopwrite against, so we disable nopwrite.
	*
	* When syncing, the db_blkptr is initially the BP of the previous
	* txg. We can not nopwrite against it because it will be changed
	* (this is similar to the non-late-arrival case where the dbuf is
	* dirty in a future txg).
	*
	* Then dbuf_write_ready() sets bp_blkptr to the location we will write.
	* We can not nopwrite against it because although the BP will not
	* (typically) be changed, the data has not yet been persisted to this
	* location.
	*
	* Finally, when dbuf_write_done() is called, it is theoretically
	* possible to always nopwrite, because the data that was written in
	* this txg is the same data that we are trying to write. However we
	* would need to check that this dbuf is not dirty in any future
	* txg's (as we do in the normal dmu_sync() path). For simplicity, we
	* don't nopwrite in this case.
	*/
	zp->zp_nopwrite = B_FALSE;

	zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp,
	abd_get_from_buf(zgd->zgd_db->db_data, zgd->zgd_db->db_size),
	zgd->zgd_db->db_size, zgd->zgd_db->db_size, zp,
	dmu_sync_late_arrival_ready, NULL, NULL, dmu_sync_late_arrival_done,
	dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb));

	return (0);
	}

	/*
	* Intent log support: sync the block associated with db to disk.
	* N.B. and XXX: the caller is responsible for making sure that the
	* data isn't changing while dmu_sync() is writing it.
	*
	* Return values:
	*
	* EEXIST: this txg has already been synced, so there's nothing to do.
	* The caller should not log the write.
	*
	* ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
	* The caller should not log the write.
	*
	* EALREADY: this block is already in the process of being synced.
	* The caller should track its progress (somehow).
	*
	* EIO: could not do the I/O.
	* The caller should do a txg_wait_synced().
	*
	* 0: the I/O has been initiated.
	* The caller should log this blkptr in the done callback.
	* It is possible that the I/O will fail, in which case
	* the error will be reported to the done callback and
	* propagated to pio from zio_done().
	*/
	int
	dmu_sync(zio_t pio, uint64_t txg, dmu_sync_cb_t done, zgd_t *zgd)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zgd->zgd_db;
	objset_t *os = db->db_objset;
	dsl_dataset_t *ds = os->os_dsl_dataset;
	dbuf_dirty_record_t dr, dr_next;
	dmu_sync_arg_t *dsa;
	zbookmark_phys_t zb;
	zio_prop_t zp;
	dnode_t *dn;

	ASSERT(pio != NULL);
	ASSERT(txg != 0);

	SET_BOOKMARK(&zb, ds->ds_object,
	db->db.db_object, db->db_level, db->db_blkid);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	dmu_write_policy(os, dn, db->db_level, WP_DMU_SYNC, &zp);
	DB_DNODE_EXIT(db);

	/*
	* If we're frozen (running ziltest), we always need to generate a bp.
	*/
	if (txg > spa_freeze_txg(os->os_spa))
	return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));

	/*
	* Grabbing db_mtx now provides a barrier between dbuf_sync_leaf()
	* and us. If we determine that this txg is not yet syncing,
	* but it begins to sync a moment later, that's OK because the
	* sync thread will block in dbuf_sync_leaf() until we drop db_mtx.
	*/
	mutex_enter(&db->db_mtx);

	if (txg <= spa_last_synced_txg(os->os_spa)) {
	/*
	* This txg has already synced. There's nothing to do.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(EEXIST));
	}

	if (txg <= spa_syncing_txg(os->os_spa)) {
	/*
	* This txg is currently syncing, so we can't mess with
	* the dirty record anymore; just write a new log block.
	*/
	mutex_exit(&db->db_mtx);
	return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));
	}

	dr = dbuf_find_dirty_eq(db, txg);

	if (dr == NULL) {
	/*
	* There's no dr for this dbuf, so it must have been freed.
	* There's no need to log writes to freed blocks, so we're done.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(ENOENT));
	}

	dr_next = list_next(&db->db_dirty_records, dr);
	ASSERT(dr_next == NULL \|\| dr_next->dr_txg < txg);

	if (db->db_blkptr != NULL) {
	/*
	* We need to fill in zgd_bp with the current blkptr so that
	* the nopwrite code can check if we're writing the same
	* data that's already on disk. We can only nopwrite if we
	* are sure that after making the copy, db_blkptr will not
	* change until our i/o completes. We ensure this by
	* holding the db_mtx, and only allowing nopwrite if the
	* block is not already dirty (see below). This is verified
	* by dmu_sync_done(), which VERIFYs that the db_blkptr has
	* not changed.
	*/
	zgd->zgd_bp = db->db_blkptr;
	}

	/*
	* Assume the on-disk data is X, the current syncing data (in
	* txg - 1) is Y, and the current in-memory data is Z (currently
	* in dmu_sync).
	*
	* We usually want to perform a nopwrite if X and Z are the
	* same. However, if Y is different (i.e. the BP is going to
	* change before this write takes effect), then a nopwrite will
	* be incorrect - we would override with X, which could have
	* been freed when Y was written.
	*
	* (Note that this is not a concern when we are nop-writing from
	* syncing context, because X and Y must be identical, because
	* all previous txgs have been synced.)
	*
	* Therefore, we disable nopwrite if the current BP could change
	* before this TXG. There are two ways it could change: by
	* being dirty (dr_next is non-NULL), or by being freed
	* (dnode_block_freed()). This behavior is verified by
	* zio_done(), which VERIFYs that the override BP is identical
	* to the on-disk BP.
	*/
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	if (dr_next != NULL \|\| dnode_block_freed(dn, db->db_blkid))
	zp.zp_nopwrite = B_FALSE;
	DB_DNODE_EXIT(db);

	ASSERT(dr->dr_txg == txg);
	if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC \|\|
	dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
	/*
	* We have already issued a sync write for this buffer,
	* or this buffer has already been synced. It could not
	* have been dirtied since, or we would have cleared the state.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(EALREADY));
	}

	ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
	dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
	mutex_exit(&db->db_mtx);

	dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
	dsa->dsa_dr = dr;
	dsa->dsa_done = done;
	dsa->dsa_zgd = zgd;
	dsa->dsa_tx = NULL;

	zio_nowait(arc_write(pio, os->os_spa, txg,
	zgd->zgd_bp, dr->dt.dl.dr_data, dbuf_is_l2cacheable(db),
	&zp, dmu_sync_ready, NULL, NULL, dmu_sync_done, dsa,
	ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));

	return (0);
	}

	int
	dmu_object_set_nlevels(objset_t os, uint64_t object, int nlevels, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	err = dnode_set_nlevels(dn, nlevels, tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	err = dnode_set_blksz(dn, size, ibs, tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	dnode_new_blkid(dn, maxblkid, tx, B_FALSE, B_TRUE);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	return (0);
	}

	void
	dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	/*
	* Send streams include each object's checksum function. This
	* check ensures that the receiving system can understand the
	* checksum function transmitted.
	*/
	ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS);

	VERIFY0(dnode_hold(os, object, FTAG, &dn));
	ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS);
	dn->dn_checksum = checksum;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);
	}

	void
	dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	/*
	* Send streams include each object's compression function. This
	* check ensures that the receiving system can understand the
	* compression function transmitted.
	*/
	ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS);

	VERIFY0(dnode_hold(os, object, FTAG, &dn));
	dn->dn_compress = compress;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);
	}

	/*
	* When the "redundant_metadata" property is set to "most", only indirect
	* blocks of this level and higher will have an additional ditto block.
	*/
	static const int zfs_redundant_metadata_most_ditto_level = 2;

	void
	dmu_write_policy(objset_t os, dnode_t dn, int level, int wp, zio_prop_t *zp)
	{
	dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET;
	boolean_t ismd = (level > 0 \|\| DMU_OT_IS_METADATA(type) \|\|
	(wp & WP_SPILL));
	enum zio_checksum checksum = os->os_checksum;
	enum zio_compress compress = os->os_compress;
	uint8_t complevel = os->os_complevel;
	enum zio_checksum dedup_checksum = os->os_dedup_checksum;
	boolean_t dedup = B_FALSE;
	boolean_t nopwrite = B_FALSE;
	boolean_t dedup_verify = os->os_dedup_verify;
	boolean_t encrypt = B_FALSE;
	int copies = os->os_copies;

	/*
	* We maintain different write policies for each of the following
	* types of data:
	* 1. metadata
	* 2. preallocated blocks (i.e. level-0 blocks of a dump device)
	* 3. all other level 0 blocks
	*/
	if (ismd) {
	/*
	* XXX -- we should design a compression algorithm
	* that specializes in arrays of bps.
	*/
	compress = zio_compress_select(os->os_spa,
	ZIO_COMPRESS_ON, ZIO_COMPRESS_ON);

	/*
	* Metadata always gets checksummed. If the data
	* checksum is multi-bit correctable, and it's not a
	* ZBT-style checksum, then it's suitable for metadata
	* as well. Otherwise, the metadata checksum defaults
	* to fletcher4.
	*/
	if (!(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_METADATA) \|\|
	(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_EMBEDDED))
	checksum = ZIO_CHECKSUM_FLETCHER_4;

	if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL \|\|
	(os->os_redundant_metadata ==
	ZFS_REDUNDANT_METADATA_MOST &&
	(level >= zfs_redundant_metadata_most_ditto_level \|\|
	DMU_OT_IS_METADATA(type) \|\| (wp & WP_SPILL))))
	copies++;
	} else if (wp & WP_NOFILL) {
	ASSERT(level == 0);

	/*
	* If we're writing preallocated blocks, we aren't actually
	* writing them so don't set any policy properties. These
	* blocks are currently only used by an external subsystem
	* outside of zfs (i.e. dump) and not written by the zio
	* pipeline.
	*/
	compress = ZIO_COMPRESS_OFF;
	checksum = ZIO_CHECKSUM_OFF;
	} else {
	compress = zio_compress_select(os->os_spa, dn->dn_compress,
	compress);
	complevel = zio_complevel_select(os->os_spa, compress,
	complevel, complevel);

	checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ?
	zio_checksum_select(dn->dn_checksum, checksum) :
	dedup_checksum;

	/*
	* Determine dedup setting. If we are in dmu_sync(),
	* we won't actually dedup now because that's all
	* done in syncing context; but we do want to use the
	* dedup checksum. If the checksum is not strong
	* enough to ensure unique signatures, force
	* dedup_verify.
	*/
	if (dedup_checksum != ZIO_CHECKSUM_OFF) {
	dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE;
	if (!(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP))
	dedup_verify = B_TRUE;
	}

	/*
	* Enable nopwrite if we have secure enough checksum
	* algorithm (see comment in zio_nop_write) and
	* compression is enabled. We don't enable nopwrite if
	* dedup is enabled as the two features are mutually
	* exclusive.
	*/
	nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE) &&
	compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled);
	}

	/*
	* All objects in an encrypted objset are protected from modification
	* via a MAC. Encrypted objects store their IV and salt in the last DVA
	* in the bp, so we cannot use all copies. Encrypted objects are also
	* not subject to nopwrite since writing the same data will still
	* result in a new ciphertext. Only encrypted blocks can be dedup'd
	* to avoid ambiguity in the dedup code since the DDT does not store
	* object types.
	*/
	if (os->os_encrypted && (wp & WP_NOFILL) == 0) {
	encrypt = B_TRUE;

	if (DMU_OT_IS_ENCRYPTED(type)) {
	copies = MIN(copies, SPA_DVAS_PER_BP - 1);
	nopwrite = B_FALSE;
	} else {
	dedup = B_FALSE;
	}

	if (level <= 0 &&
	(type == DMU_OT_DNODE \|\| type == DMU_OT_OBJSET)) {
	compress = ZIO_COMPRESS_EMPTY;
	}
	}

	zp->zp_compress = compress;
	zp->zp_complevel = complevel;
	zp->zp_checksum = checksum;
	zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type;
	zp->zp_level = level;
	zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa));
	zp->zp_dedup = dedup;
	zp->zp_dedup_verify = dedup && dedup_verify;
	zp->zp_nopwrite = nopwrite;
	zp->zp_encrypt = encrypt;
	zp->zp_byteorder = ZFS_HOST_BYTEORDER;
	memset(zp->zp_salt, 0, ZIO_DATA_SALT_LEN);
	memset(zp->zp_iv, 0, ZIO_DATA_IV_LEN);
	memset(zp->zp_mac, 0, ZIO_DATA_MAC_LEN);
	zp->zp_zpl_smallblk = DMU_OT_IS_FILE(zp->zp_type) ?
	os->os_zpl_special_smallblock : 0;

	ASSERT3U(zp->zp_compress, !=, ZIO_COMPRESS_INHERIT);
	}

	/*
	* This function is only called from zfs_holey_common() for zpl_llseek()
	* in order to determine the location of holes. In order to accurately
	* report holes all dirty data must be synced to disk. This causes extremely
	* poor performance when seeking for holes in a dirty file. As a compromise,
	* only provide hole data when the dnode is clean. When a dnode is dirty
	* report the dnode as having no holes which is always a safe thing to do.
	*/
	int
	dmu_offset_next(objset_t os, uint64_t object, boolean_t hole, uint64_t off)
	{
	dnode_t *dn;
	int err;

	restart:
	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	if (dnode_is_dirty(dn)) {
	/*
	* If the zfs_dmu_offset_next_sync module option is enabled
	* then strict hole reporting has been requested. Dirty
	* dnodes must be synced to disk to accurately report all
	* holes. When disabled dirty dnodes are reported to not
	* have any holes which is always safe.
	*
	* When called by zfs_holey_common() the zp->z_rangelock
	* is held to prevent zfs_write() and mmap writeback from
	* re-dirtying the dnode after txg_wait_synced().
	*/
	if (zfs_dmu_offset_next_sync) {
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	txg_wait_synced(dmu_objset_pool(os), 0);
	goto restart;
	}

	err = SET_ERROR(EBUSY);
	} else {
	err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK \|
	(hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
	}

	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);

	return (err);
	}

	void
	__dmu_object_info_from_dnode(dnode_t dn, dmu_object_info_t doi)
	{
	dnode_phys_t *dnp = dn->dn_phys;

	doi->doi_data_block_size = dn->dn_datablksz;
	doi->doi_metadata_block_size = dn->dn_indblkshift ?
	1ULL << dn->dn_indblkshift : 0;
	doi->doi_type = dn->dn_type;
	doi->doi_bonus_type = dn->dn_bonustype;
	doi->doi_bonus_size = dn->dn_bonuslen;
	doi->doi_dnodesize = dn->dn_num_slots << DNODE_SHIFT;
	doi->doi_indirection = dn->dn_nlevels;
	doi->doi_checksum = dn->dn_checksum;
	doi->doi_compress = dn->dn_compress;
	doi->doi_nblkptr = dn->dn_nblkptr;
	doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9;
	doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
	doi->doi_fill_count = 0;
	for (int i = 0; i < dnp->dn_nblkptr; i++)
	doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]);
	}

	void
	dmu_object_info_from_dnode(dnode_t dn, dmu_object_info_t doi)
	{
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	mutex_enter(&dn->dn_mtx);

	__dmu_object_info_from_dnode(dn, doi);

	mutex_exit(&dn->dn_mtx);
	rw_exit(&dn->dn_struct_rwlock);
	}

	/*
	* Get information on a DMU object.
	* If doi is NULL, just indicates whether the object exists.
	*/
	int
	dmu_object_info(objset_t os, uint64_t object, dmu_object_info_t doi)
	{
	dnode_t *dn;
	int err = dnode_hold(os, object, FTAG, &dn);

	if (err)
	return (err);

	if (doi != NULL)
	dmu_object_info_from_dnode(dn, doi);

	dnode_rele(dn, FTAG);
	return (0);
	}

	/*
	* As above, but faster; can be used when you have a held dbuf in hand.
	*/
	void
	dmu_object_info_from_db(dmu_buf_t db_fake, dmu_object_info_t doi)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	DB_DNODE_ENTER(db);
	dmu_object_info_from_dnode(DB_DNODE(db), doi);
	DB_DNODE_EXIT(db);
	}

	/*
	* Faster still when you only care about the size.
	*/
	void
	dmu_object_size_from_db(dmu_buf_t db_fake, uint32_t blksize,
	u_longlong_t *nblk512)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	*blksize = dn->dn_datablksz;
	/* add in number of slots used for the dnode itself */
	*nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
	SPA_MINBLOCKSHIFT) + dn->dn_num_slots;
	DB_DNODE_EXIT(db);
	}

	void
	dmu_object_dnsize_from_db(dmu_buf_t db_fake, int dnsize)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	*dnsize = dn->dn_num_slots << DNODE_SHIFT;
	DB_DNODE_EXIT(db);
	}

	void
	byteswap_uint64_array(void *vbuf, size_t size)
	{
	uint64_t *buf = vbuf;
	size_t count = size >> 3;
	int i;

	ASSERT((size & 7) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_64(buf[i]);
	}

	void
	byteswap_uint32_array(void *vbuf, size_t size)
	{
	uint32_t *buf = vbuf;
	size_t count = size >> 2;
	int i;

	ASSERT((size & 3) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_32(buf[i]);
	}

	void
	byteswap_uint16_array(void *vbuf, size_t size)
	{
	uint16_t *buf = vbuf;
	size_t count = size >> 1;
	int i;

	ASSERT((size & 1) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_16(buf[i]);
	}

	void
	byteswap_uint8_array(void *vbuf, size_t size)
	{
	(void) vbuf, (void) size;
	}

	void
	dmu_init(void)
	{
	abd_init();
	zfs_dbgmsg_init();
	sa_cache_init();
	dmu_objset_init();
	dnode_init();
	zfetch_init();
	dmu_tx_init();
	l2arc_init();
	arc_init();
	dbuf_init();
	}

	void
	dmu_fini(void)
	{
	arc_fini(); /* arc depends on l2arc, so arc must go first */
	l2arc_fini();
	dmu_tx_fini();
	zfetch_fini();
	dbuf_fini();
	dnode_fini();
	dmu_objset_fini();
	sa_cache_fini();
	zfs_dbgmsg_fini();
	abd_fini();
	}

	EXPORT_SYMBOL(dmu_bonus_hold);
	EXPORT_SYMBOL(dmu_bonus_hold_by_dnode);
	EXPORT_SYMBOL(dmu_buf_hold_array_by_bonus);
	EXPORT_SYMBOL(dmu_buf_rele_array);
	EXPORT_SYMBOL(dmu_prefetch);
	EXPORT_SYMBOL(dmu_free_range);
	EXPORT_SYMBOL(dmu_free_long_range);
	EXPORT_SYMBOL(dmu_free_long_object);
	EXPORT_SYMBOL(dmu_read);
	EXPORT_SYMBOL(dmu_read_by_dnode);
	EXPORT_SYMBOL(dmu_write);
	EXPORT_SYMBOL(dmu_write_by_dnode);
	EXPORT_SYMBOL(dmu_prealloc);
	EXPORT_SYMBOL(dmu_object_info);
	EXPORT_SYMBOL(dmu_object_info_from_dnode);
	EXPORT_SYMBOL(dmu_object_info_from_db);
	EXPORT_SYMBOL(dmu_object_size_from_db);
	EXPORT_SYMBOL(dmu_object_dnsize_from_db);
	EXPORT_SYMBOL(dmu_object_set_nlevels);
	EXPORT_SYMBOL(dmu_object_set_blocksize);
	EXPORT_SYMBOL(dmu_object_set_maxblkid);
	EXPORT_SYMBOL(dmu_object_set_checksum);
	EXPORT_SYMBOL(dmu_object_set_compress);
	EXPORT_SYMBOL(dmu_offset_next);
	EXPORT_SYMBOL(dmu_write_policy);
	EXPORT_SYMBOL(dmu_sync);
	EXPORT_SYMBOL(dmu_request_arcbuf);
	EXPORT_SYMBOL(dmu_return_arcbuf);
	EXPORT_SYMBOL(dmu_assign_arcbuf_by_dnode);
	EXPORT_SYMBOL(dmu_assign_arcbuf_by_dbuf);
	EXPORT_SYMBOL(dmu_buf_hold);
	EXPORT_SYMBOL(dmu_ot);

	ZFS_MODULE_PARAM(zfs, zfs_, nopwrite_enabled, INT, ZMOD_RW,
	"Enable NOP writes");

	ZFS_MODULE_PARAM(zfs, zfs_, per_txg_dirty_frees_percent, ULONG, ZMOD_RW,
	"Percentage of dirtied blocks from frees in one TXG");

	ZFS_MODULE_PARAM(zfs, zfs_, dmu_offset_next_sync, INT, ZMOD_RW,
	"Enable forcing txg sync to find holes");

	/* CSTYLED */
	-ZFS_MODULE_PARAM(zfs, , dmu_prefetch_max, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, , dmu_prefetch_max, UINT, ZMOD_RW,
	"Limit one prefetch call to this size");
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_object.c b/sys/contrib/openzfs/module/zfs/dmu_object.c
	index 640af59be550..6fe4480d6d9a 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_object.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_object.c
	@@ -1,523 +1,523 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	*/

	#include <sys/dbuf.h>
	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/dnode.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_dataset.h>

	/*
	* Each of the concurrent object allocators will grab
	* 2^dmu_object_alloc_chunk_shift dnode slots at a time. The default is to
	* grab 128 slots, which is 4 blocks worth. This was experimentally
	* determined to be the lowest value that eliminates the measurable effect
	* of lock contention from this code path.
	*/
	-int dmu_object_alloc_chunk_shift = 7;
	+uint_t dmu_object_alloc_chunk_shift = 7;

	static uint64_t
	dmu_object_alloc_impl(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dnode_t *allocated_dnode, const void tag, dmu_tx_t *tx)
	{
	uint64_t object;
	uint64_t L1_dnode_count = DNODES_PER_BLOCK <<
	(DMU_META_DNODE(os)->dn_indblkshift - SPA_BLKPTRSHIFT);
	dnode_t *dn = NULL;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	boolean_t restarted = B_FALSE;
	uint64_t *cpuobj = NULL;
	- int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	+ uint_t dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	int error;

	cpuobj = &os->os_obj_next_percpu[CPU_SEQID_UNSTABLE %
	os->os_obj_next_percpu_len];

	if (dn_slots == 0) {
	dn_slots = DNODE_MIN_SLOTS;
	} else {
	ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
	ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);
	}

	/*
	* The "chunk" of dnodes that is assigned to a CPU-specific
	* allocator needs to be at least one block's worth, to avoid
	* lock contention on the dbuf. It can be at most one L1 block's
	* worth, so that the "rescan after polishing off a L1's worth"
	* logic below will be sure to kick in.
	*/
	if (dnodes_per_chunk < DNODES_PER_BLOCK)
	dnodes_per_chunk = DNODES_PER_BLOCK;
	if (dnodes_per_chunk > L1_dnode_count)
	dnodes_per_chunk = L1_dnode_count;

	/*
	* The caller requested the dnode be returned as a performance
	* optimization in order to avoid releasing the hold only to
	* immediately reacquire it. Since they caller is responsible
	* for releasing the hold they must provide the tag.
	*/
	if (allocated_dnode != NULL) {
	ASSERT3P(tag, !=, NULL);
	} else {
	ASSERT3P(tag, ==, NULL);
	tag = FTAG;
	}

	object = *cpuobj;
	for (;;) {
	/*
	* If we finished a chunk of dnodes, get a new one from
	* the global allocator.
	*/
	if ((P2PHASE(object, dnodes_per_chunk) == 0) \|\|
	(P2PHASE(object + dn_slots - 1, dnodes_per_chunk) <
	dn_slots)) {
	DNODE_STAT_BUMP(dnode_alloc_next_chunk);
	mutex_enter(&os->os_obj_lock);
	ASSERT0(P2PHASE(os->os_obj_next_chunk,
	dnodes_per_chunk));
	object = os->os_obj_next_chunk;

	/*
	* Each time we polish off a L1 bp worth of dnodes
	* (2^12 objects), move to another L1 bp that's
	* still reasonably sparse (at most 1/4 full). Look
	* from the beginning at most once per txg. If we
	* still can't allocate from that L1 block, search
	* for an empty L0 block, which will quickly skip
	* to the end of the metadnode if no nearby L0
	* blocks are empty. This fallback avoids a
	* pathology where full dnode blocks containing
	* large dnodes appear sparse because they have a
	* low blk_fill, leading to many failed allocation
	* attempts. In the long term a better mechanism to
	* search for sparse metadnode regions, such as
	* spacemaps, could be implemented.
	*
	* os_scan_dnodes is set during txg sync if enough
	* objects have been freed since the previous
	* rescan to justify backfilling again.
	*
	* Note that dmu_traverse depends on the behavior
	* that we use multiple blocks of the dnode object
	* before going back to reuse objects. Any change
	* to this algorithm should preserve that property
	* or find another solution to the issues described
	* in traverse_visitbp.
	*/
	if (P2PHASE(object, L1_dnode_count) == 0) {
	uint64_t offset;
	uint64_t blkfill;
	int minlvl;
	if (os->os_rescan_dnodes) {
	offset = 0;
	os->os_rescan_dnodes = B_FALSE;
	} else {
	offset = object << DNODE_SHIFT;
	}
	blkfill = restarted ? 1 : DNODES_PER_BLOCK >> 2;
	minlvl = restarted ? 1 : 2;
	restarted = B_TRUE;
	error = dnode_next_offset(DMU_META_DNODE(os),
	DNODE_FIND_HOLE, &offset, minlvl,
	blkfill, 0);
	if (error == 0) {
	object = offset >> DNODE_SHIFT;
	}
	}
	/*
	* Note: if "restarted", we may find a L0 that
	* is not suitably aligned.
	*/
	os->os_obj_next_chunk =
	P2ALIGN(object, dnodes_per_chunk) +
	dnodes_per_chunk;
	(void) atomic_swap_64(cpuobj, object);
	mutex_exit(&os->os_obj_lock);
	}

	/*
	* The value of (*cpuobj) before adding dn_slots is the object
	* ID assigned to us. The value afterwards is the object ID
	* assigned to whoever wants to do an allocation next.
	*/
	object = atomic_add_64_nv(cpuobj, dn_slots) - dn_slots;

	/*
	* XXX We should check for an i/o error here and return
	* up to our caller. Actually we should pre-read it in
	* dmu_tx_assign(), but there is currently no mechanism
	* to do so.
	*/
	error = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE,
	dn_slots, tag, &dn);
	if (error == 0) {
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	/*
	* Another thread could have allocated it; check
	* again now that we have the struct lock.
	*/
	if (dn->dn_type == DMU_OT_NONE) {
	dnode_allocate(dn, ot, blocksize,
	indirect_blockshift, bonustype,
	bonuslen, dn_slots, tx);
	rw_exit(&dn->dn_struct_rwlock);
	dmu_tx_add_new_object(tx, dn);

	/*
	* Caller requested the allocated dnode be
	* returned and is responsible for the hold.
	*/
	if (allocated_dnode != NULL)
	*allocated_dnode = dn;
	else
	dnode_rele(dn, tag);

	return (object);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, tag);
	DNODE_STAT_BUMP(dnode_alloc_race);
	}

	/*
	* Skip to next known valid starting point on error. This
	* is the start of the next block of dnodes.
	*/
	if (dmu_object_next(os, &object, B_TRUE, 0) != 0) {
	object = P2ROUNDUP(object + 1, DNODES_PER_BLOCK);
	DNODE_STAT_BUMP(dnode_alloc_next_block);
	}
	(void) atomic_swap_64(cpuobj, object);
	}
	}

	uint64_t
	dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize,
	dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	bonuslen, 0, NULL, NULL, tx);
	}

	uint64_t
	dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	dmu_tx_t *tx)
	{
	return dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	bonustype, bonuslen, 0, NULL, NULL, tx);
	}

	uint64_t
	dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, int blocksize,
	dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx)
	{
	return (dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	bonuslen, dnodesize, NULL, NULL, tx));
	}

	/*
	* Allocate a new object and return a pointer to the newly allocated dnode
	* via the allocated_dnode argument. The returned dnode will be held and
	* the caller is responsible for releasing the hold by calling dnode_rele().
	*/
	uint64_t
	dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dnode_t *allocated_dnode, const void tag, dmu_tx_t *tx)
	{
	return (dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	bonustype, bonuslen, dnodesize, allocated_dnode, tag, tx));
	}

	int
	dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return (dmu_object_claim_dnsize(os, object, ot, blocksize, bonustype,
	bonuslen, 0, tx));
	}

	int
	dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
	dn_slots = DNODE_MIN_SLOTS;
	ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
	ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);

	if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
	return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, dn_slots,
	FTAG, &dn);
	if (err)
	return (err);

	dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, dn_slots, tx);
	dmu_tx_add_new_object(tx, dn);

	dnode_rele(dn, FTAG);

	return (0);
	}

	int
	dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return (dmu_object_reclaim_dnsize(os, object, ot, blocksize, bonustype,
	bonuslen, DNODE_MIN_SIZE, B_FALSE, tx));
	}

	int
	dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize,
	boolean_t keep_spill, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
	dn_slots = DNODE_MIN_SLOTS;

	if (object == DMU_META_DNODE_OBJECT)
	return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, dn_slots,
	keep_spill, tx);

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_rm_spill(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	if (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	dbuf_rm_spill(dn, tx);
	dnode_rm_spill(dn, tx);
	}
	rw_exit(&dn->dn_struct_rwlock);

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_free(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	ASSERT(object != DMU_META_DNODE_OBJECT \|\| dmu_tx_private_ok(tx));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	ASSERT(dn->dn_type != DMU_OT_NONE);
	/*
	* If we don't create this free range, we'll leak indirect blocks when
	* we get to freeing the dnode in syncing context.
	*/
	dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
	dnode_free(dn, tx);
	dnode_rele(dn, FTAG);

	return (0);
	}

	/*
	* Return (in *objectp) the next object which is allocated (or a hole)
	* after *object, taking into account only objects that may have been modified
	* after the specified txg.
	*/
	int
	dmu_object_next(objset_t os, uint64_t objectp, boolean_t hole, uint64_t txg)
	{
	uint64_t offset;
	uint64_t start_obj;
	struct dsl_dataset *ds = os->os_dsl_dataset;
	int error;

	if (*objectp == 0) {
	start_obj = 1;
	} else if (ds && dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_LARGE_DNODE)) {
	uint64_t i = *objectp + 1;
	uint64_t last_obj = *objectp \| (DNODES_PER_BLOCK - 1);
	dmu_object_info_t doi;

	/*
	* Scan through the remaining meta dnode block. The contents
	* of each slot in the block are known so it can be quickly
	* checked. If the block is exhausted without a match then
	* hand off to dnode_next_offset() for further scanning.
	*/
	while (i <= last_obj) {
	error = dmu_object_info(os, i, &doi);
	if (error == ENOENT) {
	if (hole) {
	*objectp = i;
	return (0);
	} else {
	i++;
	}
	} else if (error == EEXIST) {
	i++;
	} else if (error == 0) {
	if (hole) {
	i += doi.doi_dnodesize >> DNODE_SHIFT;
	} else {
	*objectp = i;
	return (0);
	}
	} else {
	return (error);
	}
	}

	start_obj = i;
	} else {
	start_obj = *objectp + 1;
	}

	offset = start_obj << DNODE_SHIFT;

	error = dnode_next_offset(DMU_META_DNODE(os),
	(hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg);

	*objectp = offset >> DNODE_SHIFT;

	return (error);
	}

	/*
	* Turn this object from old_type into DMU_OTN_ZAP_METADATA, and bump the
	* refcount on SPA_FEATURE_EXTENSIBLE_DATASET.
	*
	* Only for use from syncing context, on MOS objects.
	*/
	void
	dmu_object_zapify(objset_t *mos, uint64_t object, dmu_object_type_t old_type,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	if (dn->dn_type == DMU_OTN_ZAP_METADATA) {
	dnode_rele(dn, FTAG);
	return;
	}
	ASSERT3U(dn->dn_type, ==, old_type);
	ASSERT0(dn->dn_maxblkid);

	/*
	* We must initialize the ZAP data before changing the type,
	* so that concurrent calls to *_is_zapified() can determine if
	* the object has been completely zapified by checking the type.
	*/
	mzap_create_impl(dn, 0, 0, tx);

	dn->dn_next_type[tx->tx_txg & TXG_MASK] = dn->dn_type =
	DMU_OTN_ZAP_METADATA;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);

	spa_feature_incr(dmu_objset_spa(mos),
	SPA_FEATURE_EXTENSIBLE_DATASET, tx);
	}

	void
	dmu_object_free_zapified(objset_t mos, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	dmu_object_type_t t;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	t = dn->dn_type;
	dnode_rele(dn, FTAG);

	if (t == DMU_OTN_ZAP_METADATA) {
	spa_feature_decr(dmu_objset_spa(mos),
	SPA_FEATURE_EXTENSIBLE_DATASET, tx);
	}
	VERIFY0(dmu_object_free(mos, object, tx));
	}

	EXPORT_SYMBOL(dmu_object_alloc);
	EXPORT_SYMBOL(dmu_object_alloc_ibs);
	EXPORT_SYMBOL(dmu_object_alloc_dnsize);
	EXPORT_SYMBOL(dmu_object_alloc_hold);
	EXPORT_SYMBOL(dmu_object_claim);
	EXPORT_SYMBOL(dmu_object_claim_dnsize);
	EXPORT_SYMBOL(dmu_object_reclaim);
	EXPORT_SYMBOL(dmu_object_reclaim_dnsize);
	EXPORT_SYMBOL(dmu_object_rm_spill);
	EXPORT_SYMBOL(dmu_object_free);
	EXPORT_SYMBOL(dmu_object_next);
	EXPORT_SYMBOL(dmu_object_zapify);
	EXPORT_SYMBOL(dmu_object_free_zapified);

	/* BEGIN CSTYLED */
	-ZFS_MODULE_PARAM(zfs, , dmu_object_alloc_chunk_shift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, , dmu_object_alloc_chunk_shift, UINT, ZMOD_RW,
	"CPU-specific allocator grabs 2^N objects at once");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_recv.c b/sys/contrib/openzfs/module/zfs/dmu_recv.c
	index 55d03677feaa..a9e4a6745905 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_recv.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_recv.c
	@@ -1,3730 +1,3743 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2022 Axcient.
	*/

	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zvol.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_znode.h>
	#include <zfs_fletcher.h>
	#include <sys/avl.h>
	#include <sys/ddt.h>
	#include <sys/zfs_onexit.h>
	#include <sys/dsl_destroy.h>
	#include <sys/blkptr.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/zfeature.h>
	#include <sys/bqueue.h>
	#include <sys/objlist.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#endif
	#include <sys/zfs_file.h>

	-static int zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
	-static int zfs_recv_queue_ff = 20;
	-static int zfs_recv_write_batch_size = 1024 * 1024;
	+static uint_t zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
	+static uint_t zfs_recv_queue_ff = 20;
	+static uint_t zfs_recv_write_batch_size = 1024 * 1024;
	static int zfs_recv_best_effort_corrective = 0;

	static const void *const dmu_recv_tag = "dmu_recv_tag";
	const char *const recv_clone_name = "%recv";

	static int receive_read_payload_and_next_header(dmu_recv_cookie_t *ra, int len,
	void *buf);

	struct receive_record_arg {
	dmu_replay_record_t header;
	void payload; / Pointer to a buffer containing the payload */
	/*
	* If the record is a WRITE or SPILL, pointer to the abd containing the
	* payload.
	*/
	abd_t *abd;
	int payload_size;
	uint64_t bytes_read; /* bytes read from stream when record created */
	boolean_t eos_marker; /* Marks the end of the stream */
	bqueue_node_t node;
	};

	struct receive_writer_arg {
	objset_t *os;
	boolean_t byteswap;
	bqueue_t q;

	/*
	* These three members are used to signal to the main thread when
	* we're done.
	*/
	kmutex_t mutex;
	kcondvar_t cv;
	boolean_t done;

	int err;
	const char *tofs;
	boolean_t heal;
	boolean_t resumable;
	boolean_t raw; /* DMU_BACKUP_FEATURE_RAW set */
	boolean_t spill; /* DRR_FLAG_SPILL_BLOCK set */
	boolean_t full; /* this is a full send stream */
	uint64_t last_object;
	uint64_t last_offset;
	uint64_t max_object; /* highest object ID referenced in stream */
	uint64_t bytes_read; /* bytes read when current record created */

	list_t write_batch;

	/* Encryption parameters for the last received DRR_OBJECT_RANGE */
	boolean_t or_crypt_params_present;
	uint64_t or_firstobj;
	uint64_t or_numslots;
	uint8_t or_salt[ZIO_DATA_SALT_LEN];
	uint8_t or_iv[ZIO_DATA_IV_LEN];
	uint8_t or_mac[ZIO_DATA_MAC_LEN];
	boolean_t or_byteorder;
	zio_t *heal_pio;
	};

	typedef struct dmu_recv_begin_arg {
	const char *drba_origin;
	dmu_recv_cookie_t *drba_cookie;
	cred_t *drba_cred;
	proc_t *drba_proc;
	dsl_crypto_params_t *drba_dcp;
	} dmu_recv_begin_arg_t;

	static void
	byteswap_record(dmu_replay_record_t *drr)
	{
	#define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X))
	#define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X))
	drr->drr_type = BSWAP_32(drr->drr_type);
	drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen);

	switch (drr->drr_type) {
	case DRR_BEGIN:
	DO64(drr_begin.drr_magic);
	DO64(drr_begin.drr_versioninfo);
	DO64(drr_begin.drr_creation_time);
	DO32(drr_begin.drr_type);
	DO32(drr_begin.drr_flags);
	DO64(drr_begin.drr_toguid);
	DO64(drr_begin.drr_fromguid);
	break;
	case DRR_OBJECT:
	DO64(drr_object.drr_object);
	DO32(drr_object.drr_type);
	DO32(drr_object.drr_bonustype);
	DO32(drr_object.drr_blksz);
	DO32(drr_object.drr_bonuslen);
	DO32(drr_object.drr_raw_bonuslen);
	DO64(drr_object.drr_toguid);
	DO64(drr_object.drr_maxblkid);
	break;
	case DRR_FREEOBJECTS:
	DO64(drr_freeobjects.drr_firstobj);
	DO64(drr_freeobjects.drr_numobjs);
	DO64(drr_freeobjects.drr_toguid);
	break;
	case DRR_WRITE:
	DO64(drr_write.drr_object);
	DO32(drr_write.drr_type);
	DO64(drr_write.drr_offset);
	DO64(drr_write.drr_logical_size);
	DO64(drr_write.drr_toguid);
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum);
	DO64(drr_write.drr_key.ddk_prop);
	DO64(drr_write.drr_compressed_size);
	break;
	case DRR_WRITE_EMBEDDED:
	DO64(drr_write_embedded.drr_object);
	DO64(drr_write_embedded.drr_offset);
	DO64(drr_write_embedded.drr_length);
	DO64(drr_write_embedded.drr_toguid);
	DO32(drr_write_embedded.drr_lsize);
	DO32(drr_write_embedded.drr_psize);
	break;
	case DRR_FREE:
	DO64(drr_free.drr_object);
	DO64(drr_free.drr_offset);
	DO64(drr_free.drr_length);
	DO64(drr_free.drr_toguid);
	break;
	case DRR_SPILL:
	DO64(drr_spill.drr_object);
	DO64(drr_spill.drr_length);
	DO64(drr_spill.drr_toguid);
	DO64(drr_spill.drr_compressed_size);
	DO32(drr_spill.drr_type);
	break;
	case DRR_OBJECT_RANGE:
	DO64(drr_object_range.drr_firstobj);
	DO64(drr_object_range.drr_numslots);
	DO64(drr_object_range.drr_toguid);
	break;
	case DRR_REDACT:
	DO64(drr_redact.drr_object);
	DO64(drr_redact.drr_offset);
	DO64(drr_redact.drr_length);
	DO64(drr_redact.drr_toguid);
	break;
	case DRR_END:
	DO64(drr_end.drr_toguid);
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum);
	break;
	default:
	break;
	}

	if (drr->drr_type != DRR_BEGIN) {
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum);
	}

	#undef DO64
	#undef DO32
	}

	static boolean_t
	redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
	{
	for (int i = 0; i < num_snaps; i++) {
	if (snaps[i] == guid)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Check that the new stream we're trying to receive is redacted with respect to
	* a subset of the snapshots that the origin was redacted with respect to. For
	* the reasons behind this, see the man page on redacted zfs sends and receives.
	*/
	static boolean_t
	compatible_redact_snaps(uint64_t *origin_snaps, uint64_t origin_num_snaps,
	uint64_t *redact_snaps, uint64_t num_redact_snaps)
	{
	/*
	* Short circuit the comparison; if we are redacted with respect to
	* more snapshots than the origin, we can't be redacted with respect
	* to a subset.
	*/
	if (num_redact_snaps > origin_num_snaps) {
	return (B_FALSE);
	}

	for (int i = 0; i < num_redact_snaps; i++) {
	if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
	redact_snaps[i])) {
	return (B_FALSE);
	}
	}
	return (B_TRUE);
	}

	static boolean_t
	redact_check(dmu_recv_begin_arg_t drba, dsl_dataset_t origin)
	{
	uint64_t *origin_snaps;
	uint64_t origin_num_snaps;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	struct drr_begin *drrb = drc->drc_drrb;
	int featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	int err = 0;
	boolean_t ret = B_TRUE;
	uint64_t *redact_snaps;
	uint_t numredactsnaps;

	/*
	* If this is a full send stream, we're safe no matter what.
	*/
	if (drrb->drr_fromguid == 0)
	return (ret);

	VERIFY(dsl_dataset_get_uint64_array_feature(origin,
	SPA_FEATURE_REDACTED_DATASETS, &origin_num_snaps, &origin_snaps));

	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_FROM_SNAPS, &redact_snaps, &numredactsnaps) ==
	0) {
	/*
	* If the send stream was sent from the redaction bookmark or
	* the redacted version of the dataset, then we're safe. Verify
	* that this is from the a compatible redaction bookmark or
	* redacted dataset.
	*/
	if (!compatible_redact_snaps(origin_snaps, origin_num_snaps,
	redact_snaps, numredactsnaps)) {
	err = EINVAL;
	}
	} else if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	/*
	* If the stream is redacted, it must be redacted with respect
	* to a subset of what the origin is redacted with respect to.
	* See case number 2 in the zfs man page section on redacted zfs
	* send.
	*/
	err = nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps);

	if (err != 0 \|\| !compatible_redact_snaps(origin_snaps,
	origin_num_snaps, redact_snaps, numredactsnaps)) {
	err = EINVAL;
	}
	} else if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
	drrb->drr_toguid)) {
	/*
	* If the stream isn't redacted but the origin is, this must be
	* one of the snapshots the origin is redacted with respect to.
	* See case number 1 in the zfs man page section on redacted zfs
	* send.
	*/
	err = EINVAL;
	}

	if (err != 0)
	ret = B_FALSE;
	return (ret);
	}

	/*
	* If we previously received a stream with --large-block, we don't support
	* receiving an incremental on top of it without --large-block. This avoids
	* forcing a read-modify-write or trying to re-aggregate a string of WRITE
	* records.
	*/
	static int
	recv_check_large_blocks(dsl_dataset_t *ds, uint64_t featureflags)
	{
	if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_LARGE_BLOCKS) &&
	!(featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS))
	return (SET_ERROR(ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH));
	return (0);
	}

	static int
	recv_begin_check_existing_impl(dmu_recv_begin_arg_t drba, dsl_dataset_t ds,
	uint64_t fromguid, uint64_t featureflags)
	{
	uint64_t obj;
	uint64_t children;
	int error;
	dsl_dataset_t *snap;
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	boolean_t encrypted = ds->ds_dir->dd_crypto_obj != 0;
	boolean_t raw = (featureflags & DMU_BACKUP_FEATURE_RAW) != 0;
	boolean_t embed = (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) != 0;

	/* Temporary clone name must not exist. */
	error = zap_lookup(dp->dp_meta_objset,
	dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name,
	8, 1, &obj);
	if (error != ENOENT)
	return (error == 0 ? SET_ERROR(EBUSY) : error);

	/* Resume state must not be set. */
	if (dsl_dataset_has_resume_receive_state(ds))
	return (SET_ERROR(EBUSY));

	/* New snapshot name must not exist if we're not healing it. */
	error = zap_lookup(dp->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj,
	drba->drba_cookie->drc_tosnap, 8, 1, &obj);
	if (drba->drba_cookie->drc_heal) {
	if (error != 0)
	return (error);
	} else if (error != ENOENT) {
	return (error == 0 ? SET_ERROR(EEXIST) : error);
	}

	/* Must not have children if receiving a ZVOL. */
	error = zap_count(dp->dp_meta_objset,
	dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, &children);
	if (error != 0)
	return (error);
	if (drba->drba_cookie->drc_drrb->drr_type != DMU_OST_ZFS &&
	children > 0)
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));

	/*
	* Check snapshot limit before receiving. We'll recheck again at the
	* end, but might as well abort before receiving if we're already over
	* the limit.
	*
	* Note that we do not check the file system limit with
	* dsl_dir_fscount_check because the temporary %clones don't count
	* against that limit.
	*/
	error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT,
	NULL, drba->drba_cred, drba->drba_proc);
	if (error != 0)
	return (error);

	if (drba->drba_cookie->drc_heal) {
	/* Encryption is incompatible with embedded data. */
	if (encrypted && embed)
	return (SET_ERROR(EINVAL));

	/* Healing is not supported when in 'force' mode. */
	if (drba->drba_cookie->drc_force)
	return (SET_ERROR(EINVAL));

	/* Must have keys loaded if doing encrypted non-raw recv. */
	if (encrypted && !raw) {
	if (spa_keystore_lookup_key(dp->dp_spa, ds->ds_object,
	NULL, NULL) != 0)
	return (SET_ERROR(EACCES));
	}

	error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap);
	if (error != 0)
	return (error);

	/*
	* When not doing best effort corrective recv healing can only
	* be done if the send stream is for the same snapshot as the
	* one we are trying to heal.
	*/
	if (zfs_recv_best_effort_corrective == 0 &&
	drba->drba_cookie->drc_drrb->drr_toguid !=
	dsl_dataset_phys(snap)->ds_guid) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	dsl_dataset_rele(snap, FTAG);
	} else if (fromguid != 0) {
	/* Sanity check the incremental recv */
	uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;

	/* Can't perform a raw receive on top of a non-raw receive */
	if (!encrypted && raw)
	return (SET_ERROR(EINVAL));

	/* Encryption is incompatible with embedded data */
	if (encrypted && embed)
	return (SET_ERROR(EINVAL));

	/* Find snapshot in this dir that matches fromguid. */
	while (obj != 0) {
	error = dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap);
	if (error != 0)
	return (SET_ERROR(ENODEV));
	if (snap->ds_dir != ds->ds_dir) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ENODEV));
	}
	if (dsl_dataset_phys(snap)->ds_guid == fromguid)
	break;
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_dataset_rele(snap, FTAG);
	}
	if (obj == 0)
	return (SET_ERROR(ENODEV));

	if (drba->drba_cookie->drc_force) {
	drba->drba_cookie->drc_fromsnapobj = obj;
	} else {
	/*
	* If we are not forcing, there must be no
	* changes since fromsnap. Raw sends have an
	* additional constraint that requires that
	* no "noop" snapshots exist between fromsnap
	* and tosnap for the IVset checking code to
	* work properly.
	*/
	if (dsl_dataset_modified_since_snap(ds, snap) \|\|
	(raw &&
	dsl_dataset_phys(ds)->ds_prev_snap_obj !=
	snap->ds_object)) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ETXTBSY));
	}
	drba->drba_cookie->drc_fromsnapobj =
	ds->ds_prev->ds_object;
	}

	if (dsl_dataset_feature_is_active(snap,
	SPA_FEATURE_REDACTED_DATASETS) && !redact_check(drba,
	snap)) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = recv_check_large_blocks(snap, featureflags);
	if (error != 0) {
	dsl_dataset_rele(snap, FTAG);
	return (error);
	}

	dsl_dataset_rele(snap, FTAG);
	} else {
	/* If full and not healing then must be forced. */
	if (!drba->drba_cookie->drc_force)
	return (SET_ERROR(EEXIST));

	/*
	* We don't support using zfs recv -F to blow away
	* encrypted filesystems. This would require the
	* dsl dir to point to the old encryption key and
	* the new one at the same time during the receive.
	*/
	if ((!encrypted && raw) \|\| encrypted)
	return (SET_ERROR(EINVAL));

	/*
	* Perform the same encryption checks we would if
	* we were creating a new dataset from scratch.
	*/
	if (!raw) {
	boolean_t will_encrypt;

	error = dmu_objset_create_crypt_check(
	ds->ds_dir->dd_parent, drba->drba_dcp,
	&will_encrypt);
	if (error != 0)
	return (error);

	if (will_encrypt && embed)
	return (SET_ERROR(EINVAL));
	}
	}

	return (0);
	}

	/*
	* Check that any feature flags used in the data stream we're receiving are
	* supported by the pool we are receiving into.
	*
	* Note that some of the features we explicitly check here have additional
	* (implicit) features they depend on, but those dependencies are enforced
	* through the zfeature_register() calls declaring the features that we
	* explicitly check.
	*/
	static int
	recv_begin_check_feature_flags_impl(uint64_t featureflags, spa_t *spa)
	{
	/*
	* Check if there are any unsupported feature flags.
	*/
	if (!DMU_STREAM_SUPPORTED(featureflags)) {
	return (SET_ERROR(ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE));
	}

	/* Verify pool version supports SA if SA_SPILL feature set */
	if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) &&
	spa_version(spa) < SPA_VERSION_SA)
	return (SET_ERROR(ENOTSUP));

	/*
	* LZ4 compressed, ZSTD compressed, embedded, mooched, large blocks,
	* and large_dnodes in the stream can only be used if those pool
	* features are enabled because we don't attempt to decompress /
	* un-embed / un-mooch / split up the blocks / dnodes during the
	* receive process.
	*/
	if ((featureflags & DMU_BACKUP_FEATURE_LZ4) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_ZSTD) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_ZSTD_COMPRESS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
	return (SET_ERROR(ENOTSUP));

	/*
	* Receiving redacted streams requires that redacted datasets are
	* enabled.
	*/
	if ((featureflags & DMU_BACKUP_FEATURE_REDACTED) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_REDACTED_DATASETS))
	return (SET_ERROR(ENOTSUP));

	return (0);
	}

	static int
	dmu_recv_begin_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	struct drr_begin *drrb = drba->drba_cookie->drc_drrb;
	uint64_t fromguid = drrb->drr_fromguid;
	int flags = drrb->drr_flags;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	int error;
	uint64_t featureflags = drba->drba_cookie->drc_featureflags;
	dsl_dataset_t *ds;
	const char *tofs = drba->drba_cookie->drc_tofs;

	/* already checked */
	ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
	ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING));

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM \|\|
	drrb->drr_type >= DMU_OST_NUMTYPES \|\|
	((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL))
	return (SET_ERROR(EINVAL));

	error = recv_begin_check_feature_flags_impl(featureflags, dp->dp_spa);
	if (error != 0)
	return (error);

	/* Resumable receives require extensible datasets */
	if (drba->drba_cookie->drc_resumable &&
	!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET))
	return (SET_ERROR(ENOTSUP));

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	/* raw receives require the encryption feature */
	if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_ENCRYPTION))
	return (SET_ERROR(ENOTSUP));

	/* embedded data is incompatible with encryption and raw recv */
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
	return (SET_ERROR(EINVAL));

	/* raw receives require spill block allocation flag */
	if (!(flags & DRR_FLAG_SPILL_BLOCK))
	return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
	} else {
	/*
	* We support unencrypted datasets below encrypted ones now,
	* so add the DS_HOLD_FLAG_DECRYPT flag only if we are dealing
	* with a dataset we may encrypt.
	*/
	if (drba->drba_dcp != NULL &&
	drba->drba_dcp->cp_crypt != ZIO_CRYPT_OFF) {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}
	}

	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error == 0) {
	/* target fs already exists; recv into temp clone */

	/* Can't recv a clone into an existing fs */
	if (flags & DRR_FLAG_CLONE \|\| drba->drba_origin) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = recv_begin_check_existing_impl(drba, ds, fromguid,
	featureflags);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	} else if (error == ENOENT) {
	/* target fs does not exist; must be a full backup or clone */
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	objset_t *os;

	/* healing recv must be done "into" an existing snapshot */
	if (drba->drba_cookie->drc_heal == B_TRUE)
	return (SET_ERROR(ENOTSUP));

	/*
	* If it's a non-clone incremental, we are missing the
	* target fs, so fail the recv.
	*/
	if (fromguid != 0 && !((flags & DRR_FLAG_CLONE) \|\|
	drba->drba_origin))
	return (SET_ERROR(ENOENT));

	/*
	* If we're receiving a full send as a clone, and it doesn't
	* contain all the necessary free records and freeobject
	* records, reject it.
	*/
	if (fromguid == 0 && drba->drba_origin != NULL &&
	!(flags & DRR_FLAG_FREERECORDS))
	return (SET_ERROR(EINVAL));

	/* Open the parent of tofs */
	ASSERT3U(strlen(tofs), <, sizeof (buf));
	(void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1);
	error = dsl_dataset_hold(dp, buf, FTAG, &ds);
	if (error != 0)
	return (error);

	if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
	drba->drba_origin == NULL) {
	boolean_t will_encrypt;

	/*
	* Check that we aren't breaking any encryption rules
	* and that we have all the parameters we need to
	* create an encrypted dataset if necessary. If we are
	* making an encrypted dataset the stream can't have
	* embedded data.
	*/
	error = dmu_objset_create_crypt_check(ds->ds_dir,
	drba->drba_dcp, &will_encrypt);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (will_encrypt &&
	(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	/*
	* Check filesystem and snapshot limits before receiving. We'll
	* recheck snapshot limits again at the end (we create the
	* filesystems and increment those counts during begin_sync).
	*/
	error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
	ZFS_PROP_FILESYSTEM_LIMIT, NULL,
	drba->drba_cred, drba->drba_proc);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
	ZFS_PROP_SNAPSHOT_LIMIT, NULL,
	drba->drba_cred, drba->drba_proc);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	/* can't recv below anything but filesystems (eg. no ZVOLs) */
	error = dmu_objset_from_ds(ds, &os);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
	}

	if (drba->drba_origin != NULL) {
	dsl_dataset_t *origin;
	error = dsl_dataset_hold_flags(dp, drba->drba_origin,
	dsflags, FTAG, &origin);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	if (!origin->ds_is_snapshot) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}
	if (dsl_dataset_phys(origin)->ds_guid != fromguid &&
	fromguid != 0) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENODEV));
	}

	if (origin->ds_dir->dd_crypto_obj != 0 &&
	(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If the origin is redacted we need to verify that this
	* send stream can safely be received on top of the
	* origin.
	*/
	if (dsl_dataset_feature_is_active(origin,
	SPA_FEATURE_REDACTED_DATASETS)) {
	if (!redact_check(drba, origin)) {
	dsl_dataset_rele_flags(origin, dsflags,
	FTAG);
	dsl_dataset_rele_flags(ds, dsflags,
	FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	error = recv_check_large_blocks(ds, featureflags);
	if (error != 0) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (error);
	}

	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	}

	dsl_dataset_rele(ds, FTAG);
	error = 0;
	}
	return (error);
	}

	static void
	dmu_recv_begin_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	objset_t *mos = dp->dp_meta_objset;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	struct drr_begin *drrb = drc->drc_drrb;
	const char *tofs = drc->drc_tofs;
	uint64_t featureflags = drc->drc_featureflags;
	dsl_dataset_t ds, newds;
	objset_t *os;
	uint64_t dsobj;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	int error;
	uint64_t crflags = 0;
	dsl_crypto_params_t dummy_dcp = { 0 };
	dsl_crypto_params_t *dcp = drba->drba_dcp;

	if (drrb->drr_flags & DRR_FLAG_CI_DATA)
	crflags \|= DS_FLAG_CI_DATASET;

	if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0)
	dsflags \|= DS_HOLD_FLAG_DECRYPT;

	/*
	* Raw, non-incremental recvs always use a dummy dcp with
	* the raw cmd set. Raw incremental recvs do not use a dcp
	* since the encryption parameters are already set in stone.
	*/
	if (dcp == NULL && drrb->drr_fromguid == 0 &&
	drba->drba_origin == NULL) {
	ASSERT3P(dcp, ==, NULL);
	dcp = &dummy_dcp;

	if (featureflags & DMU_BACKUP_FEATURE_RAW)
	dcp->cp_cmd = DCP_CMD_RAW_RECV;
	}

	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error == 0) {
	/* Create temporary clone unless we're doing corrective recv */
	dsl_dataset_t *snap = NULL;

	if (drba->drba_cookie->drc_fromsnapobj != 0) {
	VERIFY0(dsl_dataset_hold_obj(dp,
	drba->drba_cookie->drc_fromsnapobj, FTAG, &snap));
	ASSERT3P(dcp, ==, NULL);
	}
	if (drc->drc_heal) {
	/* When healing we want to use the provided snapshot */
	VERIFY0(dsl_dataset_snap_lookup(ds, drc->drc_tosnap,
	&dsobj));
	} else {
	dsobj = dsl_dataset_create_sync(ds->ds_dir,
	recv_clone_name, snap, crflags, drba->drba_cred,
	dcp, tx);
	}
	if (drba->drba_cookie->drc_fromsnapobj != 0)
	dsl_dataset_rele(snap, FTAG);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	} else {
	dsl_dir_t *dd;
	const char *tail;
	dsl_dataset_t *origin = NULL;

	VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail));

	if (drba->drba_origin != NULL) {
	VERIFY0(dsl_dataset_hold(dp, drba->drba_origin,
	FTAG, &origin));
	ASSERT3P(dcp, ==, NULL);
	}

	/* Create new dataset. */
	dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1,
	origin, crflags, drba->drba_cred, dcp, tx);
	if (origin != NULL)
	dsl_dataset_rele(origin, FTAG);
	dsl_dir_rele(dd, FTAG);
	drc->drc_newfs = B_TRUE;
	}
	VERIFY0(dsl_dataset_own_obj_force(dp, dsobj, dsflags, dmu_recv_tag,
	&newds));
	if (dsl_dataset_feature_is_active(newds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	/*
	* If the origin dataset is redacted, the child will be redacted
	* when we create it. We clear the new dataset's
	* redaction info; if it should be redacted, we'll fill
	* in its information later.
	*/
	dsl_dataset_deactivate_feature(newds,
	SPA_FEATURE_REDACTED_DATASETS, tx);
	}
	VERIFY0(dmu_objset_from_ds(newds, &os));

	if (drc->drc_resumable) {
	dsl_dataset_zapify(newds, tx);
	if (drrb->drr_fromguid != 0) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID,
	8, 1, &drrb->drr_fromguid, tx));
	}
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID,
	8, 1, &drrb->drr_toguid, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME,
	1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx));
	uint64_t one = 1;
	uint64_t zero = 0;
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT,
	8, 1, &one, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET,
	8, 1, &zero, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES,
	8, 1, &zero, tx));
	if (featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_LARGEBLOCK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_COMPRESSED) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_COMPRESSOK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_RAWOK,
	8, 1, &one, tx));
	}

	uint64_t *redact_snaps;
	uint_t numredactsnaps;
	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_FROM_SNAPS, &redact_snaps,
	&numredactsnaps) == 0) {
	VERIFY0(zap_add(mos, dsobj,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS,
	sizeof (*redact_snaps), numredactsnaps,
	redact_snaps, tx));
	}
	}

	/*
	* Usually the os->os_encrypted value is tied to the presence of a
	* DSL Crypto Key object in the dd. However, that will not be received
	* until dmu_recv_stream(), so we set the value manually for now.
	*/
	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	os->os_encrypted = B_TRUE;
	drba->drba_cookie->drc_raw = B_TRUE;
	}

	if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	uint64_t *redact_snaps;
	uint_t numredactsnaps;
	VERIFY0(nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps));
	dsl_dataset_activate_redaction(newds, redact_snaps,
	numredactsnaps, tx);
	}

	dmu_buf_will_dirty(newds->ds_dbuf, tx);
	dsl_dataset_phys(newds)->ds_flags \|= DS_FLAG_INCONSISTENT;

	/*
	* If we actually created a non-clone, we need to create the objset
	* in our new dataset. If this is a raw send we postpone this until
	* dmu_recv_stream() so that we can allocate the metadnode with the
	* properties from the DRR_BEGIN payload.
	*/
	rrw_enter(&newds->ds_bp_rwlock, RW_READER, FTAG);
	if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds)) &&
	(featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
	!drc->drc_heal) {
	(void) dmu_objset_create_impl(dp->dp_spa,
	newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx);
	}
	rrw_exit(&newds->ds_bp_rwlock, FTAG);

	drba->drba_cookie->drc_ds = newds;
	drba->drba_cookie->drc_os = os;

	spa_history_log_internal_ds(newds, "receive", tx, " ");
	}

	static int
	dmu_recv_resume_begin_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	struct drr_begin *drrb = drc->drc_drrb;
	int error;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	dsl_dataset_t *ds;
	const char *tofs = drc->drc_tofs;

	/* already checked */
	ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
	ASSERT(drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING);

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM \|\|
	drrb->drr_type >= DMU_OST_NUMTYPES)
	return (SET_ERROR(EINVAL));

	/*
	* This is mostly a sanity check since we should have already done these
	* checks during a previous attempt to receive the data.
	*/
	error = recv_begin_check_feature_flags_impl(drc->drc_featureflags,
	dp->dp_spa);
	if (error != 0)
	return (error);

	/* 6 extra bytes for /%recv */
	char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];

	(void) snprintf(recvname, sizeof (recvname), "%s/%s",
	tofs, recv_clone_name);

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	/* raw receives require spill block allocation flag */
	if (!(drrb->drr_flags & DRR_FLAG_SPILL_BLOCK))
	return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
	} else {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}

	+ boolean_t recvexist = B_TRUE;
	if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) {
	/* %recv does not exist; continue in tofs */
	+ recvexist = B_FALSE;
	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error != 0)
	return (error);
	}

	+ /*
	+ * Resume of full/newfs recv on existing dataset should be done with
	+ * force flag
	+ */
	+ if (recvexist && drrb->drr_fromguid == 0 && !drc->drc_force) {
	+ dsl_dataset_rele_flags(ds, dsflags, FTAG);
	+ return (SET_ERROR(ZFS_ERR_RESUME_EXISTS));
	+ }
	+
	/* check that ds is marked inconsistent */
	if (!DS_IS_INCONSISTENT(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/* check that there is resuming data, and that the toguid matches */
	if (!dsl_dataset_is_zapified(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}
	uint64_t val;
	error = zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val);
	if (error != 0 \|\| drrb->drr_toguid != val) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Check if the receive is still running. If so, it will be owned.
	* Note that nothing else can own the dataset (e.g. after the receive
	* fails) because it will be marked inconsistent.
	*/
	if (dsl_dataset_has_owner(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EBUSY));
	}

	/* There should not be any snapshots of this fs yet. */
	if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Note: resume point will be checked when we process the first WRITE
	* record.
	*/

	/* check that the origin matches */
	val = 0;
	(void) zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val);
	if (drrb->drr_fromguid != val) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (ds->ds_prev != NULL && drrb->drr_fromguid != 0)
	drc->drc_fromsnapobj = ds->ds_prev->ds_object;

	/*
	* If we're resuming, and the send is redacted, then the original send
	* must have been redacted, and must have been redacted with respect to
	* the same snapshots.
	*/
	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	uint64_t num_ds_redact_snaps;
	uint64_t *ds_redact_snaps;

	uint_t num_stream_redact_snaps;
	uint64_t *stream_redact_snaps;

	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &stream_redact_snaps,
	&num_stream_redact_snaps) != 0) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (!dsl_dataset_get_uint64_array_feature(ds,
	SPA_FEATURE_REDACTED_DATASETS, &num_ds_redact_snaps,
	&ds_redact_snaps)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	for (int i = 0; i < num_ds_redact_snaps; i++) {
	if (!redact_snaps_contains(ds_redact_snaps,
	num_ds_redact_snaps, stream_redact_snaps[i])) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}
	}

	error = recv_check_large_blocks(ds, drc->drc_featureflags);
	if (error != 0) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (error);
	}

	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (0);
	}

	static void
	dmu_recv_resume_begin_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	const char *tofs = drba->drba_cookie->drc_tofs;
	uint64_t featureflags = drba->drba_cookie->drc_featureflags;
	dsl_dataset_t *ds;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	/* 6 extra bytes for /%recv */
	char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];

	(void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs,
	recv_clone_name);

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	drba->drba_cookie->drc_raw = B_TRUE;
	} else {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}

	if (dsl_dataset_own_force(dp, recvname, dsflags, dmu_recv_tag, &ds)
	!= 0) {
	/* %recv does not exist; continue in tofs */
	VERIFY0(dsl_dataset_own_force(dp, tofs, dsflags, dmu_recv_tag,
	&ds));
	drba->drba_cookie->drc_newfs = B_TRUE;
	}

	ASSERT(DS_IS_INCONSISTENT(ds));
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds)) \|\|
	drba->drba_cookie->drc_raw);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	drba->drba_cookie->drc_ds = ds;
	VERIFY0(dmu_objset_from_ds(ds, &drba->drba_cookie->drc_os));
	drba->drba_cookie->drc_should_save = B_TRUE;

	spa_history_log_internal_ds(ds, "resume receive", tx, " ");
	}

	/*
	* NB: callers MUST call dmu_recv_stream() if dmu_recv_begin()
	* succeeds; otherwise we will leak the holds on the datasets.
	*/
	int
	dmu_recv_begin(char tofs, char tosnap, dmu_replay_record_t *drr_begin,
	boolean_t force, boolean_t heal, boolean_t resumable, nvlist_t *localprops,
	nvlist_t hidden_args, char origin, dmu_recv_cookie_t *drc,
	zfs_file_t fp, offset_t voffp)
	{
	dmu_recv_begin_arg_t drba = { 0 };
	int err;

	memset(drc, 0, sizeof (dmu_recv_cookie_t));
	drc->drc_drr_begin = drr_begin;
	drc->drc_drrb = &drr_begin->drr_u.drr_begin;
	drc->drc_tosnap = tosnap;
	drc->drc_tofs = tofs;
	drc->drc_force = force;
	drc->drc_heal = heal;
	drc->drc_resumable = resumable;
	drc->drc_cred = CRED();
	drc->drc_proc = curproc;
	drc->drc_clone = (origin != NULL);

	if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
	drc->drc_byteswap = B_TRUE;
	(void) fletcher_4_incremental_byteswap(drr_begin,
	sizeof (dmu_replay_record_t), &drc->drc_cksum);
	byteswap_record(drr_begin);
	} else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) {
	(void) fletcher_4_incremental_native(drr_begin,
	sizeof (dmu_replay_record_t), &drc->drc_cksum);
	} else {
	return (SET_ERROR(EINVAL));
	}

	drc->drc_fp = fp;
	drc->drc_voff = *voffp;
	drc->drc_featureflags =
	DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo);

	uint32_t payloadlen = drc->drc_drr_begin->drr_payloadlen;
	void *payload = NULL;
	if (payloadlen != 0)
	payload = kmem_alloc(payloadlen, KM_SLEEP);

	err = receive_read_payload_and_next_header(drc, payloadlen,
	payload);
	if (err != 0) {
	kmem_free(payload, payloadlen);
	return (err);
	}
	if (payloadlen != 0) {
	err = nvlist_unpack(payload, payloadlen, &drc->drc_begin_nvl,
	KM_SLEEP);
	kmem_free(payload, payloadlen);
	if (err != 0) {
	kmem_free(drc->drc_next_rrd,
	sizeof (*drc->drc_next_rrd));
	return (err);
	}
	}

	if (drc->drc_drrb->drr_flags & DRR_FLAG_SPILL_BLOCK)
	drc->drc_spill = B_TRUE;

	drba.drba_origin = origin;
	drba.drba_cookie = drc;
	drba.drba_cred = CRED();
	drba.drba_proc = curproc;

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
	err = dsl_sync_task(tofs,
	dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync,
	&drba, 5, ZFS_SPACE_CHECK_NORMAL);
	} else {
	/*
	* For non-raw, non-incremental, non-resuming receives the
	* user can specify encryption parameters on the command line
	* with "zfs recv -o". For these receives we create a dcp and
	* pass it to the sync task. Creating the dcp will implicitly
	* remove the encryption params from the localprops nvlist,
	* which avoids errors when trying to set these normally
	* read-only properties. Any other kind of receive that
	* attempts to set these properties will fail as a result.
	*/
	if ((DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RAW) == 0 &&
	origin == NULL && drc->drc_drrb->drr_fromguid == 0) {
	err = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
	localprops, hidden_args, &drba.drba_dcp);
	}

	if (err == 0) {
	err = dsl_sync_task(tofs,
	dmu_recv_begin_check, dmu_recv_begin_sync,
	&drba, 5, ZFS_SPACE_CHECK_NORMAL);
	dsl_crypto_params_free(drba.drba_dcp, !!err);
	}
	}

	if (err != 0) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	nvlist_free(drc->drc_begin_nvl);
	}
	return (err);
	}

	/*
	* Holds data need for corrective recv callback
	*/
	typedef struct cr_cb_data {
	uint64_t size;
	zbookmark_phys_t zb;
	spa_t *spa;
	} cr_cb_data_t;

	static void
	corrective_read_done(zio_t *zio)
	{
	cr_cb_data_t *data = zio->io_private;
	/* Corruption corrected; update error log if needed */
	if (zio->io_error == 0)
	spa_remove_error(data->spa, &data->zb);
	kmem_free(data, sizeof (cr_cb_data_t));
	abd_free(zio->io_abd);
	}

	/*
	* zio_rewrite the data pointed to by bp with the data from the rrd's abd.
	*/
	static int
	do_corrective_recv(struct receive_writer_arg rwa, struct drr_write drrw,
	struct receive_record_arg rrd, blkptr_t bp)
	{
	int err;
	zio_t *io;
	zbookmark_phys_t zb;
	dnode_t *dn;
	abd_t *abd = rrd->abd;
	zio_cksum_t bp_cksum = bp->blk_cksum;
	enum zio_flag flags = ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_CANFAIL;

	if (rwa->raw)
	flags \|= ZIO_FLAG_RAW;

	err = dnode_hold(rwa->os, drrw->drr_object, FTAG, &dn);
	if (err != 0)
	return (err);
	SET_BOOKMARK(&zb, dmu_objset_id(rwa->os), drrw->drr_object, 0,
	dbuf_whichblock(dn, 0, drrw->drr_offset));
	dnode_rele(dn, FTAG);

	if (!rwa->raw && DRR_WRITE_COMPRESSED(drrw)) {
	/* Decompress the stream data */
	abd_t *dabd = abd_alloc_linear(
	drrw->drr_logical_size, B_FALSE);
	err = zio_decompress_data(drrw->drr_compressiontype,
	abd, abd_to_buf(dabd), abd_get_size(abd),
	abd_get_size(dabd), NULL);

	if (err != 0) {
	abd_free(dabd);
	return (err);
	}
	/* Swap in the newly decompressed data into the abd */
	abd_free(abd);
	abd = dabd;
	}

	if (!rwa->raw && BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
	/* Recompress the data */
	abd_t *cabd = abd_alloc_linear(BP_GET_PSIZE(bp),
	B_FALSE);
	uint64_t csize = zio_compress_data(BP_GET_COMPRESS(bp),
	abd, abd_to_buf(cabd), abd_get_size(abd),
	rwa->os->os_complevel);
	abd_zero_off(cabd, csize, BP_GET_PSIZE(bp) - csize);
	/* Swap in newly compressed data into the abd */
	abd_free(abd);
	abd = cabd;
	flags \|= ZIO_FLAG_RAW_COMPRESS;
	}

	/*
	* The stream is not encrypted but the data on-disk is.
	* We need to re-encrypt the buf using the same
	* encryption type, salt, iv, and mac that was used to encrypt
	* the block previosly.
	*/
	if (!rwa->raw && BP_USES_CRYPT(bp)) {
	dsl_dataset_t *ds;
	dsl_crypto_key_t *dck = NULL;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;
	dsl_pool_t *dp = dmu_objset_pool(rwa->os);
	abd_t *eabd = abd_alloc_linear(BP_GET_PSIZE(bp), B_FALSE);

	zio_crypt_decode_params_bp(bp, salt, iv);
	zio_crypt_decode_mac_bp(bp, mac);

	dsl_pool_config_enter(dp, FTAG);
	err = dsl_dataset_hold_flags(dp, rwa->tofs,
	DS_HOLD_FLAG_DECRYPT, FTAG, &ds);
	if (err != 0) {
	dsl_pool_config_exit(dp, FTAG);
	abd_free(eabd);
	return (SET_ERROR(EACCES));
	}

	/* Look up the key from the spa's keystore */
	err = spa_keystore_lookup_key(rwa->os->os_spa,
	zb.zb_objset, FTAG, &dck);
	if (err != 0) {
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT,
	FTAG);
	dsl_pool_config_exit(dp, FTAG);
	abd_free(eabd);
	return (SET_ERROR(EACCES));
	}

	err = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
	BP_GET_TYPE(bp), BP_SHOULD_BYTESWAP(bp), salt, iv,
	mac, abd_get_size(abd), abd, eabd, &no_crypt);

	spa_keystore_dsl_key_rele(rwa->os->os_spa, dck, FTAG);
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
	dsl_pool_config_exit(dp, FTAG);

	ASSERT0(no_crypt);
	if (err != 0) {
	abd_free(eabd);
	return (err);
	}
	/* Swap in the newly encrypted data into the abd */
	abd_free(abd);
	abd = eabd;

	/*
	* We want to prevent zio_rewrite() from trying to
	* encrypt the data again
	*/
	flags \|= ZIO_FLAG_RAW_ENCRYPT;
	}
	rrd->abd = abd;

	io = zio_rewrite(NULL, rwa->os->os_spa, bp->blk_birth, bp, abd,
	BP_GET_PSIZE(bp), NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, flags, &zb);

	ASSERT(abd_get_size(abd) == BP_GET_LSIZE(bp) \|\|
	abd_get_size(abd) == BP_GET_PSIZE(bp));

	/* compute new bp checksum value and make sure it matches the old one */
	zio_checksum_compute(io, BP_GET_CHECKSUM(bp), abd, abd_get_size(abd));
	if (!ZIO_CHECKSUM_EQUAL(bp_cksum, io->io_bp->blk_cksum)) {
	zio_destroy(io);
	if (zfs_recv_best_effort_corrective != 0)
	return (0);
	return (SET_ERROR(ECKSUM));
	}

	/* Correct the corruption in place */
	err = zio_wait(io);
	if (err == 0) {
	cr_cb_data_t *cb_data =
	kmem_alloc(sizeof (cr_cb_data_t), KM_SLEEP);
	cb_data->spa = rwa->os->os_spa;
	cb_data->size = drrw->drr_logical_size;
	cb_data->zb = zb;
	/* Test if healing worked by re-reading the bp */
	err = zio_wait(zio_read(rwa->heal_pio, rwa->os->os_spa, bp,
	abd_alloc_for_io(drrw->drr_logical_size, B_FALSE),
	drrw->drr_logical_size, corrective_read_done,
	cb_data, ZIO_PRIORITY_ASYNC_READ, flags, NULL));
	}
	if (err != 0 && zfs_recv_best_effort_corrective != 0)
	err = 0;

	return (err);
	}

	static int
	receive_read(dmu_recv_cookie_t drc, int len, void buf)
	{
	int done = 0;

	/*
	* The code doesn't rely on this (lengths being multiples of 8). See
	* comment in dump_bytes.
	*/
	ASSERT(len % 8 == 0 \|\|
	(drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) != 0);

	while (done < len) {
	ssize_t resid;
	zfs_file_t *fp = drc->drc_fp;
	int err = zfs_file_read(fp, (char *)buf + done,
	len - done, &resid);
	if (resid == len - done) {
	/*
	* Note: ECKSUM or ZFS_ERR_STREAM_TRUNCATED indicates
	* that the receive was interrupted and can
	* potentially be resumed.
	*/
	err = SET_ERROR(ZFS_ERR_STREAM_TRUNCATED);
	}
	drc->drc_voff += len - done - resid;
	done = len - resid;
	if (err != 0)
	return (err);
	}

	drc->drc_bytes_read += len;

	ASSERT3U(done, ==, len);
	return (0);
	}

	static inline uint8_t
	deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size)
	{
	if (bonus_type == DMU_OT_SA) {
	return (1);
	} else {
	return (1 +
	((DN_OLD_MAX_BONUSLEN -
	MIN(DN_OLD_MAX_BONUSLEN, bonus_size)) >> SPA_BLKPTRSHIFT));
	}
	}

	static void
	save_resume_state(struct receive_writer_arg *rwa,
	uint64_t object, uint64_t offset, dmu_tx_t *tx)
	{
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;

	if (!rwa->resumable)
	return;

	/*
	* We use ds_resume_bytes[] != 0 to indicate that we need to
	* update this on disk, so it must not be 0.
	*/
	ASSERT(rwa->bytes_read != 0);

	/*
	* We only resume from write records, which have a valid
	* (non-meta-dnode) object number.
	*/
	ASSERT(object != 0);

	/*
	* For resuming to work correctly, we must receive records in order,
	* sorted by object,offset. This is checked by the callers, but
	* assert it here for good measure.
	*/
	ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]);
	ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] \|\|
	offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]);
	ASSERT3U(rwa->bytes_read, >=,
	rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]);

	rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object;
	rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset;
	rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read;
	}

	static int
	receive_object_is_same_generation(objset_t *os, uint64_t object,
	dmu_object_type_t old_bonus_type, dmu_object_type_t new_bonus_type,
	const void new_bonus, boolean_t samegenp)
	{
	zfs_file_info_t zoi;
	int err;

	dmu_buf_t *old_bonus_dbuf;
	err = dmu_bonus_hold(os, object, FTAG, &old_bonus_dbuf);
	if (err != 0)
	return (err);
	err = dmu_get_file_info(os, old_bonus_type, old_bonus_dbuf->db_data,
	&zoi);
	dmu_buf_rele(old_bonus_dbuf, FTAG);
	if (err != 0)
	return (err);
	uint64_t old_gen = zoi.zfi_generation;

	err = dmu_get_file_info(os, new_bonus_type, new_bonus, &zoi);
	if (err != 0)
	return (err);
	uint64_t new_gen = zoi.zfi_generation;

	*samegenp = (old_gen == new_gen);
	return (0);
	}

	static int
	receive_handle_existing_object(const struct receive_writer_arg *rwa,
	const struct drr_object drro, const dmu_object_info_t doi,
	const void *bonus_data,
	uint64_t object_to_hold, uint32_t new_blksz)
	{
	uint32_t indblksz = drro->drr_indblkshift ?
	1ULL << drro->drr_indblkshift : 0;
	int nblkptr = deduce_nblkptr(drro->drr_bonustype,
	drro->drr_bonuslen);
	uint8_t dn_slots = drro->drr_dn_slots != 0 ?
	drro->drr_dn_slots : DNODE_MIN_SLOTS;
	boolean_t do_free_range = B_FALSE;
	int err;

	*object_to_hold = drro->drr_object;

	/* nblkptr should be bounded by the bonus size and type */
	if (rwa->raw && nblkptr != drro->drr_nblkptr)
	return (SET_ERROR(EINVAL));

	/*
	* After the previous send stream, the sending system may
	* have freed this object, and then happened to re-allocate
	* this object number in a later txg. In this case, we are
	* receiving a different logical file, and the block size may
	* appear to be different. i.e. we may have a different
	* block size for this object than what the send stream says.
	* In this case we need to remove the object's contents,
	* so that its structure can be changed and then its contents
	* entirely replaced by subsequent WRITE records.
	*
	* If this is a -L (--large-block) incremental stream, and
	* the previous stream was not -L, the block size may appear
	* to increase. i.e. we may have a smaller block size for
	* this object than what the send stream says. In this case
	* we need to keep the object's contents and block size
	* intact, so that we don't lose parts of the object's
	* contents that are not changed by this incremental send
	* stream.
	*
	* We can distinguish between the two above cases by using
	* the ZPL's generation number (see
	* receive_object_is_same_generation()). However, we only
	* want to rely on the generation number when absolutely
	* necessary, because with raw receives, the generation is
	* encrypted. We also want to minimize dependence on the
	* ZPL, so that other types of datasets can also be received
	* (e.g. ZVOLs, although note that ZVOLS currently do not
	* reallocate their objects or change their structure).
	* Therefore, we check a number of different cases where we
	* know it is safe to discard the object's contents, before
	* using the ZPL's generation number to make the above
	* distinction.
	*/
	if (drro->drr_blksz != doi->doi_data_block_size) {
	if (rwa->raw) {
	/*
	* RAW streams always have large blocks, so
	* we are sure that the data is not needed
	* due to changing --large-block to be on.
	* Which is fortunate since the bonus buffer
	* (which contains the ZPL generation) is
	* encrypted, and the key might not be
	* loaded.
	*/
	do_free_range = B_TRUE;
	} else if (rwa->full) {
	/*
	* This is a full send stream, so it always
	* replaces what we have. Even if the
	* generation numbers happen to match, this
	* can not actually be the same logical file.
	* This is relevant when receiving a full
	* send as a clone.
	*/
	do_free_range = B_TRUE;
	} else if (drro->drr_type !=
	DMU_OT_PLAIN_FILE_CONTENTS \|\|
	doi->doi_type != DMU_OT_PLAIN_FILE_CONTENTS) {
	/*
	* PLAIN_FILE_CONTENTS are the only type of
	* objects that have ever been stored with
	* large blocks, so we don't need the special
	* logic below. ZAP blocks can shrink (when
	* there's only one block), so we don't want
	* to hit the error below about block size
	* only increasing.
	*/
	do_free_range = B_TRUE;
	} else if (doi->doi_max_offset <=
	doi->doi_data_block_size) {
	/*
	* There is only one block. We can free it,
	* because its contents will be replaced by a
	* WRITE record. This can not be the no-L ->
	* -L case, because the no-L case would have
	* resulted in multiple blocks. If we
	* supported -L -> no-L, it would not be safe
	* to free the file's contents. Fortunately,
	* that is not allowed (see
	* recv_check_large_blocks()).
	*/
	do_free_range = B_TRUE;
	} else {
	boolean_t is_same_gen;
	err = receive_object_is_same_generation(rwa->os,
	drro->drr_object, doi->doi_bonus_type,
	drro->drr_bonustype, bonus_data, &is_same_gen);
	if (err != 0)
	return (SET_ERROR(EINVAL));

	if (is_same_gen) {
	/*
	* This is the same logical file, and
	* the block size must be increasing.
	* It could only decrease if
	* --large-block was changed to be
	* off, which is checked in
	* recv_check_large_blocks().
	*/
	if (drro->drr_blksz <=
	doi->doi_data_block_size)
	return (SET_ERROR(EINVAL));
	/*
	* We keep the existing blocksize and
	* contents.
	*/
	*new_blksz =
	doi->doi_data_block_size;
	} else {
	do_free_range = B_TRUE;
	}
	}
	}

	/* nblkptr can only decrease if the object was reallocated */
	if (nblkptr < doi->doi_nblkptr)
	do_free_range = B_TRUE;

	/* number of slots can only change on reallocation */
	if (dn_slots != doi->doi_dnodesize >> DNODE_SHIFT)
	do_free_range = B_TRUE;

	/*
	* For raw sends we also check a few other fields to
	* ensure we are preserving the objset structure exactly
	* as it was on the receive side:
	* - A changed indirect block size
	* - A smaller nlevels
	*/
	if (rwa->raw) {
	if (indblksz != doi->doi_metadata_block_size)
	do_free_range = B_TRUE;
	if (drro->drr_nlevels < doi->doi_indirection)
	do_free_range = B_TRUE;
	}

	if (do_free_range) {
	err = dmu_free_long_range(rwa->os, drro->drr_object,
	0, DMU_OBJECT_END);
	if (err != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* The dmu does not currently support decreasing nlevels
	* or changing the number of dnode slots on an object. For
	* non-raw sends, this does not matter and the new object
	* can just use the previous one's nlevels. For raw sends,
	* however, the structure of the received dnode (including
	* nlevels and dnode slots) must match that of the send
	* side. Therefore, instead of using dmu_object_reclaim(),
	* we must free the object completely and call
	* dmu_object_claim_dnsize() instead.
	*/
	if ((rwa->raw && drro->drr_nlevels < doi->doi_indirection) \|\|
	dn_slots != doi->doi_dnodesize >> DNODE_SHIFT) {
	err = dmu_free_long_object(rwa->os, drro->drr_object);
	if (err != 0)
	return (SET_ERROR(EINVAL));

	txg_wait_synced(dmu_objset_pool(rwa->os), 0);
	*object_to_hold = DMU_NEW_OBJECT;
	}

	/*
	* For raw receives, free everything beyond the new incoming
	* maxblkid. Normally this would be done with a DRR_FREE
	* record that would come after this DRR_OBJECT record is
	* processed. However, for raw receives we manually set the
	* maxblkid from the drr_maxblkid and so we must first free
	* everything above that blkid to ensure the DMU is always
	* consistent with itself. We will never free the first block
	* of the object here because a maxblkid of 0 could indicate
	* an object with a single block or one with no blocks. This
	* free may be skipped when dmu_free_long_range() was called
	* above since it covers the entire object's contents.
	*/
	if (rwa->raw && *object_to_hold != DMU_NEW_OBJECT && !do_free_range) {
	err = dmu_free_long_range(rwa->os, drro->drr_object,
	(drro->drr_maxblkid + 1) * doi->doi_data_block_size,
	DMU_OBJECT_END);
	if (err != 0)
	return (SET_ERROR(EINVAL));
	}
	return (0);
	}

	noinline static int
	receive_object(struct receive_writer_arg rwa, struct drr_object drro,
	void *data)
	{
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	int err;
	uint32_t new_blksz = drro->drr_blksz;
	uint8_t dn_slots = drro->drr_dn_slots != 0 ?
	drro->drr_dn_slots : DNODE_MIN_SLOTS;

	if (drro->drr_type == DMU_OT_NONE \|\|
	!DMU_OT_IS_VALID(drro->drr_type) \|\|
	!DMU_OT_IS_VALID(drro->drr_bonustype) \|\|
	drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS \|\|
	drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS \|\|
	P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) \|\|
	drro->drr_blksz < SPA_MINBLOCKSIZE \|\|
	drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) \|\|
	drro->drr_bonuslen >
	DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(rwa->os))) \|\|
	dn_slots >
	(spa_maxdnodesize(dmu_objset_spa(rwa->os)) >> DNODE_SHIFT)) {
	return (SET_ERROR(EINVAL));
	}

	if (rwa->raw) {
	/*
	* We should have received a DRR_OBJECT_RANGE record
	* containing this block and stored it in rwa.
	*/
	if (drro->drr_object < rwa->or_firstobj \|\|
	drro->drr_object >= rwa->or_firstobj + rwa->or_numslots \|\|
	drro->drr_raw_bonuslen < drro->drr_bonuslen \|\|
	drro->drr_indblkshift > SPA_MAXBLOCKSHIFT \|\|
	drro->drr_nlevels > DN_MAX_LEVELS \|\|
	drro->drr_nblkptr > DN_MAX_NBLKPTR \|\|
	DN_SLOTS_TO_BONUSLEN(dn_slots) <
	drro->drr_raw_bonuslen)
	return (SET_ERROR(EINVAL));
	} else {
	/*
	* The DRR_OBJECT_SPILL flag is valid when the DRR_BEGIN
	* record indicates this by setting DRR_FLAG_SPILL_BLOCK.
	*/
	if (((drro->drr_flags & ~(DRR_OBJECT_SPILL))) \|\|
	(!rwa->spill && DRR_OBJECT_HAS_SPILL(drro->drr_flags))) {
	return (SET_ERROR(EINVAL));
	}

	if (drro->drr_raw_bonuslen != 0 \|\| drro->drr_nblkptr != 0 \|\|
	drro->drr_indblkshift != 0 \|\| drro->drr_nlevels != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	err = dmu_object_info(rwa->os, drro->drr_object, &doi);

	if (err != 0 && err != ENOENT && err != EEXIST)
	return (SET_ERROR(EINVAL));

	if (drro->drr_object > rwa->max_object)
	rwa->max_object = drro->drr_object;

	/*
	* If we are losing blkptrs or changing the block size this must
	* be a new file instance. We must clear out the previous file
	* contents before we can change this type of metadata in the dnode.
	* Raw receives will also check that the indirect structure of the
	* dnode hasn't changed.
	*/
	uint64_t object_to_hold;
	if (err == 0) {
	err = receive_handle_existing_object(rwa, drro, &doi, data,
	&object_to_hold, &new_blksz);
	+ if (err != 0)
	+ return (err);
	} else if (err == EEXIST) {
	/*
	* The object requested is currently an interior slot of a
	* multi-slot dnode. This will be resolved when the next txg
	* is synced out, since the send stream will have told us
	* to free this slot when we freed the associated dnode
	* earlier in the stream.
	*/
	txg_wait_synced(dmu_objset_pool(rwa->os), 0);

	if (dmu_object_info(rwa->os, drro->drr_object, NULL) != ENOENT)
	return (SET_ERROR(EINVAL));

	/* object was freed and we are about to allocate a new one */
	object_to_hold = DMU_NEW_OBJECT;
	} else {
	/* object is free and we are about to allocate a new one */
	object_to_hold = DMU_NEW_OBJECT;
	}

	/*
	* If this is a multi-slot dnode there is a chance that this
	* object will expand into a slot that is already used by
	* another object from the previous snapshot. We must free
	* these objects before we attempt to allocate the new dnode.
	*/
	if (dn_slots > 1) {
	boolean_t need_sync = B_FALSE;

	for (uint64_t slot = drro->drr_object + 1;
	slot < drro->drr_object + dn_slots;
	slot++) {
	dmu_object_info_t slot_doi;

	err = dmu_object_info(rwa->os, slot, &slot_doi);
	if (err == ENOENT \|\| err == EEXIST)
	continue;
	else if (err != 0)
	return (err);

	err = dmu_free_long_object(rwa->os, slot);
	if (err != 0)
	return (err);

	need_sync = B_TRUE;
	}

	if (need_sync)
	txg_wait_synced(dmu_objset_pool(rwa->os), 0);
	}

	tx = dmu_tx_create(rwa->os);
	dmu_tx_hold_bonus(tx, object_to_hold);
	dmu_tx_hold_write(tx, object_to_hold, 0, 0);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	return (err);
	}

	if (object_to_hold == DMU_NEW_OBJECT) {
	/* Currently free, wants to be allocated */
	err = dmu_object_claim_dnsize(rwa->os, drro->drr_object,
	drro->drr_type, new_blksz,
	drro->drr_bonustype, drro->drr_bonuslen,
	dn_slots << DNODE_SHIFT, tx);
	} else if (drro->drr_type != doi.doi_type \|\|
	new_blksz != doi.doi_data_block_size \|\|
	drro->drr_bonustype != doi.doi_bonus_type \|\|
	drro->drr_bonuslen != doi.doi_bonus_size) {
	/* Currently allocated, but with different properties */
	err = dmu_object_reclaim_dnsize(rwa->os, drro->drr_object,
	drro->drr_type, new_blksz,
	drro->drr_bonustype, drro->drr_bonuslen,
	dn_slots << DNODE_SHIFT, rwa->spill ?
	DRR_OBJECT_HAS_SPILL(drro->drr_flags) : B_FALSE, tx);
	} else if (rwa->spill && !DRR_OBJECT_HAS_SPILL(drro->drr_flags)) {
	/*
	* Currently allocated, the existing version of this object
	* may reference a spill block that is no longer allocated
	* at the source and needs to be freed.
	*/
	err = dmu_object_rm_spill(rwa->os, drro->drr_object, tx);
	}

	if (err != 0) {
	dmu_tx_commit(tx);
	return (SET_ERROR(EINVAL));
	}

	if (rwa->or_crypt_params_present) {
	/*
	* Set the crypt params for the buffer associated with this
	* range of dnodes. This causes the blkptr_t to have the
	* same crypt params (byteorder, salt, iv, mac) as on the
	* sending side.
	*
	* Since we are committing this tx now, it is possible for
	* the dnode block to end up on-disk with the incorrect MAC,
	* if subsequent objects in this block are received in a
	* different txg. However, since the dataset is marked as
	* inconsistent, no code paths will do a non-raw read (or
	* decrypt the block / verify the MAC). The receive code and
	* scrub code can safely do raw reads and verify the
	* checksum. They don't need to verify the MAC.
	*/
	dmu_buf_t *db = NULL;
	uint64_t offset = rwa->or_firstobj * DNODE_MIN_SIZE;

	err = dmu_buf_hold_by_dnode(DMU_META_DNODE(rwa->os),
	offset, FTAG, &db, DMU_READ_PREFETCH \| DMU_READ_NO_DECRYPT);
	if (err != 0) {
	dmu_tx_commit(tx);
	return (SET_ERROR(EINVAL));
	}

	dmu_buf_set_crypt_params(db, rwa->or_byteorder,
	rwa->or_salt, rwa->or_iv, rwa->or_mac, tx);

	dmu_buf_rele(db, FTAG);

	rwa->or_crypt_params_present = B_FALSE;
	}

	dmu_object_set_checksum(rwa->os, drro->drr_object,
	drro->drr_checksumtype, tx);
	dmu_object_set_compress(rwa->os, drro->drr_object,
	drro->drr_compress, tx);

	/* handle more restrictive dnode structuring for raw recvs */
	if (rwa->raw) {
	/*
	* Set the indirect block size, block shift, nlevels.
	* This will not fail because we ensured all of the
	* blocks were freed earlier if this is a new object.
	* For non-new objects block size and indirect block
	* shift cannot change and nlevels can only increase.
	*/
	ASSERT3U(new_blksz, ==, drro->drr_blksz);
	VERIFY0(dmu_object_set_blocksize(rwa->os, drro->drr_object,
	drro->drr_blksz, drro->drr_indblkshift, tx));
	VERIFY0(dmu_object_set_nlevels(rwa->os, drro->drr_object,
	drro->drr_nlevels, tx));

	/*
	* Set the maxblkid. This will always succeed because
	* we freed all blocks beyond the new maxblkid above.
	*/
	VERIFY0(dmu_object_set_maxblkid(rwa->os, drro->drr_object,
	drro->drr_maxblkid, tx));
	}

	if (data != NULL) {
	dmu_buf_t *db;
	dnode_t *dn;
	uint32_t flags = DMU_READ_NO_PREFETCH;

	if (rwa->raw)
	flags \|= DMU_READ_NO_DECRYPT;

	VERIFY0(dnode_hold(rwa->os, drro->drr_object, FTAG, &dn));
	VERIFY0(dmu_bonus_hold_by_dnode(dn, FTAG, &db, flags));

	dmu_buf_will_dirty(db, tx);

	ASSERT3U(db->db_size, >=, drro->drr_bonuslen);
	memcpy(db->db_data, data, DRR_OBJECT_PAYLOAD_SIZE(drro));

	/*
	* Raw bonus buffers have their byteorder determined by the
	* DRR_OBJECT_RANGE record.
	*/
	if (rwa->byteswap && !rwa->raw) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drro->drr_bonustype);
	dmu_ot_byteswap[byteswap].ob_func(db->db_data,
	DRR_OBJECT_PAYLOAD_SIZE(drro));
	}
	dmu_buf_rele(db, FTAG);
	dnode_rele(dn, FTAG);
	}
	dmu_tx_commit(tx);

	return (0);
	}

	noinline static int
	receive_freeobjects(struct receive_writer_arg *rwa,
	struct drr_freeobjects *drrfo)
	{
	uint64_t obj;
	int next_err = 0;

	if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj)
	return (SET_ERROR(EINVAL));

	for (obj = drrfo->drr_firstobj == 0 ? 1 : drrfo->drr_firstobj;
	obj < drrfo->drr_firstobj + drrfo->drr_numobjs &&
	obj < DN_MAX_OBJECT && next_err == 0;
	next_err = dmu_object_next(rwa->os, &obj, FALSE, 0)) {
	dmu_object_info_t doi;
	int err;

	err = dmu_object_info(rwa->os, obj, &doi);
	if (err == ENOENT)
	continue;
	else if (err != 0)
	return (err);

	err = dmu_free_long_object(rwa->os, obj);

	if (err != 0)
	return (err);
	}
	if (next_err != ESRCH)
	return (next_err);
	return (0);
	}

	/*
	* Note: if this fails, the caller will clean up any records left on the
	* rwa->write_batch list.
	*/
	static int
	flush_write_batch_impl(struct receive_writer_arg *rwa)
	{
	dnode_t *dn;
	int err;

	if (dnode_hold(rwa->os, rwa->last_object, FTAG, &dn) != 0)
	return (SET_ERROR(EINVAL));

	struct receive_record_arg *last_rrd = list_tail(&rwa->write_batch);
	struct drr_write *last_drrw = &last_rrd->header.drr_u.drr_write;

	struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
	struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;

	ASSERT3U(rwa->last_object, ==, last_drrw->drr_object);
	ASSERT3U(rwa->last_offset, ==, last_drrw->drr_offset);

	dmu_tx_t *tx = dmu_tx_create(rwa->os);
	dmu_tx_hold_write_by_dnode(tx, dn, first_drrw->drr_offset,
	last_drrw->drr_offset - first_drrw->drr_offset +
	last_drrw->drr_logical_size);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	struct receive_record_arg *rrd;
	while ((rrd = list_head(&rwa->write_batch)) != NULL) {
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;
	abd_t *abd = rrd->abd;

	ASSERT3U(drrw->drr_object, ==, rwa->last_object);

	if (drrw->drr_logical_size != dn->dn_datablksz) {
	/*
	* The WRITE record is larger than the object's block
	* size. We must be receiving an incremental
	* large-block stream into a dataset that previously did
	* a non-large-block receive. Lightweight writes must
	* be exactly one block, so we need to decompress the
	* data (if compressed) and do a normal dmu_write().
	*/
	ASSERT3U(drrw->drr_logical_size, >, dn->dn_datablksz);
	if (DRR_WRITE_COMPRESSED(drrw)) {
	abd_t *decomp_abd =
	abd_alloc_linear(drrw->drr_logical_size,
	B_FALSE);

	err = zio_decompress_data(
	drrw->drr_compressiontype,
	abd, abd_to_buf(decomp_abd),
	abd_get_size(abd),
	abd_get_size(decomp_abd), NULL);

	if (err == 0) {
	dmu_write_by_dnode(dn,
	drrw->drr_offset,
	drrw->drr_logical_size,
	abd_to_buf(decomp_abd), tx);
	}
	abd_free(decomp_abd);
	} else {
	dmu_write_by_dnode(dn,
	drrw->drr_offset,
	drrw->drr_logical_size,
	abd_to_buf(abd), tx);
	}
	if (err == 0)
	abd_free(abd);
	} else {
	zio_prop_t zp;
	dmu_write_policy(rwa->os, dn, 0, 0, &zp);

	enum zio_flag zio_flags = 0;

	if (rwa->raw) {
	zp.zp_encrypt = B_TRUE;
	zp.zp_compress = drrw->drr_compressiontype;
	zp.zp_byteorder = ZFS_HOST_BYTEORDER ^
	!!DRR_IS_RAW_BYTESWAPPED(drrw->drr_flags) ^
	rwa->byteswap;
	memcpy(zp.zp_salt, drrw->drr_salt,
	ZIO_DATA_SALT_LEN);
	memcpy(zp.zp_iv, drrw->drr_iv,
	ZIO_DATA_IV_LEN);
	memcpy(zp.zp_mac, drrw->drr_mac,
	ZIO_DATA_MAC_LEN);
	if (DMU_OT_IS_ENCRYPTED(zp.zp_type)) {
	zp.zp_nopwrite = B_FALSE;
	zp.zp_copies = MIN(zp.zp_copies,
	SPA_DVAS_PER_BP - 1);
	}
	zio_flags \|= ZIO_FLAG_RAW;
	} else if (DRR_WRITE_COMPRESSED(drrw)) {
	ASSERT3U(drrw->drr_compressed_size, >, 0);
	ASSERT3U(drrw->drr_logical_size, >=,
	drrw->drr_compressed_size);
	zp.zp_compress = drrw->drr_compressiontype;
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	} else if (rwa->byteswap) {
	/*
	* Note: compressed blocks never need to be
	* byteswapped, because WRITE records for
	* metadata blocks are never compressed. The
	* exception is raw streams, which are written
	* in the original byteorder, and the byteorder
	* bit is preserved in the BP by setting
	* zp_byteorder above.
	*/
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrw->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(
	abd_to_buf(abd),
	DRR_WRITE_PAYLOAD_SIZE(drrw));
	}

	/*
	* Since this data can't be read until the receive
	* completes, we can do a "lightweight" write for
	* improved performance.
	*/
	err = dmu_lightweight_write_by_dnode(dn,
	drrw->drr_offset, abd, &zp, zio_flags, tx);
	}

	if (err != 0) {
	/*
	* This rrd is left on the list, so the caller will
	* free it (and the abd).
	*/
	break;
	}

	/*
	* Note: If the receive fails, we want the resume stream to
	* start with the same record that we last successfully
	* received (as opposed to the next record), so that we can
	* verify that we are resuming from the correct location.
	*/
	save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx);

	list_remove(&rwa->write_batch, rrd);
	kmem_free(rrd, sizeof (*rrd));
	}

	dmu_tx_commit(tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	noinline static int
	flush_write_batch(struct receive_writer_arg *rwa)
	{
	if (list_is_empty(&rwa->write_batch))
	return (0);
	int err = rwa->err;
	if (err == 0)
	err = flush_write_batch_impl(rwa);
	if (err != 0) {
	struct receive_record_arg *rrd;
	while ((rrd = list_remove_head(&rwa->write_batch)) != NULL) {
	abd_free(rrd->abd);
	kmem_free(rrd, sizeof (*rrd));
	}
	}
	ASSERT(list_is_empty(&rwa->write_batch));
	return (err);
	}

	noinline static int
	receive_process_write_record(struct receive_writer_arg *rwa,
	struct receive_record_arg *rrd)
	{
	int err = 0;

	ASSERT3U(rrd->header.drr_type, ==, DRR_WRITE);
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;

	if (drrw->drr_offset + drrw->drr_logical_size < drrw->drr_offset \|\|
	!DMU_OT_IS_VALID(drrw->drr_type))
	return (SET_ERROR(EINVAL));

	if (rwa->heal) {
	blkptr_t *bp;
	dmu_buf_t *dbp;
	dnode_t *dn;
	int flags = DB_RF_CANFAIL;

	if (rwa->raw)
	flags \|= DB_RF_NO_DECRYPT;

	if (rwa->byteswap) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrw->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(rrd->abd),
	DRR_WRITE_PAYLOAD_SIZE(drrw));
	}

	err = dmu_buf_hold_noread(rwa->os, drrw->drr_object,
	drrw->drr_offset, FTAG, &dbp);
	if (err != 0)
	return (err);

	/* Try to read the object to see if it needs healing */
	err = dbuf_read((dmu_buf_impl_t *)dbp, NULL, flags);
	/*
	* We only try to heal when dbuf_read() returns a ECKSUMs.
	* Other errors (even EIO) get returned to caller.
	* EIO indicates that the device is not present/accessible,
	* so writing to it will likely fail.
	* If the block is healthy, we don't want to overwrite it
	* unnecessarily.
	*/
	if (err != ECKSUM) {
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}
	dn = dmu_buf_dnode_enter(dbp);
	/* Make sure the on-disk block and recv record sizes match */
	if (drrw->drr_logical_size !=
	dn->dn_datablkszsec << SPA_MINBLOCKSHIFT) {
	err = ENOTSUP;
	dmu_buf_dnode_exit(dbp);
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}
	/* Get the block pointer for the corrupted block */
	bp = dmu_buf_get_blkptr(dbp);
	err = do_corrective_recv(rwa, drrw, rrd, bp);
	dmu_buf_dnode_exit(dbp);
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}

	/*
	* For resuming to work, records must be in increasing order
	* by (object, offset).
	*/
	if (drrw->drr_object < rwa->last_object \|\|
	(drrw->drr_object == rwa->last_object &&
	drrw->drr_offset < rwa->last_offset)) {
	return (SET_ERROR(EINVAL));
	}

	struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
	struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
	uint64_t batch_size =
	MIN(zfs_recv_write_batch_size, DMU_MAX_ACCESS / 2);
	if (first_rrd != NULL &&
	(drrw->drr_object != first_drrw->drr_object \|\|
	drrw->drr_offset >= first_drrw->drr_offset + batch_size)) {
	err = flush_write_batch(rwa);
	if (err != 0)
	return (err);
	}

	rwa->last_object = drrw->drr_object;
	rwa->last_offset = drrw->drr_offset;

	if (rwa->last_object > rwa->max_object)
	rwa->max_object = rwa->last_object;

	list_insert_tail(&rwa->write_batch, rrd);
	/*
	* Return EAGAIN to indicate that we will use this rrd again,
	* so the caller should not free it
	*/
	return (EAGAIN);
	}

	static int
	receive_write_embedded(struct receive_writer_arg *rwa,
	struct drr_write_embedded drrwe, void data)
	{
	dmu_tx_t *tx;
	int err;

	if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_psize > BPE_PAYLOAD_SIZE)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES)
	return (SET_ERROR(EINVAL));
	if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
	return (SET_ERROR(EINVAL));
	if (rwa->raw)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_object > rwa->max_object)
	rwa->max_object = drrwe->drr_object;

	tx = dmu_tx_create(rwa->os);

	dmu_tx_hold_write(tx, drrwe->drr_object,
	drrwe->drr_offset, drrwe->drr_length);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	return (err);
	}

	dmu_write_embedded(rwa->os, drrwe->drr_object,
	drrwe->drr_offset, data, drrwe->drr_etype,
	drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize,
	rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx);

	/* See comment in restore_write. */
	save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx);
	dmu_tx_commit(tx);
	return (0);
	}

	static int
	receive_spill(struct receive_writer_arg rwa, struct drr_spill drrs,
	abd_t *abd)
	{
	dmu_buf_t db, db_spill;
	int err;

	if (drrs->drr_length < SPA_MINBLOCKSIZE \|\|
	drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os)))
	return (SET_ERROR(EINVAL));

	/*
	* This is an unmodified spill block which was added to the stream
	* to resolve an issue with incorrectly removing spill blocks. It
	* should be ignored by current versions of the code which support
	* the DRR_FLAG_SPILL_BLOCK flag.
	*/
	if (rwa->spill && DRR_SPILL_IS_UNMODIFIED(drrs->drr_flags)) {
	abd_free(abd);
	return (0);
	}

	if (rwa->raw) {
	if (!DMU_OT_IS_VALID(drrs->drr_type) \|\|
	drrs->drr_compressiontype >= ZIO_COMPRESS_FUNCTIONS \|\|
	drrs->drr_compressed_size == 0)
	return (SET_ERROR(EINVAL));
	}

	if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0)
	return (SET_ERROR(EINVAL));

	if (drrs->drr_object > rwa->max_object)
	rwa->max_object = drrs->drr_object;

	VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db));
	if ((err = dmu_spill_hold_by_bonus(db, DMU_READ_NO_DECRYPT, FTAG,
	&db_spill)) != 0) {
	dmu_buf_rele(db, FTAG);
	return (err);
	}

	dmu_tx_t *tx = dmu_tx_create(rwa->os);

	dmu_tx_hold_spill(tx, db->db_object);

	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_buf_rele(db, FTAG);
	dmu_buf_rele(db_spill, FTAG);
	dmu_tx_abort(tx);
	return (err);
	}

	/*
	* Spill blocks may both grow and shrink. When a change in size
	* occurs any existing dbuf must be updated to match the logical
	* size of the provided arc_buf_t.
	*/
	if (db_spill->db_size != drrs->drr_length) {
	dmu_buf_will_fill(db_spill, tx);
	VERIFY0(dbuf_spill_set_blksz(db_spill,
	drrs->drr_length, tx));
	}

	arc_buf_t *abuf;
	if (rwa->raw) {
	boolean_t byteorder = ZFS_HOST_BYTEORDER ^
	!!DRR_IS_RAW_BYTESWAPPED(drrs->drr_flags) ^
	rwa->byteswap;

	abuf = arc_loan_raw_buf(dmu_objset_spa(rwa->os),
	drrs->drr_object, byteorder, drrs->drr_salt,
	drrs->drr_iv, drrs->drr_mac, drrs->drr_type,
	drrs->drr_compressed_size, drrs->drr_length,
	drrs->drr_compressiontype, 0);
	} else {
	abuf = arc_loan_buf(dmu_objset_spa(rwa->os),
	DMU_OT_IS_METADATA(drrs->drr_type),
	drrs->drr_length);
	if (rwa->byteswap) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrs->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(abd),
	DRR_SPILL_PAYLOAD_SIZE(drrs));
	}
	}

	memcpy(abuf->b_data, abd_to_buf(abd), DRR_SPILL_PAYLOAD_SIZE(drrs));
	abd_free(abd);
	dbuf_assign_arcbuf((dmu_buf_impl_t *)db_spill, abuf, tx);

	dmu_buf_rele(db, FTAG);
	dmu_buf_rele(db_spill, FTAG);

	dmu_tx_commit(tx);
	return (0);
	}

	noinline static int
	receive_free(struct receive_writer_arg rwa, struct drr_free drrf)
	{
	int err;

	if (drrf->drr_length != -1ULL &&
	drrf->drr_offset + drrf->drr_length < drrf->drr_offset)
	return (SET_ERROR(EINVAL));

	if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0)
	return (SET_ERROR(EINVAL));

	if (drrf->drr_object > rwa->max_object)
	rwa->max_object = drrf->drr_object;

	err = dmu_free_long_range(rwa->os, drrf->drr_object,
	drrf->drr_offset, drrf->drr_length);

	return (err);
	}

	static int
	receive_object_range(struct receive_writer_arg *rwa,
	struct drr_object_range *drror)
	{
	/*
	* By default, we assume this block is in our native format
	* (ZFS_HOST_BYTEORDER). We then take into account whether
	* the send stream is byteswapped (rwa->byteswap). Finally,
	* we need to byteswap again if this particular block was
	* in non-native format on the send side.
	*/
	boolean_t byteorder = ZFS_HOST_BYTEORDER ^ rwa->byteswap ^
	!!DRR_IS_RAW_BYTESWAPPED(drror->drr_flags);

	/*
	* Since dnode block sizes are constant, we should not need to worry
	* about making sure that the dnode block size is the same on the
	* sending and receiving sides for the time being. For non-raw sends,
	* this does not matter (and in fact we do not send a DRR_OBJECT_RANGE
	* record at all). Raw sends require this record type because the
	* encryption parameters are used to protect an entire block of bonus
	* buffers. If the size of dnode blocks ever becomes variable,
	* handling will need to be added to ensure that dnode block sizes
	* match on the sending and receiving side.
	*/
	if (drror->drr_numslots != DNODES_PER_BLOCK \|\|
	P2PHASE(drror->drr_firstobj, DNODES_PER_BLOCK) != 0 \|\|
	!rwa->raw)
	return (SET_ERROR(EINVAL));

	if (drror->drr_firstobj > rwa->max_object)
	rwa->max_object = drror->drr_firstobj;

	/*
	* The DRR_OBJECT_RANGE handling must be deferred to receive_object()
	* so that the block of dnodes is not written out when it's empty,
	* and converted to a HOLE BP.
	*/
	rwa->or_crypt_params_present = B_TRUE;
	rwa->or_firstobj = drror->drr_firstobj;
	rwa->or_numslots = drror->drr_numslots;
	memcpy(rwa->or_salt, drror->drr_salt, ZIO_DATA_SALT_LEN);
	memcpy(rwa->or_iv, drror->drr_iv, ZIO_DATA_IV_LEN);
	memcpy(rwa->or_mac, drror->drr_mac, ZIO_DATA_MAC_LEN);
	rwa->or_byteorder = byteorder;

	return (0);
	}

	/*
	* Until we have the ability to redact large ranges of data efficiently, we
	* process these records as frees.
	*/
	noinline static int
	receive_redact(struct receive_writer_arg rwa, struct drr_redact drrr)
	{
	struct drr_free drrf = {0};
	drrf.drr_length = drrr->drr_length;
	drrf.drr_object = drrr->drr_object;
	drrf.drr_offset = drrr->drr_offset;
	drrf.drr_toguid = drrr->drr_toguid;
	return (receive_free(rwa, &drrf));
	}

	/* used to destroy the drc_ds on error */
	static void
	dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc)
	{
	dsl_dataset_t *ds = drc->drc_ds;
	ds_hold_flags_t dsflags;

	dsflags = (drc->drc_raw) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	/*
	* Wait for the txg sync before cleaning up the receive. For
	* resumable receives, this ensures that our resume state has
	* been written out to disk. For raw receives, this ensures
	* that the user accounting code will not attempt to do anything
	* after we stopped receiving the dataset.
	*/
	txg_wait_synced(ds->ds_dir->dd_pool, 0);
	ds->ds_objset->os_raw_receive = B_FALSE;

	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	if (drc->drc_resumable && drc->drc_should_save &&
	!BP_IS_HOLE(dsl_dataset_get_blkptr(ds))) {
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
	} else {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	dsl_dataset_name(ds, name);
	dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
	if (!drc->drc_heal)
	(void) dsl_destroy_head(name);
	}
	}

	static void
	receive_cksum(dmu_recv_cookie_t drc, int len, void buf)
	{
	if (drc->drc_byteswap) {
	(void) fletcher_4_incremental_byteswap(buf, len,
	&drc->drc_cksum);
	} else {
	(void) fletcher_4_incremental_native(buf, len, &drc->drc_cksum);
	}
	}

	/*
	* Read the payload into a buffer of size len, and update the current record's
	* payload field.
	* Allocate drc->drc_next_rrd and read the next record's header into
	* drc->drc_next_rrd->header.
	* Verify checksum of payload and next record.
	*/
	static int
	receive_read_payload_and_next_header(dmu_recv_cookie_t drc, int len, void buf)
	{
	int err;

	if (len != 0) {
	ASSERT3U(len, <=, SPA_MAXBLOCKSIZE);
	err = receive_read(drc, len, buf);
	if (err != 0)
	return (err);
	receive_cksum(drc, len, buf);

	/* note: rrd is NULL when reading the begin record's payload */
	if (drc->drc_rrd != NULL) {
	drc->drc_rrd->payload = buf;
	drc->drc_rrd->payload_size = len;
	drc->drc_rrd->bytes_read = drc->drc_bytes_read;
	}
	} else {
	ASSERT3P(buf, ==, NULL);
	}

	drc->drc_prev_cksum = drc->drc_cksum;

	drc->drc_next_rrd = kmem_zalloc(sizeof (*drc->drc_next_rrd), KM_SLEEP);
	err = receive_read(drc, sizeof (drc->drc_next_rrd->header),
	&drc->drc_next_rrd->header);
	drc->drc_next_rrd->bytes_read = drc->drc_bytes_read;

	if (err != 0) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (err);
	}
	if (drc->drc_next_rrd->header.drr_type == DRR_BEGIN) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (SET_ERROR(EINVAL));
	}

	/*
	* Note: checksum is of everything up to but not including the
	* checksum itself.
	*/
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	receive_cksum(drc,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	&drc->drc_next_rrd->header);

	zio_cksum_t cksum_orig =
	drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
	zio_cksum_t *cksump =
	&drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;

	if (drc->drc_byteswap)
	byteswap_record(&drc->drc_next_rrd->header);

	if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) &&
	!ZIO_CHECKSUM_EQUAL(drc->drc_cksum, *cksump)) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (SET_ERROR(ECKSUM));
	}

	receive_cksum(drc, sizeof (cksum_orig), &cksum_orig);

	return (0);
	}

	/*
	* Issue the prefetch reads for any necessary indirect blocks.
	*
	* We use the object ignore list to tell us whether or not to issue prefetches
	* for a given object. We do this for both correctness (in case the blocksize
	* of an object has changed) and performance (if the object doesn't exist, don't
	* needlessly try to issue prefetches). We also trim the list as we go through
	* the stream to prevent it from growing to an unbounded size.
	*
	* The object numbers within will always be in sorted order, and any write
	* records we see will also be in sorted order, but they're not sorted with
	* respect to each other (i.e. we can get several object records before
	* receiving each object's write records). As a result, once we've reached a
	* given object number, we can safely remove any reference to lower object
	* numbers in the ignore list. In practice, we receive up to 32 object records
	* before receiving write records, so the list can have up to 32 nodes in it.
	*/
	static void
	receive_read_prefetch(dmu_recv_cookie_t *drc, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	if (!objlist_exists(drc->drc_ignore_objlist, object)) {
	dmu_prefetch(drc->drc_os, object, 1, offset, length,
	ZIO_PRIORITY_SYNC_READ);
	}
	}

	/*
	* Read records off the stream, issuing any necessary prefetches.
	*/
	static int
	receive_read_record(dmu_recv_cookie_t *drc)
	{
	int err;

	switch (drc->drc_rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro =
	&drc->drc_rrd->header.drr_u.drr_object;
	uint32_t size = DRR_OBJECT_PAYLOAD_SIZE(drro);
	void *buf = NULL;
	dmu_object_info_t doi;

	if (size != 0)
	buf = kmem_zalloc(size, KM_SLEEP);

	err = receive_read_payload_and_next_header(drc, size, buf);
	if (err != 0) {
	kmem_free(buf, size);
	return (err);
	}
	err = dmu_object_info(drc->drc_os, drro->drr_object, &doi);
	/*
	* See receive_read_prefetch for an explanation why we're
	* storing this object in the ignore_obj_list.
	*/
	if (err == ENOENT \|\| err == EEXIST \|\|
	(err == 0 && doi.doi_data_block_size != drro->drr_blksz)) {
	objlist_insert(drc->drc_ignore_objlist,
	drro->drr_object);
	err = 0;
	}
	return (err);
	}
	case DRR_FREEOBJECTS:
	{
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);
	}
	case DRR_WRITE:
	{
	struct drr_write *drrw = &drc->drc_rrd->header.drr_u.drr_write;
	int size = DRR_WRITE_PAYLOAD_SIZE(drrw);
	abd_t *abd = abd_alloc_linear(size, B_FALSE);
	err = receive_read_payload_and_next_header(drc, size,
	abd_to_buf(abd));
	if (err != 0) {
	abd_free(abd);
	return (err);
	}
	drc->drc_rrd->abd = abd;
	receive_read_prefetch(drc, drrw->drr_object, drrw->drr_offset,
	drrw->drr_logical_size);
	return (err);
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&drc->drc_rrd->header.drr_u.drr_write_embedded;
	uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8);
	void *buf = kmem_zalloc(size, KM_SLEEP);

	err = receive_read_payload_and_next_header(drc, size, buf);
	if (err != 0) {
	kmem_free(buf, size);
	return (err);
	}

	receive_read_prefetch(drc, drrwe->drr_object, drrwe->drr_offset,
	drrwe->drr_length);
	return (err);
	}
	case DRR_FREE:
	case DRR_REDACT:
	{
	/*
	* It might be beneficial to prefetch indirect blocks here, but
	* we don't really have the data to decide for sure.
	*/
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);
	}
	case DRR_END:
	{
	struct drr_end *drre = &drc->drc_rrd->header.drr_u.drr_end;
	if (!ZIO_CHECKSUM_EQUAL(drc->drc_prev_cksum,
	drre->drr_checksum))
	return (SET_ERROR(ECKSUM));
	return (0);
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &drc->drc_rrd->header.drr_u.drr_spill;
	int size = DRR_SPILL_PAYLOAD_SIZE(drrs);
	abd_t *abd = abd_alloc_linear(size, B_FALSE);
	err = receive_read_payload_and_next_header(drc, size,
	abd_to_buf(abd));
	if (err != 0)
	abd_free(abd);
	else
	drc->drc_rrd->abd = abd;
	return (err);
	}
	case DRR_OBJECT_RANGE:
	{
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);

	}
	default:
	return (SET_ERROR(EINVAL));
	}
	}



	static void
	dprintf_drr(struct receive_record_arg *rrd, int err)
	{
	#ifdef ZFS_DEBUG
	switch (rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro = &rrd->header.drr_u.drr_object;
	dprintf("drr_type = OBJECT obj = %llu type = %u "
	"bonustype = %u blksz = %u bonuslen = %u cksumtype = %u "
	"compress = %u dn_slots = %u err = %d\n",
	(u_longlong_t)drro->drr_object, drro->drr_type,
	drro->drr_bonustype, drro->drr_blksz, drro->drr_bonuslen,
	drro->drr_checksumtype, drro->drr_compress,
	drro->drr_dn_slots, err);
	break;
	}
	case DRR_FREEOBJECTS:
	{
	struct drr_freeobjects *drrfo =
	&rrd->header.drr_u.drr_freeobjects;
	dprintf("drr_type = FREEOBJECTS firstobj = %llu "
	"numobjs = %llu err = %d\n",
	(u_longlong_t)drrfo->drr_firstobj,
	(u_longlong_t)drrfo->drr_numobjs, err);
	break;
	}
	case DRR_WRITE:
	{
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;
	dprintf("drr_type = WRITE obj = %llu type = %u offset = %llu "
	"lsize = %llu cksumtype = %u flags = %u "
	"compress = %u psize = %llu err = %d\n",
	(u_longlong_t)drrw->drr_object, drrw->drr_type,
	(u_longlong_t)drrw->drr_offset,
	(u_longlong_t)drrw->drr_logical_size,
	drrw->drr_checksumtype, drrw->drr_flags,
	drrw->drr_compressiontype,
	(u_longlong_t)drrw->drr_compressed_size, err);
	break;
	}
	case DRR_WRITE_BYREF:
	{
	struct drr_write_byref *drrwbr =
	&rrd->header.drr_u.drr_write_byref;
	dprintf("drr_type = WRITE_BYREF obj = %llu offset = %llu "
	"length = %llu toguid = %llx refguid = %llx "
	"refobject = %llu refoffset = %llu cksumtype = %u "
	"flags = %u err = %d\n",
	(u_longlong_t)drrwbr->drr_object,
	(u_longlong_t)drrwbr->drr_offset,
	(u_longlong_t)drrwbr->drr_length,
	(u_longlong_t)drrwbr->drr_toguid,
	(u_longlong_t)drrwbr->drr_refguid,
	(u_longlong_t)drrwbr->drr_refobject,
	(u_longlong_t)drrwbr->drr_refoffset,
	drrwbr->drr_checksumtype, drrwbr->drr_flags, err);
	break;
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&rrd->header.drr_u.drr_write_embedded;
	dprintf("drr_type = WRITE_EMBEDDED obj = %llu offset = %llu "
	"length = %llu compress = %u etype = %u lsize = %u "
	"psize = %u err = %d\n",
	(u_longlong_t)drrwe->drr_object,
	(u_longlong_t)drrwe->drr_offset,
	(u_longlong_t)drrwe->drr_length,
	drrwe->drr_compression, drrwe->drr_etype,
	drrwe->drr_lsize, drrwe->drr_psize, err);
	break;
	}
	case DRR_FREE:
	{
	struct drr_free *drrf = &rrd->header.drr_u.drr_free;
	dprintf("drr_type = FREE obj = %llu offset = %llu "
	"length = %lld err = %d\n",
	(u_longlong_t)drrf->drr_object,
	(u_longlong_t)drrf->drr_offset,
	(longlong_t)drrf->drr_length,
	err);
	break;
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
	dprintf("drr_type = SPILL obj = %llu length = %llu "
	"err = %d\n", (u_longlong_t)drrs->drr_object,
	(u_longlong_t)drrs->drr_length, err);
	break;
	}
	case DRR_OBJECT_RANGE:
	{
	struct drr_object_range *drror =
	&rrd->header.drr_u.drr_object_range;
	dprintf("drr_type = OBJECT_RANGE firstobj = %llu "
	"numslots = %llu flags = %u err = %d\n",
	(u_longlong_t)drror->drr_firstobj,
	(u_longlong_t)drror->drr_numslots,
	drror->drr_flags, err);
	break;
	}
	default:
	return;
	}
	#endif
	}

	/*
	* Commit the records to the pool.
	*/
	static int
	receive_process_record(struct receive_writer_arg *rwa,
	struct receive_record_arg *rrd)
	{
	int err;

	/* Processing in order, therefore bytes_read should be increasing. */
	ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read);
	rwa->bytes_read = rrd->bytes_read;

	/* We can only heal write records; other ones get ignored */
	if (rwa->heal && rrd->header.drr_type != DRR_WRITE) {
	if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}
	return (0);
	}

	if (!rwa->heal && rrd->header.drr_type != DRR_WRITE) {
	err = flush_write_batch(rwa);
	if (err != 0) {
	if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}

	return (err);
	}
	}

	switch (rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro = &rrd->header.drr_u.drr_object;
	err = receive_object(rwa, drro, rrd->payload);
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	break;
	}
	case DRR_FREEOBJECTS:
	{
	struct drr_freeobjects *drrfo =
	&rrd->header.drr_u.drr_freeobjects;
	err = receive_freeobjects(rwa, drrfo);
	break;
	}
	case DRR_WRITE:
	{
	err = receive_process_write_record(rwa, rrd);
	if (rwa->heal) {
	/*
	* If healing - always free the abd after processing
	*/
	abd_free(rrd->abd);
	rrd->abd = NULL;
	} else if (err != EAGAIN) {
	/*
	* On success, a non-healing
	* receive_process_write_record() returns
	* EAGAIN to indicate that we do not want to free
	* the rrd or arc_buf.
	*/
	ASSERT(err != 0);
	abd_free(rrd->abd);
	rrd->abd = NULL;
	}
	break;
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&rrd->header.drr_u.drr_write_embedded;
	err = receive_write_embedded(rwa, drrwe, rrd->payload);
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	break;
	}
	case DRR_FREE:
	{
	struct drr_free *drrf = &rrd->header.drr_u.drr_free;
	err = receive_free(rwa, drrf);
	break;
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
	err = receive_spill(rwa, drrs, rrd->abd);
	if (err != 0)
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	break;
	}
	case DRR_OBJECT_RANGE:
	{
	struct drr_object_range *drror =
	&rrd->header.drr_u.drr_object_range;
	err = receive_object_range(rwa, drror);
	break;
	}
	case DRR_REDACT:
	{
	struct drr_redact *drrr = &rrd->header.drr_u.drr_redact;
	err = receive_redact(rwa, drrr);
	break;
	}
	default:
	err = (SET_ERROR(EINVAL));
	}

	if (err != 0)
	dprintf_drr(rrd, err);

	return (err);
	}

	/*
	* dmu_recv_stream's worker thread; pull records off the queue, and then call
	* receive_process_record When we're done, signal the main thread and exit.
	*/
	static __attribute__((noreturn)) void
	receive_writer_thread(void *arg)
	{
	struct receive_writer_arg *rwa = arg;
	struct receive_record_arg *rrd;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker;
	rrd = bqueue_dequeue(&rwa->q)) {
	/*
	* If there's an error, the main thread will stop putting things
	* on the queue, but we need to clear everything in it before we
	* can exit.
	*/
	int err = 0;
	if (rwa->err == 0) {
	err = receive_process_record(rwa, rrd);
	} else if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}
	/*
	* EAGAIN indicates that this record has been saved (on
	* raw->write_batch), and will be used again, so we don't
	* free it.
	* When healing data we always need to free the record.
	*/
	if (err != EAGAIN \|\| rwa->heal) {
	if (rwa->err == 0)
	rwa->err = err;
	kmem_free(rrd, sizeof (*rrd));
	}
	}
	kmem_free(rrd, sizeof (*rrd));

	if (rwa->heal) {
	zio_wait(rwa->heal_pio);
	} else {
	int err = flush_write_batch(rwa);
	if (rwa->err == 0)
	rwa->err = err;
	}
	mutex_enter(&rwa->mutex);
	rwa->done = B_TRUE;
	cv_signal(&rwa->cv);
	mutex_exit(&rwa->mutex);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	static int
	resume_check(dmu_recv_cookie_t drc, nvlist_t begin_nvl)
	{
	uint64_t val;
	objset_t *mos = dmu_objset_pool(drc->drc_os)->dp_meta_objset;
	uint64_t dsobj = dmu_objset_id(drc->drc_os);
	uint64_t resume_obj, resume_off;

	if (nvlist_lookup_uint64(begin_nvl,
	"resume_object", &resume_obj) != 0 \|\|
	nvlist_lookup_uint64(begin_nvl,
	"resume_offset", &resume_off) != 0) {
	return (SET_ERROR(EINVAL));
	}
	VERIFY0(zap_lookup(mos, dsobj,
	DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val));
	if (resume_obj != val)
	return (SET_ERROR(EINVAL));
	VERIFY0(zap_lookup(mos, dsobj,
	DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val));
	if (resume_off != val)
	return (SET_ERROR(EINVAL));

	return (0);
	}

	/*
	* Read in the stream's records, one by one, and apply them to the pool. There
	* are two threads involved; the thread that calls this function will spin up a
	* worker thread, read the records off the stream one by one, and issue
	* prefetches for any necessary indirect blocks. It will then push the records
	* onto an internal blocking queue. The worker thread will pull the records off
	* the queue, and actually write the data into the DMU. This way, the worker
	* thread doesn't have to wait for reads to complete, since everything it needs
	* (the indirect blocks) will be prefetched.
	*
	* NB: callers must call dmu_recv_end() if this succeeds.
	*/
	int
	dmu_recv_stream(dmu_recv_cookie_t drc, offset_t voffp)
	{
	int err = 0;
	struct receive_writer_arg rwa = kmem_zalloc(sizeof (rwa), KM_SLEEP);

	if (dsl_dataset_has_resume_receive_state(drc->drc_ds)) {
	uint64_t bytes = 0;
	(void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset,
	drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES,
	sizeof (bytes), 1, &bytes);
	drc->drc_bytes_read += bytes;
	}

	drc->drc_ignore_objlist = objlist_create();

	/* these were verified in dmu_recv_begin */
	ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==,
	DMU_SUBSTREAM);
	ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES);

	ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT);
	ASSERT0(drc->drc_os->os_encrypted &&
	(drc->drc_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA));

	/* handle DSL encryption key payload */
	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	nvlist_t *keynvl = NULL;

	ASSERT(drc->drc_os->os_encrypted);
	ASSERT(drc->drc_raw);

	err = nvlist_lookup_nvlist(drc->drc_begin_nvl, "crypt_keydata",
	&keynvl);
	if (err != 0)
	goto out;

	if (!drc->drc_heal) {
	/*
	* If this is a new dataset we set the key immediately.
	* Otherwise we don't want to change the key until we
	* are sure the rest of the receive succeeded so we
	* stash the keynvl away until then.
	*/
	err = dsl_crypto_recv_raw(spa_name(drc->drc_os->os_spa),
	drc->drc_ds->ds_object, drc->drc_fromsnapobj,
	drc->drc_drrb->drr_type, keynvl, drc->drc_newfs);
	if (err != 0)
	goto out;
	}

	/* see comment in dmu_recv_end_sync() */
	drc->drc_ivset_guid = 0;
	(void) nvlist_lookup_uint64(keynvl, "to_ivset_guid",
	&drc->drc_ivset_guid);

	if (!drc->drc_newfs)
	drc->drc_keynvl = fnvlist_dup(keynvl);
	}

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
	err = resume_check(drc, drc->drc_begin_nvl);
	if (err != 0)
	goto out;
	}

	/*
	* If we failed before this point we will clean up any new resume
	* state that was created. Now that we've gotten past the initial
	* checks we are ok to retain that resume state.
	*/
	drc->drc_should_save = B_TRUE;

	(void) bqueue_init(&rwa->q, zfs_recv_queue_ff,
	MAX(zfs_recv_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct receive_record_arg, node));
	cv_init(&rwa->cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&rwa->mutex, NULL, MUTEX_DEFAULT, NULL);
	rwa->os = drc->drc_os;
	rwa->byteswap = drc->drc_byteswap;
	rwa->heal = drc->drc_heal;
	rwa->tofs = drc->drc_tofs;
	rwa->resumable = drc->drc_resumable;
	rwa->raw = drc->drc_raw;
	rwa->spill = drc->drc_spill;
	rwa->full = (drc->drc_drr_begin->drr_u.drr_begin.drr_fromguid == 0);
	rwa->os->os_raw_receive = drc->drc_raw;
	if (drc->drc_heal) {
	rwa->heal_pio = zio_root(drc->drc_os->os_spa, NULL, NULL,
	ZIO_FLAG_GODFATHER);
	}
	list_create(&rwa->write_batch, sizeof (struct receive_record_arg),
	offsetof(struct receive_record_arg, node.bqn_node));

	(void) thread_create(NULL, 0, receive_writer_thread, rwa, 0, curproc,
	TS_RUN, minclsyspri);
	/*
	* We're reading rwa->err without locks, which is safe since we are the
	* only reader, and the worker thread is the only writer. It's ok if we
	* miss a write for an iteration or two of the loop, since the writer
	* thread will keep freeing records we send it until we send it an eos
	* marker.
	*
	* We can leave this loop in 3 ways: First, if rwa->err is
	* non-zero. In that case, the writer thread will free the rrd we just
	* pushed. Second, if we're interrupted; in that case, either it's the
	* first loop and drc->drc_rrd was never allocated, or it's later, and
	* drc->drc_rrd has been handed off to the writer thread who will free
	* it. Finally, if receive_read_record fails or we're at the end of the
	* stream, then we free drc->drc_rrd and exit.
	*/
	while (rwa->err == 0) {
	if (issig(JUSTLOOKING) && issig(FORREAL)) {
	err = SET_ERROR(EINTR);
	break;
	}

	ASSERT3P(drc->drc_rrd, ==, NULL);
	drc->drc_rrd = drc->drc_next_rrd;
	drc->drc_next_rrd = NULL;
	/* Allocates and loads header into drc->drc_next_rrd */
	err = receive_read_record(drc);

	if (drc->drc_rrd->header.drr_type == DRR_END \|\| err != 0) {
	kmem_free(drc->drc_rrd, sizeof (*drc->drc_rrd));
	drc->drc_rrd = NULL;
	break;
	}

	bqueue_enqueue(&rwa->q, drc->drc_rrd,
	sizeof (struct receive_record_arg) +
	drc->drc_rrd->payload_size);
	drc->drc_rrd = NULL;
	}

	ASSERT3P(drc->drc_rrd, ==, NULL);
	drc->drc_rrd = kmem_zalloc(sizeof (*drc->drc_rrd), KM_SLEEP);
	drc->drc_rrd->eos_marker = B_TRUE;
	bqueue_enqueue_flush(&rwa->q, drc->drc_rrd, 1);

	mutex_enter(&rwa->mutex);
	while (!rwa->done) {
	/*
	* We need to use cv_wait_sig() so that any process that may
	* be sleeping here can still fork.
	*/
	(void) cv_wait_sig(&rwa->cv, &rwa->mutex);
	}
	mutex_exit(&rwa->mutex);

	/*
	* If we are receiving a full stream as a clone, all object IDs which
	* are greater than the maximum ID referenced in the stream are
	* by definition unused and must be freed.
	*/
	if (drc->drc_clone && drc->drc_drrb->drr_fromguid == 0) {
	uint64_t obj = rwa->max_object + 1;
	int free_err = 0;
	int next_err = 0;

	while (next_err == 0) {
	free_err = dmu_free_long_object(rwa->os, obj);
	if (free_err != 0 && free_err != ENOENT)
	break;

	next_err = dmu_object_next(rwa->os, &obj, FALSE, 0);
	}

	if (err == 0) {
	if (free_err != 0 && free_err != ENOENT)
	err = free_err;
	else if (next_err != ESRCH)
	err = next_err;
	}
	}

	cv_destroy(&rwa->cv);
	mutex_destroy(&rwa->mutex);
	bqueue_destroy(&rwa->q);
	list_destroy(&rwa->write_batch);
	if (err == 0)
	err = rwa->err;

	out:
	/*
	* If we hit an error before we started the receive_writer_thread
	* we need to clean up the next_rrd we create by processing the
	* DRR_BEGIN record.
	*/
	if (drc->drc_next_rrd != NULL)
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));

	/*
	* The objset will be invalidated by dmu_recv_end() when we do
	* dsl_dataset_clone_swap_sync_impl().
	*/
	drc->drc_os = NULL;

	kmem_free(rwa, sizeof (*rwa));
	nvlist_free(drc->drc_begin_nvl);

	if (err != 0) {
	/*
	* Clean up references. If receive is not resumable,
	* destroy what we created, so we don't leave it in
	* the inconsistent state.
	*/
	dmu_recv_cleanup_ds(drc);
	nvlist_free(drc->drc_keynvl);
	}

	objlist_destroy(drc->drc_ignore_objlist);
	drc->drc_ignore_objlist = NULL;
	*voffp = drc->drc_voff;
	return (err);
	}

	static int
	dmu_recv_end_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_cookie_t *drc = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	int error;

	ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag);

	if (drc->drc_heal) {
	error = 0;
	} else if (!drc->drc_newfs) {
	dsl_dataset_t *origin_head;

	error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head);
	if (error != 0)
	return (error);
	if (drc->drc_force) {
	/*
	* We will destroy any snapshots in tofs (i.e. before
	* origin_head) that are after the origin (which is
	* the snap before drc_ds, because drc_ds can not
	* have any snaps of its own).
	*/
	uint64_t obj;

	obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
	while (obj !=
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
	dsl_dataset_t *snap;
	error = dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap);
	if (error != 0)
	break;
	if (snap->ds_dir != origin_head->ds_dir)
	error = SET_ERROR(EINVAL);
	if (error == 0) {
	error = dsl_destroy_snapshot_check_impl(
	snap, B_FALSE);
	}
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_dataset_rele(snap, FTAG);
	if (error != 0)
	break;
	}
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	}
	if (drc->drc_keynvl != NULL) {
	error = dsl_crypto_recv_raw_key_check(drc->drc_ds,
	drc->drc_keynvl, tx);
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	}

	error = dsl_dataset_clone_swap_check_impl(drc->drc_ds,
	origin_head, drc->drc_force, drc->drc_owner, tx);
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	error = dsl_dataset_snapshot_check_impl(origin_head,
	drc->drc_tosnap, tx, B_TRUE, 1,
	drc->drc_cred, drc->drc_proc);
	dsl_dataset_rele(origin_head, FTAG);
	if (error != 0)
	return (error);

	error = dsl_destroy_head_check_impl(drc->drc_ds, 1);
	} else {
	error = dsl_dataset_snapshot_check_impl(drc->drc_ds,
	drc->drc_tosnap, tx, B_TRUE, 1,
	drc->drc_cred, drc->drc_proc);
	}
	return (error);
	}

	static void
	dmu_recv_end_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_cookie_t *drc = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0;
	uint64_t newsnapobj = 0;

	spa_history_log_internal_ds(drc->drc_ds, "finish receiving",
	tx, "snap=%s", drc->drc_tosnap);
	drc->drc_ds->ds_objset->os_raw_receive = B_FALSE;

	if (drc->drc_heal) {
	if (drc->drc_keynvl != NULL) {
	nvlist_free(drc->drc_keynvl);
	drc->drc_keynvl = NULL;
	}
	} else if (!drc->drc_newfs) {
	dsl_dataset_t *origin_head;

	VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG,
	&origin_head));

	if (drc->drc_force) {
	/*
	* Destroy any snapshots of drc_tofs (origin_head)
	* after the origin (the snap before drc_ds).
	*/
	uint64_t obj;

	obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
	while (obj !=
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
	dsl_dataset_t *snap;
	VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap));
	ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir);
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_destroy_snapshot_sync_impl(snap,
	B_FALSE, tx);
	dsl_dataset_rele(snap, FTAG);
	}
	}
	if (drc->drc_keynvl != NULL) {
	dsl_crypto_recv_raw_key_sync(drc->drc_ds,
	drc->drc_keynvl, tx);
	nvlist_free(drc->drc_keynvl);
	drc->drc_keynvl = NULL;
	}

	VERIFY3P(drc->drc_ds->ds_prev, ==,
	origin_head->ds_prev);

	dsl_dataset_clone_swap_sync_impl(drc->drc_ds,
	origin_head, tx);
	/*
	* The objset was evicted by dsl_dataset_clone_swap_sync_impl,
	* so drc_os is no longer valid.
	*/
	drc->drc_os = NULL;

	dsl_dataset_snapshot_sync_impl(origin_head,
	drc->drc_tosnap, tx);

	/* set snapshot's creation time and guid */
	dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx);
	dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time =
	drc->drc_drrb->drr_creation_time;
	dsl_dataset_phys(origin_head->ds_prev)->ds_guid =
	drc->drc_drrb->drr_toguid;
	dsl_dataset_phys(origin_head->ds_prev)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
	dsl_dataset_phys(origin_head)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	newsnapobj =
	dsl_dataset_phys(origin_head)->ds_prev_snap_obj;

	dsl_dataset_rele(origin_head, FTAG);
	dsl_destroy_head_sync_impl(drc->drc_ds, tx);

	if (drc->drc_owner != NULL)
	VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner);
	} else {
	dsl_dataset_t *ds = drc->drc_ds;

	dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx);

	/* set snapshot's creation time and guid */
	dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
	dsl_dataset_phys(ds->ds_prev)->ds_creation_time =
	drc->drc_drrb->drr_creation_time;
	dsl_dataset_phys(ds->ds_prev)->ds_guid =
	drc->drc_drrb->drr_toguid;
	dsl_dataset_phys(ds->ds_prev)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT;
	if (dsl_dataset_has_resume_receive_state(ds)) {
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_FROMGUID, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OBJECT, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OFFSET, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_BYTES, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TONAME, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, tx);
	}
	newsnapobj =
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj;
	}

	/*
	* If this is a raw receive, the crypt_keydata nvlist will include
	* a to_ivset_guid for us to set on the new snapshot. This value
	* will override the value generated by the snapshot code. However,
	* this value may not be present, because older implementations of
	* the raw send code did not include this value, and we are still
	* allowed to receive them if the zfs_disable_ivset_guid_check
	* tunable is set, in which case we will leave the newly-generated
	* value.
	*/
	if (!drc->drc_heal && drc->drc_raw && drc->drc_ivset_guid != 0) {
	dmu_object_zapify(dp->dp_meta_objset, newsnapobj,
	DMU_OT_DSL_DATASET, tx);
	VERIFY0(zap_update(dp->dp_meta_objset, newsnapobj,
	DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
	&drc->drc_ivset_guid, tx));
	}

	/*
	* Release the hold from dmu_recv_begin. This must be done before
	* we return to open context, so that when we free the dataset's dnode
	* we can evict its bonus buffer. Since the dataset may be destroyed
	* at this point (and therefore won't have a valid pointer to the spa)
	* we release the key mapping manually here while we do have a valid
	* pointer, if it exists.
	*/
	if (!drc->drc_raw && encrypted) {
	(void) spa_keystore_remove_mapping(dmu_tx_pool(tx)->dp_spa,
	drc->drc_ds->ds_object, drc->drc_ds);
	}
	dsl_dataset_disown(drc->drc_ds, 0, dmu_recv_tag);
	drc->drc_ds = NULL;
	}

	static int dmu_recv_end_modified_blocks = 3;

	static int
	dmu_recv_existing_end(dmu_recv_cookie_t *drc)
	{
	#ifdef _KERNEL
	/*
	* We will be destroying the ds; make sure its origin is unmounted if
	* necessary.
	*/
	char name[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(drc->drc_ds, name);
	zfs_destroy_unmount_origin(name);
	#endif

	return (dsl_sync_task(drc->drc_tofs,
	dmu_recv_end_check, dmu_recv_end_sync, drc,
	dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
	}

	static int
	dmu_recv_new_end(dmu_recv_cookie_t *drc)
	{
	return (dsl_sync_task(drc->drc_tofs,
	dmu_recv_end_check, dmu_recv_end_sync, drc,
	dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
	}

	int
	dmu_recv_end(dmu_recv_cookie_t drc, void owner)
	{
	int error;

	drc->drc_owner = owner;

	if (drc->drc_newfs)
	error = dmu_recv_new_end(drc);
	else
	error = dmu_recv_existing_end(drc);

	if (error != 0) {
	dmu_recv_cleanup_ds(drc);
	nvlist_free(drc->drc_keynvl);
	} else if (!drc->drc_heal) {
	if (drc->drc_newfs) {
	zvol_create_minor(drc->drc_tofs);
	}
	char *snapname = kmem_asprintf("%s@%s",
	drc->drc_tofs, drc->drc_tosnap);
	zvol_create_minor(snapname);
	kmem_strfree(snapname);
	}
	return (error);
	}

	/*
	* Return TRUE if this objset is currently being received into.
	*/
	boolean_t
	dmu_objset_is_receiving(objset_t *os)
	{
	return (os->os_dsl_dataset != NULL &&
	os->os_dsl_dataset->ds_owner == dmu_recv_tag);
	}

	-ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_length, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_length, UINT, ZMOD_RW,
	"Maximum receive queue length");

	-ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_ff, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_ff, UINT, ZMOD_RW,
	"Receive queue fill fraction");

	-ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, write_batch_size, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, write_batch_size, UINT, ZMOD_RW,
	"Maximum amount of writes to batch into one transaction");

	ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, best_effort_corrective, INT, ZMOD_RW,
	"Ignore errors during corrective receive");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_send.c b/sys/contrib/openzfs/module/zfs/dmu_send.c
	index 283e2d3b37bb..4ee3ffc352b8 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_send.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_send.c
	@@ -1,3108 +1,3109 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	* Copyright 2016 RackTop Systems.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/spa_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_znode.h>
	#include <zfs_fletcher.h>
	#include <sys/avl.h>
	#include <sys/ddt.h>
	#include <sys/zfs_onexit.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dsl_destroy.h>
	#include <sys/blkptr.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/zfeature.h>
	#include <sys/bqueue.h>
	#include <sys/zvol.h>
	#include <sys/policy.h>
	#include <sys/objlist.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#endif

	/* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */
	static int zfs_send_corrupt_data = B_FALSE;
	/*
	* This tunable controls the amount of data (measured in bytes) that will be
	* prefetched by zfs send. If the main thread is blocking on reads that haven't
	* completed, this variable might need to be increased. If instead the main
	* thread is issuing new reads because the prefetches have fallen out of the
	* cache, this may need to be decreased.
	*/
	-static int zfs_send_queue_length = SPA_MAXBLOCKSIZE;
	+static uint_t zfs_send_queue_length = SPA_MAXBLOCKSIZE;
	/*
	* This tunable controls the length of the queues that zfs send worker threads
	* use to communicate. If the send_main_thread is blocking on these queues,
	* this variable may need to be increased. If there is a significant slowdown
	* at the start of a send as these threads consume all the available IO
	* resources, this variable may need to be decreased.
	*/
	-static int zfs_send_no_prefetch_queue_length = 1024 * 1024;
	+static uint_t zfs_send_no_prefetch_queue_length = 1024 * 1024;
	/*
	* These tunables control the fill fraction of the queues by zfs send. The fill
	* fraction controls the frequency with which threads have to be cv_signaled.
	* If a lot of cpu time is being spent on cv_signal, then these should be tuned
	* down. If the queues empty before the signalled thread can catch up, then
	* these should be tuned up.
	*/
	-static int zfs_send_queue_ff = 20;
	-static int zfs_send_no_prefetch_queue_ff = 20;
	+static uint_t zfs_send_queue_ff = 20;
	+static uint_t zfs_send_no_prefetch_queue_ff = 20;

	/*
	* Use this to override the recordsize calculation for fast zfs send estimates.
	*/
	-static int zfs_override_estimate_recordsize = 0;
	+static uint_t zfs_override_estimate_recordsize = 0;

	/* Set this tunable to FALSE to disable setting of DRR_FLAG_FREERECORDS */
	static const boolean_t zfs_send_set_freerecords_bit = B_TRUE;

	/* Set this tunable to FALSE is disable sending unmodified spill blocks. */
	static int zfs_send_unmodified_spill_blocks = B_TRUE;

	static inline boolean_t
	overflow_multiply(uint64_t a, uint64_t b, uint64_t *c)
	{
	uint64_t temp = a * b;
	if (b != 0 && temp / b != a)
	return (B_FALSE);
	*c = temp;
	return (B_TRUE);
	}

	struct send_thread_arg {
	bqueue_t q;
	objset_t os; / Objset to traverse */
	uint64_t fromtxg; /* Traverse from this txg */
	int flags; /* flags to pass to traverse_dataset */
	int error_code;
	boolean_t cancel;
	zbookmark_phys_t resume;
	uint64_t *num_blocks_visited;
	};

	struct redact_list_thread_arg {
	boolean_t cancel;
	bqueue_t q;
	zbookmark_phys_t resume;
	redaction_list_t *rl;
	boolean_t mark_redact;
	int error_code;
	uint64_t *num_blocks_visited;
	};

	struct send_merge_thread_arg {
	bqueue_t q;
	objset_t *os;
	struct redact_list_thread_arg *from_arg;
	struct send_thread_arg *to_arg;
	struct redact_list_thread_arg *redact_arg;
	int error;
	boolean_t cancel;
	};

	struct send_range {
	boolean_t eos_marker; /* Marks the end of the stream */
	uint64_t object;
	uint64_t start_blkid;
	uint64_t end_blkid;
	bqueue_node_t ln;
	enum type {DATA, HOLE, OBJECT, OBJECT_RANGE, REDACT,
	PREVIOUSLY_REDACTED} type;
	union {
	struct srd {
	dmu_object_type_t obj_type;
	uint32_t datablksz; // logical size
	uint32_t datasz; // payload size
	blkptr_t bp;
	arc_buf_t *abuf;
	abd_t *abd;
	kmutex_t lock;
	kcondvar_t cv;
	boolean_t io_outstanding;
	boolean_t io_compressed;
	int io_err;
	} data;
	struct srh {
	uint32_t datablksz;
	} hole;
	struct sro {
	/*
	* This is a pointer because embedding it in the
	* struct causes these structures to be massively larger
	* for all range types; this makes the code much less
	* memory efficient.
	*/
	dnode_phys_t *dnp;
	blkptr_t bp;
	} object;
	struct srr {
	uint32_t datablksz;
	} redact;
	struct sror {
	blkptr_t bp;
	} object_range;
	} sru;
	};

	/*
	* The list of data whose inclusion in a send stream can be pending from
	* one call to backup_cb to another. Multiple calls to dump_free(),
	* dump_freeobjects(), and dump_redact() can be aggregated into a single
	* DRR_FREE, DRR_FREEOBJECTS, or DRR_REDACT replay record.
	*/
	typedef enum {
	PENDING_NONE,
	PENDING_FREE,
	PENDING_FREEOBJECTS,
	PENDING_REDACT
	} dmu_pendop_t;

	typedef struct dmu_send_cookie {
	dmu_replay_record_t *dsc_drr;
	dmu_send_outparams_t *dsc_dso;
	offset_t *dsc_off;
	objset_t *dsc_os;
	zio_cksum_t dsc_zc;
	uint64_t dsc_toguid;
	uint64_t dsc_fromtxg;
	int dsc_err;
	dmu_pendop_t dsc_pending_op;
	uint64_t dsc_featureflags;
	uint64_t dsc_last_data_object;
	uint64_t dsc_last_data_offset;
	uint64_t dsc_resume_object;
	uint64_t dsc_resume_offset;
	boolean_t dsc_sent_begin;
	boolean_t dsc_sent_end;
	} dmu_send_cookie_t;

	static int do_dump(dmu_send_cookie_t dscp, struct send_range range);

	static void
	range_free(struct send_range *range)
	{
	if (range->type == OBJECT) {
	size_t size = sizeof (dnode_phys_t) *
	(range->sru.object.dnp->dn_extra_slots + 1);
	kmem_free(range->sru.object.dnp, size);
	} else if (range->type == DATA) {
	mutex_enter(&range->sru.data.lock);
	while (range->sru.data.io_outstanding)
	cv_wait(&range->sru.data.cv, &range->sru.data.lock);
	if (range->sru.data.abd != NULL)
	abd_free(range->sru.data.abd);
	if (range->sru.data.abuf != NULL) {
	arc_buf_destroy(range->sru.data.abuf,
	&range->sru.data.abuf);
	}
	mutex_exit(&range->sru.data.lock);

	cv_destroy(&range->sru.data.cv);
	mutex_destroy(&range->sru.data.lock);
	}
	kmem_free(range, sizeof (*range));
	}

	/*
	* For all record types except BEGIN, fill in the checksum (overlaid in
	* drr_u.drr_checksum.drr_checksum). The checksum verifies everything
	* up to the start of the checksum itself.
	*/
	static int
	dump_record(dmu_send_cookie_t dscp, void payload, int payload_len)
	{
	dmu_send_outparams_t *dso = dscp->dsc_dso;
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	(void) fletcher_4_incremental_native(dscp->dsc_drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	&dscp->dsc_zc);
	if (dscp->dsc_drr->drr_type == DRR_BEGIN) {
	dscp->dsc_sent_begin = B_TRUE;
	} else {
	ASSERT(ZIO_CHECKSUM_IS_ZERO(&dscp->dsc_drr->drr_u.
	drr_checksum.drr_checksum));
	dscp->dsc_drr->drr_u.drr_checksum.drr_checksum = dscp->dsc_zc;
	}
	if (dscp->dsc_drr->drr_type == DRR_END) {
	dscp->dsc_sent_end = B_TRUE;
	}
	(void) fletcher_4_incremental_native(&dscp->dsc_drr->
	drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), &dscp->dsc_zc);
	*dscp->dsc_off += sizeof (dmu_replay_record_t);
	dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, dscp->dsc_drr,
	sizeof (dmu_replay_record_t), dso->dso_arg);
	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));
	if (payload_len != 0) {
	*dscp->dsc_off += payload_len;
	/*
	* payload is null when dso_dryrun == B_TRUE (i.e. when we're
	* doing a send size calculation)
	*/
	if (payload != NULL) {
	(void) fletcher_4_incremental_native(
	payload, payload_len, &dscp->dsc_zc);
	}

	/*
	* The code does not rely on this (len being a multiple of 8).
	* We keep this assertion because of the corresponding assertion
	* in receive_read(). Keeping this assertion ensures that we do
	* not inadvertently break backwards compatibility (causing the
	* assertion in receive_read() to trigger on old software).
	*
	* Raw sends cannot be received on old software, and so can
	* bypass this assertion.
	*/

	ASSERT((payload_len % 8 == 0) \|\|
	(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW));

	dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, payload,
	payload_len, dso->dso_arg);
	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));
	}
	return (0);
	}

	/*
	* Fill in the drr_free struct, or perform aggregation if the previous record is
	* also a free record, and the two are adjacent.
	*
	* Note that we send free records even for a full send, because we want to be
	* able to receive a full send as a clone, which requires a list of all the free
	* and freeobject records that were generated on the source.
	*/
	static int
	dump_free(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	struct drr_free *drrf = &(dscp->dsc_drr->drr_u.drr_free);

	/*
	* When we receive a free record, dbuf_free_range() assumes
	* that the receiving system doesn't have any dbufs in the range
	* being freed. This is always true because there is a one-record
	* constraint: we only send one WRITE record for any given
	* object,offset. We know that the one-record constraint is
	* true because we always send data in increasing order by
	* object,offset.
	*
	* If the increasing-order constraint ever changes, we should find
	* another way to assert that the one-record constraint is still
	* satisfied.
	*/
	ASSERT(object > dscp->dsc_last_data_object \|\|
	(object == dscp->dsc_last_data_object &&
	offset > dscp->dsc_last_data_offset));

	/*
	* If there is a pending op, but it's not PENDING_FREE, push it out,
	* since free block aggregation can only be done for blocks of the
	* same type (i.e., DRR_FREE records can only be aggregated with
	* other DRR_FREE records. DRR_FREEOBJECTS records can only be
	* aggregated with other DRR_FREEOBJECTS records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_FREE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_FREE) {
	/*
	* Check to see whether this free block can be aggregated
	* with pending one.
	*/
	if (drrf->drr_object == object && drrf->drr_offset +
	drrf->drr_length == offset) {
	if (offset + length < offset \|\| length == UINT64_MAX)
	drrf->drr_length = UINT64_MAX;
	else
	drrf->drr_length += length;
	return (0);
	} else {
	/* not a continuation. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}
	/* create a FREE record and make it pending */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_FREE;
	drrf->drr_object = object;
	drrf->drr_offset = offset;
	if (offset + length < offset)
	drrf->drr_length = DMU_OBJECT_END;
	else
	drrf->drr_length = length;
	drrf->drr_toguid = dscp->dsc_toguid;
	if (length == DMU_OBJECT_END) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	} else {
	dscp->dsc_pending_op = PENDING_FREE;
	}

	return (0);
	}

	/*
	* Fill in the drr_redact struct, or perform aggregation if the previous record
	* is also a redaction record, and the two are adjacent.
	*/
	static int
	dump_redact(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	struct drr_redact *drrr = &dscp->dsc_drr->drr_u.drr_redact;

	/*
	* If there is a pending op, but it's not PENDING_REDACT, push it out,
	* since free block aggregation can only be done for blocks of the
	* same type (i.e., DRR_REDACT records can only be aggregated with
	* other DRR_REDACT records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_REDACT) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_REDACT) {
	/*
	* Check to see whether this redacted block can be aggregated
	* with pending one.
	*/
	if (drrr->drr_object == object && drrr->drr_offset +
	drrr->drr_length == offset) {
	drrr->drr_length += length;
	return (0);
	} else {
	/* not a continuation. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}
	/* create a REDACT record and make it pending */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_REDACT;
	drrr->drr_object = object;
	drrr->drr_offset = offset;
	drrr->drr_length = length;
	drrr->drr_toguid = dscp->dsc_toguid;
	dscp->dsc_pending_op = PENDING_REDACT;

	return (0);
	}

	static int
	dmu_dump_write(dmu_send_cookie_t *dscp, dmu_object_type_t type, uint64_t object,
	uint64_t offset, int lsize, int psize, const blkptr_t *bp,
	boolean_t io_compressed, void *data)
	{
	uint64_t payload_size;
	boolean_t raw = (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
	struct drr_write *drrw = &(dscp->dsc_drr->drr_u.drr_write);

	/*
	* We send data in increasing object, offset order.
	* See comment in dump_free() for details.
	*/
	ASSERT(object > dscp->dsc_last_data_object \|\|
	(object == dscp->dsc_last_data_object &&
	offset > dscp->dsc_last_data_offset));
	dscp->dsc_last_data_object = object;
	dscp->dsc_last_data_offset = offset + lsize - 1;

	/*
	* If there is any kind of pending aggregation (currently either
	* a grouping of free objects or free blocks), push it out to
	* the stream, since aggregation can't be done across operations
	* of different types.
	*/
	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	/* write a WRITE record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_WRITE;
	drrw->drr_object = object;
	drrw->drr_type = type;
	drrw->drr_offset = offset;
	drrw->drr_toguid = dscp->dsc_toguid;
	drrw->drr_logical_size = lsize;

	/* only set the compression fields if the buf is compressed or raw */
	boolean_t compressed =
	(bp != NULL ? BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
	io_compressed : lsize != psize);
	if (raw \|\| compressed) {
	ASSERT(raw \|\| dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_COMPRESSED);
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT3S(psize, >, 0);

	if (raw) {
	ASSERT(BP_IS_PROTECTED(bp));

	/*
	* This is a raw protected block so we need to pass
	* along everything the receiving side will need to
	* interpret this block, including the byteswap, salt,
	* IV, and MAC.
	*/
	if (BP_SHOULD_BYTESWAP(bp))
	drrw->drr_flags \|= DRR_RAW_BYTESWAP;
	zio_crypt_decode_params_bp(bp, drrw->drr_salt,
	drrw->drr_iv);
	zio_crypt_decode_mac_bp(bp, drrw->drr_mac);
	} else {
	/* this is a compressed block */
	ASSERT(dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_COMPRESSED);
	ASSERT(!BP_SHOULD_BYTESWAP(bp));
	ASSERT(!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)));
	ASSERT3U(BP_GET_COMPRESS(bp), !=, ZIO_COMPRESS_OFF);
	ASSERT3S(lsize, >=, psize);
	}

	/* set fields common to compressed and raw sends */
	drrw->drr_compressiontype = BP_GET_COMPRESS(bp);
	drrw->drr_compressed_size = psize;
	payload_size = drrw->drr_compressed_size;
	} else {
	payload_size = drrw->drr_logical_size;
	}

	if (bp == NULL \|\| BP_IS_EMBEDDED(bp) \|\| (BP_IS_PROTECTED(bp) && !raw)) {
	/*
	* There's no pre-computed checksum for partial-block writes,
	* embedded BP's, or encrypted BP's that are being sent as
	* plaintext, so (like fletcher4-checksummed blocks) userland
	* will have to compute a dedup-capable checksum itself.
	*/
	drrw->drr_checksumtype = ZIO_CHECKSUM_OFF;
	} else {
	drrw->drr_checksumtype = BP_GET_CHECKSUM(bp);
	if (zio_checksum_table[drrw->drr_checksumtype].ci_flags &
	ZCHECKSUM_FLAG_DEDUP)
	drrw->drr_flags \|= DRR_CHECKSUM_DEDUP;
	DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp));
	DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp));
	DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp));
	DDK_SET_CRYPT(&drrw->drr_key, BP_IS_PROTECTED(bp));
	drrw->drr_key.ddk_cksum = bp->blk_cksum;
	}

	if (dump_record(dscp, data, payload_size) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_write_embedded(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	int blksz, const blkptr_t *bp)
	{
	char buf[BPE_PAYLOAD_SIZE];
	struct drr_write_embedded *drrw =
	&(dscp->dsc_drr->drr_u.drr_write_embedded);

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	ASSERT(BP_IS_EMBEDDED(bp));

	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_WRITE_EMBEDDED;
	drrw->drr_object = object;
	drrw->drr_offset = offset;
	drrw->drr_length = blksz;
	drrw->drr_toguid = dscp->dsc_toguid;
	drrw->drr_compression = BP_GET_COMPRESS(bp);
	drrw->drr_etype = BPE_GET_ETYPE(bp);
	drrw->drr_lsize = BPE_GET_LSIZE(bp);
	drrw->drr_psize = BPE_GET_PSIZE(bp);

	decode_embedded_bp_compressed(bp, buf);

	if (dump_record(dscp, buf, P2ROUNDUP(drrw->drr_psize, 8)) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_spill(dmu_send_cookie_t dscp, const blkptr_t bp, uint64_t object,
	void *data)
	{
	struct drr_spill *drrs = &(dscp->dsc_drr->drr_u.drr_spill);
	uint64_t blksz = BP_GET_LSIZE(bp);
	uint64_t payload_size = blksz;

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	/* write a SPILL record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_SPILL;
	drrs->drr_object = object;
	drrs->drr_length = blksz;
	drrs->drr_toguid = dscp->dsc_toguid;

	/* See comment in dump_dnode() for full details */
	if (zfs_send_unmodified_spill_blocks &&
	(bp->blk_birth <= dscp->dsc_fromtxg)) {
	drrs->drr_flags \|= DRR_SPILL_UNMODIFIED;
	}

	/* handle raw send fields */
	if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	ASSERT(BP_IS_PROTECTED(bp));

	if (BP_SHOULD_BYTESWAP(bp))
	drrs->drr_flags \|= DRR_RAW_BYTESWAP;
	drrs->drr_compressiontype = BP_GET_COMPRESS(bp);
	drrs->drr_compressed_size = BP_GET_PSIZE(bp);
	zio_crypt_decode_params_bp(bp, drrs->drr_salt, drrs->drr_iv);
	zio_crypt_decode_mac_bp(bp, drrs->drr_mac);
	payload_size = drrs->drr_compressed_size;
	}

	if (dump_record(dscp, data, payload_size) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_freeobjects(dmu_send_cookie_t *dscp, uint64_t firstobj, uint64_t numobjs)
	{
	struct drr_freeobjects *drrfo = &(dscp->dsc_drr->drr_u.drr_freeobjects);
	uint64_t maxobj = DNODES_PER_BLOCK *
	(DMU_META_DNODE(dscp->dsc_os)->dn_maxblkid + 1);

	/*
	* ZoL < 0.7 does not handle large FREEOBJECTS records correctly,
	* leading to zfs recv never completing. to avoid this issue, don't
	* send FREEOBJECTS records for object IDs which cannot exist on the
	* receiving side.
	*/
	if (maxobj > 0) {
	if (maxobj <= firstobj)
	return (0);

	if (maxobj < firstobj + numobjs)
	numobjs = maxobj - firstobj;
	}

	/*
	* If there is a pending op, but it's not PENDING_FREEOBJECTS,
	* push it out, since free block aggregation can only be done for
	* blocks of the same type (i.e., DRR_FREE records can only be
	* aggregated with other DRR_FREE records. DRR_FREEOBJECTS records
	* can only be aggregated with other DRR_FREEOBJECTS records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_FREEOBJECTS) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_FREEOBJECTS) {
	/*
	* See whether this free object array can be aggregated
	* with pending one
	*/
	if (drrfo->drr_firstobj + drrfo->drr_numobjs == firstobj) {
	drrfo->drr_numobjs += numobjs;
	return (0);
	} else {
	/* can't be aggregated. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}

	/* write a FREEOBJECTS record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_FREEOBJECTS;
	drrfo->drr_firstobj = firstobj;
	drrfo->drr_numobjs = numobjs;
	drrfo->drr_toguid = dscp->dsc_toguid;

	dscp->dsc_pending_op = PENDING_FREEOBJECTS;

	return (0);
	}

	static int
	dump_dnode(dmu_send_cookie_t dscp, const blkptr_t bp, uint64_t object,
	dnode_phys_t *dnp)
	{
	struct drr_object *drro = &(dscp->dsc_drr->drr_u.drr_object);
	int bonuslen;

	if (object < dscp->dsc_resume_object) {
	/*
	* Note: when resuming, we will visit all the dnodes in
	* the block of dnodes that we are resuming from. In
	* this case it's unnecessary to send the dnodes prior to
	* the one we are resuming from. We should be at most one
	* block's worth of dnodes behind the resume point.
	*/
	ASSERT3U(dscp->dsc_resume_object - object, <,
	1 << (DNODE_BLOCK_SHIFT - DNODE_SHIFT));
	return (0);
	}

	if (dnp == NULL \|\| dnp->dn_type == DMU_OT_NONE)
	return (dump_freeobjects(dscp, object, 1));

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	/* write an OBJECT record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_OBJECT;
	drro->drr_object = object;
	drro->drr_type = dnp->dn_type;
	drro->drr_bonustype = dnp->dn_bonustype;
	drro->drr_blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
	drro->drr_bonuslen = dnp->dn_bonuslen;
	drro->drr_dn_slots = dnp->dn_extra_slots + 1;
	drro->drr_checksumtype = dnp->dn_checksum;
	drro->drr_compress = dnp->dn_compress;
	drro->drr_toguid = dscp->dsc_toguid;

	if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
	drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE)
	drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE;

	bonuslen = P2ROUNDUP(dnp->dn_bonuslen, 8);

	if ((dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
	ASSERT(BP_IS_ENCRYPTED(bp));

	if (BP_SHOULD_BYTESWAP(bp))
	drro->drr_flags \|= DRR_RAW_BYTESWAP;

	/* needed for reconstructing dnp on recv side */
	drro->drr_maxblkid = dnp->dn_maxblkid;
	drro->drr_indblkshift = dnp->dn_indblkshift;
	drro->drr_nlevels = dnp->dn_nlevels;
	drro->drr_nblkptr = dnp->dn_nblkptr;

	/*
	* Since we encrypt the entire bonus area, the (raw) part
	* beyond the bonuslen is actually nonzero, so we need
	* to send it.
	*/
	if (bonuslen != 0) {
	if (drro->drr_bonuslen > DN_MAX_BONUS_LEN(dnp))
	return (SET_ERROR(EINVAL));
	drro->drr_raw_bonuslen = DN_MAX_BONUS_LEN(dnp);
	bonuslen = drro->drr_raw_bonuslen;
	}
	}

	/*
	* DRR_OBJECT_SPILL is set for every dnode which references a
	* spill block. This allows the receiving pool to definitively
	* determine when a spill block should be kept or freed.
	*/
	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR)
	drro->drr_flags \|= DRR_OBJECT_SPILL;

	if (dump_record(dscp, DN_BONUS(dnp), bonuslen) != 0)
	return (SET_ERROR(EINTR));

	/* Free anything past the end of the file. */
	if (dump_free(dscp, object, (dnp->dn_maxblkid + 1) *
	(dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), DMU_OBJECT_END) != 0)
	return (SET_ERROR(EINTR));

	/*
	* Send DRR_SPILL records for unmodified spill blocks. This is useful
	* because changing certain attributes of the object (e.g. blocksize)
	* can cause old versions of ZFS to incorrectly remove a spill block.
	* Including these records in the stream forces an up to date version
	* to always be written ensuring they're never lost. Current versions
	* of the code which understand the DRR_FLAG_SPILL_BLOCK feature can
	* ignore these unmodified spill blocks.
	*/
	if (zfs_send_unmodified_spill_blocks &&
	(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) &&
	(DN_SPILL_BLKPTR(dnp)->blk_birth <= dscp->dsc_fromtxg)) {
	struct send_range record;
	blkptr_t *bp = DN_SPILL_BLKPTR(dnp);

	memset(&record, 0, sizeof (struct send_range));
	record.type = DATA;
	record.object = object;
	record.eos_marker = B_FALSE;
	record.start_blkid = DMU_SPILL_BLKID;
	record.end_blkid = record.start_blkid + 1;
	record.sru.data.bp = *bp;
	record.sru.data.obj_type = dnp->dn_type;
	record.sru.data.datablksz = BP_GET_LSIZE(bp);

	if (do_dump(dscp, &record) != 0)
	return (SET_ERROR(EINTR));
	}

	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));

	return (0);
	}

	static int
	dump_object_range(dmu_send_cookie_t dscp, const blkptr_t bp,
	uint64_t firstobj, uint64_t numslots)
	{
	struct drr_object_range *drror =
	&(dscp->dsc_drr->drr_u.drr_object_range);

	/* we only use this record type for raw sends */
	ASSERT(BP_IS_PROTECTED(bp));
	ASSERT(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_DNODE);
	ASSERT0(BP_GET_LEVEL(bp));

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_OBJECT_RANGE;
	drror->drr_firstobj = firstobj;
	drror->drr_numslots = numslots;
	drror->drr_toguid = dscp->dsc_toguid;
	if (BP_SHOULD_BYTESWAP(bp))
	drror->drr_flags \|= DRR_RAW_BYTESWAP;
	zio_crypt_decode_params_bp(bp, drror->drr_salt, drror->drr_iv);
	zio_crypt_decode_mac_bp(bp, drror->drr_mac);

	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static boolean_t
	send_do_embed(const blkptr_t *bp, uint64_t featureflags)
	{
	if (!BP_IS_EMBEDDED(bp))
	return (B_FALSE);

	/*
	* Compression function must be legacy, or explicitly enabled.
	*/
	if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS &&
	!(featureflags & DMU_BACKUP_FEATURE_LZ4)))
	return (B_FALSE);

	/*
	* If we have not set the ZSTD feature flag, we can't send ZSTD
	* compressed embedded blocks, as the receiver may not support them.
	*/
	if ((BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD &&
	!(featureflags & DMU_BACKUP_FEATURE_ZSTD)))
	return (B_FALSE);

	/*
	* Embed type must be explicitly enabled.
	*/
	switch (BPE_GET_ETYPE(bp)) {
	case BP_EMBEDDED_TYPE_DATA:
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
	return (B_TRUE);
	break;
	default:
	return (B_FALSE);
	}
	return (B_FALSE);
	}

	/*
	* This function actually handles figuring out what kind of record needs to be
	* dumped, and calling the appropriate helper function. In most cases,
	* the data has already been read by send_reader_thread().
	*/
	static int
	do_dump(dmu_send_cookie_t dscp, struct send_range range)
	{
	int err = 0;
	switch (range->type) {
	case OBJECT:
	err = dump_dnode(dscp, &range->sru.object.bp, range->object,
	range->sru.object.dnp);
	return (err);
	case OBJECT_RANGE: {
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
	return (0);
	}
	uint64_t epb = BP_GET_LSIZE(&range->sru.object_range.bp) >>
	DNODE_SHIFT;
	uint64_t firstobj = range->start_blkid * epb;
	err = dump_object_range(dscp, &range->sru.object_range.bp,
	firstobj, epb);
	break;
	}
	case REDACT: {
	struct srr *srrp = &range->sru.redact;
	err = dump_redact(dscp, range->object, range->start_blkid *
	srrp->datablksz, (range->end_blkid - range->start_blkid) *
	srrp->datablksz);
	return (err);
	}
	case DATA: {
	struct srd *srdp = &range->sru.data;
	blkptr_t *bp = &srdp->bp;
	spa_t *spa =
	dmu_objset_spa(dscp->dsc_os);

	ASSERT3U(srdp->datablksz, ==, BP_GET_LSIZE(bp));
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	if (BP_GET_TYPE(bp) == DMU_OT_SA) {
	arc_flags_t aflags = ARC_FLAG_WAIT;
	enum zio_flag zioflags = ZIO_FLAG_CANFAIL;

	if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	ASSERT(BP_IS_PROTECTED(bp));
	zioflags \|= ZIO_FLAG_RAW;
	}

	zbookmark_phys_t zb;
	ASSERT3U(range->start_blkid, ==, DMU_SPILL_BLKID);
	zb.zb_objset = dmu_objset_id(dscp->dsc_os);
	zb.zb_object = range->object;
	zb.zb_level = 0;
	zb.zb_blkid = range->start_blkid;

	arc_buf_t *abuf = NULL;
	if (!dscp->dsc_dso->dso_dryrun && arc_read(NULL, spa,
	bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
	zioflags, &aflags, &zb) != 0)
	return (SET_ERROR(EIO));

	err = dump_spill(dscp, bp, zb.zb_object,
	(abuf == NULL ? NULL : abuf->b_data));
	if (abuf != NULL)
	arc_buf_destroy(abuf, &abuf);
	return (err);
	}
	if (send_do_embed(bp, dscp->dsc_featureflags)) {
	err = dump_write_embedded(dscp, range->object,
	range->start_blkid * srdp->datablksz,
	srdp->datablksz, bp);
	return (err);
	}
	ASSERT(range->object > dscp->dsc_resume_object \|\|
	(range->object == dscp->dsc_resume_object &&
	range->start_blkid * srdp->datablksz >=
	dscp->dsc_resume_offset));
	/* it's a level-0 block of a regular object */

	mutex_enter(&srdp->lock);
	while (srdp->io_outstanding)
	cv_wait(&srdp->cv, &srdp->lock);
	err = srdp->io_err;
	mutex_exit(&srdp->lock);

	if (err != 0) {
	if (zfs_send_corrupt_data &&
	!dscp->dsc_dso->dso_dryrun) {
	/*
	* Send a block filled with 0x"zfs badd bloc"
	*/
	srdp->abuf = arc_alloc_buf(spa, &srdp->abuf,
	ARC_BUFC_DATA, srdp->datablksz);
	uint64_t *ptr;
	for (ptr = srdp->abuf->b_data;
	(char )ptr < (char )srdp->abuf->b_data +
	srdp->datablksz; ptr++)
	*ptr = 0x2f5baddb10cULL;
	} else {
	return (SET_ERROR(EIO));
	}
	}

	ASSERT(dscp->dsc_dso->dso_dryrun \|\|
	srdp->abuf != NULL \|\| srdp->abd != NULL);

	uint64_t offset = range->start_blkid * srdp->datablksz;

	char *data = NULL;
	if (srdp->abd != NULL) {
	data = abd_to_buf(srdp->abd);
	ASSERT3P(srdp->abuf, ==, NULL);
	} else if (srdp->abuf != NULL) {
	data = srdp->abuf->b_data;
	}

	/*
	* If we have large blocks stored on disk but the send flags
	* don't allow us to send large blocks, we split the data from
	* the arc buf into chunks.
	*/
	if (srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
	!(dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_LARGE_BLOCKS)) {
	while (srdp->datablksz > 0 && err == 0) {
	int n = MIN(srdp->datablksz,
	SPA_OLD_MAXBLOCKSIZE);
	err = dmu_dump_write(dscp, srdp->obj_type,
	range->object, offset, n, n, NULL, B_FALSE,
	data);
	offset += n;
	/*
	* When doing dry run, data==NULL is used as a
	* sentinel value by
	* dmu_dump_write()->dump_record().
	*/
	if (data != NULL)
	data += n;
	srdp->datablksz -= n;
	}
	} else {
	err = dmu_dump_write(dscp, srdp->obj_type,
	range->object, offset,
	srdp->datablksz, srdp->datasz, bp,
	srdp->io_compressed, data);
	}
	return (err);
	}
	case HOLE: {
	struct srh *srhp = &range->sru.hole;
	if (range->object == DMU_META_DNODE_OBJECT) {
	uint32_t span = srhp->datablksz >> DNODE_SHIFT;
	uint64_t first_obj = range->start_blkid * span;
	uint64_t numobj = range->end_blkid * span - first_obj;
	return (dump_freeobjects(dscp, first_obj, numobj));
	}
	uint64_t offset = 0;

	/*
	* If this multiply overflows, we don't need to send this block.
	* Even if it has a birth time, it can never not be a hole, so
	* we don't need to send records for it.
	*/
	if (!overflow_multiply(range->start_blkid, srhp->datablksz,
	&offset)) {
	return (0);
	}
	uint64_t len = 0;

	if (!overflow_multiply(range->end_blkid, srhp->datablksz, &len))
	len = UINT64_MAX;
	len = len - offset;
	return (dump_free(dscp, range->object, offset, len));
	}
	default:
	panic("Invalid range type in do_dump: %d", range->type);
	}
	return (err);
	}

	static struct send_range *
	range_alloc(enum type type, uint64_t object, uint64_t start_blkid,
	uint64_t end_blkid, boolean_t eos)
	{
	struct send_range range = kmem_alloc(sizeof (range), KM_SLEEP);
	range->type = type;
	range->object = object;
	range->start_blkid = start_blkid;
	range->end_blkid = end_blkid;
	range->eos_marker = eos;
	if (type == DATA) {
	range->sru.data.abd = NULL;
	range->sru.data.abuf = NULL;
	mutex_init(&range->sru.data.lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&range->sru.data.cv, NULL, CV_DEFAULT, NULL);
	range->sru.data.io_outstanding = 0;
	range->sru.data.io_err = 0;
	range->sru.data.io_compressed = B_FALSE;
	}
	return (range);
	}

	/*
	* This is the callback function to traverse_dataset that acts as a worker
	* thread for dmu_send_impl.
	*/
	static int
	send_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const struct dnode_phys dnp, void *arg)
	{
	(void) zilog;
	struct send_thread_arg *sta = arg;
	struct send_range *record;

	ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT \|\|
	zb->zb_object >= sta->resume.zb_object);

	/*
	* All bps of an encrypted os should have the encryption bit set.
	* If this is not true it indicates tampering and we report an error.
	*/
	if (sta->os->os_encrypted &&
	!BP_IS_HOLE(bp) && !BP_USES_CRYPT(bp)) {
	spa_log_error(spa, zb);
	zfs_panic_recover("unencrypted block in encrypted "
	"object set %llu", dmu_objset_id(sta->os));
	return (SET_ERROR(EIO));
	}

	if (sta->cancel)
	return (SET_ERROR(EINTR));
	if (zb->zb_object != DMU_META_DNODE_OBJECT &&
	DMU_OBJECT_IS_SPECIAL(zb->zb_object))
	return (0);
	atomic_inc_64(sta->num_blocks_visited);

	if (zb->zb_level == ZB_DNODE_LEVEL) {
	if (zb->zb_object == DMU_META_DNODE_OBJECT)
	return (0);
	record = range_alloc(OBJECT, zb->zb_object, 0, 0, B_FALSE);
	record->sru.object.bp = *bp;
	size_t size = sizeof (dnp) (dnp->dn_extra_slots + 1);
	record->sru.object.dnp = kmem_alloc(size, KM_SLEEP);
	memcpy(record->sru.object.dnp, dnp, size);
	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}
	if (zb->zb_level == 0 && zb->zb_object == DMU_META_DNODE_OBJECT &&
	!BP_IS_HOLE(bp)) {
	record = range_alloc(OBJECT_RANGE, 0, zb->zb_blkid,
	zb->zb_blkid + 1, B_FALSE);
	record->sru.object_range.bp = *bp;
	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}
	if (zb->zb_level < 0 \|\| (zb->zb_level > 0 && !BP_IS_HOLE(bp)))
	return (0);
	if (zb->zb_object == DMU_META_DNODE_OBJECT && !BP_IS_HOLE(bp))
	return (0);

	uint64_t span = bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level);
	uint64_t start;

	/*
	* If this multiply overflows, we don't need to send this block.
	* Even if it has a birth time, it can never not be a hole, so
	* we don't need to send records for it.
	*/
	if (!overflow_multiply(span, zb->zb_blkid, &start) \|\| (!(zb->zb_blkid ==
	DMU_SPILL_BLKID \|\| DMU_OT_IS_METADATA(dnp->dn_type)) &&
	span * zb->zb_blkid > dnp->dn_maxblkid)) {
	ASSERT(BP_IS_HOLE(bp));
	return (0);
	}

	if (zb->zb_blkid == DMU_SPILL_BLKID)
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);

	enum type record_type = DATA;
	if (BP_IS_HOLE(bp))
	record_type = HOLE;
	else if (BP_IS_REDACTED(bp))
	record_type = REDACT;
	else
	record_type = DATA;

	record = range_alloc(record_type, zb->zb_object, start,
	(start + span < start ? 0 : start + span), B_FALSE);

	uint64_t datablksz = (zb->zb_blkid == DMU_SPILL_BLKID ?
	BP_GET_LSIZE(bp) : dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);

	if (BP_IS_HOLE(bp)) {
	record->sru.hole.datablksz = datablksz;
	} else if (BP_IS_REDACTED(bp)) {
	record->sru.redact.datablksz = datablksz;
	} else {
	record->sru.data.datablksz = datablksz;
	record->sru.data.obj_type = dnp->dn_type;
	record->sru.data.bp = *bp;
	}

	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}

	struct redact_list_cb_arg {
	uint64_t *num_blocks_visited;
	bqueue_t *q;
	boolean_t *cancel;
	boolean_t mark_redact;
	};

	static int
	redact_list_cb(redact_block_phys_t rb, void arg)
	{
	struct redact_list_cb_arg *rlcap = arg;

	atomic_inc_64(rlcap->num_blocks_visited);
	if (*rlcap->cancel)
	return (-1);

	struct send_range *data = range_alloc(REDACT, rb->rbp_object,
	rb->rbp_blkid, rb->rbp_blkid + redact_block_get_count(rb), B_FALSE);
	ASSERT3U(data->end_blkid, >, rb->rbp_blkid);
	if (rlcap->mark_redact) {
	data->type = REDACT;
	data->sru.redact.datablksz = redact_block_get_size(rb);
	} else {
	data->type = PREVIOUSLY_REDACTED;
	}
	bqueue_enqueue(rlcap->q, data, sizeof (*data));

	return (0);
	}

	/*
	* This function kicks off the traverse_dataset. It also handles setting the
	* error code of the thread in case something goes wrong, and pushes the End of
	* Stream record when the traverse_dataset call has finished.
	*/
	static __attribute__((noreturn)) void
	send_traverse_thread(void *arg)
	{
	struct send_thread_arg *st_arg = arg;
	int err = 0;
	struct send_range *data;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	err = traverse_dataset_resume(st_arg->os->os_dsl_dataset,
	st_arg->fromtxg, &st_arg->resume,
	st_arg->flags, send_cb, st_arg);

	if (err != EINTR)
	st_arg->error_code = err;
	data = range_alloc(DATA, 0, 0, 0, B_TRUE);
	bqueue_enqueue_flush(&st_arg->q, data, sizeof (*data));
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	/*
	* Utility function that causes End of Stream records to compare after of all
	* others, so that other threads' comparison logic can stay simple.
	*/
	static int __attribute__((unused))
	send_range_after(const struct send_range from, const struct send_range to)
	{
	if (from->eos_marker == B_TRUE)
	return (1);
	if (to->eos_marker == B_TRUE)
	return (-1);

	uint64_t from_obj = from->object;
	uint64_t from_end_obj = from->object + 1;
	uint64_t to_obj = to->object;
	uint64_t to_end_obj = to->object + 1;
	if (from_obj == 0) {
	ASSERT(from->type == HOLE \|\| from->type == OBJECT_RANGE);
	from_obj = from->start_blkid << DNODES_PER_BLOCK_SHIFT;
	from_end_obj = from->end_blkid << DNODES_PER_BLOCK_SHIFT;
	}
	if (to_obj == 0) {
	ASSERT(to->type == HOLE \|\| to->type == OBJECT_RANGE);
	to_obj = to->start_blkid << DNODES_PER_BLOCK_SHIFT;
	to_end_obj = to->end_blkid << DNODES_PER_BLOCK_SHIFT;
	}

	if (from_end_obj <= to_obj)
	return (-1);
	if (from_obj >= to_end_obj)
	return (1);
	int64_t cmp = TREE_CMP(to->type == OBJECT_RANGE, from->type ==
	OBJECT_RANGE);
	if (unlikely(cmp))
	return (cmp);
	cmp = TREE_CMP(to->type == OBJECT, from->type == OBJECT);
	if (unlikely(cmp))
	return (cmp);
	if (from->end_blkid <= to->start_blkid)
	return (-1);
	if (from->start_blkid >= to->end_blkid)
	return (1);
	return (0);
	}

	/*
	* Pop the new data off the queue, check that the records we receive are in
	* the right order, but do not free the old data. This is used so that the
	* records can be sent on to the main thread without copying the data.
	*/
	static struct send_range *
	get_next_range_nofree(bqueue_t bq, struct send_range prev)
	{
	struct send_range *next = bqueue_dequeue(bq);
	ASSERT3S(send_range_after(prev, next), ==, -1);
	return (next);
	}

	/*
	* Pop the new data off the queue, check that the records we receive are in
	* the right order, and free the old data.
	*/
	static struct send_range *
	get_next_range(bqueue_t bq, struct send_range prev)
	{
	struct send_range *next = get_next_range_nofree(bq, prev);
	range_free(prev);
	return (next);
	}

	static __attribute__((noreturn)) void
	redact_list_thread(void *arg)
	{
	struct redact_list_thread_arg *rlt_arg = arg;
	struct send_range *record;
	fstrans_cookie_t cookie = spl_fstrans_mark();
	if (rlt_arg->rl != NULL) {
	struct redact_list_cb_arg rlcba = {0};
	rlcba.cancel = &rlt_arg->cancel;
	rlcba.q = &rlt_arg->q;
	rlcba.num_blocks_visited = rlt_arg->num_blocks_visited;
	rlcba.mark_redact = rlt_arg->mark_redact;
	int err = dsl_redaction_list_traverse(rlt_arg->rl,
	&rlt_arg->resume, redact_list_cb, &rlcba);
	if (err != EINTR)
	rlt_arg->error_code = err;
	}
	record = range_alloc(DATA, 0, 0, 0, B_TRUE);
	bqueue_enqueue_flush(&rlt_arg->q, record, sizeof (*record));
	spl_fstrans_unmark(cookie);

	thread_exit();
	}

	/*
	* Compare the start point of the two provided ranges. End of stream ranges
	* compare last, objects compare before any data or hole inside that object and
	* multi-object holes that start at the same object.
	*/
	static int
	send_range_start_compare(struct send_range r1, struct send_range r2)
	{
	uint64_t r1_objequiv = r1->object;
	uint64_t r1_l0equiv = r1->start_blkid;
	uint64_t r2_objequiv = r2->object;
	uint64_t r2_l0equiv = r2->start_blkid;
	int64_t cmp = TREE_CMP(r1->eos_marker, r2->eos_marker);
	if (unlikely(cmp))
	return (cmp);
	if (r1->object == 0) {
	r1_objequiv = r1->start_blkid * DNODES_PER_BLOCK;
	r1_l0equiv = 0;
	}
	if (r2->object == 0) {
	r2_objequiv = r2->start_blkid * DNODES_PER_BLOCK;
	r2_l0equiv = 0;
	}

	cmp = TREE_CMP(r1_objequiv, r2_objequiv);
	if (likely(cmp))
	return (cmp);
	cmp = TREE_CMP(r2->type == OBJECT_RANGE, r1->type == OBJECT_RANGE);
	if (unlikely(cmp))
	return (cmp);
	cmp = TREE_CMP(r2->type == OBJECT, r1->type == OBJECT);
	if (unlikely(cmp))
	return (cmp);

	return (TREE_CMP(r1_l0equiv, r2_l0equiv));
	}

	enum q_idx {
	REDACT_IDX = 0,
	TO_IDX,
	FROM_IDX,
	NUM_THREADS
	};

	/*
	* This function returns the next range the send_merge_thread should operate on.
	* The inputs are two arrays; the first one stores the range at the front of the
	* queues stored in the second one. The ranges are sorted in descending
	* priority order; the metadata from earlier ranges overrules metadata from
	* later ranges. out_mask is used to return which threads the ranges came from;
	* bit i is set if ranges[i] started at the same place as the returned range.
	*
	* This code is not hardcoded to compare a specific number of threads; it could
	* be used with any number, just by changing the q_idx enum.
	*
	* The "next range" is the one with the earliest start; if two starts are equal,
	* the highest-priority range is the next to operate on. If a higher-priority
	* range starts in the middle of the first range, then the first range will be
	* truncated to end where the higher-priority range starts, and we will operate
	* on that one next time. In this way, we make sure that each block covered by
	* some range gets covered by a returned range, and each block covered is
	* returned using the metadata of the highest-priority range it appears in.
	*
	* For example, if the three ranges at the front of the queues were [2,4),
	* [3,5), and [1,3), then the ranges returned would be [1,2) with the metadata
	* from the third range, [2,4) with the metadata from the first range, and then
	* [4,5) with the metadata from the second.
	*/
	static struct send_range *
	find_next_range(struct send_range ranges, bqueue_t qs, uint64_t *out_mask)
	{
	int idx = 0; // index of the range with the earliest start
	int i;
	uint64_t bmask = 0;
	for (i = 1; i < NUM_THREADS; i++) {
	if (send_range_start_compare(ranges[i], ranges[idx]) < 0)
	idx = i;
	}
	if (ranges[idx]->eos_marker) {
	struct send_range *ret = range_alloc(DATA, 0, 0, 0, B_TRUE);
	*out_mask = 0;
	return (ret);
	}
	/*
	* Find all the ranges that start at that same point.
	*/
	for (i = 0; i < NUM_THREADS; i++) {
	if (send_range_start_compare(ranges[i], ranges[idx]) == 0)
	bmask \|= 1 << i;
	}
	*out_mask = bmask;
	/*
	* OBJECT_RANGE records only come from the TO thread, and should always
	* be treated as overlapping with nothing and sent on immediately. They
	* are only used in raw sends, and are never redacted.
	*/
	if (ranges[idx]->type == OBJECT_RANGE) {
	ASSERT3U(idx, ==, TO_IDX);
	ASSERT3U(*out_mask, ==, 1 << TO_IDX);
	struct send_range *ret = ranges[idx];
	ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
	return (ret);
	}
	/*
	* Find the first start or end point after the start of the first range.
	*/
	uint64_t first_change = ranges[idx]->end_blkid;
	for (i = 0; i < NUM_THREADS; i++) {
	if (i == idx \|\| ranges[i]->eos_marker \|\|
	ranges[i]->object > ranges[idx]->object \|\|
	ranges[i]->object == DMU_META_DNODE_OBJECT)
	continue;
	ASSERT3U(ranges[i]->object, ==, ranges[idx]->object);
	if (first_change > ranges[i]->start_blkid &&
	(bmask & (1 << i)) == 0)
	first_change = ranges[i]->start_blkid;
	else if (first_change > ranges[i]->end_blkid)
	first_change = ranges[i]->end_blkid;
	}
	/*
	* Update all ranges to no longer overlap with the range we're
	* returning. All such ranges must start at the same place as the range
	* being returned, and end at or after first_change. Thus we update
	* their start to first_change. If that makes them size 0, then free
	* them and pull a new range from that thread.
	*/
	for (i = 0; i < NUM_THREADS; i++) {
	if (i == idx \|\| (bmask & (1 << i)) == 0)
	continue;
	ASSERT3U(first_change, >, ranges[i]->start_blkid);
	ranges[i]->start_blkid = first_change;
	ASSERT3U(ranges[i]->start_blkid, <=, ranges[i]->end_blkid);
	if (ranges[i]->start_blkid == ranges[i]->end_blkid)
	ranges[i] = get_next_range(qs[i], ranges[i]);
	}
	/*
	* Short-circuit the simple case; if the range doesn't overlap with
	* anything else, or it only overlaps with things that start at the same
	* place and are longer, send it on.
	*/
	if (first_change == ranges[idx]->end_blkid) {
	struct send_range *ret = ranges[idx];
	ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
	return (ret);
	}

	/*
	* Otherwise, return a truncated copy of ranges[idx] and move the start
	* of ranges[idx] back to first_change.
	*/
	struct send_range ret = kmem_alloc(sizeof (ret), KM_SLEEP);
	ret = ranges[idx];
	ret->end_blkid = first_change;
	ranges[idx]->start_blkid = first_change;
	return (ret);
	}

	#define FROM_AND_REDACT_BITS ((1 << REDACT_IDX) \| (1 << FROM_IDX))

	/*
	* Merge the results from the from thread and the to thread, and then hand the
	* records off to send_prefetch_thread to prefetch them. If this is not a
	* send from a redaction bookmark, the from thread will push an end of stream
	* record and stop, and we'll just send everything that was changed in the
	* to_ds since the ancestor's creation txg. If it is, then since
	* traverse_dataset has a canonical order, we can compare each change as
	* they're pulled off the queues. That will give us a stream that is
	* appropriately sorted, and covers all records. In addition, we pull the
	* data from the redact_list_thread and use that to determine which blocks
	* should be redacted.
	*/
	static __attribute__((noreturn)) void
	send_merge_thread(void *arg)
	{
	struct send_merge_thread_arg *smt_arg = arg;
	struct send_range *front_ranges[NUM_THREADS];
	bqueue_t *queues[NUM_THREADS];
	int err = 0;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	if (smt_arg->redact_arg == NULL) {
	front_ranges[REDACT_IDX] =
	kmem_zalloc(sizeof (struct send_range), KM_SLEEP);
	front_ranges[REDACT_IDX]->eos_marker = B_TRUE;
	front_ranges[REDACT_IDX]->type = REDACT;
	queues[REDACT_IDX] = NULL;
	} else {
	front_ranges[REDACT_IDX] =
	bqueue_dequeue(&smt_arg->redact_arg->q);
	queues[REDACT_IDX] = &smt_arg->redact_arg->q;
	}
	front_ranges[TO_IDX] = bqueue_dequeue(&smt_arg->to_arg->q);
	queues[TO_IDX] = &smt_arg->to_arg->q;
	front_ranges[FROM_IDX] = bqueue_dequeue(&smt_arg->from_arg->q);
	queues[FROM_IDX] = &smt_arg->from_arg->q;
	uint64_t mask = 0;
	struct send_range *range;
	for (range = find_next_range(front_ranges, queues, &mask);
	!range->eos_marker && err == 0 && !smt_arg->cancel;
	range = find_next_range(front_ranges, queues, &mask)) {
	/*
	* If the range in question was in both the from redact bookmark
	* and the bookmark we're using to redact, then don't send it.
	* It's already redacted on the receiving system, so a redaction
	* record would be redundant.
	*/
	if ((mask & FROM_AND_REDACT_BITS) == FROM_AND_REDACT_BITS) {
	ASSERT3U(range->type, ==, REDACT);
	range_free(range);
	continue;
	}
	bqueue_enqueue(&smt_arg->q, range, sizeof (*range));

	if (smt_arg->to_arg->error_code != 0) {
	err = smt_arg->to_arg->error_code;
	} else if (smt_arg->from_arg->error_code != 0) {
	err = smt_arg->from_arg->error_code;
	} else if (smt_arg->redact_arg != NULL &&
	smt_arg->redact_arg->error_code != 0) {
	err = smt_arg->redact_arg->error_code;
	}
	}
	if (smt_arg->cancel && err == 0)
	err = SET_ERROR(EINTR);
	smt_arg->error = err;
	if (smt_arg->error != 0) {
	smt_arg->to_arg->cancel = B_TRUE;
	smt_arg->from_arg->cancel = B_TRUE;
	if (smt_arg->redact_arg != NULL)
	smt_arg->redact_arg->cancel = B_TRUE;
	}
	for (int i = 0; i < NUM_THREADS; i++) {
	while (!front_ranges[i]->eos_marker) {
	front_ranges[i] = get_next_range(queues[i],
	front_ranges[i]);
	}
	range_free(front_ranges[i]);
	}
	if (range == NULL)
	range = kmem_zalloc(sizeof (*range), KM_SLEEP);
	range->eos_marker = B_TRUE;
	bqueue_enqueue_flush(&smt_arg->q, range, 1);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	struct send_reader_thread_arg {
	struct send_merge_thread_arg *smta;
	bqueue_t q;
	boolean_t cancel;
	boolean_t issue_reads;
	uint64_t featureflags;
	int error;
	};

	static void
	dmu_send_read_done(zio_t *zio)
	{
	struct send_range *range = zio->io_private;

	mutex_enter(&range->sru.data.lock);
	if (zio->io_error != 0) {
	abd_free(range->sru.data.abd);
	range->sru.data.abd = NULL;
	range->sru.data.io_err = zio->io_error;
	}

	ASSERT(range->sru.data.io_outstanding);
	range->sru.data.io_outstanding = B_FALSE;
	cv_broadcast(&range->sru.data.cv);
	mutex_exit(&range->sru.data.lock);
	}

	static void
	issue_data_read(struct send_reader_thread_arg srta, struct send_range range)
	{
	struct srd *srdp = &range->sru.data;
	blkptr_t *bp = &srdp->bp;
	objset_t *os = srta->smta->os;

	ASSERT3U(range->type, ==, DATA);
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	/*
	* If we have large blocks stored on disk but
	* the send flags don't allow us to send large
	* blocks, we split the data from the arc buf
	* into chunks.
	*/
	boolean_t split_large_blocks =
	srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
	!(srta->featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS);
	/*
	* We should only request compressed data from the ARC if all
	* the following are true:
	* - stream compression was requested
	* - we aren't splitting large blocks into smaller chunks
	* - the data won't need to be byteswapped before sending
	* - this isn't an embedded block
	* - this isn't metadata (if receiving on a different endian
	* system it can be byteswapped more easily)
	*/
	boolean_t request_compressed =
	(srta->featureflags & DMU_BACKUP_FEATURE_COMPRESSED) &&
	!split_large_blocks && !BP_SHOULD_BYTESWAP(bp) &&
	!BP_IS_EMBEDDED(bp) && !DMU_OT_IS_METADATA(BP_GET_TYPE(bp));

	enum zio_flag zioflags = ZIO_FLAG_CANFAIL;

	if (srta->featureflags & DMU_BACKUP_FEATURE_RAW) {
	zioflags \|= ZIO_FLAG_RAW;
	srdp->io_compressed = B_TRUE;
	} else if (request_compressed) {
	zioflags \|= ZIO_FLAG_RAW_COMPRESS;
	srdp->io_compressed = B_TRUE;
	}

	srdp->datasz = (zioflags & ZIO_FLAG_RAW_COMPRESS) ?
	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp);

	if (!srta->issue_reads)
	return;
	if (BP_IS_REDACTED(bp))
	return;
	if (send_do_embed(bp, srta->featureflags))
	return;

	zbookmark_phys_t zb = {
	.zb_objset = dmu_objset_id(os),
	.zb_object = range->object,
	.zb_level = 0,
	.zb_blkid = range->start_blkid,
	};

	arc_flags_t aflags = ARC_FLAG_CACHED_ONLY;

	int arc_err = arc_read(NULL, os->os_spa, bp,
	arc_getbuf_func, &srdp->abuf, ZIO_PRIORITY_ASYNC_READ,
	zioflags, &aflags, &zb);
	/*
	* If the data is not already cached in the ARC, we read directly
	* from zio. This avoids the performance overhead of adding a new
	* entry to the ARC, and we also avoid polluting the ARC cache with
	* data that is not likely to be used in the future.
	*/
	if (arc_err != 0) {
	srdp->abd = abd_alloc_linear(srdp->datasz, B_FALSE);
	srdp->io_outstanding = B_TRUE;
	zio_nowait(zio_read(NULL, os->os_spa, bp, srdp->abd,
	srdp->datasz, dmu_send_read_done, range,
	ZIO_PRIORITY_ASYNC_READ, zioflags, &zb));
	}
	}

	/*
	* Create a new record with the given values.
	*/
	static void
	enqueue_range(struct send_reader_thread_arg srta, bqueue_t q, dnode_t *dn,
	uint64_t blkid, uint64_t count, const blkptr_t *bp, uint32_t datablksz)
	{
	enum type range_type = (bp == NULL \|\| BP_IS_HOLE(bp) ? HOLE :
	(BP_IS_REDACTED(bp) ? REDACT : DATA));

	struct send_range *range = range_alloc(range_type, dn->dn_object,
	blkid, blkid + count, B_FALSE);

	if (blkid == DMU_SPILL_BLKID)
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);

	switch (range_type) {
	case HOLE:
	range->sru.hole.datablksz = datablksz;
	break;
	case DATA:
	ASSERT3U(count, ==, 1);
	range->sru.data.datablksz = datablksz;
	range->sru.data.obj_type = dn->dn_type;
	range->sru.data.bp = *bp;
	issue_data_read(srta, range);
	break;
	case REDACT:
	range->sru.redact.datablksz = datablksz;
	break;
	default:
	break;
	}
	bqueue_enqueue(q, range, datablksz);
	}

	/*
	* This thread is responsible for two things: First, it retrieves the correct
	* blkptr in the to ds if we need to send the data because of something from
	* the from thread. As a result of this, we're the first ones to discover that
	* some indirect blocks can be discarded because they're not holes. Second,
	* it issues prefetches for the data we need to send.
	*/
	static __attribute__((noreturn)) void
	send_reader_thread(void *arg)
	{
	struct send_reader_thread_arg *srta = arg;
	struct send_merge_thread_arg *smta = srta->smta;
	bqueue_t *inq = &smta->q;
	bqueue_t *outq = &srta->q;
	objset_t *os = smta->os;
	fstrans_cookie_t cookie = spl_fstrans_mark();
	struct send_range *range = bqueue_dequeue(inq);
	int err = 0;

	/*
	* If the record we're analyzing is from a redaction bookmark from the
	* fromds, then we need to know whether or not it exists in the tods so
	* we know whether to create records for it or not. If it does, we need
	* the datablksz so we can generate an appropriate record for it.
	* Finally, if it isn't redacted, we need the blkptr so that we can send
	* a WRITE record containing the actual data.
	*/
	uint64_t last_obj = UINT64_MAX;
	uint64_t last_obj_exists = B_TRUE;
	while (!range->eos_marker && !srta->cancel && smta->error == 0 &&
	err == 0) {
	switch (range->type) {
	case DATA:
	issue_data_read(srta, range);
	bqueue_enqueue(outq, range, range->sru.data.datablksz);
	range = get_next_range_nofree(inq, range);
	break;
	case HOLE:
	case OBJECT:
	case OBJECT_RANGE:
	case REDACT: // Redacted blocks must exist
	bqueue_enqueue(outq, range, sizeof (*range));
	range = get_next_range_nofree(inq, range);
	break;
	case PREVIOUSLY_REDACTED: {
	/*
	* This entry came from the "from bookmark" when
	* sending from a bookmark that has a redaction
	* list. We need to check if this object/blkid
	* exists in the target ("to") dataset, and if
	* not then we drop this entry. We also need
	* to fill in the block pointer so that we know
	* what to prefetch.
	*
	* To accomplish the above, we first cache whether or
	* not the last object we examined exists. If it
	* doesn't, we can drop this record. If it does, we hold
	* the dnode and use it to call dbuf_dnode_findbp. We do
	* this instead of dbuf_bookmark_findbp because we will
	* often operate on large ranges, and holding the dnode
	* once is more efficient.
	*/
	boolean_t object_exists = B_TRUE;
	/*
	* If the data is redacted, we only care if it exists,
	* so that we don't send records for objects that have
	* been deleted.
	*/
	dnode_t *dn;
	if (range->object == last_obj && !last_obj_exists) {
	/*
	* If we're still examining the same object as
	* previously, and it doesn't exist, we don't
	* need to call dbuf_bookmark_findbp.
	*/
	object_exists = B_FALSE;
	} else {
	err = dnode_hold(os, range->object, FTAG, &dn);
	if (err == ENOENT) {
	object_exists = B_FALSE;
	err = 0;
	}
	last_obj = range->object;
	last_obj_exists = object_exists;
	}

	if (err != 0) {
	break;
	} else if (!object_exists) {
	/*
	* The block was modified, but doesn't
	* exist in the to dataset; if it was
	* deleted in the to dataset, then we'll
	* visit the hole bp for it at some point.
	*/
	range = get_next_range(inq, range);
	continue;
	}
	uint64_t file_max =
	(dn->dn_maxblkid < range->end_blkid ?
	dn->dn_maxblkid : range->end_blkid);
	/*
	* The object exists, so we need to try to find the
	* blkptr for each block in the range we're processing.
	*/
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	for (uint64_t blkid = range->start_blkid;
	blkid < file_max; blkid++) {
	blkptr_t bp;
	uint32_t datablksz =
	dn->dn_phys->dn_datablkszsec <<
	SPA_MINBLOCKSHIFT;
	uint64_t offset = blkid * datablksz;
	/*
	* This call finds the next non-hole block in
	* the object. This is to prevent a
	* performance problem where we're unredacting
	* a large hole. Using dnode_next_offset to
	* skip over the large hole avoids iterating
	* over every block in it.
	*/
	err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK,
	&offset, 1, 1, 0);
	if (err == ESRCH) {
	offset = UINT64_MAX;
	err = 0;
	} else if (err != 0) {
	break;
	}
	if (offset != blkid * datablksz) {
	/*
	* if there is a hole from here
	* (blkid) to offset
	*/
	offset = MIN(offset, file_max *
	datablksz);
	uint64_t nblks = (offset / datablksz) -
	blkid;
	enqueue_range(srta, outq, dn, blkid,
	nblks, NULL, datablksz);
	blkid += nblks;
	}
	if (blkid >= file_max)
	break;
	err = dbuf_dnode_findbp(dn, 0, blkid, &bp,
	NULL, NULL);
	if (err != 0)
	break;
	ASSERT(!BP_IS_HOLE(&bp));
	enqueue_range(srta, outq, dn, blkid, 1, &bp,
	datablksz);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	range = get_next_range(inq, range);
	}
	}
	}
	if (srta->cancel \|\| err != 0) {
	smta->cancel = B_TRUE;
	srta->error = err;
	} else if (smta->error != 0) {
	srta->error = smta->error;
	}
	while (!range->eos_marker)
	range = get_next_range(inq, range);

	bqueue_enqueue_flush(outq, range, 1);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	#define NUM_SNAPS_NOT_REDACTED UINT64_MAX

	struct dmu_send_params {
	/* Pool args */
	const void *tag; // Tag dp was held with, will be used to release dp.
	dsl_pool_t *dp;
	/* To snapshot args */
	const char *tosnap;
	dsl_dataset_t *to_ds;
	/* From snapshot args */
	zfs_bookmark_phys_t ancestor_zb;
	uint64_t *fromredactsnaps;
	/* NUM_SNAPS_NOT_REDACTED if not sending from redaction bookmark */
	uint64_t numfromredactsnaps;
	/* Stream params */
	boolean_t is_clone;
	boolean_t embedok;
	boolean_t large_block_ok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t resumeobj;
	uint64_t resumeoff;
	uint64_t saved_guid;
	zfs_bookmark_phys_t *redactbook;
	/* Stream output params */
	dmu_send_outparams_t *dso;

	/* Stream progress params */
	offset_t *off;
	int outfd;
	char saved_toname[MAXNAMELEN];
	};

	static int
	setup_featureflags(struct dmu_send_params dspp, objset_t os,
	uint64_t *featureflags)
	{
	dsl_dataset_t *to_ds = dspp->to_ds;
	dsl_pool_t *dp = dspp->dp;
	#ifdef _KERNEL
	if (dmu_objset_type(os) == DMU_OST_ZFS) {
	uint64_t version;
	if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0)
	return (SET_ERROR(EINVAL));

	if (version >= ZPL_VERSION_SA)
	*featureflags \|= DMU_BACKUP_FEATURE_SA_SPILL;
	}
	#endif

	/* raw sends imply large_block_ok */
	if ((dspp->rawok \|\| dspp->large_block_ok) &&
	dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_BLOCKS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LARGE_BLOCKS;
	}

	/* encrypted datasets will not have embedded blocks */
	if ((dspp->embedok \|\| dspp->rawok) && !os->os_encrypted &&
	spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) {
	*featureflags \|= DMU_BACKUP_FEATURE_EMBED_DATA;
	}

	/* raw send implies compressok */
	if (dspp->compressok \|\| dspp->rawok)
	*featureflags \|= DMU_BACKUP_FEATURE_COMPRESSED;

	if (dspp->rawok && os->os_encrypted)
	*featureflags \|= DMU_BACKUP_FEATURE_RAW;

	if ((*featureflags &
	(DMU_BACKUP_FEATURE_EMBED_DATA \| DMU_BACKUP_FEATURE_COMPRESSED \|
	DMU_BACKUP_FEATURE_RAW)) != 0 &&
	spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LZ4;
	}

	/*
	* We specifically do not include DMU_BACKUP_FEATURE_EMBED_DATA here to
	* allow sending ZSTD compressed datasets to a receiver that does not
	* support ZSTD
	*/
	if ((*featureflags &
	(DMU_BACKUP_FEATURE_COMPRESSED \| DMU_BACKUP_FEATURE_RAW)) != 0 &&
	dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_ZSTD_COMPRESS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_ZSTD;
	}

	if (dspp->resumeobj != 0 \|\| dspp->resumeoff != 0) {
	*featureflags \|= DMU_BACKUP_FEATURE_RESUMING;
	}

	if (dspp->redactbook != NULL) {
	*featureflags \|= DMU_BACKUP_FEATURE_REDACTED;
	}

	if (dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_DNODE)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LARGE_DNODE;
	}
	return (0);
	}

	static dmu_replay_record_t *
	create_begin_record(struct dmu_send_params dspp, objset_t os,
	uint64_t featureflags)
	{
	dmu_replay_record_t *drr = kmem_zalloc(sizeof (dmu_replay_record_t),
	KM_SLEEP);
	drr->drr_type = DRR_BEGIN;

	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	dsl_dataset_t *to_ds = dspp->to_ds;

	drrb->drr_magic = DMU_BACKUP_MAGIC;
	drrb->drr_creation_time = dsl_dataset_phys(to_ds)->ds_creation_time;
	drrb->drr_type = dmu_objset_type(os);
	drrb->drr_toguid = dsl_dataset_phys(to_ds)->ds_guid;
	drrb->drr_fromguid = dspp->ancestor_zb.zbm_guid;

	DMU_SET_STREAM_HDRTYPE(drrb->drr_versioninfo, DMU_SUBSTREAM);
	DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, featureflags);

	if (dspp->is_clone)
	drrb->drr_flags \|= DRR_FLAG_CLONE;
	if (dsl_dataset_phys(dspp->to_ds)->ds_flags & DS_FLAG_CI_DATASET)
	drrb->drr_flags \|= DRR_FLAG_CI_DATA;
	if (zfs_send_set_freerecords_bit)
	drrb->drr_flags \|= DRR_FLAG_FREERECORDS;
	drr->drr_u.drr_begin.drr_flags \|= DRR_FLAG_SPILL_BLOCK;

	if (dspp->savedok) {
	drrb->drr_toguid = dspp->saved_guid;
	strlcpy(drrb->drr_toname, dspp->saved_toname,
	sizeof (drrb->drr_toname));
	} else {
	dsl_dataset_name(to_ds, drrb->drr_toname);
	if (!to_ds->ds_is_snapshot) {
	(void) strlcat(drrb->drr_toname, "@--head--",
	sizeof (drrb->drr_toname));
	}
	}
	return (drr);
	}

	static void
	setup_to_thread(struct send_thread_arg to_arg, objset_t to_os,
	dmu_sendstatus_t *dssp, uint64_t fromtxg, boolean_t rawok)
	{
	VERIFY0(bqueue_init(&to_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	to_arg->error_code = 0;
	to_arg->cancel = B_FALSE;
	to_arg->os = to_os;
	to_arg->fromtxg = fromtxg;
	to_arg->flags = TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA;
	if (rawok)
	to_arg->flags \|= TRAVERSE_NO_DECRYPT;
	if (zfs_send_corrupt_data)
	to_arg->flags \|= TRAVERSE_HARD;
	to_arg->num_blocks_visited = &dssp->dss_blocks;
	(void) thread_create(NULL, 0, send_traverse_thread, to_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_from_thread(struct redact_list_thread_arg *from_arg,
	redaction_list_t from_rl, dmu_sendstatus_t dssp)
	{
	VERIFY0(bqueue_init(&from_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	from_arg->error_code = 0;
	from_arg->cancel = B_FALSE;
	from_arg->rl = from_rl;
	from_arg->mark_redact = B_FALSE;
	from_arg->num_blocks_visited = &dssp->dss_blocks;
	/*
	* If from_ds is null, send_traverse_thread just returns success and
	* enqueues an eos marker.
	*/
	(void) thread_create(NULL, 0, redact_list_thread, from_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_redact_list_thread(struct redact_list_thread_arg *rlt_arg,
	struct dmu_send_params dspp, redaction_list_t rl, dmu_sendstatus_t *dssp)
	{
	if (dspp->redactbook == NULL)
	return;

	rlt_arg->cancel = B_FALSE;
	VERIFY0(bqueue_init(&rlt_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	rlt_arg->error_code = 0;
	rlt_arg->mark_redact = B_TRUE;
	rlt_arg->rl = rl;
	rlt_arg->num_blocks_visited = &dssp->dss_blocks;

	(void) thread_create(NULL, 0, redact_list_thread, rlt_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_merge_thread(struct send_merge_thread_arg *smt_arg,
	struct dmu_send_params dspp, struct redact_list_thread_arg from_arg,
	struct send_thread_arg to_arg, struct redact_list_thread_arg rlt_arg,
	objset_t *os)
	{
	VERIFY0(bqueue_init(&smt_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	smt_arg->cancel = B_FALSE;
	smt_arg->error = 0;
	smt_arg->from_arg = from_arg;
	smt_arg->to_arg = to_arg;
	if (dspp->redactbook != NULL)
	smt_arg->redact_arg = rlt_arg;

	smt_arg->os = os;
	(void) thread_create(NULL, 0, send_merge_thread, smt_arg, 0, curproc,
	TS_RUN, minclsyspri);
	}

	static void
	setup_reader_thread(struct send_reader_thread_arg *srt_arg,
	struct dmu_send_params dspp, struct send_merge_thread_arg smt_arg,
	uint64_t featureflags)
	{
	VERIFY0(bqueue_init(&srt_arg->q, zfs_send_queue_ff,
	MAX(zfs_send_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	srt_arg->smta = smt_arg;
	srt_arg->issue_reads = !dspp->dso->dso_dryrun;
	srt_arg->featureflags = featureflags;
	(void) thread_create(NULL, 0, send_reader_thread, srt_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static int
	setup_resume_points(struct dmu_send_params *dspp,
	struct send_thread_arg to_arg, struct redact_list_thread_arg from_arg,
	struct redact_list_thread_arg *rlt_arg,
	struct send_merge_thread_arg smt_arg, boolean_t resuming, objset_t os,
	redaction_list_t redact_rl, nvlist_t nvl)
	{
	(void) smt_arg;
	dsl_dataset_t *to_ds = dspp->to_ds;
	int err = 0;

	uint64_t obj = 0;
	uint64_t blkid = 0;
	if (resuming) {
	obj = dspp->resumeobj;
	dmu_object_info_t to_doi;
	err = dmu_object_info(os, obj, &to_doi);
	if (err != 0)
	return (err);

	blkid = dspp->resumeoff / to_doi.doi_data_block_size;
	}
	/*
	* If we're resuming a redacted send, we can skip to the appropriate
	* point in the redaction bookmark by binary searching through it.
	*/
	if (redact_rl != NULL) {
	SET_BOOKMARK(&rlt_arg->resume, to_ds->ds_object, obj, 0, blkid);
	}

	SET_BOOKMARK(&to_arg->resume, to_ds->ds_object, obj, 0, blkid);
	if (nvlist_exists(nvl, BEGINNV_REDACT_FROM_SNAPS)) {
	uint64_t objset = dspp->ancestor_zb.zbm_redaction_obj;
	/*
	* Note: If the resume point is in an object whose
	* blocksize is different in the from vs to snapshots,
	* we will have divided by the "wrong" blocksize.
	* However, in this case fromsnap's send_cb() will
	* detect that the blocksize has changed and therefore
	* ignore this object.
	*
	* If we're resuming a send from a redaction bookmark,
	* we still cannot accidentally suggest blocks behind
	* the to_ds. In addition, we know that any blocks in
	* the object in the to_ds will have to be sent, since
	* the size changed. Therefore, we can't cause any harm
	* this way either.
	*/
	SET_BOOKMARK(&from_arg->resume, objset, obj, 0, blkid);
	}
	if (resuming) {
	fnvlist_add_uint64(nvl, BEGINNV_RESUME_OBJECT, dspp->resumeobj);
	fnvlist_add_uint64(nvl, BEGINNV_RESUME_OFFSET, dspp->resumeoff);
	}
	return (0);
	}

	static dmu_sendstatus_t *
	setup_send_progress(struct dmu_send_params *dspp)
	{
	dmu_sendstatus_t dssp = kmem_zalloc(sizeof (dssp), KM_SLEEP);
	dssp->dss_outfd = dspp->outfd;
	dssp->dss_off = dspp->off;
	dssp->dss_proc = curproc;
	mutex_enter(&dspp->to_ds->ds_sendstream_lock);
	list_insert_head(&dspp->to_ds->ds_sendstreams, dssp);
	mutex_exit(&dspp->to_ds->ds_sendstream_lock);
	return (dssp);
	}

	/*
	* Actually do the bulk of the work in a zfs send.
	*
	* The idea is that we want to do a send from ancestor_zb to to_ds. We also
	* want to not send any data that has been modified by all the datasets in
	* redactsnaparr, and store the list of blocks that are redacted in this way in
	* a bookmark named redactbook, created on the to_ds. We do this by creating
	* several worker threads, whose function is described below.
	*
	* There are three cases.
	* The first case is a redacted zfs send. In this case there are 5 threads.
	* The first thread is the to_ds traversal thread: it calls dataset_traverse on
	* the to_ds and finds all the blocks that have changed since ancestor_zb (if
	* it's a full send, that's all blocks in the dataset). It then sends those
	* blocks on to the send merge thread. The redact list thread takes the data
	* from the redaction bookmark and sends those blocks on to the send merge
	* thread. The send merge thread takes the data from the to_ds traversal
	* thread, and combines it with the redaction records from the redact list
	* thread. If a block appears in both the to_ds's data and the redaction data,
	* the send merge thread will mark it as redacted and send it on to the prefetch
	* thread. Otherwise, the send merge thread will send the block on to the
	* prefetch thread unchanged. The prefetch thread will issue prefetch reads for
	* any data that isn't redacted, and then send the data on to the main thread.
	* The main thread behaves the same as in a normal send case, issuing demand
	* reads for data blocks and sending out records over the network
	*
	* The graphic below diagrams the flow of data in the case of a redacted zfs
	* send. Each box represents a thread, and each line represents the flow of
	* data.
	*
	* Records from the \|
	* redaction bookmark \|
	* +--------------------+ \| +---------------------------+
	* \| \| v \| Send Merge Thread \|
	* \| Redact List Thread +----------> Apply redaction marks to \|
	* \| \| \| records as specified by \|
	* +--------------------+ \| redaction ranges \|
	* +----^---------------+------+
	* \| \| Merged data
	* \| \|
	* \| +------------v--------+
	* \| \| Prefetch Thread \|
	* +--------------------+ \| \| Issues prefetch \|
	* \| to_ds Traversal \| \| \| reads of data blocks\|
	* \| Thread (finds +---------------+ +------------+--------+
	* \| candidate blocks) \| Blocks modified \| Prefetched data
	* +--------------------+ by to_ds since \|
	* ancestor_zb +------------v----+
	* \| Main Thread \| File Descriptor
	* \| Sends data over +->(to zfs receive)
	* \| wire \|
	* +-----------------+
	*
	* The second case is an incremental send from a redaction bookmark. The to_ds
	* traversal thread and the main thread behave the same as in the redacted
	* send case. The new thread is the from bookmark traversal thread. It
	* iterates over the redaction list in the redaction bookmark, and enqueues
	* records for each block that was redacted in the original send. The send
	* merge thread now has to merge the data from the two threads. For details
	* about that process, see the header comment of send_merge_thread(). Any data
	* it decides to send on will be prefetched by the prefetch thread. Note that
	* you can perform a redacted send from a redaction bookmark; in that case,
	* the data flow behaves very similarly to the flow in the redacted send case,
	* except with the addition of the bookmark traversal thread iterating over the
	* redaction bookmark. The send_merge_thread also has to take on the
	* responsibility of merging the redact list thread's records, the bookmark
	* traversal thread's records, and the to_ds records.
	*
	* +---------------------+
	* \| \|
	* \| Redact List Thread +--------------+
	* \| \| \|
	* +---------------------+ \|
	* Blocks in redaction list \| Ranges modified by every secure snap
	* of from bookmark \| (or EOS if not readcted)
	* \|
	* +---------------------+ \| +----v----------------------+
	* \| bookmark Traversal \| v \| Send Merge Thread \|
	* \| Thread (finds +---------> Merges bookmark, rlt, and \|
	* \| candidate blocks) \| \| to_ds send records \|
	* +---------------------+ +----^---------------+------+
	* \| \| Merged data
	* \| +------------v--------+
	* \| \| Prefetch Thread \|
	* +--------------------+ \| \| Issues prefetch \|
	* \| to_ds Traversal \| \| \| reads of data blocks\|
	* \| Thread (finds +---------------+ +------------+--------+
	* \| candidate blocks) \| Blocks modified \| Prefetched data
	* +--------------------+ by to_ds since +------------v----+
	* ancestor_zb \| Main Thread \| File Descriptor
	* \| Sends data over +->(to zfs receive)
	* \| wire \|
	* +-----------------+
	*
	* The final case is a simple zfs full or incremental send. The to_ds traversal
	* thread behaves the same as always. The redact list thread is never started.
	* The send merge thread takes all the blocks that the to_ds traversal thread
	* sends it, prefetches the data, and sends the blocks on to the main thread.
	* The main thread sends the data over the wire.
	*
	* To keep performance acceptable, we want to prefetch the data in the worker
	* threads. While the to_ds thread could simply use the TRAVERSE_PREFETCH
	* feature built into traverse_dataset, the combining and deletion of records
	* due to redaction and sends from redaction bookmarks mean that we could
	* issue many unnecessary prefetches. As a result, we only prefetch data
	* after we've determined that the record is not going to be redacted. To
	* prevent the prefetching from getting too far ahead of the main thread, the
	* blocking queues that are used for communication are capped not by the
	* number of entries in the queue, but by the sum of the size of the
	* prefetches associated with them. The limit on the amount of data that the
	* thread can prefetch beyond what the main thread has reached is controlled
	* by the global variable zfs_send_queue_length. In addition, to prevent poor
	* performance in the beginning of a send, we also limit the distance ahead
	* that the traversal threads can be. That distance is controlled by the
	* zfs_send_no_prefetch_queue_length tunable.
	*
	* Note: Releases dp using the specified tag.
	*/
	static int
	dmu_send_impl(struct dmu_send_params *dspp)
	{
	objset_t *os;
	dmu_replay_record_t *drr;
	dmu_sendstatus_t *dssp;
	dmu_send_cookie_t dsc = {0};
	int err;
	uint64_t fromtxg = dspp->ancestor_zb.zbm_creation_txg;
	uint64_t featureflags = 0;
	struct redact_list_thread_arg *from_arg;
	struct send_thread_arg *to_arg;
	struct redact_list_thread_arg *rlt_arg;
	struct send_merge_thread_arg *smt_arg;
	struct send_reader_thread_arg *srt_arg;
	struct send_range *range;
	redaction_list_t *from_rl = NULL;
	redaction_list_t *redact_rl = NULL;
	boolean_t resuming = (dspp->resumeobj != 0 \|\| dspp->resumeoff != 0);
	boolean_t book_resuming = resuming;

	dsl_dataset_t *to_ds = dspp->to_ds;
	zfs_bookmark_phys_t *ancestor_zb = &dspp->ancestor_zb;
	dsl_pool_t *dp = dspp->dp;
	const void *tag = dspp->tag;

	err = dmu_objset_from_ds(to_ds, &os);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	/*
	* If this is a non-raw send of an encrypted ds, we can ensure that
	* the objset_phys_t is authenticated. This is safe because this is
	* either a snapshot or we have owned the dataset, ensuring that
	* it can't be modified.
	*/
	if (!dspp->rawok && os->os_encrypted &&
	arc_is_unauthenticated(os->os_phys_buf)) {
	zbookmark_phys_t zb;

	SET_BOOKMARK(&zb, to_ds->ds_object, ZB_ROOT_OBJECT,
	ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
	err = arc_untransform(os->os_phys_buf, os->os_spa,
	&zb, B_FALSE);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	ASSERT0(arc_is_unauthenticated(os->os_phys_buf));
	}

	if ((err = setup_featureflags(dspp, os, &featureflags)) != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	/*
	* If we're doing a redacted send, hold the bookmark's redaction list.
	*/
	if (dspp->redactbook != NULL) {
	err = dsl_redaction_list_hold_obj(dp,
	dspp->redactbook->zbm_redaction_obj, FTAG,
	&redact_rl);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (SET_ERROR(EINVAL));
	}
	dsl_redaction_list_long_hold(dp, redact_rl, FTAG);
	}

	/*
	* If we're sending from a redaction bookmark, hold the redaction list
	* so that we can consider sending the redacted blocks.
	*/
	if (ancestor_zb->zbm_redaction_obj != 0) {
	err = dsl_redaction_list_hold_obj(dp,
	ancestor_zb->zbm_redaction_obj, FTAG, &from_rl);
	if (err != 0) {
	if (redact_rl != NULL) {
	dsl_redaction_list_long_rele(redact_rl, FTAG);
	dsl_redaction_list_rele(redact_rl, FTAG);
	}
	dsl_pool_rele(dp, tag);
	return (SET_ERROR(EINVAL));
	}
	dsl_redaction_list_long_hold(dp, from_rl, FTAG);
	}

	dsl_dataset_long_hold(to_ds, FTAG);

	from_arg = kmem_zalloc(sizeof (*from_arg), KM_SLEEP);
	to_arg = kmem_zalloc(sizeof (*to_arg), KM_SLEEP);
	rlt_arg = kmem_zalloc(sizeof (*rlt_arg), KM_SLEEP);
	smt_arg = kmem_zalloc(sizeof (*smt_arg), KM_SLEEP);
	srt_arg = kmem_zalloc(sizeof (*srt_arg), KM_SLEEP);

	drr = create_begin_record(dspp, os, featureflags);
	dssp = setup_send_progress(dspp);

	dsc.dsc_drr = drr;
	dsc.dsc_dso = dspp->dso;
	dsc.dsc_os = os;
	dsc.dsc_off = dspp->off;
	dsc.dsc_toguid = dsl_dataset_phys(to_ds)->ds_guid;
	dsc.dsc_fromtxg = fromtxg;
	dsc.dsc_pending_op = PENDING_NONE;
	dsc.dsc_featureflags = featureflags;
	dsc.dsc_resume_object = dspp->resumeobj;
	dsc.dsc_resume_offset = dspp->resumeoff;

	dsl_pool_rele(dp, tag);

	void *payload = NULL;
	size_t payload_len = 0;
	nvlist_t *nvl = fnvlist_alloc();

	/*
	* If we're doing a redacted send, we include the snapshots we're
	* redacted with respect to so that the target system knows what send
	* streams can be correctly received on top of this dataset. If we're
	* instead sending a redacted dataset, we include the snapshots that the
	* dataset was created with respect to.
	*/
	if (dspp->redactbook != NULL) {
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS,
	redact_rl->rl_phys->rlp_snaps,
	redact_rl->rl_phys->rlp_num_snaps);
	} else if (dsl_dataset_feature_is_active(to_ds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	uint64_t *tods_guids;
	uint64_t length;
	VERIFY(dsl_dataset_get_uint64_array_feature(to_ds,
	SPA_FEATURE_REDACTED_DATASETS, &length, &tods_guids));
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS, tods_guids,
	length);
	}

	/*
	* If we're sending from a redaction bookmark, then we should retrieve
	* the guids of that bookmark so we can send them over the wire.
	*/
	if (from_rl != NULL) {
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
	from_rl->rl_phys->rlp_snaps,
	from_rl->rl_phys->rlp_num_snaps);
	}

	/*
	* If the snapshot we're sending from is redacted, include the redaction
	* list in the stream.
	*/
	if (dspp->numfromredactsnaps != NUM_SNAPS_NOT_REDACTED) {
	ASSERT3P(from_rl, ==, NULL);
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
	dspp->fromredactsnaps, (uint_t)dspp->numfromredactsnaps);
	if (dspp->numfromredactsnaps > 0) {
	kmem_free(dspp->fromredactsnaps,
	dspp->numfromredactsnaps * sizeof (uint64_t));
	dspp->fromredactsnaps = NULL;
	}
	}

	if (resuming \|\| book_resuming) {
	err = setup_resume_points(dspp, to_arg, from_arg,
	rlt_arg, smt_arg, resuming, os, redact_rl, nvl);
	if (err != 0)
	goto out;
	}

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	uint64_t ivset_guid = (ancestor_zb != NULL) ?
	ancestor_zb->zbm_ivset_guid : 0;
	nvlist_t *keynvl = NULL;
	ASSERT(os->os_encrypted);

	err = dsl_crypto_populate_key_nvlist(os, ivset_guid,
	&keynvl);
	if (err != 0) {
	fnvlist_free(nvl);
	goto out;
	}

	fnvlist_add_nvlist(nvl, "crypt_keydata", keynvl);
	fnvlist_free(keynvl);
	}

	if (!nvlist_empty(nvl)) {
	payload = fnvlist_pack(nvl, &payload_len);
	drr->drr_payloadlen = payload_len;
	}

	fnvlist_free(nvl);
	err = dump_record(&dsc, payload, payload_len);
	fnvlist_pack_free(payload, payload_len);
	if (err != 0) {
	err = dsc.dsc_err;
	goto out;
	}

	setup_to_thread(to_arg, os, dssp, fromtxg, dspp->rawok);
	setup_from_thread(from_arg, from_rl, dssp);
	setup_redact_list_thread(rlt_arg, dspp, redact_rl, dssp);
	setup_merge_thread(smt_arg, dspp, from_arg, to_arg, rlt_arg, os);
	setup_reader_thread(srt_arg, dspp, smt_arg, featureflags);

	range = bqueue_dequeue(&srt_arg->q);
	while (err == 0 && !range->eos_marker) {
	err = do_dump(&dsc, range);
	range = get_next_range(&srt_arg->q, range);
	if (issig(JUSTLOOKING) && issig(FORREAL))
	err = SET_ERROR(EINTR);
	}

	/*
	* If we hit an error or are interrupted, cancel our worker threads and
	* clear the queue of any pending records. The threads will pass the
	* cancel up the tree of worker threads, and each one will clean up any
	* pending records before exiting.
	*/
	if (err != 0) {
	srt_arg->cancel = B_TRUE;
	while (!range->eos_marker) {
	range = get_next_range(&srt_arg->q, range);
	}
	}
	range_free(range);

	bqueue_destroy(&srt_arg->q);
	bqueue_destroy(&smt_arg->q);
	if (dspp->redactbook != NULL)
	bqueue_destroy(&rlt_arg->q);
	bqueue_destroy(&to_arg->q);
	bqueue_destroy(&from_arg->q);

	if (err == 0 && srt_arg->error != 0)
	err = srt_arg->error;

	if (err != 0)
	goto out;

	if (dsc.dsc_pending_op != PENDING_NONE)
	if (dump_record(&dsc, NULL, 0) != 0)
	err = SET_ERROR(EINTR);

	if (err != 0) {
	if (err == EINTR && dsc.dsc_err != 0)
	err = dsc.dsc_err;
	goto out;
	}

	/*
	* Send the DRR_END record if this is not a saved stream.
	* Otherwise, the omitted DRR_END record will signal to
	* the receive side that the stream is incomplete.
	*/
	if (!dspp->savedok) {
	memset(drr, 0, sizeof (dmu_replay_record_t));
	drr->drr_type = DRR_END;
	drr->drr_u.drr_end.drr_checksum = dsc.dsc_zc;
	drr->drr_u.drr_end.drr_toguid = dsc.dsc_toguid;

	if (dump_record(&dsc, NULL, 0) != 0)
	err = dsc.dsc_err;
	}
	out:
	mutex_enter(&to_ds->ds_sendstream_lock);
	list_remove(&to_ds->ds_sendstreams, dssp);
	mutex_exit(&to_ds->ds_sendstream_lock);

	VERIFY(err != 0 \|\| (dsc.dsc_sent_begin &&
	(dsc.dsc_sent_end \|\| dspp->savedok)));

	kmem_free(drr, sizeof (dmu_replay_record_t));
	kmem_free(dssp, sizeof (dmu_sendstatus_t));
	kmem_free(from_arg, sizeof (*from_arg));
	kmem_free(to_arg, sizeof (*to_arg));
	kmem_free(rlt_arg, sizeof (*rlt_arg));
	kmem_free(smt_arg, sizeof (*smt_arg));
	kmem_free(srt_arg, sizeof (*srt_arg));

	dsl_dataset_long_rele(to_ds, FTAG);
	if (from_rl != NULL) {
	dsl_redaction_list_long_rele(from_rl, FTAG);
	dsl_redaction_list_rele(from_rl, FTAG);
	}
	if (redact_rl != NULL) {
	dsl_redaction_list_long_rele(redact_rl, FTAG);
	dsl_redaction_list_rele(redact_rl, FTAG);
	}

	return (err);
	}

	int
	dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
	boolean_t embedok, boolean_t large_block_ok, boolean_t compressok,
	boolean_t rawok, boolean_t savedok, int outfd, offset_t *off,
	dmu_send_outparams_t *dsop)
	{
	int err;
	dsl_dataset_t *fromds;
	ds_hold_flags_t dsflags;
	struct dmu_send_params dspp = {0};
	dspp.embedok = embedok;
	dspp.large_block_ok = large_block_ok;
	dspp.compressok = compressok;
	dspp.outfd = outfd;
	dspp.off = off;
	dspp.dso = dsop;
	dspp.tag = FTAG;
	dspp.rawok = rawok;
	dspp.savedok = savedok;

	dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	err = dsl_pool_hold(pool, FTAG, &dspp.dp);
	if (err != 0)
	return (err);

	err = dsl_dataset_hold_obj_flags(dspp.dp, tosnap, dsflags, FTAG,
	&dspp.to_ds);
	if (err != 0) {
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}

	if (fromsnap != 0) {
	err = dsl_dataset_hold_obj_flags(dspp.dp, fromsnap, dsflags,
	FTAG, &fromds);
	if (err != 0) {
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}
	dspp.ancestor_zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
	dspp.ancestor_zb.zbm_creation_txg =
	dsl_dataset_phys(fromds)->ds_creation_txg;
	dspp.ancestor_zb.zbm_creation_time =
	dsl_dataset_phys(fromds)->ds_creation_time;

	if (dsl_dataset_is_zapified(fromds)) {
	(void) zap_lookup(dspp.dp->dp_meta_objset,
	fromds->ds_object, DS_FIELD_IVSET_GUID, 8, 1,
	&dspp.ancestor_zb.zbm_ivset_guid);
	}

	/* See dmu_send for the reasons behind this. */
	uint64_t *fromredact;

	if (!dsl_dataset_get_uint64_array_feature(fromds,
	SPA_FEATURE_REDACTED_DATASETS,
	&dspp.numfromredactsnaps,
	&fromredact)) {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	} else if (dspp.numfromredactsnaps > 0) {
	uint64_t size = dspp.numfromredactsnaps *
	sizeof (uint64_t);
	dspp.fromredactsnaps = kmem_zalloc(size, KM_SLEEP);
	memcpy(dspp.fromredactsnaps, fromredact, size);
	}

	boolean_t is_before =
	dsl_dataset_is_before(dspp.to_ds, fromds, 0);
	dspp.is_clone = (dspp.to_ds->ds_dir !=
	fromds->ds_dir);
	dsl_dataset_rele(fromds, FTAG);
	if (!is_before) {
	dsl_pool_rele(dspp.dp, FTAG);
	err = SET_ERROR(EXDEV);
	} else {
	err = dmu_send_impl(&dspp);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dmu_send_impl(&dspp);
	}
	dsl_dataset_rele(dspp.to_ds, FTAG);
	return (err);
	}

	int
	dmu_send(const char tosnap, const char fromsnap, boolean_t embedok,
	boolean_t large_block_ok, boolean_t compressok, boolean_t rawok,
	boolean_t savedok, uint64_t resumeobj, uint64_t resumeoff,
	const char redactbook, int outfd, offset_t off,
	dmu_send_outparams_t *dsop)
	{
	int err = 0;
	ds_hold_flags_t dsflags;
	boolean_t owned = B_FALSE;
	dsl_dataset_t *fromds = NULL;
	zfs_bookmark_phys_t book = {0};
	struct dmu_send_params dspp = {0};

	dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	dspp.tosnap = tosnap;
	dspp.embedok = embedok;
	dspp.large_block_ok = large_block_ok;
	dspp.compressok = compressok;
	dspp.outfd = outfd;
	dspp.off = off;
	dspp.dso = dsop;
	dspp.tag = FTAG;
	dspp.resumeobj = resumeobj;
	dspp.resumeoff = resumeoff;
	dspp.rawok = rawok;
	dspp.savedok = savedok;

	if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL)
	return (SET_ERROR(EINVAL));

	err = dsl_pool_hold(tosnap, FTAG, &dspp.dp);
	if (err != 0)
	return (err);

	if (strchr(tosnap, '@') == NULL && spa_writeable(dspp.dp->dp_spa)) {
	/*
	* We are sending a filesystem or volume. Ensure
	* that it doesn't change by owning the dataset.
	*/

	if (savedok) {
	/*
	* We are looking for the dataset that represents the
	* partially received send stream. If this stream was
	* received as a new snapshot of an existing dataset,
	* this will be saved in a hidden clone named
	* "<pool>/<dataset>/%recv". Otherwise, the stream
	* will be saved in the live dataset itself. In
	* either case we need to use dsl_dataset_own_force()
	* because the stream is marked as inconsistent,
	* which would normally make it unavailable to be
	* owned.
	*/
	char *name = kmem_asprintf("%s/%s", tosnap,
	recv_clone_name);
	err = dsl_dataset_own_force(dspp.dp, name, dsflags,
	FTAG, &dspp.to_ds);
	if (err == ENOENT) {
	err = dsl_dataset_own_force(dspp.dp, tosnap,
	dsflags, FTAG, &dspp.to_ds);
	}

	if (err == 0) {
	err = zap_lookup(dspp.dp->dp_meta_objset,
	dspp.to_ds->ds_object,
	DS_FIELD_RESUME_TOGUID, 8, 1,
	&dspp.saved_guid);
	}

	if (err == 0) {
	err = zap_lookup(dspp.dp->dp_meta_objset,
	dspp.to_ds->ds_object,
	DS_FIELD_RESUME_TONAME, 1,
	sizeof (dspp.saved_toname),
	dspp.saved_toname);
	}
	if (err != 0)
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);

	kmem_strfree(name);
	} else {
	err = dsl_dataset_own(dspp.dp, tosnap, dsflags,
	FTAG, &dspp.to_ds);
	}
	owned = B_TRUE;
	} else {
	err = dsl_dataset_hold_flags(dspp.dp, tosnap, dsflags, FTAG,
	&dspp.to_ds);
	}

	if (err != 0) {
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}

	if (redactbook != NULL) {
	char path[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(path, tosnap, sizeof (path));
	char *at = strchr(path, '@');
	if (at == NULL) {
	err = EINVAL;
	} else {
	(void) snprintf(at, sizeof (path) - (at - path), "#%s",
	redactbook);
	err = dsl_bookmark_lookup(dspp.dp, path,
	NULL, &book);
	dspp.redactbook = &book;
	}
	}

	if (err != 0) {
	dsl_pool_rele(dspp.dp, FTAG);
	if (owned)
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
	else
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	return (err);
	}

	if (fromsnap != NULL) {
	zfs_bookmark_phys_t *zb = &dspp.ancestor_zb;
	int fsnamelen;
	if (strpbrk(tosnap, "@#") != NULL)
	fsnamelen = strpbrk(tosnap, "@#") - tosnap;
	else
	fsnamelen = strlen(tosnap);

	/*
	* If the fromsnap is in a different filesystem, then
	* mark the send stream as a clone.
	*/
	if (strncmp(tosnap, fromsnap, fsnamelen) != 0 \|\|
	(fromsnap[fsnamelen] != '@' &&
	fromsnap[fsnamelen] != '#')) {
	dspp.is_clone = B_TRUE;
	}

	if (strchr(fromsnap, '@') != NULL) {
	err = dsl_dataset_hold(dspp.dp, fromsnap, FTAG,
	&fromds);

	if (err != 0) {
	ASSERT3P(fromds, ==, NULL);
	} else {
	/*
	* We need to make a deep copy of the redact
	* snapshots of the from snapshot, because the
	* array will be freed when we evict from_ds.
	*/
	uint64_t *fromredact;
	if (!dsl_dataset_get_uint64_array_feature(
	fromds, SPA_FEATURE_REDACTED_DATASETS,
	&dspp.numfromredactsnaps,
	&fromredact)) {
	dspp.numfromredactsnaps =
	NUM_SNAPS_NOT_REDACTED;
	} else if (dspp.numfromredactsnaps > 0) {
	uint64_t size =
	dspp.numfromredactsnaps *
	sizeof (uint64_t);
	dspp.fromredactsnaps = kmem_zalloc(size,
	KM_SLEEP);
	memcpy(dspp.fromredactsnaps, fromredact,
	size);
	}
	if (!dsl_dataset_is_before(dspp.to_ds, fromds,
	0)) {
	err = SET_ERROR(EXDEV);
	} else {
	zb->zbm_creation_txg =
	dsl_dataset_phys(fromds)->
	ds_creation_txg;
	zb->zbm_creation_time =
	dsl_dataset_phys(fromds)->
	ds_creation_time;
	zb->zbm_guid =
	dsl_dataset_phys(fromds)->ds_guid;
	zb->zbm_redaction_obj = 0;

	if (dsl_dataset_is_zapified(fromds)) {
	(void) zap_lookup(
	dspp.dp->dp_meta_objset,
	fromds->ds_object,
	DS_FIELD_IVSET_GUID, 8, 1,
	&zb->zbm_ivset_guid);
	}
	}
	dsl_dataset_rele(fromds, FTAG);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dsl_bookmark_lookup(dspp.dp, fromsnap, dspp.to_ds,
	zb);
	if (err == EXDEV && zb->zbm_redaction_obj != 0 &&
	zb->zbm_guid ==
	dsl_dataset_phys(dspp.to_ds)->ds_guid)
	err = 0;
	}

	if (err == 0) {
	/* dmu_send_impl will call dsl_pool_rele for us. */
	err = dmu_send_impl(&dspp);
	} else {
	dsl_pool_rele(dspp.dp, FTAG);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dmu_send_impl(&dspp);
	}
	if (owned)
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
	else
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	return (err);
	}

	static int
	dmu_adjust_send_estimate_for_indirects(dsl_dataset_t *ds, uint64_t uncompressed,
	uint64_t compressed, boolean_t stream_compressed, uint64_t *sizep)
	{
	int err = 0;
	uint64_t size;
	/*
	* Assume that space (both on-disk and in-stream) is dominated by
	* data. We will adjust for indirect blocks and the copies property,
	* but ignore per-object space used (eg, dnodes and DRR_OBJECT records).
	*/

	uint64_t recordsize;
	uint64_t record_count;
	objset_t *os;
	VERIFY0(dmu_objset_from_ds(ds, &os));

	/* Assume all (uncompressed) blocks are recordsize. */
	if (zfs_override_estimate_recordsize != 0) {
	recordsize = zfs_override_estimate_recordsize;
	} else if (os->os_phys->os_type == DMU_OST_ZVOL) {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &recordsize);
	} else {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_RECORDSIZE), &recordsize);
	}
	if (err != 0)
	return (err);
	record_count = uncompressed / recordsize;

	/*
	* If we're estimating a send size for a compressed stream, use the
	* compressed data size to estimate the stream size. Otherwise, use the
	* uncompressed data size.
	*/
	size = stream_compressed ? compressed : uncompressed;

	/*
	* Subtract out approximate space used by indirect blocks.
	* Assume most space is used by data blocks (non-indirect, non-dnode).
	* Assume no ditto blocks or internal fragmentation.
	*
	* Therefore, space used by indirect blocks is sizeof(blkptr_t) per
	* block.
	*/
	size -= record_count * sizeof (blkptr_t);

	/* Add in the space for the record associated with each block. */
	size += record_count * sizeof (dmu_replay_record_t);

	*sizep = size;

	return (0);
	}

	int
	dmu_send_estimate_fast(dsl_dataset_t origds, dsl_dataset_t fromds,
	zfs_bookmark_phys_t *frombook, boolean_t stream_compressed,
	boolean_t saved, uint64_t *sizep)
	{
	int err;
	dsl_dataset_t *ds = origds;
	uint64_t uncomp, comp;

	ASSERT(dsl_pool_config_held(origds->ds_dir->dd_pool));
	ASSERT(fromds == NULL \|\| frombook == NULL);

	/*
	* If this is a saved send we may actually be sending
	* from the %recv clone used for resuming.
	*/
	if (saved) {
	objset_t *mos = origds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t guid;
	char dsname[ZFS_MAX_DATASET_NAME_LEN + 6];

	dsl_dataset_name(origds, dsname);
	(void) strcat(dsname, "/");
	- (void) strcat(dsname, recv_clone_name);
	+ (void) strlcat(dsname, recv_clone_name,
	+ sizeof (dsname) - strlen(dsname));

	err = dsl_dataset_hold(origds->ds_dir->dd_pool,
	dsname, FTAG, &ds);
	if (err != ENOENT && err != 0) {
	return (err);
	} else if (err == ENOENT) {
	ds = origds;
	}

	/* check that this dataset has partially received data */
	err = zap_lookup(mos, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, 8, 1, &guid);
	if (err != 0) {
	err = SET_ERROR(err == ENOENT ? EINVAL : err);
	goto out;
	}

	err = zap_lookup(mos, ds->ds_object,
	DS_FIELD_RESUME_TONAME, 1, sizeof (dsname), dsname);
	if (err != 0) {
	err = SET_ERROR(err == ENOENT ? EINVAL : err);
	goto out;
	}
	}

	/* tosnap must be a snapshot or the target of a saved send */
	if (!ds->ds_is_snapshot && ds == origds)
	return (SET_ERROR(EINVAL));

	if (fromds != NULL) {
	uint64_t used;
	if (!fromds->ds_is_snapshot) {
	err = SET_ERROR(EINVAL);
	goto out;
	}

	if (!dsl_dataset_is_before(ds, fromds, 0)) {
	err = SET_ERROR(EXDEV);
	goto out;
	}

	err = dsl_dataset_space_written(fromds, ds, &used, &comp,
	&uncomp);
	if (err != 0)
	goto out;
	} else if (frombook != NULL) {
	uint64_t used;
	err = dsl_dataset_space_written_bookmark(frombook, ds, &used,
	&comp, &uncomp);
	if (err != 0)
	goto out;
	} else {
	uncomp = dsl_dataset_phys(ds)->ds_uncompressed_bytes;
	comp = dsl_dataset_phys(ds)->ds_compressed_bytes;
	}

	err = dmu_adjust_send_estimate_for_indirects(ds, uncomp, comp,
	stream_compressed, sizep);
	/*
	* Add the size of the BEGIN and END records to the estimate.
	*/
	sizep += 2 sizeof (dmu_replay_record_t);

	out:
	if (ds != origds)
	dsl_dataset_rele(ds, FTAG);
	return (err);
	}

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, corrupt_data, INT, ZMOD_RW,
	"Allow sending corrupt data");

	-ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_length, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_length, UINT, ZMOD_RW,
	"Maximum send queue length");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, unmodified_spill_blocks, INT, ZMOD_RW,
	"Send unmodified spill blocks");

	-ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_length, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_length, UINT, ZMOD_RW,
	"Maximum send queue length for non-prefetch queues");

	-ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_ff, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_ff, UINT, ZMOD_RW,
	"Send queue fill fraction");

	-ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_ff, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_ff, UINT, ZMOD_RW,
	"Send queue fill fraction for non-prefetch queues");

	-ZFS_MODULE_PARAM(zfs_send, zfs_, override_estimate_recordsize, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_send, zfs_, override_estimate_recordsize, UINT, ZMOD_RW,
	"Override block size estimate with fixed size");
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_traverse.c b/sys/contrib/openzfs/module/zfs/dmu_traverse.c
	index 44b3c9fc5c9d..2ed75640f68d 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_traverse.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_traverse.c
	@@ -1,826 +1,826 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_pool.h>
	#include <sys/dnode.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/dmu_impl.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/callb.h>
	#include <sys/zfeature.h>

	static int32_t zfs_pd_bytes_max = 50 * 1024 * 1024; /* 50MB */
	static int32_t send_holes_without_birth_time = 1;
	-static int32_t zfs_traverse_indirect_prefetch_limit = 32;
	+static uint_t zfs_traverse_indirect_prefetch_limit = 32;

	typedef struct prefetch_data {
	kmutex_t pd_mtx;
	kcondvar_t pd_cv;
	int32_t pd_bytes_fetched;
	int pd_flags;
	boolean_t pd_cancel;
	boolean_t pd_exited;
	zbookmark_phys_t pd_resume;
	} prefetch_data_t;

	typedef struct traverse_data {
	spa_t *td_spa;
	uint64_t td_objset;
	blkptr_t *td_rootbp;
	uint64_t td_min_txg;
	zbookmark_phys_t *td_resume;
	int td_flags;
	prefetch_data_t *td_pfd;
	boolean_t td_paused;
	uint64_t td_hole_birth_enabled_txg;
	blkptr_cb_t *td_func;
	void *td_arg;
	boolean_t td_realloc_possible;
	} traverse_data_t;

	static int traverse_dnode(traverse_data_t td, const blkptr_t bp,
	const dnode_phys_t *dnp, uint64_t objset, uint64_t object);
	static void prefetch_dnode_metadata(traverse_data_t td, const dnode_phys_t ,
	uint64_t objset, uint64_t object);

	static int
	traverse_zil_block(zilog_t zilog, const blkptr_t bp, void *arg,
	uint64_t claim_txg)
	{
	traverse_data_t *td = arg;
	zbookmark_phys_t zb;

	if (BP_IS_HOLE(bp))
	return (0);

	if (claim_txg == 0 && bp->blk_birth >= spa_min_claim_txg(td->td_spa))
	return (-1);

	SET_BOOKMARK(&zb, td->td_objset, ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
	bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);

	(void) td->td_func(td->td_spa, zilog, bp, &zb, NULL, td->td_arg);

	return (0);
	}

	static int
	traverse_zil_record(zilog_t zilog, const lr_t lrc, void *arg,
	uint64_t claim_txg)
	{
	traverse_data_t *td = arg;

	if (lrc->lrc_txtype == TX_WRITE) {
	lr_write_t lr = (lr_write_t )lrc;
	blkptr_t *bp = &lr->lr_blkptr;
	zbookmark_phys_t zb;

	if (BP_IS_HOLE(bp))
	return (0);

	if (claim_txg == 0 \|\| bp->blk_birth < claim_txg)
	return (0);

	SET_BOOKMARK(&zb, td->td_objset, lr->lr_foid,
	ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp));

	(void) td->td_func(td->td_spa, zilog, bp, &zb, NULL,
	td->td_arg);
	}
	return (0);
	}

	static void
	traverse_zil(traverse_data_t td, zil_header_t zh)
	{
	uint64_t claim_txg = zh->zh_claim_txg;

	/*
	* We only want to visit blocks that have been claimed but not yet
	* replayed; plus blocks that are already stable in read-only mode.
	*/
	if (claim_txg == 0 && spa_writeable(td->td_spa))
	return;

	zilog_t *zilog = zil_alloc(spa_get_dsl(td->td_spa)->dp_meta_objset, zh);
	(void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, td,
	claim_txg, !(td->td_flags & TRAVERSE_NO_DECRYPT));
	zil_free(zilog);
	}

	typedef enum resume_skip {
	RESUME_SKIP_ALL,
	RESUME_SKIP_NONE,
	RESUME_SKIP_CHILDREN
	} resume_skip_t;

	/*
	* Returns RESUME_SKIP_ALL if td indicates that we are resuming a traversal and
	* the block indicated by zb does not need to be visited at all. Returns
	* RESUME_SKIP_CHILDREN if we are resuming a post traversal and we reach the
	* resume point. This indicates that this block should be visited but not its
	* children (since they must have been visited in a previous traversal).
	* Otherwise returns RESUME_SKIP_NONE.
	*/
	static resume_skip_t
	resume_skip_check(traverse_data_t td, const dnode_phys_t dnp,
	const zbookmark_phys_t *zb)
	{
	if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume)) {
	/*
	* If we already visited this bp & everything below,
	* don't bother doing it again.
	*/
	if (zbookmark_subtree_completed(dnp, zb, td->td_resume))
	return (RESUME_SKIP_ALL);

	/*
	* If we found the block we're trying to resume from, zero
	* the bookmark out to indicate that we have resumed.
	*/
	if (memcmp(zb, td->td_resume, sizeof (*zb)) == 0) {
	memset(td->td_resume, 0, sizeof (*zb));
	if (td->td_flags & TRAVERSE_POST)
	return (RESUME_SKIP_CHILDREN);
	}
	}
	return (RESUME_SKIP_NONE);
	}

	/*
	* Returns B_TRUE, if prefetch read is issued, otherwise B_FALSE.
	*/
	static boolean_t
	traverse_prefetch_metadata(traverse_data_t *td,
	const blkptr_t bp, const zbookmark_phys_t zb)
	{
	arc_flags_t flags = ARC_FLAG_NOWAIT \| ARC_FLAG_PREFETCH;
	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE;

	if (!(td->td_flags & TRAVERSE_PREFETCH_METADATA))
	return (B_FALSE);
	/*
	* If we are in the process of resuming, don't prefetch, because
	* some children will not be needed (and in fact may have already
	* been freed).
	*/
	if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume))
	return (B_FALSE);
	if (BP_IS_HOLE(bp) \|\| bp->blk_birth <= td->td_min_txg)
	return (B_FALSE);
	if (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE)
	return (B_FALSE);
	ASSERT(!BP_IS_REDACTED(bp));

	if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
	zio_flags \|= ZIO_FLAG_RAW;

	(void) arc_read(NULL, td->td_spa, bp, NULL, NULL,
	ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
	return (B_TRUE);
	}

	static boolean_t
	prefetch_needed(prefetch_data_t pfd, const blkptr_t bp)
	{
	ASSERT(pfd->pd_flags & TRAVERSE_PREFETCH_DATA);
	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp) \|\|
	BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG \|\| BP_IS_REDACTED(bp))
	return (B_FALSE);
	return (B_TRUE);
	}

	static int
	traverse_visitbp(traverse_data_t td, const dnode_phys_t dnp,
	const blkptr_t bp, const zbookmark_phys_t zb)
	{
	int err = 0;
	arc_buf_t *buf = NULL;
	prefetch_data_t *pd = td->td_pfd;

	switch (resume_skip_check(td, dnp, zb)) {
	case RESUME_SKIP_ALL:
	return (0);
	case RESUME_SKIP_CHILDREN:
	goto post;
	case RESUME_SKIP_NONE:
	break;
	default:
	ASSERT(0);
	}

	if (bp->blk_birth == 0) {
	/*
	* Since this block has a birth time of 0 it must be one of
	* two things: a hole created before the
	* SPA_FEATURE_HOLE_BIRTH feature was enabled, or a hole
	* which has always been a hole in an object.
	*
	* If a file is written sparsely, then the unwritten parts of
	* the file were "always holes" -- that is, they have been
	* holes since this object was allocated. However, we (and
	* our callers) can not necessarily tell when an object was
	* allocated. Therefore, if it's possible that this object
	* was freed and then its object number reused, we need to
	* visit all the holes with birth==0.
	*
	* If it isn't possible that the object number was reused,
	* then if SPA_FEATURE_HOLE_BIRTH was enabled before we wrote
	* all the blocks we will visit as part of this traversal,
	* then this hole must have always existed, so we can skip
	* it. We visit blocks born after (exclusive) td_min_txg.
	*
	* Note that the meta-dnode cannot be reallocated.
	*/
	if (!send_holes_without_birth_time &&
	(!td->td_realloc_possible \|\|
	zb->zb_object == DMU_META_DNODE_OBJECT) &&
	td->td_hole_birth_enabled_txg <= td->td_min_txg)
	return (0);
	} else if (bp->blk_birth <= td->td_min_txg) {
	return (0);
	}

	if (pd != NULL && !pd->pd_exited && prefetch_needed(pd, bp)) {
	uint64_t size = BP_GET_LSIZE(bp);
	mutex_enter(&pd->pd_mtx);
	ASSERT(pd->pd_bytes_fetched >= 0);
	while (pd->pd_bytes_fetched < size && !pd->pd_exited)
	cv_wait_sig(&pd->pd_cv, &pd->pd_mtx);
	pd->pd_bytes_fetched -= size;
	cv_broadcast(&pd->pd_cv);
	mutex_exit(&pd->pd_mtx);
	}

	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp)) {
	err = td->td_func(td->td_spa, NULL, bp, zb, dnp, td->td_arg);
	if (err != 0)
	goto post;
	return (0);
	}

	if (td->td_flags & TRAVERSE_PRE) {
	err = td->td_func(td->td_spa, NULL, bp, zb, dnp,
	td->td_arg);
	if (err == TRAVERSE_VISIT_NO_CHILDREN)
	return (0);
	if (err != 0)
	goto post;
	}

	if (BP_GET_LEVEL(bp) > 0) {
	uint32_t flags = ARC_FLAG_WAIT;
	int32_t i, ptidx, pidx;
	uint32_t prefetchlimit;
	int32_t epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
	zbookmark_phys_t *czb;

	ASSERT(!BP_IS_PROTECTED(bp));

	err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
	if (err != 0)
	goto post;

	czb = kmem_alloc(sizeof (zbookmark_phys_t), KM_SLEEP);

	/*
	* When performing a traversal it is beneficial to
	* asynchronously read-ahead the upcoming indirect
	* blocks since they will be needed shortly. However,
	* since a 128k indirect (non-L0) block may contain up
	* to 1024 128-byte block pointers, its preferable to not
	* prefetch them all at once. Issuing a large number of
	* async reads may effect performance, and the earlier
	* the indirect blocks are prefetched the less likely
	* they are to still be resident in the ARC when needed.
	* Therefore, prefetching indirect blocks is limited to
	* zfs_traverse_indirect_prefetch_limit=32 blocks by
	* default.
	*
	* pidx: Index for which next prefetch to be issued.
	* ptidx: Index at which next prefetch to be triggered.
	*/
	ptidx = 0;
	pidx = 1;
	prefetchlimit = zfs_traverse_indirect_prefetch_limit;
	for (i = 0; i < epb; i++) {
	if (prefetchlimit && i == ptidx) {
	ASSERT3S(ptidx, <=, pidx);
	for (uint32_t prefetched = 0; pidx < epb &&
	prefetched < prefetchlimit; pidx++) {
	SET_BOOKMARK(czb, zb->zb_objset,
	zb->zb_object, zb->zb_level - 1,
	zb->zb_blkid * epb + pidx);
	if (traverse_prefetch_metadata(td,
	&((blkptr_t *)buf->b_data)[pidx],
	czb) == B_TRUE) {
	prefetched++;
	if (prefetched ==
	MAX(prefetchlimit / 2, 1))
	ptidx = pidx;
	}
	}
	}

	/* recursively visitbp() blocks below this */
	SET_BOOKMARK(czb, zb->zb_objset, zb->zb_object,
	zb->zb_level - 1,
	zb->zb_blkid * epb + i);
	err = traverse_visitbp(td, dnp,
	&((blkptr_t *)buf->b_data)[i], czb);
	if (err != 0)
	break;
	}

	kmem_free(czb, sizeof (zbookmark_phys_t));

	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
	uint32_t flags = ARC_FLAG_WAIT;
	uint32_t zio_flags = ZIO_FLAG_CANFAIL;
	int32_t i;
	int32_t epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
	dnode_phys_t *child_dnp;

	/*
	* dnode blocks might have their bonus buffers encrypted, so
	* we must be careful to honor TRAVERSE_NO_DECRYPT
	*/
	if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
	zio_flags \|= ZIO_FLAG_RAW;

	err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
	if (err != 0)
	goto post;

	child_dnp = buf->b_data;

	for (i = 0; i < epb; i += child_dnp[i].dn_extra_slots + 1) {
	prefetch_dnode_metadata(td, &child_dnp[i],
	zb->zb_objset, zb->zb_blkid * epb + i);
	}

	/* recursively visitbp() blocks below this */
	for (i = 0; i < epb; i += child_dnp[i].dn_extra_slots + 1) {
	err = traverse_dnode(td, bp, &child_dnp[i],
	zb->zb_objset, zb->zb_blkid * epb + i);
	if (err != 0)
	break;
	}
	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
	uint32_t zio_flags = ZIO_FLAG_CANFAIL;
	arc_flags_t flags = ARC_FLAG_WAIT;
	objset_phys_t *osp;

	if ((td->td_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
	zio_flags \|= ZIO_FLAG_RAW;

	err = arc_read(NULL, td->td_spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, zb);
	if (err != 0)
	goto post;

	osp = buf->b_data;
	prefetch_dnode_metadata(td, &osp->os_meta_dnode, zb->zb_objset,
	DMU_META_DNODE_OBJECT);
	/*
	* See the block comment above for the goal of this variable.
	* If the maxblkid of the meta-dnode is 0, then we know that
	* we've never had more than DNODES_PER_BLOCK objects in the
	* dataset, which means we can't have reused any object ids.
	*/
	if (osp->os_meta_dnode.dn_maxblkid == 0)
	td->td_realloc_possible = B_FALSE;

	if (OBJSET_BUF_HAS_USERUSED(buf)) {
	if (OBJSET_BUF_HAS_PROJECTUSED(buf))
	prefetch_dnode_metadata(td,
	&osp->os_projectused_dnode,
	zb->zb_objset, DMU_PROJECTUSED_OBJECT);
	prefetch_dnode_metadata(td, &osp->os_groupused_dnode,
	zb->zb_objset, DMU_GROUPUSED_OBJECT);
	prefetch_dnode_metadata(td, &osp->os_userused_dnode,
	zb->zb_objset, DMU_USERUSED_OBJECT);
	}

	err = traverse_dnode(td, bp, &osp->os_meta_dnode, zb->zb_objset,
	DMU_META_DNODE_OBJECT);
	if (err == 0 && OBJSET_BUF_HAS_USERUSED(buf)) {
	if (OBJSET_BUF_HAS_PROJECTUSED(buf))
	err = traverse_dnode(td, bp,
	&osp->os_projectused_dnode, zb->zb_objset,
	DMU_PROJECTUSED_OBJECT);
	if (err == 0)
	err = traverse_dnode(td, bp,
	&osp->os_groupused_dnode, zb->zb_objset,
	DMU_GROUPUSED_OBJECT);
	if (err == 0)
	err = traverse_dnode(td, bp,
	&osp->os_userused_dnode, zb->zb_objset,
	DMU_USERUSED_OBJECT);
	}
	}

	if (buf)
	arc_buf_destroy(buf, &buf);

	post:
	if (err == 0 && (td->td_flags & TRAVERSE_POST))
	err = td->td_func(td->td_spa, NULL, bp, zb, dnp, td->td_arg);

	if ((td->td_flags & TRAVERSE_HARD) && (err == EIO \|\| err == ECKSUM)) {
	/*
	* Ignore this disk error as requested by the HARD flag,
	* and continue traversal.
	*/
	err = 0;
	}

	/*
	* If we are stopping here, set td_resume.
	*/
	if (td->td_resume != NULL && err != 0 && !td->td_paused) {
	td->td_resume->zb_objset = zb->zb_objset;
	td->td_resume->zb_object = zb->zb_object;
	td->td_resume->zb_level = 0;
	/*
	* If we have stopped on an indirect block (e.g. due to
	* i/o error), we have not visited anything below it.
	* Set the bookmark to the first level-0 block that we need
	* to visit. This way, the resuming code does not need to
	* deal with resuming from indirect blocks.
	*
	* Note, if zb_level <= 0, dnp may be NULL, so we don't want
	* to dereference it.
	*/
	td->td_resume->zb_blkid = zb->zb_blkid;
	if (zb->zb_level > 0) {
	td->td_resume->zb_blkid <<= zb->zb_level *
	(dnp->dn_indblkshift - SPA_BLKPTRSHIFT);
	}
	td->td_paused = B_TRUE;
	}

	return (err);
	}

	static void
	prefetch_dnode_metadata(traverse_data_t td, const dnode_phys_t dnp,
	uint64_t objset, uint64_t object)
	{
	int j;
	zbookmark_phys_t czb;

	for (j = 0; j < dnp->dn_nblkptr; j++) {
	SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
	traverse_prefetch_metadata(td, &dnp->dn_blkptr[j], &czb);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
	traverse_prefetch_metadata(td, DN_SPILL_BLKPTR(dnp), &czb);
	}
	}

	static int
	traverse_dnode(traverse_data_t td, const blkptr_t bp, const dnode_phys_t *dnp,
	uint64_t objset, uint64_t object)
	{
	int j, err = 0;
	zbookmark_phys_t czb;

	if (object != DMU_META_DNODE_OBJECT && td->td_resume != NULL &&
	object < td->td_resume->zb_object)
	return (0);

	if (td->td_flags & TRAVERSE_PRE) {
	SET_BOOKMARK(&czb, objset, object, ZB_DNODE_LEVEL,
	ZB_DNODE_BLKID);
	err = td->td_func(td->td_spa, NULL, bp, &czb, dnp,
	td->td_arg);
	if (err == TRAVERSE_VISIT_NO_CHILDREN)
	return (0);
	if (err != 0)
	return (err);
	}

	for (j = 0; j < dnp->dn_nblkptr; j++) {
	SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
	err = traverse_visitbp(td, dnp, &dnp->dn_blkptr[j], &czb);
	if (err != 0)
	break;
	}

	if (err == 0 && (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
	SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
	err = traverse_visitbp(td, dnp, DN_SPILL_BLKPTR(dnp), &czb);
	}

	if (err == 0 && (td->td_flags & TRAVERSE_POST)) {
	SET_BOOKMARK(&czb, objset, object, ZB_DNODE_LEVEL,
	ZB_DNODE_BLKID);
	err = td->td_func(td->td_spa, NULL, bp, &czb, dnp,
	td->td_arg);
	if (err == TRAVERSE_VISIT_NO_CHILDREN)
	return (0);
	if (err != 0)
	return (err);
	}
	return (err);
	}

	static int
	traverse_prefetcher(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	(void) zilog, (void) dnp;
	prefetch_data_t *pfd = arg;
	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE;
	arc_flags_t aflags = ARC_FLAG_NOWAIT \| ARC_FLAG_PREFETCH \|
	ARC_FLAG_PRESCIENT_PREFETCH;

	ASSERT(pfd->pd_bytes_fetched >= 0);
	if (zb->zb_level == ZB_DNODE_LEVEL)
	return (0);
	if (pfd->pd_cancel)
	return (SET_ERROR(EINTR));

	if (!prefetch_needed(pfd, bp))
	return (0);

	mutex_enter(&pfd->pd_mtx);
	while (!pfd->pd_cancel && pfd->pd_bytes_fetched >= zfs_pd_bytes_max)
	cv_wait_sig(&pfd->pd_cv, &pfd->pd_mtx);
	pfd->pd_bytes_fetched += BP_GET_LSIZE(bp);
	cv_broadcast(&pfd->pd_cv);
	mutex_exit(&pfd->pd_mtx);

	if ((pfd->pd_flags & TRAVERSE_NO_DECRYPT) && BP_IS_PROTECTED(bp))
	zio_flags \|= ZIO_FLAG_RAW;

	(void) arc_read(NULL, spa, bp, NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
	zio_flags, &aflags, zb);

	return (0);
	}

	static void
	traverse_prefetch_thread(void *arg)
	{
	traverse_data_t *td_main = arg;
	traverse_data_t td = *td_main;
	zbookmark_phys_t czb;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	td.td_func = traverse_prefetcher;
	td.td_arg = td_main->td_pfd;
	td.td_pfd = NULL;
	td.td_resume = &td_main->td_pfd->pd_resume;

	SET_BOOKMARK(&czb, td.td_objset,
	ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
	(void) traverse_visitbp(&td, NULL, td.td_rootbp, &czb);

	mutex_enter(&td_main->td_pfd->pd_mtx);
	td_main->td_pfd->pd_exited = B_TRUE;
	cv_broadcast(&td_main->td_pfd->pd_cv);
	mutex_exit(&td_main->td_pfd->pd_mtx);
	spl_fstrans_unmark(cookie);
	}

	/*
	* NB: dataset must not be changing on-disk (eg, is a snapshot or we are
	* in syncing context).
	*/
	static int
	traverse_impl(spa_t spa, dsl_dataset_t ds, uint64_t objset, blkptr_t *rootbp,
	uint64_t txg_start, zbookmark_phys_t *resume, int flags,
	blkptr_cb_t func, void *arg)
	{
	traverse_data_t *td;
	prefetch_data_t *pd;
	zbookmark_phys_t *czb;
	int err;

	ASSERT(ds == NULL \|\| objset == ds->ds_object);
	ASSERT(!(flags & TRAVERSE_PRE) \|\| !(flags & TRAVERSE_POST));

	td = kmem_alloc(sizeof (traverse_data_t), KM_SLEEP);
	pd = kmem_zalloc(sizeof (prefetch_data_t), KM_SLEEP);
	czb = kmem_alloc(sizeof (zbookmark_phys_t), KM_SLEEP);

	td->td_spa = spa;
	td->td_objset = objset;
	td->td_rootbp = rootbp;
	td->td_min_txg = txg_start;
	td->td_resume = resume;
	td->td_func = func;
	td->td_arg = arg;
	td->td_pfd = pd;
	td->td_flags = flags;
	td->td_paused = B_FALSE;
	td->td_realloc_possible = (txg_start == 0 ? B_FALSE : B_TRUE);

	if (spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
	VERIFY(spa_feature_enabled_txg(spa,
	SPA_FEATURE_HOLE_BIRTH, &td->td_hole_birth_enabled_txg));
	} else {
	td->td_hole_birth_enabled_txg = UINT64_MAX;
	}

	pd->pd_flags = flags;
	if (resume != NULL)
	pd->pd_resume = *resume;
	mutex_init(&pd->pd_mtx, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&pd->pd_cv, NULL, CV_DEFAULT, NULL);

	SET_BOOKMARK(czb, td->td_objset,
	ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);

	/* See comment on ZIL traversal in dsl_scan_visitds. */
	if (ds != NULL && !ds->ds_is_snapshot && !BP_IS_HOLE(rootbp)) {
	enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
	uint32_t flags = ARC_FLAG_WAIT;
	objset_phys_t *osp;
	arc_buf_t *buf;
	ASSERT(!BP_IS_REDACTED(rootbp));

	if ((td->td_flags & TRAVERSE_NO_DECRYPT) &&
	BP_IS_PROTECTED(rootbp))
	zio_flags \|= ZIO_FLAG_RAW;

	err = arc_read(NULL, td->td_spa, rootbp, arc_getbuf_func,
	&buf, ZIO_PRIORITY_ASYNC_READ, zio_flags, &flags, czb);
	if (err != 0) {
	/*
	* If both TRAVERSE_HARD and TRAVERSE_PRE are set,
	* continue to visitbp so that td_func can be called
	* in pre stage, and err will reset to zero.
	*/
	if (!(td->td_flags & TRAVERSE_HARD) \|\|
	!(td->td_flags & TRAVERSE_PRE))
	goto out;
	} else {
	osp = buf->b_data;
	traverse_zil(td, &osp->os_zil_header);
	arc_buf_destroy(buf, &buf);
	}
	}

	if (!(flags & TRAVERSE_PREFETCH_DATA) \|\|
	taskq_dispatch(spa->spa_prefetch_taskq, traverse_prefetch_thread,
	td, TQ_NOQUEUE) == TASKQID_INVALID)
	pd->pd_exited = B_TRUE;

	err = traverse_visitbp(td, NULL, rootbp, czb);

	mutex_enter(&pd->pd_mtx);
	pd->pd_cancel = B_TRUE;
	cv_broadcast(&pd->pd_cv);
	while (!pd->pd_exited)
	cv_wait_sig(&pd->pd_cv, &pd->pd_mtx);
	mutex_exit(&pd->pd_mtx);
	out:
	mutex_destroy(&pd->pd_mtx);
	cv_destroy(&pd->pd_cv);

	kmem_free(czb, sizeof (zbookmark_phys_t));
	kmem_free(pd, sizeof (struct prefetch_data));
	kmem_free(td, sizeof (struct traverse_data));

	return (err);
	}

	/*
	* NB: dataset must not be changing on-disk (eg, is a snapshot or we are
	* in syncing context).
	*/
	int
	traverse_dataset_resume(dsl_dataset_t *ds, uint64_t txg_start,
	zbookmark_phys_t *resume,
	int flags, blkptr_cb_t func, void *arg)
	{
	return (traverse_impl(ds->ds_dir->dd_pool->dp_spa, ds, ds->ds_object,
	&dsl_dataset_phys(ds)->ds_bp, txg_start, resume, flags, func, arg));
	}

	int
	traverse_dataset(dsl_dataset_t *ds, uint64_t txg_start,
	int flags, blkptr_cb_t func, void *arg)
	{
	return (traverse_dataset_resume(ds, txg_start, NULL, flags, func, arg));
	}

	int
	traverse_dataset_destroyed(spa_t spa, blkptr_t blkptr,
	uint64_t txg_start, zbookmark_phys_t *resume, int flags,
	blkptr_cb_t func, void *arg)
	{
	return (traverse_impl(spa, NULL, ZB_DESTROYED_OBJSET,
	blkptr, txg_start, resume, flags, func, arg));
	}

	/*
	* NB: pool must not be changing on-disk (eg, from zdb or sync context).
	*/
	int
	traverse_pool(spa_t *spa, uint64_t txg_start, int flags,
	blkptr_cb_t func, void *arg)
	{
	int err;
	dsl_pool_t *dp = spa_get_dsl(spa);
	objset_t *mos = dp->dp_meta_objset;
	boolean_t hard = (flags & TRAVERSE_HARD);

	/* visit the MOS */
	err = traverse_impl(spa, NULL, 0, spa_get_rootblkptr(spa),
	txg_start, NULL, flags, func, arg);
	if (err != 0)
	return (err);

	/* visit each dataset */
	for (uint64_t obj = 1; err == 0;
	err = dmu_object_next(mos, &obj, B_FALSE, txg_start)) {
	dmu_object_info_t doi;

	err = dmu_object_info(mos, obj, &doi);
	if (err != 0) {
	if (hard)
	continue;
	break;
	}

	if (doi.doi_bonus_type == DMU_OT_DSL_DATASET) {
	dsl_dataset_t *ds;
	uint64_t txg = txg_start;

	dsl_pool_config_enter(dp, FTAG);
	err = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
	dsl_pool_config_exit(dp, FTAG);
	if (err != 0) {
	if (hard)
	continue;
	break;
	}
	if (dsl_dataset_phys(ds)->ds_prev_snap_txg > txg)
	txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
	err = traverse_dataset(ds, txg, flags, func, arg);
	dsl_dataset_rele(ds, FTAG);
	if (err != 0)
	break;
	}
	}
	if (err == ESRCH)
	err = 0;
	return (err);
	}

	EXPORT_SYMBOL(traverse_dataset);
	EXPORT_SYMBOL(traverse_pool);

	ZFS_MODULE_PARAM(zfs, zfs_, pd_bytes_max, INT, ZMOD_RW,
	"Max number of bytes to prefetch");

	-ZFS_MODULE_PARAM(zfs, zfs_, traverse_indirect_prefetch_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, traverse_indirect_prefetch_limit, UINT, ZMOD_RW,
	"Traverse prefetch number of blocks pointed by indirect block");

	#if defined(_KERNEL)
	module_param_named(ignore_hole_birth, send_holes_without_birth_time, int, 0644);
	MODULE_PARM_DESC(ignore_hole_birth,
	"Alias for send_holes_without_birth_time");
	#endif

	/* CSTYLED */
	ZFS_MODULE_PARAM(zfs, , send_holes_without_birth_time, INT, ZMOD_RW,
	"Ignore hole_birth txg for zfs send");
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_dataset.c b/sys/contrib/openzfs/module/zfs/dsl_dataset.c
	index 94b77aa1b743..7a066b786cd0 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_dataset.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_dataset.c
	@@ -1,5017 +1,5019 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright (c) 2014 RackTop Systems.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright 2016, OmniTI Computer Consulting, Inc. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2020 The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	*/

	#include <sys/dmu_objset.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/arc.h>
	#include <sys/zio.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/unique.h>
	#include <sys/zfs_context.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_onexit.h>
	#include <sys/zvol.h>
	#include <sys/dsl_scan.h>
	#include <sys/dsl_deadlist.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_userhold.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/policy.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/zio_compress.h>
	#include <zfs_fletcher.h>
	#include <sys/zio_checksum.h>

	/*
	* The SPA supports block sizes up to 16MB. However, very large blocks
	* can have an impact on i/o latency (e.g. tying up a spinning disk for
	* ~300ms), and also potentially on the memory allocator. Therefore,
	* we did not allow the recordsize to be set larger than zfs_max_recordsize
	* (former default: 1MB). Larger blocks could be created by changing this
	* tunable, and pools with larger blocks could always be imported and used,
	* regardless of this setting.
	*
	* We do, however, still limit it by default to 1M on x86_32, because Linux's
	* 3/1 memory split doesn't leave much room for 16M chunks.
	*/
	#ifdef _ILP32
	-int zfs_max_recordsize = 1 * 1024 * 1024;
	+uint_t zfs_max_recordsize = 1 * 1024 * 1024;
	#else
	-int zfs_max_recordsize = 16 * 1024 * 1024;
	+uint_t zfs_max_recordsize = 16 * 1024 * 1024;
	#endif
	static int zfs_allow_redacted_dataset_mount = 0;

	int zfs_snapshot_history_enabled = 1;

	#define SWITCH64(x, y) \
	{ \
	uint64_t __tmp = (x); \
	(x) = (y); \
	(y) = __tmp; \
	}

	#define DS_REF_MAX (1ULL << 62)

	static void dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds,
	uint64_t obj, dmu_tx_t *tx);
	static void dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t *ds,
	dmu_tx_t *tx);

	static void unload_zfeature(dsl_dataset_t *ds, spa_feature_t f);

	-extern int spa_asize_inflation;
	+extern uint_t spa_asize_inflation;

	static zil_header_t zero_zil;

	/*
	* Figure out how much of this delta should be propagated to the dsl_dir
	* layer. If there's a refreservation, that space has already been
	* partially accounted for in our ancestors.
	*/
	static int64_t
	parent_delta(dsl_dataset_t *ds, int64_t delta)
	{
	dsl_dataset_phys_t *ds_phys;
	uint64_t old_bytes, new_bytes;

	if (ds->ds_reserved == 0)
	return (delta);

	ds_phys = dsl_dataset_phys(ds);
	old_bytes = MAX(ds_phys->ds_unique_bytes, ds->ds_reserved);
	new_bytes = MAX(ds_phys->ds_unique_bytes + delta, ds->ds_reserved);

	ASSERT3U(ABS((int64_t)(new_bytes - old_bytes)), <=, ABS(delta));
	return (new_bytes - old_bytes);
	}

	void
	dsl_dataset_block_born(dsl_dataset_t ds, const blkptr_t bp, dmu_tx_t *tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	int used = bp_get_dsize_sync(spa, bp);
	int compressed = BP_GET_PSIZE(bp);
	int uncompressed = BP_GET_UCSIZE(bp);
	int64_t delta;
	spa_feature_t f;

	dprintf_bp(bp, "ds=%p", ds);

	ASSERT(dmu_tx_is_syncing(tx));
	/* It could have been compressed away to nothing */
	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp))
	return;
	ASSERT(BP_GET_TYPE(bp) != DMU_OT_NONE);
	ASSERT(DMU_OT_IS_VALID(BP_GET_TYPE(bp)));
	if (ds == NULL) {
	dsl_pool_mos_diduse_space(tx->tx_pool,
	used, compressed, uncompressed);
	return;
	}

	ASSERT3U(bp->blk_birth, >, dsl_dataset_phys(ds)->ds_prev_snap_txg);
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	mutex_enter(&ds->ds_lock);
	delta = parent_delta(ds, used);
	dsl_dataset_phys(ds)->ds_referenced_bytes += used;
	dsl_dataset_phys(ds)->ds_compressed_bytes += compressed;
	dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed;
	dsl_dataset_phys(ds)->ds_unique_bytes += used;

	if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE) {
	ds->ds_feature_activation[SPA_FEATURE_LARGE_BLOCKS] =
	(void *)B_TRUE;
	}


	f = zio_checksum_to_feature(BP_GET_CHECKSUM(bp));
	if (f != SPA_FEATURE_NONE) {
	ASSERT3S(spa_feature_table[f].fi_type, ==,
	ZFEATURE_TYPE_BOOLEAN);
	ds->ds_feature_activation[f] = (void *)B_TRUE;
	}

	f = zio_compress_to_feature(BP_GET_COMPRESS(bp));
	if (f != SPA_FEATURE_NONE) {
	ASSERT3S(spa_feature_table[f].fi_type, ==,
	ZFEATURE_TYPE_BOOLEAN);
	ds->ds_feature_activation[f] = (void *)B_TRUE;
	}

	/*
	* Track block for livelist, but ignore embedded blocks because
	* they do not need to be freed.
	*/
	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
	bp->blk_birth > ds->ds_dir->dd_origin_txg &&
	!(BP_IS_EMBEDDED(bp))) {
	ASSERT(dsl_dir_is_clone(ds->ds_dir));
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LIVELIST));
	bplist_append(&ds->ds_dir->dd_pending_allocs, bp);
	}

	mutex_exit(&ds->ds_lock);
	dsl_dir_diduse_transfer_space(ds->ds_dir, delta,
	compressed, uncompressed, used,
	DD_USED_REFRSRV, DD_USED_HEAD, tx);
	}

	/*
	* Called when the specified segment has been remapped, and is thus no
	* longer referenced in the head dataset. The vdev must be indirect.
	*
	* If the segment is referenced by a snapshot, put it on the remap deadlist.
	* Otherwise, add this segment to the obsolete spacemap.
	*/
	void
	dsl_dataset_block_remapped(dsl_dataset_t *ds, uint64_t vdev, uint64_t offset,
	uint64_t size, uint64_t birth, dmu_tx_t *tx)
	{
	spa_t *spa = ds->ds_dir->dd_pool->dp_spa;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(birth <= tx->tx_txg);
	ASSERT(!ds->ds_is_snapshot);

	if (birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) {
	spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx);
	} else {
	blkptr_t fakebp;
	dva_t *dva = &fakebp.blk_dva[0];

	ASSERT(ds != NULL);

	mutex_enter(&ds->ds_remap_deadlist_lock);
	if (!dsl_dataset_remap_deadlist_exists(ds)) {
	dsl_dataset_create_remap_deadlist(ds, tx);
	}
	mutex_exit(&ds->ds_remap_deadlist_lock);

	BP_ZERO(&fakebp);
	fakebp.blk_birth = birth;
	DVA_SET_VDEV(dva, vdev);
	DVA_SET_OFFSET(dva, offset);
	DVA_SET_ASIZE(dva, size);
	dsl_deadlist_insert(&ds->ds_remap_deadlist, &fakebp, B_FALSE,
	tx);
	}
	}

	int
	dsl_dataset_block_kill(dsl_dataset_t ds, const blkptr_t bp, dmu_tx_t *tx,
	boolean_t async)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	int used = bp_get_dsize_sync(spa, bp);
	int compressed = BP_GET_PSIZE(bp);
	int uncompressed = BP_GET_UCSIZE(bp);

	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp))
	return (0);

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(bp->blk_birth <= tx->tx_txg);

	if (ds == NULL) {
	dsl_free(tx->tx_pool, tx->tx_txg, bp);
	dsl_pool_mos_diduse_space(tx->tx_pool,
	-used, -compressed, -uncompressed);
	return (used);
	}
	ASSERT3P(tx->tx_pool, ==, ds->ds_dir->dd_pool);

	ASSERT(!ds->ds_is_snapshot);
	dmu_buf_will_dirty(ds->ds_dbuf, tx);

	/*
	* Track block for livelist, but ignore embedded blocks because
	* they do not need to be freed.
	*/
	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
	bp->blk_birth > ds->ds_dir->dd_origin_txg &&
	!(BP_IS_EMBEDDED(bp))) {
	ASSERT(dsl_dir_is_clone(ds->ds_dir));
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LIVELIST));
	bplist_append(&ds->ds_dir->dd_pending_frees, bp);
	}

	if (bp->blk_birth > dsl_dataset_phys(ds)->ds_prev_snap_txg) {
	int64_t delta;

	dprintf_bp(bp, "freeing ds=%llu", (u_longlong_t)ds->ds_object);
	dsl_free(tx->tx_pool, tx->tx_txg, bp);

	mutex_enter(&ds->ds_lock);
	ASSERT(dsl_dataset_phys(ds)->ds_unique_bytes >= used \|\|
	!DS_UNIQUE_IS_ACCURATE(ds));
	delta = parent_delta(ds, -used);
	dsl_dataset_phys(ds)->ds_unique_bytes -= used;
	mutex_exit(&ds->ds_lock);
	dsl_dir_diduse_transfer_space(ds->ds_dir,
	delta, -compressed, -uncompressed, -used,
	DD_USED_REFRSRV, DD_USED_HEAD, tx);
	} else {
	dprintf_bp(bp, "putting on dead list: %s", "");
	if (async) {
	/*
	* We are here as part of zio's write done callback,
	* which means we're a zio interrupt thread. We can't
	* call dsl_deadlist_insert() now because it may block
	* waiting for I/O. Instead, put bp on the deferred
	* queue and let dsl_pool_sync() finish the job.
	*/
	bplist_append(&ds->ds_pending_deadlist, bp);
	} else {
	dsl_deadlist_insert(&ds->ds_deadlist, bp, B_FALSE, tx);
	}
	ASSERT3U(ds->ds_prev->ds_object, ==,
	dsl_dataset_phys(ds)->ds_prev_snap_obj);
	ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_num_children > 0);
	/* if (bp->blk_birth > prev prev snap txg) prev unique += bs */
	if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
	ds->ds_object && bp->blk_birth >
	dsl_dataset_phys(ds->ds_prev)->ds_prev_snap_txg) {
	dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
	mutex_enter(&ds->ds_prev->ds_lock);
	dsl_dataset_phys(ds->ds_prev)->ds_unique_bytes += used;
	mutex_exit(&ds->ds_prev->ds_lock);
	}
	if (bp->blk_birth > ds->ds_dir->dd_origin_txg) {
	dsl_dir_transfer_space(ds->ds_dir, used,
	DD_USED_HEAD, DD_USED_SNAP, tx);
	}
	}

	dsl_bookmark_block_killed(ds, bp, tx);

	mutex_enter(&ds->ds_lock);
	ASSERT3U(dsl_dataset_phys(ds)->ds_referenced_bytes, >=, used);
	dsl_dataset_phys(ds)->ds_referenced_bytes -= used;
	ASSERT3U(dsl_dataset_phys(ds)->ds_compressed_bytes, >=, compressed);
	dsl_dataset_phys(ds)->ds_compressed_bytes -= compressed;
	ASSERT3U(dsl_dataset_phys(ds)->ds_uncompressed_bytes, >=, uncompressed);
	dsl_dataset_phys(ds)->ds_uncompressed_bytes -= uncompressed;
	mutex_exit(&ds->ds_lock);

	return (used);
	}

	struct feature_type_uint64_array_arg {
	uint64_t length;
	uint64_t *array;
	};

	static void
	unload_zfeature(dsl_dataset_t *ds, spa_feature_t f)
	{
	switch (spa_feature_table[f].fi_type) {
	case ZFEATURE_TYPE_BOOLEAN:
	break;
	case ZFEATURE_TYPE_UINT64_ARRAY:
	{
	struct feature_type_uint64_array_arg *ftuaa = ds->ds_feature[f];
	kmem_free(ftuaa->array, ftuaa->length * sizeof (uint64_t));
	kmem_free(ftuaa, sizeof (*ftuaa));
	break;
	}
	default:
	panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
	}
	}

	static int
	load_zfeature(objset_t mos, dsl_dataset_t ds, spa_feature_t f)
	{
	int err = 0;
	switch (spa_feature_table[f].fi_type) {
	case ZFEATURE_TYPE_BOOLEAN:
	err = zap_contains(mos, ds->ds_object,
	spa_feature_table[f].fi_guid);
	if (err == 0) {
	ds->ds_feature[f] = (void *)B_TRUE;
	} else {
	ASSERT3U(err, ==, ENOENT);
	err = 0;
	}
	break;
	case ZFEATURE_TYPE_UINT64_ARRAY:
	{
	uint64_t int_size, num_int;
	uint64_t *data;
	err = zap_length(mos, ds->ds_object,
	spa_feature_table[f].fi_guid, &int_size, &num_int);
	if (err != 0) {
	ASSERT3U(err, ==, ENOENT);
	err = 0;
	break;
	}
	ASSERT3U(int_size, ==, sizeof (uint64_t));
	data = kmem_alloc(int_size * num_int, KM_SLEEP);
	VERIFY0(zap_lookup(mos, ds->ds_object,
	spa_feature_table[f].fi_guid, int_size, num_int, data));
	struct feature_type_uint64_array_arg *ftuaa =
	kmem_alloc(sizeof (*ftuaa), KM_SLEEP);
	ftuaa->length = num_int;
	ftuaa->array = data;
	ds->ds_feature[f] = ftuaa;
	break;
	}
	default:
	panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
	}
	return (err);
	}

	/*
	* We have to release the fsid synchronously or we risk that a subsequent
	* mount of the same dataset will fail to unique_insert the fsid. This
	* failure would manifest itself as the fsid of this dataset changing
	* between mounts which makes NFS clients quite unhappy.
	*/
	static void
	dsl_dataset_evict_sync(void *dbu)
	{
	dsl_dataset_t *ds = dbu;

	ASSERT(ds->ds_owner == NULL);

	unique_remove(ds->ds_fsid_guid);
	}

	static void
	dsl_dataset_evict_async(void *dbu)
	{
	dsl_dataset_t *ds = dbu;

	ASSERT(ds->ds_owner == NULL);

	ds->ds_dbuf = NULL;

	if (ds->ds_objset != NULL)
	dmu_objset_evict(ds->ds_objset);

	if (ds->ds_prev) {
	dsl_dataset_rele(ds->ds_prev, ds);
	ds->ds_prev = NULL;
	}

	dsl_bookmark_fini_ds(ds);

	bplist_destroy(&ds->ds_pending_deadlist);
	if (dsl_deadlist_is_open(&ds->ds_deadlist))
	dsl_deadlist_close(&ds->ds_deadlist);
	if (dsl_deadlist_is_open(&ds->ds_remap_deadlist))
	dsl_deadlist_close(&ds->ds_remap_deadlist);
	if (ds->ds_dir)
	dsl_dir_async_rele(ds->ds_dir, ds);

	ASSERT(!list_link_active(&ds->ds_synced_link));

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (dsl_dataset_feature_is_active(ds, f))
	unload_zfeature(ds, f);
	}

	list_destroy(&ds->ds_prop_cbs);
	mutex_destroy(&ds->ds_lock);
	mutex_destroy(&ds->ds_opening_lock);
	mutex_destroy(&ds->ds_sendstream_lock);
	mutex_destroy(&ds->ds_remap_deadlist_lock);
	zfs_refcount_destroy(&ds->ds_longholds);
	rrw_destroy(&ds->ds_bp_rwlock);

	kmem_free(ds, sizeof (dsl_dataset_t));
	}

	int
	dsl_dataset_get_snapname(dsl_dataset_t *ds)
	{
	dsl_dataset_phys_t *headphys;
	int err;
	dmu_buf_t *headdbuf;
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	objset_t *mos = dp->dp_meta_objset;

	if (ds->ds_snapname[0])
	return (0);
	if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0)
	return (0);

	err = dmu_bonus_hold(mos, dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj,
	FTAG, &headdbuf);
	if (err != 0)
	return (err);
	headphys = headdbuf->db_data;
	err = zap_value_search(dp->dp_meta_objset,
	headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname);
	if (err != 0 && zfs_recover == B_TRUE) {
	err = 0;
	(void) snprintf(ds->ds_snapname, sizeof (ds->ds_snapname),
	"SNAPOBJ=%llu-ERR=%d",
	(unsigned long long)ds->ds_object, err);
	}
	dmu_buf_rele(headdbuf, FTAG);
	return (err);
	}

	int
	dsl_dataset_snap_lookup(dsl_dataset_t ds, const char name, uint64_t *value)
	{
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
	matchtype_t mt = 0;
	int err;

	if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
	mt = MT_NORMALIZE;

	err = zap_lookup_norm(mos, snapobj, name, 8, 1,
	value, mt, NULL, 0, NULL);
	if (err == ENOTSUP && (mt & MT_NORMALIZE))
	err = zap_lookup(mos, snapobj, name, 8, 1, value);
	return (err);
	}

	int
	dsl_dataset_snap_remove(dsl_dataset_t ds, const char name, dmu_tx_t *tx,
	boolean_t adj_cnt)
	{
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
	matchtype_t mt = 0;
	int err;

	dsl_dir_snap_cmtime_update(ds->ds_dir, tx);

	if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
	mt = MT_NORMALIZE;

	err = zap_remove_norm(mos, snapobj, name, mt, tx);
	if (err == ENOTSUP && (mt & MT_NORMALIZE))
	err = zap_remove(mos, snapobj, name, tx);

	if (err == 0 && adj_cnt)
	dsl_fs_ss_count_adjust(ds->ds_dir, -1,
	DD_FIELD_SNAPSHOT_COUNT, tx);

	return (err);
	}

	boolean_t
	dsl_dataset_try_add_ref(dsl_pool_t dp, dsl_dataset_t ds, const void *tag)
	{
	dmu_buf_t *dbuf = ds->ds_dbuf;
	boolean_t result = B_FALSE;

	if (dbuf != NULL && dmu_buf_try_add_ref(dbuf, dp->dp_meta_objset,
	ds->ds_object, DMU_BONUS_BLKID, tag)) {

	if (ds == dmu_buf_get_user(dbuf))
	result = B_TRUE;
	else
	dmu_buf_rele(dbuf, tag);
	}

	return (result);
	}

	int
	dsl_dataset_hold_obj(dsl_pool_t dp, uint64_t dsobj, const void tag,
	dsl_dataset_t **dsp)
	{
	objset_t *mos = dp->dp_meta_objset;
	dmu_buf_t *dbuf;
	dsl_dataset_t *ds;
	int err;
	dmu_object_info_t doi;

	ASSERT(dsl_pool_config_held(dp));

	err = dmu_bonus_hold(mos, dsobj, tag, &dbuf);
	if (err != 0)
	return (err);

	/* Make sure dsobj has the correct object type. */
	dmu_object_info_from_db(dbuf, &doi);
	if (doi.doi_bonus_type != DMU_OT_DSL_DATASET) {
	dmu_buf_rele(dbuf, tag);
	return (SET_ERROR(EINVAL));
	}

	ds = dmu_buf_get_user(dbuf);
	if (ds == NULL) {
	dsl_dataset_t *winner = NULL;

	ds = kmem_zalloc(sizeof (dsl_dataset_t), KM_SLEEP);
	ds->ds_dbuf = dbuf;
	ds->ds_object = dsobj;
	ds->ds_is_snapshot = dsl_dataset_phys(ds)->ds_num_children != 0;
	list_link_init(&ds->ds_synced_link);

	err = dsl_dir_hold_obj(dp, dsl_dataset_phys(ds)->ds_dir_obj,
	NULL, ds, &ds->ds_dir);
	if (err != 0) {
	kmem_free(ds, sizeof (dsl_dataset_t));
	dmu_buf_rele(dbuf, tag);
	return (err);
	}

	mutex_init(&ds->ds_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ds->ds_opening_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ds->ds_sendstream_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ds->ds_remap_deadlist_lock,
	NULL, MUTEX_DEFAULT, NULL);
	rrw_init(&ds->ds_bp_rwlock, B_FALSE);
	zfs_refcount_create(&ds->ds_longholds);

	bplist_create(&ds->ds_pending_deadlist);

	list_create(&ds->ds_sendstreams, sizeof (dmu_sendstatus_t),
	offsetof(dmu_sendstatus_t, dss_link));

	list_create(&ds->ds_prop_cbs, sizeof (dsl_prop_cb_record_t),
	offsetof(dsl_prop_cb_record_t, cbr_ds_node));

	if (doi.doi_type == DMU_OTN_ZAP_METADATA) {
	spa_feature_t f;

	for (f = 0; f < SPA_FEATURES; f++) {
	if (!(spa_feature_table[f].fi_flags &
	ZFEATURE_FLAG_PER_DATASET))
	continue;
	err = load_zfeature(mos, ds, f);
	}
	}

	if (!ds->ds_is_snapshot) {
	ds->ds_snapname[0] = '\0';
	if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
	err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj,
	ds, &ds->ds_prev);
	}
	+ if (err != 0)
	+ goto after_dsl_bookmark_fini;
	err = dsl_bookmark_init_ds(ds);
	} else {
	if (zfs_flags & ZFS_DEBUG_SNAPNAMES)
	err = dsl_dataset_get_snapname(ds);
	if (err == 0 &&
	dsl_dataset_phys(ds)->ds_userrefs_obj != 0) {
	err = zap_count(
	ds->ds_dir->dd_pool->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_userrefs_obj,
	&ds->ds_userrefs);
	}
	}

	if (err == 0 && !ds->ds_is_snapshot) {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
	&ds->ds_reserved);
	if (err == 0) {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_REFQUOTA),
	&ds->ds_quota);
	}
	} else {
	ds->ds_reserved = ds->ds_quota = 0;
	}

	if (err == 0 && ds->ds_dir->dd_crypto_obj != 0 &&
	ds->ds_is_snapshot &&
	zap_contains(mos, dsobj, DS_FIELD_IVSET_GUID) != 0) {
	dp->dp_spa->spa_errata =
	ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
	}

	dsl_deadlist_open(&ds->ds_deadlist,
	mos, dsl_dataset_phys(ds)->ds_deadlist_obj);
	uint64_t remap_deadlist_obj =
	dsl_dataset_get_remap_deadlist_object(ds);
	if (remap_deadlist_obj != 0) {
	dsl_deadlist_open(&ds->ds_remap_deadlist, mos,
	remap_deadlist_obj);
	}

	dmu_buf_init_user(&ds->ds_dbu, dsl_dataset_evict_sync,
	dsl_dataset_evict_async, &ds->ds_dbuf);
	if (err == 0)
	winner = dmu_buf_set_user_ie(dbuf, &ds->ds_dbu);

	if (err != 0 \|\| winner != NULL) {
	- bplist_destroy(&ds->ds_pending_deadlist);
	dsl_deadlist_close(&ds->ds_deadlist);
	if (dsl_deadlist_is_open(&ds->ds_remap_deadlist))
	dsl_deadlist_close(&ds->ds_remap_deadlist);
	dsl_bookmark_fini_ds(ds);
	+after_dsl_bookmark_fini:
	if (ds->ds_prev)
	dsl_dataset_rele(ds->ds_prev, ds);
	dsl_dir_rele(ds->ds_dir, ds);
	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (dsl_dataset_feature_is_active(ds, f))
	unload_zfeature(ds, f);
	}

	list_destroy(&ds->ds_prop_cbs);
	list_destroy(&ds->ds_sendstreams);
	+ bplist_destroy(&ds->ds_pending_deadlist);
	mutex_destroy(&ds->ds_lock);
	mutex_destroy(&ds->ds_opening_lock);
	mutex_destroy(&ds->ds_sendstream_lock);
	mutex_destroy(&ds->ds_remap_deadlist_lock);
	zfs_refcount_destroy(&ds->ds_longholds);
	rrw_destroy(&ds->ds_bp_rwlock);
	kmem_free(ds, sizeof (dsl_dataset_t));
	if (err != 0) {
	dmu_buf_rele(dbuf, tag);
	return (err);
	}
	ds = winner;
	} else {
	ds->ds_fsid_guid =
	unique_insert(dsl_dataset_phys(ds)->ds_fsid_guid);
	if (ds->ds_fsid_guid !=
	dsl_dataset_phys(ds)->ds_fsid_guid) {
	zfs_dbgmsg("ds_fsid_guid changed from "
	"%llx to %llx for pool %s dataset id %llu",
	(long long)
	dsl_dataset_phys(ds)->ds_fsid_guid,
	(long long)ds->ds_fsid_guid,
	spa_name(dp->dp_spa),
	(u_longlong_t)dsobj);
	}
	}
	}

	ASSERT3P(ds->ds_dbuf, ==, dbuf);
	ASSERT3P(dsl_dataset_phys(ds), ==, dbuf->db_data);
	ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0 \|\|
	spa_version(dp->dp_spa) < SPA_VERSION_ORIGIN \|\|
	dp->dp_origin_snap == NULL \|\| ds == dp->dp_origin_snap);
	*dsp = ds;

	return (0);
	}

	int
	dsl_dataset_create_key_mapping(dsl_dataset_t *ds)
	{
	dsl_dir_t *dd = ds->ds_dir;

	if (dd->dd_crypto_obj == 0)
	return (0);

	return (spa_keystore_create_mapping(dd->dd_pool->dp_spa,
	ds, ds, &ds->ds_key_mapping));
	}

	int
	dsl_dataset_hold_obj_flags(dsl_pool_t *dp, uint64_t dsobj,
	ds_hold_flags_t flags, const void tag, dsl_dataset_t *dsp)
	{
	int err;

	err = dsl_dataset_hold_obj(dp, dsobj, tag, dsp);
	if (err != 0)
	return (err);

	ASSERT3P(*dsp, !=, NULL);

	if (flags & DS_HOLD_FLAG_DECRYPT) {
	err = dsl_dataset_create_key_mapping(*dsp);
	if (err != 0)
	dsl_dataset_rele(*dsp, tag);
	}

	return (err);
	}

	int
	dsl_dataset_hold_flags(dsl_pool_t dp, const char name, ds_hold_flags_t flags,
	const void tag, dsl_dataset_t *dsp)
	{
	dsl_dir_t *dd;
	const char *snapname;
	uint64_t obj;
	int err = 0;
	dsl_dataset_t *ds;

	err = dsl_dir_hold(dp, name, FTAG, &dd, &snapname);
	if (err != 0)
	return (err);

	ASSERT(dsl_pool_config_held(dp));
	obj = dsl_dir_phys(dd)->dd_head_dataset_obj;
	if (obj != 0)
	err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag, &ds);
	else
	err = SET_ERROR(ENOENT);

	/* we may be looking for a snapshot */
	if (err == 0 && snapname != NULL) {
	dsl_dataset_t *snap_ds;

	if (*snapname++ != '@') {
	dsl_dataset_rele_flags(ds, flags, tag);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(ENOENT));
	}

	dprintf("looking for snapshot '%s'\n", snapname);
	err = dsl_dataset_snap_lookup(ds, snapname, &obj);
	if (err == 0) {
	err = dsl_dataset_hold_obj_flags(dp, obj, flags, tag,
	&snap_ds);
	}
	dsl_dataset_rele_flags(ds, flags, tag);

	if (err == 0) {
	mutex_enter(&snap_ds->ds_lock);
	if (snap_ds->ds_snapname[0] == 0)
	(void) strlcpy(snap_ds->ds_snapname, snapname,
	sizeof (snap_ds->ds_snapname));
	mutex_exit(&snap_ds->ds_lock);
	ds = snap_ds;
	}
	}
	if (err == 0)
	*dsp = ds;
	dsl_dir_rele(dd, FTAG);
	return (err);
	}

	int
	dsl_dataset_hold(dsl_pool_t dp, const char name, const void *tag,
	dsl_dataset_t **dsp)
	{
	return (dsl_dataset_hold_flags(dp, name, 0, tag, dsp));
	}

	static int
	dsl_dataset_own_obj_impl(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags,
	const void tag, boolean_t override, dsl_dataset_t *dsp)
	{
	int err = dsl_dataset_hold_obj_flags(dp, dsobj, flags, tag, dsp);
	if (err != 0)
	return (err);
	if (!dsl_dataset_tryown(*dsp, tag, override)) {
	dsl_dataset_rele_flags(*dsp, flags, tag);
	*dsp = NULL;
	return (SET_ERROR(EBUSY));
	}
	return (0);
	}


	int
	dsl_dataset_own_obj(dsl_pool_t *dp, uint64_t dsobj, ds_hold_flags_t flags,
	const void tag, dsl_dataset_t *dsp)
	{
	return (dsl_dataset_own_obj_impl(dp, dsobj, flags, tag, B_FALSE, dsp));
	}

	int
	dsl_dataset_own_obj_force(dsl_pool_t *dp, uint64_t dsobj,
	ds_hold_flags_t flags, const void tag, dsl_dataset_t *dsp)
	{
	return (dsl_dataset_own_obj_impl(dp, dsobj, flags, tag, B_TRUE, dsp));
	}

	static int
	dsl_dataset_own_impl(dsl_pool_t dp, const char name, ds_hold_flags_t flags,
	const void tag, boolean_t override, dsl_dataset_t *dsp)
	{
	int err = dsl_dataset_hold_flags(dp, name, flags, tag, dsp);
	if (err != 0)
	return (err);
	if (!dsl_dataset_tryown(*dsp, tag, override)) {
	dsl_dataset_rele_flags(*dsp, flags, tag);
	return (SET_ERROR(EBUSY));
	}
	return (0);
	}

	int
	dsl_dataset_own_force(dsl_pool_t dp, const char name, ds_hold_flags_t flags,
	const void tag, dsl_dataset_t *dsp)
	{
	return (dsl_dataset_own_impl(dp, name, flags, tag, B_TRUE, dsp));
	}

	int
	dsl_dataset_own(dsl_pool_t dp, const char name, ds_hold_flags_t flags,
	const void tag, dsl_dataset_t *dsp)
	{
	return (dsl_dataset_own_impl(dp, name, flags, tag, B_FALSE, dsp));
	}

	/*
	* See the comment above dsl_pool_hold() for details. In summary, a long
	* hold is used to prevent destruction of a dataset while the pool hold
	* is dropped, allowing other concurrent operations (e.g. spa_sync()).
	*
	* The dataset and pool must be held when this function is called. After it
	* is called, the pool hold may be released while the dataset is still held
	* and accessed.
	*/
	void
	dsl_dataset_long_hold(dsl_dataset_t ds, const void tag)
	{
	ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
	(void) zfs_refcount_add(&ds->ds_longholds, tag);
	}

	void
	dsl_dataset_long_rele(dsl_dataset_t ds, const void tag)
	{
	(void) zfs_refcount_remove(&ds->ds_longholds, tag);
	}

	/* Return B_TRUE if there are any long holds on this dataset. */
	boolean_t
	dsl_dataset_long_held(dsl_dataset_t *ds)
	{
	return (!zfs_refcount_is_zero(&ds->ds_longholds));
	}

	void
	dsl_dataset_name(dsl_dataset_t ds, char name)
	{
	if (ds == NULL) {
	(void) strlcpy(name, "mos", ZFS_MAX_DATASET_NAME_LEN);
	} else {
	dsl_dir_name(ds->ds_dir, name);
	VERIFY0(dsl_dataset_get_snapname(ds));
	if (ds->ds_snapname[0]) {
	VERIFY3U(strlcat(name, "@", ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	/*
	* We use a "recursive" mutex so that we
	* can call dprintf_ds() with ds_lock held.
	*/
	if (!MUTEX_HELD(&ds->ds_lock)) {
	mutex_enter(&ds->ds_lock);
	VERIFY3U(strlcat(name, ds->ds_snapname,
	ZFS_MAX_DATASET_NAME_LEN), <,
	ZFS_MAX_DATASET_NAME_LEN);
	mutex_exit(&ds->ds_lock);
	} else {
	VERIFY3U(strlcat(name, ds->ds_snapname,
	ZFS_MAX_DATASET_NAME_LEN), <,
	ZFS_MAX_DATASET_NAME_LEN);
	}
	}
	}
	}

	int
	dsl_dataset_namelen(dsl_dataset_t *ds)
	{
	VERIFY0(dsl_dataset_get_snapname(ds));
	mutex_enter(&ds->ds_lock);
	int len = strlen(ds->ds_snapname);
	mutex_exit(&ds->ds_lock);
	/* add '@' if ds is a snap */
	if (len > 0)
	len++;
	len += dsl_dir_namelen(ds->ds_dir);
	return (len);
	}

	void
	dsl_dataset_rele(dsl_dataset_t ds, const void tag)
	{
	dmu_buf_rele(ds->ds_dbuf, tag);
	}

	void
	dsl_dataset_remove_key_mapping(dsl_dataset_t *ds)
	{
	dsl_dir_t *dd = ds->ds_dir;

	if (dd == NULL \|\| dd->dd_crypto_obj == 0)
	return;

	(void) spa_keystore_remove_mapping(dd->dd_pool->dp_spa,
	ds->ds_object, ds);
	}

	void
	dsl_dataset_rele_flags(dsl_dataset_t *ds, ds_hold_flags_t flags,
	const void *tag)
	{
	if (flags & DS_HOLD_FLAG_DECRYPT)
	dsl_dataset_remove_key_mapping(ds);

	dsl_dataset_rele(ds, tag);
	}

	void
	dsl_dataset_disown(dsl_dataset_t ds, ds_hold_flags_t flags, const void tag)
	{
	ASSERT3P(ds->ds_owner, ==, tag);
	ASSERT(ds->ds_dbuf != NULL);

	mutex_enter(&ds->ds_lock);
	ds->ds_owner = NULL;
	mutex_exit(&ds->ds_lock);
	dsl_dataset_long_rele(ds, tag);
	dsl_dataset_rele_flags(ds, flags, tag);
	}

	boolean_t
	dsl_dataset_tryown(dsl_dataset_t ds, const void tag, boolean_t override)
	{
	boolean_t gotit = FALSE;

	ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
	mutex_enter(&ds->ds_lock);
	if (ds->ds_owner == NULL && (override \|\| !(DS_IS_INCONSISTENT(ds) \|\|
	(dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_REDACTED_DATASETS) &&
	!zfs_allow_redacted_dataset_mount)))) {
	ds->ds_owner = tag;
	dsl_dataset_long_hold(ds, tag);
	gotit = TRUE;
	}
	mutex_exit(&ds->ds_lock);
	return (gotit);
	}

	boolean_t
	dsl_dataset_has_owner(dsl_dataset_t *ds)
	{
	boolean_t rv;
	mutex_enter(&ds->ds_lock);
	rv = (ds->ds_owner != NULL);
	mutex_exit(&ds->ds_lock);
	return (rv);
	}

	static boolean_t
	zfeature_active(spa_feature_t f, void *arg)
	{
	switch (spa_feature_table[f].fi_type) {
	case ZFEATURE_TYPE_BOOLEAN: {
	boolean_t val = (boolean_t)(uintptr_t)arg;
	ASSERT(val == B_FALSE \|\| val == B_TRUE);
	return (val);
	}
	case ZFEATURE_TYPE_UINT64_ARRAY:
	/*
	* In this case, arg is a uint64_t array. The feature is active
	* if the array is non-null.
	*/
	return (arg != NULL);
	default:
	panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
	return (B_FALSE);
	}
	}

	boolean_t
	dsl_dataset_feature_is_active(dsl_dataset_t *ds, spa_feature_t f)
	{
	return (zfeature_active(f, ds->ds_feature[f]));
	}

	/*
	* The buffers passed out by this function are references to internal buffers;
	* they should not be freed by callers of this function, and they should not be
	* used after the dataset has been released.
	*/
	boolean_t
	dsl_dataset_get_uint64_array_feature(dsl_dataset_t *ds, spa_feature_t f,
	uint64_t outlength, uint64_t *outp)
	{
	VERIFY(spa_feature_table[f].fi_type & ZFEATURE_TYPE_UINT64_ARRAY);
	if (!dsl_dataset_feature_is_active(ds, f)) {
	return (B_FALSE);
	}
	struct feature_type_uint64_array_arg *ftuaa = ds->ds_feature[f];
	*outp = ftuaa->array;
	*outlength = ftuaa->length;
	return (B_TRUE);
	}

	void
	dsl_dataset_activate_feature(uint64_t dsobj, spa_feature_t f, void *arg,
	dmu_tx_t *tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset;
	uint64_t zero = 0;

	VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET);

	spa_feature_incr(spa, f, tx);
	dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);

	switch (spa_feature_table[f].fi_type) {
	case ZFEATURE_TYPE_BOOLEAN:
	ASSERT3S((boolean_t)(uintptr_t)arg, ==, B_TRUE);
	VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid,
	sizeof (zero), 1, &zero, tx));
	break;
	case ZFEATURE_TYPE_UINT64_ARRAY:
	{
	struct feature_type_uint64_array_arg *ftuaa = arg;
	VERIFY0(zap_add(mos, dsobj, spa_feature_table[f].fi_guid,
	sizeof (uint64_t), ftuaa->length, ftuaa->array, tx));
	break;
	}
	default:
	panic("Invalid zfeature type %d", spa_feature_table[f].fi_type);
	}
	}

	static void
	dsl_dataset_deactivate_feature_impl(dsl_dataset_t *ds, spa_feature_t f,
	dmu_tx_t *tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset;
	uint64_t dsobj = ds->ds_object;

	VERIFY(spa_feature_table[f].fi_flags & ZFEATURE_FLAG_PER_DATASET);

	VERIFY0(zap_remove(mos, dsobj, spa_feature_table[f].fi_guid, tx));
	spa_feature_decr(spa, f, tx);
	ds->ds_feature[f] = NULL;
	}

	void
	dsl_dataset_deactivate_feature(dsl_dataset_t ds, spa_feature_t f, dmu_tx_t tx)
	{
	unload_zfeature(ds, f);
	dsl_dataset_deactivate_feature_impl(ds, f, tx);
	}

	uint64_t
	dsl_dataset_create_sync_dd(dsl_dir_t dd, dsl_dataset_t origin,
	dsl_crypto_params_t dcp, uint64_t flags, dmu_tx_t tx)
	{
	dsl_pool_t *dp = dd->dd_pool;
	dmu_buf_t *dbuf;
	dsl_dataset_phys_t *dsphys;
	uint64_t dsobj;
	objset_t *mos = dp->dp_meta_objset;

	if (origin == NULL)
	origin = dp->dp_origin_snap;

	ASSERT(origin == NULL \|\| origin->ds_dir->dd_pool == dp);
	ASSERT(origin == NULL \|\| dsl_dataset_phys(origin)->ds_num_children > 0);
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(dsl_dir_phys(dd)->dd_head_dataset_obj == 0);

	dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
	DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
	VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
	dmu_buf_will_dirty(dbuf, tx);
	dsphys = dbuf->db_data;
	memset(dsphys, 0, sizeof (dsl_dataset_phys_t));
	dsphys->ds_dir_obj = dd->dd_object;
	dsphys->ds_flags = flags;
	dsphys->ds_fsid_guid = unique_create();
	(void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
	sizeof (dsphys->ds_guid));
	dsphys->ds_snapnames_zapobj =
	zap_create_norm(mos, U8_TEXTPREP_TOUPPER, DMU_OT_DSL_DS_SNAP_MAP,
	DMU_OT_NONE, 0, tx);
	dsphys->ds_creation_time = gethrestime_sec();
	dsphys->ds_creation_txg = tx->tx_txg == TXG_INITIAL ? 1 : tx->tx_txg;

	if (origin == NULL) {
	dsphys->ds_deadlist_obj = dsl_deadlist_alloc(mos, tx);
	} else {
	dsl_dataset_t ohds; / head of the origin snapshot */

	dsphys->ds_prev_snap_obj = origin->ds_object;
	dsphys->ds_prev_snap_txg =
	dsl_dataset_phys(origin)->ds_creation_txg;
	dsphys->ds_referenced_bytes =
	dsl_dataset_phys(origin)->ds_referenced_bytes;
	dsphys->ds_compressed_bytes =
	dsl_dataset_phys(origin)->ds_compressed_bytes;
	dsphys->ds_uncompressed_bytes =
	dsl_dataset_phys(origin)->ds_uncompressed_bytes;
	rrw_enter(&origin->ds_bp_rwlock, RW_READER, FTAG);
	dsphys->ds_bp = dsl_dataset_phys(origin)->ds_bp;
	rrw_exit(&origin->ds_bp_rwlock, FTAG);

	/*
	* Inherit flags that describe the dataset's contents
	* (INCONSISTENT) or properties (Case Insensitive).
	*/
	dsphys->ds_flags \|= dsl_dataset_phys(origin)->ds_flags &
	(DS_FLAG_INCONSISTENT \| DS_FLAG_CI_DATASET);

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (zfeature_active(f, origin->ds_feature[f])) {
	dsl_dataset_activate_feature(dsobj, f,
	origin->ds_feature[f], tx);
	}
	}

	dmu_buf_will_dirty(origin->ds_dbuf, tx);
	dsl_dataset_phys(origin)->ds_num_children++;

	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dir_phys(origin->ds_dir)->dd_head_dataset_obj,
	FTAG, &ohds));
	dsphys->ds_deadlist_obj = dsl_deadlist_clone(&ohds->ds_deadlist,
	dsphys->ds_prev_snap_txg, dsphys->ds_prev_snap_obj, tx);
	dsl_dataset_rele(ohds, FTAG);

	if (spa_version(dp->dp_spa) >= SPA_VERSION_NEXT_CLONES) {
	if (dsl_dataset_phys(origin)->ds_next_clones_obj == 0) {
	dsl_dataset_phys(origin)->ds_next_clones_obj =
	zap_create(mos,
	DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
	}
	VERIFY0(zap_add_int(mos,
	dsl_dataset_phys(origin)->ds_next_clones_obj,
	dsobj, tx));
	}

	dmu_buf_will_dirty(dd->dd_dbuf, tx);
	dsl_dir_phys(dd)->dd_origin_obj = origin->ds_object;
	if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
	if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) {
	dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
	dsl_dir_phys(origin->ds_dir)->dd_clones =
	zap_create(mos,
	DMU_OT_DSL_CLONES, DMU_OT_NONE, 0, tx);
	}
	VERIFY0(zap_add_int(mos,
	dsl_dir_phys(origin->ds_dir)->dd_clones,
	dsobj, tx));
	}
	}

	/* handle encryption */
	dsl_dataset_create_crypt_sync(dsobj, dd, origin, dcp, tx);

	if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
	dsphys->ds_flags \|= DS_FLAG_UNIQUE_ACCURATE;

	dmu_buf_rele(dbuf, FTAG);

	dmu_buf_will_dirty(dd->dd_dbuf, tx);
	dsl_dir_phys(dd)->dd_head_dataset_obj = dsobj;

	return (dsobj);
	}

	static void
	dsl_dataset_zero_zil(dsl_dataset_t ds, dmu_tx_t tx)
	{
	objset_t *os;

	VERIFY0(dmu_objset_from_ds(ds, &os));
	if (memcmp(&os->os_zil_header, &zero_zil, sizeof (zero_zil)) != 0) {
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	zio_t *zio;

	memset(&os->os_zil_header, 0, sizeof (os->os_zil_header));
	if (os->os_encrypted)
	os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;

	zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	dsl_dataset_sync(ds, zio, tx);
	VERIFY0(zio_wait(zio));

	/* dsl_dataset_sync_done will drop this reference. */
	dmu_buf_add_ref(ds->ds_dbuf, ds);
	dsl_dataset_sync_done(ds, tx);
	}
	}

	uint64_t
	dsl_dataset_create_sync(dsl_dir_t pdd, const char lastname,
	dsl_dataset_t origin, uint64_t flags, cred_t cr,
	dsl_crypto_params_t dcp, dmu_tx_t tx)
	{
	dsl_pool_t *dp = pdd->dd_pool;
	uint64_t dsobj, ddobj;
	dsl_dir_t *dd;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(lastname[0] != '@');
	/*
	* Filesystems will eventually have their origin set to dp_origin_snap,
	* but that's taken care of in dsl_dataset_create_sync_dd. When
	* creating a filesystem, this function is called with origin equal to
	* NULL.
	*/
	if (origin != NULL)
	ASSERT3P(origin, !=, dp->dp_origin_snap);

	ddobj = dsl_dir_create_sync(dp, pdd, lastname, tx);
	VERIFY0(dsl_dir_hold_obj(dp, ddobj, lastname, FTAG, &dd));

	dsobj = dsl_dataset_create_sync_dd(dd, origin, dcp,
	flags & ~DS_CREATE_FLAG_NODIRTY, tx);

	dsl_deleg_set_create_perms(dd, tx, cr);

	/*
	* If we are creating a clone and the livelist feature is enabled,
	* add the entry DD_FIELD_LIVELIST to ZAP.
	*/
	if (origin != NULL &&
	spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LIVELIST)) {
	objset_t *mos = dd->dd_pool->dp_meta_objset;
	dsl_dir_zapify(dd, tx);
	uint64_t obj = dsl_deadlist_alloc(mos, tx);
	VERIFY0(zap_add(mos, dd->dd_object, DD_FIELD_LIVELIST,
	sizeof (uint64_t), 1, &obj, tx));
	spa_feature_incr(dp->dp_spa, SPA_FEATURE_LIVELIST, tx);
	}

	/*
	* Since we're creating a new node we know it's a leaf, so we can
	* initialize the counts if the limit feature is active.
	*/
	if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) {
	uint64_t cnt = 0;
	objset_t *os = dd->dd_pool->dp_meta_objset;

	dsl_dir_zapify(dd, tx);
	VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT,
	sizeof (cnt), 1, &cnt, tx));
	VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT,
	sizeof (cnt), 1, &cnt, tx));
	}

	dsl_dir_rele(dd, FTAG);

	/*
	* If we are creating a clone, make sure we zero out any stale
	* data from the origin snapshots zil header.
	*/
	if (origin != NULL && !(flags & DS_CREATE_FLAG_NODIRTY)) {
	dsl_dataset_t *ds;

	VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
	dsl_dataset_zero_zil(ds, tx);
	dsl_dataset_rele(ds, FTAG);
	}

	return (dsobj);
	}

	/*
	* The unique space in the head dataset can be calculated by subtracting
	* the space used in the most recent snapshot, that is still being used
	* in this file system, from the space currently in use. To figure out
	* the space in the most recent snapshot still in use, we need to take
	* the total space used in the snapshot and subtract out the space that
	* has been freed up since the snapshot was taken.
	*/
	void
	dsl_dataset_recalc_head_uniq(dsl_dataset_t *ds)
	{
	uint64_t mrs_used;
	uint64_t dlused, dlcomp, dluncomp;

	ASSERT(!ds->ds_is_snapshot);

	if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0)
	mrs_used = dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes;
	else
	mrs_used = 0;

	dsl_deadlist_space(&ds->ds_deadlist, &dlused, &dlcomp, &dluncomp);

	ASSERT3U(dlused, <=, mrs_used);
	dsl_dataset_phys(ds)->ds_unique_bytes =
	dsl_dataset_phys(ds)->ds_referenced_bytes - (mrs_used - dlused);

	if (spa_version(ds->ds_dir->dd_pool->dp_spa) >=
	SPA_VERSION_UNIQUE_ACCURATE)
	dsl_dataset_phys(ds)->ds_flags \|= DS_FLAG_UNIQUE_ACCURATE;
	}

	void
	dsl_dataset_remove_from_next_clones(dsl_dataset_t *ds, uint64_t obj,
	dmu_tx_t *tx)
	{
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t count __maybe_unused;
	int err;

	ASSERT(dsl_dataset_phys(ds)->ds_num_children >= 2);
	err = zap_remove_int(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
	obj, tx);
	/*
	* The err should not be ENOENT, but a bug in a previous version
	* of the code could cause upgrade_clones_cb() to not set
	* ds_next_snap_obj when it should, leading to a missing entry.
	* If we knew that the pool was created after
	* SPA_VERSION_NEXT_CLONES, we could assert that it isn't
	* ENOENT. However, at least we can check that we don't have
	* too many entries in the next_clones_obj even after failing to
	* remove this one.
	*/
	if (err != ENOENT)
	VERIFY0(err);
	ASSERT0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
	&count));
	ASSERT3U(count, <=, dsl_dataset_phys(ds)->ds_num_children - 2);
	}


	blkptr_t *
	dsl_dataset_get_blkptr(dsl_dataset_t *ds)
	{
	return (&dsl_dataset_phys(ds)->ds_bp);
	}

	spa_t *
	dsl_dataset_get_spa(dsl_dataset_t *ds)
	{
	return (ds->ds_dir->dd_pool->dp_spa);
	}

	void
	dsl_dataset_dirty(dsl_dataset_t ds, dmu_tx_t tx)
	{
	dsl_pool_t *dp;

	if (ds == NULL) /* this is the meta-objset */
	return;

	ASSERT(ds->ds_objset != NULL);

	if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0)
	panic("dirtying snapshot!");

	/* Must not dirty a dataset in the same txg where it got snapshotted. */
	ASSERT3U(tx->tx_txg, >, dsl_dataset_phys(ds)->ds_prev_snap_txg);

	dp = ds->ds_dir->dd_pool;
	if (txg_list_add(&dp->dp_dirty_datasets, ds, tx->tx_txg)) {
	objset_t *os = ds->ds_objset;

	/* up the hold count until we can be written out */
	dmu_buf_add_ref(ds->ds_dbuf, ds);

	/* if this dataset is encrypted, grab a reference to the DCK */
	if (ds->ds_dir->dd_crypto_obj != 0 &&
	!os->os_raw_receive &&
	!os->os_next_write_raw[tx->tx_txg & TXG_MASK]) {
	ASSERT3P(ds->ds_key_mapping, !=, NULL);
	key_mapping_add_ref(ds->ds_key_mapping, ds);
	}
	}
	}

	static int
	dsl_dataset_snapshot_reserve_space(dsl_dataset_t ds, dmu_tx_t tx)
	{
	uint64_t asize;

	if (!dmu_tx_is_syncing(tx))
	return (0);

	/*
	* If there's an fs-only reservation, any blocks that might become
	* owned by the snapshot dataset must be accommodated by space
	* outside of the reservation.
	*/
	ASSERT(ds->ds_reserved == 0 \|\| DS_UNIQUE_IS_ACCURATE(ds));
	asize = MIN(dsl_dataset_phys(ds)->ds_unique_bytes, ds->ds_reserved);
	if (asize > dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE))
	return (SET_ERROR(ENOSPC));

	/*
	* Propagate any reserved space for this snapshot to other
	* snapshot checks in this sync group.
	*/
	if (asize > 0)
	dsl_dir_willuse_space(ds->ds_dir, asize, tx);

	return (0);
	}

	int
	dsl_dataset_snapshot_check_impl(dsl_dataset_t ds, const char snapname,
	dmu_tx_t tx, boolean_t recv, uint64_t cnt, cred_t cr, proc_t *proc)
	{
	int error;
	uint64_t value;

	ds->ds_trysnap_txg = tx->tx_txg;

	if (!dmu_tx_is_syncing(tx))
	return (0);

	/*
	* We don't allow multiple snapshots of the same txg. If there
	* is already one, try again.
	*/
	if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg)
	return (SET_ERROR(EAGAIN));

	/*
	* Check for conflicting snapshot name.
	*/
	error = dsl_dataset_snap_lookup(ds, snapname, &value);
	if (error == 0)
	return (SET_ERROR(EEXIST));
	if (error != ENOENT)
	return (error);

	/*
	* We don't allow taking snapshots of inconsistent datasets, such as
	* those into which we are currently receiving. However, if we are
	* creating this snapshot as part of a receive, this check will be
	* executed atomically with respect to the completion of the receive
	* itself but prior to the clearing of DS_FLAG_INCONSISTENT; in this
	* case we ignore this, knowing it will be fixed up for us shortly in
	* dmu_recv_end_sync().
	*/
	if (!recv && DS_IS_INCONSISTENT(ds))
	return (SET_ERROR(EBUSY));

	/*
	* Skip the check for temporary snapshots or if we have already checked
	* the counts in dsl_dataset_snapshot_check. This means we really only
	* check the count here when we're receiving a stream.
	*/
	if (cnt != 0 && cr != NULL) {
	error = dsl_fs_ss_limit_check(ds->ds_dir, cnt,
	ZFS_PROP_SNAPSHOT_LIMIT, NULL, cr, proc);
	if (error != 0)
	return (error);
	}

	error = dsl_dataset_snapshot_reserve_space(ds, tx);
	if (error != 0)
	return (error);

	return (0);
	}

	int
	dsl_dataset_snapshot_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_snapshot_arg_t *ddsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	nvpair_t *pair;
	int rv = 0;

	/*
	* Pre-compute how many total new snapshots will be created for each
	* level in the tree and below. This is needed for validating the
	* snapshot limit when either taking a recursive snapshot or when
	* taking multiple snapshots.
	*
	* The problem is that the counts are not actually adjusted when
	* we are checking, only when we finally sync. For a single snapshot,
	* this is easy, the count will increase by 1 at each node up the tree,
	* but its more complicated for the recursive/multiple snapshot case.
	*
	* The dsl_fs_ss_limit_check function does recursively check the count
	* at each level up the tree but since it is validating each snapshot
	* independently we need to be sure that we are validating the complete
	* count for the entire set of snapshots. We do this by rolling up the
	* counts for each component of the name into an nvlist and then
	* checking each of those cases with the aggregated count.
	*
	* This approach properly handles not only the recursive snapshot
	* case (where we get all of those on the ddsa_snaps list) but also
	* the sibling case (e.g. snapshot a/b and a/c so that we will also
	* validate the limit on 'a' using a count of 2).
	*
	* We validate the snapshot names in the third loop and only report
	* name errors once.
	*/
	if (dmu_tx_is_syncing(tx)) {
	char *nm;
	nvlist_t *cnt_track = NULL;
	cnt_track = fnvlist_alloc();

	nm = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	/* Rollup aggregated counts into the cnt_track list */
	for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
	char *pdelim;
	uint64_t val;

	(void) strlcpy(nm, nvpair_name(pair), MAXPATHLEN);
	pdelim = strchr(nm, '@');
	if (pdelim == NULL)
	continue;
	*pdelim = '\0';

	do {
	if (nvlist_lookup_uint64(cnt_track, nm,
	&val) == 0) {
	/* update existing entry */
	fnvlist_add_uint64(cnt_track, nm,
	val + 1);
	} else {
	/* add to list */
	fnvlist_add_uint64(cnt_track, nm, 1);
	}

	pdelim = strrchr(nm, '/');
	if (pdelim != NULL)
	*pdelim = '\0';
	} while (pdelim != NULL);
	}

	kmem_free(nm, MAXPATHLEN);

	/* Check aggregated counts at each level */
	for (pair = nvlist_next_nvpair(cnt_track, NULL);
	pair != NULL; pair = nvlist_next_nvpair(cnt_track, pair)) {
	int error = 0;
	char *name;
	uint64_t cnt = 0;
	dsl_dataset_t *ds;

	name = nvpair_name(pair);
	cnt = fnvpair_value_uint64(pair);
	ASSERT(cnt > 0);

	error = dsl_dataset_hold(dp, name, FTAG, &ds);
	if (error == 0) {
	error = dsl_fs_ss_limit_check(ds->ds_dir, cnt,
	ZFS_PROP_SNAPSHOT_LIMIT, NULL,
	ddsa->ddsa_cr, ddsa->ddsa_proc);
	dsl_dataset_rele(ds, FTAG);
	}

	if (error != 0) {
	if (ddsa->ddsa_errors != NULL)
	fnvlist_add_int32(ddsa->ddsa_errors,
	name, error);
	rv = error;
	/* only report one error for this check */
	break;
	}
	}
	nvlist_free(cnt_track);
	}

	for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
	pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
	int error = 0;
	dsl_dataset_t *ds;
	char name, atp = NULL;
	char dsname[ZFS_MAX_DATASET_NAME_LEN];

	name = nvpair_name(pair);
	if (strlen(name) >= ZFS_MAX_DATASET_NAME_LEN)
	error = SET_ERROR(ENAMETOOLONG);
	if (error == 0) {
	atp = strchr(name, '@');
	if (atp == NULL)
	error = SET_ERROR(EINVAL);
	if (error == 0)
	(void) strlcpy(dsname, name, atp - name + 1);
	}
	if (error == 0)
	error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
	if (error == 0) {
	/* passing 0/NULL skips dsl_fs_ss_limit_check */
	error = dsl_dataset_snapshot_check_impl(ds,
	atp + 1, tx, B_FALSE, 0, NULL, NULL);
	dsl_dataset_rele(ds, FTAG);
	}

	if (error != 0) {
	if (ddsa->ddsa_errors != NULL) {
	fnvlist_add_int32(ddsa->ddsa_errors,
	name, error);
	}
	rv = error;
	}
	}

	return (rv);
	}

	void
	dsl_dataset_snapshot_sync_impl(dsl_dataset_t ds, const char snapname,
	dmu_tx_t *tx)
	{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dmu_buf_t *dbuf;
	dsl_dataset_phys_t *dsphys;
	uint64_t dsobj, crtxg;
	objset_t *mos = dp->dp_meta_objset;
	static zil_header_t zero_zil __maybe_unused;
	objset_t *os __maybe_unused;

	ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));

	/*
	* If we are on an old pool, the zil must not be active, in which
	* case it will be zeroed. Usually zil_suspend() accomplishes this.
	*/
	ASSERT(spa_version(dmu_tx_pool(tx)->dp_spa) >= SPA_VERSION_FAST_SNAP \|\|
	dmu_objset_from_ds(ds, &os) != 0 \|\|
	memcmp(&os->os_phys->os_zil_header, &zero_zil,
	sizeof (zero_zil)) == 0);

	/* Should not snapshot a dirty dataset. */
	ASSERT(!txg_list_member(&ds->ds_dir->dd_pool->dp_dirty_datasets,
	ds, tx->tx_txg));

	dsl_fs_ss_count_adjust(ds->ds_dir, 1, DD_FIELD_SNAPSHOT_COUNT, tx);

	/*
	* The origin's ds_creation_txg has to be < TXG_INITIAL
	*/
	if (strcmp(snapname, ORIGIN_DIR_NAME) == 0)
	crtxg = 1;
	else
	crtxg = tx->tx_txg;

	dsobj = dmu_object_alloc(mos, DMU_OT_DSL_DATASET, 0,
	DMU_OT_DSL_DATASET, sizeof (dsl_dataset_phys_t), tx);
	VERIFY0(dmu_bonus_hold(mos, dsobj, FTAG, &dbuf));
	dmu_buf_will_dirty(dbuf, tx);
	dsphys = dbuf->db_data;
	memset(dsphys, 0, sizeof (dsl_dataset_phys_t));
	dsphys->ds_dir_obj = ds->ds_dir->dd_object;
	dsphys->ds_fsid_guid = unique_create();
	(void) random_get_pseudo_bytes((void*)&dsphys->ds_guid,
	sizeof (dsphys->ds_guid));
	dsphys->ds_prev_snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	dsphys->ds_prev_snap_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
	dsphys->ds_next_snap_obj = ds->ds_object;
	dsphys->ds_num_children = 1;
	dsphys->ds_creation_time = gethrestime_sec();
	dsphys->ds_creation_txg = crtxg;
	dsphys->ds_deadlist_obj = dsl_dataset_phys(ds)->ds_deadlist_obj;
	dsphys->ds_referenced_bytes = dsl_dataset_phys(ds)->ds_referenced_bytes;
	dsphys->ds_compressed_bytes = dsl_dataset_phys(ds)->ds_compressed_bytes;
	dsphys->ds_uncompressed_bytes =
	dsl_dataset_phys(ds)->ds_uncompressed_bytes;
	dsphys->ds_flags = dsl_dataset_phys(ds)->ds_flags;
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp;
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	dmu_buf_rele(dbuf, FTAG);

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (zfeature_active(f, ds->ds_feature[f])) {
	dsl_dataset_activate_feature(dsobj, f,
	ds->ds_feature[f], tx);
	}
	}

	ASSERT3U(ds->ds_prev != 0, ==,
	dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);
	if (ds->ds_prev) {
	uint64_t next_clones_obj =
	dsl_dataset_phys(ds->ds_prev)->ds_next_clones_obj;
	ASSERT(dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
	ds->ds_object \|\|
	dsl_dataset_phys(ds->ds_prev)->ds_num_children > 1);
	if (dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj ==
	ds->ds_object) {
	dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
	ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, ==,
	dsl_dataset_phys(ds->ds_prev)->ds_creation_txg);
	dsl_dataset_phys(ds->ds_prev)->ds_next_snap_obj = dsobj;
	} else if (next_clones_obj != 0) {
	dsl_dataset_remove_from_next_clones(ds->ds_prev,
	dsphys->ds_next_snap_obj, tx);
	VERIFY0(zap_add_int(mos,
	next_clones_obj, dsobj, tx));
	}
	}

	/*
	* If we have a reference-reservation on this dataset, we will
	* need to increase the amount of refreservation being charged
	* since our unique space is going to zero.
	*/
	if (ds->ds_reserved) {
	int64_t delta;
	ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
	delta = MIN(dsl_dataset_phys(ds)->ds_unique_bytes,
	ds->ds_reserved);
	dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV,
	delta, 0, 0, tx);
	}

	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_deadlist_obj =
	dsl_deadlist_clone(&ds->ds_deadlist, UINT64_MAX,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, tx);
	dsl_deadlist_close(&ds->ds_deadlist);
	dsl_deadlist_open(&ds->ds_deadlist, mos,
	dsl_dataset_phys(ds)->ds_deadlist_obj);
	dsl_deadlist_add_key(&ds->ds_deadlist,
	dsl_dataset_phys(ds)->ds_prev_snap_txg, tx);
	dsl_bookmark_snapshotted(ds, tx);

	if (dsl_dataset_remap_deadlist_exists(ds)) {
	uint64_t remap_deadlist_obj =
	dsl_dataset_get_remap_deadlist_object(ds);
	/*
	* Move the remap_deadlist to the snapshot. The head
	* will create a new remap deadlist on demand, from
	* dsl_dataset_block_remapped().
	*/
	dsl_dataset_unset_remap_deadlist_object(ds, tx);
	dsl_deadlist_close(&ds->ds_remap_deadlist);

	dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_REMAP_DEADLIST,
	sizeof (remap_deadlist_obj), 1, &remap_deadlist_obj, tx));
	}

	/*
	* Create a ivset guid for this snapshot if the dataset is
	* encrypted. This may be overridden by a raw receive. A
	* previous implementation of this code did not have this
	* field as part of the on-disk format for ZFS encryption
	* (see errata #4). As part of the remediation for this
	* issue, we ask the user to enable the bookmark_v2 feature
	* which is now a dependency of the encryption feature. We
	* use this as a heuristic to determine when the user has
	* elected to correct any datasets created with the old code.
	* As a result, we only do this step if the bookmark_v2
	* feature is enabled, which limits the number of states a
	* given pool / dataset can be in with regards to terms of
	* correcting the issue.
	*/
	if (ds->ds_dir->dd_crypto_obj != 0 &&
	spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_BOOKMARK_V2)) {
	uint64_t ivset_guid = unique_create();

	dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_IVSET_GUID,
	sizeof (ivset_guid), 1, &ivset_guid, tx));
	}

	ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_txg, <, tx->tx_txg);
	dsl_dataset_phys(ds)->ds_prev_snap_obj = dsobj;
	dsl_dataset_phys(ds)->ds_prev_snap_txg = crtxg;
	dsl_dataset_phys(ds)->ds_unique_bytes = 0;

	if (spa_version(dp->dp_spa) >= SPA_VERSION_UNIQUE_ACCURATE)
	dsl_dataset_phys(ds)->ds_flags \|= DS_FLAG_UNIQUE_ACCURATE;

	VERIFY0(zap_add(mos, dsl_dataset_phys(ds)->ds_snapnames_zapobj,
	snapname, 8, 1, &dsobj, tx));

	if (ds->ds_prev)
	dsl_dataset_rele(ds->ds_prev, ds);
	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, ds, &ds->ds_prev));

	dsl_scan_ds_snapshotted(ds, tx);

	dsl_dir_snap_cmtime_update(ds->ds_dir, tx);

	if (zfs_snapshot_history_enabled)
	spa_history_log_internal_ds(ds->ds_prev, "snapshot", tx, " ");
	}

	void
	dsl_dataset_snapshot_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_snapshot_arg_t *ddsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	nvpair_t *pair;

	for (pair = nvlist_next_nvpair(ddsa->ddsa_snaps, NULL);
	pair != NULL; pair = nvlist_next_nvpair(ddsa->ddsa_snaps, pair)) {
	dsl_dataset_t *ds;
	char name, atp;
	char dsname[ZFS_MAX_DATASET_NAME_LEN];

	name = nvpair_name(pair);
	atp = strchr(name, '@');
	(void) strlcpy(dsname, name, atp - name + 1);
	VERIFY0(dsl_dataset_hold(dp, dsname, FTAG, &ds));

	dsl_dataset_snapshot_sync_impl(ds, atp + 1, tx);
	if (ddsa->ddsa_props != NULL) {
	dsl_props_set_sync_impl(ds->ds_prev,
	ZPROP_SRC_LOCAL, ddsa->ddsa_props, tx);
	}
	dsl_dataset_rele(ds, FTAG);
	}
	}

	/*
	* The snapshots must all be in the same pool.
	* All-or-nothing: if there are any failures, nothing will be modified.
	*/
	int
	dsl_dataset_snapshot(nvlist_t snaps, nvlist_t props, nvlist_t *errors)
	{
	dsl_dataset_snapshot_arg_t ddsa;
	nvpair_t *pair;
	boolean_t needsuspend;
	int error;
	spa_t *spa;
	char *firstname;
	nvlist_t *suspended = NULL;

	pair = nvlist_next_nvpair(snaps, NULL);
	if (pair == NULL)
	return (0);
	firstname = nvpair_name(pair);

	error = spa_open(firstname, &spa, FTAG);
	if (error != 0)
	return (error);
	needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
	spa_close(spa, FTAG);

	if (needsuspend) {
	suspended = fnvlist_alloc();
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	char *snapname = nvpair_name(pair);
	char *atp;
	void *cookie;

	atp = strchr(snapname, '@');
	if (atp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}
	(void) strlcpy(fsname, snapname, atp - snapname + 1);

	error = zil_suspend(fsname, &cookie);
	if (error != 0)
	break;
	fnvlist_add_uint64(suspended, fsname,
	(uintptr_t)cookie);
	}
	}

	ddsa.ddsa_snaps = snaps;
	ddsa.ddsa_props = props;
	ddsa.ddsa_errors = errors;
	ddsa.ddsa_cr = CRED();
	ddsa.ddsa_proc = curproc;

	if (error == 0) {
	error = dsl_sync_task(firstname, dsl_dataset_snapshot_check,
	dsl_dataset_snapshot_sync, &ddsa,
	fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL);
	}

	if (suspended != NULL) {
	for (pair = nvlist_next_nvpair(suspended, NULL); pair != NULL;
	pair = nvlist_next_nvpair(suspended, pair)) {
	zil_resume((void *)(uintptr_t)
	fnvpair_value_uint64(pair));
	}
	fnvlist_free(suspended);
	}

	if (error == 0) {
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	zvol_create_minor(nvpair_name(pair));
	}
	}

	return (error);
	}

	typedef struct dsl_dataset_snapshot_tmp_arg {
	const char *ddsta_fsname;
	const char *ddsta_snapname;
	minor_t ddsta_cleanup_minor;
	const char *ddsta_htag;
	} dsl_dataset_snapshot_tmp_arg_t;

	static int
	dsl_dataset_snapshot_tmp_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	int error;

	error = dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds);
	if (error != 0)
	return (error);

	/* NULL cred means no limit check for tmp snapshot */
	error = dsl_dataset_snapshot_check_impl(ds, ddsta->ddsta_snapname,
	tx, B_FALSE, 0, NULL, NULL);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	error = dsl_dataset_user_hold_check_one(NULL, ddsta->ddsta_htag,
	B_TRUE, tx);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	static void
	dsl_dataset_snapshot_tmp_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_snapshot_tmp_arg_t *ddsta = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds = NULL;

	VERIFY0(dsl_dataset_hold(dp, ddsta->ddsta_fsname, FTAG, &ds));

	dsl_dataset_snapshot_sync_impl(ds, ddsta->ddsta_snapname, tx);
	dsl_dataset_user_hold_sync_one(ds->ds_prev, ddsta->ddsta_htag,
	ddsta->ddsta_cleanup_minor, gethrestime_sec(), tx);
	dsl_destroy_snapshot_sync_impl(ds->ds_prev, B_TRUE, tx);

	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dataset_snapshot_tmp(const char fsname, const char snapname,
	minor_t cleanup_minor, const char *htag)
	{
	dsl_dataset_snapshot_tmp_arg_t ddsta;
	int error;
	spa_t *spa;
	boolean_t needsuspend;
	void *cookie;

	ddsta.ddsta_fsname = fsname;
	ddsta.ddsta_snapname = snapname;
	ddsta.ddsta_cleanup_minor = cleanup_minor;
	ddsta.ddsta_htag = htag;

	error = spa_open(fsname, &spa, FTAG);
	if (error != 0)
	return (error);
	needsuspend = (spa_version(spa) < SPA_VERSION_FAST_SNAP);
	spa_close(spa, FTAG);

	if (needsuspend) {
	error = zil_suspend(fsname, &cookie);
	if (error != 0)
	return (error);
	}

	error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check,
	dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED);

	if (needsuspend)
	zil_resume(cookie);
	return (error);
	}

	void
	dsl_dataset_sync(dsl_dataset_t ds, zio_t zio, dmu_tx_t *tx)
	{
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(ds->ds_objset != NULL);
	ASSERT(dsl_dataset_phys(ds)->ds_next_snap_obj == 0);

	/*
	* in case we had to change ds_fsid_guid when we opened it,
	* sync it out now.
	*/
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid;

	if (ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] != 0) {
	VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_RESUME_OBJECT, 8, 1,
	&ds->ds_resume_object[tx->tx_txg & TXG_MASK], tx));
	VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_RESUME_OFFSET, 8, 1,
	&ds->ds_resume_offset[tx->tx_txg & TXG_MASK], tx));
	VERIFY0(zap_update(tx->tx_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_RESUME_BYTES, 8, 1,
	&ds->ds_resume_bytes[tx->tx_txg & TXG_MASK], tx));
	ds->ds_resume_object[tx->tx_txg & TXG_MASK] = 0;
	ds->ds_resume_offset[tx->tx_txg & TXG_MASK] = 0;
	ds->ds_resume_bytes[tx->tx_txg & TXG_MASK] = 0;
	}

	dmu_objset_sync(ds->ds_objset, zio, tx);

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (zfeature_active(f, ds->ds_feature_activation[f])) {
	if (zfeature_active(f, ds->ds_feature[f]))
	continue;
	dsl_dataset_activate_feature(ds->ds_object, f,
	ds->ds_feature_activation[f], tx);
	ds->ds_feature[f] = ds->ds_feature_activation[f];
	}
	}
	}

	/*
	* Check if the percentage of blocks shared between the clone and the
	* snapshot (as opposed to those that are clone only) is below a certain
	* threshold
	*/
	static boolean_t
	dsl_livelist_should_disable(dsl_dataset_t *ds)
	{
	uint64_t used, referenced;
	int percent_shared;

	used = dsl_dir_get_usedds(ds->ds_dir);
	referenced = dsl_get_referenced(ds);
	- ASSERT3U(referenced, >=, 0);
	- ASSERT3U(used, >=, 0);
	if (referenced == 0)
	return (B_FALSE);
	percent_shared = (100 * (referenced - used)) / referenced;
	if (percent_shared <= zfs_livelist_min_percent_shared)
	return (B_TRUE);
	return (B_FALSE);
	}

	/*
	* Check if it is possible to combine two livelist entries into one.
	* This is the case if the combined number of 'live' blkptrs (ALLOCs that
	* don't have a matching FREE) is under the maximum sublist size.
	* We check this by subtracting twice the total number of frees from the total
	* number of blkptrs. FREEs are counted twice because each FREE blkptr
	* will cancel out an ALLOC blkptr when the livelist is processed.
	*/
	static boolean_t
	dsl_livelist_should_condense(dsl_deadlist_entry_t *first,
	dsl_deadlist_entry_t *next)
	{
	uint64_t total_free = first->dle_bpobj.bpo_phys->bpo_num_freed +
	next->dle_bpobj.bpo_phys->bpo_num_freed;
	uint64_t total_entries = first->dle_bpobj.bpo_phys->bpo_num_blkptrs +
	next->dle_bpobj.bpo_phys->bpo_num_blkptrs;
	if ((total_entries - (2 * total_free)) < zfs_livelist_max_entries)
	return (B_TRUE);
	return (B_FALSE);
	}

	typedef struct try_condense_arg {
	spa_t *spa;
	dsl_dataset_t *ds;
	} try_condense_arg_t;

	/*
	* Iterate over the livelist entries, searching for a pair to condense.
	* A nonzero return value means stop, 0 means keep looking.
	*/
	static int
	dsl_livelist_try_condense(void arg, dsl_deadlist_entry_t first)
	{
	try_condense_arg_t *tca = arg;
	spa_t *spa = tca->spa;
	dsl_dataset_t *ds = tca->ds;
	dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist;
	dsl_deadlist_entry_t *next;

	/* The condense thread has not yet been created at import */
	if (spa->spa_livelist_condense_zthr == NULL)
	return (1);

	/* A condense is already in progress */
	if (spa->spa_to_condense.ds != NULL)
	return (1);

	next = AVL_NEXT(&ll->dl_tree, &first->dle_node);
	/* The livelist has only one entry - don't condense it */
	if (next == NULL)
	return (1);

	/* Next is the newest entry - don't condense it */
	if (AVL_NEXT(&ll->dl_tree, &next->dle_node) == NULL)
	return (1);

	/* This pair is not ready to condense but keep looking */
	if (!dsl_livelist_should_condense(first, next))
	return (0);

	/*
	* Add a ref to prevent the dataset from being evicted while
	* the condense zthr or synctask are running. Ref will be
	* released at the end of the condense synctask
	*/
	dmu_buf_add_ref(ds->ds_dbuf, spa);

	spa->spa_to_condense.ds = ds;
	spa->spa_to_condense.first = first;
	spa->spa_to_condense.next = next;
	spa->spa_to_condense.syncing = B_FALSE;
	spa->spa_to_condense.cancelled = B_FALSE;

	zthr_wakeup(spa->spa_livelist_condense_zthr);
	return (1);
	}

	static void
	dsl_flush_pending_livelist(dsl_dataset_t ds, dmu_tx_t tx)
	{
	dsl_dir_t *dd = ds->ds_dir;
	spa_t *spa = ds->ds_dir->dd_pool->dp_spa;
	dsl_deadlist_entry_t *last = dsl_deadlist_last(&dd->dd_livelist);

	/* Check if we need to add a new sub-livelist */
	if (last == NULL) {
	/* The livelist is empty */
	dsl_deadlist_add_key(&dd->dd_livelist,
	tx->tx_txg - 1, tx);
	} else if (spa_sync_pass(spa) == 1) {
	/*
	* Check if the newest entry is full. If it is, make a new one.
	* We only do this once per sync because we could overfill a
	* sublist in one sync pass and don't want to add another entry
	* for a txg that is already represented. This ensures that
	* blkptrs born in the same txg are stored in the same sublist.
	*/
	bpobj_t bpobj = last->dle_bpobj;
	uint64_t all = bpobj.bpo_phys->bpo_num_blkptrs;
	uint64_t free = bpobj.bpo_phys->bpo_num_freed;
	uint64_t alloc = all - free;
	if (alloc > zfs_livelist_max_entries) {
	dsl_deadlist_add_key(&dd->dd_livelist,
	tx->tx_txg - 1, tx);
	}
	}

	/* Insert each entry into the on-disk livelist */
	bplist_iterate(&dd->dd_pending_allocs,
	dsl_deadlist_insert_alloc_cb, &dd->dd_livelist, tx);
	bplist_iterate(&dd->dd_pending_frees,
	dsl_deadlist_insert_free_cb, &dd->dd_livelist, tx);

	/* Attempt to condense every pair of adjacent entries */
	try_condense_arg_t arg = {
	.spa = spa,
	.ds = ds
	};
	dsl_deadlist_iterate(&dd->dd_livelist, dsl_livelist_try_condense,
	&arg);
	}

	void
	dsl_dataset_sync_done(dsl_dataset_t ds, dmu_tx_t tx)
	{
	objset_t *os = ds->ds_objset;

	bplist_iterate(&ds->ds_pending_deadlist,
	dsl_deadlist_insert_alloc_cb, &ds->ds_deadlist, tx);

	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist)) {
	dsl_flush_pending_livelist(ds, tx);
	if (dsl_livelist_should_disable(ds)) {
	dsl_dir_remove_livelist(ds->ds_dir, tx, B_TRUE);
	}
	}

	dsl_bookmark_sync_done(ds, tx);

	multilist_destroy(&os->os_synced_dnodes);

	if (os->os_encrypted)
	os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_FALSE;
	else
	ASSERT0(os->os_next_write_raw[tx->tx_txg & TXG_MASK]);

	ASSERT(!dmu_objset_is_dirty(os, dmu_tx_get_txg(tx)));

	dmu_buf_rele(ds->ds_dbuf, ds);
	}

	int
	get_clones_stat_impl(dsl_dataset_t ds, nvlist_t val)
	{
	uint64_t count = 0;
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	zap_cursor_t zc;
	zap_attribute_t za;

	ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));

	/*
	* There may be missing entries in ds_next_clones_obj
	* due to a bug in a previous version of the code.
	* Only trust it if it has the right number of entries.
	*/
	if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) {
	VERIFY0(zap_count(mos, dsl_dataset_phys(ds)->ds_next_clones_obj,
	&count));
	}
	if (count != dsl_dataset_phys(ds)->ds_num_children - 1) {
	return (SET_ERROR(ENOENT));
	}
	for (zap_cursor_init(&zc, mos,
	dsl_dataset_phys(ds)->ds_next_clones_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	dsl_dataset_t *clone;
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,
	za.za_first_integer, FTAG, &clone));
	dsl_dir_name(clone->ds_dir, buf);
	fnvlist_add_boolean(val, buf);
	dsl_dataset_rele(clone, FTAG);
	}
	zap_cursor_fini(&zc);
	return (0);
	}

	void
	get_clones_stat(dsl_dataset_t ds, nvlist_t nv)
	{
	nvlist_t *propval = fnvlist_alloc();
	nvlist_t *val = fnvlist_alloc();

	if (get_clones_stat_impl(ds, val) == 0) {
	fnvlist_add_nvlist(propval, ZPROP_VALUE, val);
	fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_CLONES),
	propval);
	}

	nvlist_free(val);
	nvlist_free(propval);
	}

	static char *
	get_receive_resume_token_impl(dsl_dataset_t *ds)
	{
	if (!dsl_dataset_has_resume_receive_state(ds))
	return (NULL);

	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	char *str;
	void *packed;
	uint8_t *compressed;
	uint64_t val;
	nvlist_t *token_nv = fnvlist_alloc();
	size_t packed_size, compressed_size;

	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val) == 0) {
	fnvlist_add_uint64(token_nv, "fromguid", val);
	}
	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val) == 0) {
	fnvlist_add_uint64(token_nv, "object", val);
	}
	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val) == 0) {
	fnvlist_add_uint64(token_nv, "offset", val);
	}
	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_BYTES, sizeof (val), 1, &val) == 0) {
	fnvlist_add_uint64(token_nv, "bytes", val);
	}
	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val) == 0) {
	fnvlist_add_uint64(token_nv, "toguid", val);
	}
	char buf[MAXNAMELEN];
	if (zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TONAME, 1, sizeof (buf), buf) == 0) {
	fnvlist_add_string(token_nv, "toname", buf);
	}
	if (zap_contains(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_LARGEBLOCK) == 0) {
	fnvlist_add_boolean(token_nv, "largeblockok");
	}
	if (zap_contains(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_EMBEDOK) == 0) {
	fnvlist_add_boolean(token_nv, "embedok");
	}
	if (zap_contains(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_COMPRESSOK) == 0) {
	fnvlist_add_boolean(token_nv, "compressok");
	}
	if (zap_contains(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_RAWOK) == 0) {
	fnvlist_add_boolean(token_nv, "rawok");
	}
	if (dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	uint64_t num_redact_snaps = 0;
	uint64_t *redact_snaps = NULL;
	VERIFY3B(dsl_dataset_get_uint64_array_feature(ds,
	SPA_FEATURE_REDACTED_DATASETS, &num_redact_snaps,
	&redact_snaps), ==, B_TRUE);
	fnvlist_add_uint64_array(token_nv, "redact_snaps",
	redact_snaps, num_redact_snaps);
	}
	if (zap_contains(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS) == 0) {
	uint64_t num_redact_snaps = 0, int_size = 0;
	uint64_t *redact_snaps = NULL;
	VERIFY0(zap_length(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, &int_size,
	&num_redact_snaps));
	ASSERT3U(int_size, ==, sizeof (uint64_t));

	redact_snaps = kmem_alloc(int_size * num_redact_snaps,
	KM_SLEEP);
	VERIFY0(zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, int_size,
	num_redact_snaps, redact_snaps));
	fnvlist_add_uint64_array(token_nv, "book_redact_snaps",
	redact_snaps, num_redact_snaps);
	kmem_free(redact_snaps, int_size * num_redact_snaps);
	}
	packed = fnvlist_pack(token_nv, &packed_size);
	fnvlist_free(token_nv);
	compressed = kmem_alloc(packed_size, KM_SLEEP);

	compressed_size = gzip_compress(packed, compressed,
	packed_size, packed_size, 6);

	zio_cksum_t cksum;
	fletcher_4_native_varsize(compressed, compressed_size, &cksum);

	size_t alloc_size = compressed_size * 2 + 1;
	str = kmem_alloc(alloc_size, KM_SLEEP);
	for (int i = 0; i < compressed_size; i++) {
	size_t offset = i * 2;
	(void) snprintf(str + offset, alloc_size - offset,
	"%02x", compressed[i]);
	}
	str[compressed_size * 2] = '\0';
	char *propval = kmem_asprintf("%u-%llx-%llx-%s",
	ZFS_SEND_RESUME_TOKEN_VERSION,
	(longlong_t)cksum.zc_word[0],
	(longlong_t)packed_size, str);
	kmem_free(packed, packed_size);
	kmem_free(str, alloc_size);
	kmem_free(compressed, packed_size);
	return (propval);
	}

	/*
	* Returns a string that represents the receive resume state token. It should
	* be freed with strfree(). NULL is returned if no resume state is present.
	*/
	char *
	get_receive_resume_token(dsl_dataset_t *ds)
	{
	/*
	* A failed "newfs" (e.g. full) resumable receive leaves
	* the stats set on this dataset. Check here for the prop.
	*/
	char *token = get_receive_resume_token_impl(ds);
	if (token != NULL)
	return (token);
	/*
	* A failed incremental resumable receive leaves the
	* stats set on our child named "%recv". Check the child
	* for the prop.
	*/
	/* 6 extra bytes for /%recv */
	char name[ZFS_MAX_DATASET_NAME_LEN + 6];
	dsl_dataset_t *recv_ds;
	dsl_dataset_name(ds, name);
	if (strlcat(name, "/", sizeof (name)) < sizeof (name) &&
	strlcat(name, recv_clone_name, sizeof (name)) < sizeof (name) &&
	dsl_dataset_hold(ds->ds_dir->dd_pool, name, FTAG, &recv_ds) == 0) {
	token = get_receive_resume_token_impl(recv_ds);
	dsl_dataset_rele(recv_ds, FTAG);
	}
	return (token);
	}

	uint64_t
	dsl_get_refratio(dsl_dataset_t *ds)
	{
	uint64_t ratio = dsl_dataset_phys(ds)->ds_compressed_bytes == 0 ? 100 :
	(dsl_dataset_phys(ds)->ds_uncompressed_bytes * 100 /
	dsl_dataset_phys(ds)->ds_compressed_bytes);
	return (ratio);
	}

	uint64_t
	dsl_get_logicalreferenced(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_uncompressed_bytes);
	}

	uint64_t
	dsl_get_compressratio(dsl_dataset_t *ds)
	{
	if (ds->ds_is_snapshot) {
	return (dsl_get_refratio(ds));
	} else {
	dsl_dir_t *dd = ds->ds_dir;
	mutex_enter(&dd->dd_lock);
	uint64_t val = dsl_dir_get_compressratio(dd);
	mutex_exit(&dd->dd_lock);
	return (val);
	}
	}

	uint64_t
	dsl_get_used(dsl_dataset_t *ds)
	{
	if (ds->ds_is_snapshot) {
	return (dsl_dataset_phys(ds)->ds_unique_bytes);
	} else {
	dsl_dir_t *dd = ds->ds_dir;
	mutex_enter(&dd->dd_lock);
	uint64_t val = dsl_dir_get_used(dd);
	mutex_exit(&dd->dd_lock);
	return (val);
	}
	}

	uint64_t
	dsl_get_creation(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_creation_time);
	}

	uint64_t
	dsl_get_creationtxg(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_creation_txg);
	}

	uint64_t
	dsl_get_refquota(dsl_dataset_t *ds)
	{
	return (ds->ds_quota);
	}

	uint64_t
	dsl_get_refreservation(dsl_dataset_t *ds)
	{
	return (ds->ds_reserved);
	}

	uint64_t
	dsl_get_guid(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_guid);
	}

	uint64_t
	dsl_get_unique(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_unique_bytes);
	}

	uint64_t
	dsl_get_objsetid(dsl_dataset_t *ds)
	{
	return (ds->ds_object);
	}

	uint64_t
	dsl_get_userrefs(dsl_dataset_t *ds)
	{
	return (ds->ds_userrefs);
	}

	uint64_t
	dsl_get_defer_destroy(dsl_dataset_t *ds)
	{
	return (DS_IS_DEFER_DESTROY(ds) ? 1 : 0);
	}

	uint64_t
	dsl_get_referenced(dsl_dataset_t *ds)
	{
	return (dsl_dataset_phys(ds)->ds_referenced_bytes);
	}

	uint64_t
	dsl_get_numclones(dsl_dataset_t *ds)
	{
	ASSERT(ds->ds_is_snapshot);
	return (dsl_dataset_phys(ds)->ds_num_children - 1);
	}

	uint64_t
	dsl_get_inconsistent(dsl_dataset_t *ds)
	{
	return ((dsl_dataset_phys(ds)->ds_flags & DS_FLAG_INCONSISTENT) ?
	1 : 0);
	}

	uint64_t
	dsl_get_redacted(dsl_dataset_t *ds)
	{
	return (dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_REDACTED_DATASETS));
	}

	uint64_t
	dsl_get_available(dsl_dataset_t *ds)
	{
	uint64_t refdbytes = dsl_get_referenced(ds);
	uint64_t availbytes = dsl_dir_space_available(ds->ds_dir,
	NULL, 0, TRUE);
	if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) {
	availbytes +=
	ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes;
	}
	if (ds->ds_quota != 0) {
	/*
	* Adjust available bytes according to refquota
	*/
	if (refdbytes < ds->ds_quota) {
	availbytes = MIN(availbytes,
	ds->ds_quota - refdbytes);
	} else {
	availbytes = 0;
	}
	}
	return (availbytes);
	}

	int
	dsl_get_written(dsl_dataset_t ds, uint64_t written)
	{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dsl_dataset_t *prev;
	int err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
	if (err == 0) {
	uint64_t comp, uncomp;
	err = dsl_dataset_space_written(prev, ds, written,
	&comp, &uncomp);
	dsl_dataset_rele(prev, FTAG);
	}
	return (err);
	}

	/*
	* 'snap' should be a buffer of size ZFS_MAX_DATASET_NAME_LEN.
	*/
	int
	dsl_get_prev_snap(dsl_dataset_t ds, char snap)
	{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	if (ds->ds_prev != NULL && ds->ds_prev != dp->dp_origin_snap) {
	dsl_dataset_name(ds->ds_prev, snap);
	return (0);
	} else {
	return (SET_ERROR(ENOENT));
	}
	}

	void
	dsl_get_redact_snaps(dsl_dataset_t ds, nvlist_t propval)
	{
	uint64_t nsnaps;
	uint64_t *snaps;
	if (dsl_dataset_get_uint64_array_feature(ds,
	SPA_FEATURE_REDACTED_DATASETS, &nsnaps, &snaps)) {
	fnvlist_add_uint64_array(propval, ZPROP_VALUE, snaps,
	nsnaps);
	}
	}

	/*
	* Returns the mountpoint property and source for the given dataset in the value
	* and source buffers. The value buffer must be at least as large as MAXPATHLEN
	* and the source buffer as least as large a ZFS_MAX_DATASET_NAME_LEN.
	* Returns 0 on success and an error on failure.
	*/
	int
	dsl_get_mountpoint(dsl_dataset_t ds, const char dsname, char *value,
	char *source)
	{
	int error;
	dsl_pool_t *dp = ds->ds_dir->dd_pool;

	/* Retrieve the mountpoint value stored in the zap object */
	error = dsl_prop_get_ds(ds, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 1,
	ZAP_MAXVALUELEN, value, source);
	if (error != 0) {
	return (error);
	}

	/*
	* Process the dsname and source to find the full mountpoint string.
	* Can be skipped for 'legacy' or 'none'.
	*/
	if (value[0] == '/') {
	char *buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
	char *root = buf;
	const char *relpath;

	/*
	* If we inherit the mountpoint, even from a dataset
	* with a received value, the source will be the path of
	* the dataset we inherit from. If source is
	* ZPROP_SOURCE_VAL_RECVD, the received value is not
	* inherited.
	*/
	if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) {
	relpath = "";
	} else {
	ASSERT0(strncmp(dsname, source, strlen(source)));
	relpath = dsname + strlen(source);
	if (relpath[0] == '/')
	relpath++;
	}

	spa_altroot(dp->dp_spa, root, ZAP_MAXVALUELEN);

	/*
	* Special case an alternate root of '/'. This will
	* avoid having multiple leading slashes in the
	* mountpoint path.
	*/
	if (strcmp(root, "/") == 0)
	root++;

	/*
	* If the mountpoint is '/' then skip over this
	* if we are obtaining either an alternate root or
	* an inherited mountpoint.
	*/
	char *mnt = value;
	if (value[1] == '\0' && (root[0] != '\0' \|\|
	relpath[0] != '\0'))
	mnt = value + 1;

	+ mnt = kmem_strdup(mnt);
	+
	if (relpath[0] == '\0') {
	(void) snprintf(value, ZAP_MAXVALUELEN, "%s%s",
	root, mnt);
	} else {
	(void) snprintf(value, ZAP_MAXVALUELEN, "%s%s%s%s",
	root, mnt, relpath[0] == '@' ? "" : "/",
	relpath);
	}
	kmem_free(buf, ZAP_MAXVALUELEN);
	+ kmem_strfree(mnt);
	}

	return (0);
	}

	void
	dsl_dataset_stats(dsl_dataset_t ds, nvlist_t nv)
	{
	dsl_pool_t *dp __maybe_unused = ds->ds_dir->dd_pool;

	ASSERT(dsl_pool_config_held(dp));

	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRATIO,
	dsl_get_refratio(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALREFERENCED,
	dsl_get_logicalreferenced(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO,
	dsl_get_compressratio(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED,
	dsl_get_used(ds));

	if (ds->ds_is_snapshot) {
	get_clones_stat(ds, nv);
	} else {
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	if (dsl_get_prev_snap(ds, buf) == 0)
	dsl_prop_nvlist_add_string(nv, ZFS_PROP_PREV_SNAP,
	buf);
	dsl_dir_stats(ds->ds_dir, nv);
	}

	nvlist_t *propval = fnvlist_alloc();
	dsl_get_redact_snaps(ds, propval);
	fnvlist_add_nvlist(nv, zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS),
	propval);
	nvlist_free(propval);

	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_AVAILABLE,
	dsl_get_available(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFERENCED,
	dsl_get_referenced(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATION,
	dsl_get_creation(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_CREATETXG,
	dsl_get_creationtxg(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFQUOTA,
	dsl_get_refquota(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_REFRESERVATION,
	dsl_get_refreservation(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_GUID,
	dsl_get_guid(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_UNIQUE,
	dsl_get_unique(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_OBJSETID,
	dsl_get_objsetid(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERREFS,
	dsl_get_userrefs(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_DEFER_DESTROY,
	dsl_get_defer_destroy(ds));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_SNAPSHOTS_CHANGED,
	dsl_dir_snap_cmtime(ds->ds_dir).tv_sec);
	dsl_dataset_crypt_stats(ds, nv);

	if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
	uint64_t written;
	if (dsl_get_written(ds, &written) == 0) {
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_WRITTEN,
	written);
	}
	}

	if (!dsl_dataset_is_snapshot(ds)) {
	char *token = get_receive_resume_token(ds);
	if (token != NULL) {
	dsl_prop_nvlist_add_string(nv,
	ZFS_PROP_RECEIVE_RESUME_TOKEN, token);
	kmem_strfree(token);
	}
	}
	}

	void
	dsl_dataset_fast_stat(dsl_dataset_t ds, dmu_objset_stats_t stat)
	{
	dsl_pool_t *dp __maybe_unused = ds->ds_dir->dd_pool;
	ASSERT(dsl_pool_config_held(dp));

	stat->dds_creation_txg = dsl_get_creationtxg(ds);
	stat->dds_inconsistent = dsl_get_inconsistent(ds);
	stat->dds_guid = dsl_get_guid(ds);
	stat->dds_redacted = dsl_get_redacted(ds);
	stat->dds_origin[0] = '\0';
	if (ds->ds_is_snapshot) {
	stat->dds_is_snapshot = B_TRUE;
	stat->dds_num_clones = dsl_get_numclones(ds);
	} else {
	stat->dds_is_snapshot = B_FALSE;
	stat->dds_num_clones = 0;

	if (dsl_dir_is_clone(ds->ds_dir)) {
	dsl_dir_get_origin(ds->ds_dir, stat->dds_origin);
	}
	}
	}

	uint64_t
	dsl_dataset_fsid_guid(dsl_dataset_t *ds)
	{
	return (ds->ds_fsid_guid);
	}

	void
	dsl_dataset_space(dsl_dataset_t *ds,
	uint64_t refdbytesp, uint64_t availbytesp,
	uint64_t usedobjsp, uint64_t availobjsp)
	{
	*refdbytesp = dsl_dataset_phys(ds)->ds_referenced_bytes;
	*availbytesp = dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE);
	if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes)
	*availbytesp +=
	ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes;
	if (ds->ds_quota != 0) {
	/*
	* Adjust available bytes according to refquota
	*/
	if (*refdbytesp < ds->ds_quota)
	availbytesp = MIN(availbytesp,
	ds->ds_quota - *refdbytesp);
	else
	*availbytesp = 0;
	}
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	*usedobjsp = BP_GET_FILL(&dsl_dataset_phys(ds)->ds_bp);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	availobjsp = DN_MAX_OBJECT - usedobjsp;
	}

	boolean_t
	dsl_dataset_modified_since_snap(dsl_dataset_t ds, dsl_dataset_t snap)
	{
	dsl_pool_t *dp __maybe_unused = ds->ds_dir->dd_pool;
	uint64_t birth;

	ASSERT(dsl_pool_config_held(dp));
	if (snap == NULL)
	return (B_FALSE);
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	birth = dsl_dataset_get_blkptr(ds)->blk_birth;
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	if (birth > dsl_dataset_phys(snap)->ds_creation_txg) {
	objset_t os, os_snap;
	/*
	* It may be that only the ZIL differs, because it was
	* reset in the head. Don't count that as being
	* modified.
	*/
	if (dmu_objset_from_ds(ds, &os) != 0)
	return (B_TRUE);
	if (dmu_objset_from_ds(snap, &os_snap) != 0)
	return (B_TRUE);
	return (memcmp(&os->os_phys->os_meta_dnode,
	&os_snap->os_phys->os_meta_dnode,
	sizeof (os->os_phys->os_meta_dnode)) != 0);
	}
	return (B_FALSE);
	}

	static int
	dsl_dataset_rename_snapshot_check_impl(dsl_pool_t *dp,
	dsl_dataset_t hds, void arg)
	{
	(void) dp;
	dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
	int error;
	uint64_t val;

	error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
	if (error != 0) {
	/* ignore nonexistent snapshots */
	return (error == ENOENT ? 0 : error);
	}

	/* new name should not exist */
	error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_newsnapname, &val);
	if (error == 0)
	error = SET_ERROR(EEXIST);
	else if (error == ENOENT)
	error = 0;

	/* dataset name + 1 for the "@" + the new snapshot name must fit */
	if (dsl_dir_namelen(hds->ds_dir) + 1 +
	strlen(ddrsa->ddrsa_newsnapname) >= ZFS_MAX_DATASET_NAME_LEN)
	error = SET_ERROR(ENAMETOOLONG);

	return (error);
	}

	int
	dsl_dataset_rename_snapshot_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *hds;
	int error;

	error = dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds);
	if (error != 0)
	return (error);

	if (ddrsa->ddrsa_recursive) {
	error = dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
	dsl_dataset_rename_snapshot_check_impl, ddrsa,
	DS_FIND_CHILDREN);
	} else {
	error = dsl_dataset_rename_snapshot_check_impl(dp, hds, ddrsa);
	}
	dsl_dataset_rele(hds, FTAG);
	return (error);
	}

	static int
	dsl_dataset_rename_snapshot_sync_impl(dsl_pool_t *dp,
	dsl_dataset_t hds, void arg)
	{
	dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
	dsl_dataset_t *ds;
	uint64_t val;
	dmu_tx_t *tx = ddrsa->ddrsa_tx;
	int error;

	error = dsl_dataset_snap_lookup(hds, ddrsa->ddrsa_oldsnapname, &val);
	ASSERT(error == 0 \|\| error == ENOENT);
	if (error == ENOENT) {
	/* ignore nonexistent snapshots */
	return (0);
	}

	VERIFY0(dsl_dataset_hold_obj(dp, val, FTAG, &ds));

	/* log before we change the name */
	spa_history_log_internal_ds(ds, "rename", tx,
	"-> @%s", ddrsa->ddrsa_newsnapname);

	VERIFY0(dsl_dataset_snap_remove(hds, ddrsa->ddrsa_oldsnapname, tx,
	B_FALSE));
	mutex_enter(&ds->ds_lock);
	(void) strlcpy(ds->ds_snapname, ddrsa->ddrsa_newsnapname,
	sizeof (ds->ds_snapname));
	mutex_exit(&ds->ds_lock);
	VERIFY0(zap_add(dp->dp_meta_objset,
	dsl_dataset_phys(hds)->ds_snapnames_zapobj,
	ds->ds_snapname, 8, 1, &ds->ds_object, tx));
	zvol_rename_minors(dp->dp_spa, ddrsa->ddrsa_oldsnapname,
	ddrsa->ddrsa_newsnapname, B_TRUE);

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	void
	dsl_dataset_rename_snapshot_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_rename_snapshot_arg_t *ddrsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *hds = NULL;

	VERIFY0(dsl_dataset_hold(dp, ddrsa->ddrsa_fsname, FTAG, &hds));
	ddrsa->ddrsa_tx = tx;
	if (ddrsa->ddrsa_recursive) {
	VERIFY0(dmu_objset_find_dp(dp, hds->ds_dir->dd_object,
	dsl_dataset_rename_snapshot_sync_impl, ddrsa,
	DS_FIND_CHILDREN));
	} else {
	VERIFY0(dsl_dataset_rename_snapshot_sync_impl(dp, hds, ddrsa));
	}
	dsl_dataset_rele(hds, FTAG);
	}

	int
	dsl_dataset_rename_snapshot(const char *fsname,
	const char oldsnapname, const char newsnapname, boolean_t recursive)
	{
	dsl_dataset_rename_snapshot_arg_t ddrsa;

	ddrsa.ddrsa_fsname = fsname;
	ddrsa.ddrsa_oldsnapname = oldsnapname;
	ddrsa.ddrsa_newsnapname = newsnapname;
	ddrsa.ddrsa_recursive = recursive;

	return (dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check,
	dsl_dataset_rename_snapshot_sync, &ddrsa,
	1, ZFS_SPACE_CHECK_RESERVED));
	}

	/*
	* If we're doing an ownership handoff, we need to make sure that there is
	* only one long hold on the dataset. We're not allowed to change anything here
	* so we don't permanently release the long hold or regular hold here. We want
	* to do this only when syncing to avoid the dataset unexpectedly going away
	* when we release the long hold.
	*/
	static int
	dsl_dataset_handoff_check(dsl_dataset_t ds, void owner, dmu_tx_t *tx)
	{
	boolean_t held = B_FALSE;

	if (!dmu_tx_is_syncing(tx))
	return (0);

	dsl_dir_t *dd = ds->ds_dir;
	mutex_enter(&dd->dd_activity_lock);
	uint64_t holds = zfs_refcount_count(&ds->ds_longholds) -
	(owner != NULL ? 1 : 0);
	/*
	* The value of dd_activity_waiters can chance as soon as we drop the
	* lock, but we're fine with that; new waiters coming in or old
	* waiters leaving doesn't cause problems, since we're going to cancel
	* waiters later anyway. The goal of this check is to verify that no
	* non-waiters have long-holds, and all new long-holds will be
	* prevented because we're holding the pool config as writer.
	*/
	if (holds != dd->dd_activity_waiters)
	held = B_TRUE;
	mutex_exit(&dd->dd_activity_lock);

	if (held)
	return (SET_ERROR(EBUSY));

	return (0);
	}

	int
	dsl_dataset_rollback_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_rollback_arg_t *ddra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	int64_t unused_refres_delta;
	int error;

	error = dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds);
	if (error != 0)
	return (error);

	/* must not be a snapshot */
	if (ds->ds_is_snapshot) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/* must have a most recent snapshot */
	if (dsl_dataset_phys(ds)->ds_prev_snap_txg < TXG_INITIAL) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ESRCH));
	}

	/*
	* No rollback to a snapshot created in the current txg, because
	* the rollback may dirty the dataset and create blocks that are
	* not reachable from the rootbp while having a birth txg that
	* falls into the snapshot's range.
	*/
	if (dmu_tx_is_syncing(tx) &&
	dsl_dataset_phys(ds)->ds_prev_snap_txg >= tx->tx_txg) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EAGAIN));
	}

	/*
	* If the expected target snapshot is specified, then check that
	* the latest snapshot is it.
	*/
	if (ddra->ddra_tosnap != NULL) {
	dsl_dataset_t *snapds;

	/* Check if the target snapshot exists at all. */
	error = dsl_dataset_hold(dp, ddra->ddra_tosnap, FTAG, &snapds);
	if (error != 0) {
	/*
	* ESRCH is used to signal that the target snapshot does
	* not exist, while ENOENT is used to report that
	* the rolled back dataset does not exist.
	* ESRCH is also used to cover other cases where the
	* target snapshot is not related to the dataset being
	* rolled back such as being in a different pool.
	*/
	if (error == ENOENT \|\| error == EXDEV)
	error = SET_ERROR(ESRCH);
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	ASSERT(snapds->ds_is_snapshot);

	/* Check if the snapshot is the latest snapshot indeed. */
	if (snapds != ds->ds_prev) {
	/*
	* Distinguish between the case where the only problem
	* is intervening snapshots (EEXIST) vs the snapshot
	* not being a valid target for rollback (ESRCH).
	*/
	if (snapds->ds_dir == ds->ds_dir \|\|
	(dsl_dir_is_clone(ds->ds_dir) &&
	dsl_dir_phys(ds->ds_dir)->dd_origin_obj ==
	snapds->ds_object)) {
	error = SET_ERROR(EEXIST);
	} else {
	error = SET_ERROR(ESRCH);
	}
	dsl_dataset_rele(snapds, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	dsl_dataset_rele(snapds, FTAG);
	}

	/* must not have any bookmarks after the most recent snapshot */
	if (dsl_bookmark_latest_txg(ds) >
	dsl_dataset_phys(ds)->ds_prev_snap_txg) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EEXIST));
	}

	error = dsl_dataset_handoff_check(ds, ddra->ddra_owner, tx);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	/*
	* Check if the snap we are rolling back to uses more than
	* the refquota.
	*/
	if (ds->ds_quota != 0 &&
	dsl_dataset_phys(ds->ds_prev)->ds_referenced_bytes > ds->ds_quota) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EDQUOT));
	}

	/*
	* When we do the clone swap, we will temporarily use more space
	* due to the refreservation (the head will no longer have any
	* unique space, so the entire amount of the refreservation will need
	* to be free). We will immediately destroy the clone, freeing
	* this space, but the freeing happens over many txg's.
	*/
	unused_refres_delta = (int64_t)MIN(ds->ds_reserved,
	dsl_dataset_phys(ds)->ds_unique_bytes);

	if (unused_refres_delta > 0 &&
	unused_refres_delta >
	dsl_dir_space_available(ds->ds_dir, NULL, 0, TRUE)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOSPC));
	}

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	void
	dsl_dataset_rollback_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_rollback_arg_t *ddra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t ds, clone;
	uint64_t cloneobj;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];

	VERIFY0(dsl_dataset_hold(dp, ddra->ddra_fsname, FTAG, &ds));

	dsl_dataset_name(ds->ds_prev, namebuf);
	fnvlist_add_string(ddra->ddra_result, "target", namebuf);

	cloneobj = dsl_dataset_create_sync(ds->ds_dir, "%rollback",
	ds->ds_prev, DS_CREATE_FLAG_NODIRTY, kcred, NULL, tx);

	VERIFY0(dsl_dataset_hold_obj(dp, cloneobj, FTAG, &clone));

	dsl_dataset_clone_swap_sync_impl(clone, ds, tx);
	dsl_dataset_zero_zil(ds, tx);

	dsl_destroy_head_sync_impl(clone, tx);

	dsl_dataset_rele(clone, FTAG);
	dsl_dataset_rele(ds, FTAG);
	}

	/*
	* Rolls back the given filesystem or volume to the most recent snapshot.
	* The name of the most recent snapshot will be returned under key "target"
	* in the result nvlist.
	*
	* If owner != NULL:
	* - The existing dataset MUST be owned by the specified owner at entry
	* - Upon return, dataset will still be held by the same owner, whether we
	* succeed or not.
	*
	* This mode is required any time the existing filesystem is mounted. See
	* notes above zfs_suspend_fs() for further details.
	*/
	int
	dsl_dataset_rollback(const char fsname, const char tosnap, void *owner,
	nvlist_t *result)
	{
	dsl_dataset_rollback_arg_t ddra;

	ddra.ddra_fsname = fsname;
	ddra.ddra_tosnap = tosnap;
	ddra.ddra_owner = owner;
	ddra.ddra_result = result;

	return (dsl_sync_task(fsname, dsl_dataset_rollback_check,
	dsl_dataset_rollback_sync, &ddra,
	1, ZFS_SPACE_CHECK_RESERVED));
	}

	struct promotenode {
	list_node_t link;
	dsl_dataset_t *ds;
	};

	static int snaplist_space(list_t l, uint64_t mintxg, uint64_t spacep);
	static int promote_hold(dsl_dataset_promote_arg_t ddpa, dsl_pool_t dp,
	const void *tag);
	static void promote_rele(dsl_dataset_promote_arg_t ddpa, const void tag);

	int
	dsl_dataset_promote_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_promote_arg_t *ddpa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *hds;
	struct promotenode *snap;
	- dsl_dataset_t origin_ds, origin_head;
	int err;
	uint64_t unused;
	uint64_t ss_mv_cnt;
	size_t max_snap_len;
	boolean_t conflicting_snaps;

	err = promote_hold(ddpa, dp, FTAG);
	if (err != 0)
	return (err);

	hds = ddpa->ddpa_clone;
	max_snap_len = MAXNAMELEN - strlen(ddpa->ddpa_clonename) - 1;

	if (dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE) {
	promote_rele(ddpa, FTAG);
	return (SET_ERROR(EXDEV));
	}

	snap = list_head(&ddpa->shared_snaps);
	- origin_head = snap->ds;
	if (snap == NULL) {
	err = SET_ERROR(ENOENT);
	goto out;
	}
	- origin_ds = snap->ds;
	+ dsl_dataset_t *const origin_ds = snap->ds;

	/*
	* Encrypted clones share a DSL Crypto Key with their origin's dsl dir.
	* When doing a promote we must make sure the encryption root for
	* both the target and the target's origin does not change to avoid
	* needing to rewrap encryption keys
	*/
	err = dsl_dataset_promote_crypt_check(hds->ds_dir, origin_ds->ds_dir);
	if (err != 0)
	goto out;

	/*
	* Compute and check the amount of space to transfer. Since this is
	* so expensive, don't do the preliminary check.
	*/
	if (!dmu_tx_is_syncing(tx)) {
	promote_rele(ddpa, FTAG);
	return (0);
	}

	/* compute origin's new unique space */
	snap = list_tail(&ddpa->clone_snaps);
	ASSERT(snap != NULL);
	ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==,
	origin_ds->ds_object);
	dsl_deadlist_space_range(&snap->ds->ds_deadlist,
	dsl_dataset_phys(origin_ds)->ds_prev_snap_txg, UINT64_MAX,
	&ddpa->unique, &unused, &unused);

	/*
	* Walk the snapshots that we are moving
	*
	* Compute space to transfer. Consider the incremental changes
	* to used by each snapshot:
	* (my used) = (prev's used) + (blocks born) - (blocks killed)
	* So each snapshot gave birth to:
	* (blocks born) = (my used) - (prev's used) + (blocks killed)
	* So a sequence would look like:
	* (uN - u(N-1) + kN) + ... + (u1 - u0 + k1) + (u0 - 0 + k0)
	* Which simplifies to:
	* uN + kN + kN-1 + ... + k1 + k0
	* Note however, if we stop before we reach the ORIGIN we get:
	* uN + kN + kN-1 + ... + kM - uM-1
	*/
	conflicting_snaps = B_FALSE;
	ss_mv_cnt = 0;
	ddpa->used = dsl_dataset_phys(origin_ds)->ds_referenced_bytes;
	ddpa->comp = dsl_dataset_phys(origin_ds)->ds_compressed_bytes;
	ddpa->uncomp = dsl_dataset_phys(origin_ds)->ds_uncompressed_bytes;
	for (snap = list_head(&ddpa->shared_snaps); snap;
	snap = list_next(&ddpa->shared_snaps, snap)) {
	uint64_t val, dlused, dlcomp, dluncomp;
	dsl_dataset_t *ds = snap->ds;

	ss_mv_cnt++;

	/*
	* If there are long holds, we won't be able to evict
	* the objset.
	*/
	if (dsl_dataset_long_held(ds)) {
	err = SET_ERROR(EBUSY);
	goto out;
	}

	/* Check that the snapshot name does not conflict */
	VERIFY0(dsl_dataset_get_snapname(ds));
	if (strlen(ds->ds_snapname) >= max_snap_len) {
	err = SET_ERROR(ENAMETOOLONG);
	goto out;
	}
	err = dsl_dataset_snap_lookup(hds, ds->ds_snapname, &val);
	if (err == 0) {
	fnvlist_add_boolean(ddpa->err_ds,
	snap->ds->ds_snapname);
	conflicting_snaps = B_TRUE;
	} else if (err != ENOENT) {
	goto out;
	}

	/* The very first snapshot does not have a deadlist */
	if (dsl_dataset_phys(ds)->ds_prev_snap_obj == 0)
	continue;

	dsl_deadlist_space(&ds->ds_deadlist,
	&dlused, &dlcomp, &dluncomp);
	ddpa->used += dlused;
	ddpa->comp += dlcomp;
	ddpa->uncomp += dluncomp;
	}

	/*
	* Check that bookmarks that are being transferred don't have
	* name conflicts.
	*/
	- for (dsl_bookmark_node_t *dbn = avl_first(&origin_head->ds_bookmarks);
	+ for (dsl_bookmark_node_t *dbn = avl_first(&origin_ds->ds_bookmarks);
	dbn != NULL && dbn->dbn_phys.zbm_creation_txg <=
	dsl_dataset_phys(origin_ds)->ds_creation_txg;
	- dbn = AVL_NEXT(&origin_head->ds_bookmarks, dbn)) {
	+ dbn = AVL_NEXT(&origin_ds->ds_bookmarks, dbn)) {
	if (strlen(dbn->dbn_name) >= max_snap_len) {
	err = SET_ERROR(ENAMETOOLONG);
	goto out;
	}
	zfs_bookmark_phys_t bm;
	err = dsl_bookmark_lookup_impl(ddpa->ddpa_clone,
	dbn->dbn_name, &bm);

	if (err == 0) {
	fnvlist_add_boolean(ddpa->err_ds, dbn->dbn_name);
	conflicting_snaps = B_TRUE;
	} else if (err == ESRCH) {
	err = 0;
	} else if (err != 0) {
	goto out;
	}
	}

	/*
	* In order to return the full list of conflicting snapshots, we check
	* whether there was a conflict after traversing all of them.
	*/
	if (conflicting_snaps) {
	err = SET_ERROR(EEXIST);
	goto out;
	}

	/*
	* If we are a clone of a clone then we never reached ORIGIN,
	* so we need to subtract out the clone origin's used space.
	*/
	if (ddpa->origin_origin) {
	ddpa->used -=
	dsl_dataset_phys(ddpa->origin_origin)->ds_referenced_bytes;
	ddpa->comp -=
	dsl_dataset_phys(ddpa->origin_origin)->ds_compressed_bytes;
	ddpa->uncomp -=
	dsl_dataset_phys(ddpa->origin_origin)->
	ds_uncompressed_bytes;
	}

	/* Check that there is enough space and limit headroom here */
	err = dsl_dir_transfer_possible(origin_ds->ds_dir, hds->ds_dir,
	0, ss_mv_cnt, ddpa->used, ddpa->cr, ddpa->proc);
	if (err != 0)
	goto out;

	/*
	* Compute the amounts of space that will be used by snapshots
	* after the promotion (for both origin and clone). For each,
	* it is the amount of space that will be on all of their
	* deadlists (that was not born before their new origin).
	*/
	if (dsl_dir_phys(hds->ds_dir)->dd_flags & DD_FLAG_USED_BREAKDOWN) {
	uint64_t space;

	/*
	* Note, typically this will not be a clone of a clone,
	* so dd_origin_txg will be < TXG_INITIAL, so
	* these snaplist_space() -> dsl_deadlist_space_range()
	* calls will be fast because they do not have to
	* iterate over all bps.
	*/
	snap = list_head(&ddpa->origin_snaps);
	if (snap == NULL) {
	err = SET_ERROR(ENOENT);
	goto out;
	}
	err = snaplist_space(&ddpa->shared_snaps,
	snap->ds->ds_dir->dd_origin_txg, &ddpa->cloneusedsnap);
	if (err != 0)
	goto out;

	err = snaplist_space(&ddpa->clone_snaps,
	snap->ds->ds_dir->dd_origin_txg, &space);
	if (err != 0)
	goto out;
	ddpa->cloneusedsnap += space;
	}
	if (dsl_dir_phys(origin_ds->ds_dir)->dd_flags &
	DD_FLAG_USED_BREAKDOWN) {
	err = snaplist_space(&ddpa->origin_snaps,
	dsl_dataset_phys(origin_ds)->ds_creation_txg,
	&ddpa->originusedsnap);
	if (err != 0)
	goto out;
	}

	out:
	promote_rele(ddpa, FTAG);
	return (err);
	}

	void
	dsl_dataset_promote_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_promote_arg_t *ddpa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *hds;
	struct promotenode *snap;
	dsl_dataset_t *origin_ds;
	dsl_dataset_t *origin_head;
	dsl_dir_t *dd;
	dsl_dir_t *odd = NULL;
	uint64_t oldnext_obj;
	int64_t delta;

	ASSERT(nvlist_empty(ddpa->err_ds));

	VERIFY0(promote_hold(ddpa, dp, FTAG));
	hds = ddpa->ddpa_clone;

	ASSERT0(dsl_dataset_phys(hds)->ds_flags & DS_FLAG_NOPROMOTE);

	snap = list_head(&ddpa->shared_snaps);
	origin_ds = snap->ds;
	dd = hds->ds_dir;

	snap = list_head(&ddpa->origin_snaps);
	origin_head = snap->ds;

	/*
	* We need to explicitly open odd, since origin_ds's dd will be
	* changing.
	*/
	VERIFY0(dsl_dir_hold_obj(dp, origin_ds->ds_dir->dd_object,
	NULL, FTAG, &odd));

	dsl_dataset_promote_crypt_sync(hds->ds_dir, odd, tx);

	/* change origin's next snap */
	dmu_buf_will_dirty(origin_ds->ds_dbuf, tx);
	oldnext_obj = dsl_dataset_phys(origin_ds)->ds_next_snap_obj;
	snap = list_tail(&ddpa->clone_snaps);
	ASSERT3U(dsl_dataset_phys(snap->ds)->ds_prev_snap_obj, ==,
	origin_ds->ds_object);
	dsl_dataset_phys(origin_ds)->ds_next_snap_obj = snap->ds->ds_object;

	/* change the origin's next clone */
	if (dsl_dataset_phys(origin_ds)->ds_next_clones_obj) {
	dsl_dataset_remove_from_next_clones(origin_ds,
	snap->ds->ds_object, tx);
	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dataset_phys(origin_ds)->ds_next_clones_obj,
	oldnext_obj, tx));
	}

	/* change origin */
	dmu_buf_will_dirty(dd->dd_dbuf, tx);
	ASSERT3U(dsl_dir_phys(dd)->dd_origin_obj, ==, origin_ds->ds_object);
	dsl_dir_phys(dd)->dd_origin_obj = dsl_dir_phys(odd)->dd_origin_obj;
	dd->dd_origin_txg = origin_head->ds_dir->dd_origin_txg;
	dmu_buf_will_dirty(odd->dd_dbuf, tx);
	dsl_dir_phys(odd)->dd_origin_obj = origin_ds->ds_object;
	origin_head->ds_dir->dd_origin_txg =
	dsl_dataset_phys(origin_ds)->ds_creation_txg;

	/* change dd_clone entries */
	if (spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
	VERIFY0(zap_remove_int(dp->dp_meta_objset,
	dsl_dir_phys(odd)->dd_clones, hds->ds_object, tx));
	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones,
	hds->ds_object, tx));

	VERIFY0(zap_remove_int(dp->dp_meta_objset,
	dsl_dir_phys(ddpa->origin_origin->ds_dir)->dd_clones,
	origin_head->ds_object, tx));
	if (dsl_dir_phys(dd)->dd_clones == 0) {
	dsl_dir_phys(dd)->dd_clones =
	zap_create(dp->dp_meta_objset, DMU_OT_DSL_CLONES,
	DMU_OT_NONE, 0, tx);
	}
	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_clones, origin_head->ds_object, tx));
	}

	/*
	* Move bookmarks to this dir.
	*/
	dsl_bookmark_node_t *dbn_next;
	for (dsl_bookmark_node_t *dbn = avl_first(&origin_head->ds_bookmarks);
	dbn != NULL && dbn->dbn_phys.zbm_creation_txg <=
	dsl_dataset_phys(origin_ds)->ds_creation_txg;
	dbn = dbn_next) {
	dbn_next = AVL_NEXT(&origin_head->ds_bookmarks, dbn);

	avl_remove(&origin_head->ds_bookmarks, dbn);
	VERIFY0(zap_remove(dp->dp_meta_objset,
	origin_head->ds_bookmarks_obj, dbn->dbn_name, tx));

	dsl_bookmark_node_add(hds, dbn, tx);
	}

	dsl_bookmark_next_changed(hds, origin_ds, tx);

	/* move snapshots to this dir */
	for (snap = list_head(&ddpa->shared_snaps); snap;
	snap = list_next(&ddpa->shared_snaps, snap)) {
	dsl_dataset_t *ds = snap->ds;

	/*
	* Property callbacks are registered to a particular
	* dsl_dir. Since ours is changing, evict the objset
	* so that they will be unregistered from the old dsl_dir.
	*/
	if (ds->ds_objset) {
	dmu_objset_evict(ds->ds_objset);
	ds->ds_objset = NULL;
	}

	/* move snap name entry */
	VERIFY0(dsl_dataset_get_snapname(ds));
	VERIFY0(dsl_dataset_snap_remove(origin_head,
	ds->ds_snapname, tx, B_TRUE));
	VERIFY0(zap_add(dp->dp_meta_objset,
	dsl_dataset_phys(hds)->ds_snapnames_zapobj, ds->ds_snapname,
	8, 1, &ds->ds_object, tx));
	dsl_fs_ss_count_adjust(hds->ds_dir, 1,
	DD_FIELD_SNAPSHOT_COUNT, tx);

	/* change containing dsl_dir */
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	ASSERT3U(dsl_dataset_phys(ds)->ds_dir_obj, ==, odd->dd_object);
	dsl_dataset_phys(ds)->ds_dir_obj = dd->dd_object;
	ASSERT3P(ds->ds_dir, ==, odd);
	dsl_dir_rele(ds->ds_dir, ds);
	VERIFY0(dsl_dir_hold_obj(dp, dd->dd_object,
	NULL, ds, &ds->ds_dir));

	/* move any clone references */
	if (dsl_dataset_phys(ds)->ds_next_clones_obj &&
	spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) {
	zap_cursor_t zc;
	zap_attribute_t za;

	for (zap_cursor_init(&zc, dp->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_next_clones_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	dsl_dataset_t *cnds;
	uint64_t o;

	if (za.za_first_integer == oldnext_obj) {
	/*
	* We've already moved the
	* origin's reference.
	*/
	continue;
	}

	VERIFY0(dsl_dataset_hold_obj(dp,
	za.za_first_integer, FTAG, &cnds));
	o = dsl_dir_phys(cnds->ds_dir)->
	dd_head_dataset_obj;

	VERIFY0(zap_remove_int(dp->dp_meta_objset,
	dsl_dir_phys(odd)->dd_clones, o, tx));
	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_clones, o, tx));
	dsl_dataset_rele(cnds, FTAG);
	}
	zap_cursor_fini(&zc);
	}

	ASSERT(!dsl_prop_hascb(ds));
	}

	/*
	* Change space accounting.
	* Note, pa->*usedsnap and dd_used_breakdown[SNAP] will either
	* both be valid, or both be 0 (resulting in delta == 0). This
	* is true for each of {clone,origin} independently.
	*/

	delta = ddpa->cloneusedsnap -
	dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP];
	ASSERT3S(delta, >=, 0);
	ASSERT3U(ddpa->used, >=, delta);
	dsl_dir_diduse_space(dd, DD_USED_SNAP, delta, 0, 0, tx);
	dsl_dir_diduse_space(dd, DD_USED_HEAD,
	ddpa->used - delta, ddpa->comp, ddpa->uncomp, tx);

	delta = ddpa->originusedsnap -
	dsl_dir_phys(odd)->dd_used_breakdown[DD_USED_SNAP];
	ASSERT3S(delta, <=, 0);
	ASSERT3U(ddpa->used, >=, -delta);
	dsl_dir_diduse_space(odd, DD_USED_SNAP, delta, 0, 0, tx);
	dsl_dir_diduse_space(odd, DD_USED_HEAD,
	-ddpa->used - delta, -ddpa->comp, -ddpa->uncomp, tx);

	dsl_dataset_phys(origin_ds)->ds_unique_bytes = ddpa->unique;

	/*
	* Since livelists are specific to a clone's origin txg, they
	* are no longer accurate. Destroy the livelist from the clone being
	* promoted. If the origin dataset is a clone, destroy its livelist
	* as well.
	*/
	dsl_dir_remove_livelist(dd, tx, B_TRUE);
	dsl_dir_remove_livelist(odd, tx, B_TRUE);

	/* log history record */
	spa_history_log_internal_ds(hds, "promote", tx, " ");

	dsl_dir_rele(odd, FTAG);
	promote_rele(ddpa, FTAG);

	/*
	* Transfer common error blocks from old head to new head.
	*/
	if (spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_HEAD_ERRLOG)) {
	uint64_t old_head = origin_head->ds_object;
	uint64_t new_head = hds->ds_object;
	spa_swap_errlog(dp->dp_spa, new_head, old_head, tx);
	}
	}

	/*
	* Make a list of dsl_dataset_t's for the snapshots between first_obj
	* (exclusive) and last_obj (inclusive). The list will be in reverse
	* order (last_obj will be the list_head()). If first_obj == 0, do all
	* snapshots back to this dataset's origin.
	*/
	static int
	snaplist_make(dsl_pool_t *dp,
	uint64_t first_obj, uint64_t last_obj, list_t l, const void tag)
	{
	uint64_t obj = last_obj;

	list_create(l, sizeof (struct promotenode),
	offsetof(struct promotenode, link));

	while (obj != first_obj) {
	dsl_dataset_t *ds;
	struct promotenode *snap;
	int err;

	err = dsl_dataset_hold_obj(dp, obj, tag, &ds);
	ASSERT(err != ENOENT);
	if (err != 0)
	return (err);

	if (first_obj == 0)
	first_obj = dsl_dir_phys(ds->ds_dir)->dd_origin_obj;

	snap = kmem_alloc(sizeof (*snap), KM_SLEEP);
	snap->ds = ds;
	list_insert_tail(l, snap);
	obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	}

	return (0);
	}

	static int
	snaplist_space(list_t l, uint64_t mintxg, uint64_t spacep)
	{
	struct promotenode *snap;

	*spacep = 0;
	for (snap = list_head(l); snap; snap = list_next(l, snap)) {
	uint64_t used, comp, uncomp;
	dsl_deadlist_space_range(&snap->ds->ds_deadlist,
	mintxg, UINT64_MAX, &used, &comp, &uncomp);
	*spacep += used;
	}
	return (0);
	}

	static void
	snaplist_destroy(list_t l, const void tag)
	{
	struct promotenode *snap;

	if (l == NULL \|\| !list_link_active(&l->list_head))
	return;

	while ((snap = list_tail(l)) != NULL) {
	list_remove(l, snap);
	dsl_dataset_rele(snap->ds, tag);
	kmem_free(snap, sizeof (*snap));
	}
	list_destroy(l);
	}

	static int
	promote_hold(dsl_dataset_promote_arg_t ddpa, dsl_pool_t dp, const void *tag)
	{
	int error;
	dsl_dir_t *dd;
	struct promotenode *snap;

	error = dsl_dataset_hold(dp, ddpa->ddpa_clonename, tag,
	&ddpa->ddpa_clone);
	if (error != 0)
	return (error);
	dd = ddpa->ddpa_clone->ds_dir;

	if (ddpa->ddpa_clone->ds_is_snapshot \|\|
	!dsl_dir_is_clone(dd)) {
	dsl_dataset_rele(ddpa->ddpa_clone, tag);
	return (SET_ERROR(EINVAL));
	}

	error = snaplist_make(dp, 0, dsl_dir_phys(dd)->dd_origin_obj,
	&ddpa->shared_snaps, tag);
	if (error != 0)
	goto out;

	error = snaplist_make(dp, 0, ddpa->ddpa_clone->ds_object,
	&ddpa->clone_snaps, tag);
	if (error != 0)
	goto out;

	snap = list_head(&ddpa->shared_snaps);
	ASSERT3U(snap->ds->ds_object, ==, dsl_dir_phys(dd)->dd_origin_obj);
	error = snaplist_make(dp, dsl_dir_phys(dd)->dd_origin_obj,
	dsl_dir_phys(snap->ds->ds_dir)->dd_head_dataset_obj,
	&ddpa->origin_snaps, tag);
	if (error != 0)
	goto out;

	if (dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj != 0) {
	error = dsl_dataset_hold_obj(dp,
	dsl_dir_phys(snap->ds->ds_dir)->dd_origin_obj,
	tag, &ddpa->origin_origin);
	if (error != 0)
	goto out;
	}
	out:
	if (error != 0)
	promote_rele(ddpa, tag);
	return (error);
	}

	static void
	promote_rele(dsl_dataset_promote_arg_t ddpa, const void tag)
	{
	snaplist_destroy(&ddpa->shared_snaps, tag);
	snaplist_destroy(&ddpa->clone_snaps, tag);
	snaplist_destroy(&ddpa->origin_snaps, tag);
	if (ddpa->origin_origin != NULL)
	dsl_dataset_rele(ddpa->origin_origin, tag);
	dsl_dataset_rele(ddpa->ddpa_clone, tag);
	}

	/*
	* Promote a clone.
	*
	* If it fails due to a conflicting snapshot name, "conflsnap" will be filled
	* in with the name. (It must be at least ZFS_MAX_DATASET_NAME_LEN bytes long.)
	*/
	int
	dsl_dataset_promote(const char name, char conflsnap)
	{
	dsl_dataset_promote_arg_t ddpa = { 0 };
	uint64_t numsnaps;
	int error;
	nvpair_t *snap_pair;
	objset_t *os;

	/*
	* We will modify space proportional to the number of
	* snapshots. Compute numsnaps.
	*/
	error = dmu_objset_hold(name, FTAG, &os);
	if (error != 0)
	return (error);
	error = zap_count(dmu_objset_pool(os)->dp_meta_objset,
	dsl_dataset_phys(dmu_objset_ds(os))->ds_snapnames_zapobj,
	&numsnaps);
	dmu_objset_rele(os, FTAG);
	if (error != 0)
	return (error);

	ddpa.ddpa_clonename = name;
	ddpa.err_ds = fnvlist_alloc();
	ddpa.cr = CRED();
	ddpa.proc = curproc;

	error = dsl_sync_task(name, dsl_dataset_promote_check,
	dsl_dataset_promote_sync, &ddpa,
	2 + numsnaps, ZFS_SPACE_CHECK_RESERVED);

	/*
	* Return the first conflicting snapshot found.
	*/
	snap_pair = nvlist_next_nvpair(ddpa.err_ds, NULL);
	if (snap_pair != NULL && conflsnap != NULL)
	(void) strlcpy(conflsnap, nvpair_name(snap_pair),
	ZFS_MAX_DATASET_NAME_LEN);

	fnvlist_free(ddpa.err_ds);
	return (error);
	}

	int
	dsl_dataset_clone_swap_check_impl(dsl_dataset_t *clone,
	dsl_dataset_t origin_head, boolean_t force, void owner, dmu_tx_t *tx)
	{
	/*
	* "slack" factor for received datasets with refquota set on them.
	* See the bottom of this function for details on its use.
	*/
	uint64_t refquota_slack = (uint64_t)DMU_MAX_ACCESS *
	spa_asize_inflation;
	int64_t unused_refres_delta;

	/* they should both be heads */
	if (clone->ds_is_snapshot \|\|
	origin_head->ds_is_snapshot)
	return (SET_ERROR(EINVAL));

	/* if we are not forcing, the branch point should be just before them */
	if (!force && clone->ds_prev != origin_head->ds_prev)
	return (SET_ERROR(EINVAL));

	/* clone should be the clone (unless they are unrelated) */
	if (clone->ds_prev != NULL &&
	clone->ds_prev != clone->ds_dir->dd_pool->dp_origin_snap &&
	origin_head->ds_dir != clone->ds_prev->ds_dir)
	return (SET_ERROR(EINVAL));

	/* the clone should be a child of the origin */
	if (clone->ds_dir->dd_parent != origin_head->ds_dir)
	return (SET_ERROR(EINVAL));

	/* origin_head shouldn't be modified unless 'force' */
	if (!force &&
	dsl_dataset_modified_since_snap(origin_head, origin_head->ds_prev))
	return (SET_ERROR(ETXTBSY));

	/* origin_head should have no long holds (e.g. is not mounted) */
	if (dsl_dataset_handoff_check(origin_head, owner, tx))
	return (SET_ERROR(EBUSY));

	/* check amount of any unconsumed refreservation */
	unused_refres_delta =
	(int64_t)MIN(origin_head->ds_reserved,
	dsl_dataset_phys(origin_head)->ds_unique_bytes) -
	(int64_t)MIN(origin_head->ds_reserved,
	dsl_dataset_phys(clone)->ds_unique_bytes);

	if (unused_refres_delta > 0 &&
	unused_refres_delta >
	dsl_dir_space_available(origin_head->ds_dir, NULL, 0, TRUE))
	return (SET_ERROR(ENOSPC));

	/*
	* The clone can't be too much over the head's refquota.
	*
	* To ensure that the entire refquota can be used, we allow one
	* transaction to exceed the refquota. Therefore, this check
	* needs to also allow for the space referenced to be more than the
	* refquota. The maximum amount of space that one transaction can use
	* on disk is DMU_MAX_ACCESS * spa_asize_inflation. Allowing this
	* overage ensures that we are able to receive a filesystem that
	* exceeds the refquota on the source system.
	*
	* So that overage is the refquota_slack we use below.
	*/
	if (origin_head->ds_quota != 0 &&
	dsl_dataset_phys(clone)->ds_referenced_bytes >
	origin_head->ds_quota + refquota_slack)
	return (SET_ERROR(EDQUOT));

	return (0);
	}

	static void
	dsl_dataset_swap_remap_deadlists(dsl_dataset_t *clone,
	dsl_dataset_t origin, dmu_tx_t tx)
	{
	uint64_t clone_remap_dl_obj, origin_remap_dl_obj;
	dsl_pool_t *dp = dmu_tx_pool(tx);

	ASSERT(dsl_pool_sync_context(dp));

	clone_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(clone);
	origin_remap_dl_obj = dsl_dataset_get_remap_deadlist_object(origin);

	if (clone_remap_dl_obj != 0) {
	dsl_deadlist_close(&clone->ds_remap_deadlist);
	dsl_dataset_unset_remap_deadlist_object(clone, tx);
	}
	if (origin_remap_dl_obj != 0) {
	dsl_deadlist_close(&origin->ds_remap_deadlist);
	dsl_dataset_unset_remap_deadlist_object(origin, tx);
	}

	if (clone_remap_dl_obj != 0) {
	dsl_dataset_set_remap_deadlist_object(origin,
	clone_remap_dl_obj, tx);
	dsl_deadlist_open(&origin->ds_remap_deadlist,
	dp->dp_meta_objset, clone_remap_dl_obj);
	}
	if (origin_remap_dl_obj != 0) {
	dsl_dataset_set_remap_deadlist_object(clone,
	origin_remap_dl_obj, tx);
	dsl_deadlist_open(&clone->ds_remap_deadlist,
	dp->dp_meta_objset, origin_remap_dl_obj);
	}
	}

	void
	dsl_dataset_clone_swap_sync_impl(dsl_dataset_t *clone,
	dsl_dataset_t origin_head, dmu_tx_t tx)
	{
	dsl_pool_t *dp = dmu_tx_pool(tx);
	int64_t unused_refres_delta;

	ASSERT(clone->ds_reserved == 0);
	/*
	* NOTE: On DEBUG kernels there could be a race between this and
	* the check function if spa_asize_inflation is adjusted...
	*/
	ASSERT(origin_head->ds_quota == 0 \|\|
	dsl_dataset_phys(clone)->ds_unique_bytes <= origin_head->ds_quota +
	DMU_MAX_ACCESS * spa_asize_inflation);
	ASSERT3P(clone->ds_prev, ==, origin_head->ds_prev);

	dsl_dir_cancel_waiters(origin_head->ds_dir);

	/*
	* Swap per-dataset feature flags.
	*/
	for (spa_feature_t f = 0; f < SPA_FEATURES; f++) {
	if (!(spa_feature_table[f].fi_flags &
	ZFEATURE_FLAG_PER_DATASET)) {
	ASSERT(!dsl_dataset_feature_is_active(clone, f));
	ASSERT(!dsl_dataset_feature_is_active(origin_head, f));
	continue;
	}

	boolean_t clone_inuse = dsl_dataset_feature_is_active(clone, f);
	void *clone_feature = clone->ds_feature[f];
	boolean_t origin_head_inuse =
	dsl_dataset_feature_is_active(origin_head, f);
	void *origin_head_feature = origin_head->ds_feature[f];

	if (clone_inuse)
	dsl_dataset_deactivate_feature_impl(clone, f, tx);
	if (origin_head_inuse)
	dsl_dataset_deactivate_feature_impl(origin_head, f, tx);

	if (clone_inuse) {
	dsl_dataset_activate_feature(origin_head->ds_object, f,
	clone_feature, tx);
	origin_head->ds_feature[f] = clone_feature;
	}
	if (origin_head_inuse) {
	dsl_dataset_activate_feature(clone->ds_object, f,
	origin_head_feature, tx);
	clone->ds_feature[f] = origin_head_feature;
	}
	}

	dmu_buf_will_dirty(clone->ds_dbuf, tx);
	dmu_buf_will_dirty(origin_head->ds_dbuf, tx);

	if (clone->ds_objset != NULL) {
	dmu_objset_evict(clone->ds_objset);
	clone->ds_objset = NULL;
	}

	if (origin_head->ds_objset != NULL) {
	dmu_objset_evict(origin_head->ds_objset);
	origin_head->ds_objset = NULL;
	}

	unused_refres_delta =
	(int64_t)MIN(origin_head->ds_reserved,
	dsl_dataset_phys(origin_head)->ds_unique_bytes) -
	(int64_t)MIN(origin_head->ds_reserved,
	dsl_dataset_phys(clone)->ds_unique_bytes);

	/*
	* Reset origin's unique bytes.
	*/
	{
	dsl_dataset_t *origin = clone->ds_prev;
	uint64_t comp, uncomp;

	dmu_buf_will_dirty(origin->ds_dbuf, tx);
	dsl_deadlist_space_range(&clone->ds_deadlist,
	dsl_dataset_phys(origin)->ds_prev_snap_txg, UINT64_MAX,
	&dsl_dataset_phys(origin)->ds_unique_bytes, &comp, &uncomp);
	}

	/* swap blkptrs */
	{
	rrw_enter(&clone->ds_bp_rwlock, RW_WRITER, FTAG);
	rrw_enter(&origin_head->ds_bp_rwlock, RW_WRITER, FTAG);
	blkptr_t tmp;
	tmp = dsl_dataset_phys(origin_head)->ds_bp;
	dsl_dataset_phys(origin_head)->ds_bp =
	dsl_dataset_phys(clone)->ds_bp;
	dsl_dataset_phys(clone)->ds_bp = tmp;
	rrw_exit(&origin_head->ds_bp_rwlock, FTAG);
	rrw_exit(&clone->ds_bp_rwlock, FTAG);
	}

	/* set dd__bytes /
	{
	int64_t dused, dcomp, duncomp;
	uint64_t cdl_used, cdl_comp, cdl_uncomp;
	uint64_t odl_used, odl_comp, odl_uncomp;

	ASSERT3U(dsl_dir_phys(clone->ds_dir)->
	dd_used_breakdown[DD_USED_SNAP], ==, 0);

	dsl_deadlist_space(&clone->ds_deadlist,
	&cdl_used, &cdl_comp, &cdl_uncomp);
	dsl_deadlist_space(&origin_head->ds_deadlist,
	&odl_used, &odl_comp, &odl_uncomp);

	dused = dsl_dataset_phys(clone)->ds_referenced_bytes +
	cdl_used -
	(dsl_dataset_phys(origin_head)->ds_referenced_bytes +
	odl_used);
	dcomp = dsl_dataset_phys(clone)->ds_compressed_bytes +
	cdl_comp -
	(dsl_dataset_phys(origin_head)->ds_compressed_bytes +
	odl_comp);
	duncomp = dsl_dataset_phys(clone)->ds_uncompressed_bytes +
	cdl_uncomp -
	(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes +
	odl_uncomp);

	dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_HEAD,
	dused, dcomp, duncomp, tx);
	dsl_dir_diduse_space(clone->ds_dir, DD_USED_HEAD,
	-dused, -dcomp, -duncomp, tx);

	/*
	* The difference in the space used by snapshots is the
	* difference in snapshot space due to the head's
	* deadlist (since that's the only thing that's
	* changing that affects the snapused).
	*/
	dsl_deadlist_space_range(&clone->ds_deadlist,
	origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
	&cdl_used, &cdl_comp, &cdl_uncomp);
	dsl_deadlist_space_range(&origin_head->ds_deadlist,
	origin_head->ds_dir->dd_origin_txg, UINT64_MAX,
	&odl_used, &odl_comp, &odl_uncomp);
	dsl_dir_transfer_space(origin_head->ds_dir, cdl_used - odl_used,
	DD_USED_HEAD, DD_USED_SNAP, tx);
	}

	/* swap ds__bytes /
	SWITCH64(dsl_dataset_phys(origin_head)->ds_referenced_bytes,
	dsl_dataset_phys(clone)->ds_referenced_bytes);
	SWITCH64(dsl_dataset_phys(origin_head)->ds_compressed_bytes,
	dsl_dataset_phys(clone)->ds_compressed_bytes);
	SWITCH64(dsl_dataset_phys(origin_head)->ds_uncompressed_bytes,
	dsl_dataset_phys(clone)->ds_uncompressed_bytes);
	SWITCH64(dsl_dataset_phys(origin_head)->ds_unique_bytes,
	dsl_dataset_phys(clone)->ds_unique_bytes);

	/* apply any parent delta for change in unconsumed refreservation */
	dsl_dir_diduse_space(origin_head->ds_dir, DD_USED_REFRSRV,
	unused_refres_delta, 0, 0, tx);

	/*
	* Swap deadlists.
	*/
	dsl_deadlist_close(&clone->ds_deadlist);
	dsl_deadlist_close(&origin_head->ds_deadlist);
	SWITCH64(dsl_dataset_phys(origin_head)->ds_deadlist_obj,
	dsl_dataset_phys(clone)->ds_deadlist_obj);
	dsl_deadlist_open(&clone->ds_deadlist, dp->dp_meta_objset,
	dsl_dataset_phys(clone)->ds_deadlist_obj);
	dsl_deadlist_open(&origin_head->ds_deadlist, dp->dp_meta_objset,
	dsl_dataset_phys(origin_head)->ds_deadlist_obj);
	dsl_dataset_swap_remap_deadlists(clone, origin_head, tx);

	/*
	* If there is a bookmark at the origin, its "next dataset" is
	* changing, so we need to reset its FBN.
	*/
	dsl_bookmark_next_changed(origin_head, origin_head->ds_prev, tx);

	dsl_scan_ds_clone_swapped(origin_head, clone, tx);

	/*
	* Destroy any livelists associated with the clone or the origin,
	* since after the swap the corresponding livelists are no longer
	* valid.
	*/
	dsl_dir_remove_livelist(clone->ds_dir, tx, B_TRUE);
	dsl_dir_remove_livelist(origin_head->ds_dir, tx, B_TRUE);

	spa_history_log_internal_ds(clone, "clone swap", tx,
	"parent=%s", origin_head->ds_dir->dd_myname);
	}

	/*
	* Given a pool name and a dataset object number in that pool,
	* return the name of that dataset.
	*/
	int
	dsl_dsobj_to_dsname(char pname, uint64_t obj, char buf)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	int error;

	error = dsl_pool_hold(pname, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
	if (error == 0) {
	dsl_dataset_name(ds, buf);
	dsl_dataset_rele(ds, FTAG);
	}
	dsl_pool_rele(dp, FTAG);

	return (error);
	}

	int
	dsl_dataset_check_quota(dsl_dataset_t *ds, boolean_t check_quota,
	uint64_t asize, uint64_t inflight, uint64_t used, uint64_t ref_rsrv)
	{
	int error = 0;

	ASSERT3S(asize, >, 0);

	/*
	* *ref_rsrv is the portion of asize that will come from any
	* unconsumed refreservation space.
	*/
	*ref_rsrv = 0;

	mutex_enter(&ds->ds_lock);
	/*
	* Make a space adjustment for reserved bytes.
	*/
	if (ds->ds_reserved > dsl_dataset_phys(ds)->ds_unique_bytes) {
	ASSERT3U(*used, >=,
	ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes);
	*used -=
	(ds->ds_reserved - dsl_dataset_phys(ds)->ds_unique_bytes);
	*ref_rsrv =
	asize - MIN(asize, parent_delta(ds, asize + inflight));
	}

	if (!check_quota \|\| ds->ds_quota == 0) {
	mutex_exit(&ds->ds_lock);
	return (0);
	}
	/*
	* If they are requesting more space, and our current estimate
	* is over quota, they get to try again unless the actual
	* on-disk is over quota and there are no pending changes (which
	* may free up space for us).
	*/
	if (dsl_dataset_phys(ds)->ds_referenced_bytes + inflight >=
	ds->ds_quota) {
	if (inflight > 0 \|\|
	dsl_dataset_phys(ds)->ds_referenced_bytes < ds->ds_quota)
	error = SET_ERROR(ERESTART);
	else
	error = SET_ERROR(EDQUOT);
	}
	mutex_exit(&ds->ds_lock);

	return (error);
	}

	typedef struct dsl_dataset_set_qr_arg {
	const char *ddsqra_name;
	zprop_source_t ddsqra_source;
	uint64_t ddsqra_value;
	} dsl_dataset_set_qr_arg_t;


	static int
	dsl_dataset_set_refquota_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	int error;
	uint64_t newval;

	if (spa_version(dp->dp_spa) < SPA_VERSION_REFQUOTA)
	return (SET_ERROR(ENOTSUP));

	error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
	if (error != 0)
	return (error);

	if (ds->ds_is_snapshot) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = dsl_prop_predict(ds->ds_dir,
	zfs_prop_to_name(ZFS_PROP_REFQUOTA),
	ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (newval == 0) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	if (newval < dsl_dataset_phys(ds)->ds_referenced_bytes \|\|
	newval < ds->ds_reserved) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOSPC));
	}

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	static void
	dsl_dataset_set_refquota_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds = NULL;
	uint64_t newval;

	VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));

	dsl_prop_set_sync_impl(ds,
	zfs_prop_to_name(ZFS_PROP_REFQUOTA),
	ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1,
	&ddsqra->ddsqra_value, tx);

	VERIFY0(dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_REFQUOTA), &newval));

	if (ds->ds_quota != newval) {
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	ds->ds_quota = newval;
	}
	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
	uint64_t refquota)
	{
	dsl_dataset_set_qr_arg_t ddsqra;

	ddsqra.ddsqra_name = dsname;
	ddsqra.ddsqra_source = source;
	ddsqra.ddsqra_value = refquota;

	return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check,
	dsl_dataset_set_refquota_sync, &ddsqra, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	static int
	dsl_dataset_set_refreservation_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	int error;
	uint64_t newval, unique;

	if (spa_version(dp->dp_spa) < SPA_VERSION_REFRESERVATION)
	return (SET_ERROR(ENOTSUP));

	error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
	if (error != 0)
	return (error);

	if (ds->ds_is_snapshot) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = dsl_prop_predict(ds->ds_dir,
	zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
	ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	/*
	* If we are doing the preliminary check in open context, the
	* space estimates may be inaccurate.
	*/
	if (!dmu_tx_is_syncing(tx)) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	mutex_enter(&ds->ds_lock);
	if (!DS_UNIQUE_IS_ACCURATE(ds))
	dsl_dataset_recalc_head_uniq(ds);
	unique = dsl_dataset_phys(ds)->ds_unique_bytes;
	mutex_exit(&ds->ds_lock);

	if (MAX(unique, newval) > MAX(unique, ds->ds_reserved)) {
	uint64_t delta = MAX(unique, newval) -
	MAX(unique, ds->ds_reserved);

	if (delta >
	dsl_dir_space_available(ds->ds_dir, NULL, 0, B_TRUE) \|\|
	(ds->ds_quota > 0 && newval > ds->ds_quota)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOSPC));
	}
	}

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	void
	dsl_dataset_set_refreservation_sync_impl(dsl_dataset_t *ds,
	zprop_source_t source, uint64_t value, dmu_tx_t *tx)
	{
	uint64_t newval;
	uint64_t unique;
	int64_t delta;

	dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_REFRESERVATION),
	source, sizeof (value), 1, &value, tx);

	VERIFY0(dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &newval));

	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	mutex_enter(&ds->ds_dir->dd_lock);
	mutex_enter(&ds->ds_lock);
	ASSERT(DS_UNIQUE_IS_ACCURATE(ds));
	unique = dsl_dataset_phys(ds)->ds_unique_bytes;
	delta = MAX(0, (int64_t)(newval - unique)) -
	MAX(0, (int64_t)(ds->ds_reserved - unique));
	ds->ds_reserved = newval;
	mutex_exit(&ds->ds_lock);

	dsl_dir_diduse_space(ds->ds_dir, DD_USED_REFRSRV, delta, 0, 0, tx);
	mutex_exit(&ds->ds_dir->dd_lock);
	}

	static void
	dsl_dataset_set_refreservation_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds = NULL;

	VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));
	dsl_dataset_set_refreservation_sync_impl(ds,
	ddsqra->ddsqra_source, ddsqra->ddsqra_value, tx);
	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
	uint64_t refreservation)
	{
	dsl_dataset_set_qr_arg_t ddsqra;

	ddsqra.ddsqra_name = dsname;
	ddsqra.ddsqra_source = source;
	ddsqra.ddsqra_value = refreservation;

	return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check,
	dsl_dataset_set_refreservation_sync, &ddsqra, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	typedef struct dsl_dataset_set_compression_arg {
	const char *ddsca_name;
	zprop_source_t ddsca_source;
	uint64_t ddsca_value;
	} dsl_dataset_set_compression_arg_t;

	static int
	dsl_dataset_set_compression_check(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_compression_arg_t *ddsca = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);

	uint64_t compval = ZIO_COMPRESS_ALGO(ddsca->ddsca_value);
	spa_feature_t f = zio_compress_to_feature(compval);

	if (f == SPA_FEATURE_NONE)
	return (SET_ERROR(EINVAL));

	if (!spa_feature_is_enabled(dp->dp_spa, f))
	return (SET_ERROR(ENOTSUP));

	return (0);
	}

	static void
	dsl_dataset_set_compression_sync(void arg, dmu_tx_t tx)
	{
	dsl_dataset_set_compression_arg_t *ddsca = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds = NULL;

	uint64_t compval = ZIO_COMPRESS_ALGO(ddsca->ddsca_value);
	spa_feature_t f = zio_compress_to_feature(compval);
	ASSERT3S(f, !=, SPA_FEATURE_NONE);
	ASSERT3S(spa_feature_table[f].fi_type, ==, ZFEATURE_TYPE_BOOLEAN);

	VERIFY0(dsl_dataset_hold(dp, ddsca->ddsca_name, FTAG, &ds));
	if (zfeature_active(f, ds->ds_feature[f]) != B_TRUE) {
	ds->ds_feature_activation[f] = (void *)B_TRUE;
	dsl_dataset_activate_feature(ds->ds_object, f,
	ds->ds_feature_activation[f], tx);
	ds->ds_feature[f] = ds->ds_feature_activation[f];
	}
	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dataset_set_compression(const char *dsname, zprop_source_t source,
	uint64_t compression)
	{
	dsl_dataset_set_compression_arg_t ddsca;

	/*
	* The sync task is only required for zstd in order to activate
	* the feature flag when the property is first set.
	*/
	if (ZIO_COMPRESS_ALGO(compression) != ZIO_COMPRESS_ZSTD)
	return (0);

	ddsca.ddsca_name = dsname;
	ddsca.ddsca_source = source;
	ddsca.ddsca_value = compression;

	return (dsl_sync_task(dsname, dsl_dataset_set_compression_check,
	dsl_dataset_set_compression_sync, &ddsca, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	/*
	* Return (in *usedp) the amount of space referenced by "new" that was not
	* referenced at the time the bookmark corresponds to. "New" may be a
	* snapshot or a head. The bookmark must be before new, in
	* new's filesystem (or its origin) -- caller verifies this.
	*
	* The written space is calculated by considering two components: First, we
	* ignore any freed space, and calculate the written as new's used space
	* minus old's used space. Next, we add in the amount of space that was freed
	* between the two time points, thus reducing new's used space relative to
	* old's. Specifically, this is the space that was born before
	* zbm_creation_txg, and freed before new (ie. on new's deadlist or a
	* previous deadlist).
	*
	* space freed [---------------------]
	* snapshots ---O-------O--------O-------O------
	* bookmark new
	*
	* Note, the bookmark's zbm_*_bytes_refd must be valid, but if the HAS_FBN
	* flag is not set, we will calculate the freed_before_next based on the
	* next snapshot's deadlist, rather than using zbm_*_freed_before_next_snap.
	*/
	static int
	dsl_dataset_space_written_impl(zfs_bookmark_phys_t *bmp,
	dsl_dataset_t new, uint64_t usedp, uint64_t compp, uint64_t uncompp)
	{
	int err = 0;
	dsl_pool_t *dp = new->ds_dir->dd_pool;

	ASSERT(dsl_pool_config_held(dp));
	if (dsl_dataset_is_snapshot(new)) {
	ASSERT3U(bmp->zbm_creation_txg, <,
	dsl_dataset_phys(new)->ds_creation_txg);
	}

	*usedp = 0;
	*usedp += dsl_dataset_phys(new)->ds_referenced_bytes;
	*usedp -= bmp->zbm_referenced_bytes_refd;

	*compp = 0;
	*compp += dsl_dataset_phys(new)->ds_compressed_bytes;
	*compp -= bmp->zbm_compressed_bytes_refd;

	*uncompp = 0;
	*uncompp += dsl_dataset_phys(new)->ds_uncompressed_bytes;
	*uncompp -= bmp->zbm_uncompressed_bytes_refd;

	dsl_dataset_t *snap = new;

	while (dsl_dataset_phys(snap)->ds_prev_snap_txg >
	bmp->zbm_creation_txg) {
	uint64_t used, comp, uncomp;

	dsl_deadlist_space_range(&snap->ds_deadlist,
	0, bmp->zbm_creation_txg,
	&used, &comp, &uncomp);
	*usedp += used;
	*compp += comp;
	*uncompp += uncomp;

	uint64_t snapobj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	if (snap != new)
	dsl_dataset_rele(snap, FTAG);
	err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &snap);
	if (err != 0)
	break;
	}

	/*
	* We might not have the FBN if we are calculating written from
	* a snapshot (because we didn't know the correct "next" snapshot
	* until now).
	*/
	if (bmp->zbm_flags & ZBM_FLAG_HAS_FBN) {
	*usedp += bmp->zbm_referenced_freed_before_next_snap;
	*compp += bmp->zbm_compressed_freed_before_next_snap;
	*uncompp += bmp->zbm_uncompressed_freed_before_next_snap;
	} else {
	ASSERT3U(dsl_dataset_phys(snap)->ds_prev_snap_txg, ==,
	bmp->zbm_creation_txg);
	uint64_t used, comp, uncomp;
	dsl_deadlist_space(&snap->ds_deadlist, &used, &comp, &uncomp);
	*usedp += used;
	*compp += comp;
	*uncompp += uncomp;
	}
	if (snap != new)
	dsl_dataset_rele(snap, FTAG);
	return (err);
	}

	/*
	* Return (in *usedp) the amount of space written in new that was not
	* present at the time the bookmark corresponds to. New may be a
	* snapshot or the head. Old must be a bookmark before new, in
	* new's filesystem (or its origin) -- caller verifies this.
	*/
	int
	dsl_dataset_space_written_bookmark(zfs_bookmark_phys_t *bmp,
	dsl_dataset_t new, uint64_t usedp, uint64_t compp, uint64_t uncompp)
	{
	if (!(bmp->zbm_flags & ZBM_FLAG_HAS_FBN))
	return (SET_ERROR(ENOTSUP));
	return (dsl_dataset_space_written_impl(bmp, new,
	usedp, compp, uncompp));
	}

	/*
	* Return (in *usedp) the amount of space written in new that is not
	* present in oldsnap. New may be a snapshot or the head. Old must be
	* a snapshot before new, in new's filesystem (or its origin). If not then
	* fail and return EINVAL.
	*/
	int
	dsl_dataset_space_written(dsl_dataset_t oldsnap, dsl_dataset_t new,
	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	{
	if (!dsl_dataset_is_before(new, oldsnap, 0))
	return (SET_ERROR(EINVAL));

	zfs_bookmark_phys_t zbm = { 0 };
	dsl_dataset_phys_t *dsp = dsl_dataset_phys(oldsnap);
	zbm.zbm_guid = dsp->ds_guid;
	zbm.zbm_creation_txg = dsp->ds_creation_txg;
	zbm.zbm_creation_time = dsp->ds_creation_time;
	zbm.zbm_referenced_bytes_refd = dsp->ds_referenced_bytes;
	zbm.zbm_compressed_bytes_refd = dsp->ds_compressed_bytes;
	zbm.zbm_uncompressed_bytes_refd = dsp->ds_uncompressed_bytes;

	/*
	* If oldsnap is the origin (or origin's origin, ...) of new,
	* we can't easily calculate the effective FBN. Therefore,
	* we do not set ZBM_FLAG_HAS_FBN, so that the _impl will calculate
	* it relative to the correct "next": the next snapshot towards "new",
	* rather than the next snapshot in oldsnap's dsl_dir.
	*/
	return (dsl_dataset_space_written_impl(&zbm, new,
	usedp, compp, uncompp));
	}

	/*
	* Return (in *usedp) the amount of space that will be reclaimed if firstsnap,
	* lastsnap, and all snapshots in between are deleted.
	*
	* blocks that would be freed [---------------------------]
	* snapshots ---O-------O--------O-------O--------O
	* firstsnap lastsnap
	*
	* This is the set of blocks that were born after the snap before firstsnap,
	* (birth > firstsnap->prev_snap_txg) and died before the snap after the
	* last snap (ie, is on lastsnap->ds_next->ds_deadlist or an earlier deadlist).
	* We calculate this by iterating over the relevant deadlists (from the snap
	* after lastsnap, backward to the snap after firstsnap), summing up the
	* space on the deadlist that was born after the snap before firstsnap.
	*/
	int
	dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap,
	dsl_dataset_t *lastsnap,
	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	{
	int err = 0;
	uint64_t snapobj;
	dsl_pool_t *dp = firstsnap->ds_dir->dd_pool;

	ASSERT(firstsnap->ds_is_snapshot);
	ASSERT(lastsnap->ds_is_snapshot);

	/*
	* Check that the snapshots are in the same dsl_dir, and firstsnap
	* is before lastsnap.
	*/
	if (firstsnap->ds_dir != lastsnap->ds_dir \|\|
	dsl_dataset_phys(firstsnap)->ds_creation_txg >
	dsl_dataset_phys(lastsnap)->ds_creation_txg)
	return (SET_ERROR(EINVAL));

	usedp = compp = *uncompp = 0;

	snapobj = dsl_dataset_phys(lastsnap)->ds_next_snap_obj;
	while (snapobj != firstsnap->ds_object) {
	dsl_dataset_t *ds;
	uint64_t used, comp, uncomp;

	err = dsl_dataset_hold_obj(dp, snapobj, FTAG, &ds);
	if (err != 0)
	break;

	dsl_deadlist_space_range(&ds->ds_deadlist,
	dsl_dataset_phys(firstsnap)->ds_prev_snap_txg, UINT64_MAX,
	&used, &comp, &uncomp);
	*usedp += used;
	*compp += comp;
	*uncompp += uncomp;

	snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	ASSERT3U(snapobj, !=, 0);
	dsl_dataset_rele(ds, FTAG);
	}
	return (err);
	}

	/*
	* Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
	* For example, they could both be snapshots of the same filesystem, and
	* 'earlier' is before 'later'. Or 'earlier' could be the origin of
	* 'later's filesystem. Or 'earlier' could be an older snapshot in the origin's
	* filesystem. Or 'earlier' could be the origin's origin.
	*
	* If non-zero, earlier_txg is used instead of earlier's ds_creation_txg.
	*/
	boolean_t
	dsl_dataset_is_before(dsl_dataset_t later, dsl_dataset_t earlier,
	uint64_t earlier_txg)
	{
	dsl_pool_t *dp = later->ds_dir->dd_pool;
	int error;
	boolean_t ret;

	ASSERT(dsl_pool_config_held(dp));
	ASSERT(earlier->ds_is_snapshot \|\| earlier_txg != 0);

	if (earlier_txg == 0)
	earlier_txg = dsl_dataset_phys(earlier)->ds_creation_txg;

	if (later->ds_is_snapshot &&
	earlier_txg >= dsl_dataset_phys(later)->ds_creation_txg)
	return (B_FALSE);

	if (later->ds_dir == earlier->ds_dir)
	return (B_TRUE);

	/*
	* We check dd_origin_obj explicitly here rather than using
	* dsl_dir_is_clone() so that we will return TRUE if "earlier"
	* is $ORIGIN@$ORIGIN. dsl_dataset_space_written() depends on
	* this behavior.
	*/
	if (dsl_dir_phys(later->ds_dir)->dd_origin_obj == 0)
	return (B_FALSE);

	dsl_dataset_t *origin;
	error = dsl_dataset_hold_obj(dp,
	dsl_dir_phys(later->ds_dir)->dd_origin_obj, FTAG, &origin);
	if (error != 0)
	return (B_FALSE);
	if (dsl_dataset_phys(origin)->ds_creation_txg == earlier_txg &&
	origin->ds_dir == earlier->ds_dir) {
	dsl_dataset_rele(origin, FTAG);
	return (B_TRUE);
	}
	ret = dsl_dataset_is_before(origin, earlier, earlier_txg);
	dsl_dataset_rele(origin, FTAG);
	return (ret);
	}

	void
	dsl_dataset_zapify(dsl_dataset_t ds, dmu_tx_t tx)
	{
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	dmu_object_zapify(mos, ds->ds_object, DMU_OT_DSL_DATASET, tx);
	}

	boolean_t
	dsl_dataset_is_zapified(dsl_dataset_t *ds)
	{
	dmu_object_info_t doi;

	dmu_object_info_from_db(ds->ds_dbuf, &doi);
	return (doi.doi_type == DMU_OTN_ZAP_METADATA);
	}

	boolean_t
	dsl_dataset_has_resume_receive_state(dsl_dataset_t *ds)
	{
	return (dsl_dataset_is_zapified(ds) &&
	zap_contains(ds->ds_dir->dd_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_RESUME_TOGUID) == 0);
	}

	uint64_t
	dsl_dataset_get_remap_deadlist_object(dsl_dataset_t *ds)
	{
	uint64_t remap_deadlist_obj;
	int err;

	if (!dsl_dataset_is_zapified(ds))
	return (0);

	err = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object,
	DS_FIELD_REMAP_DEADLIST, sizeof (remap_deadlist_obj), 1,
	&remap_deadlist_obj);

	if (err != 0) {
	VERIFY3S(err, ==, ENOENT);
	return (0);
	}

	ASSERT(remap_deadlist_obj != 0);
	return (remap_deadlist_obj);
	}

	boolean_t
	dsl_dataset_remap_deadlist_exists(dsl_dataset_t *ds)
	{
	EQUIV(dsl_deadlist_is_open(&ds->ds_remap_deadlist),
	dsl_dataset_get_remap_deadlist_object(ds) != 0);
	return (dsl_deadlist_is_open(&ds->ds_remap_deadlist));
	}

	static void
	dsl_dataset_set_remap_deadlist_object(dsl_dataset_t *ds, uint64_t obj,
	dmu_tx_t *tx)
	{
	ASSERT(obj != 0);
	dsl_dataset_zapify(ds, tx);
	VERIFY0(zap_add(ds->ds_dir->dd_pool->dp_meta_objset, ds->ds_object,
	DS_FIELD_REMAP_DEADLIST, sizeof (obj), 1, &obj, tx));
	}

	static void
	dsl_dataset_unset_remap_deadlist_object(dsl_dataset_t ds, dmu_tx_t tx)
	{
	VERIFY0(zap_remove(ds->ds_dir->dd_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_REMAP_DEADLIST, tx));
	}

	void
	dsl_dataset_destroy_remap_deadlist(dsl_dataset_t ds, dmu_tx_t tx)
	{
	uint64_t remap_deadlist_object;
	spa_t *spa = ds->ds_dir->dd_pool->dp_spa;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(dsl_dataset_remap_deadlist_exists(ds));

	remap_deadlist_object = ds->ds_remap_deadlist.dl_object;
	dsl_deadlist_close(&ds->ds_remap_deadlist);
	dsl_deadlist_free(spa_meta_objset(spa), remap_deadlist_object, tx);
	dsl_dataset_unset_remap_deadlist_object(ds, tx);
	spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	}

	void
	dsl_dataset_create_remap_deadlist(dsl_dataset_t ds, dmu_tx_t tx)
	{
	uint64_t remap_deadlist_obj;
	spa_t *spa = ds->ds_dir->dd_pool->dp_spa;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(MUTEX_HELD(&ds->ds_remap_deadlist_lock));
	/*
	* Currently we only create remap deadlists when there are indirect
	* vdevs with referenced mappings.
	*/
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));

	remap_deadlist_obj = dsl_deadlist_clone(
	&ds->ds_deadlist, UINT64_MAX,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, tx);
	dsl_dataset_set_remap_deadlist_object(ds,
	remap_deadlist_obj, tx);
	dsl_deadlist_open(&ds->ds_remap_deadlist, spa_meta_objset(spa),
	remap_deadlist_obj);
	spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	}

	void
	dsl_dataset_activate_redaction(dsl_dataset_t ds, uint64_t redact_snaps,
	uint64_t num_redact_snaps, dmu_tx_t *tx)
	{
	uint64_t dsobj = ds->ds_object;
	struct feature_type_uint64_array_arg *ftuaa =
	kmem_zalloc(sizeof (*ftuaa), KM_SLEEP);
	ftuaa->length = (int64_t)num_redact_snaps;
	if (num_redact_snaps > 0) {
	ftuaa->array = kmem_alloc(num_redact_snaps * sizeof (uint64_t),
	KM_SLEEP);
	memcpy(ftuaa->array, redact_snaps, num_redact_snaps *
	sizeof (uint64_t));
	}
	dsl_dataset_activate_feature(dsobj, SPA_FEATURE_REDACTED_DATASETS,
	ftuaa, tx);
	ds->ds_feature[SPA_FEATURE_REDACTED_DATASETS] = ftuaa;
	}

	/*
	* Find and return (in *oldest_dsobj) the oldest snapshot of the dsobj
	* dataset whose birth time is >= min_txg.
	*/
	int
	dsl_dataset_oldest_snapshot(spa_t *spa, uint64_t head_ds, uint64_t min_txg,
	uint64_t *oldest_dsobj)
	{
	dsl_dataset_t *ds;
	dsl_pool_t *dp = spa->spa_dsl_pool;

	int error = dsl_dataset_hold_obj(dp, head_ds, FTAG, &ds);
	if (error != 0)
	return (error);

	uint64_t prev_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	uint64_t prev_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;

	while (prev_obj != 0 && min_txg < prev_obj_txg) {
	dsl_dataset_rele(ds, FTAG);
	if ((error = dsl_dataset_hold_obj(dp, prev_obj,
	FTAG, &ds)) != 0)
	return (error);
	prev_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
	prev_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	}
	*oldest_dsobj = ds->ds_object;
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	-ZFS_MODULE_PARAM(zfs, zfs_, max_recordsize, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, max_recordsize, UINT, ZMOD_RW,
	"Max allowed record size");

	ZFS_MODULE_PARAM(zfs, zfs_, allow_redacted_dataset_mount, INT, ZMOD_RW,
	"Allow mounting of redacted datasets");

	ZFS_MODULE_PARAM(zfs, zfs_, snapshot_history_enabled, INT, ZMOD_RW,
	"Include snapshot events in pool history/events");

	EXPORT_SYMBOL(dsl_dataset_hold);
	EXPORT_SYMBOL(dsl_dataset_hold_flags);
	EXPORT_SYMBOL(dsl_dataset_hold_obj);
	EXPORT_SYMBOL(dsl_dataset_hold_obj_flags);
	EXPORT_SYMBOL(dsl_dataset_own);
	EXPORT_SYMBOL(dsl_dataset_own_obj);
	EXPORT_SYMBOL(dsl_dataset_name);
	EXPORT_SYMBOL(dsl_dataset_rele);
	EXPORT_SYMBOL(dsl_dataset_rele_flags);
	EXPORT_SYMBOL(dsl_dataset_disown);
	EXPORT_SYMBOL(dsl_dataset_tryown);
	EXPORT_SYMBOL(dsl_dataset_create_sync);
	EXPORT_SYMBOL(dsl_dataset_create_sync_dd);
	EXPORT_SYMBOL(dsl_dataset_snapshot_check);
	EXPORT_SYMBOL(dsl_dataset_snapshot_sync);
	EXPORT_SYMBOL(dsl_dataset_promote);
	EXPORT_SYMBOL(dsl_dataset_user_hold);
	EXPORT_SYMBOL(dsl_dataset_user_release);
	EXPORT_SYMBOL(dsl_dataset_get_holds);
	EXPORT_SYMBOL(dsl_dataset_get_blkptr);
	EXPORT_SYMBOL(dsl_dataset_get_spa);
	EXPORT_SYMBOL(dsl_dataset_modified_since_snap);
	EXPORT_SYMBOL(dsl_dataset_space_written);
	EXPORT_SYMBOL(dsl_dataset_space_wouldfree);
	EXPORT_SYMBOL(dsl_dataset_sync);
	EXPORT_SYMBOL(dsl_dataset_block_born);
	EXPORT_SYMBOL(dsl_dataset_block_kill);
	EXPORT_SYMBOL(dsl_dataset_dirty);
	EXPORT_SYMBOL(dsl_dataset_stats);
	EXPORT_SYMBOL(dsl_dataset_fast_stat);
	EXPORT_SYMBOL(dsl_dataset_space);
	EXPORT_SYMBOL(dsl_dataset_fsid_guid);
	EXPORT_SYMBOL(dsl_dsobj_to_dsname);
	EXPORT_SYMBOL(dsl_dataset_check_quota);
	EXPORT_SYMBOL(dsl_dataset_clone_swap_check_impl);
	EXPORT_SYMBOL(dsl_dataset_clone_swap_sync_impl);
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_deadlist.c b/sys/contrib/openzfs/module/zfs/dsl_deadlist.c
	index 7270924ec11f..1ecae0fe3865 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_deadlist.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_deadlist.c
	@@ -1,1041 +1,1046 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	*/

	#include <sys/dmu.h>
	#include <sys/zap.h>
	#include <sys/zfs_context.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dataset.h>

	/*
	* Deadlist concurrency:
	*
	* Deadlists can only be modified from the syncing thread.
	*
	* Except for dsl_deadlist_insert(), it can only be modified with the
	* dp_config_rwlock held with RW_WRITER.
	*
	* The accessors (dsl_deadlist_space() and dsl_deadlist_space_range()) can
	* be called concurrently, from open context, with the dl_config_rwlock held
	* with RW_READER.
	*
	* Therefore, we only need to provide locking between dsl_deadlist_insert() and
	* the accessors, protecting:
	* dl_phys->dl_used,comp,uncomp
	* and protecting the dl_tree from being loaded.
	* The locking is provided by dl_lock. Note that locking on the bpobj_t
	* provides its own locking, and dl_oldfmt is immutable.
	*/

	/*
	* Livelist Overview
	* ================
	*
	* Livelists use the same 'deadlist_t' struct as deadlists and are also used
	* to track blkptrs over the lifetime of a dataset. Livelists however, belong
	* to clones and track the blkptrs that are clone-specific (were born after
	* the clone's creation). The exception is embedded block pointers which are
	* not included in livelists because they do not need to be freed.
	*
	* When it comes time to delete the clone, the livelist provides a quick
	* reference as to what needs to be freed. For this reason, livelists also track
	* when clone-specific blkptrs are freed before deletion to prevent double
	* frees. Each blkptr in a livelist is marked as a FREE or an ALLOC and the
	* deletion algorithm iterates backwards over the livelist, matching
	* FREE/ALLOC pairs and then freeing those ALLOCs which remain. livelists
	* are also updated in the case when blkptrs are remapped: the old version
	* of the blkptr is cancelled out with a FREE and the new version is tracked
	* with an ALLOC.
	*
	* To bound the amount of memory required for deletion, livelists over a
	* certain size are spread over multiple entries. Entries are grouped by
	* birth txg so we can be sure the ALLOC/FREE pair for a given blkptr will
	* be in the same entry. This allows us to delete livelists incrementally
	* over multiple syncs, one entry at a time.
	*
	* During the lifetime of the clone, livelists can get extremely large.
	* Their size is managed by periodic condensing (preemptively cancelling out
	* FREE/ALLOC pairs). Livelists are disabled when a clone is promoted or when
	* the shared space between the clone and its origin is so small that it
	* doesn't make sense to use livelists anymore.
	*/

	/*
	* The threshold sublist size at which we create a new sub-livelist for the
	* next txg. However, since blkptrs of the same transaction group must be in
	* the same sub-list, the actual sublist size may exceed this. When picking the
	* size we had to balance the fact that larger sublists mean fewer sublists
	* (decreasing the cost of insertion) against the consideration that sublists
	* will be loaded into memory and shouldn't take up an inordinate amount of
	* space. We settled on ~500000 entries, corresponding to roughly 128M.
	*/
	unsigned long zfs_livelist_max_entries = 500000;

	/*
	* We can approximate how much of a performance gain a livelist will give us
	* based on the percentage of blocks shared between the clone and its origin.
	* 0 percent shared means that the clone has completely diverged and that the
	* old method is maximally effective: every read from the block tree will
	* result in lots of frees. Livelists give us gains when they track blocks
	* scattered across the tree, when one read in the old method might only
	* result in a few frees. Once the clone has been overwritten enough,
	* writes are no longer sparse and we'll no longer get much of a benefit from
	* tracking them with a livelist. We chose a lower limit of 75 percent shared
	* (25 percent overwritten). This means that 1/4 of all block pointers will be
	* freed (e.g. each read frees 256, out of a max of 1024) so we expect livelists
	* to make deletion 4x faster. Once the amount of shared space drops below this
	* threshold, the clone will revert to the old deletion method.
	*/
	int zfs_livelist_min_percent_shared = 75;

	static int
	dsl_deadlist_compare(const void arg1, const void arg2)
	{
	const dsl_deadlist_entry_t *dle1 = arg1;
	const dsl_deadlist_entry_t *dle2 = arg2;

	return (TREE_CMP(dle1->dle_mintxg, dle2->dle_mintxg));
	}

	static int
	dsl_deadlist_cache_compare(const void arg1, const void arg2)
	{
	const dsl_deadlist_cache_entry_t *dlce1 = arg1;
	const dsl_deadlist_cache_entry_t *dlce2 = arg2;

	return (TREE_CMP(dlce1->dlce_mintxg, dlce2->dlce_mintxg));
	}

	static void
	dsl_deadlist_load_tree(dsl_deadlist_t *dl)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	int error;

	ASSERT(MUTEX_HELD(&dl->dl_lock));

	ASSERT(!dl->dl_oldfmt);
	if (dl->dl_havecache) {
	/*
	* After loading the tree, the caller may modify the tree,
	* e.g. to add or remove nodes, or to make a node no longer
	* refer to the empty_bpobj. These changes would make the
	* dl_cache incorrect. Therefore we discard the cache here,
	* so that it can't become incorrect.
	*/
	dsl_deadlist_cache_entry_t *dlce;
	void *cookie = NULL;
	while ((dlce = avl_destroy_nodes(&dl->dl_cache, &cookie))
	!= NULL) {
	kmem_free(dlce, sizeof (*dlce));
	}
	avl_destroy(&dl->dl_cache);
	dl->dl_havecache = B_FALSE;
	}
	if (dl->dl_havetree)
	return;

	avl_create(&dl->dl_tree, dsl_deadlist_compare,
	sizeof (dsl_deadlist_entry_t),
	offsetof(dsl_deadlist_entry_t, dle_node));
	for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	dsl_deadlist_entry_t dle = kmem_alloc(sizeof (dle), KM_SLEEP);
	dle->dle_mintxg = zfs_strtonum(za.za_name, NULL);

	/*
	* Prefetch all the bpobj's so that we do that i/o
	* in parallel. Then open them all in a second pass.
	*/
	dle->dle_bpobj.bpo_object = za.za_first_integer;
	dmu_prefetch(dl->dl_os, dle->dle_bpobj.bpo_object,
	0, 0, 0, ZIO_PRIORITY_SYNC_READ);

	avl_add(&dl->dl_tree, dle);
	}
	VERIFY3U(error, ==, ENOENT);
	zap_cursor_fini(&zc);

	for (dsl_deadlist_entry_t *dle = avl_first(&dl->dl_tree);
	dle != NULL; dle = AVL_NEXT(&dl->dl_tree, dle)) {
	VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os,
	dle->dle_bpobj.bpo_object));
	}
	dl->dl_havetree = B_TRUE;
	}

	/*
	* Load only the non-empty bpobj's into the dl_cache. The cache is an analog
	* of the dl_tree, but contains only non-empty_bpobj nodes from the ZAP. It
	* is used only for gathering space statistics. The dl_cache has two
	* advantages over the dl_tree:
	*
	* 1. Loading the dl_cache is ~5x faster than loading the dl_tree (if it's
	* mostly empty_bpobj's), due to less CPU overhead to open the empty_bpobj
	* many times and to inquire about its (zero) space stats many times.
	*
	* 2. The dl_cache uses less memory than the dl_tree. We only need to load
	* the dl_tree of snapshots when deleting a snapshot, after which we free the
	* dl_tree with dsl_deadlist_discard_tree
	*/
	static void
	dsl_deadlist_load_cache(dsl_deadlist_t *dl)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	int error;

	ASSERT(MUTEX_HELD(&dl->dl_lock));

	ASSERT(!dl->dl_oldfmt);
	if (dl->dl_havecache)
	return;

	uint64_t empty_bpobj = dmu_objset_pool(dl->dl_os)->dp_empty_bpobj;

	avl_create(&dl->dl_cache, dsl_deadlist_cache_compare,
	sizeof (dsl_deadlist_cache_entry_t),
	offsetof(dsl_deadlist_cache_entry_t, dlce_node));
	for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	if (za.za_first_integer == empty_bpobj)
	continue;
	dsl_deadlist_cache_entry_t *dlce =
	kmem_zalloc(sizeof (*dlce), KM_SLEEP);
	dlce->dlce_mintxg = zfs_strtonum(za.za_name, NULL);

	/*
	* Prefetch all the bpobj's so that we do that i/o
	* in parallel. Then open them all in a second pass.
	*/
	dlce->dlce_bpobj = za.za_first_integer;
	dmu_prefetch(dl->dl_os, dlce->dlce_bpobj,
	0, 0, 0, ZIO_PRIORITY_SYNC_READ);
	avl_add(&dl->dl_cache, dlce);
	}
	VERIFY3U(error, ==, ENOENT);
	zap_cursor_fini(&zc);

	for (dsl_deadlist_cache_entry_t *dlce = avl_first(&dl->dl_cache);
	dlce != NULL; dlce = AVL_NEXT(&dl->dl_cache, dlce)) {
	bpobj_t bpo;
	VERIFY0(bpobj_open(&bpo, dl->dl_os, dlce->dlce_bpobj));

	VERIFY0(bpobj_space(&bpo,
	&dlce->dlce_bytes, &dlce->dlce_comp, &dlce->dlce_uncomp));
	bpobj_close(&bpo);
	}
	dl->dl_havecache = B_TRUE;
	}

	/*
	* Discard the tree to save memory.
	*/
	void
	dsl_deadlist_discard_tree(dsl_deadlist_t *dl)
	{
	mutex_enter(&dl->dl_lock);

	if (!dl->dl_havetree) {
	mutex_exit(&dl->dl_lock);
	return;
	}
	dsl_deadlist_entry_t *dle;
	void *cookie = NULL;
	while ((dle = avl_destroy_nodes(&dl->dl_tree, &cookie)) != NULL) {
	bpobj_close(&dle->dle_bpobj);
	kmem_free(dle, sizeof (*dle));
	}
	avl_destroy(&dl->dl_tree);

	dl->dl_havetree = B_FALSE;
	mutex_exit(&dl->dl_lock);
	}

	void
	dsl_deadlist_iterate(dsl_deadlist_t dl, deadlist_iter_t func, void args)
	{
	dsl_deadlist_entry_t *dle;

	ASSERT(dsl_deadlist_is_open(dl));

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);
	mutex_exit(&dl->dl_lock);
	for (dle = avl_first(&dl->dl_tree); dle != NULL;
	dle = AVL_NEXT(&dl->dl_tree, dle)) {
	if (func(args, dle) != 0)
	break;
	}
	}

	void
	dsl_deadlist_open(dsl_deadlist_t dl, objset_t os, uint64_t object)
	{
	dmu_object_info_t doi;

	ASSERT(!dsl_deadlist_is_open(dl));

	mutex_init(&dl->dl_lock, NULL, MUTEX_DEFAULT, NULL);
	dl->dl_os = os;
	dl->dl_object = object;
	VERIFY0(dmu_bonus_hold(os, object, dl, &dl->dl_dbuf));
	dmu_object_info_from_db(dl->dl_dbuf, &doi);
	if (doi.doi_type == DMU_OT_BPOBJ) {
	dmu_buf_rele(dl->dl_dbuf, dl);
	dl->dl_dbuf = NULL;
	dl->dl_oldfmt = B_TRUE;
	VERIFY0(bpobj_open(&dl->dl_bpobj, os, object));
	return;
	}

	dl->dl_oldfmt = B_FALSE;
	dl->dl_phys = dl->dl_dbuf->db_data;
	dl->dl_havetree = B_FALSE;
	dl->dl_havecache = B_FALSE;
	}

	boolean_t
	dsl_deadlist_is_open(dsl_deadlist_t *dl)
	{
	return (dl->dl_os != NULL);
	}

	void
	dsl_deadlist_close(dsl_deadlist_t *dl)
	{
	ASSERT(dsl_deadlist_is_open(dl));
	mutex_destroy(&dl->dl_lock);

	if (dl->dl_oldfmt) {
	dl->dl_oldfmt = B_FALSE;
	bpobj_close(&dl->dl_bpobj);
	dl->dl_os = NULL;
	dl->dl_object = 0;
	return;
	}

	if (dl->dl_havetree) {
	dsl_deadlist_entry_t *dle;
	void *cookie = NULL;
	while ((dle = avl_destroy_nodes(&dl->dl_tree, &cookie))
	!= NULL) {
	bpobj_close(&dle->dle_bpobj);
	kmem_free(dle, sizeof (*dle));
	}
	avl_destroy(&dl->dl_tree);
	}
	if (dl->dl_havecache) {
	dsl_deadlist_cache_entry_t *dlce;
	void *cookie = NULL;
	while ((dlce = avl_destroy_nodes(&dl->dl_cache, &cookie))
	!= NULL) {
	kmem_free(dlce, sizeof (*dlce));
	}
	avl_destroy(&dl->dl_cache);
	}
	dmu_buf_rele(dl->dl_dbuf, dl);
	dl->dl_dbuf = NULL;
	dl->dl_phys = NULL;
	dl->dl_os = NULL;
	dl->dl_object = 0;
	}

	uint64_t
	dsl_deadlist_alloc(objset_t os, dmu_tx_t tx)
	{
	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
	return (bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx));
	return (zap_create(os, DMU_OT_DEADLIST, DMU_OT_DEADLIST_HDR,
	sizeof (dsl_deadlist_phys_t), tx));
	}

	void
	dsl_deadlist_free(objset_t os, uint64_t dlobj, dmu_tx_t tx)
	{
	dmu_object_info_t doi;
	zap_cursor_t zc;
	zap_attribute_t za;
	int error;

	VERIFY0(dmu_object_info(os, dlobj, &doi));
	if (doi.doi_type == DMU_OT_BPOBJ) {
	bpobj_free(os, dlobj, tx);
	return;
	}

	for (zap_cursor_init(&zc, os, dlobj);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t obj = za.za_first_integer;
	if (obj == dmu_objset_pool(os)->dp_empty_bpobj)
	bpobj_decr_empty(os, tx);
	else
	bpobj_free(os, obj, tx);
	}
	VERIFY3U(error, ==, ENOENT);
	zap_cursor_fini(&zc);
	VERIFY0(dmu_object_free(os, dlobj, tx));
	}

	static void
	dle_enqueue(dsl_deadlist_t dl, dsl_deadlist_entry_t dle,
	const blkptr_t bp, boolean_t bp_freed, dmu_tx_t tx)
	{
	ASSERT(MUTEX_HELD(&dl->dl_lock));
	if (dle->dle_bpobj.bpo_object ==
	dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
	uint64_t obj = bpobj_alloc(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
	bpobj_close(&dle->dle_bpobj);
	bpobj_decr_empty(dl->dl_os, tx);
	VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
	VERIFY0(zap_update_int_key(dl->dl_os, dl->dl_object,
	dle->dle_mintxg, obj, tx));
	}
	bpobj_enqueue(&dle->dle_bpobj, bp, bp_freed, tx);
	}

	static void
	dle_enqueue_subobj(dsl_deadlist_t dl, dsl_deadlist_entry_t dle,
	uint64_t obj, dmu_tx_t *tx)
	{
	ASSERT(MUTEX_HELD(&dl->dl_lock));
	if (dle->dle_bpobj.bpo_object !=
	dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
	bpobj_enqueue_subobj(&dle->dle_bpobj, obj, tx);
	} else {
	bpobj_close(&dle->dle_bpobj);
	bpobj_decr_empty(dl->dl_os, tx);
	VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
	VERIFY0(zap_update_int_key(dl->dl_os, dl->dl_object,
	dle->dle_mintxg, obj, tx));
	}
	}

	void
	dsl_deadlist_insert(dsl_deadlist_t dl, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	dsl_deadlist_entry_t dle_tofind;
	dsl_deadlist_entry_t *dle;
	avl_index_t where;

	if (dl->dl_oldfmt) {
	bpobj_enqueue(&dl->dl_bpobj, bp, bp_freed, tx);
	return;
	}

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);

	dmu_buf_will_dirty(dl->dl_dbuf, tx);

	int sign = bp_freed ? -1 : +1;
	dl->dl_phys->dl_used +=
	sign * bp_get_dsize_sync(dmu_objset_spa(dl->dl_os), bp);
	dl->dl_phys->dl_comp += sign * BP_GET_PSIZE(bp);
	dl->dl_phys->dl_uncomp += sign * BP_GET_UCSIZE(bp);

	dle_tofind.dle_mintxg = bp->blk_birth;
	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
	if (dle == NULL)
	dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
	else
	dle = AVL_PREV(&dl->dl_tree, dle);

	if (dle == NULL) {
	zfs_panic_recover("blkptr at %p has invalid BLK_BIRTH %llu",
	bp, (longlong_t)bp->blk_birth);
	dle = avl_first(&dl->dl_tree);
	}

	ASSERT3P(dle, !=, NULL);
	dle_enqueue(dl, dle, bp, bp_freed, tx);
	mutex_exit(&dl->dl_lock);
	}

	int
	dsl_deadlist_insert_alloc_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	dsl_deadlist_t *dl = arg;
	dsl_deadlist_insert(dl, bp, B_FALSE, tx);
	return (0);
	}

	int
	dsl_deadlist_insert_free_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	dsl_deadlist_t *dl = arg;
	dsl_deadlist_insert(dl, bp, B_TRUE, tx);
	return (0);
	}

	/*
	* Insert new key in deadlist, which must be > all current entries.
	* mintxg is not inclusive.
	*/
	void
	dsl_deadlist_add_key(dsl_deadlist_t dl, uint64_t mintxg, dmu_tx_t tx)
	{
	uint64_t obj;
	dsl_deadlist_entry_t *dle;

	if (dl->dl_oldfmt)
	return;

	dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
	dle->dle_mintxg = mintxg;

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);

	obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
	VERIFY0(bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
	avl_add(&dl->dl_tree, dle);

	VERIFY0(zap_add_int_key(dl->dl_os, dl->dl_object,
	mintxg, obj, tx));
	mutex_exit(&dl->dl_lock);
	}

	/*
	* Remove this key, merging its entries into the previous key.
	*/
	void
	dsl_deadlist_remove_key(dsl_deadlist_t dl, uint64_t mintxg, dmu_tx_t tx)
	{
	dsl_deadlist_entry_t dle_tofind;
	dsl_deadlist_entry_t dle, dle_prev;

	if (dl->dl_oldfmt)
	return;
	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);

	dle_tofind.dle_mintxg = mintxg;
	dle = avl_find(&dl->dl_tree, &dle_tofind, NULL);
	ASSERT3P(dle, !=, NULL);
	dle_prev = AVL_PREV(&dl->dl_tree, dle);

	dle_enqueue_subobj(dl, dle_prev, dle->dle_bpobj.bpo_object, tx);

	avl_remove(&dl->dl_tree, dle);
	bpobj_close(&dle->dle_bpobj);
	kmem_free(dle, sizeof (*dle));

	VERIFY0(zap_remove_int(dl->dl_os, dl->dl_object, mintxg, tx));
	mutex_exit(&dl->dl_lock);
	}

	/*
	* Remove a deadlist entry and all of its contents by removing the entry from
	* the deadlist's avl tree, freeing the entry's bpobj and adjusting the
	* deadlist's space accounting accordingly.
	*/
	void
	dsl_deadlist_remove_entry(dsl_deadlist_t dl, uint64_t mintxg, dmu_tx_t tx)
	{
	uint64_t used, comp, uncomp;
	dsl_deadlist_entry_t dle_tofind;
	dsl_deadlist_entry_t *dle;
	objset_t *os = dl->dl_os;

	if (dl->dl_oldfmt)
	return;

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);

	dle_tofind.dle_mintxg = mintxg;
	dle = avl_find(&dl->dl_tree, &dle_tofind, NULL);
	VERIFY3P(dle, !=, NULL);

	avl_remove(&dl->dl_tree, dle);
	VERIFY0(zap_remove_int(os, dl->dl_object, mintxg, tx));
	VERIFY0(bpobj_space(&dle->dle_bpobj, &used, &comp, &uncomp));
	dmu_buf_will_dirty(dl->dl_dbuf, tx);
	dl->dl_phys->dl_used -= used;
	dl->dl_phys->dl_comp -= comp;
	dl->dl_phys->dl_uncomp -= uncomp;
	if (dle->dle_bpobj.bpo_object == dmu_objset_pool(os)->dp_empty_bpobj) {
	bpobj_decr_empty(os, tx);
	} else {
	bpobj_free(os, dle->dle_bpobj.bpo_object, tx);
	}
	bpobj_close(&dle->dle_bpobj);
	kmem_free(dle, sizeof (*dle));
	mutex_exit(&dl->dl_lock);
	}

	/*
	* Clear out the contents of a deadlist_entry by freeing its bpobj,
	* replacing it with an empty bpobj and adjusting the deadlist's
	* space accounting
	*/
	void
	dsl_deadlist_clear_entry(dsl_deadlist_entry_t dle, dsl_deadlist_t dl,
	dmu_tx_t *tx)
	{
	uint64_t new_obj, used, comp, uncomp;
	objset_t *os = dl->dl_os;

	mutex_enter(&dl->dl_lock);
	VERIFY0(zap_remove_int(os, dl->dl_object, dle->dle_mintxg, tx));
	VERIFY0(bpobj_space(&dle->dle_bpobj, &used, &comp, &uncomp));
	dmu_buf_will_dirty(dl->dl_dbuf, tx);
	dl->dl_phys->dl_used -= used;
	dl->dl_phys->dl_comp -= comp;
	dl->dl_phys->dl_uncomp -= uncomp;
	if (dle->dle_bpobj.bpo_object == dmu_objset_pool(os)->dp_empty_bpobj)
	bpobj_decr_empty(os, tx);
	else
	bpobj_free(os, dle->dle_bpobj.bpo_object, tx);
	bpobj_close(&dle->dle_bpobj);
	new_obj = bpobj_alloc_empty(os, SPA_OLD_MAXBLOCKSIZE, tx);
	VERIFY0(bpobj_open(&dle->dle_bpobj, os, new_obj));
	VERIFY0(zap_add_int_key(os, dl->dl_object, dle->dle_mintxg,
	new_obj, tx));
	ASSERT(bpobj_is_empty(&dle->dle_bpobj));
	mutex_exit(&dl->dl_lock);
	}

	/*
	* Return the first entry in deadlist's avl tree
	*/
	dsl_deadlist_entry_t *
	dsl_deadlist_first(dsl_deadlist_t *dl)
	{
	dsl_deadlist_entry_t *dle;

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);
	dle = avl_first(&dl->dl_tree);
	mutex_exit(&dl->dl_lock);

	return (dle);
	}

	/*
	* Return the last entry in deadlist's avl tree
	*/
	dsl_deadlist_entry_t *
	dsl_deadlist_last(dsl_deadlist_t *dl)
	{
	dsl_deadlist_entry_t *dle;

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);
	dle = avl_last(&dl->dl_tree);
	mutex_exit(&dl->dl_lock);

	return (dle);
	}

	/*
	* Walk ds's snapshots to regenerate generate ZAP & AVL.
	*/
	static void
	dsl_deadlist_regenerate(objset_t *os, uint64_t dlobj,
	uint64_t mrs_obj, dmu_tx_t *tx)
	{
	dsl_deadlist_t dl = { 0 };
	dsl_pool_t *dp = dmu_objset_pool(os);

	dsl_deadlist_open(&dl, os, dlobj);
	if (dl.dl_oldfmt) {
	dsl_deadlist_close(&dl);
	return;
	}

	while (mrs_obj != 0) {
	dsl_dataset_t *ds;
	VERIFY0(dsl_dataset_hold_obj(dp, mrs_obj, FTAG, &ds));
	dsl_deadlist_add_key(&dl,
	dsl_dataset_phys(ds)->ds_prev_snap_txg, tx);
	mrs_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	dsl_dataset_rele(ds, FTAG);
	}
	dsl_deadlist_close(&dl);
	}

	uint64_t
	dsl_deadlist_clone(dsl_deadlist_t *dl, uint64_t maxtxg,
	uint64_t mrs_obj, dmu_tx_t *tx)
	{
	dsl_deadlist_entry_t *dle;
	uint64_t newobj;

	newobj = dsl_deadlist_alloc(dl->dl_os, tx);

	if (dl->dl_oldfmt) {
	dsl_deadlist_regenerate(dl->dl_os, newobj, mrs_obj, tx);
	return (newobj);
	}

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_tree(dl);

	for (dle = avl_first(&dl->dl_tree); dle;
	dle = AVL_NEXT(&dl->dl_tree, dle)) {
	uint64_t obj;

	if (dle->dle_mintxg >= maxtxg)
	break;

	obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
	VERIFY0(zap_add_int_key(dl->dl_os, newobj,
	dle->dle_mintxg, obj, tx));
	}
	mutex_exit(&dl->dl_lock);
	return (newobj);
	}

	void
	dsl_deadlist_space(dsl_deadlist_t *dl,
	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	{
	ASSERT(dsl_deadlist_is_open(dl));
	if (dl->dl_oldfmt) {
	VERIFY0(bpobj_space(&dl->dl_bpobj,
	usedp, compp, uncompp));
	return;
	}

	mutex_enter(&dl->dl_lock);
	*usedp = dl->dl_phys->dl_used;
	*compp = dl->dl_phys->dl_comp;
	*uncompp = dl->dl_phys->dl_uncomp;
	mutex_exit(&dl->dl_lock);
	}

	/*
	* return space used in the range (mintxg, maxtxg].
	* Includes maxtxg, does not include mintxg.
	* mintxg and maxtxg must both be keys in the deadlist (unless maxtxg is
	* UINT64_MAX).
	*/
	void
	dsl_deadlist_space_range(dsl_deadlist_t *dl, uint64_t mintxg, uint64_t maxtxg,
	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	{
	dsl_deadlist_cache_entry_t *dlce;
	dsl_deadlist_cache_entry_t dlce_tofind;
	avl_index_t where;

	if (dl->dl_oldfmt) {
	VERIFY0(bpobj_space_range(&dl->dl_bpobj,
	mintxg, maxtxg, usedp, compp, uncompp));
	return;
	}

	usedp = compp = *uncompp = 0;

	mutex_enter(&dl->dl_lock);
	dsl_deadlist_load_cache(dl);
	dlce_tofind.dlce_mintxg = mintxg;
	dlce = avl_find(&dl->dl_cache, &dlce_tofind, &where);

	/*
	* If this mintxg doesn't exist, it may be an empty_bpobj which
	* is omitted from the sparse tree. Start at the next non-empty
	* entry.
	*/
	if (dlce == NULL)
	dlce = avl_nearest(&dl->dl_cache, where, AVL_AFTER);

	for (; dlce && dlce->dlce_mintxg < maxtxg;
	dlce = AVL_NEXT(&dl->dl_tree, dlce)) {
	*usedp += dlce->dlce_bytes;
	*compp += dlce->dlce_comp;
	*uncompp += dlce->dlce_uncomp;
	}

	mutex_exit(&dl->dl_lock);
	}

	static void
	dsl_deadlist_insert_bpobj(dsl_deadlist_t *dl, uint64_t obj, uint64_t birth,
	dmu_tx_t *tx)
	{
	dsl_deadlist_entry_t dle_tofind;
	dsl_deadlist_entry_t *dle;
	avl_index_t where;
	uint64_t used, comp, uncomp;
	bpobj_t bpo;

	ASSERT(MUTEX_HELD(&dl->dl_lock));

	VERIFY0(bpobj_open(&bpo, dl->dl_os, obj));
	VERIFY0(bpobj_space(&bpo, &used, &comp, &uncomp));
	bpobj_close(&bpo);

	dsl_deadlist_load_tree(dl);

	dmu_buf_will_dirty(dl->dl_dbuf, tx);
	dl->dl_phys->dl_used += used;
	dl->dl_phys->dl_comp += comp;
	dl->dl_phys->dl_uncomp += uncomp;

	dle_tofind.dle_mintxg = birth;
	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
	if (dle == NULL)
	dle = avl_nearest(&dl->dl_tree, where, AVL_BEFORE);
	dle_enqueue_subobj(dl, dle, obj, tx);
	}

	static int
	dsl_deadlist_insert_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	dsl_deadlist_t *dl = arg;
	dsl_deadlist_insert(dl, bp, bp_freed, tx);
	return (0);
	}

	/*
	* Merge the deadlist pointed to by 'obj' into dl. obj will be left as
	* an empty deadlist.
	*/
	void
	dsl_deadlist_merge(dsl_deadlist_t dl, uint64_t obj, dmu_tx_t tx)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	dmu_buf_t *bonus;
	dsl_deadlist_phys_t *dlp;
	dmu_object_info_t doi;
	int error;

	VERIFY0(dmu_object_info(dl->dl_os, obj, &doi));
	if (doi.doi_type == DMU_OT_BPOBJ) {
	bpobj_t bpo;
	VERIFY0(bpobj_open(&bpo, dl->dl_os, obj));
	VERIFY0(bpobj_iterate(&bpo, dsl_deadlist_insert_cb, dl, tx));
	bpobj_close(&bpo);
	return;
	}

	mutex_enter(&dl->dl_lock);
	for (zap_cursor_init(&zc, dl->dl_os, obj);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t mintxg = zfs_strtonum(za.za_name, NULL);
	dsl_deadlist_insert_bpobj(dl, za.za_first_integer, mintxg, tx);
	VERIFY0(zap_remove_int(dl->dl_os, obj, mintxg, tx));
	}
	VERIFY3U(error, ==, ENOENT);
	zap_cursor_fini(&zc);

	VERIFY0(dmu_bonus_hold(dl->dl_os, obj, FTAG, &bonus));
	dlp = bonus->db_data;
	dmu_buf_will_dirty(bonus, tx);
	memset(dlp, 0, sizeof (*dlp));
	dmu_buf_rele(bonus, FTAG);
	mutex_exit(&dl->dl_lock);
	}

	/*
	* Remove entries on dl that are born > mintxg, and put them on the bpobj.
	*/
	void
	dsl_deadlist_move_bpobj(dsl_deadlist_t dl, bpobj_t bpo, uint64_t mintxg,
	dmu_tx_t *tx)
	{
	dsl_deadlist_entry_t dle_tofind;
	dsl_deadlist_entry_t *dle;
	avl_index_t where;

	ASSERT(!dl->dl_oldfmt);

	mutex_enter(&dl->dl_lock);
	dmu_buf_will_dirty(dl->dl_dbuf, tx);
	dsl_deadlist_load_tree(dl);

	dle_tofind.dle_mintxg = mintxg;
	dle = avl_find(&dl->dl_tree, &dle_tofind, &where);
	if (dle == NULL)
	dle = avl_nearest(&dl->dl_tree, where, AVL_AFTER);
	while (dle) {
	uint64_t used, comp, uncomp;
	dsl_deadlist_entry_t *dle_next;

	bpobj_enqueue_subobj(bpo, dle->dle_bpobj.bpo_object, tx);

	VERIFY0(bpobj_space(&dle->dle_bpobj,
	&used, &comp, &uncomp));
	ASSERT3U(dl->dl_phys->dl_used, >=, used);
	ASSERT3U(dl->dl_phys->dl_comp, >=, comp);
	ASSERT3U(dl->dl_phys->dl_uncomp, >=, uncomp);
	dl->dl_phys->dl_used -= used;
	dl->dl_phys->dl_comp -= comp;
	dl->dl_phys->dl_uncomp -= uncomp;

	VERIFY0(zap_remove_int(dl->dl_os, dl->dl_object,
	dle->dle_mintxg, tx));

	dle_next = AVL_NEXT(&dl->dl_tree, dle);
	avl_remove(&dl->dl_tree, dle);
	bpobj_close(&dle->dle_bpobj);
	kmem_free(dle, sizeof (*dle));
	dle = dle_next;
	}
	mutex_exit(&dl->dl_lock);
	}

	typedef struct livelist_entry {
	blkptr_t le_bp;
	uint32_t le_refcnt;
	avl_node_t le_node;
	} livelist_entry_t;

	static int
	livelist_compare(const void larg, const void rarg)
	{
	const blkptr_t l = &((livelist_entry_t )larg)->le_bp;
	const blkptr_t r = &((livelist_entry_t )rarg)->le_bp;

	/* Sort them according to dva[0] */
	uint64_t l_dva0_vdev = DVA_GET_VDEV(&l->blk_dva[0]);
	uint64_t r_dva0_vdev = DVA_GET_VDEV(&r->blk_dva[0]);

	if (l_dva0_vdev != r_dva0_vdev)
	return (TREE_CMP(l_dva0_vdev, r_dva0_vdev));

	/* if vdevs are equal, sort by offsets. */
	uint64_t l_dva0_offset = DVA_GET_OFFSET(&l->blk_dva[0]);
	uint64_t r_dva0_offset = DVA_GET_OFFSET(&r->blk_dva[0]);
	if (l_dva0_offset == r_dva0_offset)
	ASSERT3U(l->blk_birth, ==, r->blk_birth);
	return (TREE_CMP(l_dva0_offset, r_dva0_offset));
	}

	struct livelist_iter_arg {
	avl_tree_t *avl;
	bplist_t *to_free;
	zthr_t *t;
	};

	/*
	* Expects an AVL tree which is incrementally filled will FREE blkptrs
	* and used to match up ALLOC/FREE pairs. ALLOC'd blkptrs without a
	* corresponding FREE are stored in the supplied bplist.
	*
	* Note that multiple FREE and ALLOC entries for the same blkptr may
	* be encountered when dedup is involved. For this reason we keep a
	* refcount for all the FREE entries of each blkptr and ensure that
	* each of those FREE entries has a corresponding ALLOC preceding it.
	*/
	static int
	dsl_livelist_iterate(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	struct livelist_iter_arg *lia = arg;
	avl_tree_t *avl = lia->avl;
	bplist_t *to_free = lia->to_free;
	zthr_t *t = lia->t;
	ASSERT(tx == NULL);

	if ((t != NULL) && (zthr_has_waiters(t) \|\| zthr_iscancelled(t)))
	return (SET_ERROR(EINTR));

	livelist_entry_t node;
	node.le_bp = *bp;
	livelist_entry_t *found = avl_find(avl, &node, NULL);
	if (bp_freed) {
	if (found == NULL) {
	/* first free entry for this blkptr */
	livelist_entry_t *e =
	kmem_alloc(sizeof (livelist_entry_t), KM_SLEEP);
	e->le_bp = *bp;
	e->le_refcnt = 1;
	avl_add(avl, e);
	} else {
	/* dedup block free */
	ASSERT(BP_GET_DEDUP(bp));
	ASSERT3U(BP_GET_CHECKSUM(bp), ==,
	BP_GET_CHECKSUM(&found->le_bp));
	ASSERT3U(found->le_refcnt + 1, >, found->le_refcnt);
	found->le_refcnt++;
	}
	} else {
	if (found == NULL) {
	/* block is currently marked as allocated */
	bplist_append(to_free, bp);
	} else {
	/* alloc matches a free entry */
	ASSERT3U(found->le_refcnt, !=, 0);
	found->le_refcnt--;
	if (found->le_refcnt == 0) {
	/* all tracked free pairs have been matched */
	avl_remove(avl, found);
	kmem_free(found, sizeof (livelist_entry_t));
	} else {
	/*
	* This is definitely a deduped blkptr so
	* let's validate it.
	*/
	ASSERT(BP_GET_DEDUP(bp));
	ASSERT3U(BP_GET_CHECKSUM(bp), ==,
	BP_GET_CHECKSUM(&found->le_bp));
	}
	}
	}
	return (0);
	}

	/*
	* Accepts a bpobj and a bplist. Will insert into the bplist the blkptrs
	* which have an ALLOC entry but no matching FREE
	*/
	int
	dsl_process_sub_livelist(bpobj_t bpobj, bplist_t to_free, zthr_t *t,
	uint64_t *size)
	{
	avl_tree_t avl;
	avl_create(&avl, livelist_compare, sizeof (livelist_entry_t),
	offsetof(livelist_entry_t, le_node));

	/* process the sublist */
	struct livelist_iter_arg arg = {
	.avl = &avl,
	.to_free = to_free,
	.t = t
	};
	int err = bpobj_iterate_nofree(bpobj, dsl_livelist_iterate, &arg, size);
	+ VERIFY(err != 0 \|\| avl_numnodes(&avl) == 0);

	- VERIFY0(avl_numnodes(&avl));
	+ void *cookie = NULL;
	+ livelist_entry_t *le = NULL;
	+ while ((le = avl_destroy_nodes(&avl, &cookie)) != NULL) {
	+ kmem_free(le, sizeof (livelist_entry_t));
	+ }
	avl_destroy(&avl);
	return (err);
	}

	ZFS_MODULE_PARAM(zfs_livelist, zfs_livelist_, max_entries, ULONG, ZMOD_RW,
	"Size to start the next sub-livelist in a livelist");

	ZFS_MODULE_PARAM(zfs_livelist, zfs_livelist_, min_percent_shared, INT, ZMOD_RW,
	"Threshold at which livelist is disabled");
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_dir.c b/sys/contrib/openzfs/module/zfs/dsl_dir.c
	index a4db3ee2f306..d93c7f08c1c2 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_dir.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_dir.c
	@@ -1,2478 +1,2476 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2013 Martin Matuska. All rights reserved.
	* Copyright (c) 2014 Joyent, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_deleg.h>
	#include <sys/dmu_impl.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/metaslab.h>
	#include <sys/zap.h>
	#include <sys/zio.h>
	#include <sys/arc.h>
	#include <sys/sunddi.h>
	#include <sys/zfeature.h>
	#include <sys/policy.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_znode.h>
	#include <sys/zvol.h>
	#include <sys/zthr.h>
	#include "zfs_namecheck.h"
	#include "zfs_prop.h"

	/*
	* Filesystem and Snapshot Limits
	* ------------------------------
	*
	* These limits are used to restrict the number of filesystems and/or snapshots
	* that can be created at a given level in the tree or below. A typical
	* use-case is with a delegated dataset where the administrator wants to ensure
	* that a user within the zone is not creating too many additional filesystems
	* or snapshots, even though they're not exceeding their space quota.
	*
	* The filesystem and snapshot counts are stored as extensible properties. This
	* capability is controlled by a feature flag and must be enabled to be used.
	* Once enabled, the feature is not active until the first limit is set. At
	* that point, future operations to create/destroy filesystems or snapshots
	* will validate and update the counts.
	*
	* Because the count properties will not exist before the feature is active,
	* the counts are updated when a limit is first set on an uninitialized
	* dsl_dir node in the tree (The filesystem/snapshot count on a node includes
	* all of the nested filesystems/snapshots. Thus, a new leaf node has a
	* filesystem count of 0 and a snapshot count of 0. Non-existent filesystem and
	* snapshot count properties on a node indicate uninitialized counts on that
	* node.) When first setting a limit on an uninitialized node, the code starts
	* at the filesystem with the new limit and descends into all sub-filesystems
	* to add the count properties.
	*
	* In practice this is lightweight since a limit is typically set when the
	* filesystem is created and thus has no children. Once valid, changing the
	* limit value won't require a re-traversal since the counts are already valid.
	* When recursively fixing the counts, if a node with a limit is encountered
	* during the descent, the counts are known to be valid and there is no need to
	* descend into that filesystem's children. The counts on filesystems above the
	* one with the new limit will still be uninitialized, unless a limit is
	* eventually set on one of those filesystems. The counts are always recursively
	* updated when a limit is set on a dataset, unless there is already a limit.
	* When a new limit value is set on a filesystem with an existing limit, it is
	* possible for the new limit to be less than the current count at that level
	* since a user who can change the limit is also allowed to exceed the limit.
	*
	* Once the feature is active, then whenever a filesystem or snapshot is
	* created, the code recurses up the tree, validating the new count against the
	* limit at each initialized level. In practice, most levels will not have a
	* limit set. If there is a limit at any initialized level up the tree, the
	* check must pass or the creation will fail. Likewise, when a filesystem or
	* snapshot is destroyed, the counts are recursively adjusted all the way up
	* the initialized nodes in the tree. Renaming a filesystem into different point
	* in the tree will first validate, then update the counts on each branch up to
	* the common ancestor. A receive will also validate the counts and then update
	* them.
	*
	* An exception to the above behavior is that the limit is not enforced if the
	* user has permission to modify the limit. This is primarily so that
	* recursive snapshots in the global zone always work. We want to prevent a
	* denial-of-service in which a lower level delegated dataset could max out its
	* limit and thus block recursive snapshots from being taken in the global zone.
	* Because of this, it is possible for the snapshot count to be over the limit
	* and snapshots taken in the global zone could cause a lower level dataset to
	* hit or exceed its limit. The administrator taking the global zone recursive
	* snapshot should be aware of this side-effect and behave accordingly.
	* For consistency, the filesystem limit is also not enforced if the user can
	* modify the limit.
	*
	* The filesystem and snapshot limits are validated by dsl_fs_ss_limit_check()
	* and updated by dsl_fs_ss_count_adjust(). A new limit value is setup in
	* dsl_dir_activate_fs_ss_limit() and the counts are adjusted, if necessary, by
	* dsl_dir_init_fs_ss_count().
	*/

	static uint64_t dsl_dir_space_towrite(dsl_dir_t *dd);

	typedef struct ddulrt_arg {
	dsl_dir_t *ddulrta_dd;
	uint64_t ddlrta_txg;
	} ddulrt_arg_t;

	static void
	dsl_dir_evict_async(void *dbu)
	{
	dsl_dir_t *dd = dbu;
	int t;
	dsl_pool_t *dp __maybe_unused = dd->dd_pool;

	dd->dd_dbuf = NULL;

	for (t = 0; t < TXG_SIZE; t++) {
	ASSERT(!txg_list_member(&dp->dp_dirty_dirs, dd, t));
	ASSERT(dd->dd_tempreserved[t] == 0);
	ASSERT(dd->dd_space_towrite[t] == 0);
	}

	if (dd->dd_parent)
	dsl_dir_async_rele(dd->dd_parent, dd);

	spa_async_close(dd->dd_pool->dp_spa, dd);

	if (dsl_deadlist_is_open(&dd->dd_livelist))
	dsl_dir_livelist_close(dd);

	dsl_prop_fini(dd);
	cv_destroy(&dd->dd_activity_cv);
	mutex_destroy(&dd->dd_activity_lock);
	mutex_destroy(&dd->dd_lock);
	kmem_free(dd, sizeof (dsl_dir_t));
	}

	int
	dsl_dir_hold_obj(dsl_pool_t *dp, uint64_t ddobj,
	const char tail, const void tag, dsl_dir_t **ddp)
	{
	dmu_buf_t *dbuf;
	dsl_dir_t *dd;
	dmu_object_info_t doi;
	int err;

	ASSERT(dsl_pool_config_held(dp));

	err = dmu_bonus_hold(dp->dp_meta_objset, ddobj, tag, &dbuf);
	if (err != 0)
	return (err);
	dd = dmu_buf_get_user(dbuf);

	dmu_object_info_from_db(dbuf, &doi);
	ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_DSL_DIR);
	ASSERT3U(doi.doi_bonus_size, >=, sizeof (dsl_dir_phys_t));

	if (dd == NULL) {
	dsl_dir_t *winner;

	dd = kmem_zalloc(sizeof (dsl_dir_t), KM_SLEEP);
	dd->dd_object = ddobj;
	dd->dd_dbuf = dbuf;
	dd->dd_pool = dp;

	mutex_init(&dd->dd_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&dd->dd_activity_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&dd->dd_activity_cv, NULL, CV_DEFAULT, NULL);
	dsl_prop_init(dd);

	if (dsl_dir_is_zapified(dd)) {
	err = zap_lookup(dp->dp_meta_objset,
	ddobj, DD_FIELD_CRYPTO_KEY_OBJ,
	sizeof (uint64_t), 1, &dd->dd_crypto_obj);
	if (err == 0) {
	/* check for on-disk format errata */
	if (dsl_dir_incompatible_encryption_version(
	dd)) {
	dp->dp_spa->spa_errata =
	ZPOOL_ERRATA_ZOL_6845_ENCRYPTION;
	}
	} else if (err != ENOENT) {
	goto errout;
	}
	}

	if (dsl_dir_phys(dd)->dd_parent_obj) {
	err = dsl_dir_hold_obj(dp,
	dsl_dir_phys(dd)->dd_parent_obj, NULL, dd,
	&dd->dd_parent);
	if (err != 0)
	goto errout;
	if (tail) {
	#ifdef ZFS_DEBUG
	uint64_t foundobj;

	err = zap_lookup(dp->dp_meta_objset,
	dsl_dir_phys(dd->dd_parent)->
	dd_child_dir_zapobj, tail,
	sizeof (foundobj), 1, &foundobj);
	ASSERT(err \|\| foundobj == ddobj);
	#endif
	(void) strlcpy(dd->dd_myname, tail,
	sizeof (dd->dd_myname));
	} else {
	err = zap_value_search(dp->dp_meta_objset,
	dsl_dir_phys(dd->dd_parent)->
	dd_child_dir_zapobj,
	ddobj, 0, dd->dd_myname);
	}
	if (err != 0)
	goto errout;
	} else {
	(void) strlcpy(dd->dd_myname, spa_name(dp->dp_spa),
	sizeof (dd->dd_myname));
	}

	if (dsl_dir_is_clone(dd)) {
	dmu_buf_t *origin_bonus;
	dsl_dataset_phys_t *origin_phys;

	/*
	* We can't open the origin dataset, because
	* that would require opening this dsl_dir.
	* Just look at its phys directly instead.
	*/
	err = dmu_bonus_hold(dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_origin_obj, FTAG,
	&origin_bonus);
	if (err != 0)
	goto errout;
	origin_phys = origin_bonus->db_data;
	dd->dd_origin_txg =
	origin_phys->ds_creation_txg;
	dmu_buf_rele(origin_bonus, FTAG);
	if (dsl_dir_is_zapified(dd)) {
	uint64_t obj;
	err = zap_lookup(dp->dp_meta_objset,
	dd->dd_object, DD_FIELD_LIVELIST,
	sizeof (uint64_t), 1, &obj);
	if (err == 0)
	dsl_dir_livelist_open(dd, obj);
	else if (err != ENOENT)
	goto errout;
	}
	}

	if (dsl_dir_is_zapified(dd)) {
	inode_timespec_t t = {0};
	- zap_lookup(dp->dp_meta_objset, ddobj,
	+ (void) zap_lookup(dp->dp_meta_objset, ddobj,
	DD_FIELD_SNAPSHOTS_CHANGED,
	sizeof (uint64_t),
	sizeof (inode_timespec_t) / sizeof (uint64_t),
	&t);
	dd->dd_snap_cmtime = t;
	}

	dmu_buf_init_user(&dd->dd_dbu, NULL, dsl_dir_evict_async,
	&dd->dd_dbuf);
	winner = dmu_buf_set_user_ie(dbuf, &dd->dd_dbu);
	if (winner != NULL) {
	if (dd->dd_parent)
	dsl_dir_rele(dd->dd_parent, dd);
	if (dsl_deadlist_is_open(&dd->dd_livelist))
	dsl_dir_livelist_close(dd);
	dsl_prop_fini(dd);
	cv_destroy(&dd->dd_activity_cv);
	mutex_destroy(&dd->dd_activity_lock);
	mutex_destroy(&dd->dd_lock);
	kmem_free(dd, sizeof (dsl_dir_t));
	dd = winner;
	} else {
	spa_open_ref(dp->dp_spa, dd);
	}
	}

	/*
	* The dsl_dir_t has both open-to-close and instantiate-to-evict
	* holds on the spa. We need the open-to-close holds because
	* otherwise the spa_refcnt wouldn't change when we open a
	* dir which the spa also has open, so we could incorrectly
	* think it was OK to unload/export/destroy the pool. We need
	* the instantiate-to-evict hold because the dsl_dir_t has a
	* pointer to the dd_pool, which has a pointer to the spa_t.
	*/
	spa_open_ref(dp->dp_spa, tag);
	ASSERT3P(dd->dd_pool, ==, dp);
	ASSERT3U(dd->dd_object, ==, ddobj);
	ASSERT3P(dd->dd_dbuf, ==, dbuf);
	*ddp = dd;
	return (0);

	errout:
	if (dd->dd_parent)
	dsl_dir_rele(dd->dd_parent, dd);
	if (dsl_deadlist_is_open(&dd->dd_livelist))
	dsl_dir_livelist_close(dd);
	dsl_prop_fini(dd);
	cv_destroy(&dd->dd_activity_cv);
	mutex_destroy(&dd->dd_activity_lock);
	mutex_destroy(&dd->dd_lock);
	kmem_free(dd, sizeof (dsl_dir_t));
	dmu_buf_rele(dbuf, tag);
	return (err);
	}

	void
	dsl_dir_rele(dsl_dir_t dd, const void tag)
	{
	dprintf_dd(dd, "%s\n", "");
	spa_close(dd->dd_pool->dp_spa, tag);
	dmu_buf_rele(dd->dd_dbuf, tag);
	}

	/*
	* Remove a reference to the given dsl dir that is being asynchronously
	* released. Async releases occur from a taskq performing eviction of
	* dsl datasets and dirs. This process is identical to a normal release
	* with the exception of using the async API for releasing the reference on
	* the spa.
	*/
	void
	dsl_dir_async_rele(dsl_dir_t dd, const void tag)
	{
	dprintf_dd(dd, "%s\n", "");
	spa_async_close(dd->dd_pool->dp_spa, tag);
	dmu_buf_rele(dd->dd_dbuf, tag);
	}

	/* buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes */
	void
	dsl_dir_name(dsl_dir_t dd, char buf)
	{
	if (dd->dd_parent) {
	dsl_dir_name(dd->dd_parent, buf);
	VERIFY3U(strlcat(buf, "/", ZFS_MAX_DATASET_NAME_LEN), <,
	ZFS_MAX_DATASET_NAME_LEN);
	} else {
	buf[0] = '\0';
	}
	if (!MUTEX_HELD(&dd->dd_lock)) {
	/*
	* recursive mutex so that we can use
	* dprintf_dd() with dd_lock held
	*/
	mutex_enter(&dd->dd_lock);
	VERIFY3U(strlcat(buf, dd->dd_myname, ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	mutex_exit(&dd->dd_lock);
	} else {
	VERIFY3U(strlcat(buf, dd->dd_myname, ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	}
	}

	/* Calculate name length, avoiding all the strcat calls of dsl_dir_name */
	int
	dsl_dir_namelen(dsl_dir_t *dd)
	{
	int result = 0;

	if (dd->dd_parent) {
	/* parent's name + 1 for the "/" */
	result = dsl_dir_namelen(dd->dd_parent) + 1;
	}

	if (!MUTEX_HELD(&dd->dd_lock)) {
	/* see dsl_dir_name */
	mutex_enter(&dd->dd_lock);
	result += strlen(dd->dd_myname);
	mutex_exit(&dd->dd_lock);
	} else {
	result += strlen(dd->dd_myname);
	}

	return (result);
	}

	static int
	getcomponent(const char path, char component, const char **nextp)
	{
	char *p;

	if ((path == NULL) \|\| (path[0] == '\0'))
	return (SET_ERROR(ENOENT));
	/* This would be a good place to reserve some namespace... */
	p = strpbrk(path, "/@");
	if (p && (p[1] == '/' \|\| p[1] == '@')) {
	/* two separators in a row */
	return (SET_ERROR(EINVAL));
	}
	if (p == NULL \|\| p == path) {
	/*
	* if the first thing is an @ or /, it had better be an
	* @ and it had better not have any more ats or slashes,
	* and it had better have something after the @.
	*/
	if (p != NULL &&
	(p[0] != '@' \|\| strpbrk(path+1, "/@") \|\| p[1] == '\0'))
	return (SET_ERROR(EINVAL));
	if (strlen(path) >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));
	(void) strlcpy(component, path, ZFS_MAX_DATASET_NAME_LEN);
	p = NULL;
	} else if (p[0] == '/') {
	if (p - path >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));
	- (void) strncpy(component, path, p - path);
	- component[p - path] = '\0';
	+ (void) strlcpy(component, path, p - path + 1);
	p++;
	} else if (p[0] == '@') {
	/*
	* if the next separator is an @, there better not be
	* any more slashes.
	*/
	if (strchr(path, '/'))
	return (SET_ERROR(EINVAL));
	if (p - path >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));
	- (void) strncpy(component, path, p - path);
	- component[p - path] = '\0';
	+ (void) strlcpy(component, path, p - path + 1);
	} else {
	panic("invalid p=%p", (void *)p);
	}
	*nextp = p;
	return (0);
	}

	/*
	* Return the dsl_dir_t, and possibly the last component which couldn't
	* be found in *tail. The name must be in the specified dsl_pool_t. This
	* thread must hold the dp_config_rwlock for the pool. Returns NULL if the
	* path is bogus, or if tail==NULL and we couldn't parse the whole name.
	* (*tail)[0] == '@' means that the last component is a snapshot.
	*/
	int
	dsl_dir_hold(dsl_pool_t dp, const char name, const void *tag,
	dsl_dir_t ddp, const char tailp)
	{
	char *buf;
	const char spaname, next, *nextnext = NULL;
	int err;
	dsl_dir_t *dd;
	uint64_t ddobj;

	buf = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
	err = getcomponent(name, buf, &next);
	if (err != 0)
	goto error;

	/* Make sure the name is in the specified pool. */
	spaname = spa_name(dp->dp_spa);
	if (strcmp(buf, spaname) != 0) {
	err = SET_ERROR(EXDEV);
	goto error;
	}

	ASSERT(dsl_pool_config_held(dp));

	err = dsl_dir_hold_obj(dp, dp->dp_root_dir_obj, NULL, tag, &dd);
	if (err != 0) {
	goto error;
	}

	while (next != NULL) {
	dsl_dir_t *child_dd;
	err = getcomponent(next, buf, &nextnext);
	if (err != 0)
	break;
	ASSERT(next[0] != '\0');
	if (next[0] == '@')
	break;
	dprintf("looking up %s in obj%lld\n",
	buf, (longlong_t)dsl_dir_phys(dd)->dd_child_dir_zapobj);

	err = zap_lookup(dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_child_dir_zapobj,
	buf, sizeof (ddobj), 1, &ddobj);
	if (err != 0) {
	if (err == ENOENT)
	err = 0;
	break;
	}

	err = dsl_dir_hold_obj(dp, ddobj, buf, tag, &child_dd);
	if (err != 0)
	break;
	dsl_dir_rele(dd, tag);
	dd = child_dd;
	next = nextnext;
	}

	if (err != 0) {
	dsl_dir_rele(dd, tag);
	goto error;
	}

	/*
	* It's an error if there's more than one component left, or
	* tailp==NULL and there's any component left.
	*/
	if (next != NULL &&
	(tailp == NULL \|\| (nextnext && nextnext[0] != '\0'))) {
	/* bad path name */
	dsl_dir_rele(dd, tag);
	dprintf("next=%p (%s) tail=%p\n", next, next?next:"", tailp);
	err = SET_ERROR(ENOENT);
	}
	if (tailp != NULL)
	*tailp = next;
	if (err == 0)
	*ddp = dd;
	error:
	kmem_free(buf, ZFS_MAX_DATASET_NAME_LEN);
	return (err);
	}

	/*
	* If the counts are already initialized for this filesystem and its
	* descendants then do nothing, otherwise initialize the counts.
	*
	* The counts on this filesystem, and those below, may be uninitialized due to
	* either the use of a pre-existing pool which did not support the
	* filesystem/snapshot limit feature, or one in which the feature had not yet
	* been enabled.
	*
	* Recursively descend the filesystem tree and update the filesystem/snapshot
	* counts on each filesystem below, then update the cumulative count on the
	* current filesystem. If the filesystem already has a count set on it,
	* then we know that its counts, and the counts on the filesystems below it,
	* are already correct, so we don't have to update this filesystem.
	*/
	static void
	dsl_dir_init_fs_ss_count(dsl_dir_t dd, dmu_tx_t tx)
	{
	uint64_t my_fs_cnt = 0;
	uint64_t my_ss_cnt = 0;
	dsl_pool_t *dp = dd->dd_pool;
	objset_t *os = dp->dp_meta_objset;
	zap_cursor_t *zc;
	zap_attribute_t *za;
	dsl_dataset_t *ds;

	ASSERT(spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT));
	ASSERT(dsl_pool_config_held(dp));
	ASSERT(dmu_tx_is_syncing(tx));

	dsl_dir_zapify(dd, tx);

	/*
	* If the filesystem count has already been initialized then we
	* don't need to recurse down any further.
	*/
	if (zap_contains(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT) == 0)
	return;

	zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
	za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);

	/* Iterate my child dirs */
	for (zap_cursor_init(zc, os, dsl_dir_phys(dd)->dd_child_dir_zapobj);
	zap_cursor_retrieve(zc, za) == 0; zap_cursor_advance(zc)) {
	dsl_dir_t *chld_dd;
	uint64_t count;

	VERIFY0(dsl_dir_hold_obj(dp, za->za_first_integer, NULL, FTAG,
	&chld_dd));

	/*
	* Ignore hidden ($FREE, $MOS & $ORIGIN) objsets.
	*/
	if (chld_dd->dd_myname[0] == '$') {
	dsl_dir_rele(chld_dd, FTAG);
	continue;
	}

	my_fs_cnt++; /* count this child */

	dsl_dir_init_fs_ss_count(chld_dd, tx);

	VERIFY0(zap_lookup(os, chld_dd->dd_object,
	DD_FIELD_FILESYSTEM_COUNT, sizeof (count), 1, &count));
	my_fs_cnt += count;
	VERIFY0(zap_lookup(os, chld_dd->dd_object,
	DD_FIELD_SNAPSHOT_COUNT, sizeof (count), 1, &count));
	my_ss_cnt += count;

	dsl_dir_rele(chld_dd, FTAG);
	}
	zap_cursor_fini(zc);
	/* Count my snapshots (we counted children's snapshots above) */
	VERIFY0(dsl_dataset_hold_obj(dd->dd_pool,
	dsl_dir_phys(dd)->dd_head_dataset_obj, FTAG, &ds));

	for (zap_cursor_init(zc, os, dsl_dataset_phys(ds)->ds_snapnames_zapobj);
	zap_cursor_retrieve(zc, za) == 0;
	zap_cursor_advance(zc)) {
	/* Don't count temporary snapshots */
	if (za->za_name[0] != '%')
	my_ss_cnt++;
	}
	zap_cursor_fini(zc);

	dsl_dataset_rele(ds, FTAG);

	kmem_free(zc, sizeof (zap_cursor_t));
	kmem_free(za, sizeof (zap_attribute_t));

	/* we're in a sync task, update counts */
	dmu_buf_will_dirty(dd->dd_dbuf, tx);
	VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT,
	sizeof (my_fs_cnt), 1, &my_fs_cnt, tx));
	VERIFY0(zap_add(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT,
	sizeof (my_ss_cnt), 1, &my_ss_cnt, tx));
	}

	static int
	dsl_dir_actv_fs_ss_limit_check(void arg, dmu_tx_t tx)
	{
	char ddname = (char )arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	dsl_dir_t *dd;
	int error;

	error = dsl_dataset_hold(dp, ddname, FTAG, &ds);
	if (error != 0)
	return (error);

	if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	dd = ds->ds_dir;
	if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_FS_SS_LIMIT) &&
	dsl_dir_is_zapified(dd) &&
	zap_contains(dp->dp_meta_objset, dd->dd_object,
	DD_FIELD_FILESYSTEM_COUNT) == 0) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EALREADY));
	}

	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	static void
	dsl_dir_actv_fs_ss_limit_sync(void arg, dmu_tx_t tx)
	{
	char ddname = (char )arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	spa_t *spa;

	VERIFY0(dsl_dataset_hold(dp, ddname, FTAG, &ds));

	spa = dsl_dataset_get_spa(ds);

	if (!spa_feature_is_active(spa, SPA_FEATURE_FS_SS_LIMIT)) {
	/*
	* Since the feature was not active and we're now setting a
	* limit, increment the feature-active counter so that the
	* feature becomes active for the first time.
	*
	* We are already in a sync task so we can update the MOS.
	*/
	spa_feature_incr(spa, SPA_FEATURE_FS_SS_LIMIT, tx);
	}

	/*
	* Since we are now setting a non-UINT64_MAX limit on the filesystem,
	* we need to ensure the counts are correct. Descend down the tree from
	* this point and update all of the counts to be accurate.
	*/
	dsl_dir_init_fs_ss_count(ds->ds_dir, tx);

	dsl_dataset_rele(ds, FTAG);
	}

	/*
	* Make sure the feature is enabled and activate it if necessary.
	* Since we're setting a limit, ensure the on-disk counts are valid.
	* This is only called by the ioctl path when setting a limit value.
	*
	* We do not need to validate the new limit, since users who can change the
	* limit are also allowed to exceed the limit.
	*/
	int
	dsl_dir_activate_fs_ss_limit(const char *ddname)
	{
	int error;

	error = dsl_sync_task(ddname, dsl_dir_actv_fs_ss_limit_check,
	dsl_dir_actv_fs_ss_limit_sync, (void *)ddname, 0,
	ZFS_SPACE_CHECK_RESERVED);

	if (error == EALREADY)
	error = 0;

	return (error);
	}

	/*
	* Used to determine if the filesystem_limit or snapshot_limit should be
	* enforced. We allow the limit to be exceeded if the user has permission to
	* write the property value. We pass in the creds that we got in the open
	* context since we will always be the GZ root in syncing context. We also have
	* to handle the case where we are allowed to change the limit on the current
	* dataset, but there may be another limit in the tree above.
	*
	* We can never modify these two properties within a non-global zone. In
	* addition, the other checks are modeled on zfs_secpolicy_write_perms. We
	* can't use that function since we are already holding the dp_config_rwlock.
	* In addition, we already have the dd and dealing with snapshots is simplified
	* in this code.
	*/

	typedef enum {
	ENFORCE_ALWAYS,
	ENFORCE_NEVER,
	ENFORCE_ABOVE
	} enforce_res_t;

	static enforce_res_t
	dsl_enforce_ds_ss_limits(dsl_dir_t *dd, zfs_prop_t prop,
	cred_t cr, proc_t proc)
	{
	enforce_res_t enforce = ENFORCE_ALWAYS;
	uint64_t obj;
	dsl_dataset_t *ds;
	uint64_t zoned;
	const char *zonedstr;

	ASSERT(prop == ZFS_PROP_FILESYSTEM_LIMIT \|\|
	prop == ZFS_PROP_SNAPSHOT_LIMIT);

	#ifdef _KERNEL
	if (crgetzoneid(cr) != GLOBAL_ZONEID)
	return (ENFORCE_ALWAYS);

	/*
	* We are checking the saved credentials of the user process, which is
	* not the current process. Note that we can't use secpolicy_zfs(),
	* because it only works if the cred is that of the current process (on
	* Linux).
	*/
	if (secpolicy_zfs_proc(cr, proc) == 0)
	return (ENFORCE_NEVER);
	#else
	(void) proc;
	#endif

	if ((obj = dsl_dir_phys(dd)->dd_head_dataset_obj) == 0)
	return (ENFORCE_ALWAYS);

	ASSERT(dsl_pool_config_held(dd->dd_pool));

	if (dsl_dataset_hold_obj(dd->dd_pool, obj, FTAG, &ds) != 0)
	return (ENFORCE_ALWAYS);

	zonedstr = zfs_prop_to_name(ZFS_PROP_ZONED);
	if (dsl_prop_get_ds(ds, zonedstr, 8, 1, &zoned, NULL) \|\| zoned) {
	/* Only root can access zoned fs's from the GZ */
	enforce = ENFORCE_ALWAYS;
	} else {
	if (dsl_deleg_access_impl(ds, zfs_prop_to_name(prop), cr) == 0)
	enforce = ENFORCE_ABOVE;
	}

	dsl_dataset_rele(ds, FTAG);
	return (enforce);
	}

	/*
	* Check if adding additional child filesystem(s) would exceed any filesystem
	* limits or adding additional snapshot(s) would exceed any snapshot limits.
	* The prop argument indicates which limit to check.
	*
	* Note that all filesystem limits up to the root (or the highest
	* initialized) filesystem or the given ancestor must be satisfied.
	*/
	int
	dsl_fs_ss_limit_check(dsl_dir_t *dd, uint64_t delta, zfs_prop_t prop,
	dsl_dir_t ancestor, cred_t cr, proc_t *proc)
	{
	objset_t *os = dd->dd_pool->dp_meta_objset;
	uint64_t limit, count;
	const char *count_prop;
	enforce_res_t enforce;
	int err = 0;

	ASSERT(dsl_pool_config_held(dd->dd_pool));
	ASSERT(prop == ZFS_PROP_FILESYSTEM_LIMIT \|\|
	prop == ZFS_PROP_SNAPSHOT_LIMIT);

	/*
	* If we're allowed to change the limit, don't enforce the limit
	* e.g. this can happen if a snapshot is taken by an administrative
	* user in the global zone (i.e. a recursive snapshot by root).
	* However, we must handle the case of delegated permissions where we
	* are allowed to change the limit on the current dataset, but there
	* is another limit in the tree above.
	*/
	enforce = dsl_enforce_ds_ss_limits(dd, prop, cr, proc);
	if (enforce == ENFORCE_NEVER)
	return (0);

	/*
	* e.g. if renaming a dataset with no snapshots, count adjustment
	* is 0.
	*/
	if (delta == 0)
	return (0);

	if (prop == ZFS_PROP_SNAPSHOT_LIMIT) {
	/*
	* We don't enforce the limit for temporary snapshots. This is
	* indicated by a NULL cred_t argument.
	*/
	if (cr == NULL)
	return (0);

	count_prop = DD_FIELD_SNAPSHOT_COUNT;
	} else {
	count_prop = DD_FIELD_FILESYSTEM_COUNT;
	}

	/*
	* If an ancestor has been provided, stop checking the limit once we
	* hit that dir. We need this during rename so that we don't overcount
	* the check once we recurse up to the common ancestor.
	*/
	if (ancestor == dd)
	return (0);

	/*
	* If we hit an uninitialized node while recursing up the tree, we can
	* stop since we know there is no limit here (or above). The counts are
	* not valid on this node and we know we won't touch this node's counts.
	*/
	if (!dsl_dir_is_zapified(dd))
	return (0);
	err = zap_lookup(os, dd->dd_object,
	count_prop, sizeof (count), 1, &count);
	if (err == ENOENT)
	return (0);
	if (err != 0)
	return (err);

	err = dsl_prop_get_dd(dd, zfs_prop_to_name(prop), 8, 1, &limit, NULL,
	B_FALSE);
	if (err != 0)
	return (err);

	/* Is there a limit which we've hit? */
	if (enforce == ENFORCE_ALWAYS && (count + delta) > limit)
	return (SET_ERROR(EDQUOT));

	if (dd->dd_parent != NULL)
	err = dsl_fs_ss_limit_check(dd->dd_parent, delta, prop,
	ancestor, cr, proc);

	return (err);
	}

	/*
	* Adjust the filesystem or snapshot count for the specified dsl_dir_t and all
	* parents. When a new filesystem/snapshot is created, increment the count on
	* all parents, and when a filesystem/snapshot is destroyed, decrement the
	* count.
	*/
	void
	dsl_fs_ss_count_adjust(dsl_dir_t dd, int64_t delta, const char prop,
	dmu_tx_t *tx)
	{
	int err;
	objset_t *os = dd->dd_pool->dp_meta_objset;
	uint64_t count;

	ASSERT(dsl_pool_config_held(dd->dd_pool));
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(strcmp(prop, DD_FIELD_FILESYSTEM_COUNT) == 0 \|\|
	strcmp(prop, DD_FIELD_SNAPSHOT_COUNT) == 0);

	/*
	* We don't do accounting for hidden ($FREE, $MOS & $ORIGIN) objsets.
	*/
	if (dd->dd_myname[0] == '$' && strcmp(prop,
	DD_FIELD_FILESYSTEM_COUNT) == 0) {
	return;
	}

	/*
	* e.g. if renaming a dataset with no snapshots, count adjustment is 0
	*/
	if (delta == 0)
	return;

	/*
	* If we hit an uninitialized node while recursing up the tree, we can
	* stop since we know the counts are not valid on this node and we
	* know we shouldn't touch this node's counts. An uninitialized count
	* on the node indicates that either the feature has not yet been
	* activated or there are no limits on this part of the tree.
	*/
	if (!dsl_dir_is_zapified(dd) \|\| (err = zap_lookup(os, dd->dd_object,
	prop, sizeof (count), 1, &count)) == ENOENT)
	return;
	VERIFY0(err);

	count += delta;
	/* Use a signed verify to make sure we're not neg. */
	VERIFY3S(count, >=, 0);

	VERIFY0(zap_update(os, dd->dd_object, prop, sizeof (count), 1, &count,
	tx));

	/* Roll up this additional count into our ancestors */
	if (dd->dd_parent != NULL)
	dsl_fs_ss_count_adjust(dd->dd_parent, delta, prop, tx);
	}

	uint64_t
	dsl_dir_create_sync(dsl_pool_t dp, dsl_dir_t pds, const char *name,
	dmu_tx_t *tx)
	{
	objset_t *mos = dp->dp_meta_objset;
	uint64_t ddobj;
	dsl_dir_phys_t *ddphys;
	dmu_buf_t *dbuf;

	ddobj = dmu_object_alloc(mos, DMU_OT_DSL_DIR, 0,
	DMU_OT_DSL_DIR, sizeof (dsl_dir_phys_t), tx);
	if (pds) {
	VERIFY0(zap_add(mos, dsl_dir_phys(pds)->dd_child_dir_zapobj,
	name, sizeof (uint64_t), 1, &ddobj, tx));
	} else {
	/* it's the root dir */
	VERIFY0(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1, &ddobj, tx));
	}
	VERIFY0(dmu_bonus_hold(mos, ddobj, FTAG, &dbuf));
	dmu_buf_will_dirty(dbuf, tx);
	ddphys = dbuf->db_data;

	ddphys->dd_creation_time = gethrestime_sec();
	if (pds) {
	ddphys->dd_parent_obj = pds->dd_object;

	/* update the filesystem counts */
	dsl_fs_ss_count_adjust(pds, 1, DD_FIELD_FILESYSTEM_COUNT, tx);
	}
	ddphys->dd_props_zapobj = zap_create(mos,
	DMU_OT_DSL_PROPS, DMU_OT_NONE, 0, tx);
	ddphys->dd_child_dir_zapobj = zap_create(mos,
	DMU_OT_DSL_DIR_CHILD_MAP, DMU_OT_NONE, 0, tx);
	if (spa_version(dp->dp_spa) >= SPA_VERSION_USED_BREAKDOWN)
	ddphys->dd_flags \|= DD_FLAG_USED_BREAKDOWN;

	dmu_buf_rele(dbuf, FTAG);

	return (ddobj);
	}

	boolean_t
	dsl_dir_is_clone(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_origin_obj &&
	(dd->dd_pool->dp_origin_snap == NULL \|\|
	dsl_dir_phys(dd)->dd_origin_obj !=
	dd->dd_pool->dp_origin_snap->ds_object));
	}

	uint64_t
	dsl_dir_get_used(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_used_bytes);
	}

	uint64_t
	dsl_dir_get_compressed(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_compressed_bytes);
	}

	uint64_t
	dsl_dir_get_quota(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_quota);
	}

	uint64_t
	dsl_dir_get_reservation(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_reserved);
	}

	uint64_t
	dsl_dir_get_compressratio(dsl_dir_t *dd)
	{
	/* a fixed point number, 100x the ratio */
	return (dsl_dir_phys(dd)->dd_compressed_bytes == 0 ? 100 :
	(dsl_dir_phys(dd)->dd_uncompressed_bytes * 100 /
	dsl_dir_phys(dd)->dd_compressed_bytes));
	}

	uint64_t
	dsl_dir_get_logicalused(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_uncompressed_bytes);
	}

	uint64_t
	dsl_dir_get_usedsnap(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_SNAP]);
	}

	uint64_t
	dsl_dir_get_usedds(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_HEAD]);
	}

	uint64_t
	dsl_dir_get_usedrefreserv(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_REFRSRV]);
	}

	uint64_t
	dsl_dir_get_usedchild(dsl_dir_t *dd)
	{
	return (dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_CHILD] +
	dsl_dir_phys(dd)->dd_used_breakdown[DD_USED_CHILD_RSRV]);
	}

	void
	dsl_dir_get_origin(dsl_dir_t dd, char buf)
	{
	dsl_dataset_t *ds;
	VERIFY0(dsl_dataset_hold_obj(dd->dd_pool,
	dsl_dir_phys(dd)->dd_origin_obj, FTAG, &ds));

	dsl_dataset_name(ds, buf);

	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dir_get_filesystem_count(dsl_dir_t dd, uint64_t count)
	{
	if (dsl_dir_is_zapified(dd)) {
	objset_t *os = dd->dd_pool->dp_meta_objset;
	return (zap_lookup(os, dd->dd_object, DD_FIELD_FILESYSTEM_COUNT,
	sizeof (*count), 1, count));
	} else {
	return (SET_ERROR(ENOENT));
	}
	}

	int
	dsl_dir_get_snapshot_count(dsl_dir_t dd, uint64_t count)
	{
	if (dsl_dir_is_zapified(dd)) {
	objset_t *os = dd->dd_pool->dp_meta_objset;
	return (zap_lookup(os, dd->dd_object, DD_FIELD_SNAPSHOT_COUNT,
	sizeof (*count), 1, count));
	} else {
	return (SET_ERROR(ENOENT));
	}
	}

	void
	dsl_dir_stats(dsl_dir_t dd, nvlist_t nv)
	{
	mutex_enter(&dd->dd_lock);
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_QUOTA,
	dsl_dir_get_quota(dd));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_RESERVATION,
	dsl_dir_get_reservation(dd));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_LOGICALUSED,
	dsl_dir_get_logicalused(dd));
	if (dsl_dir_phys(dd)->dd_flags & DD_FLAG_USED_BREAKDOWN) {
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDSNAP,
	dsl_dir_get_usedsnap(dd));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDDS,
	dsl_dir_get_usedds(dd));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDREFRESERV,
	dsl_dir_get_usedrefreserv(dd));
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USEDCHILD,
	dsl_dir_get_usedchild(dd));
	}
	mutex_exit(&dd->dd_lock);

	uint64_t count;
	if (dsl_dir_get_filesystem_count(dd, &count) == 0) {
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_FILESYSTEM_COUNT,
	count);
	}
	if (dsl_dir_get_snapshot_count(dd, &count) == 0) {
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_SNAPSHOT_COUNT,
	count);
	}

	if (dsl_dir_is_clone(dd)) {
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dir_get_origin(dd, buf);
	dsl_prop_nvlist_add_string(nv, ZFS_PROP_ORIGIN, buf);
	}

	}

	void
	dsl_dir_dirty(dsl_dir_t dd, dmu_tx_t tx)
	{
	dsl_pool_t *dp = dd->dd_pool;

	ASSERT(dsl_dir_phys(dd));

	if (txg_list_add(&dp->dp_dirty_dirs, dd, tx->tx_txg)) {
	/* up the hold count until we can be written out */
	dmu_buf_add_ref(dd->dd_dbuf, dd);
	}
	}

	static int64_t
	parent_delta(dsl_dir_t *dd, uint64_t used, int64_t delta)
	{
	uint64_t old_accounted = MAX(used, dsl_dir_phys(dd)->dd_reserved);
	uint64_t new_accounted =
	MAX(used + delta, dsl_dir_phys(dd)->dd_reserved);
	return (new_accounted - old_accounted);
	}

	void
	dsl_dir_sync(dsl_dir_t dd, dmu_tx_t tx)
	{
	ASSERT(dmu_tx_is_syncing(tx));

	mutex_enter(&dd->dd_lock);
	ASSERT0(dd->dd_tempreserved[tx->tx_txg & TXG_MASK]);
	dprintf_dd(dd, "txg=%llu towrite=%lluK\n", (u_longlong_t)tx->tx_txg,
	(u_longlong_t)dd->dd_space_towrite[tx->tx_txg & TXG_MASK] / 1024);
	dd->dd_space_towrite[tx->tx_txg & TXG_MASK] = 0;
	mutex_exit(&dd->dd_lock);

	/* release the hold from dsl_dir_dirty */
	dmu_buf_rele(dd->dd_dbuf, dd);
	}

	static uint64_t
	dsl_dir_space_towrite(dsl_dir_t *dd)
	{
	uint64_t space = 0;

	ASSERT(MUTEX_HELD(&dd->dd_lock));

	for (int i = 0; i < TXG_SIZE; i++) {
	space += dd->dd_space_towrite[i & TXG_MASK];
	ASSERT3U(dd->dd_space_towrite[i & TXG_MASK], >=, 0);
	}
	return (space);
	}

	/*
	* How much space would dd have available if ancestor had delta applied
	* to it? If ondiskonly is set, we're only interested in what's
	* on-disk, not estimated pending changes.
	*/
	uint64_t
	dsl_dir_space_available(dsl_dir_t *dd,
	dsl_dir_t *ancestor, int64_t delta, int ondiskonly)
	{
	uint64_t parentspace, myspace, quota, used;

	/*
	* If there are no restrictions otherwise, assume we have
	* unlimited space available.
	*/
	quota = UINT64_MAX;
	parentspace = UINT64_MAX;

	if (dd->dd_parent != NULL) {
	parentspace = dsl_dir_space_available(dd->dd_parent,
	ancestor, delta, ondiskonly);
	}

	mutex_enter(&dd->dd_lock);
	if (dsl_dir_phys(dd)->dd_quota != 0)
	quota = dsl_dir_phys(dd)->dd_quota;
	used = dsl_dir_phys(dd)->dd_used_bytes;
	if (!ondiskonly)
	used += dsl_dir_space_towrite(dd);

	if (dd->dd_parent == NULL) {
	uint64_t poolsize = dsl_pool_adjustedsize(dd->dd_pool,
	ZFS_SPACE_CHECK_NORMAL);
	quota = MIN(quota, poolsize);
	}

	if (dsl_dir_phys(dd)->dd_reserved > used && parentspace != UINT64_MAX) {
	/*
	* We have some space reserved, in addition to what our
	* parent gave us.
	*/
	parentspace += dsl_dir_phys(dd)->dd_reserved - used;
	}

	if (dd == ancestor) {
	ASSERT(delta <= 0);
	ASSERT(used >= -delta);
	used += delta;
	if (parentspace != UINT64_MAX)
	parentspace -= delta;
	}

	if (used > quota) {
	/* over quota */
	myspace = 0;
	} else {
	/*
	* the lesser of the space provided by our parent and
	* the space left in our quota
	*/
	myspace = MIN(parentspace, quota - used);
	}

	mutex_exit(&dd->dd_lock);

	return (myspace);
	}

	struct tempreserve {
	list_node_t tr_node;
	dsl_dir_t *tr_ds;
	uint64_t tr_size;
	};

	static int
	dsl_dir_tempreserve_impl(dsl_dir_t *dd, uint64_t asize, boolean_t netfree,
	boolean_t ignorequota, list_t *tr_list,
	dmu_tx_t *tx, boolean_t first)
	{
	uint64_t txg;
	uint64_t quota;
	struct tempreserve *tr;
	int retval;
	uint64_t ref_rsrv;

	top_of_function:
	txg = tx->tx_txg;
	retval = EDQUOT;
	ref_rsrv = 0;

	ASSERT3U(txg, !=, 0);
	ASSERT3S(asize, >, 0);

	mutex_enter(&dd->dd_lock);

	/*
	* Check against the dsl_dir's quota. We don't add in the delta
	* when checking for over-quota because they get one free hit.
	*/
	uint64_t est_inflight = dsl_dir_space_towrite(dd);
	for (int i = 0; i < TXG_SIZE; i++)
	est_inflight += dd->dd_tempreserved[i];
	uint64_t used_on_disk = dsl_dir_phys(dd)->dd_used_bytes;

	/*
	* On the first iteration, fetch the dataset's used-on-disk and
	* refreservation values. Also, if checkrefquota is set, test if
	* allocating this space would exceed the dataset's refquota.
	*/
	if (first && tx->tx_objset) {
	int error;
	dsl_dataset_t *ds = tx->tx_objset->os_dsl_dataset;

	error = dsl_dataset_check_quota(ds, !netfree,
	asize, est_inflight, &used_on_disk, &ref_rsrv);
	if (error != 0) {
	mutex_exit(&dd->dd_lock);
	DMU_TX_STAT_BUMP(dmu_tx_quota);
	return (error);
	}
	}

	/*
	* If this transaction will result in a net free of space,
	* we want to let it through.
	*/
	if (ignorequota \|\| netfree \|\| dsl_dir_phys(dd)->dd_quota == 0)
	quota = UINT64_MAX;
	else
	quota = dsl_dir_phys(dd)->dd_quota;

	/*
	* Adjust the quota against the actual pool size at the root
	* minus any outstanding deferred frees.
	* To ensure that it's possible to remove files from a full
	* pool without inducing transient overcommits, we throttle
	* netfree transactions against a quota that is slightly larger,
	* but still within the pool's allocation slop. In cases where
	* we're very close to full, this will allow a steady trickle of
	* removes to get through.
	*/
	if (dd->dd_parent == NULL) {
	uint64_t avail = dsl_pool_unreserved_space(dd->dd_pool,
	(netfree) ?
	ZFS_SPACE_CHECK_RESERVED : ZFS_SPACE_CHECK_NORMAL);

	if (avail < quota) {
	quota = avail;
	retval = SET_ERROR(ENOSPC);
	}
	}

	/*
	* If they are requesting more space, and our current estimate
	* is over quota, they get to try again unless the actual
	* on-disk is over quota and there are no pending changes
	* or deferred frees (which may free up space for us).
	*/
	if (used_on_disk + est_inflight >= quota) {
	if (est_inflight > 0 \|\| used_on_disk < quota) {
	retval = SET_ERROR(ERESTART);
	} else {
	ASSERT3U(used_on_disk, >=, quota);

	if (retval == ENOSPC && (used_on_disk - quota) <
	dsl_pool_deferred_space(dd->dd_pool)) {
	retval = SET_ERROR(ERESTART);
	}
	}

	dprintf_dd(dd, "failing: used=%lluK inflight = %lluK "
	"quota=%lluK tr=%lluK err=%d\n",
	(u_longlong_t)used_on_disk>>10,
	(u_longlong_t)est_inflight>>10,
	(u_longlong_t)quota>>10, (u_longlong_t)asize>>10, retval);
	mutex_exit(&dd->dd_lock);
	DMU_TX_STAT_BUMP(dmu_tx_quota);
	return (retval);
	}

	/* We need to up our estimated delta before dropping dd_lock */
	dd->dd_tempreserved[txg & TXG_MASK] += asize;

	uint64_t parent_rsrv = parent_delta(dd, used_on_disk + est_inflight,
	asize - ref_rsrv);
	mutex_exit(&dd->dd_lock);

	tr = kmem_zalloc(sizeof (struct tempreserve), KM_SLEEP);
	tr->tr_ds = dd;
	tr->tr_size = asize;
	list_insert_tail(tr_list, tr);

	/* see if it's OK with our parent */
	if (dd->dd_parent != NULL && parent_rsrv != 0) {
	/*
	* Recurse on our parent without recursion. This has been
	* observed to be potentially large stack usage even within
	* the test suite. Largest seen stack was 7632 bytes on linux.
	*/

	dd = dd->dd_parent;
	asize = parent_rsrv;
	ignorequota = (dsl_dir_phys(dd)->dd_head_dataset_obj == 0);
	first = B_FALSE;
	goto top_of_function;

	} else {
	return (0);
	}
	}

	/*
	* Reserve space in this dsl_dir, to be used in this tx's txg.
	* After the space has been dirtied (and dsl_dir_willuse_space()
	* has been called), the reservation should be canceled, using
	* dsl_dir_tempreserve_clear().
	*/
	int
	dsl_dir_tempreserve_space(dsl_dir_t *dd, uint64_t lsize, uint64_t asize,
	boolean_t netfree, void *tr_cookiep, dmu_tx_t tx)
	{
	int err;
	list_t *tr_list;

	if (asize == 0) {
	*tr_cookiep = NULL;
	return (0);
	}

	tr_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
	list_create(tr_list, sizeof (struct tempreserve),
	offsetof(struct tempreserve, tr_node));
	ASSERT3S(asize, >, 0);

	err = arc_tempreserve_space(dd->dd_pool->dp_spa, lsize, tx->tx_txg);
	if (err == 0) {
	struct tempreserve *tr;

	tr = kmem_zalloc(sizeof (struct tempreserve), KM_SLEEP);
	tr->tr_size = lsize;
	list_insert_tail(tr_list, tr);
	} else {
	if (err == EAGAIN) {
	/*
	* If arc_memory_throttle() detected that pageout
	* is running and we are low on memory, we delay new
	* non-pageout transactions to give pageout an
	* advantage.
	*
	* It is unfortunate to be delaying while the caller's
	* locks are held.
	*/
	txg_delay(dd->dd_pool, tx->tx_txg,
	MSEC2NSEC(10), MSEC2NSEC(10));
	err = SET_ERROR(ERESTART);
	}
	}

	if (err == 0) {
	err = dsl_dir_tempreserve_impl(dd, asize, netfree,
	B_FALSE, tr_list, tx, B_TRUE);
	}

	if (err != 0)
	dsl_dir_tempreserve_clear(tr_list, tx);
	else
	*tr_cookiep = tr_list;

	return (err);
	}

	/*
	* Clear a temporary reservation that we previously made with
	* dsl_dir_tempreserve_space().
	*/
	void
	dsl_dir_tempreserve_clear(void tr_cookie, dmu_tx_t tx)
	{
	int txgidx = tx->tx_txg & TXG_MASK;
	list_t *tr_list = tr_cookie;
	struct tempreserve *tr;

	ASSERT3U(tx->tx_txg, !=, 0);

	if (tr_cookie == NULL)
	return;

	while ((tr = list_head(tr_list)) != NULL) {
	if (tr->tr_ds) {
	mutex_enter(&tr->tr_ds->dd_lock);
	ASSERT3U(tr->tr_ds->dd_tempreserved[txgidx], >=,
	tr->tr_size);
	tr->tr_ds->dd_tempreserved[txgidx] -= tr->tr_size;
	mutex_exit(&tr->tr_ds->dd_lock);
	} else {
	arc_tempreserve_clear(tr->tr_size);
	}
	list_remove(tr_list, tr);
	kmem_free(tr, sizeof (struct tempreserve));
	}

	kmem_free(tr_list, sizeof (list_t));
	}

	/*
	* This should be called from open context when we think we're going to write
	* or free space, for example when dirtying data. Be conservative; it's okay
	* to write less space or free more, but we don't want to write more or free
	* less than the amount specified.
	*
	* NOTE: The behavior of this function is identical to the Illumos / FreeBSD
	* version however it has been adjusted to use an iterative rather than
	* recursive algorithm to minimize stack usage.
	*/
	void
	dsl_dir_willuse_space(dsl_dir_t dd, int64_t space, dmu_tx_t tx)
	{
	int64_t parent_space;
	uint64_t est_used;

	do {
	mutex_enter(&dd->dd_lock);
	if (space > 0)
	dd->dd_space_towrite[tx->tx_txg & TXG_MASK] += space;

	est_used = dsl_dir_space_towrite(dd) +
	dsl_dir_phys(dd)->dd_used_bytes;
	parent_space = parent_delta(dd, est_used, space);
	mutex_exit(&dd->dd_lock);

	/* Make sure that we clean up dd_space_to* */
	dsl_dir_dirty(dd, tx);

	dd = dd->dd_parent;
	space = parent_space;
	} while (space && dd);
	}

	/* call from syncing context when we actually write/free space for this dd */
	void
	dsl_dir_diduse_space(dsl_dir_t *dd, dd_used_t type,
	int64_t used, int64_t compressed, int64_t uncompressed, dmu_tx_t *tx)
	{
	int64_t accounted_delta;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(type < DD_USED_NUM);

	dmu_buf_will_dirty(dd->dd_dbuf, tx);

	/*
	* dsl_dataset_set_refreservation_sync_impl() calls this with
	* dd_lock held, so that it can atomically update
	* ds->ds_reserved and the dsl_dir accounting, so that
	* dsl_dataset_check_quota() can see dataset and dir accounting
	* consistently.
	*/
	boolean_t needlock = !MUTEX_HELD(&dd->dd_lock);
	if (needlock)
	mutex_enter(&dd->dd_lock);
	dsl_dir_phys_t *ddp = dsl_dir_phys(dd);
	accounted_delta = parent_delta(dd, ddp->dd_used_bytes, used);
	ASSERT(used >= 0 \|\| ddp->dd_used_bytes >= -used);
	ASSERT(compressed >= 0 \|\| ddp->dd_compressed_bytes >= -compressed);
	ASSERT(uncompressed >= 0 \|\|
	ddp->dd_uncompressed_bytes >= -uncompressed);
	ddp->dd_used_bytes += used;
	ddp->dd_uncompressed_bytes += uncompressed;
	ddp->dd_compressed_bytes += compressed;

	if (ddp->dd_flags & DD_FLAG_USED_BREAKDOWN) {
	ASSERT(used >= 0 \|\| ddp->dd_used_breakdown[type] >= -used);
	ddp->dd_used_breakdown[type] += used;
	#ifdef ZFS_DEBUG
	{
	dd_used_t t;
	uint64_t u = 0;
	for (t = 0; t < DD_USED_NUM; t++)
	u += ddp->dd_used_breakdown[t];
	ASSERT3U(u, ==, ddp->dd_used_bytes);
	}
	#endif
	}
	if (needlock)
	mutex_exit(&dd->dd_lock);

	if (dd->dd_parent != NULL) {
	dsl_dir_diduse_transfer_space(dd->dd_parent,
	accounted_delta, compressed, uncompressed,
	used, DD_USED_CHILD_RSRV, DD_USED_CHILD, tx);
	}
	}

	void
	dsl_dir_transfer_space(dsl_dir_t *dd, int64_t delta,
	dd_used_t oldtype, dd_used_t newtype, dmu_tx_t *tx)
	{
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(oldtype < DD_USED_NUM);
	ASSERT(newtype < DD_USED_NUM);

	dsl_dir_phys_t *ddp = dsl_dir_phys(dd);
	if (delta == 0 \|\|
	!(ddp->dd_flags & DD_FLAG_USED_BREAKDOWN))
	return;

	dmu_buf_will_dirty(dd->dd_dbuf, tx);
	mutex_enter(&dd->dd_lock);
	ASSERT(delta > 0 ?
	ddp->dd_used_breakdown[oldtype] >= delta :
	ddp->dd_used_breakdown[newtype] >= -delta);
	ASSERT(ddp->dd_used_bytes >= ABS(delta));
	ddp->dd_used_breakdown[oldtype] -= delta;
	ddp->dd_used_breakdown[newtype] += delta;
	mutex_exit(&dd->dd_lock);
	}

	void
	dsl_dir_diduse_transfer_space(dsl_dir_t *dd, int64_t used,
	int64_t compressed, int64_t uncompressed, int64_t tonew,
	dd_used_t oldtype, dd_used_t newtype, dmu_tx_t *tx)
	{
	int64_t accounted_delta;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(oldtype < DD_USED_NUM);
	ASSERT(newtype < DD_USED_NUM);

	dmu_buf_will_dirty(dd->dd_dbuf, tx);

	mutex_enter(&dd->dd_lock);
	dsl_dir_phys_t *ddp = dsl_dir_phys(dd);
	accounted_delta = parent_delta(dd, ddp->dd_used_bytes, used);
	ASSERT(used >= 0 \|\| ddp->dd_used_bytes >= -used);
	ASSERT(compressed >= 0 \|\| ddp->dd_compressed_bytes >= -compressed);
	ASSERT(uncompressed >= 0 \|\|
	ddp->dd_uncompressed_bytes >= -uncompressed);
	ddp->dd_used_bytes += used;
	ddp->dd_uncompressed_bytes += uncompressed;
	ddp->dd_compressed_bytes += compressed;

	if (ddp->dd_flags & DD_FLAG_USED_BREAKDOWN) {
	ASSERT(tonew - used <= 0 \|\|
	ddp->dd_used_breakdown[oldtype] >= tonew - used);
	ASSERT(tonew >= 0 \|\|
	ddp->dd_used_breakdown[newtype] >= -tonew);
	ddp->dd_used_breakdown[oldtype] -= tonew - used;
	ddp->dd_used_breakdown[newtype] += tonew;
	#ifdef ZFS_DEBUG
	{
	dd_used_t t;
	uint64_t u = 0;
	for (t = 0; t < DD_USED_NUM; t++)
	u += ddp->dd_used_breakdown[t];
	ASSERT3U(u, ==, ddp->dd_used_bytes);
	}
	#endif
	}
	mutex_exit(&dd->dd_lock);

	if (dd->dd_parent != NULL) {
	dsl_dir_diduse_transfer_space(dd->dd_parent,
	accounted_delta, compressed, uncompressed,
	used, DD_USED_CHILD_RSRV, DD_USED_CHILD, tx);
	}
	}

	typedef struct dsl_dir_set_qr_arg {
	const char *ddsqra_name;
	zprop_source_t ddsqra_source;
	uint64_t ddsqra_value;
	} dsl_dir_set_qr_arg_t;

	static int
	dsl_dir_set_quota_check(void arg, dmu_tx_t tx)
	{
	dsl_dir_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	int error;
	uint64_t towrite, newval;

	error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
	if (error != 0)
	return (error);

	error = dsl_prop_predict(ds->ds_dir, "quota",
	ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (newval == 0) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	mutex_enter(&ds->ds_dir->dd_lock);
	/*
	* If we are doing the preliminary check in open context, and
	* there are pending changes, then don't fail it, since the
	* pending changes could under-estimate the amount of space to be
	* freed up.
	*/
	towrite = dsl_dir_space_towrite(ds->ds_dir);
	if ((dmu_tx_is_syncing(tx) \|\| towrite == 0) &&
	(newval < dsl_dir_phys(ds->ds_dir)->dd_reserved \|\|
	newval < dsl_dir_phys(ds->ds_dir)->dd_used_bytes + towrite)) {
	error = SET_ERROR(ENOSPC);
	}
	mutex_exit(&ds->ds_dir->dd_lock);
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	static void
	dsl_dir_set_quota_sync(void arg, dmu_tx_t tx)
	{
	dsl_dir_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	uint64_t newval;

	VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));

	if (spa_version(dp->dp_spa) >= SPA_VERSION_RECVD_PROPS) {
	dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_QUOTA),
	ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1,
	&ddsqra->ddsqra_value, tx);

	VERIFY0(dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_QUOTA), &newval));
	} else {
	newval = ddsqra->ddsqra_value;
	spa_history_log_internal_ds(ds, "set", tx, "%s=%lld",
	zfs_prop_to_name(ZFS_PROP_QUOTA), (longlong_t)newval);
	}

	dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
	mutex_enter(&ds->ds_dir->dd_lock);
	dsl_dir_phys(ds->ds_dir)->dd_quota = newval;
	mutex_exit(&ds->ds_dir->dd_lock);
	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dir_set_quota(const char *ddname, zprop_source_t source, uint64_t quota)
	{
	dsl_dir_set_qr_arg_t ddsqra;

	ddsqra.ddsqra_name = ddname;
	ddsqra.ddsqra_source = source;
	ddsqra.ddsqra_value = quota;

	return (dsl_sync_task(ddname, dsl_dir_set_quota_check,
	dsl_dir_set_quota_sync, &ddsqra, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	static int
	dsl_dir_set_reservation_check(void arg, dmu_tx_t tx)
	{
	dsl_dir_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	dsl_dir_t *dd;
	uint64_t newval, used, avail;
	int error;

	error = dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds);
	if (error != 0)
	return (error);
	dd = ds->ds_dir;

	/*
	* If we are doing the preliminary check in open context, the
	* space estimates may be inaccurate.
	*/
	if (!dmu_tx_is_syncing(tx)) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	error = dsl_prop_predict(ds->ds_dir,
	zfs_prop_to_name(ZFS_PROP_RESERVATION),
	ddsqra->ddsqra_source, ddsqra->ddsqra_value, &newval);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	mutex_enter(&dd->dd_lock);
	used = dsl_dir_phys(dd)->dd_used_bytes;
	mutex_exit(&dd->dd_lock);

	if (dd->dd_parent) {
	avail = dsl_dir_space_available(dd->dd_parent,
	NULL, 0, FALSE);
	} else {
	avail = dsl_pool_adjustedsize(dd->dd_pool,
	ZFS_SPACE_CHECK_NORMAL) - used;
	}

	if (MAX(used, newval) > MAX(used, dsl_dir_phys(dd)->dd_reserved)) {
	uint64_t delta = MAX(used, newval) -
	MAX(used, dsl_dir_phys(dd)->dd_reserved);

	if (delta > avail \|\|
	(dsl_dir_phys(dd)->dd_quota > 0 &&
	newval > dsl_dir_phys(dd)->dd_quota))
	error = SET_ERROR(ENOSPC);
	}

	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	void
	dsl_dir_set_reservation_sync_impl(dsl_dir_t dd, uint64_t value, dmu_tx_t tx)
	{
	uint64_t used;
	int64_t delta;

	dmu_buf_will_dirty(dd->dd_dbuf, tx);

	mutex_enter(&dd->dd_lock);
	used = dsl_dir_phys(dd)->dd_used_bytes;
	delta = MAX(used, value) - MAX(used, dsl_dir_phys(dd)->dd_reserved);
	dsl_dir_phys(dd)->dd_reserved = value;

	if (dd->dd_parent != NULL) {
	/* Roll up this additional usage into our ancestors */
	dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD_RSRV,
	delta, 0, 0, tx);
	}
	mutex_exit(&dd->dd_lock);
	}

	static void
	dsl_dir_set_reservation_sync(void arg, dmu_tx_t tx)
	{
	dsl_dir_set_qr_arg_t *ddsqra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	uint64_t newval;

	VERIFY0(dsl_dataset_hold(dp, ddsqra->ddsqra_name, FTAG, &ds));

	if (spa_version(dp->dp_spa) >= SPA_VERSION_RECVD_PROPS) {
	dsl_prop_set_sync_impl(ds,
	zfs_prop_to_name(ZFS_PROP_RESERVATION),
	ddsqra->ddsqra_source, sizeof (ddsqra->ddsqra_value), 1,
	&ddsqra->ddsqra_value, tx);

	VERIFY0(dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_RESERVATION), &newval));
	} else {
	newval = ddsqra->ddsqra_value;
	spa_history_log_internal_ds(ds, "set", tx, "%s=%lld",
	zfs_prop_to_name(ZFS_PROP_RESERVATION),
	(longlong_t)newval);
	}

	dsl_dir_set_reservation_sync_impl(ds->ds_dir, newval, tx);
	dsl_dataset_rele(ds, FTAG);
	}

	int
	dsl_dir_set_reservation(const char *ddname, zprop_source_t source,
	uint64_t reservation)
	{
	dsl_dir_set_qr_arg_t ddsqra;

	ddsqra.ddsqra_name = ddname;
	ddsqra.ddsqra_source = source;
	ddsqra.ddsqra_value = reservation;

	return (dsl_sync_task(ddname, dsl_dir_set_reservation_check,
	dsl_dir_set_reservation_sync, &ddsqra, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	static dsl_dir_t *
	closest_common_ancestor(dsl_dir_t ds1, dsl_dir_t ds2)
	{
	for (; ds1; ds1 = ds1->dd_parent) {
	dsl_dir_t *dd;
	for (dd = ds2; dd; dd = dd->dd_parent) {
	if (ds1 == dd)
	return (dd);
	}
	}
	return (NULL);
	}

	/*
	* If delta is applied to dd, how much of that delta would be applied to
	* ancestor? Syncing context only.
	*/
	static int64_t
	would_change(dsl_dir_t dd, int64_t delta, dsl_dir_t ancestor)
	{
	if (dd == ancestor)
	return (delta);

	mutex_enter(&dd->dd_lock);
	delta = parent_delta(dd, dsl_dir_phys(dd)->dd_used_bytes, delta);
	mutex_exit(&dd->dd_lock);
	return (would_change(dd->dd_parent, delta, ancestor));
	}

	typedef struct dsl_dir_rename_arg {
	const char *ddra_oldname;
	const char *ddra_newname;
	cred_t *ddra_cred;
	proc_t *ddra_proc;
	} dsl_dir_rename_arg_t;

	typedef struct dsl_valid_rename_arg {
	int char_delta;
	int nest_delta;
	} dsl_valid_rename_arg_t;

	static int
	dsl_valid_rename(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) dp;
	dsl_valid_rename_arg_t *dvra = arg;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];

	dsl_dataset_name(ds, namebuf);

	ASSERT3U(strnlen(namebuf, ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	int namelen = strlen(namebuf) + dvra->char_delta;
	int depth = get_dataset_depth(namebuf) + dvra->nest_delta;

	if (namelen >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));
	if (dvra->nest_delta > 0 && depth >= zfs_max_dataset_nesting)
	return (SET_ERROR(ENAMETOOLONG));
	return (0);
	}

	static int
	dsl_dir_rename_check(void arg, dmu_tx_t tx)
	{
	dsl_dir_rename_arg_t *ddra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t dd, newparent;
	dsl_valid_rename_arg_t dvra;
	dsl_dataset_t *parentds;
	objset_t *parentos;
	const char *mynewname;
	int error;

	/* target dir should exist */
	error = dsl_dir_hold(dp, ddra->ddra_oldname, FTAG, &dd, NULL);
	if (error != 0)
	return (error);

	/* new parent should exist */
	error = dsl_dir_hold(dp, ddra->ddra_newname, FTAG,
	&newparent, &mynewname);
	if (error != 0) {
	dsl_dir_rele(dd, FTAG);
	return (error);
	}

	/* can't rename to different pool */
	if (dd->dd_pool != newparent->dd_pool) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(EXDEV));
	}

	/* new name should not already exist */
	if (mynewname == NULL) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(EEXIST));
	}

	/* can't rename below anything but filesystems (eg. no ZVOLs) */
	error = dsl_dataset_hold_obj(newparent->dd_pool,
	dsl_dir_phys(newparent)->dd_head_dataset_obj, FTAG, &parentds);
	if (error != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (error);
	}
	error = dmu_objset_from_ds(parentds, &parentos);
	if (error != 0) {
	dsl_dataset_rele(parentds, FTAG);
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (error);
	}
	if (dmu_objset_type(parentos) != DMU_OST_ZFS) {
	dsl_dataset_rele(parentds, FTAG);
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
	}
	dsl_dataset_rele(parentds, FTAG);

	ASSERT3U(strnlen(ddra->ddra_newname, ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	ASSERT3U(strnlen(ddra->ddra_oldname, ZFS_MAX_DATASET_NAME_LEN),
	<, ZFS_MAX_DATASET_NAME_LEN);
	dvra.char_delta = strlen(ddra->ddra_newname)
	- strlen(ddra->ddra_oldname);
	dvra.nest_delta = get_dataset_depth(ddra->ddra_newname)
	- get_dataset_depth(ddra->ddra_oldname);

	/* if the name length is growing, validate child name lengths */
	if (dvra.char_delta > 0 \|\| dvra.nest_delta > 0) {
	error = dmu_objset_find_dp(dp, dd->dd_object, dsl_valid_rename,
	&dvra, DS_FIND_CHILDREN \| DS_FIND_SNAPSHOTS);
	if (error != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (error);
	}
	}

	if (dmu_tx_is_syncing(tx)) {
	if (spa_feature_is_active(dp->dp_spa,
	SPA_FEATURE_FS_SS_LIMIT)) {
	/*
	* Although this is the check function and we don't
	* normally make on-disk changes in check functions,
	* we need to do that here.
	*
	* Ensure this portion of the tree's counts have been
	* initialized in case the new parent has limits set.
	*/
	dsl_dir_init_fs_ss_count(dd, tx);
	}
	}

	if (newparent != dd->dd_parent) {
	/* is there enough space? */
	uint64_t myspace =
	MAX(dsl_dir_phys(dd)->dd_used_bytes,
	dsl_dir_phys(dd)->dd_reserved);
	objset_t *os = dd->dd_pool->dp_meta_objset;
	uint64_t fs_cnt = 0;
	uint64_t ss_cnt = 0;

	if (dsl_dir_is_zapified(dd)) {
	int err;

	err = zap_lookup(os, dd->dd_object,
	DD_FIELD_FILESYSTEM_COUNT, sizeof (fs_cnt), 1,
	&fs_cnt);
	if (err != ENOENT && err != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (err);
	}

	/*
	* have to add 1 for the filesystem itself that we're
	* moving
	*/
	fs_cnt++;

	err = zap_lookup(os, dd->dd_object,
	DD_FIELD_SNAPSHOT_COUNT, sizeof (ss_cnt), 1,
	&ss_cnt);
	if (err != ENOENT && err != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (err);
	}
	}

	/* check for encryption errors */
	error = dsl_dir_rename_crypt_check(dd, newparent);
	if (error != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(EACCES));
	}

	/* no rename into our descendant */
	if (closest_common_ancestor(dd, newparent) == dd) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = dsl_dir_transfer_possible(dd->dd_parent,
	newparent, fs_cnt, ss_cnt, myspace,
	ddra->ddra_cred, ddra->ddra_proc);
	if (error != 0) {
	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (error);
	}
	}

	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	return (0);
	}

	static void
	dsl_dir_rename_sync(void arg, dmu_tx_t tx)
	{
	dsl_dir_rename_arg_t *ddra = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t dd, newparent;
	const char *mynewname;
	objset_t *mos = dp->dp_meta_objset;

	VERIFY0(dsl_dir_hold(dp, ddra->ddra_oldname, FTAG, &dd, NULL));
	VERIFY0(dsl_dir_hold(dp, ddra->ddra_newname, FTAG, &newparent,
	&mynewname));

	/* Log this before we change the name. */
	spa_history_log_internal_dd(dd, "rename", tx,
	"-> %s", ddra->ddra_newname);

	if (newparent != dd->dd_parent) {
	objset_t *os = dd->dd_pool->dp_meta_objset;
	uint64_t fs_cnt = 0;
	uint64_t ss_cnt = 0;

	/*
	* We already made sure the dd counts were initialized in the
	* check function.
	*/
	if (spa_feature_is_active(dp->dp_spa,
	SPA_FEATURE_FS_SS_LIMIT)) {
	VERIFY0(zap_lookup(os, dd->dd_object,
	DD_FIELD_FILESYSTEM_COUNT, sizeof (fs_cnt), 1,
	&fs_cnt));
	/* add 1 for the filesystem itself that we're moving */
	fs_cnt++;

	VERIFY0(zap_lookup(os, dd->dd_object,
	DD_FIELD_SNAPSHOT_COUNT, sizeof (ss_cnt), 1,
	&ss_cnt));
	}

	dsl_fs_ss_count_adjust(dd->dd_parent, -fs_cnt,
	DD_FIELD_FILESYSTEM_COUNT, tx);
	dsl_fs_ss_count_adjust(newparent, fs_cnt,
	DD_FIELD_FILESYSTEM_COUNT, tx);

	dsl_fs_ss_count_adjust(dd->dd_parent, -ss_cnt,
	DD_FIELD_SNAPSHOT_COUNT, tx);
	dsl_fs_ss_count_adjust(newparent, ss_cnt,
	DD_FIELD_SNAPSHOT_COUNT, tx);

	dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD,
	-dsl_dir_phys(dd)->dd_used_bytes,
	-dsl_dir_phys(dd)->dd_compressed_bytes,
	-dsl_dir_phys(dd)->dd_uncompressed_bytes, tx);
	dsl_dir_diduse_space(newparent, DD_USED_CHILD,
	dsl_dir_phys(dd)->dd_used_bytes,
	dsl_dir_phys(dd)->dd_compressed_bytes,
	dsl_dir_phys(dd)->dd_uncompressed_bytes, tx);

	if (dsl_dir_phys(dd)->dd_reserved >
	dsl_dir_phys(dd)->dd_used_bytes) {
	uint64_t unused_rsrv = dsl_dir_phys(dd)->dd_reserved -
	dsl_dir_phys(dd)->dd_used_bytes;

	dsl_dir_diduse_space(dd->dd_parent, DD_USED_CHILD_RSRV,
	-unused_rsrv, 0, 0, tx);
	dsl_dir_diduse_space(newparent, DD_USED_CHILD_RSRV,
	unused_rsrv, 0, 0, tx);
	}
	}

	dmu_buf_will_dirty(dd->dd_dbuf, tx);

	/* remove from old parent zapobj */
	VERIFY0(zap_remove(mos,
	dsl_dir_phys(dd->dd_parent)->dd_child_dir_zapobj,
	dd->dd_myname, tx));

	(void) strlcpy(dd->dd_myname, mynewname,
	sizeof (dd->dd_myname));
	dsl_dir_rele(dd->dd_parent, dd);
	dsl_dir_phys(dd)->dd_parent_obj = newparent->dd_object;
	VERIFY0(dsl_dir_hold_obj(dp,
	newparent->dd_object, NULL, dd, &dd->dd_parent));

	/* add to new parent zapobj */
	VERIFY0(zap_add(mos, dsl_dir_phys(newparent)->dd_child_dir_zapobj,
	dd->dd_myname, 8, 1, &dd->dd_object, tx));

	/* TODO: A rename callback to avoid these layering violations. */
	zfsvfs_update_fromname(ddra->ddra_oldname, ddra->ddra_newname);
	zvol_rename_minors(dp->dp_spa, ddra->ddra_oldname,
	ddra->ddra_newname, B_TRUE);

	dsl_prop_notify_all(dd);

	dsl_dir_rele(newparent, FTAG);
	dsl_dir_rele(dd, FTAG);
	}

	int
	dsl_dir_rename(const char oldname, const char newname)
	{
	dsl_dir_rename_arg_t ddra;

	ddra.ddra_oldname = oldname;
	ddra.ddra_newname = newname;
	ddra.ddra_cred = CRED();
	ddra.ddra_proc = curproc;

	return (dsl_sync_task(oldname,
	dsl_dir_rename_check, dsl_dir_rename_sync, &ddra,
	3, ZFS_SPACE_CHECK_RESERVED));
	}

	int
	dsl_dir_transfer_possible(dsl_dir_t sdd, dsl_dir_t tdd,
	uint64_t fs_cnt, uint64_t ss_cnt, uint64_t space,
	cred_t cr, proc_t proc)
	{
	dsl_dir_t *ancestor;
	int64_t adelta;
	uint64_t avail;
	int err;

	ancestor = closest_common_ancestor(sdd, tdd);
	adelta = would_change(sdd, -space, ancestor);
	avail = dsl_dir_space_available(tdd, ancestor, adelta, FALSE);
	if (avail < space)
	return (SET_ERROR(ENOSPC));

	err = dsl_fs_ss_limit_check(tdd, fs_cnt, ZFS_PROP_FILESYSTEM_LIMIT,
	ancestor, cr, proc);
	if (err != 0)
	return (err);
	err = dsl_fs_ss_limit_check(tdd, ss_cnt, ZFS_PROP_SNAPSHOT_LIMIT,
	ancestor, cr, proc);
	if (err != 0)
	return (err);

	return (0);
	}

	inode_timespec_t
	dsl_dir_snap_cmtime(dsl_dir_t *dd)
	{
	inode_timespec_t t;

	mutex_enter(&dd->dd_lock);
	t = dd->dd_snap_cmtime;
	mutex_exit(&dd->dd_lock);

	return (t);
	}

	void
	dsl_dir_snap_cmtime_update(dsl_dir_t dd, dmu_tx_t tx)
	{
	dsl_pool_t *dp = dmu_tx_pool(tx);
	inode_timespec_t t;
	gethrestime(&t);

	mutex_enter(&dd->dd_lock);
	dd->dd_snap_cmtime = t;
	if (spa_feature_is_enabled(dp->dp_spa,
	SPA_FEATURE_EXTENSIBLE_DATASET)) {
	objset_t *mos = dd->dd_pool->dp_meta_objset;
	uint64_t ddobj = dd->dd_object;
	dsl_dir_zapify(dd, tx);
	VERIFY0(zap_update(mos, ddobj,
	DD_FIELD_SNAPSHOTS_CHANGED,
	sizeof (uint64_t),
	sizeof (inode_timespec_t) / sizeof (uint64_t),
	&t, tx));
	}
	mutex_exit(&dd->dd_lock);
	}

	void
	dsl_dir_zapify(dsl_dir_t dd, dmu_tx_t tx)
	{
	objset_t *mos = dd->dd_pool->dp_meta_objset;
	dmu_object_zapify(mos, dd->dd_object, DMU_OT_DSL_DIR, tx);
	}

	boolean_t
	dsl_dir_is_zapified(dsl_dir_t *dd)
	{
	dmu_object_info_t doi;

	dmu_object_info_from_db(dd->dd_dbuf, &doi);
	return (doi.doi_type == DMU_OTN_ZAP_METADATA);
	}

	void
	dsl_dir_livelist_open(dsl_dir_t *dd, uint64_t obj)
	{
	objset_t *mos = dd->dd_pool->dp_meta_objset;
	ASSERT(spa_feature_is_active(dd->dd_pool->dp_spa,
	SPA_FEATURE_LIVELIST));
	dsl_deadlist_open(&dd->dd_livelist, mos, obj);
	bplist_create(&dd->dd_pending_allocs);
	bplist_create(&dd->dd_pending_frees);
	}

	void
	dsl_dir_livelist_close(dsl_dir_t *dd)
	{
	dsl_deadlist_close(&dd->dd_livelist);
	bplist_destroy(&dd->dd_pending_allocs);
	bplist_destroy(&dd->dd_pending_frees);
	}

	void
	dsl_dir_remove_livelist(dsl_dir_t dd, dmu_tx_t tx, boolean_t total)
	{
	uint64_t obj;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	spa_t *spa = dp->dp_spa;
	livelist_condense_entry_t to_condense = spa->spa_to_condense;

	if (!dsl_deadlist_is_open(&dd->dd_livelist))
	return;

	/*
	* If the livelist being removed is set to be condensed, stop the
	* condense zthr and indicate the cancellation in the spa_to_condense
	* struct in case the condense no-wait synctask has already started
	*/
	zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
	if (ll_condense_thread != NULL &&
	(to_condense.ds != NULL) && (to_condense.ds->ds_dir == dd)) {
	/*
	* We use zthr_wait_cycle_done instead of zthr_cancel
	* because we don't want to destroy the zthr, just have
	* it skip its current task.
	*/
	spa->spa_to_condense.cancelled = B_TRUE;
	zthr_wait_cycle_done(ll_condense_thread);
	/*
	* If we've returned from zthr_wait_cycle_done without
	* clearing the to_condense data structure it's either
	* because the no-wait synctask has started (which is
	* indicated by 'syncing' field of to_condense) and we
	* can expect it to clear to_condense on its own.
	* Otherwise, we returned before the zthr ran. The
	* checkfunc will now fail as cancelled == B_TRUE so we
	* can safely NULL out ds, allowing a different dir's
	* livelist to be condensed.
	*
	* We can be sure that the to_condense struct will not
	* be repopulated at this stage because both this
	* function and dsl_livelist_try_condense execute in
	* syncing context.
	*/
	if ((spa->spa_to_condense.ds != NULL) &&
	!spa->spa_to_condense.syncing) {
	dmu_buf_rele(spa->spa_to_condense.ds->ds_dbuf,
	spa);
	spa->spa_to_condense.ds = NULL;
	}
	}

	dsl_dir_livelist_close(dd);
	VERIFY0(zap_lookup(dp->dp_meta_objset, dd->dd_object,
	DD_FIELD_LIVELIST, sizeof (uint64_t), 1, &obj));
	VERIFY0(zap_remove(dp->dp_meta_objset, dd->dd_object,
	DD_FIELD_LIVELIST, tx));
	if (total) {
	dsl_deadlist_free(dp->dp_meta_objset, obj, tx);
	spa_feature_decr(spa, SPA_FEATURE_LIVELIST, tx);
	}
	}

	static int
	dsl_dir_activity_in_progress(dsl_dir_t dd, dsl_dataset_t ds,
	zfs_wait_activity_t activity, boolean_t *in_progress)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&dd->dd_activity_lock));

	switch (activity) {
	case ZFS_WAIT_DELETEQ: {
	#ifdef _KERNEL
	objset_t *os;
	error = dmu_objset_from_ds(ds, &os);
	if (error != 0)
	break;

	mutex_enter(&os->os_user_ptr_lock);
	void *user = dmu_objset_get_user(os);
	mutex_exit(&os->os_user_ptr_lock);
	if (dmu_objset_type(os) != DMU_OST_ZFS \|\|
	user == NULL \|\| zfs_get_vfs_flag_unmounted(os)) {
	*in_progress = B_FALSE;
	return (0);
	}

	uint64_t readonly = B_FALSE;
	error = zfs_get_temporary_prop(ds, ZFS_PROP_READONLY, &readonly,
	NULL);

	if (error != 0)
	break;

	if (readonly \|\| !spa_writeable(dd->dd_pool->dp_spa)) {
	*in_progress = B_FALSE;
	return (0);
	}

	uint64_t count, unlinked_obj;
	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
	&unlinked_obj);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	break;
	}
	error = zap_count(os, unlinked_obj, &count);

	if (error == 0)
	*in_progress = (count != 0);
	break;
	#else
	/*
	* The delete queue is ZPL specific, and libzpool doesn't have
	* it. It doesn't make sense to wait for it.
	*/
	(void) ds;
	*in_progress = B_FALSE;
	break;
	#endif
	}
	default:
	panic("unrecognized value for activity %d", activity);
	}

	return (error);
	}

	int
	dsl_dir_wait(dsl_dir_t dd, dsl_dataset_t ds, zfs_wait_activity_t activity,
	boolean_t *waited)
	{
	int error = 0;
	boolean_t in_progress;
	dsl_pool_t *dp = dd->dd_pool;
	for (;;) {
	dsl_pool_config_enter(dp, FTAG);
	error = dsl_dir_activity_in_progress(dd, ds, activity,
	&in_progress);
	dsl_pool_config_exit(dp, FTAG);
	if (error != 0 \|\| !in_progress)
	break;

	*waited = B_TRUE;

	if (cv_wait_sig(&dd->dd_activity_cv, &dd->dd_activity_lock) ==
	0 \|\| dd->dd_activity_cancelled) {
	error = SET_ERROR(EINTR);
	break;
	}
	}
	return (error);
	}

	void
	dsl_dir_cancel_waiters(dsl_dir_t *dd)
	{
	mutex_enter(&dd->dd_activity_lock);
	dd->dd_activity_cancelled = B_TRUE;
	cv_broadcast(&dd->dd_activity_cv);
	while (dd->dd_activity_waiters > 0)
	cv_wait(&dd->dd_activity_cv, &dd->dd_activity_lock);
	mutex_exit(&dd->dd_activity_lock);
	}

	#if defined(_KERNEL)
	EXPORT_SYMBOL(dsl_dir_set_quota);
	EXPORT_SYMBOL(dsl_dir_set_reservation);
	#endif
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_pool.c b/sys/contrib/openzfs/module/zfs/dsl_pool.c
	index 7a589cadbc32..4fd3722a051e 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_pool.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_pool.c
	@@ -1,1494 +1,1494 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
	*/

	#include <sys/dsl_pool.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_scan.h>
	#include <sys/dnode.h>
	#include <sys/dmu_tx.h>
	#include <sys/dmu_objset.h>
	#include <sys/arc.h>
	#include <sys/zap.h>
	#include <sys/zio.h>
	#include <sys/zfs_context.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_znode.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/bptree.h>
	#include <sys/zfeature.h>
	#include <sys/zil_impl.h>
	#include <sys/dsl_userhold.h>
	#include <sys/trace_zfs.h>
	#include <sys/mmp.h>

	/*
	* ZFS Write Throttle
	* ------------------
	*
	* ZFS must limit the rate of incoming writes to the rate at which it is able
	* to sync data modifications to the backend storage. Throttling by too much
	* creates an artificial limit; throttling by too little can only be sustained
	* for short periods and would lead to highly lumpy performance. On a per-pool
	* basis, ZFS tracks the amount of modified (dirty) data. As operations change
	* data, the amount of dirty data increases; as ZFS syncs out data, the amount
	* of dirty data decreases. When the amount of dirty data exceeds a
	* predetermined threshold further modifications are blocked until the amount
	* of dirty data decreases (as data is synced out).
	*
	* The limit on dirty data is tunable, and should be adjusted according to
	* both the IO capacity and available memory of the system. The larger the
	* window, the more ZFS is able to aggregate and amortize metadata (and data)
	* changes. However, memory is a limited resource, and allowing for more dirty
	* data comes at the cost of keeping other useful data in memory (for example
	* ZFS data cached by the ARC).
	*
	* Implementation
	*
	* As buffers are modified dsl_pool_willuse_space() increments both the per-
	* txg (dp_dirty_pertxg[]) and poolwide (dp_dirty_total) accounting of
	* dirty space used; dsl_pool_dirty_space() decrements those values as data
	* is synced out from dsl_pool_sync(). While only the poolwide value is
	* relevant, the per-txg value is useful for debugging. The tunable
	* zfs_dirty_data_max determines the dirty space limit. Once that value is
	* exceeded, new writes are halted until space frees up.
	*
	* The zfs_dirty_data_sync_percent tunable dictates the threshold at which we
	* ensure that there is a txg syncing (see the comment in txg.c for a full
	* description of transaction group stages).
	*
	* The IO scheduler uses both the dirty space limit and current amount of
	* dirty data as inputs. Those values affect the number of concurrent IOs ZFS
	* issues. See the comment in vdev_queue.c for details of the IO scheduler.
	*
	* The delay is also calculated based on the amount of dirty data. See the
	* comment above dmu_tx_delay() for details.
	*/

	/*
	* zfs_dirty_data_max will be set to zfs_dirty_data_max_percent% of all memory,
	* capped at zfs_dirty_data_max_max. It can also be overridden with a module
	* parameter.
	*/
	unsigned long zfs_dirty_data_max = 0;
	unsigned long zfs_dirty_data_max_max = 0;
	-int zfs_dirty_data_max_percent = 10;
	-int zfs_dirty_data_max_max_percent = 25;
	+uint_t zfs_dirty_data_max_percent = 10;
	+uint_t zfs_dirty_data_max_max_percent = 25;

	/*
	* The upper limit of TX_WRITE log data. Write operations are throttled
	* when approaching the limit until log data is cleared out after txg sync.
	* It only counts TX_WRITE log with WR_COPIED or WR_NEED_COPY.
	*/
	unsigned long zfs_wrlog_data_max = 0;

	/*
	* If there's at least this much dirty data (as a percentage of
	* zfs_dirty_data_max), push out a txg. This should be less than
	* zfs_vdev_async_write_active_min_dirty_percent.
	*/
	-static int zfs_dirty_data_sync_percent = 20;
	+static uint_t zfs_dirty_data_sync_percent = 20;

	/*
	* Once there is this amount of dirty data, the dmu_tx_delay() will kick in
	* and delay each transaction.
	* This value should be >= zfs_vdev_async_write_active_max_dirty_percent.
	*/
	-int zfs_delay_min_dirty_percent = 60;
	+uint_t zfs_delay_min_dirty_percent = 60;

	/*
	* This controls how quickly the delay approaches infinity.
	* Larger values cause it to delay more for a given amount of dirty data.
	* Therefore larger values will cause there to be less dirty data for a
	* given throughput.
	*
	* For the smoothest delay, this value should be about 1 billion divided
	* by the maximum number of operations per second. This will smoothly
	* handle between 10x and 1/10th this number.
	*
	* Note: zfs_delay_scale * zfs_dirty_data_max must be < 2^64, due to the
	* multiply in dmu_tx_delay().
	*/
	unsigned long zfs_delay_scale = 1000 * 1000 * 1000 / 2000;

	/*
	* This determines the number of threads used by the dp_sync_taskq.
	*/
	static int zfs_sync_taskq_batch_pct = 75;

	/*
	* These tunables determine the behavior of how zil_itxg_clean() is
	* called via zil_clean() in the context of spa_sync(). When an itxg
	* list needs to be cleaned, TQ_NOSLEEP will be used when dispatching.
	* If the dispatch fails, the call to zil_itxg_clean() will occur
	* synchronously in the context of spa_sync(), which can negatively
	* impact the performance of spa_sync() (e.g. in the case of the itxg
	* list having a large number of itxs that needs to be cleaned).
	*
	* Thus, these tunables can be used to manipulate the behavior of the
	* taskq used by zil_clean(); they determine the number of taskq entries
	* that are pre-populated when the taskq is first created (via the
	* "zfs_zil_clean_taskq_minalloc" tunable) and the maximum number of
	* taskq entries that are cached after an on-demand allocation (via the
	* "zfs_zil_clean_taskq_maxalloc").
	*
	* The idea being, we want to try reasonably hard to ensure there will
	* already be a taskq entry pre-allocated by the time that it is needed
	* by zil_clean(). This way, we can avoid the possibility of an
	* on-demand allocation of a new taskq entry from failing, which would
	* result in zil_itxg_clean() being called synchronously from zil_clean()
	* (which can adversely affect performance of spa_sync()).
	*
	* Additionally, the number of threads used by the taskq can be
	* configured via the "zfs_zil_clean_taskq_nthr_pct" tunable.
	*/
	static int zfs_zil_clean_taskq_nthr_pct = 100;
	static int zfs_zil_clean_taskq_minalloc = 1024;
	static int zfs_zil_clean_taskq_maxalloc = 1024 * 1024;

	int
	dsl_pool_open_special_dir(dsl_pool_t dp, const char name, dsl_dir_t **ddp)
	{
	uint64_t obj;
	int err;

	err = zap_lookup(dp->dp_meta_objset,
	dsl_dir_phys(dp->dp_root_dir)->dd_child_dir_zapobj,
	name, sizeof (obj), 1, &obj);
	if (err)
	return (err);

	return (dsl_dir_hold_obj(dp, obj, name, dp, ddp));
	}

	static dsl_pool_t *
	dsl_pool_open_impl(spa_t *spa, uint64_t txg)
	{
	dsl_pool_t *dp;
	blkptr_t *bp = spa_get_rootblkptr(spa);

	dp = kmem_zalloc(sizeof (dsl_pool_t), KM_SLEEP);
	dp->dp_spa = spa;
	dp->dp_meta_rootbp = *bp;
	rrw_init(&dp->dp_config_rwlock, B_TRUE);
	txg_init(dp, txg);
	mmp_init(spa);

	txg_list_create(&dp->dp_dirty_datasets, spa,
	offsetof(dsl_dataset_t, ds_dirty_link));
	txg_list_create(&dp->dp_dirty_zilogs, spa,
	offsetof(zilog_t, zl_dirty_link));
	txg_list_create(&dp->dp_dirty_dirs, spa,
	offsetof(dsl_dir_t, dd_dirty_link));
	txg_list_create(&dp->dp_sync_tasks, spa,
	offsetof(dsl_sync_task_t, dst_node));
	txg_list_create(&dp->dp_early_sync_tasks, spa,
	offsetof(dsl_sync_task_t, dst_node));

	dp->dp_sync_taskq = taskq_create("dp_sync_taskq",
	zfs_sync_taskq_batch_pct, minclsyspri, 1, INT_MAX,
	TASKQ_THREADS_CPU_PCT);

	dp->dp_zil_clean_taskq = taskq_create("dp_zil_clean_taskq",
	zfs_zil_clean_taskq_nthr_pct, minclsyspri,
	zfs_zil_clean_taskq_minalloc,
	zfs_zil_clean_taskq_maxalloc,
	TASKQ_PREPOPULATE \| TASKQ_THREADS_CPU_PCT);

	mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&dp->dp_spaceavail_cv, NULL, CV_DEFAULT, NULL);

	aggsum_init(&dp->dp_wrlog_total, 0);
	for (int i = 0; i < TXG_SIZE; i++) {
	aggsum_init(&dp->dp_wrlog_pertxg[i], 0);
	}

	dp->dp_zrele_taskq = taskq_create("z_zrele", 100, defclsyspri,
	boot_ncpus * 8, INT_MAX, TASKQ_PREPOPULATE \| TASKQ_DYNAMIC \|
	TASKQ_THREADS_CPU_PCT);
	dp->dp_unlinked_drain_taskq = taskq_create("z_unlinked_drain",
	100, defclsyspri, boot_ncpus, INT_MAX,
	TASKQ_PREPOPULATE \| TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);

	return (dp);
	}

	int
	dsl_pool_init(spa_t spa, uint64_t txg, dsl_pool_t *dpp)
	{
	int err;
	dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);

	/*
	* Initialize the caller's dsl_pool_t structure before we actually open
	* the meta objset. This is done because a self-healing write zio may
	* be issued as part of dmu_objset_open_impl() and the spa needs its
	* dsl_pool_t initialized in order to handle the write.
	*/
	*dpp = dp;

	err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
	&dp->dp_meta_objset);
	if (err != 0) {
	dsl_pool_close(dp);
	*dpp = NULL;
	}

	return (err);
	}

	int
	dsl_pool_open(dsl_pool_t *dp)
	{
	int err;
	dsl_dir_t *dd;
	dsl_dataset_t *ds;
	uint64_t obj;

	rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
	&dp->dp_root_dir_obj);
	if (err)
	goto out;

	err = dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
	NULL, dp, &dp->dp_root_dir);
	if (err)
	goto out;

	err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
	if (err)
	goto out;

	if (spa_version(dp->dp_spa) >= SPA_VERSION_ORIGIN) {
	err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
	if (err)
	goto out;
	err = dsl_dataset_hold_obj(dp,
	dsl_dir_phys(dd)->dd_head_dataset_obj, FTAG, &ds);
	if (err == 0) {
	err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, dp,
	&dp->dp_origin_snap);
	dsl_dataset_rele(ds, FTAG);
	}
	dsl_dir_rele(dd, dp);
	if (err)
	goto out;
	}

	if (spa_version(dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
	err = dsl_pool_open_special_dir(dp, FREE_DIR_NAME,
	&dp->dp_free_dir);
	if (err)
	goto out;

	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj);
	if (err)
	goto out;
	VERIFY0(bpobj_open(&dp->dp_free_bpobj,
	dp->dp_meta_objset, obj));
	}

	if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ, sizeof (uint64_t), 1, &obj);
	if (err == 0) {
	VERIFY0(bpobj_open(&dp->dp_obsolete_bpobj,
	dp->dp_meta_objset, obj));
	} else if (err == ENOENT) {
	/*
	* We might not have created the remap bpobj yet.
	*/
	err = 0;
	} else {
	goto out;
	}
	}

	/*
	* Note: errors ignored, because the these special dirs, used for
	* space accounting, are only created on demand.
	*/
	(void) dsl_pool_open_special_dir(dp, LEAK_DIR_NAME,
	&dp->dp_leak_dir);

	if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_ASYNC_DESTROY)) {
	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_BPTREE_OBJ, sizeof (uint64_t), 1,
	&dp->dp_bptree_obj);
	if (err != 0)
	goto out;
	}

	if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMPTY_BPOBJ)) {
	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
	&dp->dp_empty_bpobj);
	if (err != 0)
	goto out;
	}

	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_TMP_USERREFS, sizeof (uint64_t), 1,
	&dp->dp_tmp_userrefs_obj);
	if (err == ENOENT)
	err = 0;
	if (err)
	goto out;

	err = dsl_scan_init(dp, dp->dp_tx.tx_open_txg);

	out:
	rrw_exit(&dp->dp_config_rwlock, FTAG);
	return (err);
	}

	void
	dsl_pool_close(dsl_pool_t *dp)
	{
	/*
	* Drop our references from dsl_pool_open().
	*
	* Since we held the origin_snap from "syncing" context (which
	* includes pool-opening context), it actually only got a "ref"
	* and not a hold, so just drop that here.
	*/
	if (dp->dp_origin_snap != NULL)
	dsl_dataset_rele(dp->dp_origin_snap, dp);
	if (dp->dp_mos_dir != NULL)
	dsl_dir_rele(dp->dp_mos_dir, dp);
	if (dp->dp_free_dir != NULL)
	dsl_dir_rele(dp->dp_free_dir, dp);
	if (dp->dp_leak_dir != NULL)
	dsl_dir_rele(dp->dp_leak_dir, dp);
	if (dp->dp_root_dir != NULL)
	dsl_dir_rele(dp->dp_root_dir, dp);

	bpobj_close(&dp->dp_free_bpobj);
	bpobj_close(&dp->dp_obsolete_bpobj);

	/* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
	if (dp->dp_meta_objset != NULL)
	dmu_objset_evict(dp->dp_meta_objset);

	txg_list_destroy(&dp->dp_dirty_datasets);
	txg_list_destroy(&dp->dp_dirty_zilogs);
	txg_list_destroy(&dp->dp_sync_tasks);
	txg_list_destroy(&dp->dp_early_sync_tasks);
	txg_list_destroy(&dp->dp_dirty_dirs);

	taskq_destroy(dp->dp_zil_clean_taskq);
	taskq_destroy(dp->dp_sync_taskq);

	/*
	* We can't set retry to TRUE since we're explicitly specifying
	* a spa to flush. This is good enough; any missed buffers for
	* this spa won't cause trouble, and they'll eventually fall
	* out of the ARC just like any other unused buffer.
	*/
	arc_flush(dp->dp_spa, FALSE);

	mmp_fini(dp->dp_spa);
	txg_fini(dp);
	dsl_scan_fini(dp);
	dmu_buf_user_evict_wait();

	rrw_destroy(&dp->dp_config_rwlock);
	mutex_destroy(&dp->dp_lock);
	cv_destroy(&dp->dp_spaceavail_cv);

	ASSERT0(aggsum_value(&dp->dp_wrlog_total));
	aggsum_fini(&dp->dp_wrlog_total);
	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT0(aggsum_value(&dp->dp_wrlog_pertxg[i]));
	aggsum_fini(&dp->dp_wrlog_pertxg[i]);
	}

	taskq_destroy(dp->dp_unlinked_drain_taskq);
	taskq_destroy(dp->dp_zrele_taskq);
	if (dp->dp_blkstats != NULL)
	vmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
	kmem_free(dp, sizeof (dsl_pool_t));
	}

	void
	dsl_pool_create_obsolete_bpobj(dsl_pool_t dp, dmu_tx_t tx)
	{
	uint64_t obj;
	/*
	* Currently, we only create the obsolete_bpobj where there are
	* indirect vdevs with referenced mappings.
	*/
	ASSERT(spa_feature_is_active(dp->dp_spa, SPA_FEATURE_DEVICE_REMOVAL));
	/* create and open the obsolete_bpobj */
	obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
	VERIFY0(bpobj_open(&dp->dp_obsolete_bpobj, dp->dp_meta_objset, obj));
	VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
	spa_feature_incr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	}

	void
	dsl_pool_destroy_obsolete_bpobj(dsl_pool_t dp, dmu_tx_t tx)
	{
	spa_feature_decr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	VERIFY0(zap_remove(dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ, tx));
	bpobj_free(dp->dp_meta_objset,
	dp->dp_obsolete_bpobj.bpo_object, tx);
	bpobj_close(&dp->dp_obsolete_bpobj);
	}

	dsl_pool_t *
	dsl_pool_create(spa_t spa, nvlist_t zplprops __attribute__((unused)),
	dsl_crypto_params_t *dcp, uint64_t txg)
	{
	int err;
	dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
	dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
	#ifdef _KERNEL
	objset_t *os;
	#else
	objset_t *os __attribute__((unused));
	#endif
	dsl_dataset_t *ds;
	uint64_t obj;

	rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);

	/* create and open the MOS (meta-objset) */
	dp->dp_meta_objset = dmu_objset_create_impl(spa,
	NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx);
	spa->spa_meta_objset = dp->dp_meta_objset;

	/* create the pool directory */
	err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
	ASSERT0(err);

	/* Initialize scan structures */
	VERIFY0(dsl_scan_init(dp, txg));

	/* create and open the root dir */
	dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
	VERIFY0(dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
	NULL, dp, &dp->dp_root_dir));

	/* create and open the meta-objset dir */
	(void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
	VERIFY0(dsl_pool_open_special_dir(dp,
	MOS_DIR_NAME, &dp->dp_mos_dir));

	if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
	/* create and open the free dir */
	(void) dsl_dir_create_sync(dp, dp->dp_root_dir,
	FREE_DIR_NAME, tx);
	VERIFY0(dsl_pool_open_special_dir(dp,
	FREE_DIR_NAME, &dp->dp_free_dir));

	/* create and open the free_bplist */
	obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
	VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0);
	VERIFY0(bpobj_open(&dp->dp_free_bpobj,
	dp->dp_meta_objset, obj));
	}

	if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
	dsl_pool_create_origin(dp, tx);

	/*
	* Some features may be needed when creating the root dataset, so we
	* create the feature objects here.
	*/
	if (spa_version(spa) >= SPA_VERSION_FEATURES)
	spa_feature_create_zap_objects(spa, tx);

	if (dcp != NULL && dcp->cp_crypt != ZIO_CRYPT_OFF &&
	dcp->cp_crypt != ZIO_CRYPT_INHERIT)
	spa_feature_enable(spa, SPA_FEATURE_ENCRYPTION, tx);

	/* create the root dataset */
	obj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, dcp, 0, tx);

	/* create the root objset */
	VERIFY0(dsl_dataset_hold_obj_flags(dp, obj,
	DS_HOLD_FLAG_DECRYPT, FTAG, &ds));
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	os = dmu_objset_create_impl(dp->dp_spa, ds,
	dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	#ifdef _KERNEL
	zfs_create_fs(os, kcred, zplprops, tx);
	#endif
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);

	dmu_tx_commit(tx);

	rrw_exit(&dp->dp_config_rwlock, FTAG);

	return (dp);
	}

	/*
	* Account for the meta-objset space in its placeholder dsl_dir.
	*/
	void
	dsl_pool_mos_diduse_space(dsl_pool_t *dp,
	int64_t used, int64_t comp, int64_t uncomp)
	{
	ASSERT3U(comp, ==, uncomp); /* it's all metadata */
	mutex_enter(&dp->dp_lock);
	dp->dp_mos_used_delta += used;
	dp->dp_mos_compressed_delta += comp;
	dp->dp_mos_uncompressed_delta += uncomp;
	mutex_exit(&dp->dp_lock);
	}

	static void
	dsl_pool_sync_mos(dsl_pool_t dp, dmu_tx_t tx)
	{
	zio_t *zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	dmu_objset_sync(dp->dp_meta_objset, zio, tx);
	VERIFY0(zio_wait(zio));
	dmu_objset_sync_done(dp->dp_meta_objset, tx);
	taskq_wait(dp->dp_sync_taskq);
	multilist_destroy(&dp->dp_meta_objset->os_synced_dnodes);

	dprintf_bp(&dp->dp_meta_rootbp, "meta objset rootbp is %s", "");
	spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
	}

	static void
	dsl_pool_dirty_delta(dsl_pool_t *dp, int64_t delta)
	{
	ASSERT(MUTEX_HELD(&dp->dp_lock));

	if (delta < 0)
	ASSERT3U(-delta, <=, dp->dp_dirty_total);

	dp->dp_dirty_total += delta;

	/*
	* Note: we signal even when increasing dp_dirty_total.
	* This ensures forward progress -- each thread wakes the next waiter.
	*/
	if (dp->dp_dirty_total < zfs_dirty_data_max)
	cv_signal(&dp->dp_spaceavail_cv);
	}

	void
	dsl_pool_wrlog_count(dsl_pool_t *dp, int64_t size, uint64_t txg)
	{
	ASSERT3S(size, >=, 0);

	aggsum_add(&dp->dp_wrlog_pertxg[txg & TXG_MASK], size);
	aggsum_add(&dp->dp_wrlog_total, size);

	/* Choose a value slightly bigger than min dirty sync bytes */
	uint64_t sync_min =
	zfs_wrlog_data_max * (zfs_dirty_data_sync_percent + 10) / 200;
	if (aggsum_compare(&dp->dp_wrlog_pertxg[txg & TXG_MASK], sync_min) > 0)
	txg_kick(dp, txg);
	}

	boolean_t
	dsl_pool_need_wrlog_delay(dsl_pool_t *dp)
	{
	uint64_t delay_min_bytes =
	zfs_wrlog_data_max * zfs_delay_min_dirty_percent / 100;

	return (aggsum_compare(&dp->dp_wrlog_total, delay_min_bytes) > 0);
	}

	static void
	dsl_pool_wrlog_clear(dsl_pool_t *dp, uint64_t txg)
	{
	int64_t delta;
	delta = -(int64_t)aggsum_value(&dp->dp_wrlog_pertxg[txg & TXG_MASK]);
	aggsum_add(&dp->dp_wrlog_pertxg[txg & TXG_MASK], delta);
	aggsum_add(&dp->dp_wrlog_total, delta);
	/* Compact per-CPU sums after the big change. */
	(void) aggsum_value(&dp->dp_wrlog_pertxg[txg & TXG_MASK]);
	(void) aggsum_value(&dp->dp_wrlog_total);
	}

	#ifdef ZFS_DEBUG
	static boolean_t
	dsl_early_sync_task_verify(dsl_pool_t *dp, uint64_t txg)
	{
	spa_t *spa = dp->dp_spa;
	vdev_t *rvd = spa->spa_root_vdev;

	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];
	txg_list_t *tl = &vd->vdev_ms_list;
	metaslab_t *ms;

	for (ms = txg_list_head(tl, TXG_CLEAN(txg)); ms;
	ms = txg_list_next(tl, ms, TXG_CLEAN(txg))) {
	VERIFY(range_tree_is_empty(ms->ms_freeing));
	VERIFY(range_tree_is_empty(ms->ms_checkpointing));
	}
	}

	return (B_TRUE);
	}
	#else
	#define dsl_early_sync_task_verify(dp, txg) \
	((void) sizeof (dp), (void) sizeof (txg), B_TRUE)
	#endif

	void
	dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
	{
	zio_t *zio;
	dmu_tx_t *tx;
	dsl_dir_t *dd;
	dsl_dataset_t *ds;
	objset_t *mos = dp->dp_meta_objset;
	list_t synced_datasets;

	list_create(&synced_datasets, sizeof (dsl_dataset_t),
	offsetof(dsl_dataset_t, ds_synced_link));

	tx = dmu_tx_create_assigned(dp, txg);

	/*
	* Run all early sync tasks before writing out any dirty blocks.
	* For more info on early sync tasks see block comment in
	* dsl_early_sync_task().
	*/
	if (!txg_list_empty(&dp->dp_early_sync_tasks, txg)) {
	dsl_sync_task_t *dst;

	ASSERT3U(spa_sync_pass(dp->dp_spa), ==, 1);
	while ((dst =
	txg_list_remove(&dp->dp_early_sync_tasks, txg)) != NULL) {
	ASSERT(dsl_early_sync_task_verify(dp, txg));
	dsl_sync_task_sync(dst, tx);
	}
	ASSERT(dsl_early_sync_task_verify(dp, txg));
	}

	/*
	* Write out all dirty blocks of dirty datasets.
	*/
	zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
	/*
	* We must not sync any non-MOS datasets twice, because
	* we may have taken a snapshot of them. However, we
	* may sync newly-created datasets on pass 2.
	*/
	ASSERT(!list_link_active(&ds->ds_synced_link));
	list_insert_tail(&synced_datasets, ds);
	dsl_dataset_sync(ds, zio, tx);
	}
	VERIFY0(zio_wait(zio));

	/*
	* Update the long range free counter after
	* we're done syncing user data
	*/
	mutex_enter(&dp->dp_lock);
	ASSERT(spa_sync_pass(dp->dp_spa) == 1 \|\|
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] == 0);
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] = 0;
	mutex_exit(&dp->dp_lock);

	/*
	* After the data blocks have been written (ensured by the zio_wait()
	* above), update the user/group/project space accounting. This happens
	* in tasks dispatched to dp_sync_taskq, so wait for them before
	* continuing.
	*/
	for (ds = list_head(&synced_datasets); ds != NULL;
	ds = list_next(&synced_datasets, ds)) {
	dmu_objset_sync_done(ds->ds_objset, tx);
	}
	taskq_wait(dp->dp_sync_taskq);

	/*
	* Sync the datasets again to push out the changes due to
	* userspace updates. This must be done before we process the
	* sync tasks, so that any snapshots will have the correct
	* user accounting information (and we won't get confused
	* about which blocks are part of the snapshot).
	*/
	zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	while ((ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) != NULL) {
	objset_t *os = ds->ds_objset;

	ASSERT(list_link_active(&ds->ds_synced_link));
	dmu_buf_rele(ds->ds_dbuf, ds);
	dsl_dataset_sync(ds, zio, tx);

	/*
	* Release any key mappings created by calls to
	* dsl_dataset_dirty() from the userquota accounting
	* code paths.
	*/
	if (os->os_encrypted && !os->os_raw_receive &&
	!os->os_next_write_raw[txg & TXG_MASK]) {
	ASSERT3P(ds->ds_key_mapping, !=, NULL);
	key_mapping_rele(dp->dp_spa, ds->ds_key_mapping, ds);
	}
	}
	VERIFY0(zio_wait(zio));

	/*
	* Now that the datasets have been completely synced, we can
	* clean up our in-memory structures accumulated while syncing:
	*
	* - move dead blocks from the pending deadlist and livelists
	* to the on-disk versions
	* - release hold from dsl_dataset_dirty()
	* - release key mapping hold from dsl_dataset_dirty()
	*/
	while ((ds = list_remove_head(&synced_datasets)) != NULL) {
	objset_t *os = ds->ds_objset;

	if (os->os_encrypted && !os->os_raw_receive &&
	!os->os_next_write_raw[txg & TXG_MASK]) {
	ASSERT3P(ds->ds_key_mapping, !=, NULL);
	key_mapping_rele(dp->dp_spa, ds->ds_key_mapping, ds);
	}

	dsl_dataset_sync_done(ds, tx);
	}

	while ((dd = txg_list_remove(&dp->dp_dirty_dirs, txg)) != NULL) {
	dsl_dir_sync(dd, tx);
	}

	/*
	* The MOS's space is accounted for in the pool/$MOS
	* (dp_mos_dir). We can't modify the mos while we're syncing
	* it, so we remember the deltas and apply them here.
	*/
	if (dp->dp_mos_used_delta != 0 \|\| dp->dp_mos_compressed_delta != 0 \|\|
	dp->dp_mos_uncompressed_delta != 0) {
	dsl_dir_diduse_space(dp->dp_mos_dir, DD_USED_HEAD,
	dp->dp_mos_used_delta,
	dp->dp_mos_compressed_delta,
	dp->dp_mos_uncompressed_delta, tx);
	dp->dp_mos_used_delta = 0;
	dp->dp_mos_compressed_delta = 0;
	dp->dp_mos_uncompressed_delta = 0;
	}

	if (dmu_objset_is_dirty(mos, txg)) {
	dsl_pool_sync_mos(dp, tx);
	}

	/*
	* We have written all of the accounted dirty data, so our
	* dp_space_towrite should now be zero. However, some seldom-used
	* code paths do not adhere to this (e.g. dbuf_undirty()). Shore up
	* the accounting of any dirtied space now.
	*
	* Note that, besides any dirty data from datasets, the amount of
	* dirty data in the MOS is also accounted by the pool. Therefore,
	* we want to do this cleanup after dsl_pool_sync_mos() so we don't
	* attempt to update the accounting for the same dirty data twice.
	* (i.e. at this point we only update the accounting for the space
	* that we know that we "leaked").
	*/
	dsl_pool_undirty_space(dp, dp->dp_dirty_pertxg[txg & TXG_MASK], txg);

	/*
	* If we modify a dataset in the same txg that we want to destroy it,
	* its dsl_dir's dd_dbuf will be dirty, and thus have a hold on it.
	* dsl_dir_destroy_check() will fail if there are unexpected holds.
	* Therefore, we want to sync the MOS (thus syncing the dd_dbuf
	* and clearing the hold on it) before we process the sync_tasks.
	* The MOS data dirtied by the sync_tasks will be synced on the next
	* pass.
	*/
	if (!txg_list_empty(&dp->dp_sync_tasks, txg)) {
	dsl_sync_task_t *dst;
	/*
	* No more sync tasks should have been added while we
	* were syncing.
	*/
	ASSERT3U(spa_sync_pass(dp->dp_spa), ==, 1);
	while ((dst = txg_list_remove(&dp->dp_sync_tasks, txg)) != NULL)
	dsl_sync_task_sync(dst, tx);
	}

	dmu_tx_commit(tx);

	DTRACE_PROBE2(dsl_pool_sync__done, dsl_pool_t *dp, dp, uint64_t, txg);
	}

	void
	dsl_pool_sync_done(dsl_pool_t *dp, uint64_t txg)
	{
	zilog_t *zilog;

	while ((zilog = txg_list_head(&dp->dp_dirty_zilogs, txg))) {
	dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);
	/*
	* We don't remove the zilog from the dp_dirty_zilogs
	* list until after we've cleaned it. This ensures that
	* callers of zilog_is_dirty() receive an accurate
	* answer when they are racing with the spa sync thread.
	*/
	zil_clean(zilog, txg);
	(void) txg_list_remove_this(&dp->dp_dirty_zilogs, zilog, txg);
	ASSERT(!dmu_objset_is_dirty(zilog->zl_os, txg));
	dmu_buf_rele(ds->ds_dbuf, zilog);
	}

	dsl_pool_wrlog_clear(dp, txg);

	ASSERT(!dmu_objset_is_dirty(dp->dp_meta_objset, txg));
	}

	/*
	* TRUE if the current thread is the tx_sync_thread or if we
	* are being called from SPA context during pool initialization.
	*/
	int
	dsl_pool_sync_context(dsl_pool_t *dp)
	{
	return (curthread == dp->dp_tx.tx_sync_thread \|\|
	spa_is_initializing(dp->dp_spa) \|\|
	taskq_member(dp->dp_sync_taskq, curthread));
	}

	/*
	* This function returns the amount of allocatable space in the pool
	* minus whatever space is currently reserved by ZFS for specific
	* purposes. Specifically:
	*
	* 1] Any reserved SLOP space
	* 2] Any space used by the checkpoint
	* 3] Any space used for deferred frees
	*
	* The latter 2 are especially important because they are needed to
	* rectify the SPA's and DMU's different understanding of how much space
	* is used. Now the DMU is aware of that extra space tracked by the SPA
	* without having to maintain a separate special dir (e.g similar to
	* $MOS, $FREEING, and $LEAKED).
	*
	* Note: By deferred frees here, we mean the frees that were deferred
	* in spa_sync() after sync pass 1 (spa_deferred_bpobj), and not the
	* segments placed in ms_defer trees during metaslab_sync_done().
	*/
	uint64_t
	dsl_pool_adjustedsize(dsl_pool_t *dp, zfs_space_check_t slop_policy)
	{
	spa_t *spa = dp->dp_spa;
	uint64_t space, resv, adjustedsize;
	uint64_t spa_deferred_frees =
	spa->spa_deferred_bpobj.bpo_phys->bpo_bytes;

	space = spa_get_dspace(spa)
	- spa_get_checkpoint_space(spa) - spa_deferred_frees;
	resv = spa_get_slop_space(spa);

	switch (slop_policy) {
	case ZFS_SPACE_CHECK_NORMAL:
	break;
	case ZFS_SPACE_CHECK_RESERVED:
	resv >>= 1;
	break;
	case ZFS_SPACE_CHECK_EXTRA_RESERVED:
	resv >>= 2;
	break;
	case ZFS_SPACE_CHECK_NONE:
	resv = 0;
	break;
	default:
	panic("invalid slop policy value: %d", slop_policy);
	break;
	}
	adjustedsize = (space >= resv) ? (space - resv) : 0;

	return (adjustedsize);
	}

	uint64_t
	dsl_pool_unreserved_space(dsl_pool_t *dp, zfs_space_check_t slop_policy)
	{
	uint64_t poolsize = dsl_pool_adjustedsize(dp, slop_policy);
	uint64_t deferred =
	metaslab_class_get_deferred(spa_normal_class(dp->dp_spa));
	uint64_t quota = (poolsize >= deferred) ? (poolsize - deferred) : 0;
	return (quota);
	}

	uint64_t
	dsl_pool_deferred_space(dsl_pool_t *dp)
	{
	return (metaslab_class_get_deferred(spa_normal_class(dp->dp_spa)));
	}

	boolean_t
	dsl_pool_need_dirty_delay(dsl_pool_t *dp)
	{
	uint64_t delay_min_bytes =
	zfs_dirty_data_max * zfs_delay_min_dirty_percent / 100;

	mutex_enter(&dp->dp_lock);
	uint64_t dirty = dp->dp_dirty_total;
	mutex_exit(&dp->dp_lock);

	return (dirty > delay_min_bytes);
	}

	static boolean_t
	dsl_pool_need_dirty_sync(dsl_pool_t *dp, uint64_t txg)
	{
	ASSERT(MUTEX_HELD(&dp->dp_lock));

	uint64_t dirty_min_bytes =
	zfs_dirty_data_max * zfs_dirty_data_sync_percent / 100;
	uint64_t dirty = dp->dp_dirty_pertxg[txg & TXG_MASK];

	return (dirty > dirty_min_bytes);
	}

	void
	dsl_pool_dirty_space(dsl_pool_t dp, int64_t space, dmu_tx_t tx)
	{
	if (space > 0) {
	mutex_enter(&dp->dp_lock);
	dp->dp_dirty_pertxg[tx->tx_txg & TXG_MASK] += space;
	dsl_pool_dirty_delta(dp, space);
	boolean_t needsync = !dmu_tx_is_syncing(tx) &&
	dsl_pool_need_dirty_sync(dp, tx->tx_txg);
	mutex_exit(&dp->dp_lock);

	if (needsync)
	txg_kick(dp, tx->tx_txg);
	}
	}

	void
	dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg)
	{
	ASSERT3S(space, >=, 0);
	if (space == 0)
	return;

	mutex_enter(&dp->dp_lock);
	if (dp->dp_dirty_pertxg[txg & TXG_MASK] < space) {
	/* XXX writing something we didn't dirty? */
	space = dp->dp_dirty_pertxg[txg & TXG_MASK];
	}
	ASSERT3U(dp->dp_dirty_pertxg[txg & TXG_MASK], >=, space);
	dp->dp_dirty_pertxg[txg & TXG_MASK] -= space;
	ASSERT3U(dp->dp_dirty_total, >=, space);
	dsl_pool_dirty_delta(dp, -space);
	mutex_exit(&dp->dp_lock);
	}

	static int
	upgrade_clones_cb(dsl_pool_t dp, dsl_dataset_t hds, void *arg)
	{
	dmu_tx_t *tx = arg;
	dsl_dataset_t ds, prev = NULL;
	int err;

	err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
	if (err)
	return (err);

	while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
	err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
	if (err) {
	dsl_dataset_rele(ds, FTAG);
	return (err);
	}

	if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object)
	break;
	dsl_dataset_rele(ds, FTAG);
	ds = prev;
	prev = NULL;
	}

	if (prev == NULL) {
	prev = dp->dp_origin_snap;

	/*
	* The $ORIGIN can't have any data, or the accounting
	* will be wrong.
	*/
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	ASSERT0(dsl_dataset_phys(prev)->ds_bp.blk_birth);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	/* The origin doesn't get attached to itself */
	if (ds->ds_object == prev->ds_object) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_prev_snap_obj = prev->ds_object;
	dsl_dataset_phys(ds)->ds_prev_snap_txg =
	dsl_dataset_phys(prev)->ds_creation_txg;

	dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
	dsl_dir_phys(ds->ds_dir)->dd_origin_obj = prev->ds_object;

	dmu_buf_will_dirty(prev->ds_dbuf, tx);
	dsl_dataset_phys(prev)->ds_num_children++;

	if (dsl_dataset_phys(ds)->ds_next_snap_obj == 0) {
	ASSERT(ds->ds_prev == NULL);
	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj,
	ds, &ds->ds_prev));
	}
	}

	ASSERT3U(dsl_dir_phys(ds->ds_dir)->dd_origin_obj, ==, prev->ds_object);
	ASSERT3U(dsl_dataset_phys(ds)->ds_prev_snap_obj, ==, prev->ds_object);

	if (dsl_dataset_phys(prev)->ds_next_clones_obj == 0) {
	dmu_buf_will_dirty(prev->ds_dbuf, tx);
	dsl_dataset_phys(prev)->ds_next_clones_obj =
	zap_create(dp->dp_meta_objset,
	DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
	}
	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dataset_phys(prev)->ds_next_clones_obj, ds->ds_object, tx));

	dsl_dataset_rele(ds, FTAG);
	if (prev != dp->dp_origin_snap)
	dsl_dataset_rele(prev, FTAG);
	return (0);
	}

	void
	dsl_pool_upgrade_clones(dsl_pool_t dp, dmu_tx_t tx)
	{
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(dp->dp_origin_snap != NULL);

	VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj, upgrade_clones_cb,
	tx, DS_FIND_CHILDREN \| DS_FIND_SERIALIZE));
	}

	static int
	upgrade_dir_clones_cb(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	dmu_tx_t *tx = arg;
	objset_t *mos = dp->dp_meta_objset;

	if (dsl_dir_phys(ds->ds_dir)->dd_origin_obj != 0) {
	dsl_dataset_t *origin;

	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &origin));

	if (dsl_dir_phys(origin->ds_dir)->dd_clones == 0) {
	dmu_buf_will_dirty(origin->ds_dir->dd_dbuf, tx);
	dsl_dir_phys(origin->ds_dir)->dd_clones =
	zap_create(mos, DMU_OT_DSL_CLONES, DMU_OT_NONE,
	0, tx);
	}

	VERIFY0(zap_add_int(dp->dp_meta_objset,
	dsl_dir_phys(origin->ds_dir)->dd_clones,
	ds->ds_object, tx));

	dsl_dataset_rele(origin, FTAG);
	}
	return (0);
	}

	void
	dsl_pool_upgrade_dir_clones(dsl_pool_t dp, dmu_tx_t tx)
	{
	uint64_t obj;

	ASSERT(dmu_tx_is_syncing(tx));

	(void) dsl_dir_create_sync(dp, dp->dp_root_dir, FREE_DIR_NAME, tx);
	VERIFY0(dsl_pool_open_special_dir(dp,
	FREE_DIR_NAME, &dp->dp_free_dir));

	/*
	* We can't use bpobj_alloc(), because spa_version() still
	* returns the old version, and we need a new-version bpobj with
	* subobj support. So call dmu_object_alloc() directly.
	*/
	obj = dmu_object_alloc(dp->dp_meta_objset, DMU_OT_BPOBJ,
	SPA_OLD_MAXBLOCKSIZE, DMU_OT_BPOBJ_HDR, sizeof (bpobj_phys_t), tx);
	VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
	VERIFY0(bpobj_open(&dp->dp_free_bpobj, dp->dp_meta_objset, obj));

	VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	upgrade_dir_clones_cb, tx, DS_FIND_CHILDREN \| DS_FIND_SERIALIZE));
	}

	void
	dsl_pool_create_origin(dsl_pool_t dp, dmu_tx_t tx)
	{
	uint64_t dsobj;
	dsl_dataset_t *ds;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT(dp->dp_origin_snap == NULL);
	ASSERT(rrw_held(&dp->dp_config_rwlock, RW_WRITER));

	/* create the origin dir, ds, & snap-ds */
	dsobj = dsl_dataset_create_sync(dp->dp_root_dir, ORIGIN_DIR_NAME,
	NULL, 0, kcred, NULL, tx);
	VERIFY0(dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
	dsl_dataset_snapshot_sync_impl(ds, ORIGIN_DIR_NAME, tx);
	VERIFY0(dsl_dataset_hold_obj(dp, dsl_dataset_phys(ds)->ds_prev_snap_obj,
	dp, &dp->dp_origin_snap));
	dsl_dataset_rele(ds, FTAG);
	}

	taskq_t *
	dsl_pool_zrele_taskq(dsl_pool_t *dp)
	{
	return (dp->dp_zrele_taskq);
	}

	taskq_t *
	dsl_pool_unlinked_drain_taskq(dsl_pool_t *dp)
	{
	return (dp->dp_unlinked_drain_taskq);
	}

	/*
	* Walk through the pool-wide zap object of temporary snapshot user holds
	* and release them.
	*/
	void
	dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp)
	{
	zap_attribute_t za;
	zap_cursor_t zc;
	objset_t *mos = dp->dp_meta_objset;
	uint64_t zapobj = dp->dp_tmp_userrefs_obj;
	nvlist_t *holds;

	if (zapobj == 0)
	return;
	ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);

	holds = fnvlist_alloc();

	for (zap_cursor_init(&zc, mos, zapobj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	char *htag;
	nvlist_t *tags;

	htag = strchr(za.za_name, '-');
	*htag = '\0';
	++htag;
	if (nvlist_lookup_nvlist(holds, za.za_name, &tags) != 0) {
	tags = fnvlist_alloc();
	fnvlist_add_boolean(tags, htag);
	fnvlist_add_nvlist(holds, za.za_name, tags);
	fnvlist_free(tags);
	} else {
	fnvlist_add_boolean(tags, htag);
	}
	}
	dsl_dataset_user_release_tmp(dp, holds);
	fnvlist_free(holds);
	zap_cursor_fini(&zc);
	}

	/*
	* Create the pool-wide zap object for storing temporary snapshot holds.
	*/
	static void
	dsl_pool_user_hold_create_obj(dsl_pool_t dp, dmu_tx_t tx)
	{
	objset_t *mos = dp->dp_meta_objset;

	ASSERT(dp->dp_tmp_userrefs_obj == 0);
	ASSERT(dmu_tx_is_syncing(tx));

	dp->dp_tmp_userrefs_obj = zap_create_link(mos, DMU_OT_USERREFS,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_TMP_USERREFS, tx);
	}

	static int
	dsl_pool_user_hold_rele_impl(dsl_pool_t *dp, uint64_t dsobj,
	const char tag, uint64_t now, dmu_tx_t tx, boolean_t holding)
	{
	objset_t *mos = dp->dp_meta_objset;
	uint64_t zapobj = dp->dp_tmp_userrefs_obj;
	char *name;
	int error;

	ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
	ASSERT(dmu_tx_is_syncing(tx));

	/*
	* If the pool was created prior to SPA_VERSION_USERREFS, the
	* zap object for temporary holds might not exist yet.
	*/
	if (zapobj == 0) {
	if (holding) {
	dsl_pool_user_hold_create_obj(dp, tx);
	zapobj = dp->dp_tmp_userrefs_obj;
	} else {
	return (SET_ERROR(ENOENT));
	}
	}

	name = kmem_asprintf("%llx-%s", (u_longlong_t)dsobj, tag);
	if (holding)
	error = zap_add(mos, zapobj, name, 8, 1, &now, tx);
	else
	error = zap_remove(mos, zapobj, name, tx);
	kmem_strfree(name);

	return (error);
	}

	/*
	* Add a temporary hold for the given dataset object and tag.
	*/
	int
	dsl_pool_user_hold(dsl_pool_t dp, uint64_t dsobj, const char tag,
	uint64_t now, dmu_tx_t *tx)
	{
	return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, now, tx, B_TRUE));
	}

	/*
	* Release a temporary hold for the given dataset object and tag.
	*/
	int
	dsl_pool_user_release(dsl_pool_t dp, uint64_t dsobj, const char tag,
	dmu_tx_t *tx)
	{
	return (dsl_pool_user_hold_rele_impl(dp, dsobj, tag, 0,
	tx, B_FALSE));
	}

	/*
	* DSL Pool Configuration Lock
	*
	* The dp_config_rwlock protects against changes to DSL state (e.g. dataset
	* creation / destruction / rename / property setting). It must be held for
	* read to hold a dataset or dsl_dir. I.e. you must call
	* dsl_pool_config_enter() or dsl_pool_hold() before calling
	* dsl_{dataset,dir}_hold{_obj}. In most circumstances, the dp_config_rwlock
	* must be held continuously until all datasets and dsl_dirs are released.
	*
	* The only exception to this rule is that if a "long hold" is placed on
	* a dataset, then the dp_config_rwlock may be dropped while the dataset
	* is still held. The long hold will prevent the dataset from being
	* destroyed -- the destroy will fail with EBUSY. A long hold can be
	* obtained by calling dsl_dataset_long_hold(), or by "owning" a dataset
	* (by calling dsl_{dataset,objset}_{try}own{_obj}).
	*
	* Legitimate long-holders (including owners) should be long-running, cancelable
	* tasks that should cause "zfs destroy" to fail. This includes DMU
	* consumers (i.e. a ZPL filesystem being mounted or ZVOL being open),
	* "zfs send", and "zfs diff". There are several other long-holders whose
	* uses are suboptimal (e.g. "zfs promote", and zil_suspend()).
	*
	* The usual formula for long-holding would be:
	* dsl_pool_hold()
	* dsl_dataset_hold()
	* ... perform checks ...
	* dsl_dataset_long_hold()
	* dsl_pool_rele()
	* ... perform long-running task ...
	* dsl_dataset_long_rele()
	* dsl_dataset_rele()
	*
	* Note that when the long hold is released, the dataset is still held but
	* the pool is not held. The dataset may change arbitrarily during this time
	* (e.g. it could be destroyed). Therefore you shouldn't do anything to the
	* dataset except release it.
	*
	* Operations generally fall somewhere into the following taxonomy:
	*
	* Read-Only Modifying
	*
	* Dataset Layer / MOS zfs get zfs destroy
	*
	* Individual Dataset read() write()
	*
	*
	* Dataset Layer Operations
	*
	* Modifying operations should generally use dsl_sync_task(). The synctask
	* infrastructure enforces proper locking strategy with respect to the
	* dp_config_rwlock. See the comment above dsl_sync_task() for details.
	*
	* Read-only operations will manually hold the pool, then the dataset, obtain
	* information from the dataset, then release the pool and dataset.
	* dmu_objset_{hold,rele}() are convenience routines that also do the pool
	* hold/rele.
	*
	*
	* Operations On Individual Datasets
	*
	* Objects _within_ an objset should only be modified by the current 'owner'
	* of the objset to prevent incorrect concurrent modification. Thus, use
	* {dmu_objset,dsl_dataset}_own to mark some entity as the current owner,
	* and fail with EBUSY if there is already an owner. The owner can then
	* implement its own locking strategy, independent of the dataset layer's
	* locking infrastructure.
	* (E.g., the ZPL has its own set of locks to control concurrency. A regular
	* vnop will not reach into the dataset layer).
	*
	* Ideally, objects would also only be read by the objset’s owner, so that we
	* don’t observe state mid-modification.
	* (E.g. the ZPL is creating a new object and linking it into a directory; if
	* you don’t coordinate with the ZPL to hold ZPL-level locks, you could see an
	* intermediate state. The ioctl level violates this but in pretty benign
	* ways, e.g. reading the zpl props object.)
	*/

	int
	dsl_pool_hold(const char name, const void tag, dsl_pool_t **dp)
	{
	spa_t *spa;
	int error;

	error = spa_open(name, &spa, tag);
	if (error == 0) {
	*dp = spa_get_dsl(spa);
	dsl_pool_config_enter(*dp, tag);
	}
	return (error);
	}

	void
	dsl_pool_rele(dsl_pool_t dp, const void tag)
	{
	dsl_pool_config_exit(dp, tag);
	spa_close(dp->dp_spa, tag);
	}

	void
	dsl_pool_config_enter(dsl_pool_t dp, const void tag)
	{
	/*
	* We use a "reentrant" reader-writer lock, but not reentrantly.
	*
	* The rrwlock can (with the track_all flag) track all reading threads,
	* which is very useful for debugging which code path failed to release
	* the lock, and for verifying that the current thread does hold
	* the lock.
	*
	* (Unlike a rwlock, which knows that N threads hold it for
	* read, but not which threads, so rw_held(RW_READER) returns TRUE
	* if any thread holds it for read, even if this thread doesn't).
	*/
	ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
	rrw_enter(&dp->dp_config_rwlock, RW_READER, tag);
	}

	void
	dsl_pool_config_enter_prio(dsl_pool_t dp, const void tag)
	{
	ASSERT(!rrw_held(&dp->dp_config_rwlock, RW_READER));
	rrw_enter_read_prio(&dp->dp_config_rwlock, tag);
	}

	void
	dsl_pool_config_exit(dsl_pool_t dp, const void tag)
	{
	rrw_exit(&dp->dp_config_rwlock, tag);
	}

	boolean_t
	dsl_pool_config_held(dsl_pool_t *dp)
	{
	return (RRW_LOCK_HELD(&dp->dp_config_rwlock));
	}

	boolean_t
	dsl_pool_config_held_writer(dsl_pool_t *dp)
	{
	return (RRW_WRITE_HELD(&dp->dp_config_rwlock));
	}

	EXPORT_SYMBOL(dsl_pool_config_enter);
	EXPORT_SYMBOL(dsl_pool_config_exit);

	/* zfs_dirty_data_max_percent only applied at module load in arc_init(). */
	-ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_percent, INT, ZMOD_RD,
	+ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_percent, UINT, ZMOD_RD,
	"Max percent of RAM allowed to be dirty");

	/* zfs_dirty_data_max_max_percent only applied at module load in arc_init(). */
	-ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_max_percent, INT, ZMOD_RD,
	+ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_max_percent, UINT, ZMOD_RD,
	"zfs_dirty_data_max upper bound as % of RAM");

	-ZFS_MODULE_PARAM(zfs, zfs_, delay_min_dirty_percent, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, delay_min_dirty_percent, UINT, ZMOD_RW,
	"Transaction delay threshold");

	ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max, ULONG, ZMOD_RW,
	"Determines the dirty space limit");

	ZFS_MODULE_PARAM(zfs, zfs_, wrlog_data_max, ULONG, ZMOD_RW,
	"The size limit of write-transaction zil log data");

	/* zfs_dirty_data_max_max only applied at module load in arc_init(). */
	ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_max_max, ULONG, ZMOD_RD,
	"zfs_dirty_data_max upper bound in bytes");

	-ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_sync_percent, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, dirty_data_sync_percent, UINT, ZMOD_RW,
	"Dirty data txg sync threshold as a percentage of zfs_dirty_data_max");

	ZFS_MODULE_PARAM(zfs, zfs_, delay_scale, ULONG, ZMOD_RW,
	"How quickly delay approaches infinity");

	ZFS_MODULE_PARAM(zfs, zfs_, sync_taskq_batch_pct, INT, ZMOD_RW,
	"Max percent of CPUs that are used to sync dirty data");

	ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_nthr_pct, INT, ZMOD_RW,
	"Max percent of CPUs that are used per dp_sync_taskq");

	ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_minalloc, INT, ZMOD_RW,
	"Number of taskq entries that are pre-populated");

	ZFS_MODULE_PARAM(zfs_zil, zfs_zil_, clean_taskq_maxalloc, INT, ZMOD_RW,
	"Max number of taskq entries that are cached");
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_prop.c b/sys/contrib/openzfs/module/zfs/dsl_prop.c
	index 1d3d26124949..610e887b3fba 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_prop.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_prop.c
	@@ -1,1287 +1,1287 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
	* Copyright (c) 2013 Martin Matuska. All rights reserved.
	* Copyright 2019 Joyent, Inc.
	*/

	#include <sys/zfs_context.h>
	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/spa.h>
	#include <sys/zap.h>
	#include <sys/fs/zfs.h>

	#include "zfs_prop.h"

	#define ZPROP_INHERIT_SUFFIX "$inherit"
	#define ZPROP_RECVD_SUFFIX "$recvd"

	static int
	dodefault(zfs_prop_t prop, int intsz, int numints, void *buf)
	{
	/*
	* The setonce properties are read-only, BUT they still
	* have a default value that can be used as the initial
	* value.
	*/
	if (prop == ZPROP_INVAL \|\|
	(zfs_prop_readonly(prop) && !zfs_prop_setonce(prop)))
	return (SET_ERROR(ENOENT));

	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
	if (intsz != 1)
	return (SET_ERROR(EOVERFLOW));
	- (void) strncpy(buf, zfs_prop_default_string(prop),
	+ (void) strlcpy(buf, zfs_prop_default_string(prop),
	numints);
	} else {
	if (intsz != 8 \|\| numints < 1)
	return (SET_ERROR(EOVERFLOW));

	(uint64_t )buf = zfs_prop_default_numeric(prop);
	}

	return (0);
	}

	int
	dsl_prop_get_dd(dsl_dir_t dd, const char propname,
	int intsz, int numints, void buf, char setpoint, boolean_t snapshot)
	{
	int err;
	dsl_dir_t *target = dd;
	objset_t *mos = dd->dd_pool->dp_meta_objset;
	zfs_prop_t prop;
	boolean_t inheritable;
	boolean_t inheriting = B_FALSE;
	char *inheritstr;
	char *recvdstr;

	ASSERT(dsl_pool_config_held(dd->dd_pool));

	if (setpoint)
	setpoint[0] = '\0';

	prop = zfs_name_to_prop(propname);
	inheritable = (prop == ZPROP_USERPROP \|\| zfs_prop_inheritable(prop));
	inheritstr = kmem_asprintf("%s%s", propname, ZPROP_INHERIT_SUFFIX);
	recvdstr = kmem_asprintf("%s%s", propname, ZPROP_RECVD_SUFFIX);

	/*
	* Note: dd may become NULL, therefore we shouldn't dereference it
	* after this loop.
	*/
	for (; dd != NULL; dd = dd->dd_parent) {
	if (dd != target \|\| snapshot) {
	if (!inheritable) {
	err = SET_ERROR(ENOENT);
	break;
	}
	inheriting = B_TRUE;
	}

	/* Check for a local value. */
	err = zap_lookup(mos, dsl_dir_phys(dd)->dd_props_zapobj,
	propname, intsz, numints, buf);
	if (err != ENOENT) {
	if (setpoint != NULL && err == 0)
	dsl_dir_name(dd, setpoint);
	break;
	}

	/*
	* Skip the check for a received value if there is an explicit
	* inheritance entry.
	*/
	err = zap_contains(mos, dsl_dir_phys(dd)->dd_props_zapobj,
	inheritstr);
	if (err != 0 && err != ENOENT)
	break;

	if (err == ENOENT) {
	/* Check for a received value. */
	err = zap_lookup(mos, dsl_dir_phys(dd)->dd_props_zapobj,
	recvdstr, intsz, numints, buf);
	if (err != ENOENT) {
	if (setpoint != NULL && err == 0) {
	if (inheriting) {
	dsl_dir_name(dd, setpoint);
	} else {
	(void) strlcpy(setpoint,
	ZPROP_SOURCE_VAL_RECVD,
	MAXNAMELEN);
	}
	}
	break;
	}
	}

	/*
	* If we found an explicit inheritance entry, err is zero even
	* though we haven't yet found the value, so reinitializing err
	* at the end of the loop (instead of at the beginning) ensures
	* that err has a valid post-loop value.
	*/
	err = SET_ERROR(ENOENT);
	}

	if (err == ENOENT)
	err = dodefault(prop, intsz, numints, buf);

	kmem_strfree(inheritstr);
	kmem_strfree(recvdstr);

	return (err);
	}

	int
	dsl_prop_get_ds(dsl_dataset_t ds, const char propname,
	int intsz, int numints, void buf, char setpoint)
	{
	zfs_prop_t prop = zfs_name_to_prop(propname);
	boolean_t inheritable;
	uint64_t zapobj;

	ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
	inheritable = (prop == ZPROP_USERPROP \|\| zfs_prop_inheritable(prop));
	zapobj = dsl_dataset_phys(ds)->ds_props_obj;

	if (zapobj != 0) {
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	int err;

	ASSERT(ds->ds_is_snapshot);

	/* Check for a local value. */
	err = zap_lookup(mos, zapobj, propname, intsz, numints, buf);
	if (err != ENOENT) {
	if (setpoint != NULL && err == 0)
	dsl_dataset_name(ds, setpoint);
	return (err);
	}

	/*
	* Skip the check for a received value if there is an explicit
	* inheritance entry.
	*/
	if (inheritable) {
	char *inheritstr = kmem_asprintf("%s%s", propname,
	ZPROP_INHERIT_SUFFIX);
	err = zap_contains(mos, zapobj, inheritstr);
	kmem_strfree(inheritstr);
	if (err != 0 && err != ENOENT)
	return (err);
	}

	if (err == ENOENT) {
	/* Check for a received value. */
	char *recvdstr = kmem_asprintf("%s%s", propname,
	ZPROP_RECVD_SUFFIX);
	err = zap_lookup(mos, zapobj, recvdstr,
	intsz, numints, buf);
	kmem_strfree(recvdstr);
	if (err != ENOENT) {
	if (setpoint != NULL && err == 0)
	(void) strlcpy(setpoint,
	ZPROP_SOURCE_VAL_RECVD,
	MAXNAMELEN);
	return (err);
	}
	}
	}

	return (dsl_prop_get_dd(ds->ds_dir, propname,
	intsz, numints, buf, setpoint, ds->ds_is_snapshot));
	}

	static dsl_prop_record_t *
	dsl_prop_record_find(dsl_dir_t dd, const char propname)
	{
	dsl_prop_record_t *pr = NULL;

	ASSERT(MUTEX_HELD(&dd->dd_lock));

	for (pr = list_head(&dd->dd_props);
	pr != NULL; pr = list_next(&dd->dd_props, pr)) {
	if (strcmp(pr->pr_propname, propname) == 0)
	break;
	}

	return (pr);
	}

	static dsl_prop_record_t *
	dsl_prop_record_create(dsl_dir_t dd, const char propname)
	{
	dsl_prop_record_t *pr;

	ASSERT(MUTEX_HELD(&dd->dd_lock));

	pr = kmem_alloc(sizeof (dsl_prop_record_t), KM_SLEEP);
	pr->pr_propname = spa_strdup(propname);
	list_create(&pr->pr_cbs, sizeof (dsl_prop_cb_record_t),
	offsetof(dsl_prop_cb_record_t, cbr_pr_node));
	list_insert_head(&dd->dd_props, pr);

	return (pr);
	}

	void
	dsl_prop_init(dsl_dir_t *dd)
	{
	list_create(&dd->dd_props, sizeof (dsl_prop_record_t),
	offsetof(dsl_prop_record_t, pr_node));
	}

	void
	dsl_prop_fini(dsl_dir_t *dd)
	{
	dsl_prop_record_t *pr;

	while ((pr = list_remove_head(&dd->dd_props)) != NULL) {
	list_destroy(&pr->pr_cbs);
	spa_strfree((char *)pr->pr_propname);
	kmem_free(pr, sizeof (dsl_prop_record_t));
	}
	list_destroy(&dd->dd_props);
	}

	/*
	* Register interest in the named property. We'll call the callback
	* once to notify it of the current property value, and again each time
	* the property changes, until this callback is unregistered.
	*
	* Return 0 on success, errno if the prop is not an integer value.
	*/
	int
	dsl_prop_register(dsl_dataset_t ds, const char propname,
	dsl_prop_changed_cb_t callback, void cbarg)
	{
	dsl_dir_t *dd = ds->ds_dir;
	uint64_t value;
	dsl_prop_record_t *pr;
	dsl_prop_cb_record_t *cbr;
	int err;
	dsl_pool_t *dp __maybe_unused = dd->dd_pool;

	ASSERT(dsl_pool_config_held(dp));

	err = dsl_prop_get_int_ds(ds, propname, &value);
	if (err != 0)
	return (err);

	cbr = kmem_alloc(sizeof (dsl_prop_cb_record_t), KM_SLEEP);
	cbr->cbr_ds = ds;
	cbr->cbr_func = callback;
	cbr->cbr_arg = cbarg;

	mutex_enter(&dd->dd_lock);
	pr = dsl_prop_record_find(dd, propname);
	if (pr == NULL)
	pr = dsl_prop_record_create(dd, propname);
	cbr->cbr_pr = pr;
	list_insert_head(&pr->pr_cbs, cbr);
	list_insert_head(&ds->ds_prop_cbs, cbr);
	mutex_exit(&dd->dd_lock);

	cbr->cbr_func(cbr->cbr_arg, value);
	return (0);
	}

	int
	dsl_prop_get(const char dsname, const char propname,
	int intsz, int numints, void buf, char setpoint)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold(dsname, FTAG, &os);
	if (error != 0)
	return (error);

	error = dsl_prop_get_ds(dmu_objset_ds(os), propname,
	intsz, numints, buf, setpoint);

	dmu_objset_rele(os, FTAG);
	return (error);
	}

	/*
	* Get the current property value. It may have changed by the time this
	* function returns, so it is NOT safe to follow up with
	* dsl_prop_register() and assume that the value has not changed in
	* between.
	*
	* Return 0 on success, ENOENT if ddname is invalid.
	*/
	int
	dsl_prop_get_integer(const char ddname, const char propname,
	uint64_t valuep, char setpoint)
	{
	return (dsl_prop_get(ddname, propname, 8, 1, valuep, setpoint));
	}

	int
	dsl_prop_get_int_ds(dsl_dataset_t ds, const char propname,
	uint64_t *valuep)
	{
	return (dsl_prop_get_ds(ds, propname, 8, 1, valuep, NULL));
	}

	/*
	* Predict the effective value of the given special property if it were set with
	* the given value and source. This is not a general purpose function. It exists
	* only to handle the special requirements of the quota and reservation
	* properties. The fact that these properties are non-inheritable greatly
	* simplifies the prediction logic.
	*
	* Returns 0 on success, a positive error code on failure, or -1 if called with
	* a property not handled by this function.
	*/
	int
	dsl_prop_predict(dsl_dir_t dd, const char propname,
	zprop_source_t source, uint64_t value, uint64_t *newvalp)
	{
	zfs_prop_t prop = zfs_name_to_prop(propname);
	objset_t *mos;
	uint64_t zapobj;
	uint64_t version;
	char *recvdstr;
	int err = 0;

	switch (prop) {
	case ZFS_PROP_QUOTA:
	case ZFS_PROP_RESERVATION:
	case ZFS_PROP_REFQUOTA:
	case ZFS_PROP_REFRESERVATION:
	break;
	default:
	return (-1);
	}

	mos = dd->dd_pool->dp_meta_objset;
	zapobj = dsl_dir_phys(dd)->dd_props_zapobj;
	recvdstr = kmem_asprintf("%s%s", propname, ZPROP_RECVD_SUFFIX);

	version = spa_version(dd->dd_pool->dp_spa);
	if (version < SPA_VERSION_RECVD_PROPS) {
	if (source & ZPROP_SRC_NONE)
	source = ZPROP_SRC_NONE;
	else if (source & ZPROP_SRC_RECEIVED)
	source = ZPROP_SRC_LOCAL;
	}

	switch ((int)source) {
	case ZPROP_SRC_NONE:
	/* Revert to the received value, if any. */
	err = zap_lookup(mos, zapobj, recvdstr, 8, 1, newvalp);
	if (err == ENOENT)
	*newvalp = 0;
	break;
	case ZPROP_SRC_LOCAL:
	*newvalp = value;
	break;
	case ZPROP_SRC_RECEIVED:
	/*
	* If there's no local setting, then the new received value will
	* be the effective value.
	*/
	err = zap_lookup(mos, zapobj, propname, 8, 1, newvalp);
	if (err == ENOENT)
	*newvalp = value;
	break;
	case (ZPROP_SRC_NONE \| ZPROP_SRC_RECEIVED):
	/*
	* We're clearing the received value, so the local setting (if
	* it exists) remains the effective value.
	*/
	err = zap_lookup(mos, zapobj, propname, 8, 1, newvalp);
	if (err == ENOENT)
	*newvalp = 0;
	break;
	default:
	panic("unexpected property source: %d", source);
	}

	kmem_strfree(recvdstr);

	if (err == ENOENT)
	return (0);

	return (err);
	}

	/*
	* Unregister this callback. Return 0 on success, ENOENT if ddname is
	* invalid, or ENOMSG if no matching callback registered.
	*
	* NOTE: This function is no longer used internally but has been preserved
	* to prevent breaking external consumers (Lustre, etc).
	*/
	int
	dsl_prop_unregister(dsl_dataset_t ds, const char propname,
	dsl_prop_changed_cb_t callback, void cbarg)
	{
	dsl_dir_t *dd = ds->ds_dir;
	dsl_prop_cb_record_t *cbr;

	mutex_enter(&dd->dd_lock);
	for (cbr = list_head(&ds->ds_prop_cbs);
	cbr; cbr = list_next(&ds->ds_prop_cbs, cbr)) {
	if (cbr->cbr_ds == ds &&
	cbr->cbr_func == callback &&
	cbr->cbr_arg == cbarg &&
	strcmp(cbr->cbr_pr->pr_propname, propname) == 0)
	break;
	}

	if (cbr == NULL) {
	mutex_exit(&dd->dd_lock);
	return (SET_ERROR(ENOMSG));
	}

	list_remove(&ds->ds_prop_cbs, cbr);
	list_remove(&cbr->cbr_pr->pr_cbs, cbr);
	mutex_exit(&dd->dd_lock);
	kmem_free(cbr, sizeof (dsl_prop_cb_record_t));

	return (0);
	}

	/*
	* Unregister all callbacks that are registered with the
	* given callback argument.
	*/
	void
	dsl_prop_unregister_all(dsl_dataset_t ds, void cbarg)
	{
	dsl_prop_cb_record_t cbr, next_cbr;

	dsl_dir_t *dd = ds->ds_dir;

	mutex_enter(&dd->dd_lock);
	next_cbr = list_head(&ds->ds_prop_cbs);
	while (next_cbr != NULL) {
	cbr = next_cbr;
	next_cbr = list_next(&ds->ds_prop_cbs, cbr);
	if (cbr->cbr_arg == cbarg) {
	list_remove(&ds->ds_prop_cbs, cbr);
	list_remove(&cbr->cbr_pr->pr_cbs, cbr);
	kmem_free(cbr, sizeof (dsl_prop_cb_record_t));
	}
	}
	mutex_exit(&dd->dd_lock);
	}

	boolean_t
	dsl_prop_hascb(dsl_dataset_t *ds)
	{
	return (!list_is_empty(&ds->ds_prop_cbs));
	}

	static int
	dsl_prop_notify_all_cb(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) arg;
	dsl_dir_t *dd = ds->ds_dir;
	dsl_prop_record_t *pr;
	dsl_prop_cb_record_t *cbr;

	mutex_enter(&dd->dd_lock);
	for (pr = list_head(&dd->dd_props);
	pr; pr = list_next(&dd->dd_props, pr)) {
	for (cbr = list_head(&pr->pr_cbs); cbr;
	cbr = list_next(&pr->pr_cbs, cbr)) {
	uint64_t value;

	/*
	* Callback entries do not have holds on their
	* datasets so that datasets with registered
	* callbacks are still eligible for eviction.
	* Unlike operations to update properties on a
	* single dataset, we are performing a recursive
	* descent of related head datasets. The caller
	* of this function only has a dataset hold on
	* the passed in head dataset, not the snapshots
	* associated with this dataset. Without a hold,
	* the dataset pointer within callback records
	* for snapshots can be invalidated by eviction
	* at any time.
	*
	* Use dsl_dataset_try_add_ref() to verify
	* that the dataset for a snapshot has not
	* begun eviction processing and to prevent
	* eviction from occurring for the duration of
	* the callback. If the hold attempt fails,
	* this object is already being evicted and the
	* callback can be safely ignored.
	*/
	if (ds != cbr->cbr_ds &&
	!dsl_dataset_try_add_ref(dp, cbr->cbr_ds, FTAG))
	continue;

	if (dsl_prop_get_ds(cbr->cbr_ds,
	cbr->cbr_pr->pr_propname, sizeof (value), 1,
	&value, NULL) == 0)
	cbr->cbr_func(cbr->cbr_arg, value);

	if (ds != cbr->cbr_ds)
	dsl_dataset_rele(cbr->cbr_ds, FTAG);
	}
	}
	mutex_exit(&dd->dd_lock);

	return (0);
	}

	/*
	* Update all property values for ddobj & its descendants. This is used
	* when renaming the dir.
	*/
	void
	dsl_prop_notify_all(dsl_dir_t *dd)
	{
	dsl_pool_t *dp = dd->dd_pool;
	ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
	(void) dmu_objset_find_dp(dp, dd->dd_object, dsl_prop_notify_all_cb,
	NULL, DS_FIND_CHILDREN);
	}

	static void
	dsl_prop_changed_notify(dsl_pool_t *dp, uint64_t ddobj,
	const char *propname, uint64_t value, int first)
	{
	dsl_dir_t *dd;
	dsl_prop_record_t *pr;
	dsl_prop_cb_record_t *cbr;
	objset_t *mos = dp->dp_meta_objset;
	zap_cursor_t zc;
	zap_attribute_t *za;
	int err;

	ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
	err = dsl_dir_hold_obj(dp, ddobj, NULL, FTAG, &dd);
	if (err)
	return;

	if (!first) {
	/*
	* If the prop is set here, then this change is not
	* being inherited here or below; stop the recursion.
	*/
	err = zap_contains(mos, dsl_dir_phys(dd)->dd_props_zapobj,
	propname);
	if (err == 0) {
	dsl_dir_rele(dd, FTAG);
	return;
	}
	ASSERT3U(err, ==, ENOENT);
	}

	mutex_enter(&dd->dd_lock);
	pr = dsl_prop_record_find(dd, propname);
	if (pr != NULL) {
	for (cbr = list_head(&pr->pr_cbs); cbr;
	cbr = list_next(&pr->pr_cbs, cbr)) {
	uint64_t propobj;

	/*
	* cbr->cbr_ds may be invalidated due to eviction,
	* requiring the use of dsl_dataset_try_add_ref().
	* See comment block in dsl_prop_notify_all_cb()
	* for details.
	*/
	if (!dsl_dataset_try_add_ref(dp, cbr->cbr_ds, FTAG))
	continue;

	propobj = dsl_dataset_phys(cbr->cbr_ds)->ds_props_obj;

	/*
	* If the property is not set on this ds, then it is
	* inherited here; call the callback.
	*/
	if (propobj == 0 \|\|
	zap_contains(mos, propobj, propname) != 0)
	cbr->cbr_func(cbr->cbr_arg, value);

	dsl_dataset_rele(cbr->cbr_ds, FTAG);
	}
	}
	mutex_exit(&dd->dd_lock);

	za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
	for (zap_cursor_init(&zc, mos,
	dsl_dir_phys(dd)->dd_child_dir_zapobj);
	zap_cursor_retrieve(&zc, za) == 0;
	zap_cursor_advance(&zc)) {
	dsl_prop_changed_notify(dp, za->za_first_integer,
	propname, value, FALSE);
	}
	kmem_free(za, sizeof (zap_attribute_t));
	zap_cursor_fini(&zc);
	dsl_dir_rele(dd, FTAG);
	}

	void
	dsl_prop_set_sync_impl(dsl_dataset_t ds, const char propname,
	zprop_source_t source, int intsz, int numints, const void *value,
	dmu_tx_t *tx)
	{
	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t zapobj, intval, dummy, count;
	int isint;
	char valbuf[32];
	const char *valstr = NULL;
	char *inheritstr;
	char *recvdstr;
	char *tbuf = NULL;
	int err;
	uint64_t version = spa_version(ds->ds_dir->dd_pool->dp_spa);

	isint = (dodefault(zfs_name_to_prop(propname), 8, 1, &intval) == 0);

	if (ds->ds_is_snapshot) {
	ASSERT(version >= SPA_VERSION_SNAP_PROPS);
	if (dsl_dataset_phys(ds)->ds_props_obj == 0 &&
	(source & ZPROP_SRC_NONE) == 0) {
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_props_obj =
	zap_create(mos,
	DMU_OT_DSL_PROPS, DMU_OT_NONE, 0, tx);
	}
	zapobj = dsl_dataset_phys(ds)->ds_props_obj;
	} else {
	zapobj = dsl_dir_phys(ds->ds_dir)->dd_props_zapobj;
	}

	/* If we are removing objects from a non-existent ZAP just return */
	if (zapobj == 0)
	return;

	if (version < SPA_VERSION_RECVD_PROPS) {
	if (source & ZPROP_SRC_NONE)
	source = ZPROP_SRC_NONE;
	else if (source & ZPROP_SRC_RECEIVED)
	source = ZPROP_SRC_LOCAL;
	}

	inheritstr = kmem_asprintf("%s%s", propname, ZPROP_INHERIT_SUFFIX);
	recvdstr = kmem_asprintf("%s%s", propname, ZPROP_RECVD_SUFFIX);

	switch ((int)source) {
	case ZPROP_SRC_NONE:
	/*
	* revert to received value, if any (inherit -S)
	* - remove propname
	* - remove propname$inherit
	*/
	err = zap_remove(mos, zapobj, propname, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	err = zap_remove(mos, zapobj, inheritstr, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	break;
	case ZPROP_SRC_LOCAL:
	/*
	* remove propname$inherit
	* set propname -> value
	*/
	err = zap_remove(mos, zapobj, inheritstr, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	VERIFY0(zap_update(mos, zapobj, propname,
	intsz, numints, value, tx));
	break;
	case ZPROP_SRC_INHERITED:
	/*
	* explicitly inherit
	* - remove propname
	* - set propname$inherit
	*/
	err = zap_remove(mos, zapobj, propname, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	if (version >= SPA_VERSION_RECVD_PROPS &&
	dsl_prop_get_int_ds(ds, ZPROP_HAS_RECVD, &dummy) == 0) {
	dummy = 0;
	VERIFY0(zap_update(mos, zapobj, inheritstr,
	8, 1, &dummy, tx));
	}
	break;
	case ZPROP_SRC_RECEIVED:
	/*
	* set propname$recvd -> value
	*/
	err = zap_update(mos, zapobj, recvdstr,
	intsz, numints, value, tx);
	ASSERT(err == 0);
	break;
	case (ZPROP_SRC_NONE \| ZPROP_SRC_LOCAL \| ZPROP_SRC_RECEIVED):
	/*
	* clear local and received settings
	* - remove propname
	* - remove propname$inherit
	* - remove propname$recvd
	*/
	err = zap_remove(mos, zapobj, propname, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	err = zap_remove(mos, zapobj, inheritstr, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	zfs_fallthrough;
	case (ZPROP_SRC_NONE \| ZPROP_SRC_RECEIVED):
	/*
	* remove propname$recvd
	*/
	err = zap_remove(mos, zapobj, recvdstr, tx);
	ASSERT(err == 0 \|\| err == ENOENT);
	break;
	default:
	cmn_err(CE_PANIC, "unexpected property source: %d", source);
	}

	kmem_strfree(inheritstr);
	kmem_strfree(recvdstr);

	/*
	* If we are left with an empty snap zap we can destroy it.
	* This will prevent unnecessary calls to zap_lookup() in
	* the "zfs list" and "zfs get" code paths.
	*/
	if (ds->ds_is_snapshot &&
	zap_count(mos, zapobj, &count) == 0 && count == 0) {
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_props_obj = 0;
	zap_destroy(mos, zapobj, tx);
	}

	if (isint) {
	VERIFY0(dsl_prop_get_int_ds(ds, propname, &intval));

	if (ds->ds_is_snapshot) {
	dsl_prop_cb_record_t *cbr;
	/*
	* It's a snapshot; nothing can inherit this
	* property, so just look for callbacks on this
	* ds here.
	*/
	mutex_enter(&ds->ds_dir->dd_lock);
	for (cbr = list_head(&ds->ds_prop_cbs); cbr;
	cbr = list_next(&ds->ds_prop_cbs, cbr)) {
	if (strcmp(cbr->cbr_pr->pr_propname,
	propname) == 0)
	cbr->cbr_func(cbr->cbr_arg, intval);
	}
	mutex_exit(&ds->ds_dir->dd_lock);
	} else {
	dsl_prop_changed_notify(ds->ds_dir->dd_pool,
	ds->ds_dir->dd_object, propname, intval, TRUE);
	}

	(void) snprintf(valbuf, sizeof (valbuf),
	"%lld", (longlong_t)intval);
	valstr = valbuf;
	} else {
	if (source == ZPROP_SRC_LOCAL) {
	valstr = value;
	} else {
	tbuf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
	if (dsl_prop_get_ds(ds, propname, 1,
	ZAP_MAXVALUELEN, tbuf, NULL) == 0)
	valstr = tbuf;
	}
	}

	spa_history_log_internal_ds(ds, (source == ZPROP_SRC_NONE \|\|
	source == ZPROP_SRC_INHERITED) ? "inherit" : "set", tx,
	"%s=%s", propname, (valstr == NULL ? "" : valstr));

	if (tbuf != NULL)
	kmem_free(tbuf, ZAP_MAXVALUELEN);
	}

	int
	dsl_prop_set_int(const char dsname, const char propname,
	zprop_source_t source, uint64_t value)
	{
	nvlist_t *nvl = fnvlist_alloc();
	int error;

	fnvlist_add_uint64(nvl, propname, value);
	error = dsl_props_set(dsname, source, nvl);
	fnvlist_free(nvl);
	return (error);
	}

	int
	dsl_prop_set_string(const char dsname, const char propname,
	zprop_source_t source, const char *value)
	{
	nvlist_t *nvl = fnvlist_alloc();
	int error;

	fnvlist_add_string(nvl, propname, value);
	error = dsl_props_set(dsname, source, nvl);
	fnvlist_free(nvl);
	return (error);
	}

	int
	dsl_prop_inherit(const char dsname, const char propname,
	zprop_source_t source)
	{
	nvlist_t *nvl = fnvlist_alloc();
	int error;

	fnvlist_add_boolean(nvl, propname);
	error = dsl_props_set(dsname, source, nvl);
	fnvlist_free(nvl);
	return (error);
	}

	int
	dsl_props_set_check(void arg, dmu_tx_t tx)
	{
	dsl_props_set_arg_t *dpsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;
	uint64_t version;
	nvpair_t *elem = NULL;
	int err;

	err = dsl_dataset_hold(dp, dpsa->dpsa_dsname, FTAG, &ds);
	if (err != 0)
	return (err);

	version = spa_version(ds->ds_dir->dd_pool->dp_spa);
	while ((elem = nvlist_next_nvpair(dpsa->dpsa_props, elem)) != NULL) {
	if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENAMETOOLONG));
	}
	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	char *valstr = fnvpair_value_string(elem);
	if (strlen(valstr) >= (version <
	SPA_VERSION_STMF_PROP ?
	ZAP_OLDMAXVALUELEN : ZAP_MAXVALUELEN)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(E2BIG));
	}
	}
	}

	if (ds->ds_is_snapshot && version < SPA_VERSION_SNAP_PROPS) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	void
	dsl_props_set_sync_impl(dsl_dataset_t *ds, zprop_source_t source,
	nvlist_t props, dmu_tx_t tx)
	{
	nvpair_t *elem = NULL;

	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	nvpair_t *pair = elem;
	const char *name = nvpair_name(pair);

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	/*
	* This usually happens when we reuse the nvlist_t data
	* returned by the counterpart dsl_prop_get_all_impl().
	* For instance we do this to restore the original
	* received properties when an error occurs in the
	* zfs_ioc_recv() codepath.
	*/
	nvlist_t *attrs = fnvpair_value_nvlist(pair);
	pair = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE);
	}

	if (nvpair_type(pair) == DATA_TYPE_STRING) {
	const char *value = fnvpair_value_string(pair);
	dsl_prop_set_sync_impl(ds, name,
	source, 1, strlen(value) + 1, value, tx);
	} else if (nvpair_type(pair) == DATA_TYPE_UINT64) {
	uint64_t intval = fnvpair_value_uint64(pair);
	dsl_prop_set_sync_impl(ds, name,
	source, sizeof (intval), 1, &intval, tx);
	} else if (nvpair_type(pair) == DATA_TYPE_BOOLEAN) {
	dsl_prop_set_sync_impl(ds, name,
	source, 0, 0, NULL, tx);
	} else {
	panic("invalid nvpair type");
	}
	}
	}

	void
	dsl_props_set_sync(void arg, dmu_tx_t tx)
	{
	dsl_props_set_arg_t *dpsa = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dataset_t *ds;

	VERIFY0(dsl_dataset_hold(dp, dpsa->dpsa_dsname, FTAG, &ds));
	dsl_props_set_sync_impl(ds, dpsa->dpsa_source, dpsa->dpsa_props, tx);
	dsl_dataset_rele(ds, FTAG);
	}

	/*
	* All-or-nothing; if any prop can't be set, nothing will be modified.
	*/
	int
	dsl_props_set(const char dsname, zprop_source_t source, nvlist_t props)
	{
	dsl_props_set_arg_t dpsa;
	int nblks = 0;

	dpsa.dpsa_dsname = dsname;
	dpsa.dpsa_source = source;
	dpsa.dpsa_props = props;

	/*
	* If the source includes NONE, then we will only be removing entries
	* from the ZAP object. In that case don't check for ENOSPC.
	*/
	if ((source & ZPROP_SRC_NONE) == 0)
	nblks = 2 * fnvlist_num_pairs(props);

	return (dsl_sync_task(dsname, dsl_props_set_check, dsl_props_set_sync,
	&dpsa, nblks, ZFS_SPACE_CHECK_RESERVED));
	}

	typedef enum dsl_prop_getflags {
	DSL_PROP_GET_INHERITING = 0x1, /* searching parent of target ds */
	DSL_PROP_GET_SNAPSHOT = 0x2, /* snapshot dataset */
	DSL_PROP_GET_LOCAL = 0x4, /* local properties */
	DSL_PROP_GET_RECEIVED = 0x8, /* received properties */
	} dsl_prop_getflags_t;

	static int
	dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj,
	const char setpoint, dsl_prop_getflags_t flags, nvlist_t nv)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	int err = 0;

	for (zap_cursor_init(&zc, mos, propobj);
	(err = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	nvlist_t *propval;
	zfs_prop_t prop;
	char buf[ZAP_MAXNAMELEN];
	char *valstr;
	const char *suffix;
	const char *propname;
	const char *source;

	suffix = strchr(za.za_name, '$');

	if (suffix == NULL) {
	/*
	* Skip local properties if we only want received
	* properties.
	*/
	if (flags & DSL_PROP_GET_RECEIVED)
	continue;

	propname = za.za_name;
	source = setpoint;
	} else if (strcmp(suffix, ZPROP_INHERIT_SUFFIX) == 0) {
	/* Skip explicitly inherited entries. */
	continue;
	} else if (strcmp(suffix, ZPROP_RECVD_SUFFIX) == 0) {
	if (flags & DSL_PROP_GET_LOCAL)
	continue;

	- (void) strncpy(buf, za.za_name, (suffix - za.za_name));
	- buf[suffix - za.za_name] = '\0';
	+ (void) strlcpy(buf, za.za_name,
	+ MIN(sizeof (buf), suffix - za.za_name + 1));
	propname = buf;

	if (!(flags & DSL_PROP_GET_RECEIVED)) {
	/* Skip if locally overridden. */
	err = zap_contains(mos, propobj, propname);
	if (err == 0)
	continue;
	if (err != ENOENT)
	break;

	/* Skip if explicitly inherited. */
	valstr = kmem_asprintf("%s%s", propname,
	ZPROP_INHERIT_SUFFIX);
	err = zap_contains(mos, propobj, valstr);
	kmem_strfree(valstr);
	if (err == 0)
	continue;
	if (err != ENOENT)
	break;
	}

	source = ((flags & DSL_PROP_GET_INHERITING) ?
	setpoint : ZPROP_SOURCE_VAL_RECVD);
	} else {
	/*
	* For backward compatibility, skip suffixes we don't
	* recognize.
	*/
	continue;
	}

	prop = zfs_name_to_prop(propname);

	/* Skip non-inheritable properties. */
	if ((flags & DSL_PROP_GET_INHERITING) &&
	prop != ZPROP_USERPROP && !zfs_prop_inheritable(prop))
	continue;

	/* Skip properties not valid for this type. */
	if ((flags & DSL_PROP_GET_SNAPSHOT) && prop != ZPROP_USERPROP &&
	!zfs_prop_valid_for_type(prop, ZFS_TYPE_SNAPSHOT, B_FALSE))
	continue;

	/* Skip properties already defined. */
	if (nvlist_exists(nv, propname))
	continue;

	VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	if (za.za_integer_length == 1) {
	/*
	* String property
	*/
	char *tmp = kmem_alloc(za.za_num_integers,
	KM_SLEEP);
	err = zap_lookup(mos, propobj,
	za.za_name, 1, za.za_num_integers, tmp);
	if (err != 0) {
	kmem_free(tmp, za.za_num_integers);
	break;
	}
	VERIFY(nvlist_add_string(propval, ZPROP_VALUE,
	tmp) == 0);
	kmem_free(tmp, za.za_num_integers);
	} else {
	/*
	* Integer property
	*/
	ASSERT(za.za_integer_length == 8);
	(void) nvlist_add_uint64(propval, ZPROP_VALUE,
	za.za_first_integer);
	}

	VERIFY(nvlist_add_string(propval, ZPROP_SOURCE, source) == 0);
	VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0);
	nvlist_free(propval);
	}
	zap_cursor_fini(&zc);
	if (err == ENOENT)
	err = 0;
	return (err);
	}

	/*
	* Iterate over all properties for this dataset and return them in an nvlist.
	*/
	static int
	dsl_prop_get_all_ds(dsl_dataset_t ds, nvlist_t *nvp,
	dsl_prop_getflags_t flags)
	{
	dsl_dir_t *dd = ds->ds_dir;
	dsl_pool_t *dp = dd->dd_pool;
	objset_t *mos = dp->dp_meta_objset;
	int err = 0;
	char setpoint[ZFS_MAX_DATASET_NAME_LEN];

	VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	if (ds->ds_is_snapshot)
	flags \|= DSL_PROP_GET_SNAPSHOT;

	ASSERT(dsl_pool_config_held(dp));

	if (dsl_dataset_phys(ds)->ds_props_obj != 0) {
	ASSERT(flags & DSL_PROP_GET_SNAPSHOT);
	dsl_dataset_name(ds, setpoint);
	err = dsl_prop_get_all_impl(mos,
	dsl_dataset_phys(ds)->ds_props_obj, setpoint, flags, *nvp);
	if (err)
	goto out;
	}

	for (; dd != NULL; dd = dd->dd_parent) {
	if (dd != ds->ds_dir \|\| (flags & DSL_PROP_GET_SNAPSHOT)) {
	if (flags & (DSL_PROP_GET_LOCAL \|
	DSL_PROP_GET_RECEIVED))
	break;
	flags \|= DSL_PROP_GET_INHERITING;
	}
	dsl_dir_name(dd, setpoint);
	err = dsl_prop_get_all_impl(mos,
	dsl_dir_phys(dd)->dd_props_zapobj, setpoint, flags, *nvp);
	if (err)
	break;
	}

	out:
	if (err) {
	nvlist_free(*nvp);
	*nvp = NULL;
	}
	return (err);
	}

	boolean_t
	dsl_prop_get_hasrecvd(const char *dsname)
	{
	uint64_t dummy;

	return (0 ==
	dsl_prop_get_integer(dsname, ZPROP_HAS_RECVD, &dummy, NULL));
	}

	static int
	dsl_prop_set_hasrecvd_impl(const char *dsname, zprop_source_t source)
	{
	uint64_t version;
	spa_t *spa;
	int error = 0;

	VERIFY0(spa_open(dsname, &spa, FTAG));
	version = spa_version(spa);
	spa_close(spa, FTAG);

	if (version >= SPA_VERSION_RECVD_PROPS)
	error = dsl_prop_set_int(dsname, ZPROP_HAS_RECVD, source, 0);
	return (error);
	}

	/*
	* Call after successfully receiving properties to ensure that only the first
	* receive on or after SPA_VERSION_RECVD_PROPS blows away local properties.
	*/
	int
	dsl_prop_set_hasrecvd(const char *dsname)
	{
	int error = 0;
	if (!dsl_prop_get_hasrecvd(dsname))
	error = dsl_prop_set_hasrecvd_impl(dsname, ZPROP_SRC_LOCAL);
	return (error);
	}

	void
	dsl_prop_unset_hasrecvd(const char *dsname)
	{
	VERIFY0(dsl_prop_set_hasrecvd_impl(dsname, ZPROP_SRC_NONE));
	}

	int
	dsl_prop_get_all(objset_t os, nvlist_t *nvp)
	{
	return (dsl_prop_get_all_ds(os->os_dsl_dataset, nvp, 0));
	}

	int
	dsl_prop_get_received(const char dsname, nvlist_t *nvp)
	{
	objset_t *os;
	int error;

	/*
	* Received properties are not distinguishable from local properties
	* until the dataset has received properties on or after
	* SPA_VERSION_RECVD_PROPS.
	*/
	dsl_prop_getflags_t flags = (dsl_prop_get_hasrecvd(dsname) ?
	DSL_PROP_GET_RECEIVED : DSL_PROP_GET_LOCAL);

	error = dmu_objset_hold(dsname, FTAG, &os);
	if (error != 0)
	return (error);
	error = dsl_prop_get_all_ds(os->os_dsl_dataset, nvp, flags);
	dmu_objset_rele(os, FTAG);
	return (error);
	}

	void
	dsl_prop_nvlist_add_uint64(nvlist_t *nv, zfs_prop_t prop, uint64_t value)
	{
	nvlist_t *propval;
	const char *propname = zfs_prop_to_name(prop);
	uint64_t default_value;

	if (nvlist_lookup_nvlist(nv, propname, &propval) == 0) {
	VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, value) == 0);
	return;
	}

	VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	VERIFY(nvlist_add_uint64(propval, ZPROP_VALUE, value) == 0);
	/* Indicate the default source if we can. */
	if (dodefault(prop, 8, 1, &default_value) == 0 &&
	value == default_value) {
	VERIFY(nvlist_add_string(propval, ZPROP_SOURCE, "") == 0);
	}
	VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0);
	nvlist_free(propval);
	}

	void
	dsl_prop_nvlist_add_string(nvlist_t nv, zfs_prop_t prop, const char value)
	{
	nvlist_t *propval;
	const char *propname = zfs_prop_to_name(prop);

	if (nvlist_lookup_nvlist(nv, propname, &propval) == 0) {
	VERIFY(nvlist_add_string(propval, ZPROP_VALUE, value) == 0);
	return;
	}

	VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	VERIFY(nvlist_add_string(propval, ZPROP_VALUE, value) == 0);
	VERIFY(nvlist_add_nvlist(nv, propname, propval) == 0);
	nvlist_free(propval);
	}

	#if defined(_KERNEL)
	EXPORT_SYMBOL(dsl_prop_register);
	EXPORT_SYMBOL(dsl_prop_unregister);
	EXPORT_SYMBOL(dsl_prop_unregister_all);
	EXPORT_SYMBOL(dsl_prop_get);
	EXPORT_SYMBOL(dsl_prop_get_integer);
	EXPORT_SYMBOL(dsl_prop_get_all);
	EXPORT_SYMBOL(dsl_prop_get_received);
	EXPORT_SYMBOL(dsl_prop_get_ds);
	EXPORT_SYMBOL(dsl_prop_get_int_ds);
	EXPORT_SYMBOL(dsl_prop_get_dd);
	EXPORT_SYMBOL(dsl_props_set);
	EXPORT_SYMBOL(dsl_prop_set_int);
	EXPORT_SYMBOL(dsl_prop_set_string);
	EXPORT_SYMBOL(dsl_prop_inherit);
	EXPORT_SYMBOL(dsl_prop_predict);
	EXPORT_SYMBOL(dsl_prop_nvlist_add_uint64);
	EXPORT_SYMBOL(dsl_prop_nvlist_add_string);
	#endif
	diff --git a/sys/contrib/openzfs/module/zfs/dsl_scan.c b/sys/contrib/openzfs/module/zfs/dsl_scan.c
	index 28afc3dead7e..f0cd1feaf55b 100644
	--- a/sys/contrib/openzfs/module/zfs/dsl_scan.c
	+++ b/sys/contrib/openzfs/module/zfs/dsl_scan.c
	@@ -1,4494 +1,4510 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
	* Copyright 2016 Gary Mills
	* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
	* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
	* Copyright 2019 Joyent, Inc.
	*/

	#include <sys/dsl_scan.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dnode.h>
	#include <sys/dmu_tx.h>
	#include <sys/dmu_objset.h>
	#include <sys/arc.h>
	#include <sys/zap.h>
	#include <sys/zio.h>
	#include <sys/zfs_context.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_znode.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/zil_impl.h>
	#include <sys/zio_checksum.h>
	#include <sys/ddt.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/zfeature.h>
	#include <sys/abd.h>
	#include <sys/range_tree.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#endif

	/*
	* Grand theory statement on scan queue sorting
	*
	* Scanning is implemented by recursively traversing all indirection levels
	* in an object and reading all blocks referenced from said objects. This
	* results in us approximately traversing the object from lowest logical
	* offset to the highest. For best performance, we would want the logical
	* blocks to be physically contiguous. However, this is frequently not the
	* case with pools given the allocation patterns of copy-on-write filesystems.
	* So instead, we put the I/Os into a reordering queue and issue them in a
	* way that will most benefit physical disks (LBA-order).
	*
	* Queue management:
	*
	* Ideally, we would want to scan all metadata and queue up all block I/O
	* prior to starting to issue it, because that allows us to do an optimal
	* sorting job. This can however consume large amounts of memory. Therefore
	* we continuously monitor the size of the queues and constrain them to 5%
	* (zfs_scan_mem_lim_fact) of physmem. If the queues grow larger than this
	* limit, we clear out a few of the largest extents at the head of the queues
	* to make room for more scanning. Hopefully, these extents will be fairly
	* large and contiguous, allowing us to approach sequential I/O throughput
	* even without a fully sorted tree.
	*
	* Metadata scanning takes place in dsl_scan_visit(), which is called from
	* dsl_scan_sync() every spa_sync(). If we have either fully scanned all
	* metadata on the pool, or we need to make room in memory because our
	* queues are too large, dsl_scan_visit() is postponed and
	* scan_io_queues_run() is called from dsl_scan_sync() instead. This implies
	* that metadata scanning and queued I/O issuing are mutually exclusive. This
	* allows us to provide maximum sequential I/O throughput for the majority of
	* I/O's issued since sequential I/O performance is significantly negatively
	* impacted if it is interleaved with random I/O.
	*
	* Implementation Notes
	*
	* One side effect of the queued scanning algorithm is that the scanning code
	* needs to be notified whenever a block is freed. This is needed to allow
	* the scanning code to remove these I/Os from the issuing queue. Additionally,
	* we do not attempt to queue gang blocks to be issued sequentially since this
	* is very hard to do and would have an extremely limited performance benefit.
	* Instead, we simply issue gang I/Os as soon as we find them using the legacy
	* algorithm.
	*
	* Backwards compatibility
	*
	* This new algorithm is backwards compatible with the legacy on-disk data
	* structures (and therefore does not require a new feature flag).
	* Periodically during scanning (see zfs_scan_checkpoint_intval), the scan
	* will stop scanning metadata (in logical order) and wait for all outstanding
	* sorted I/O to complete. Once this is done, we write out a checkpoint
	* bookmark, indicating that we have scanned everything logically before it.
	* If the pool is imported on a machine without the new sorting algorithm,
	* the scan simply resumes from the last checkpoint using the legacy algorithm.
	*/

	typedef int (scan_cb_t)(dsl_pool_t , const blkptr_t ,
	const zbookmark_phys_t *);

	static scan_cb_t dsl_scan_scrub_cb;

	static int scan_ds_queue_compare(const void a, const void b);
	static int scan_prefetch_queue_compare(const void a, const void b);
	static void scan_ds_queue_clear(dsl_scan_t *scn);
	static void scan_ds_prefetch_queue_clear(dsl_scan_t *scn);
	static boolean_t scan_ds_queue_contains(dsl_scan_t *scn, uint64_t dsobj,
	uint64_t *txg);
	static void scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg);
	static void scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj);
	static void scan_ds_queue_sync(dsl_scan_t scn, dmu_tx_t tx);
	static uint64_t dsl_scan_count_data_disks(vdev_t *vd);

	-extern int zfs_vdev_async_write_active_min_dirty_percent;
	+extern uint_t zfs_vdev_async_write_active_min_dirty_percent;
	static int zfs_scan_blkstats = 0;

	/*
	* By default zfs will check to ensure it is not over the hard memory
	* limit before each txg. If finer-grained control of this is needed
	* this value can be set to 1 to enable checking before scanning each
	* block.
	*/
	static int zfs_scan_strict_mem_lim = B_FALSE;

	/*
	* Maximum number of parallelly executed bytes per leaf vdev. We attempt
	* to strike a balance here between keeping the vdev queues full of I/Os
	* at all times and not overflowing the queues to cause long latency,
	* which would cause long txg sync times. No matter what, we will not
	* overload the drives with I/O, since that is protected by
	* zfs_vdev_scrub_max_active.
	*/
	static unsigned long zfs_scan_vdev_limit = 4 << 20;

	-static int zfs_scan_issue_strategy = 0;
	-static int zfs_scan_legacy = B_FALSE; /* don't queue & sort zios, go direct */
	+static uint_t zfs_scan_issue_strategy = 0;
	+
	+/* don't queue & sort zios, go direct */
	+static int zfs_scan_legacy = B_FALSE;
	static unsigned long zfs_scan_max_ext_gap = 2 << 20; /* in bytes */

	/*
	* fill_weight is non-tunable at runtime, so we copy it at module init from
	* zfs_scan_fill_weight. Runtime adjustments to zfs_scan_fill_weight would
	* break queue sorting.
	*/
	-static int zfs_scan_fill_weight = 3;
	+static uint_t zfs_scan_fill_weight = 3;
	static uint64_t fill_weight;

	/* See dsl_scan_should_clear() for details on the memory limit tunables */
	static const uint64_t zfs_scan_mem_lim_min = 16 << 20; /* bytes */
	static const uint64_t zfs_scan_mem_lim_soft_max = 128 << 20; /* bytes */
	-static int zfs_scan_mem_lim_fact = 20; /* fraction of physmem */
	-static int zfs_scan_mem_lim_soft_fact = 20; /* fraction of mem lim above */
	-
	-static int zfs_scrub_min_time_ms = 1000; /* min millis to scrub per txg */
	-static int zfs_obsolete_min_time_ms = 500; /* min millis to obsolete per txg */
	-static int zfs_free_min_time_ms = 1000; /* min millis to free per txg */
	-static int zfs_resilver_min_time_ms = 3000; /* min millis to resilver per txg */
	-static int zfs_scan_checkpoint_intval = 7200; /* in seconds */
	+
	+
	+/* fraction of physmem */
	+static uint_t zfs_scan_mem_lim_fact = 20;
	+
	+/* fraction of mem lim above */
	+static uint_t zfs_scan_mem_lim_soft_fact = 20;
	+
	+/* minimum milliseconds to scrub per txg */
	+static uint_t zfs_scrub_min_time_ms = 1000;
	+
	+/* minimum milliseconds to obsolete per txg */
	+static uint_t zfs_obsolete_min_time_ms = 500;
	+
	+/* minimum milliseconds to free per txg */
	+static uint_t zfs_free_min_time_ms = 1000;
	+
	+/* minimum milliseconds to resilver per txg */
	+static uint_t zfs_resilver_min_time_ms = 3000;
	+
	+static uint_t zfs_scan_checkpoint_intval = 7200; /* in seconds */
	int zfs_scan_suspend_progress = 0; /* set to prevent scans from progressing */
	static int zfs_no_scrub_io = B_FALSE; /* set to disable scrub i/o */
	static int zfs_no_scrub_prefetch = B_FALSE; /* set to disable scrub prefetch */
	static const enum ddt_class zfs_scrub_ddt_class_max = DDT_CLASS_DUPLICATE;
	/* max number of blocks to free in a single TXG */
	static unsigned long zfs_async_block_max_blocks = ULONG_MAX;
	/* max number of dedup blocks to free in a single TXG */
	static unsigned long zfs_max_async_dedup_frees = 100000;

	/* set to disable resilver deferring */
	static int zfs_resilver_disable_defer = B_FALSE;

	/*
	* We wait a few txgs after importing a pool to begin scanning so that
	* the import / mounting code isn't held up by scrub / resilver IO.
	* Unfortunately, it is a bit difficult to determine exactly how long
	* this will take since userspace will trigger fs mounts asynchronously
	* and the kernel will create zvol minors asynchronously. As a result,
	* the value provided here is a bit arbitrary, but represents a
	* reasonable estimate of how many txgs it will take to finish fully
	* importing a pool
	*/
	#define SCAN_IMPORT_WAIT_TXGS 5

	#define DSL_SCAN_IS_SCRUB_RESILVER(scn) \
	((scn)->scn_phys.scn_func == POOL_SCAN_SCRUB \|\| \
	(scn)->scn_phys.scn_func == POOL_SCAN_RESILVER)

	/*
	* Enable/disable the processing of the free_bpobj object.
	*/
	static int zfs_free_bpobj_enabled = 1;

	/* the order has to match pool_scan_type */
	static scan_cb_t *scan_funcs[POOL_SCAN_FUNCS] = {
	NULL,
	dsl_scan_scrub_cb, /* POOL_SCAN_SCRUB */
	dsl_scan_scrub_cb, /* POOL_SCAN_RESILVER */
	};

	/* In core node for the scn->scn_queue. Represents a dataset to be scanned */
	typedef struct {
	uint64_t sds_dsobj;
	uint64_t sds_txg;
	avl_node_t sds_node;
	} scan_ds_t;

	/*
	* This controls what conditions are placed on dsl_scan_sync_state():
	* SYNC_OPTIONAL) write out scn_phys iff scn_queues_pending == 0
	* SYNC_MANDATORY) write out scn_phys always. scn_queues_pending must be 0.
	* SYNC_CACHED) if scn_queues_pending == 0, write out scn_phys. Otherwise
	* write out the scn_phys_cached version.
	* See dsl_scan_sync_state for details.
	*/
	typedef enum {
	SYNC_OPTIONAL,
	SYNC_MANDATORY,
	SYNC_CACHED
	} state_sync_type_t;

	/*
	* This struct represents the minimum information needed to reconstruct a
	* zio for sequential scanning. This is useful because many of these will
	* accumulate in the sequential IO queues before being issued, so saving
	* memory matters here.
	*/
	typedef struct scan_io {
	/* fields from blkptr_t */
	uint64_t sio_blk_prop;
	uint64_t sio_phys_birth;
	uint64_t sio_birth;
	zio_cksum_t sio_cksum;
	uint32_t sio_nr_dvas;

	/* fields from zio_t */
	uint32_t sio_flags;
	zbookmark_phys_t sio_zb;

	/* members for queue sorting */
	union {
	avl_node_t sio_addr_node; /* link into issuing queue */
	list_node_t sio_list_node; /* link for issuing to disk */
	} sio_nodes;

	/*
	* There may be up to SPA_DVAS_PER_BP DVAs here from the bp,
	* depending on how many were in the original bp. Only the
	* first DVA is really used for sorting and issuing purposes.
	* The other DVAs (if provided) simply exist so that the zio
	* layer can find additional copies to repair from in the
	* event of an error. This array must go at the end of the
	* struct to allow this for the variable number of elements.
	*/
	dva_t sio_dva[0];
	} scan_io_t;

	#define SIO_SET_OFFSET(sio, x) DVA_SET_OFFSET(&(sio)->sio_dva[0], x)
	#define SIO_SET_ASIZE(sio, x) DVA_SET_ASIZE(&(sio)->sio_dva[0], x)
	#define SIO_GET_OFFSET(sio) DVA_GET_OFFSET(&(sio)->sio_dva[0])
	#define SIO_GET_ASIZE(sio) DVA_GET_ASIZE(&(sio)->sio_dva[0])
	#define SIO_GET_END_OFFSET(sio) \
	(SIO_GET_OFFSET(sio) + SIO_GET_ASIZE(sio))
	#define SIO_GET_MUSED(sio) \
	(sizeof (scan_io_t) + ((sio)->sio_nr_dvas * sizeof (dva_t)))

	struct dsl_scan_io_queue {
	dsl_scan_t q_scn; / associated dsl_scan_t */
	vdev_t q_vd; / top-level vdev that this queue represents */
	zio_t q_zio; / scn_zio_root child for waiting on IO */

	/* trees used for sorting I/Os and extents of I/Os */
	range_tree_t *q_exts_by_addr;
	zfs_btree_t q_exts_by_size;
	avl_tree_t q_sios_by_addr;
	uint64_t q_sio_memused;
	uint64_t q_last_ext_addr;

	/* members for zio rate limiting */
	uint64_t q_maxinflight_bytes;
	uint64_t q_inflight_bytes;
	kcondvar_t q_zio_cv; /* used under vd->vdev_scan_io_queue_lock */

	/* per txg statistics */
	uint64_t q_total_seg_size_this_txg;
	uint64_t q_segs_this_txg;
	uint64_t q_total_zio_size_this_txg;
	uint64_t q_zios_this_txg;
	};

	/* private data for dsl_scan_prefetch_cb() */
	typedef struct scan_prefetch_ctx {
	zfs_refcount_t spc_refcnt; /* refcount for memory management */
	dsl_scan_t spc_scn; / dsl_scan_t for the pool */
	boolean_t spc_root; /* is this prefetch for an objset? */
	uint8_t spc_indblkshift; /* dn_indblkshift of current dnode */
	uint16_t spc_datablkszsec; /* dn_idatablkszsec of current dnode */
	} scan_prefetch_ctx_t;

	/* private data for dsl_scan_prefetch() */
	typedef struct scan_prefetch_issue_ctx {
	avl_node_t spic_avl_node; /* link into scn->scn_prefetch_queue */
	scan_prefetch_ctx_t spic_spc; / spc for the callback */
	blkptr_t spic_bp; /* bp to prefetch */
	zbookmark_phys_t spic_zb; /* bookmark to prefetch */
	} scan_prefetch_issue_ctx_t;

	static void scan_exec_io(dsl_pool_t dp, const blkptr_t bp, int zio_flags,
	const zbookmark_phys_t zb, dsl_scan_io_queue_t queue);
	static void scan_io_queue_insert_impl(dsl_scan_io_queue_t *queue,
	scan_io_t *sio);

	static dsl_scan_io_queue_t scan_io_queue_create(vdev_t vd);
	static void scan_io_queues_destroy(dsl_scan_t *scn);

	static kmem_cache_t *sio_cache[SPA_DVAS_PER_BP];

	/* sio->sio_nr_dvas must be set so we know which cache to free from */
	static void
	sio_free(scan_io_t *sio)
	{
	ASSERT3U(sio->sio_nr_dvas, >, 0);
	ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);

	kmem_cache_free(sio_cache[sio->sio_nr_dvas - 1], sio);
	}

	/* It is up to the caller to set sio->sio_nr_dvas for freeing */
	static scan_io_t *
	sio_alloc(unsigned short nr_dvas)
	{
	ASSERT3U(nr_dvas, >, 0);
	ASSERT3U(nr_dvas, <=, SPA_DVAS_PER_BP);

	return (kmem_cache_alloc(sio_cache[nr_dvas - 1], KM_SLEEP));
	}

	void
	scan_init(void)
	{
	/*
	* This is used in ext_size_compare() to weight segments
	* based on how sparse they are. This cannot be changed
	* mid-scan and the tree comparison functions don't currently
	* have a mechanism for passing additional context to the
	* compare functions. Thus we store this value globally and
	* we only allow it to be set at module initialization time
	*/
	fill_weight = zfs_scan_fill_weight;

	for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
	char name[36];

	(void) snprintf(name, sizeof (name), "sio_cache_%d", i);
	sio_cache[i] = kmem_cache_create(name,
	(sizeof (scan_io_t) + ((i + 1) * sizeof (dva_t))),
	0, NULL, NULL, NULL, NULL, NULL, 0);
	}
	}

	void
	scan_fini(void)
	{
	for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
	kmem_cache_destroy(sio_cache[i]);
	}
	}

	static inline boolean_t
	dsl_scan_is_running(const dsl_scan_t *scn)
	{
	return (scn->scn_phys.scn_state == DSS_SCANNING);
	}

	boolean_t
	dsl_scan_resilvering(dsl_pool_t *dp)
	{
	return (dsl_scan_is_running(dp->dp_scan) &&
	dp->dp_scan->scn_phys.scn_func == POOL_SCAN_RESILVER);
	}

	static inline void
	sio2bp(const scan_io_t sio, blkptr_t bp)
	{
	memset(bp, 0, sizeof (*bp));
	bp->blk_prop = sio->sio_blk_prop;
	bp->blk_phys_birth = sio->sio_phys_birth;
	bp->blk_birth = sio->sio_birth;
	bp->blk_fill = 1; /* we always only work with data pointers */
	bp->blk_cksum = sio->sio_cksum;

	ASSERT3U(sio->sio_nr_dvas, >, 0);
	ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);

	memcpy(bp->blk_dva, sio->sio_dva, sio->sio_nr_dvas * sizeof (dva_t));
	}

	static inline void
	bp2sio(const blkptr_t bp, scan_io_t sio, int dva_i)
	{
	sio->sio_blk_prop = bp->blk_prop;
	sio->sio_phys_birth = bp->blk_phys_birth;
	sio->sio_birth = bp->blk_birth;
	sio->sio_cksum = bp->blk_cksum;
	sio->sio_nr_dvas = BP_GET_NDVAS(bp);

	/*
	* Copy the DVAs to the sio. We need all copies of the block so
	* that the self healing code can use the alternate copies if the
	* first is corrupted. We want the DVA at index dva_i to be first
	* in the sio since this is the primary one that we want to issue.
	*/
	for (int i = 0, j = dva_i; i < sio->sio_nr_dvas; i++, j++) {
	sio->sio_dva[i] = bp->blk_dva[j % sio->sio_nr_dvas];
	}
	}

	int
	dsl_scan_init(dsl_pool_t *dp, uint64_t txg)
	{
	int err;
	dsl_scan_t *scn;
	spa_t *spa = dp->dp_spa;
	uint64_t f;

	scn = dp->dp_scan = kmem_zalloc(sizeof (dsl_scan_t), KM_SLEEP);
	scn->scn_dp = dp;

	/*
	* It's possible that we're resuming a scan after a reboot so
	* make sure that the scan_async_destroying flag is initialized
	* appropriately.
	*/
	ASSERT(!scn->scn_async_destroying);
	scn->scn_async_destroying = spa_feature_is_active(dp->dp_spa,
	SPA_FEATURE_ASYNC_DESTROY);

	/*
	* Calculate the max number of in-flight bytes for pool-wide
	* scanning operations (minimum 1MB). Limits for the issuing
	* phase are done per top-level vdev and are handled separately.
	*/
	scn->scn_maxinflight_bytes = MAX(zfs_scan_vdev_limit *
	dsl_scan_count_data_disks(spa->spa_root_vdev), 1ULL << 20);

	avl_create(&scn->scn_queue, scan_ds_queue_compare, sizeof (scan_ds_t),
	offsetof(scan_ds_t, sds_node));
	avl_create(&scn->scn_prefetch_queue, scan_prefetch_queue_compare,
	sizeof (scan_prefetch_issue_ctx_t),
	offsetof(scan_prefetch_issue_ctx_t, spic_avl_node));

	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	"scrub_func", sizeof (uint64_t), 1, &f);
	if (err == 0) {
	/*
	* There was an old-style scrub in progress. Restart a
	* new-style scrub from the beginning.
	*/
	scn->scn_restart_txg = txg;
	zfs_dbgmsg("old-style scrub was in progress for %s; "
	"restarting new-style scrub in txg %llu",
	spa->spa_name,
	(longlong_t)scn->scn_restart_txg);

	/*
	* Load the queue obj from the old location so that it
	* can be freed by dsl_scan_done().
	*/
	(void) zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	"scrub_queue", sizeof (uint64_t), 1,
	&scn->scn_phys.scn_queue_obj);
	} else {
	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
	&scn->scn_phys);
	/*
	* Detect if the pool contains the signature of #2094. If it
	* does properly update the scn->scn_phys structure and notify
	* the administrator by setting an errata for the pool.
	*/
	if (err == EOVERFLOW) {
	uint64_t zaptmp[SCAN_PHYS_NUMINTS + 1];
	VERIFY3S(SCAN_PHYS_NUMINTS, ==, 24);
	VERIFY3S(offsetof(dsl_scan_phys_t, scn_flags), ==,
	(23 * sizeof (uint64_t)));

	err = zap_lookup(dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SCAN,
	sizeof (uint64_t), SCAN_PHYS_NUMINTS + 1, &zaptmp);
	if (err == 0) {
	uint64_t overflow = zaptmp[SCAN_PHYS_NUMINTS];

	if (overflow & ~DSL_SCAN_FLAGS_MASK \|\|
	scn->scn_async_destroying) {
	spa->spa_errata =
	ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY;
	return (EOVERFLOW);
	}

	memcpy(&scn->scn_phys, zaptmp,
	SCAN_PHYS_NUMINTS * sizeof (uint64_t));
	scn->scn_phys.scn_flags = overflow;

	/* Required scrub already in progress. */
	if (scn->scn_phys.scn_state == DSS_FINISHED \|\|
	scn->scn_phys.scn_state == DSS_CANCELED)
	spa->spa_errata =
	ZPOOL_ERRATA_ZOL_2094_SCRUB;
	}
	}

	if (err == ENOENT)
	return (0);
	else if (err)
	return (err);

	/*
	* We might be restarting after a reboot, so jump the issued
	* counter to how far we've scanned. We know we're consistent
	* up to here.
	*/
	scn->scn_issued_before_pass = scn->scn_phys.scn_examined;

	if (dsl_scan_is_running(scn) &&
	spa_prev_software_version(dp->dp_spa) < SPA_VERSION_SCAN) {
	/*
	* A new-type scrub was in progress on an old
	* pool, and the pool was accessed by old
	* software. Restart from the beginning, since
	* the old software may have changed the pool in
	* the meantime.
	*/
	scn->scn_restart_txg = txg;
	zfs_dbgmsg("new-style scrub for %s was modified "
	"by old software; restarting in txg %llu",
	spa->spa_name,
	(longlong_t)scn->scn_restart_txg);
	} else if (dsl_scan_resilvering(dp)) {
	/*
	* If a resilver is in progress and there are already
	* errors, restart it instead of finishing this scan and
	* then restarting it. If there haven't been any errors
	* then remember that the incore DTL is valid.
	*/
	if (scn->scn_phys.scn_errors > 0) {
	scn->scn_restart_txg = txg;
	zfs_dbgmsg("resilver can't excise DTL_MISSING "
	"when finished; restarting on %s in txg "
	"%llu",
	spa->spa_name,
	(u_longlong_t)scn->scn_restart_txg);
	} else {
	/* it's safe to excise DTL when finished */
	spa->spa_scrub_started = B_TRUE;
	}
	}
	}

	memcpy(&scn->scn_phys_cached, &scn->scn_phys, sizeof (scn->scn_phys));

	/* reload the queue into the in-core state */
	if (scn->scn_phys.scn_queue_obj != 0) {
	zap_cursor_t zc;
	zap_attribute_t za;

	for (zap_cursor_init(&zc, dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	(void) zap_cursor_advance(&zc)) {
	scan_ds_queue_insert(scn,
	zfs_strtonum(za.za_name, NULL),
	za.za_first_integer);
	}
	zap_cursor_fini(&zc);
	}

	spa_scan_stat_init(spa);
	return (0);
	}

	void
	dsl_scan_fini(dsl_pool_t *dp)
	{
	if (dp->dp_scan != NULL) {
	dsl_scan_t *scn = dp->dp_scan;

	if (scn->scn_taskq != NULL)
	taskq_destroy(scn->scn_taskq);

	scan_ds_queue_clear(scn);
	avl_destroy(&scn->scn_queue);
	scan_ds_prefetch_queue_clear(scn);
	avl_destroy(&scn->scn_prefetch_queue);

	kmem_free(dp->dp_scan, sizeof (dsl_scan_t));
	dp->dp_scan = NULL;
	}
	}

	static boolean_t
	dsl_scan_restarting(dsl_scan_t scn, dmu_tx_t tx)
	{
	return (scn->scn_restart_txg != 0 &&
	scn->scn_restart_txg <= tx->tx_txg);
	}

	boolean_t
	dsl_scan_resilver_scheduled(dsl_pool_t *dp)
	{
	return ((dp->dp_scan && dp->dp_scan->scn_restart_txg != 0) \|\|
	(spa_async_tasks(dp->dp_spa) & SPA_ASYNC_RESILVER));
	}

	boolean_t
	dsl_scan_scrubbing(const dsl_pool_t *dp)
	{
	dsl_scan_phys_t *scn_phys = &dp->dp_scan->scn_phys;

	return (scn_phys->scn_state == DSS_SCANNING &&
	scn_phys->scn_func == POOL_SCAN_SCRUB);
	}

	boolean_t
	dsl_scan_is_paused_scrub(const dsl_scan_t *scn)
	{
	return (dsl_scan_scrubbing(scn->scn_dp) &&
	scn->scn_phys.scn_flags & DSF_SCRUB_PAUSED);
	}

	/*
	* Writes out a persistent dsl_scan_phys_t record to the pool directory.
	* Because we can be running in the block sorting algorithm, we do not always
	* want to write out the record, only when it is "safe" to do so. This safety
	* condition is achieved by making sure that the sorting queues are empty
	* (scn_queues_pending == 0). When this condition is not true, the sync'd state
	* is inconsistent with how much actual scanning progress has been made. The
	* kind of sync to be performed is specified by the sync_type argument. If the
	* sync is optional, we only sync if the queues are empty. If the sync is
	* mandatory, we do a hard ASSERT to make sure that the queues are empty. The
	* third possible state is a "cached" sync. This is done in response to:
	* 1) The dataset that was in the last sync'd dsl_scan_phys_t having been
	* destroyed, so we wouldn't be able to restart scanning from it.
	* 2) The snapshot that was in the last sync'd dsl_scan_phys_t having been
	* superseded by a newer snapshot.
	* 3) The dataset that was in the last sync'd dsl_scan_phys_t having been
	* swapped with its clone.
	* In all cases, a cached sync simply rewrites the last record we've written,
	* just slightly modified. For the modifications that are performed to the
	* last written dsl_scan_phys_t, see dsl_scan_ds_destroyed,
	* dsl_scan_ds_snapshotted and dsl_scan_ds_clone_swapped.
	*/
	static void
	dsl_scan_sync_state(dsl_scan_t scn, dmu_tx_t tx, state_sync_type_t sync_type)
	{
	int i;
	spa_t *spa = scn->scn_dp->dp_spa;

	ASSERT(sync_type != SYNC_MANDATORY \|\| scn->scn_queues_pending == 0);
	if (scn->scn_queues_pending == 0) {
	for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
	dsl_scan_io_queue_t *q = vd->vdev_scan_io_queue;

	if (q == NULL)
	continue;

	mutex_enter(&vd->vdev_scan_io_queue_lock);
	ASSERT3P(avl_first(&q->q_sios_by_addr), ==, NULL);
	ASSERT3P(zfs_btree_first(&q->q_exts_by_size, NULL), ==,
	NULL);
	ASSERT3P(range_tree_first(q->q_exts_by_addr), ==, NULL);
	mutex_exit(&vd->vdev_scan_io_queue_lock);
	}

	if (scn->scn_phys.scn_queue_obj != 0)
	scan_ds_queue_sync(scn, tx);
	VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
	&scn->scn_phys, tx));
	memcpy(&scn->scn_phys_cached, &scn->scn_phys,
	sizeof (scn->scn_phys));

	if (scn->scn_checkpointing)
	zfs_dbgmsg("finish scan checkpoint for %s",
	spa->spa_name);

	scn->scn_checkpointing = B_FALSE;
	scn->scn_last_checkpoint = ddi_get_lbolt();
	} else if (sync_type == SYNC_CACHED) {
	VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
	&scn->scn_phys_cached, tx));
	}
	}

	int
	dsl_scan_setup_check(void arg, dmu_tx_t tx)
	{
	(void) arg;
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
	vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;

	if (dsl_scan_is_running(scn) \|\| vdev_rebuild_active(rvd))
	return (SET_ERROR(EBUSY));

	return (0);
	}

	void
	dsl_scan_setup_sync(void arg, dmu_tx_t tx)
	{
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
	pool_scan_func_t *funcp = arg;
	dmu_object_type_t ot = 0;
	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;

	ASSERT(!dsl_scan_is_running(scn));
	ASSERT(funcp > POOL_SCAN_NONE && funcp < POOL_SCAN_FUNCS);
	memset(&scn->scn_phys, 0, sizeof (scn->scn_phys));
	scn->scn_phys.scn_func = *funcp;
	scn->scn_phys.scn_state = DSS_SCANNING;
	scn->scn_phys.scn_min_txg = 0;
	scn->scn_phys.scn_max_txg = tx->tx_txg;
	scn->scn_phys.scn_ddt_class_max = DDT_CLASSES - 1; /* the entire DDT */
	scn->scn_phys.scn_start_time = gethrestime_sec();
	scn->scn_phys.scn_errors = 0;
	scn->scn_phys.scn_to_examine = spa->spa_root_vdev->vdev_stat.vs_alloc;
	scn->scn_issued_before_pass = 0;
	scn->scn_restart_txg = 0;
	scn->scn_done_txg = 0;
	scn->scn_last_checkpoint = 0;
	scn->scn_checkpointing = B_FALSE;
	spa_scan_stat_init(spa);

	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
	scn->scn_phys.scn_ddt_class_max = zfs_scrub_ddt_class_max;

	/* rewrite all disk labels */
	vdev_config_dirty(spa->spa_root_vdev);

	if (vdev_resilver_needed(spa->spa_root_vdev,
	&scn->scn_phys.scn_min_txg, &scn->scn_phys.scn_max_txg)) {
	nvlist_t *aux = fnvlist_alloc();
	fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE,
	"healing");
	spa_event_notify(spa, NULL, aux,
	ESC_ZFS_RESILVER_START);
	nvlist_free(aux);
	} else {
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_START);
	}

	spa->spa_scrub_started = B_TRUE;
	/*
	* If this is an incremental scrub, limit the DDT scrub phase
	* to just the auto-ditto class (for correctness); the rest
	* of the scrub should go faster using top-down pruning.
	*/
	if (scn->scn_phys.scn_min_txg > TXG_INITIAL)
	scn->scn_phys.scn_ddt_class_max = DDT_CLASS_DITTO;

	/*
	* When starting a resilver clear any existing rebuild state.
	* This is required to prevent stale rebuild status from
	* being reported when a rebuild is run, then a resilver and
	* finally a scrub. In which case only the scrub status
	* should be reported by 'zpool status'.
	*/
	if (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) {
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];
	vdev_rebuild_clear_sync(
	(void *)(uintptr_t)vd->vdev_id, tx);
	}
	}
	}

	/* back to the generic stuff */

	if (zfs_scan_blkstats) {
	if (dp->dp_blkstats == NULL) {
	dp->dp_blkstats =
	vmem_alloc(sizeof (zfs_all_blkstats_t), KM_SLEEP);
	}
	memset(&dp->dp_blkstats->zab_type, 0,
	sizeof (dp->dp_blkstats->zab_type));
	} else {
	if (dp->dp_blkstats) {
	vmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
	dp->dp_blkstats = NULL;
	}
	}

	if (spa_version(spa) < SPA_VERSION_DSL_SCRUB)
	ot = DMU_OT_ZAP_OTHER;

	scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset,
	ot ? ot : DMU_OT_SCAN_QUEUE, DMU_OT_NONE, 0, tx);

	memcpy(&scn->scn_phys_cached, &scn->scn_phys, sizeof (scn->scn_phys));

	dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);

	spa_history_log_internal(spa, "scan setup", tx,
	"func=%u mintxg=%llu maxtxg=%llu",
	*funcp, (u_longlong_t)scn->scn_phys.scn_min_txg,
	(u_longlong_t)scn->scn_phys.scn_max_txg);
	}

	/*
	* Called by the ZFS_IOC_POOL_SCAN ioctl to start a scrub or resilver.
	* Can also be called to resume a paused scrub.
	*/
	int
	dsl_scan(dsl_pool_t *dp, pool_scan_func_t func)
	{
	spa_t *spa = dp->dp_spa;
	dsl_scan_t *scn = dp->dp_scan;

	/*
	* Purge all vdev caches and probe all devices. We do this here
	* rather than in sync context because this requires a writer lock
	* on the spa_config lock, which we can't do from sync context. The
	* spa_scrub_reopen flag indicates that vdev_open() should not
	* attempt to start another scrub.
	*/
	spa_vdev_state_enter(spa, SCL_NONE);
	spa->spa_scrub_reopen = B_TRUE;
	vdev_reopen(spa->spa_root_vdev);
	spa->spa_scrub_reopen = B_FALSE;
	(void) spa_vdev_state_exit(spa, NULL, 0);

	if (func == POOL_SCAN_RESILVER) {
	dsl_scan_restart_resilver(spa->spa_dsl_pool, 0);
	return (0);
	}

	if (func == POOL_SCAN_SCRUB && dsl_scan_is_paused_scrub(scn)) {
	/* got scrub start cmd, resume paused scrub */
	int err = dsl_scrub_set_pause_resume(scn->scn_dp,
	POOL_SCRUB_NORMAL);
	if (err == 0) {
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_RESUME);
	return (SET_ERROR(ECANCELED));
	}

	return (SET_ERROR(err));
	}

	return (dsl_sync_task(spa_name(spa), dsl_scan_setup_check,
	dsl_scan_setup_sync, &func, 0, ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	static void
	dsl_scan_done(dsl_scan_t scn, boolean_t complete, dmu_tx_t tx)
	{
	static const char *old_names[] = {
	"scrub_bookmark",
	"scrub_ddt_bookmark",
	"scrub_ddt_class_max",
	"scrub_queue",
	"scrub_min_txg",
	"scrub_max_txg",
	"scrub_func",
	"scrub_errors",
	NULL
	};

	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;
	int i;

	/* Remove any remnants of an old-style scrub. */
	for (i = 0; old_names[i]; i++) {
	(void) zap_remove(dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, old_names[i], tx);
	}

	if (scn->scn_phys.scn_queue_obj != 0) {
	VERIFY0(dmu_object_free(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, tx));
	scn->scn_phys.scn_queue_obj = 0;
	}
	scan_ds_queue_clear(scn);
	scan_ds_prefetch_queue_clear(scn);

	scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;

	/*
	* If we were "restarted" from a stopped state, don't bother
	* with anything else.
	*/
	if (!dsl_scan_is_running(scn)) {
	ASSERT(!scn->scn_is_sorted);
	return;
	}

	if (scn->scn_is_sorted) {
	scan_io_queues_destroy(scn);
	scn->scn_is_sorted = B_FALSE;

	if (scn->scn_taskq != NULL) {
	taskq_destroy(scn->scn_taskq);
	scn->scn_taskq = NULL;
	}
	}

	scn->scn_phys.scn_state = complete ? DSS_FINISHED : DSS_CANCELED;

	spa_notify_waiters(spa);

	if (dsl_scan_restarting(scn, tx))
	spa_history_log_internal(spa, "scan aborted, restarting", tx,
	"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
	else if (!complete)
	spa_history_log_internal(spa, "scan cancelled", tx,
	"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));
	else
	spa_history_log_internal(spa, "scan done", tx,
	"errors=%llu", (u_longlong_t)spa_get_errlog_size(spa));

	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
	spa->spa_scrub_active = B_FALSE;

	/*
	* If the scrub/resilver completed, update all DTLs to
	* reflect this. Whether it succeeded or not, vacate
	* all temporary scrub DTLs.
	*
	* As the scrub does not currently support traversing
	* data that have been freed but are part of a checkpoint,
	* we don't mark the scrub as done in the DTLs as faults
	* may still exist in those vdevs.
	*/
	if (complete &&
	!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
	scn->scn_phys.scn_max_txg, B_TRUE, B_FALSE);

	if (scn->scn_phys.scn_min_txg) {
	nvlist_t *aux = fnvlist_alloc();
	fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE,
	"healing");
	spa_event_notify(spa, NULL, aux,
	ESC_ZFS_RESILVER_FINISH);
	nvlist_free(aux);
	} else {
	spa_event_notify(spa, NULL, NULL,
	ESC_ZFS_SCRUB_FINISH);
	}
	} else {
	vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
	0, B_TRUE, B_FALSE);
	}
	spa_errlog_rotate(spa);

	/*
	* Don't clear flag until after vdev_dtl_reassess to ensure that
	* DTL_MISSING will get updated when possible.
	*/
	spa->spa_scrub_started = B_FALSE;

	/*
	* We may have finished replacing a device.
	* Let the async thread assess this and handle the detach.
	*/
	spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);

	/*
	* Clear any resilver_deferred flags in the config.
	* If there are drives that need resilvering, kick
	* off an asynchronous request to start resilver.
	* vdev_clear_resilver_deferred() may update the config
	* before the resilver can restart. In the event of
	* a crash during this period, the spa loading code
	* will find the drives that need to be resilvered
	* and start the resilver then.
	*/
	if (spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER) &&
	vdev_clear_resilver_deferred(spa->spa_root_vdev, tx)) {
	spa_history_log_internal(spa,
	"starting deferred resilver", tx, "errors=%llu",
	(u_longlong_t)spa_get_errlog_size(spa));
	spa_async_request(spa, SPA_ASYNC_RESILVER);
	}

	/* Clear recent error events (i.e. duplicate events tracking) */
	if (complete)
	zfs_ereport_clear(spa, NULL);
	}

	scn->scn_phys.scn_end_time = gethrestime_sec();

	if (spa->spa_errata == ZPOOL_ERRATA_ZOL_2094_SCRUB)
	spa->spa_errata = 0;

	ASSERT(!dsl_scan_is_running(scn));
	}

	static int
	dsl_scan_cancel_check(void arg, dmu_tx_t tx)
	{
	(void) arg;
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;

	if (!dsl_scan_is_running(scn))
	return (SET_ERROR(ENOENT));
	return (0);
	}

	static void
	dsl_scan_cancel_sync(void arg, dmu_tx_t tx)
	{
	(void) arg;
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;

	dsl_scan_done(scn, B_FALSE, tx);
	dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);
	spa_event_notify(scn->scn_dp->dp_spa, NULL, NULL, ESC_ZFS_SCRUB_ABORT);
	}

	int
	dsl_scan_cancel(dsl_pool_t *dp)
	{
	return (dsl_sync_task(spa_name(dp->dp_spa), dsl_scan_cancel_check,
	dsl_scan_cancel_sync, NULL, 3, ZFS_SPACE_CHECK_RESERVED));
	}

	static int
	dsl_scrub_pause_resume_check(void arg, dmu_tx_t tx)
	{
	pool_scrub_cmd_t *cmd = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_scan_t *scn = dp->dp_scan;

	if (*cmd == POOL_SCRUB_PAUSE) {
	/* can't pause a scrub when there is no in-progress scrub */
	if (!dsl_scan_scrubbing(dp))
	return (SET_ERROR(ENOENT));

	/* can't pause a paused scrub */
	if (dsl_scan_is_paused_scrub(scn))
	return (SET_ERROR(EBUSY));
	} else if (*cmd != POOL_SCRUB_NORMAL) {
	return (SET_ERROR(ENOTSUP));
	}

	return (0);
	}

	static void
	dsl_scrub_pause_resume_sync(void arg, dmu_tx_t tx)
	{
	pool_scrub_cmd_t *cmd = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	spa_t *spa = dp->dp_spa;
	dsl_scan_t *scn = dp->dp_scan;

	if (*cmd == POOL_SCRUB_PAUSE) {
	/* can't pause a scrub when there is no in-progress scrub */
	spa->spa_scan_pass_scrub_pause = gethrestime_sec();
	scn->scn_phys.scn_flags \|= DSF_SCRUB_PAUSED;
	scn->scn_phys_cached.scn_flags \|= DSF_SCRUB_PAUSED;
	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_PAUSED);
	spa_notify_waiters(spa);
	} else {
	ASSERT3U(*cmd, ==, POOL_SCRUB_NORMAL);
	if (dsl_scan_is_paused_scrub(scn)) {
	/*
	* We need to keep track of how much time we spend
	* paused per pass so that we can adjust the scrub rate
	* shown in the output of 'zpool status'
	*/
	spa->spa_scan_pass_scrub_spent_paused +=
	gethrestime_sec() - spa->spa_scan_pass_scrub_pause;
	spa->spa_scan_pass_scrub_pause = 0;
	scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;
	scn->scn_phys_cached.scn_flags &= ~DSF_SCRUB_PAUSED;
	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
	}
	}
	}

	/*
	* Set scrub pause/resume state if it makes sense to do so
	*/
	int
	dsl_scrub_set_pause_resume(const dsl_pool_t *dp, pool_scrub_cmd_t cmd)
	{
	return (dsl_sync_task(spa_name(dp->dp_spa),
	dsl_scrub_pause_resume_check, dsl_scrub_pause_resume_sync, &cmd, 3,
	ZFS_SPACE_CHECK_RESERVED));
	}


	/* start a new scan, or restart an existing one. */
	void
	dsl_scan_restart_resilver(dsl_pool_t *dp, uint64_t txg)
	{
	if (txg == 0) {
	dmu_tx_t *tx;
	tx = dmu_tx_create_dd(dp->dp_mos_dir);
	VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));

	txg = dmu_tx_get_txg(tx);
	dp->dp_scan->scn_restart_txg = txg;
	dmu_tx_commit(tx);
	} else {
	dp->dp_scan->scn_restart_txg = txg;
	}
	zfs_dbgmsg("restarting resilver for %s at txg=%llu",
	dp->dp_spa->spa_name, (longlong_t)txg);
	}

	void
	dsl_free(dsl_pool_t dp, uint64_t txg, const blkptr_t bp)
	{
	zio_free(dp->dp_spa, txg, bp);
	}

	void
	dsl_free_sync(zio_t pio, dsl_pool_t dp, uint64_t txg, const blkptr_t *bpp)
	{
	ASSERT(dsl_pool_sync_context(dp));
	zio_nowait(zio_free_sync(pio, dp->dp_spa, txg, bpp, pio->io_flags));
	}

	static int
	scan_ds_queue_compare(const void a, const void b)
	{
	const scan_ds_t sds_a = a, sds_b = b;

	if (sds_a->sds_dsobj < sds_b->sds_dsobj)
	return (-1);
	if (sds_a->sds_dsobj == sds_b->sds_dsobj)
	return (0);
	return (1);
	}

	static void
	scan_ds_queue_clear(dsl_scan_t *scn)
	{
	void *cookie = NULL;
	scan_ds_t *sds;
	while ((sds = avl_destroy_nodes(&scn->scn_queue, &cookie)) != NULL) {
	kmem_free(sds, sizeof (*sds));
	}
	}

	static boolean_t
	scan_ds_queue_contains(dsl_scan_t scn, uint64_t dsobj, uint64_t txg)
	{
	scan_ds_t srch, *sds;

	srch.sds_dsobj = dsobj;
	sds = avl_find(&scn->scn_queue, &srch, NULL);
	if (sds != NULL && txg != NULL)
	*txg = sds->sds_txg;
	return (sds != NULL);
	}

	static void
	scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg)
	{
	scan_ds_t *sds;
	avl_index_t where;

	sds = kmem_zalloc(sizeof (*sds), KM_SLEEP);
	sds->sds_dsobj = dsobj;
	sds->sds_txg = txg;

	VERIFY3P(avl_find(&scn->scn_queue, sds, &where), ==, NULL);
	avl_insert(&scn->scn_queue, sds, where);
	}

	static void
	scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj)
	{
	scan_ds_t srch, *sds;

	srch.sds_dsobj = dsobj;

	sds = avl_find(&scn->scn_queue, &srch, NULL);
	VERIFY(sds != NULL);
	avl_remove(&scn->scn_queue, sds);
	kmem_free(sds, sizeof (*sds));
	}

	static void
	scan_ds_queue_sync(dsl_scan_t scn, dmu_tx_t tx)
	{
	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;
	dmu_object_type_t ot = (spa_version(spa) >= SPA_VERSION_DSL_SCRUB) ?
	DMU_OT_SCAN_QUEUE : DMU_OT_ZAP_OTHER;

	ASSERT0(scn->scn_queues_pending);
	ASSERT(scn->scn_phys.scn_queue_obj != 0);

	VERIFY0(dmu_object_free(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, tx));
	scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset, ot,
	DMU_OT_NONE, 0, tx);
	for (scan_ds_t *sds = avl_first(&scn->scn_queue);
	sds != NULL; sds = AVL_NEXT(&scn->scn_queue, sds)) {
	VERIFY0(zap_add_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, sds->sds_dsobj,
	sds->sds_txg, tx));
	}
	}

	/*
	* Computes the memory limit state that we're currently in. A sorted scan
	* needs quite a bit of memory to hold the sorting queue, so we need to
	* reasonably constrain the size so it doesn't impact overall system
	* performance. We compute two limits:
	* 1) Hard memory limit: if the amount of memory used by the sorting
	* queues on a pool gets above this value, we stop the metadata
	* scanning portion and start issuing the queued up and sorted
	* I/Os to reduce memory usage.
	* This limit is calculated as a fraction of physmem (by default 5%).
	* We constrain the lower bound of the hard limit to an absolute
	* minimum of zfs_scan_mem_lim_min (default: 16 MiB). We also constrain
	* the upper bound to 5% of the total pool size - no chance we'll
	* ever need that much memory, but just to keep the value in check.
	* 2) Soft memory limit: once we hit the hard memory limit, we start
	* issuing I/O to reduce queue memory usage, but we don't want to
	* completely empty out the queues, since we might be able to find I/Os
	* that will fill in the gaps of our non-sequential IOs at some point
	* in the future. So we stop the issuing of I/Os once the amount of
	* memory used drops below the soft limit (at which point we stop issuing
	* I/O and start scanning metadata again).
	*
	* This limit is calculated by subtracting a fraction of the hard
	* limit from the hard limit. By default this fraction is 5%, so
	* the soft limit is 95% of the hard limit. We cap the size of the
	* difference between the hard and soft limits at an absolute
	* maximum of zfs_scan_mem_lim_soft_max (default: 128 MiB) - this is
	* sufficient to not cause too frequent switching between the
	* metadata scan and I/O issue (even at 2k recordsize, 128 MiB's
	* worth of queues is about 1.2 GiB of on-pool data, so scanning
	* that should take at least a decent fraction of a second).
	*/
	static boolean_t
	dsl_scan_should_clear(dsl_scan_t *scn)
	{
	spa_t *spa = scn->scn_dp->dp_spa;
	vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;
	uint64_t alloc, mlim_hard, mlim_soft, mused;

	alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	alloc += metaslab_class_get_alloc(spa_special_class(spa));
	alloc += metaslab_class_get_alloc(spa_dedup_class(spa));

	mlim_hard = MAX((physmem / zfs_scan_mem_lim_fact) * PAGESIZE,
	zfs_scan_mem_lim_min);
	mlim_hard = MIN(mlim_hard, alloc / 20);
	mlim_soft = mlim_hard - MIN(mlim_hard / zfs_scan_mem_lim_soft_fact,
	zfs_scan_mem_lim_soft_max);
	mused = 0;
	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *tvd = rvd->vdev_child[i];
	dsl_scan_io_queue_t *queue;

	mutex_enter(&tvd->vdev_scan_io_queue_lock);
	queue = tvd->vdev_scan_io_queue;
	if (queue != NULL) {
	/*
	* # of extents in exts_by_addr = # in exts_by_size.
	* B-tree efficiency is ~75%, but can be as low as 50%.
	*/
	mused += zfs_btree_numnodes(&queue->q_exts_by_size) *
	((sizeof (range_seg_gap_t) + sizeof (uint64_t)) *
	3 / 2) + queue->q_sio_memused;
	}
	mutex_exit(&tvd->vdev_scan_io_queue_lock);
	}

	dprintf("current scan memory usage: %llu bytes\n", (longlong_t)mused);

	if (mused == 0)
	ASSERT0(scn->scn_queues_pending);

	/*
	* If we are above our hard limit, we need to clear out memory.
	* If we are below our soft limit, we need to accumulate sequential IOs.
	* Otherwise, we should keep doing whatever we are currently doing.
	*/
	if (mused >= mlim_hard)
	return (B_TRUE);
	else if (mused < mlim_soft)
	return (B_FALSE);
	else
	return (scn->scn_clearing);
	}

	static boolean_t
	dsl_scan_check_suspend(dsl_scan_t scn, const zbookmark_phys_t zb)
	{
	/* we never skip user/group accounting objects */
	if (zb && (int64_t)zb->zb_object < 0)
	return (B_FALSE);

	if (scn->scn_suspending)
	return (B_TRUE); /* we're already suspending */

	if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark))
	return (B_FALSE); /* we're resuming */

	/* We only know how to resume from level-0 and objset blocks. */
	if (zb && (zb->zb_level != 0 && zb->zb_level != ZB_ROOT_LEVEL))
	return (B_FALSE);

	/*
	* We suspend if:
	* - we have scanned for at least the minimum time (default 1 sec
	* for scrub, 3 sec for resilver), and either we have sufficient
	* dirty data that we are starting to write more quickly
	* (default 30%), someone is explicitly waiting for this txg
	* to complete, or we have used up all of the time in the txg
	* timeout (default 5 sec).
	* or
	* - the spa is shutting down because this pool is being exported
	* or the machine is rebooting.
	* or
	* - the scan queue has reached its memory use limit
	*/
	uint64_t curr_time_ns = gethrtime();
	uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
	uint64_t sync_time_ns = curr_time_ns -
	scn->scn_dp->dp_spa->spa_sync_starttime;
	uint64_t dirty_min_bytes = zfs_dirty_data_max *
	zfs_vdev_async_write_active_min_dirty_percent / 100;
	- int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	+ uint_t mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;

	if ((NSEC2MSEC(scan_time_ns) > mintime &&
	(scn->scn_dp->dp_dirty_total >= dirty_min_bytes \|\|
	txg_sync_waiting(scn->scn_dp) \|\|
	NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) \|\|
	spa_shutting_down(scn->scn_dp->dp_spa) \|\|
	(zfs_scan_strict_mem_lim && dsl_scan_should_clear(scn))) {
	if (zb && zb->zb_level == ZB_ROOT_LEVEL) {
	dprintf("suspending at first available bookmark "
	"%llx/%llx/%llx/%llx\n",
	(longlong_t)zb->zb_objset,
	(longlong_t)zb->zb_object,
	(longlong_t)zb->zb_level,
	(longlong_t)zb->zb_blkid);
	SET_BOOKMARK(&scn->scn_phys.scn_bookmark,
	zb->zb_objset, 0, 0, 0);
	} else if (zb != NULL) {
	dprintf("suspending at bookmark %llx/%llx/%llx/%llx\n",
	(longlong_t)zb->zb_objset,
	(longlong_t)zb->zb_object,
	(longlong_t)zb->zb_level,
	(longlong_t)zb->zb_blkid);
	scn->scn_phys.scn_bookmark = *zb;
	} else {
	#ifdef ZFS_DEBUG
	dsl_scan_phys_t *scnp = &scn->scn_phys;
	dprintf("suspending at at DDT bookmark "
	"%llx/%llx/%llx/%llx\n",
	(longlong_t)scnp->scn_ddt_bookmark.ddb_class,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_type,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_checksum,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_cursor);
	#endif
	}
	scn->scn_suspending = B_TRUE;
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	typedef struct zil_scan_arg {
	dsl_pool_t *zsa_dp;
	zil_header_t *zsa_zh;
	} zil_scan_arg_t;

	static int
	dsl_scan_zil_block(zilog_t zilog, const blkptr_t bp, void *arg,
	uint64_t claim_txg)
	{
	(void) zilog;
	zil_scan_arg_t *zsa = arg;
	dsl_pool_t *dp = zsa->zsa_dp;
	dsl_scan_t *scn = dp->dp_scan;
	zil_header_t *zh = zsa->zsa_zh;
	zbookmark_phys_t zb;

	ASSERT(!BP_IS_REDACTED(bp));
	if (BP_IS_HOLE(bp) \|\| bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
	return (0);

	/*
	* One block ("stubby") can be allocated a long time ago; we
	* want to visit that one because it has been allocated
	* (on-disk) even if it hasn't been claimed (even though for
	* scrub there's nothing to do to it).
	*/
	if (claim_txg == 0 && bp->blk_birth >= spa_min_claim_txg(dp->dp_spa))
	return (0);

	SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
	ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);

	VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
	return (0);
	}

	static int
	dsl_scan_zil_record(zilog_t zilog, const lr_t lrc, void *arg,
	uint64_t claim_txg)
	{
	(void) zilog;
	if (lrc->lrc_txtype == TX_WRITE) {
	zil_scan_arg_t *zsa = arg;
	dsl_pool_t *dp = zsa->zsa_dp;
	dsl_scan_t *scn = dp->dp_scan;
	zil_header_t *zh = zsa->zsa_zh;
	const lr_write_t lr = (const lr_write_t )lrc;
	const blkptr_t *bp = &lr->lr_blkptr;
	zbookmark_phys_t zb;

	ASSERT(!BP_IS_REDACTED(bp));
	if (BP_IS_HOLE(bp) \|\|
	bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
	return (0);

	/*
	* birth can be < claim_txg if this record's txg is
	* already txg sync'ed (but this log block contains
	* other records that are not synced)
	*/
	if (claim_txg == 0 \|\| bp->blk_birth < claim_txg)
	return (0);

	SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
	lr->lr_foid, ZB_ZIL_LEVEL,
	lr->lr_offset / BP_GET_LSIZE(bp));

	VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
	}
	return (0);
	}

	static void
	dsl_scan_zil(dsl_pool_t dp, zil_header_t zh)
	{
	uint64_t claim_txg = zh->zh_claim_txg;
	zil_scan_arg_t zsa = { dp, zh };
	zilog_t *zilog;

	ASSERT(spa_writeable(dp->dp_spa));

	/*
	* We only want to visit blocks that have been claimed but not yet
	* replayed (or, in read-only mode, blocks that would be claimed).
	*/
	if (claim_txg == 0)
	return;

	zilog = zil_alloc(dp->dp_meta_objset, zh);

	(void) zil_parse(zilog, dsl_scan_zil_block, dsl_scan_zil_record, &zsa,
	claim_txg, B_FALSE);

	zil_free(zilog);
	}

	/*
	* We compare scan_prefetch_issue_ctx_t's based on their bookmarks. The idea
	* here is to sort the AVL tree by the order each block will be needed.
	*/
	static int
	scan_prefetch_queue_compare(const void a, const void b)
	{
	const scan_prefetch_issue_ctx_t spic_a = a, spic_b = b;
	const scan_prefetch_ctx_t *spc_a = spic_a->spic_spc;
	const scan_prefetch_ctx_t *spc_b = spic_b->spic_spc;

	return (zbookmark_compare(spc_a->spc_datablkszsec,
	spc_a->spc_indblkshift, spc_b->spc_datablkszsec,
	spc_b->spc_indblkshift, &spic_a->spic_zb, &spic_b->spic_zb));
	}

	static void
	scan_prefetch_ctx_rele(scan_prefetch_ctx_t spc, const void tag)
	{
	if (zfs_refcount_remove(&spc->spc_refcnt, tag) == 0) {
	zfs_refcount_destroy(&spc->spc_refcnt);
	kmem_free(spc, sizeof (scan_prefetch_ctx_t));
	}
	}

	static scan_prefetch_ctx_t *
	scan_prefetch_ctx_create(dsl_scan_t scn, dnode_phys_t dnp, const void *tag)
	{
	scan_prefetch_ctx_t *spc;

	spc = kmem_alloc(sizeof (scan_prefetch_ctx_t), KM_SLEEP);
	zfs_refcount_create(&spc->spc_refcnt);
	zfs_refcount_add(&spc->spc_refcnt, tag);
	spc->spc_scn = scn;
	if (dnp != NULL) {
	spc->spc_datablkszsec = dnp->dn_datablkszsec;
	spc->spc_indblkshift = dnp->dn_indblkshift;
	spc->spc_root = B_FALSE;
	} else {
	spc->spc_datablkszsec = 0;
	spc->spc_indblkshift = 0;
	spc->spc_root = B_TRUE;
	}

	return (spc);
	}

	static void
	scan_prefetch_ctx_add_ref(scan_prefetch_ctx_t spc, const void tag)
	{
	zfs_refcount_add(&spc->spc_refcnt, tag);
	}

	static void
	scan_ds_prefetch_queue_clear(dsl_scan_t *scn)
	{
	spa_t *spa = scn->scn_dp->dp_spa;
	void *cookie = NULL;
	scan_prefetch_issue_ctx_t *spic = NULL;

	mutex_enter(&spa->spa_scrub_lock);
	while ((spic = avl_destroy_nodes(&scn->scn_prefetch_queue,
	&cookie)) != NULL) {
	scan_prefetch_ctx_rele(spic->spic_spc, scn);
	kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}
	mutex_exit(&spa->spa_scrub_lock);
	}

	static boolean_t
	dsl_scan_check_prefetch_resume(scan_prefetch_ctx_t *spc,
	const zbookmark_phys_t *zb)
	{
	zbookmark_phys_t *last_zb = &spc->spc_scn->scn_prefetch_bookmark;
	dnode_phys_t tmp_dnp;
	dnode_phys_t *dnp = (spc->spc_root) ? NULL : &tmp_dnp;

	if (zb->zb_objset != last_zb->zb_objset)
	return (B_TRUE);
	if ((int64_t)zb->zb_object < 0)
	return (B_FALSE);

	tmp_dnp.dn_datablkszsec = spc->spc_datablkszsec;
	tmp_dnp.dn_indblkshift = spc->spc_indblkshift;

	if (zbookmark_subtree_completed(dnp, zb, last_zb))
	return (B_TRUE);

	return (B_FALSE);
	}

	static void
	dsl_scan_prefetch(scan_prefetch_ctx_t spc, blkptr_t bp, zbookmark_phys_t *zb)
	{
	avl_index_t idx;
	dsl_scan_t *scn = spc->spc_scn;
	spa_t *spa = scn->scn_dp->dp_spa;
	scan_prefetch_issue_ctx_t *spic;

	if (zfs_no_scrub_prefetch \|\| BP_IS_REDACTED(bp))
	return;

	if (BP_IS_HOLE(bp) \|\| bp->blk_birth <= scn->scn_phys.scn_cur_min_txg \|\|
	(BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE &&
	BP_GET_TYPE(bp) != DMU_OT_OBJSET))
	return;

	if (dsl_scan_check_prefetch_resume(spc, zb))
	return;

	scan_prefetch_ctx_add_ref(spc, scn);
	spic = kmem_alloc(sizeof (scan_prefetch_issue_ctx_t), KM_SLEEP);
	spic->spic_spc = spc;
	spic->spic_bp = *bp;
	spic->spic_zb = *zb;

	/*
	* Add the IO to the queue of blocks to prefetch. This allows us to
	* prioritize blocks that we will need first for the main traversal
	* thread.
	*/
	mutex_enter(&spa->spa_scrub_lock);
	if (avl_find(&scn->scn_prefetch_queue, spic, &idx) != NULL) {
	/* this block is already queued for prefetch */
	kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	scan_prefetch_ctx_rele(spc, scn);
	mutex_exit(&spa->spa_scrub_lock);
	return;
	}

	avl_insert(&scn->scn_prefetch_queue, spic, idx);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);
	}

	static void
	dsl_scan_prefetch_dnode(dsl_scan_t scn, dnode_phys_t dnp,
	uint64_t objset, uint64_t object)
	{
	int i;
	zbookmark_phys_t zb;
	scan_prefetch_ctx_t *spc;

	if (dnp->dn_nblkptr == 0 && !(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
	return;

	SET_BOOKMARK(&zb, objset, object, 0, 0);

	spc = scan_prefetch_ctx_create(scn, dnp, FTAG);

	for (i = 0; i < dnp->dn_nblkptr; i++) {
	zb.zb_level = BP_GET_LEVEL(&dnp->dn_blkptr[i]);
	zb.zb_blkid = i;
	dsl_scan_prefetch(spc, &dnp->dn_blkptr[i], &zb);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	zb.zb_level = 0;
	zb.zb_blkid = DMU_SPILL_BLKID;
	dsl_scan_prefetch(spc, DN_SPILL_BLKPTR(dnp), &zb);
	}

	scan_prefetch_ctx_rele(spc, FTAG);
	}

	static void
	dsl_scan_prefetch_cb(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void private)
	{
	(void) zio;
	scan_prefetch_ctx_t *spc = private;
	dsl_scan_t *scn = spc->spc_scn;
	spa_t *spa = scn->scn_dp->dp_spa;

	/* broadcast that the IO has completed for rate limiting purposes */
	mutex_enter(&spa->spa_scrub_lock);
	ASSERT3U(spa->spa_scrub_inflight, >=, BP_GET_PSIZE(bp));
	spa->spa_scrub_inflight -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);

	/* if there was an error or we are done prefetching, just cleanup */
	if (buf == NULL \|\| scn->scn_prefetch_stop)
	goto out;

	if (BP_GET_LEVEL(bp) > 0) {
	int i;
	blkptr_t *cbp;
	int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
	zbookmark_phys_t czb;

	for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
	SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
	zb->zb_level - 1, zb->zb_blkid * epb + i);
	dsl_scan_prefetch(spc, cbp, &czb);
	}
	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
	dnode_phys_t *cdnp;
	int i;
	int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;

	for (i = 0, cdnp = buf->b_data; i < epb;
	i += cdnp->dn_extra_slots + 1,
	cdnp += cdnp->dn_extra_slots + 1) {
	dsl_scan_prefetch_dnode(scn, cdnp,
	zb->zb_objset, zb->zb_blkid * epb + i);
	}
	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
	objset_phys_t *osp = buf->b_data;

	dsl_scan_prefetch_dnode(scn, &osp->os_meta_dnode,
	zb->zb_objset, DMU_META_DNODE_OBJECT);

	if (OBJSET_BUF_HAS_USERUSED(buf)) {
	dsl_scan_prefetch_dnode(scn,
	&osp->os_groupused_dnode, zb->zb_objset,
	DMU_GROUPUSED_OBJECT);
	dsl_scan_prefetch_dnode(scn,
	&osp->os_userused_dnode, zb->zb_objset,
	DMU_USERUSED_OBJECT);
	}
	}

	out:
	if (buf != NULL)
	arc_buf_destroy(buf, private);
	scan_prefetch_ctx_rele(spc, scn);
	}

	static void
	dsl_scan_prefetch_thread(void *arg)
	{
	dsl_scan_t *scn = arg;
	spa_t *spa = scn->scn_dp->dp_spa;
	scan_prefetch_issue_ctx_t *spic;

	/* loop until we are told to stop */
	while (!scn->scn_prefetch_stop) {
	arc_flags_t flags = ARC_FLAG_NOWAIT \|
	ARC_FLAG_PRESCIENT_PREFETCH \| ARC_FLAG_PREFETCH;
	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SCAN_THREAD;

	mutex_enter(&spa->spa_scrub_lock);

	/*
	* Wait until we have an IO to issue and are not above our
	* maximum in flight limit.
	*/
	while (!scn->scn_prefetch_stop &&
	(avl_numnodes(&scn->scn_prefetch_queue) == 0 \|\|
	spa->spa_scrub_inflight >= scn->scn_maxinflight_bytes)) {
	cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
	}

	/* recheck if we should stop since we waited for the cv */
	if (scn->scn_prefetch_stop) {
	mutex_exit(&spa->spa_scrub_lock);
	break;
	}

	/* remove the prefetch IO from the tree */
	spic = avl_first(&scn->scn_prefetch_queue);
	spa->spa_scrub_inflight += BP_GET_PSIZE(&spic->spic_bp);
	avl_remove(&scn->scn_prefetch_queue, spic);

	mutex_exit(&spa->spa_scrub_lock);

	if (BP_IS_PROTECTED(&spic->spic_bp)) {
	ASSERT(BP_GET_TYPE(&spic->spic_bp) == DMU_OT_DNODE \|\|
	BP_GET_TYPE(&spic->spic_bp) == DMU_OT_OBJSET);
	ASSERT3U(BP_GET_LEVEL(&spic->spic_bp), ==, 0);
	zio_flags \|= ZIO_FLAG_RAW;
	}

	/* issue the prefetch asynchronously */
	(void) arc_read(scn->scn_zio_root, scn->scn_dp->dp_spa,
	&spic->spic_bp, dsl_scan_prefetch_cb, spic->spic_spc,
	ZIO_PRIORITY_SCRUB, zio_flags, &flags, &spic->spic_zb);

	kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}

	ASSERT(scn->scn_prefetch_stop);

	/* free any prefetches we didn't get to complete */
	mutex_enter(&spa->spa_scrub_lock);
	while ((spic = avl_first(&scn->scn_prefetch_queue)) != NULL) {
	avl_remove(&scn->scn_prefetch_queue, spic);
	scan_prefetch_ctx_rele(spic->spic_spc, scn);
	kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}
	ASSERT0(avl_numnodes(&scn->scn_prefetch_queue));
	mutex_exit(&spa->spa_scrub_lock);
	}

	static boolean_t
	dsl_scan_check_resume(dsl_scan_t scn, const dnode_phys_t dnp,
	const zbookmark_phys_t *zb)
	{
	/*
	* We never skip over user/group accounting objects (obj<0)
	*/
	if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark) &&
	(int64_t)zb->zb_object >= 0) {
	/*
	* If we already visited this bp & everything below (in
	* a prior txg sync), don't bother doing it again.
	*/
	if (zbookmark_subtree_completed(dnp, zb,
	&scn->scn_phys.scn_bookmark))
	return (B_TRUE);

	/*
	* If we found the block we're trying to resume from, or
	* we went past it, zero it out to indicate that it's OK
	* to start checking for suspending again.
	*/
	if (zbookmark_subtree_tbd(dnp, zb,
	&scn->scn_phys.scn_bookmark)) {
	dprintf("resuming at %llx/%llx/%llx/%llx\n",
	(longlong_t)zb->zb_objset,
	(longlong_t)zb->zb_object,
	(longlong_t)zb->zb_level,
	(longlong_t)zb->zb_blkid);
	memset(&scn->scn_phys.scn_bookmark, 0, sizeof (*zb));
	}
	}
	return (B_FALSE);
	}

	static void dsl_scan_visitbp(blkptr_t bp, const zbookmark_phys_t zb,
	dnode_phys_t dnp, dsl_dataset_t ds, dsl_scan_t *scn,
	dmu_objset_type_t ostype, dmu_tx_t *tx);
	inline __attribute__((always_inline)) static void dsl_scan_visitdnode(
	dsl_scan_t , dsl_dataset_t ds, dmu_objset_type_t ostype,
	dnode_phys_t dnp, uint64_t object, dmu_tx_t tx);

	/*
	* Return nonzero on i/o error.
	* Return new buf to write out in *bufp.
	*/
	inline __attribute__((always_inline)) static int
	dsl_scan_recurse(dsl_scan_t scn, dsl_dataset_t ds, dmu_objset_type_t ostype,
	dnode_phys_t dnp, const blkptr_t bp,
	const zbookmark_phys_t zb, dmu_tx_t tx)
	{
	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;
	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SCAN_THREAD;
	int err;

	ASSERT(!BP_IS_REDACTED(bp));

	/*
	* There is an unlikely case of encountering dnodes with contradicting
	* dn_bonuslen and DNODE_FLAG_SPILL_BLKPTR flag before in files created
	* or modified before commit 4254acb was merged. As it is not possible
	* to know which of the two is correct, report an error.
	*/
	if (dnp != NULL &&
	dnp->dn_bonuslen > DN_MAX_BONUS_LEN(dnp)) {
	scn->scn_phys.scn_errors++;
	spa_log_error(spa, zb);
	return (SET_ERROR(EINVAL));
	}

	if (BP_GET_LEVEL(bp) > 0) {
	arc_flags_t flags = ARC_FLAG_WAIT;
	int i;
	blkptr_t *cbp;
	int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
	arc_buf_t *buf;

	err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
	if (err) {
	scn->scn_phys.scn_errors++;
	return (err);
	}
	for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
	zbookmark_phys_t czb;

	SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
	zb->zb_level - 1,
	zb->zb_blkid * epb + i);
	dsl_scan_visitbp(cbp, &czb, dnp,
	ds, scn, ostype, tx);
	}
	arc_buf_destroy(buf, &buf);
	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
	arc_flags_t flags = ARC_FLAG_WAIT;
	dnode_phys_t *cdnp;
	int i;
	int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
	arc_buf_t *buf;

	if (BP_IS_PROTECTED(bp)) {
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	zio_flags \|= ZIO_FLAG_RAW;
	}

	err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
	if (err) {
	scn->scn_phys.scn_errors++;
	return (err);
	}
	for (i = 0, cdnp = buf->b_data; i < epb;
	i += cdnp->dn_extra_slots + 1,
	cdnp += cdnp->dn_extra_slots + 1) {
	dsl_scan_visitdnode(scn, ds, ostype,
	cdnp, zb->zb_blkid * epb + i, tx);
	}

	arc_buf_destroy(buf, &buf);
	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
	arc_flags_t flags = ARC_FLAG_WAIT;
	objset_phys_t *osp;
	arc_buf_t *buf;

	err = arc_read(NULL, spa, bp, arc_getbuf_func, &buf,
	ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
	if (err) {
	scn->scn_phys.scn_errors++;
	return (err);
	}

	osp = buf->b_data;

	dsl_scan_visitdnode(scn, ds, osp->os_type,
	&osp->os_meta_dnode, DMU_META_DNODE_OBJECT, tx);

	if (OBJSET_BUF_HAS_USERUSED(buf)) {
	/*
	* We also always visit user/group/project accounting
	* objects, and never skip them, even if we are
	* suspending. This is necessary so that the
	* space deltas from this txg get integrated.
	*/
	if (OBJSET_BUF_HAS_PROJECTUSED(buf))
	dsl_scan_visitdnode(scn, ds, osp->os_type,
	&osp->os_projectused_dnode,
	DMU_PROJECTUSED_OBJECT, tx);
	dsl_scan_visitdnode(scn, ds, osp->os_type,
	&osp->os_groupused_dnode,
	DMU_GROUPUSED_OBJECT, tx);
	dsl_scan_visitdnode(scn, ds, osp->os_type,
	&osp->os_userused_dnode,
	DMU_USERUSED_OBJECT, tx);
	}
	arc_buf_destroy(buf, &buf);
	} else if (!zfs_blkptr_verify(spa, bp, B_FALSE, BLK_VERIFY_LOG)) {
	/*
	* Sanity check the block pointer contents, this is handled
	* by arc_read() for the cases above.
	*/
	scn->scn_phys.scn_errors++;
	spa_log_error(spa, zb);
	return (SET_ERROR(EINVAL));
	}

	return (0);
	}

	inline __attribute__((always_inline)) static void
	dsl_scan_visitdnode(dsl_scan_t scn, dsl_dataset_t ds,
	dmu_objset_type_t ostype, dnode_phys_t *dnp,
	uint64_t object, dmu_tx_t *tx)
	{
	int j;

	for (j = 0; j < dnp->dn_nblkptr; j++) {
	zbookmark_phys_t czb;

	SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
	dnp->dn_nlevels - 1, j);
	dsl_scan_visitbp(&dnp->dn_blkptr[j],
	&czb, dnp, ds, scn, ostype, tx);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	zbookmark_phys_t czb;
	SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
	0, DMU_SPILL_BLKID);
	dsl_scan_visitbp(DN_SPILL_BLKPTR(dnp),
	&czb, dnp, ds, scn, ostype, tx);
	}
	}

	/*
	* The arguments are in this order because mdb can only print the
	* first 5; we want them to be useful.
	*/
	static void
	dsl_scan_visitbp(blkptr_t bp, const zbookmark_phys_t zb,
	dnode_phys_t dnp, dsl_dataset_t ds, dsl_scan_t *scn,
	dmu_objset_type_t ostype, dmu_tx_t *tx)
	{
	dsl_pool_t *dp = scn->scn_dp;
	blkptr_t *bp_toread = NULL;

	if (dsl_scan_check_suspend(scn, zb))
	return;

	if (dsl_scan_check_resume(scn, dnp, zb))
	return;

	scn->scn_visited_this_txg++;

	if (BP_IS_HOLE(bp)) {
	scn->scn_holes_this_txg++;
	return;
	}

	if (BP_IS_REDACTED(bp)) {
	ASSERT(dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_REDACTED_DATASETS));
	return;
	}

	if (bp->blk_birth <= scn->scn_phys.scn_cur_min_txg) {
	scn->scn_lt_min_this_txg++;
	return;
	}

	bp_toread = kmem_alloc(sizeof (blkptr_t), KM_SLEEP);
	bp_toread = bp;

	if (dsl_scan_recurse(scn, ds, ostype, dnp, bp_toread, zb, tx) != 0)
	goto out;

	/*
	* If dsl_scan_ddt() has already visited this block, it will have
	* already done any translations or scrubbing, so don't call the
	* callback again.
	*/
	if (ddt_class_contains(dp->dp_spa,
	scn->scn_phys.scn_ddt_class_max, bp)) {
	scn->scn_ddt_contained_this_txg++;
	goto out;
	}

	/*
	* If this block is from the future (after cur_max_txg), then we
	* are doing this on behalf of a deleted snapshot, and we will
	* revisit the future block on the next pass of this dataset.
	* Don't scan it now unless we need to because something
	* under it was modified.
	*/
	if (BP_PHYSICAL_BIRTH(bp) > scn->scn_phys.scn_cur_max_txg) {
	scn->scn_gt_max_this_txg++;
	goto out;
	}

	scan_funcs[scn->scn_phys.scn_func](dp, bp, zb);

	out:
	kmem_free(bp_toread, sizeof (blkptr_t));
	}

	static void
	dsl_scan_visit_rootbp(dsl_scan_t scn, dsl_dataset_t ds, blkptr_t *bp,
	dmu_tx_t *tx)
	{
	zbookmark_phys_t zb;
	scan_prefetch_ctx_t *spc;

	SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
	ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);

	if (ZB_IS_ZERO(&scn->scn_phys.scn_bookmark)) {
	SET_BOOKMARK(&scn->scn_prefetch_bookmark,
	zb.zb_objset, 0, 0, 0);
	} else {
	scn->scn_prefetch_bookmark = scn->scn_phys.scn_bookmark;
	}

	scn->scn_objsets_visited_this_txg++;

	spc = scan_prefetch_ctx_create(scn, NULL, FTAG);
	dsl_scan_prefetch(spc, bp, &zb);
	scan_prefetch_ctx_rele(spc, FTAG);

	dsl_scan_visitbp(bp, &zb, NULL, ds, scn, DMU_OST_NONE, tx);

	dprintf_ds(ds, "finished scan%s", "");
	}

	static void
	ds_destroyed_scn_phys(dsl_dataset_t ds, dsl_scan_phys_t scn_phys)
	{
	if (scn_phys->scn_bookmark.zb_objset == ds->ds_object) {
	if (ds->ds_is_snapshot) {
	/*
	* Note:
	* - scn_cur_{min,max}_txg stays the same.
	* - Setting the flag is not really necessary if
	* scn_cur_max_txg == scn_max_txg, because there
	* is nothing after this snapshot that we care
	* about. However, we set it anyway and then
	* ignore it when we retraverse it in
	* dsl_scan_visitds().
	*/
	scn_phys->scn_bookmark.zb_objset =
	dsl_dataset_phys(ds)->ds_next_snap_obj;
	zfs_dbgmsg("destroying ds %llu on %s; currently "
	"traversing; reset zb_objset to %llu",
	(u_longlong_t)ds->ds_object,
	ds->ds_dir->dd_pool->dp_spa->spa_name,
	(u_longlong_t)dsl_dataset_phys(ds)->
	ds_next_snap_obj);
	scn_phys->scn_flags \|= DSF_VISIT_DS_AGAIN;
	} else {
	SET_BOOKMARK(&scn_phys->scn_bookmark,
	ZB_DESTROYED_OBJSET, 0, 0, 0);
	zfs_dbgmsg("destroying ds %llu on %s; currently "
	"traversing; reset bookmark to -1,0,0,0",
	(u_longlong_t)ds->ds_object,
	ds->ds_dir->dd_pool->dp_spa->spa_name);
	}
	}
	}

	/*
	* Invoked when a dataset is destroyed. We need to make sure that:
	*
	* 1) If it is the dataset that was currently being scanned, we write
	* a new dsl_scan_phys_t and marking the objset reference in it
	* as destroyed.
	* 2) Remove it from the work queue, if it was present.
	*
	* If the dataset was actually a snapshot, instead of marking the dataset
	* as destroyed, we instead substitute the next snapshot in line.
	*/
	void
	dsl_scan_ds_destroyed(dsl_dataset_t ds, dmu_tx_t tx)
	{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg;

	if (!dsl_scan_is_running(scn))
	return;

	ds_destroyed_scn_phys(ds, &scn->scn_phys);
	ds_destroyed_scn_phys(ds, &scn->scn_phys_cached);

	if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
	scan_ds_queue_remove(scn, ds->ds_object);
	if (ds->ds_is_snapshot)
	scan_ds_queue_insert(scn,
	dsl_dataset_phys(ds)->ds_next_snap_obj, mintxg);
	}

	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	ds->ds_object, &mintxg) == 0) {
	ASSERT3U(dsl_dataset_phys(ds)->ds_num_children, <=, 1);
	VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
	if (ds->ds_is_snapshot) {
	/*
	* We keep the same mintxg; it could be >
	* ds_creation_txg if the previous snapshot was
	* deleted too.
	*/
	VERIFY(zap_add_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj,
	dsl_dataset_phys(ds)->ds_next_snap_obj,
	mintxg, tx) == 0);
	zfs_dbgmsg("destroying ds %llu on %s; in queue; "
	"replacing with %llu",
	(u_longlong_t)ds->ds_object,
	dp->dp_spa->spa_name,
	(u_longlong_t)dsl_dataset_phys(ds)->
	ds_next_snap_obj);
	} else {
	zfs_dbgmsg("destroying ds %llu on %s; in queue; "
	"removing",
	(u_longlong_t)ds->ds_object,
	dp->dp_spa->spa_name);
	}
	}

	/*
	* dsl_scan_sync() should be called after this, and should sync
	* out our changed state, but just to be safe, do it here.
	*/
	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
	}

	static void
	ds_snapshotted_bookmark(dsl_dataset_t ds, zbookmark_phys_t scn_bookmark)
	{
	if (scn_bookmark->zb_objset == ds->ds_object) {
	scn_bookmark->zb_objset =
	dsl_dataset_phys(ds)->ds_prev_snap_obj;
	zfs_dbgmsg("snapshotting ds %llu on %s; currently traversing; "
	"reset zb_objset to %llu",
	(u_longlong_t)ds->ds_object,
	ds->ds_dir->dd_pool->dp_spa->spa_name,
	(u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
	}
	}

	/*
	* Called when a dataset is snapshotted. If we were currently traversing
	* this snapshot, we reset our bookmark to point at the newly created
	* snapshot. We also modify our work queue to remove the old snapshot and
	* replace with the new one.
	*/
	void
	dsl_scan_ds_snapshotted(dsl_dataset_t ds, dmu_tx_t tx)
	{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg;

	if (!dsl_scan_is_running(scn))
	return;

	ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);

	ds_snapshotted_bookmark(ds, &scn->scn_phys.scn_bookmark);
	ds_snapshotted_bookmark(ds, &scn->scn_phys_cached.scn_bookmark);

	if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
	scan_ds_queue_remove(scn, ds->ds_object);
	scan_ds_queue_insert(scn,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg);
	}

	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	ds->ds_object, &mintxg) == 0) {
	VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
	VERIFY(zap_add_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg, tx) == 0);
	zfs_dbgmsg("snapshotting ds %llu on %s; in queue; "
	"replacing with %llu",
	(u_longlong_t)ds->ds_object,
	dp->dp_spa->spa_name,
	(u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
	}

	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
	}

	static void
	ds_clone_swapped_bookmark(dsl_dataset_t ds1, dsl_dataset_t ds2,
	zbookmark_phys_t *scn_bookmark)
	{
	if (scn_bookmark->zb_objset == ds1->ds_object) {
	scn_bookmark->zb_objset = ds2->ds_object;
	zfs_dbgmsg("clone_swap ds %llu on %s; currently traversing; "
	"reset zb_objset to %llu",
	(u_longlong_t)ds1->ds_object,
	ds1->ds_dir->dd_pool->dp_spa->spa_name,
	(u_longlong_t)ds2->ds_object);
	} else if (scn_bookmark->zb_objset == ds2->ds_object) {
	scn_bookmark->zb_objset = ds1->ds_object;
	zfs_dbgmsg("clone_swap ds %llu on %s; currently traversing; "
	"reset zb_objset to %llu",
	(u_longlong_t)ds2->ds_object,
	ds2->ds_dir->dd_pool->dp_spa->spa_name,
	(u_longlong_t)ds1->ds_object);
	}
	}

	/*
	* Called when an origin dataset and its clone are swapped. If we were
	* currently traversing the dataset, we need to switch to traversing the
	* newly promoted clone.
	*/
	void
	dsl_scan_ds_clone_swapped(dsl_dataset_t ds1, dsl_dataset_t ds2, dmu_tx_t *tx)
	{
	dsl_pool_t *dp = ds1->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg1, mintxg2;
	boolean_t ds1_queued, ds2_queued;

	if (!dsl_scan_is_running(scn))
	return;

	ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys.scn_bookmark);
	ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys_cached.scn_bookmark);

	/*
	* Handle the in-memory scan queue.
	*/
	ds1_queued = scan_ds_queue_contains(scn, ds1->ds_object, &mintxg1);
	ds2_queued = scan_ds_queue_contains(scn, ds2->ds_object, &mintxg2);

	/* Sanity checking. */
	if (ds1_queued) {
	ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
	ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
	}
	if (ds2_queued) {
	ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
	ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
	}

	if (ds1_queued && ds2_queued) {
	/*
	* If both are queued, we don't need to do anything.
	* The swapping code below would not handle this case correctly,
	* since we can't insert ds2 if it is already there. That's
	* because scan_ds_queue_insert() prohibits a duplicate insert
	* and panics.
	*/
	} else if (ds1_queued) {
	scan_ds_queue_remove(scn, ds1->ds_object);
	scan_ds_queue_insert(scn, ds2->ds_object, mintxg1);
	} else if (ds2_queued) {
	scan_ds_queue_remove(scn, ds2->ds_object);
	scan_ds_queue_insert(scn, ds1->ds_object, mintxg2);
	}

	/*
	* Handle the on-disk scan queue.
	* The on-disk state is an out-of-date version of the in-memory state,
	* so the in-memory and on-disk values for ds1_queued and ds2_queued may
	* be different. Therefore we need to apply the swap logic to the
	* on-disk state independently of the in-memory state.
	*/
	ds1_queued = zap_lookup_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds1->ds_object, &mintxg1) == 0;
	ds2_queued = zap_lookup_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds2->ds_object, &mintxg2) == 0;

	/* Sanity checking. */
	if (ds1_queued) {
	ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
	ASSERT3U(mintxg1, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
	}
	if (ds2_queued) {
	ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
	ASSERT3U(mintxg2, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
	}

	if (ds1_queued && ds2_queued) {
	/*
	* If both are queued, we don't need to do anything.
	* Alternatively, we could check for EEXIST from
	* zap_add_int_key() and back out to the original state, but
	* that would be more work than checking for this case upfront.
	*/
	} else if (ds1_queued) {
	VERIFY3S(0, ==, zap_remove_int(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds1->ds_object, tx));
	VERIFY3S(0, ==, zap_add_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds2->ds_object, mintxg1, tx));
	zfs_dbgmsg("clone_swap ds %llu on %s; in queue; "
	"replacing with %llu",
	(u_longlong_t)ds1->ds_object,
	dp->dp_spa->spa_name,
	(u_longlong_t)ds2->ds_object);
	} else if (ds2_queued) {
	VERIFY3S(0, ==, zap_remove_int(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds2->ds_object, tx));
	VERIFY3S(0, ==, zap_add_int_key(dp->dp_meta_objset,
	scn->scn_phys.scn_queue_obj, ds1->ds_object, mintxg2, tx));
	zfs_dbgmsg("clone_swap ds %llu on %s; in queue; "
	"replacing with %llu",
	(u_longlong_t)ds2->ds_object,
	dp->dp_spa->spa_name,
	(u_longlong_t)ds1->ds_object);
	}

	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
	}

	static int
	enqueue_clones_cb(dsl_pool_t dp, dsl_dataset_t hds, void *arg)
	{
	uint64_t originobj = (uint64_t )arg;
	dsl_dataset_t *ds;
	int err;
	dsl_scan_t *scn = dp->dp_scan;

	if (dsl_dir_phys(hds->ds_dir)->dd_origin_obj != originobj)
	return (0);

	err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
	if (err)
	return (err);

	while (dsl_dataset_phys(ds)->ds_prev_snap_obj != originobj) {
	dsl_dataset_t *prev;
	err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);

	dsl_dataset_rele(ds, FTAG);
	if (err)
	return (err);
	ds = prev;
	}
	scan_ds_queue_insert(scn, ds->ds_object,
	dsl_dataset_phys(ds)->ds_prev_snap_txg);
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	static void
	dsl_scan_visitds(dsl_scan_t scn, uint64_t dsobj, dmu_tx_t tx)
	{
	dsl_pool_t *dp = scn->scn_dp;
	dsl_dataset_t *ds;

	VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));

	if (scn->scn_phys.scn_cur_min_txg >=
	scn->scn_phys.scn_max_txg) {
	/*
	* This can happen if this snapshot was created after the
	* scan started, and we already completed a previous snapshot
	* that was created after the scan started. This snapshot
	* only references blocks with:
	*
	* birth < our ds_creation_txg
	* cur_min_txg is no less than ds_creation_txg.
	* We have already visited these blocks.
	* or
	* birth > scn_max_txg
	* The scan requested not to visit these blocks.
	*
	* Subsequent snapshots (and clones) can reference our
	* blocks, or blocks with even higher birth times.
	* Therefore we do not need to visit them either,
	* so we do not add them to the work queue.
	*
	* Note that checking for cur_min_txg >= cur_max_txg
	* is not sufficient, because in that case we may need to
	* visit subsequent snapshots. This happens when min_txg > 0,
	* which raises cur_min_txg. In this case we will visit
	* this dataset but skip all of its blocks, because the
	* rootbp's birth time is < cur_min_txg. Then we will
	* add the next snapshots/clones to the work queue.
	*/
	char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
	dsl_dataset_name(ds, dsname);
	zfs_dbgmsg("scanning dataset %llu (%s) is unnecessary because "
	"cur_min_txg (%llu) >= max_txg (%llu)",
	(longlong_t)dsobj, dsname,
	(longlong_t)scn->scn_phys.scn_cur_min_txg,
	(longlong_t)scn->scn_phys.scn_max_txg);
	kmem_free(dsname, MAXNAMELEN);

	goto out;
	}

	/*
	* Only the ZIL in the head (non-snapshot) is valid. Even though
	* snapshots can have ZIL block pointers (which may be the same
	* BP as in the head), they must be ignored. In addition, $ORIGIN
	* doesn't have a objset (i.e. its ds_bp is a hole) so we don't
	* need to look for a ZIL in it either. So we traverse the ZIL here,
	* rather than in scan_recurse(), because the regular snapshot
	* block-sharing rules don't apply to it.
	*/
	if (!dsl_dataset_is_snapshot(ds) &&
	(dp->dp_origin_snap == NULL \|\|
	ds->ds_dir != dp->dp_origin_snap->ds_dir)) {
	objset_t *os;
	if (dmu_objset_from_ds(ds, &os) != 0) {
	goto out;
	}
	dsl_scan_zil(dp, &os->os_zil_header);
	}

	/*
	* Iterate over the bps in this ds.
	*/
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	dsl_scan_visit_rootbp(scn, ds, &dsl_dataset_phys(ds)->ds_bp, tx);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
	dsl_dataset_name(ds, dsname);
	zfs_dbgmsg("scanned dataset %llu (%s) with min=%llu max=%llu; "
	"suspending=%u",
	(longlong_t)dsobj, dsname,
	(longlong_t)scn->scn_phys.scn_cur_min_txg,
	(longlong_t)scn->scn_phys.scn_cur_max_txg,
	(int)scn->scn_suspending);
	kmem_free(dsname, ZFS_MAX_DATASET_NAME_LEN);

	if (scn->scn_suspending)
	goto out;

	/*
	* We've finished this pass over this dataset.
	*/

	/*
	* If we did not completely visit this dataset, do another pass.
	*/
	if (scn->scn_phys.scn_flags & DSF_VISIT_DS_AGAIN) {
	zfs_dbgmsg("incomplete pass on %s; visiting again",
	dp->dp_spa->spa_name);
	scn->scn_phys.scn_flags &= ~DSF_VISIT_DS_AGAIN;
	scan_ds_queue_insert(scn, ds->ds_object,
	scn->scn_phys.scn_cur_max_txg);
	goto out;
	}

	/*
	* Add descendant datasets to work queue.
	*/
	if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) {
	scan_ds_queue_insert(scn,
	dsl_dataset_phys(ds)->ds_next_snap_obj,
	dsl_dataset_phys(ds)->ds_creation_txg);
	}
	if (dsl_dataset_phys(ds)->ds_num_children > 1) {
	boolean_t usenext = B_FALSE;
	if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) {
	uint64_t count;
	/*
	* A bug in a previous version of the code could
	* cause upgrade_clones_cb() to not set
	* ds_next_snap_obj when it should, leading to a
	* missing entry. Therefore we can only use the
	* next_clones_obj when its count is correct.
	*/
	int err = zap_count(dp->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_next_clones_obj, &count);
	if (err == 0 &&
	count == dsl_dataset_phys(ds)->ds_num_children - 1)
	usenext = B_TRUE;
	}

	if (usenext) {
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, dp->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_next_clones_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	(void) zap_cursor_advance(&zc)) {
	scan_ds_queue_insert(scn,
	zfs_strtonum(za.za_name, NULL),
	dsl_dataset_phys(ds)->ds_creation_txg);
	}
	zap_cursor_fini(&zc);
	} else {
	VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	enqueue_clones_cb, &ds->ds_object,
	DS_FIND_CHILDREN));
	}
	}

	out:
	dsl_dataset_rele(ds, FTAG);
	}

	static int
	enqueue_cb(dsl_pool_t dp, dsl_dataset_t hds, void *arg)
	{
	(void) arg;
	dsl_dataset_t *ds;
	int err;
	dsl_scan_t *scn = dp->dp_scan;

	err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
	if (err)
	return (err);

	while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
	dsl_dataset_t *prev;
	err = dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
	if (err) {
	dsl_dataset_rele(ds, FTAG);
	return (err);
	}

	/*
	* If this is a clone, we don't need to worry about it for now.
	*/
	if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object) {
	dsl_dataset_rele(ds, FTAG);
	dsl_dataset_rele(prev, FTAG);
	return (0);
	}
	dsl_dataset_rele(ds, FTAG);
	ds = prev;
	}

	scan_ds_queue_insert(scn, ds->ds_object,
	dsl_dataset_phys(ds)->ds_prev_snap_txg);
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	void
	dsl_scan_ddt_entry(dsl_scan_t *scn, enum zio_checksum checksum,
	ddt_entry_t dde, dmu_tx_t tx)
	{
	(void) tx;
	const ddt_key_t *ddk = &dde->dde_key;
	ddt_phys_t *ddp = dde->dde_phys;
	blkptr_t bp;
	zbookmark_phys_t zb = { 0 };

	if (!dsl_scan_is_running(scn))
	return;

	/*
	* This function is special because it is the only thing
	* that can add scan_io_t's to the vdev scan queues from
	* outside dsl_scan_sync(). For the most part this is ok
	* as long as it is called from within syncing context.
	* However, dsl_scan_sync() expects that no new sio's will
	* be added between when all the work for a scan is done
	* and the next txg when the scan is actually marked as
	* completed. This check ensures we do not issue new sio's
	* during this period.
	*/
	if (scn->scn_done_txg != 0)
	return;

	for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0 \|\|
	ddp->ddp_phys_birth > scn->scn_phys.scn_max_txg)
	continue;
	ddt_bp_create(checksum, ddk, ddp, &bp);

	scn->scn_visited_this_txg++;
	scan_funcs[scn->scn_phys.scn_func](scn->scn_dp, &bp, &zb);
	}
	}

	/*
	* Scrub/dedup interaction.
	*
	* If there are N references to a deduped block, we don't want to scrub it
	* N times -- ideally, we should scrub it exactly once.
	*
	* We leverage the fact that the dde's replication class (enum ddt_class)
	* is ordered from highest replication class (DDT_CLASS_DITTO) to lowest
	* (DDT_CLASS_UNIQUE) so that we may walk the DDT in that order.
	*
	* To prevent excess scrubbing, the scrub begins by walking the DDT
	* to find all blocks with refcnt > 1, and scrubs each of these once.
	* Since there are two replication classes which contain blocks with
	* refcnt > 1, we scrub the highest replication class (DDT_CLASS_DITTO) first.
	* Finally the top-down scrub begins, only visiting blocks with refcnt == 1.
	*
	* There would be nothing more to say if a block's refcnt couldn't change
	* during a scrub, but of course it can so we must account for changes
	* in a block's replication class.
	*
	* Here's an example of what can occur:
	*
	* If a block has refcnt > 1 during the DDT scrub phase, but has refcnt == 1
	* when visited during the top-down scrub phase, it will be scrubbed twice.
	* This negates our scrub optimization, but is otherwise harmless.
	*
	* If a block has refcnt == 1 during the DDT scrub phase, but has refcnt > 1
	* on each visit during the top-down scrub phase, it will never be scrubbed.
	* To catch this, ddt_sync_entry() notifies the scrub code whenever a block's
	* reference class transitions to a higher level (i.e DDT_CLASS_UNIQUE to
	* DDT_CLASS_DUPLICATE); if it transitions from refcnt == 1 to refcnt > 1
	* while a scrub is in progress, it scrubs the block right then.
	*/
	static void
	dsl_scan_ddt(dsl_scan_t scn, dmu_tx_t tx)
	{
	ddt_bookmark_t *ddb = &scn->scn_phys.scn_ddt_bookmark;
	ddt_entry_t dde = {{{{0}}}};
	int error;
	uint64_t n = 0;

	while ((error = ddt_walk(scn->scn_dp->dp_spa, ddb, &dde)) == 0) {
	ddt_t *ddt;

	if (ddb->ddb_class > scn->scn_phys.scn_ddt_class_max)
	break;
	dprintf("visiting ddb=%llu/%llu/%llu/%llx\n",
	(longlong_t)ddb->ddb_class,
	(longlong_t)ddb->ddb_type,
	(longlong_t)ddb->ddb_checksum,
	(longlong_t)ddb->ddb_cursor);

	/* There should be no pending changes to the dedup table */
	ddt = scn->scn_dp->dp_spa->spa_ddt[ddb->ddb_checksum];
	ASSERT(avl_first(&ddt->ddt_tree) == NULL);

	dsl_scan_ddt_entry(scn, ddb->ddb_checksum, &dde, tx);
	n++;

	if (dsl_scan_check_suspend(scn, NULL))
	break;
	}

	zfs_dbgmsg("scanned %llu ddt entries on %s with class_max = %u; "
	"suspending=%u", (longlong_t)n, scn->scn_dp->dp_spa->spa_name,
	(int)scn->scn_phys.scn_ddt_class_max, (int)scn->scn_suspending);

	ASSERT(error == 0 \|\| error == ENOENT);
	ASSERT(error != ENOENT \|\|
	ddb->ddb_class > scn->scn_phys.scn_ddt_class_max);
	}

	static uint64_t
	dsl_scan_ds_maxtxg(dsl_dataset_t *ds)
	{
	uint64_t smt = ds->ds_dir->dd_pool->dp_scan->scn_phys.scn_max_txg;
	if (ds->ds_is_snapshot)
	return (MIN(smt, dsl_dataset_phys(ds)->ds_creation_txg));
	return (smt);
	}

	static void
	dsl_scan_visit(dsl_scan_t scn, dmu_tx_t tx)
	{
	scan_ds_t *sds;
	dsl_pool_t *dp = scn->scn_dp;

	if (scn->scn_phys.scn_ddt_bookmark.ddb_class <=
	scn->scn_phys.scn_ddt_class_max) {
	scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
	scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
	dsl_scan_ddt(scn, tx);
	if (scn->scn_suspending)
	return;
	}

	if (scn->scn_phys.scn_bookmark.zb_objset == DMU_META_OBJSET) {
	/* First do the MOS & ORIGIN */

	scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
	scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
	dsl_scan_visit_rootbp(scn, NULL,
	&dp->dp_meta_rootbp, tx);
	spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
	if (scn->scn_suspending)
	return;

	if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB) {
	VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	enqueue_cb, NULL, DS_FIND_CHILDREN));
	} else {
	dsl_scan_visitds(scn,
	dp->dp_origin_snap->ds_object, tx);
	}
	ASSERT(!scn->scn_suspending);
	} else if (scn->scn_phys.scn_bookmark.zb_objset !=
	ZB_DESTROYED_OBJSET) {
	uint64_t dsobj = scn->scn_phys.scn_bookmark.zb_objset;
	/*
	* If we were suspended, continue from here. Note if the
	* ds we were suspended on was deleted, the zb_objset may
	* be -1, so we will skip this and find a new objset
	* below.
	*/
	dsl_scan_visitds(scn, dsobj, tx);
	if (scn->scn_suspending)
	return;
	}

	/*
	* In case we suspended right at the end of the ds, zero the
	* bookmark so we don't think that we're still trying to resume.
	*/
	memset(&scn->scn_phys.scn_bookmark, 0, sizeof (zbookmark_phys_t));

	/*
	* Keep pulling things out of the dataset avl queue. Updates to the
	* persistent zap-object-as-queue happen only at checkpoints.
	*/
	while ((sds = avl_first(&scn->scn_queue)) != NULL) {
	dsl_dataset_t *ds;
	uint64_t dsobj = sds->sds_dsobj;
	uint64_t txg = sds->sds_txg;

	/* dequeue and free the ds from the queue */
	scan_ds_queue_remove(scn, dsobj);
	sds = NULL;

	/* set up min / max txg */
	VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
	if (txg != 0) {
	scn->scn_phys.scn_cur_min_txg =
	MAX(scn->scn_phys.scn_min_txg, txg);
	} else {
	scn->scn_phys.scn_cur_min_txg =
	MAX(scn->scn_phys.scn_min_txg,
	dsl_dataset_phys(ds)->ds_prev_snap_txg);
	}
	scn->scn_phys.scn_cur_max_txg = dsl_scan_ds_maxtxg(ds);
	dsl_dataset_rele(ds, FTAG);

	dsl_scan_visitds(scn, dsobj, tx);
	if (scn->scn_suspending)
	return;
	}

	/* No more objsets to fetch, we're done */
	scn->scn_phys.scn_bookmark.zb_objset = ZB_DESTROYED_OBJSET;
	ASSERT0(scn->scn_suspending);
	}

	static uint64_t
	dsl_scan_count_data_disks(vdev_t *rvd)
	{
	uint64_t i, leaves = 0;

	for (i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];
	if (vd->vdev_islog \|\| vd->vdev_isspare \|\| vd->vdev_isl2cache)
	continue;
	leaves += vdev_get_ndisks(vd) - vdev_get_nparity(vd);
	}
	return (leaves);
	}

	static void
	scan_io_queues_update_zio_stats(dsl_scan_io_queue_t q, const blkptr_t bp)
	{
	int i;
	uint64_t cur_size = 0;

	for (i = 0; i < BP_GET_NDVAS(bp); i++) {
	cur_size += DVA_GET_ASIZE(&bp->blk_dva[i]);
	}

	q->q_total_zio_size_this_txg += cur_size;
	q->q_zios_this_txg++;
	}

	static void
	scan_io_queues_update_seg_stats(dsl_scan_io_queue_t *q, uint64_t start,
	uint64_t end)
	{
	q->q_total_seg_size_this_txg += end - start;
	q->q_segs_this_txg++;
	}

	static boolean_t
	scan_io_queue_check_suspend(dsl_scan_t *scn)
	{
	/* See comment in dsl_scan_check_suspend() */
	uint64_t curr_time_ns = gethrtime();
	uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
	uint64_t sync_time_ns = curr_time_ns -
	scn->scn_dp->dp_spa->spa_sync_starttime;
	uint64_t dirty_min_bytes = zfs_dirty_data_max *
	zfs_vdev_async_write_active_min_dirty_percent / 100;
	- int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	+ uint_t mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;

	return ((NSEC2MSEC(scan_time_ns) > mintime &&
	(scn->scn_dp->dp_dirty_total >= dirty_min_bytes \|\|
	txg_sync_waiting(scn->scn_dp) \|\|
	NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) \|\|
	spa_shutting_down(scn->scn_dp->dp_spa));
	}

	/*
	* Given a list of scan_io_t's in io_list, this issues the I/Os out to
	* disk. This consumes the io_list and frees the scan_io_t's. This is
	* called when emptying queues, either when we're up against the memory
	* limit or when we have finished scanning. Returns B_TRUE if we stopped
	* processing the list before we finished. Any sios that were not issued
	* will remain in the io_list.
	*/
	static boolean_t
	scan_io_queue_issue(dsl_scan_io_queue_t queue, list_t io_list)
	{
	dsl_scan_t *scn = queue->q_scn;
	scan_io_t *sio;
	boolean_t suspended = B_FALSE;

	while ((sio = list_head(io_list)) != NULL) {
	blkptr_t bp;

	if (scan_io_queue_check_suspend(scn)) {
	suspended = B_TRUE;
	break;
	}

	sio2bp(sio, &bp);
	scan_exec_io(scn->scn_dp, &bp, sio->sio_flags,
	&sio->sio_zb, queue);
	(void) list_remove_head(io_list);
	scan_io_queues_update_zio_stats(queue, &bp);
	sio_free(sio);
	}
	return (suspended);
	}

	/*
	* This function removes sios from an IO queue which reside within a given
	* range_seg_t and inserts them (in offset order) into a list. Note that
	* we only ever return a maximum of 32 sios at once. If there are more sios
	* to process within this segment that did not make it onto the list we
	* return B_TRUE and otherwise B_FALSE.
	*/
	static boolean_t
	scan_io_queue_gather(dsl_scan_io_queue_t queue, range_seg_t rs, list_t *list)
	{
	scan_io_t srch_sio, sio, *next_sio;
	avl_index_t idx;
	uint_t num_sios = 0;
	int64_t bytes_issued = 0;

	ASSERT(rs != NULL);
	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));

	srch_sio = sio_alloc(1);
	srch_sio->sio_nr_dvas = 1;
	SIO_SET_OFFSET(srch_sio, rs_get_start(rs, queue->q_exts_by_addr));

	/*
	* The exact start of the extent might not contain any matching zios,
	* so if that's the case, examine the next one in the tree.
	*/
	sio = avl_find(&queue->q_sios_by_addr, srch_sio, &idx);
	sio_free(srch_sio);

	if (sio == NULL)
	sio = avl_nearest(&queue->q_sios_by_addr, idx, AVL_AFTER);

	while (sio != NULL && SIO_GET_OFFSET(sio) < rs_get_end(rs,
	queue->q_exts_by_addr) && num_sios <= 32) {
	ASSERT3U(SIO_GET_OFFSET(sio), >=, rs_get_start(rs,
	queue->q_exts_by_addr));
	ASSERT3U(SIO_GET_END_OFFSET(sio), <=, rs_get_end(rs,
	queue->q_exts_by_addr));

	next_sio = AVL_NEXT(&queue->q_sios_by_addr, sio);
	avl_remove(&queue->q_sios_by_addr, sio);
	if (avl_is_empty(&queue->q_sios_by_addr))
	atomic_add_64(&queue->q_scn->scn_queues_pending, -1);
	queue->q_sio_memused -= SIO_GET_MUSED(sio);

	bytes_issued += SIO_GET_ASIZE(sio);
	num_sios++;
	list_insert_tail(list, sio);
	sio = next_sio;
	}

	/*
	* We limit the number of sios we process at once to 32 to avoid
	* biting off more than we can chew. If we didn't take everything
	* in the segment we update it to reflect the work we were able to
	* complete. Otherwise, we remove it from the range tree entirely.
	*/
	if (sio != NULL && SIO_GET_OFFSET(sio) < rs_get_end(rs,
	queue->q_exts_by_addr)) {
	range_tree_adjust_fill(queue->q_exts_by_addr, rs,
	-bytes_issued);
	range_tree_resize_segment(queue->q_exts_by_addr, rs,
	SIO_GET_OFFSET(sio), rs_get_end(rs,
	queue->q_exts_by_addr) - SIO_GET_OFFSET(sio));
	queue->q_last_ext_addr = SIO_GET_OFFSET(sio);
	return (B_TRUE);
	} else {
	uint64_t rstart = rs_get_start(rs, queue->q_exts_by_addr);
	uint64_t rend = rs_get_end(rs, queue->q_exts_by_addr);
	range_tree_remove(queue->q_exts_by_addr, rstart, rend - rstart);
	queue->q_last_ext_addr = -1;
	return (B_FALSE);
	}
	}

	/*
	* This is called from the queue emptying thread and selects the next
	* extent from which we are to issue I/Os. The behavior of this function
	* depends on the state of the scan, the current memory consumption and
	* whether or not we are performing a scan shutdown.
	* 1) We select extents in an elevator algorithm (LBA-order) if the scan
	* needs to perform a checkpoint
	* 2) We select the largest available extent if we are up against the
	* memory limit.
	* 3) Otherwise we don't select any extents.
	*/
	static range_seg_t *
	scan_io_queue_fetch_ext(dsl_scan_io_queue_t *queue)
	{
	dsl_scan_t *scn = queue->q_scn;
	range_tree_t *rt = queue->q_exts_by_addr;

	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
	ASSERT(scn->scn_is_sorted);

	if (!scn->scn_checkpointing && !scn->scn_clearing)
	return (NULL);

	/*
	* During normal clearing, we want to issue our largest segments
	* first, keeping IO as sequential as possible, and leaving the
	* smaller extents for later with the hope that they might eventually
	* grow to larger sequential segments. However, when the scan is
	* checkpointing, no new extents will be added to the sorting queue,
	* so the way we are sorted now is as good as it will ever get.
	* In this case, we instead switch to issuing extents in LBA order.
	*/
	if ((zfs_scan_issue_strategy < 1 && scn->scn_checkpointing) \|\|
	zfs_scan_issue_strategy == 1)
	return (range_tree_first(rt));

	/*
	* Try to continue previous extent if it is not completed yet. After
	* shrink in scan_io_queue_gather() it may no longer be the best, but
	* otherwise we leave shorter remnant every txg.
	*/
	uint64_t start;
	- uint64_t size = 1 << rt->rt_shift;
	+ uint64_t size = 1ULL << rt->rt_shift;
	range_seg_t *addr_rs;
	if (queue->q_last_ext_addr != -1) {
	start = queue->q_last_ext_addr;
	addr_rs = range_tree_find(rt, start, size);
	if (addr_rs != NULL)
	return (addr_rs);
	}

	/*
	* Nothing to continue, so find new best extent.
	*/
	uint64_t *v = zfs_btree_first(&queue->q_exts_by_size, NULL);
	if (v == NULL)
	return (NULL);
	queue->q_last_ext_addr = start = *v << rt->rt_shift;

	/*
	* We need to get the original entry in the by_addr tree so we can
	* modify it.
	*/
	addr_rs = range_tree_find(rt, start, size);
	ASSERT3P(addr_rs, !=, NULL);
	ASSERT3U(rs_get_start(addr_rs, rt), ==, start);
	ASSERT3U(rs_get_end(addr_rs, rt), >, start);
	return (addr_rs);
	}

	static void
	scan_io_queues_run_one(void *arg)
	{
	dsl_scan_io_queue_t *queue = arg;
	kmutex_t *q_lock = &queue->q_vd->vdev_scan_io_queue_lock;
	boolean_t suspended = B_FALSE;
	range_seg_t *rs;
	scan_io_t *sio;
	zio_t *zio;
	list_t sio_list;

	ASSERT(queue->q_scn->scn_is_sorted);

	list_create(&sio_list, sizeof (scan_io_t),
	offsetof(scan_io_t, sio_nodes.sio_list_node));
	zio = zio_null(queue->q_scn->scn_zio_root, queue->q_scn->scn_dp->dp_spa,
	NULL, NULL, NULL, ZIO_FLAG_CANFAIL);
	mutex_enter(q_lock);
	queue->q_zio = zio;

	/* Calculate maximum in-flight bytes for this vdev. */
	queue->q_maxinflight_bytes = MAX(1, zfs_scan_vdev_limit *
	(vdev_get_ndisks(queue->q_vd) - vdev_get_nparity(queue->q_vd)));

	/* reset per-queue scan statistics for this txg */
	queue->q_total_seg_size_this_txg = 0;
	queue->q_segs_this_txg = 0;
	queue->q_total_zio_size_this_txg = 0;
	queue->q_zios_this_txg = 0;

	/* loop until we run out of time or sios */
	while ((rs = scan_io_queue_fetch_ext(queue)) != NULL) {
	uint64_t seg_start = 0, seg_end = 0;
	boolean_t more_left;

	ASSERT(list_is_empty(&sio_list));

	/* loop while we still have sios left to process in this rs */
	do {
	scan_io_t first_sio, last_sio;

	/*
	* We have selected which extent needs to be
	* processed next. Gather up the corresponding sios.
	*/
	more_left = scan_io_queue_gather(queue, rs, &sio_list);
	ASSERT(!list_is_empty(&sio_list));
	first_sio = list_head(&sio_list);
	last_sio = list_tail(&sio_list);

	seg_end = SIO_GET_END_OFFSET(last_sio);
	if (seg_start == 0)
	seg_start = SIO_GET_OFFSET(first_sio);

	/*
	* Issuing sios can take a long time so drop the
	* queue lock. The sio queue won't be updated by
	* other threads since we're in syncing context so
	* we can be sure that our trees will remain exactly
	* as we left them.
	*/
	mutex_exit(q_lock);
	suspended = scan_io_queue_issue(queue, &sio_list);
	mutex_enter(q_lock);

	if (suspended)
	break;
	} while (more_left);

	/* update statistics for debugging purposes */
	scan_io_queues_update_seg_stats(queue, seg_start, seg_end);

	if (suspended)
	break;
	}

	/*
	* If we were suspended in the middle of processing,
	* requeue any unfinished sios and exit.
	*/
	while ((sio = list_head(&sio_list)) != NULL) {
	list_remove(&sio_list, sio);
	scan_io_queue_insert_impl(queue, sio);
	}

	queue->q_zio = NULL;
	mutex_exit(q_lock);
	zio_nowait(zio);
	list_destroy(&sio_list);
	}

	/*
	* Performs an emptying run on all scan queues in the pool. This just
	* punches out one thread per top-level vdev, each of which processes
	* only that vdev's scan queue. We can parallelize the I/O here because
	* we know that each queue's I/Os only affect its own top-level vdev.
	*
	* This function waits for the queue runs to complete, and must be
	* called from dsl_scan_sync (or in general, syncing context).
	*/
	static void
	scan_io_queues_run(dsl_scan_t *scn)
	{
	spa_t *spa = scn->scn_dp->dp_spa;

	ASSERT(scn->scn_is_sorted);
	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (scn->scn_queues_pending == 0)
	return;

	if (scn->scn_taskq == NULL) {
	int nthreads = spa->spa_root_vdev->vdev_children;

	/*
	* We need to make this taskq always execute as many
	* threads in parallel as we have top-level vdevs and no
	* less, otherwise strange serialization of the calls to
	* scan_io_queues_run_one can occur during spa_sync runs
	* and that significantly impacts performance.
	*/
	scn->scn_taskq = taskq_create("dsl_scan_iss", nthreads,
	minclsyspri, nthreads, nthreads, TASKQ_PREPOPULATE);
	}

	for (uint64_t i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[i];

	mutex_enter(&vd->vdev_scan_io_queue_lock);
	if (vd->vdev_scan_io_queue != NULL) {
	VERIFY(taskq_dispatch(scn->scn_taskq,
	scan_io_queues_run_one, vd->vdev_scan_io_queue,
	TQ_SLEEP) != TASKQID_INVALID);
	}
	mutex_exit(&vd->vdev_scan_io_queue_lock);
	}

	/*
	* Wait for the queues to finish issuing their IOs for this run
	* before we return. There may still be IOs in flight at this
	* point.
	*/
	taskq_wait(scn->scn_taskq);
	}

	static boolean_t
	dsl_scan_async_block_should_pause(dsl_scan_t *scn)
	{
	uint64_t elapsed_nanosecs;

	if (zfs_recover)
	return (B_FALSE);

	if (zfs_async_block_max_blocks != 0 &&
	scn->scn_visited_this_txg >= zfs_async_block_max_blocks) {
	return (B_TRUE);
	}

	if (zfs_max_async_dedup_frees != 0 &&
	scn->scn_dedup_frees_this_txg >= zfs_max_async_dedup_frees) {
	return (B_TRUE);
	}

	elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
	return (elapsed_nanosecs / NANOSEC > zfs_txg_timeout \|\|
	(NSEC2MSEC(elapsed_nanosecs) > scn->scn_async_block_min_time_ms &&
	txg_sync_waiting(scn->scn_dp)) \|\|
	spa_shutting_down(scn->scn_dp->dp_spa));
	}

	static int
	dsl_scan_free_block_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	dsl_scan_t *scn = arg;

	if (!scn->scn_is_bptree \|\|
	(BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_OBJSET)) {
	if (dsl_scan_async_block_should_pause(scn))
	return (SET_ERROR(ERESTART));
	}

	zio_nowait(zio_free_sync(scn->scn_zio_root, scn->scn_dp->dp_spa,
	dmu_tx_get_txg(tx), bp, 0));
	dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
	-bp_get_dsize_sync(scn->scn_dp->dp_spa, bp),
	-BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
	scn->scn_visited_this_txg++;
	if (BP_GET_DEDUP(bp))
	scn->scn_dedup_frees_this_txg++;
	return (0);
	}

	static void
	dsl_scan_update_stats(dsl_scan_t *scn)
	{
	spa_t *spa = scn->scn_dp->dp_spa;
	uint64_t i;
	uint64_t seg_size_total = 0, zio_size_total = 0;
	uint64_t seg_count_total = 0, zio_count_total = 0;

	for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
	dsl_scan_io_queue_t *queue = vd->vdev_scan_io_queue;

	if (queue == NULL)
	continue;

	seg_size_total += queue->q_total_seg_size_this_txg;
	zio_size_total += queue->q_total_zio_size_this_txg;
	seg_count_total += queue->q_segs_this_txg;
	zio_count_total += queue->q_zios_this_txg;
	}

	if (seg_count_total == 0 \|\| zio_count_total == 0) {
	scn->scn_avg_seg_size_this_txg = 0;
	scn->scn_avg_zio_size_this_txg = 0;
	scn->scn_segs_this_txg = 0;
	scn->scn_zios_this_txg = 0;
	return;
	}

	scn->scn_avg_seg_size_this_txg = seg_size_total / seg_count_total;
	scn->scn_avg_zio_size_this_txg = zio_size_total / zio_count_total;
	scn->scn_segs_this_txg = seg_count_total;
	scn->scn_zios_this_txg = zio_count_total;
	}

	static int
	bpobj_dsl_scan_free_block_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(!bp_freed);
	return (dsl_scan_free_block_cb(arg, bp, tx));
	}

	static int
	dsl_scan_obsolete_block_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(!bp_freed);
	dsl_scan_t *scn = arg;
	const dva_t *dva = &bp->blk_dva[0];

	if (dsl_scan_async_block_should_pause(scn))
	return (SET_ERROR(ERESTART));

	spa_vdev_indirect_mark_obsolete(scn->scn_dp->dp_spa,
	DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva),
	DVA_GET_ASIZE(dva), tx);
	scn->scn_visited_this_txg++;
	return (0);
	}

	boolean_t
	dsl_scan_active(dsl_scan_t *scn)
	{
	spa_t *spa = scn->scn_dp->dp_spa;
	uint64_t used = 0, comp, uncomp;
	boolean_t clones_left;

	if (spa->spa_load_state != SPA_LOAD_NONE)
	return (B_FALSE);
	if (spa_shutting_down(spa))
	return (B_FALSE);
	if ((dsl_scan_is_running(scn) && !dsl_scan_is_paused_scrub(scn)) \|\|
	(scn->scn_async_destroying && !scn->scn_async_stalled))
	return (B_TRUE);

	if (spa_version(scn->scn_dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
	(void) bpobj_space(&scn->scn_dp->dp_free_bpobj,
	&used, &comp, &uncomp);
	}
	clones_left = spa_livelist_delete_check(spa);
	return ((used != 0) \|\| (clones_left));
	}

	static boolean_t
	dsl_scan_check_deferred(vdev_t *vd)
	{
	boolean_t need_resilver = B_FALSE;

	for (int c = 0; c < vd->vdev_children; c++) {
	need_resilver \|=
	dsl_scan_check_deferred(vd->vdev_child[c]);
	}

	if (!vdev_is_concrete(vd) \|\| vd->vdev_aux \|\|
	!vd->vdev_ops->vdev_op_leaf)
	return (need_resilver);

	if (!vd->vdev_resilver_deferred)
	need_resilver = B_TRUE;

	return (need_resilver);
	}

	static boolean_t
	dsl_scan_need_resilver(spa_t spa, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	vdev_t *vd;

	vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));

	if (vd->vdev_ops == &vdev_indirect_ops) {
	/*
	* The indirect vdev can point to multiple
	* vdevs. For simplicity, always create
	* the resilver zio_t. zio_vdev_io_start()
	* will bypass the child resilver i/o's if
	* they are on vdevs that don't have DTL's.
	*/
	return (B_TRUE);
	}

	if (DVA_GET_GANG(dva)) {
	/*
	* Gang members may be spread across multiple
	* vdevs, so the best estimate we have is the
	* scrub range, which has already been checked.
	* XXX -- it would be better to change our
	* allocation policy to ensure that all
	* gang members reside on the same vdev.
	*/
	return (B_TRUE);
	}

	/*
	* Check if the top-level vdev must resilver this offset.
	* When the offset does not intersect with a dirty leaf DTL
	* then it may be possible to skip the resilver IO. The psize
	* is provided instead of asize to simplify the check for RAIDZ.
	*/
	if (!vdev_dtl_need_resilver(vd, dva, psize, phys_birth))
	return (B_FALSE);

	/*
	* Check that this top-level vdev has a device under it which
	* is resilvering and is not deferred.
	*/
	if (!dsl_scan_check_deferred(vd))
	return (B_FALSE);

	return (B_TRUE);
	}

	static int
	dsl_process_async_destroys(dsl_pool_t dp, dmu_tx_t tx)
	{
	dsl_scan_t *scn = dp->dp_scan;
	spa_t *spa = dp->dp_spa;
	int err = 0;

	if (spa_suspend_async_destroy(spa))
	return (0);

	if (zfs_free_bpobj_enabled &&
	spa_version(spa) >= SPA_VERSION_DEADLISTS) {
	scn->scn_is_bptree = B_FALSE;
	scn->scn_async_block_min_time_ms = zfs_free_min_time_ms;
	scn->scn_zio_root = zio_root(spa, NULL,
	NULL, ZIO_FLAG_MUSTSUCCEED);
	err = bpobj_iterate(&dp->dp_free_bpobj,
	bpobj_dsl_scan_free_block_cb, scn, tx);
	VERIFY0(zio_wait(scn->scn_zio_root));
	scn->scn_zio_root = NULL;

	if (err != 0 && err != ERESTART)
	zfs_panic_recover("error %u from bpobj_iterate()", err);
	}

	if (err == 0 && spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY)) {
	ASSERT(scn->scn_async_destroying);
	scn->scn_is_bptree = B_TRUE;
	scn->scn_zio_root = zio_root(spa, NULL,
	NULL, ZIO_FLAG_MUSTSUCCEED);
	err = bptree_iterate(dp->dp_meta_objset,
	dp->dp_bptree_obj, B_TRUE, dsl_scan_free_block_cb, scn, tx);
	VERIFY0(zio_wait(scn->scn_zio_root));
	scn->scn_zio_root = NULL;

	if (err == EIO \|\| err == ECKSUM) {
	err = 0;
	} else if (err != 0 && err != ERESTART) {
	zfs_panic_recover("error %u from "
	"traverse_dataset_destroyed()", err);
	}

	if (bptree_is_empty(dp->dp_meta_objset, dp->dp_bptree_obj)) {
	/* finished; deactivate async destroy feature */
	spa_feature_decr(spa, SPA_FEATURE_ASYNC_DESTROY, tx);
	ASSERT(!spa_feature_is_active(spa,
	SPA_FEATURE_ASYNC_DESTROY));
	VERIFY0(zap_remove(dp->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_BPTREE_OBJ, tx));
	VERIFY0(bptree_free(dp->dp_meta_objset,
	dp->dp_bptree_obj, tx));
	dp->dp_bptree_obj = 0;
	scn->scn_async_destroying = B_FALSE;
	scn->scn_async_stalled = B_FALSE;
	} else {
	/*
	* If we didn't make progress, mark the async
	* destroy as stalled, so that we will not initiate
	* a spa_sync() on its behalf. Note that we only
	* check this if we are not finished, because if the
	* bptree had no blocks for us to visit, we can
	* finish without "making progress".
	*/
	scn->scn_async_stalled =
	(scn->scn_visited_this_txg == 0);
	}
	}
	if (scn->scn_visited_this_txg) {
	zfs_dbgmsg("freed %llu blocks in %llums from "
	"free_bpobj/bptree on %s in txg %llu; err=%u",
	(longlong_t)scn->scn_visited_this_txg,
	(longlong_t)
	NSEC2MSEC(gethrtime() - scn->scn_sync_start_time),
	spa->spa_name, (longlong_t)tx->tx_txg, err);
	scn->scn_visited_this_txg = 0;
	scn->scn_dedup_frees_this_txg = 0;

	/*
	* Write out changes to the DDT that may be required as a
	* result of the blocks freed. This ensures that the DDT
	* is clean when a scrub/resilver runs.
	*/
	ddt_sync(spa, tx->tx_txg);
	}
	if (err != 0)
	return (err);
	if (dp->dp_free_dir != NULL && !scn->scn_async_destroying &&
	zfs_free_leak_on_eio &&
	(dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes != 0 \|\|
	dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes != 0 \|\|
	dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes != 0)) {
	/*
	* We have finished background destroying, but there is still
	* some space left in the dp_free_dir. Transfer this leaked
	* space to the dp_leak_dir.
	*/
	if (dp->dp_leak_dir == NULL) {
	rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
	(void) dsl_dir_create_sync(dp, dp->dp_root_dir,
	LEAK_DIR_NAME, tx);
	VERIFY0(dsl_pool_open_special_dir(dp,
	LEAK_DIR_NAME, &dp->dp_leak_dir));
	rrw_exit(&dp->dp_config_rwlock, FTAG);
	}
	dsl_dir_diduse_space(dp->dp_leak_dir, DD_USED_HEAD,
	dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes,
	dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes,
	dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes, tx);
	dsl_dir_diduse_space(dp->dp_free_dir, DD_USED_HEAD,
	-dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes,
	-dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes,
	-dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes, tx);
	}

	if (dp->dp_free_dir != NULL && !scn->scn_async_destroying &&
	!spa_livelist_delete_check(spa)) {
	/* finished; verify that space accounting went to zero */
	ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_used_bytes);
	ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_compressed_bytes);
	ASSERT0(dsl_dir_phys(dp->dp_free_dir)->dd_uncompressed_bytes);
	}

	spa_notify_waiters(spa);

	EQUIV(bpobj_is_open(&dp->dp_obsolete_bpobj),
	0 == zap_contains(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ));
	if (err == 0 && bpobj_is_open(&dp->dp_obsolete_bpobj)) {
	ASSERT(spa_feature_is_active(dp->dp_spa,
	SPA_FEATURE_OBSOLETE_COUNTS));

	scn->scn_is_bptree = B_FALSE;
	scn->scn_async_block_min_time_ms = zfs_obsolete_min_time_ms;
	err = bpobj_iterate(&dp->dp_obsolete_bpobj,
	dsl_scan_obsolete_block_cb, scn, tx);
	if (err != 0 && err != ERESTART)
	zfs_panic_recover("error %u from bpobj_iterate()", err);

	if (bpobj_is_empty(&dp->dp_obsolete_bpobj))
	dsl_pool_destroy_obsolete_bpobj(dp, tx);
	}
	return (0);
	}

	/*
	* This is the primary entry point for scans that is called from syncing
	* context. Scans must happen entirely during syncing context so that we
	* can guarantee that blocks we are currently scanning will not change out
	* from under us. While a scan is active, this function controls how quickly
	* transaction groups proceed, instead of the normal handling provided by
	* txg_sync_thread().
	*/
	void
	dsl_scan_sync(dsl_pool_t dp, dmu_tx_t tx)
	{
	int err = 0;
	dsl_scan_t *scn = dp->dp_scan;
	spa_t *spa = dp->dp_spa;
	state_sync_type_t sync_type = SYNC_OPTIONAL;

	if (spa->spa_resilver_deferred &&
	!spa_feature_is_active(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER))
	spa_feature_incr(spa, SPA_FEATURE_RESILVER_DEFER, tx);

	/*
	* Check for scn_restart_txg before checking spa_load_state, so
	* that we can restart an old-style scan while the pool is being
	* imported (see dsl_scan_init). We also restart scans if there
	* is a deferred resilver and the user has manually disabled
	* deferred resilvers via the tunable.
	*/
	if (dsl_scan_restarting(scn, tx) \|\|
	(spa->spa_resilver_deferred && zfs_resilver_disable_defer)) {
	pool_scan_func_t func = POOL_SCAN_SCRUB;
	dsl_scan_done(scn, B_FALSE, tx);
	if (vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL))
	func = POOL_SCAN_RESILVER;
	zfs_dbgmsg("restarting scan func=%u on %s txg=%llu",
	func, dp->dp_spa->spa_name, (longlong_t)tx->tx_txg);
	dsl_scan_setup_sync(&func, tx);
	}

	/*
	* Only process scans in sync pass 1.
	*/
	if (spa_sync_pass(spa) > 1)
	return;

	/*
	* If the spa is shutting down, then stop scanning. This will
	* ensure that the scan does not dirty any new data during the
	* shutdown phase.
	*/
	if (spa_shutting_down(spa))
	return;

	/*
	* If the scan is inactive due to a stalled async destroy, try again.
	*/
	if (!scn->scn_async_stalled && !dsl_scan_active(scn))
	return;

	/* reset scan statistics */
	scn->scn_visited_this_txg = 0;
	scn->scn_dedup_frees_this_txg = 0;
	scn->scn_holes_this_txg = 0;
	scn->scn_lt_min_this_txg = 0;
	scn->scn_gt_max_this_txg = 0;
	scn->scn_ddt_contained_this_txg = 0;
	scn->scn_objsets_visited_this_txg = 0;
	scn->scn_avg_seg_size_this_txg = 0;
	scn->scn_segs_this_txg = 0;
	scn->scn_avg_zio_size_this_txg = 0;
	scn->scn_zios_this_txg = 0;
	scn->scn_suspending = B_FALSE;
	scn->scn_sync_start_time = gethrtime();
	spa->spa_scrub_active = B_TRUE;

	/*
	* First process the async destroys. If we suspend, don't do
	* any scrubbing or resilvering. This ensures that there are no
	* async destroys while we are scanning, so the scan code doesn't
	* have to worry about traversing it. It is also faster to free the
	* blocks than to scrub them.
	*/
	err = dsl_process_async_destroys(dp, tx);
	if (err != 0)
	return;

	if (!dsl_scan_is_running(scn) \|\| dsl_scan_is_paused_scrub(scn))
	return;

	/*
	* Wait a few txgs after importing to begin scanning so that
	* we can get the pool imported quickly.
	*/
	if (spa->spa_syncing_txg < spa->spa_first_txg + SCAN_IMPORT_WAIT_TXGS)
	return;

	/*
	* zfs_scan_suspend_progress can be set to disable scan progress.
	* We don't want to spin the txg_sync thread, so we add a delay
	* here to simulate the time spent doing a scan. This is mostly
	* useful for testing and debugging.
	*/
	if (zfs_scan_suspend_progress) {
	uint64_t scan_time_ns = gethrtime() - scn->scn_sync_start_time;
	- int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	- zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;
	+ uint_t mintime = (scn->scn_phys.scn_func ==
	+ POOL_SCAN_RESILVER) ? zfs_resilver_min_time_ms :
	+ zfs_scrub_min_time_ms;

	while (zfs_scan_suspend_progress &&
	!txg_sync_waiting(scn->scn_dp) &&
	!spa_shutting_down(scn->scn_dp->dp_spa) &&
	NSEC2MSEC(scan_time_ns) < mintime) {
	delay(hz);
	scan_time_ns = gethrtime() - scn->scn_sync_start_time;
	}
	return;
	}

	/*
	* It is possible to switch from unsorted to sorted at any time,
	* but afterwards the scan will remain sorted unless reloaded from
	* a checkpoint after a reboot.
	*/
	if (!zfs_scan_legacy) {
	scn->scn_is_sorted = B_TRUE;
	if (scn->scn_last_checkpoint == 0)
	scn->scn_last_checkpoint = ddi_get_lbolt();
	}

	/*
	* For sorted scans, determine what kind of work we will be doing
	* this txg based on our memory limitations and whether or not we
	* need to perform a checkpoint.
	*/
	if (scn->scn_is_sorted) {
	/*
	* If we are over our checkpoint interval, set scn_clearing
	* so that we can begin checkpointing immediately. The
	* checkpoint allows us to save a consistent bookmark
	* representing how much data we have scrubbed so far.
	* Otherwise, use the memory limit to determine if we should
	* scan for metadata or start issue scrub IOs. We accumulate
	* metadata until we hit our hard memory limit at which point
	* we issue scrub IOs until we are at our soft memory limit.
	*/
	if (scn->scn_checkpointing \|\|
	ddi_get_lbolt() - scn->scn_last_checkpoint >
	SEC_TO_TICK(zfs_scan_checkpoint_intval)) {
	if (!scn->scn_checkpointing)
	zfs_dbgmsg("begin scan checkpoint for %s",
	spa->spa_name);

	scn->scn_checkpointing = B_TRUE;
	scn->scn_clearing = B_TRUE;
	} else {
	boolean_t should_clear = dsl_scan_should_clear(scn);
	if (should_clear && !scn->scn_clearing) {
	zfs_dbgmsg("begin scan clearing for %s",
	spa->spa_name);
	scn->scn_clearing = B_TRUE;
	} else if (!should_clear && scn->scn_clearing) {
	zfs_dbgmsg("finish scan clearing for %s",
	spa->spa_name);
	scn->scn_clearing = B_FALSE;
	}
	}
	} else {
	ASSERT0(scn->scn_checkpointing);
	ASSERT0(scn->scn_clearing);
	}

	if (!scn->scn_clearing && scn->scn_done_txg == 0) {
	/* Need to scan metadata for more blocks to scrub */
	dsl_scan_phys_t *scnp = &scn->scn_phys;
	taskqid_t prefetch_tqid;

	/*
	* Recalculate the max number of in-flight bytes for pool-wide
	* scanning operations (minimum 1MB). Limits for the issuing
	* phase are done per top-level vdev and are handled separately.
	*/
	scn->scn_maxinflight_bytes = MAX(zfs_scan_vdev_limit *
	dsl_scan_count_data_disks(spa->spa_root_vdev), 1ULL << 20);

	if (scnp->scn_ddt_bookmark.ddb_class <=
	scnp->scn_ddt_class_max) {
	ASSERT(ZB_IS_ZERO(&scnp->scn_bookmark));
	zfs_dbgmsg("doing scan sync for %s txg %llu; "
	"ddt bm=%llu/%llu/%llu/%llx",
	spa->spa_name,
	(longlong_t)tx->tx_txg,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_class,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_type,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_checksum,
	(longlong_t)scnp->scn_ddt_bookmark.ddb_cursor);
	} else {
	zfs_dbgmsg("doing scan sync for %s txg %llu; "
	"bm=%llu/%llu/%llu/%llu",
	spa->spa_name,
	(longlong_t)tx->tx_txg,
	(longlong_t)scnp->scn_bookmark.zb_objset,
	(longlong_t)scnp->scn_bookmark.zb_object,
	(longlong_t)scnp->scn_bookmark.zb_level,
	(longlong_t)scnp->scn_bookmark.zb_blkid);
	}

	scn->scn_zio_root = zio_root(dp->dp_spa, NULL,
	NULL, ZIO_FLAG_CANFAIL);

	scn->scn_prefetch_stop = B_FALSE;
	prefetch_tqid = taskq_dispatch(dp->dp_sync_taskq,
	dsl_scan_prefetch_thread, scn, TQ_SLEEP);
	ASSERT(prefetch_tqid != TASKQID_INVALID);

	dsl_pool_config_enter(dp, FTAG);
	dsl_scan_visit(scn, tx);
	dsl_pool_config_exit(dp, FTAG);

	mutex_enter(&dp->dp_spa->spa_scrub_lock);
	scn->scn_prefetch_stop = B_TRUE;
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&dp->dp_spa->spa_scrub_lock);

	taskq_wait_id(dp->dp_sync_taskq, prefetch_tqid);
	(void) zio_wait(scn->scn_zio_root);
	scn->scn_zio_root = NULL;

	zfs_dbgmsg("scan visited %llu blocks of %s in %llums "
	"(%llu os's, %llu holes, %llu < mintxg, "
	"%llu in ddt, %llu > maxtxg)",
	(longlong_t)scn->scn_visited_this_txg,
	spa->spa_name,
	(longlong_t)NSEC2MSEC(gethrtime() -
	scn->scn_sync_start_time),
	(longlong_t)scn->scn_objsets_visited_this_txg,
	(longlong_t)scn->scn_holes_this_txg,
	(longlong_t)scn->scn_lt_min_this_txg,
	(longlong_t)scn->scn_ddt_contained_this_txg,
	(longlong_t)scn->scn_gt_max_this_txg);

	if (!scn->scn_suspending) {
	ASSERT0(avl_numnodes(&scn->scn_queue));
	scn->scn_done_txg = tx->tx_txg + 1;
	if (scn->scn_is_sorted) {
	scn->scn_checkpointing = B_TRUE;
	scn->scn_clearing = B_TRUE;
	}
	zfs_dbgmsg("scan complete for %s txg %llu",
	spa->spa_name,
	(longlong_t)tx->tx_txg);
	}
	} else if (scn->scn_is_sorted && scn->scn_queues_pending != 0) {
	ASSERT(scn->scn_clearing);

	/* need to issue scrubbing IOs from per-vdev queues */
	scn->scn_zio_root = zio_root(dp->dp_spa, NULL,
	NULL, ZIO_FLAG_CANFAIL);
	scan_io_queues_run(scn);
	(void) zio_wait(scn->scn_zio_root);
	scn->scn_zio_root = NULL;

	/* calculate and dprintf the current memory usage */
	(void) dsl_scan_should_clear(scn);
	dsl_scan_update_stats(scn);

	zfs_dbgmsg("scan issued %llu blocks for %s (%llu segs) "
	"in %llums (avg_block_size = %llu, avg_seg_size = %llu)",
	(longlong_t)scn->scn_zios_this_txg,
	spa->spa_name,
	(longlong_t)scn->scn_segs_this_txg,
	(longlong_t)NSEC2MSEC(gethrtime() -
	scn->scn_sync_start_time),
	(longlong_t)scn->scn_avg_zio_size_this_txg,
	(longlong_t)scn->scn_avg_seg_size_this_txg);
	} else if (scn->scn_done_txg != 0 && scn->scn_done_txg <= tx->tx_txg) {
	/* Finished with everything. Mark the scrub as complete */
	zfs_dbgmsg("scan issuing complete txg %llu for %s",
	(longlong_t)tx->tx_txg,
	spa->spa_name);
	ASSERT3U(scn->scn_done_txg, !=, 0);
	ASSERT0(spa->spa_scrub_inflight);
	ASSERT0(scn->scn_queues_pending);
	dsl_scan_done(scn, B_TRUE, tx);
	sync_type = SYNC_MANDATORY;
	}

	dsl_scan_sync_state(scn, tx, sync_type);
	}

	static void
	count_block_issued(spa_t spa, const blkptr_t bp, boolean_t all)
	{
	/*
	* Don't count embedded bp's, since we already did the work of
	* scanning these when we scanned the containing block.
	*/
	if (BP_IS_EMBEDDED(bp))
	return;

	/*
	* Update the spa's stats on how many bytes we have issued.
	* Sequential scrubs create a zio for each DVA of the bp. Each
	* of these will include all DVAs for repair purposes, but the
	* zio code will only try the first one unless there is an issue.
	* Therefore, we should only count the first DVA for these IOs.
	*/
	atomic_add_64(&spa->spa_scan_pass_issued,
	all ? BP_GET_ASIZE(bp) : DVA_GET_ASIZE(&bp->blk_dva[0]));
	}

	static void
	count_block(zfs_all_blkstats_t zab, const blkptr_t bp)
	{
	/*
	* If we resume after a reboot, zab will be NULL; don't record
	* incomplete stats in that case.
	*/
	if (zab == NULL)
	return;

	for (int i = 0; i < 4; i++) {
	int l = (i < 2) ? BP_GET_LEVEL(bp) : DN_MAX_LEVELS;
	int t = (i & 1) ? BP_GET_TYPE(bp) : DMU_OT_TOTAL;

	if (t & DMU_OT_NEWTYPE)
	t = DMU_OT_OTHER;
	zfs_blkstat_t *zb = &zab->zab_type[l][t];
	int equal;

	zb->zb_count++;
	zb->zb_asize += BP_GET_ASIZE(bp);
	zb->zb_lsize += BP_GET_LSIZE(bp);
	zb->zb_psize += BP_GET_PSIZE(bp);
	zb->zb_gangs += BP_COUNT_GANG(bp);

	switch (BP_GET_NDVAS(bp)) {
	case 2:
	if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[1]))
	zb->zb_ditto_2_of_2_samevdev++;
	break;
	case 3:
	equal = (DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[1])) +
	(DVA_GET_VDEV(&bp->blk_dva[0]) ==
	DVA_GET_VDEV(&bp->blk_dva[2])) +
	(DVA_GET_VDEV(&bp->blk_dva[1]) ==
	DVA_GET_VDEV(&bp->blk_dva[2]));
	if (equal == 1)
	zb->zb_ditto_2_of_3_samevdev++;
	else if (equal == 3)
	zb->zb_ditto_3_of_3_samevdev++;
	break;
	}
	}
	}

	static void
	scan_io_queue_insert_impl(dsl_scan_io_queue_t queue, scan_io_t sio)
	{
	avl_index_t idx;
	dsl_scan_t *scn = queue->q_scn;

	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));

	if (unlikely(avl_is_empty(&queue->q_sios_by_addr)))
	atomic_add_64(&scn->scn_queues_pending, 1);
	if (avl_find(&queue->q_sios_by_addr, sio, &idx) != NULL) {
	/* block is already scheduled for reading */
	sio_free(sio);
	return;
	}
	avl_insert(&queue->q_sios_by_addr, sio, idx);
	queue->q_sio_memused += SIO_GET_MUSED(sio);
	range_tree_add(queue->q_exts_by_addr, SIO_GET_OFFSET(sio),
	SIO_GET_ASIZE(sio));
	}

	/*
	* Given all the info we got from our metadata scanning process, we
	* construct a scan_io_t and insert it into the scan sorting queue. The
	* I/O must already be suitable for us to process. This is controlled
	* by dsl_scan_enqueue().
	*/
	static void
	scan_io_queue_insert(dsl_scan_io_queue_t queue, const blkptr_t bp, int dva_i,
	int zio_flags, const zbookmark_phys_t *zb)
	{
	scan_io_t *sio = sio_alloc(BP_GET_NDVAS(bp));

	ASSERT0(BP_IS_GANG(bp));
	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));

	bp2sio(bp, sio, dva_i);
	sio->sio_flags = zio_flags;
	sio->sio_zb = *zb;

	queue->q_last_ext_addr = -1;
	scan_io_queue_insert_impl(queue, sio);
	}

	/*
	* Given a set of I/O parameters as discovered by the metadata traversal
	* process, attempts to place the I/O into the sorted queues (if allowed),
	* or immediately executes the I/O.
	*/
	static void
	dsl_scan_enqueue(dsl_pool_t dp, const blkptr_t bp, int zio_flags,
	const zbookmark_phys_t *zb)
	{
	spa_t *spa = dp->dp_spa;

	ASSERT(!BP_IS_EMBEDDED(bp));

	/*
	* Gang blocks are hard to issue sequentially, so we just issue them
	* here immediately instead of queuing them.
	*/
	if (!dp->dp_scan->scn_is_sorted \|\| BP_IS_GANG(bp)) {
	scan_exec_io(dp, bp, zio_flags, zb, NULL);
	return;
	}

	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
	dva_t dva;
	vdev_t *vdev;

	dva = bp->blk_dva[i];
	vdev = vdev_lookup_top(spa, DVA_GET_VDEV(&dva));
	ASSERT(vdev != NULL);

	mutex_enter(&vdev->vdev_scan_io_queue_lock);
	if (vdev->vdev_scan_io_queue == NULL)
	vdev->vdev_scan_io_queue = scan_io_queue_create(vdev);
	ASSERT(dp->dp_scan != NULL);
	scan_io_queue_insert(vdev->vdev_scan_io_queue, bp,
	i, zio_flags, zb);
	mutex_exit(&vdev->vdev_scan_io_queue_lock);
	}
	}

	static int
	dsl_scan_scrub_cb(dsl_pool_t *dp,
	const blkptr_t bp, const zbookmark_phys_t zb)
	{
	dsl_scan_t *scn = dp->dp_scan;
	spa_t *spa = dp->dp_spa;
	uint64_t phys_birth = BP_PHYSICAL_BIRTH(bp);
	size_t psize = BP_GET_PSIZE(bp);
	boolean_t needs_io = B_FALSE;
	int zio_flags = ZIO_FLAG_SCAN_THREAD \| ZIO_FLAG_RAW \| ZIO_FLAG_CANFAIL;

	count_block(dp->dp_blkstats, bp);
	if (phys_birth <= scn->scn_phys.scn_min_txg \|\|
	phys_birth >= scn->scn_phys.scn_max_txg) {
	count_block_issued(spa, bp, B_TRUE);
	return (0);
	}

	/* Embedded BP's have phys_birth==0, so we reject them above. */
	ASSERT(!BP_IS_EMBEDDED(bp));

	ASSERT(DSL_SCAN_IS_SCRUB_RESILVER(scn));
	if (scn->scn_phys.scn_func == POOL_SCAN_SCRUB) {
	zio_flags \|= ZIO_FLAG_SCRUB;
	needs_io = B_TRUE;
	} else {
	ASSERT3U(scn->scn_phys.scn_func, ==, POOL_SCAN_RESILVER);
	zio_flags \|= ZIO_FLAG_RESILVER;
	needs_io = B_FALSE;
	}

	/* If it's an intent log block, failure is expected. */
	if (zb->zb_level == ZB_ZIL_LEVEL)
	zio_flags \|= ZIO_FLAG_SPECULATIVE;

	for (int d = 0; d < BP_GET_NDVAS(bp); d++) {
	const dva_t *dva = &bp->blk_dva[d];

	/*
	* Keep track of how much data we've examined so that
	* zpool(8) status can make useful progress reports.
	*/
	uint64_t asize = DVA_GET_ASIZE(dva);
	scn->scn_phys.scn_examined += asize;
	spa->spa_scan_pass_exam += asize;

	/* if it's a resilver, this may not be in the target range */
	if (!needs_io)
	needs_io = dsl_scan_need_resilver(spa, dva, psize,
	phys_birth);
	}

	if (needs_io && !zfs_no_scrub_io) {
	dsl_scan_enqueue(dp, bp, zio_flags, zb);
	} else {
	count_block_issued(spa, bp, B_TRUE);
	}

	/* do not relocate this block */
	return (0);
	}

	static void
	dsl_scan_scrub_done(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	dsl_scan_io_queue_t *queue = zio->io_private;

	abd_free(zio->io_abd);

	if (queue == NULL) {
	mutex_enter(&spa->spa_scrub_lock);
	ASSERT3U(spa->spa_scrub_inflight, >=, BP_GET_PSIZE(bp));
	spa->spa_scrub_inflight -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);
	} else {
	mutex_enter(&queue->q_vd->vdev_scan_io_queue_lock);
	ASSERT3U(queue->q_inflight_bytes, >=, BP_GET_PSIZE(bp));
	queue->q_inflight_bytes -= BP_GET_PSIZE(bp);
	cv_broadcast(&queue->q_zio_cv);
	mutex_exit(&queue->q_vd->vdev_scan_io_queue_lock);
	}

	if (zio->io_error && (zio->io_error != ECKSUM \|\|
	!(zio->io_flags & ZIO_FLAG_SPECULATIVE))) {
	atomic_inc_64(&spa->spa_dsl_pool->dp_scan->scn_phys.scn_errors);
	}
	}

	/*
	* Given a scanning zio's information, executes the zio. The zio need
	* not necessarily be only sortable, this function simply executes the
	* zio, no matter what it is. The optional queue argument allows the
	* caller to specify that they want per top level vdev IO rate limiting
	* instead of the legacy global limiting.
	*/
	static void
	scan_exec_io(dsl_pool_t dp, const blkptr_t bp, int zio_flags,
	const zbookmark_phys_t zb, dsl_scan_io_queue_t queue)
	{
	spa_t *spa = dp->dp_spa;
	dsl_scan_t *scn = dp->dp_scan;
	size_t size = BP_GET_PSIZE(bp);
	abd_t *data = abd_alloc_for_io(size, B_FALSE);
	zio_t *pio;

	if (queue == NULL) {
	ASSERT3U(scn->scn_maxinflight_bytes, >, 0);
	mutex_enter(&spa->spa_scrub_lock);
	while (spa->spa_scrub_inflight >= scn->scn_maxinflight_bytes)
	cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
	spa->spa_scrub_inflight += BP_GET_PSIZE(bp);
	mutex_exit(&spa->spa_scrub_lock);
	pio = scn->scn_zio_root;
	} else {
	kmutex_t *q_lock = &queue->q_vd->vdev_scan_io_queue_lock;

	ASSERT3U(queue->q_maxinflight_bytes, >, 0);
	mutex_enter(q_lock);
	while (queue->q_inflight_bytes >= queue->q_maxinflight_bytes)
	cv_wait(&queue->q_zio_cv, q_lock);
	queue->q_inflight_bytes += BP_GET_PSIZE(bp);
	pio = queue->q_zio;
	mutex_exit(q_lock);
	}

	ASSERT(pio != NULL);
	count_block_issued(spa, bp, queue == NULL);
	zio_nowait(zio_read(pio, spa, bp, data, size, dsl_scan_scrub_done,
	queue, ZIO_PRIORITY_SCRUB, zio_flags, zb));
	}

	/*
	* This is the primary extent sorting algorithm. We balance two parameters:
	* 1) how many bytes of I/O are in an extent
	* 2) how well the extent is filled with I/O (as a fraction of its total size)
	* Since we allow extents to have gaps between their constituent I/Os, it's
	* possible to have a fairly large extent that contains the same amount of
	* I/O bytes than a much smaller extent, which just packs the I/O more tightly.
	* The algorithm sorts based on a score calculated from the extent's size,
	* the relative fill volume (in %) and a "fill weight" parameter that controls
	* the split between whether we prefer larger extents or more well populated
	* extents:
	*
	* SCORE = FILL_IN_BYTES + (FILL_IN_PERCENT * FILL_IN_BYTES * FILL_WEIGHT)
	*
	* Example:
	* 1) assume extsz = 64 MiB
	* 2) assume fill = 32 MiB (extent is half full)
	* 3) assume fill_weight = 3
	* 4) SCORE = 32M + (((32M * 100) / 64M) * 3 * 32M) / 100
	* SCORE = 32M + (50 * 3 * 32M) / 100
	* SCORE = 32M + (4800M / 100)
	* SCORE = 32M + 48M
	* ^ ^
	* \| +--- final total relative fill-based score
	* +--------- final total fill-based score
	* SCORE = 80M
	*
	* As can be seen, at fill_ratio=3, the algorithm is slightly biased towards
	* extents that are more completely filled (in a 3:2 ratio) vs just larger.
	* Note that as an optimization, we replace multiplication and division by
	* 100 with bitshifting by 7 (which effectively multiplies and divides by 128).
	*
	* Since we do not care if one extent is only few percent better than another,
	* compress the score into 6 bits via binary logarithm AKA highbit64() and
	* put into otherwise unused due to ashift high bits of offset. This allows
	* to reduce q_exts_by_size B-tree elements to only 64 bits and compare them
	* with single operation. Plus it makes scrubs more sequential and reduces
	* chances that minor extent change move it within the B-tree.
	*/
	static int
	ext_size_compare(const void x, const void y)
	{
	const uint64_t a = x, b = y;

	return (TREE_CMP(a, b));
	}

	static void
	ext_size_create(range_tree_t rt, void arg)
	{
	(void) rt;
	zfs_btree_t *size_tree = arg;

	zfs_btree_create(size_tree, ext_size_compare, sizeof (uint64_t));
	}

	static void
	ext_size_destroy(range_tree_t rt, void arg)
	{
	(void) rt;
	zfs_btree_t *size_tree = arg;
	ASSERT0(zfs_btree_numnodes(size_tree));

	zfs_btree_destroy(size_tree);
	}

	static uint64_t
	ext_size_value(range_tree_t rt, range_seg_gap_t rsg)
	{
	(void) rt;
	uint64_t size = rsg->rs_end - rsg->rs_start;
	uint64_t score = rsg->rs_fill + ((((rsg->rs_fill << 7) / size) *
	fill_weight * rsg->rs_fill) >> 7);
	ASSERT3U(rt->rt_shift, >=, 8);
	return (((uint64_t)(64 - highbit64(score)) << 56) \| rsg->rs_start);
	}

	static void
	ext_size_add(range_tree_t rt, range_seg_t rs, void *arg)
	{
	zfs_btree_t *size_tree = arg;
	ASSERT3U(rt->rt_type, ==, RANGE_SEG_GAP);
	uint64_t v = ext_size_value(rt, (range_seg_gap_t *)rs);
	zfs_btree_add(size_tree, &v);
	}

	static void
	ext_size_remove(range_tree_t rt, range_seg_t rs, void *arg)
	{
	zfs_btree_t *size_tree = arg;
	ASSERT3U(rt->rt_type, ==, RANGE_SEG_GAP);
	uint64_t v = ext_size_value(rt, (range_seg_gap_t *)rs);
	zfs_btree_remove(size_tree, &v);
	}

	static void
	ext_size_vacate(range_tree_t rt, void arg)
	{
	zfs_btree_t *size_tree = arg;
	zfs_btree_clear(size_tree);
	zfs_btree_destroy(size_tree);

	ext_size_create(rt, arg);
	}

	static const range_tree_ops_t ext_size_ops = {
	.rtop_create = ext_size_create,
	.rtop_destroy = ext_size_destroy,
	.rtop_add = ext_size_add,
	.rtop_remove = ext_size_remove,
	.rtop_vacate = ext_size_vacate
	};

	/*
	* Comparator for the q_sios_by_addr tree. Sorting is simply performed
	* based on LBA-order (from lowest to highest).
	*/
	static int
	sio_addr_compare(const void x, const void y)
	{
	const scan_io_t a = x, b = y;

	return (TREE_CMP(SIO_GET_OFFSET(a), SIO_GET_OFFSET(b)));
	}

	/* IO queues are created on demand when they are needed. */
	static dsl_scan_io_queue_t *
	scan_io_queue_create(vdev_t *vd)
	{
	dsl_scan_t *scn = vd->vdev_spa->spa_dsl_pool->dp_scan;
	dsl_scan_io_queue_t q = kmem_zalloc(sizeof (q), KM_SLEEP);

	q->q_scn = scn;
	q->q_vd = vd;
	q->q_sio_memused = 0;
	q->q_last_ext_addr = -1;
	cv_init(&q->q_zio_cv, NULL, CV_DEFAULT, NULL);
	q->q_exts_by_addr = range_tree_create_gap(&ext_size_ops, RANGE_SEG_GAP,
	&q->q_exts_by_size, 0, vd->vdev_ashift, zfs_scan_max_ext_gap);
	avl_create(&q->q_sios_by_addr, sio_addr_compare,
	sizeof (scan_io_t), offsetof(scan_io_t, sio_nodes.sio_addr_node));

	return (q);
	}

	/*
	* Destroys a scan queue and all segments and scan_io_t's contained in it.
	* No further execution of I/O occurs, anything pending in the queue is
	* simply freed without being executed.
	*/
	void
	dsl_scan_io_queue_destroy(dsl_scan_io_queue_t *queue)
	{
	dsl_scan_t *scn = queue->q_scn;
	scan_io_t *sio;
	void *cookie = NULL;

	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));

	if (!avl_is_empty(&queue->q_sios_by_addr))
	atomic_add_64(&scn->scn_queues_pending, -1);
	while ((sio = avl_destroy_nodes(&queue->q_sios_by_addr, &cookie)) !=
	NULL) {
	ASSERT(range_tree_contains(queue->q_exts_by_addr,
	SIO_GET_OFFSET(sio), SIO_GET_ASIZE(sio)));
	queue->q_sio_memused -= SIO_GET_MUSED(sio);
	sio_free(sio);
	}

	ASSERT0(queue->q_sio_memused);
	range_tree_vacate(queue->q_exts_by_addr, NULL, queue);
	range_tree_destroy(queue->q_exts_by_addr);
	avl_destroy(&queue->q_sios_by_addr);
	cv_destroy(&queue->q_zio_cv);

	kmem_free(queue, sizeof (*queue));
	}

	/*
	* Properly transfers a dsl_scan_queue_t from `svd' to `tvd'. This is
	* called on behalf of vdev_top_transfer when creating or destroying
	* a mirror vdev due to zpool attach/detach.
	*/
	void
	dsl_scan_io_queue_vdev_xfer(vdev_t svd, vdev_t tvd)
	{
	mutex_enter(&svd->vdev_scan_io_queue_lock);
	mutex_enter(&tvd->vdev_scan_io_queue_lock);

	VERIFY3P(tvd->vdev_scan_io_queue, ==, NULL);
	tvd->vdev_scan_io_queue = svd->vdev_scan_io_queue;
	svd->vdev_scan_io_queue = NULL;
	if (tvd->vdev_scan_io_queue != NULL)
	tvd->vdev_scan_io_queue->q_vd = tvd;

	mutex_exit(&tvd->vdev_scan_io_queue_lock);
	mutex_exit(&svd->vdev_scan_io_queue_lock);
	}

	static void
	scan_io_queues_destroy(dsl_scan_t *scn)
	{
	vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;

	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *tvd = rvd->vdev_child[i];

	mutex_enter(&tvd->vdev_scan_io_queue_lock);
	if (tvd->vdev_scan_io_queue != NULL)
	dsl_scan_io_queue_destroy(tvd->vdev_scan_io_queue);
	tvd->vdev_scan_io_queue = NULL;
	mutex_exit(&tvd->vdev_scan_io_queue_lock);
	}
	}

	static void
	dsl_scan_freed_dva(spa_t spa, const blkptr_t bp, int dva_i)
	{
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dsl_scan_t *scn = dp->dp_scan;
	vdev_t *vdev;
	kmutex_t *q_lock;
	dsl_scan_io_queue_t *queue;
	scan_io_t srch_sio, sio;
	avl_index_t idx;
	uint64_t start, size;

	vdev = vdev_lookup_top(spa, DVA_GET_VDEV(&bp->blk_dva[dva_i]));
	ASSERT(vdev != NULL);
	q_lock = &vdev->vdev_scan_io_queue_lock;
	queue = vdev->vdev_scan_io_queue;

	mutex_enter(q_lock);
	if (queue == NULL) {
	mutex_exit(q_lock);
	return;
	}

	srch_sio = sio_alloc(BP_GET_NDVAS(bp));
	bp2sio(bp, srch_sio, dva_i);
	start = SIO_GET_OFFSET(srch_sio);
	size = SIO_GET_ASIZE(srch_sio);

	/*
	* We can find the zio in two states:
	* 1) Cold, just sitting in the queue of zio's to be issued at
	* some point in the future. In this case, all we do is
	* remove the zio from the q_sios_by_addr tree, decrement
	* its data volume from the containing range_seg_t and
	* resort the q_exts_by_size tree to reflect that the
	* range_seg_t has lost some of its 'fill'. We don't shorten
	* the range_seg_t - this is usually rare enough not to be
	* worth the extra hassle of trying keep track of precise
	* extent boundaries.
	* 2) Hot, where the zio is currently in-flight in
	* dsl_scan_issue_ios. In this case, we can't simply
	* reach in and stop the in-flight zio's, so we instead
	* block the caller. Eventually, dsl_scan_issue_ios will
	* be done with issuing the zio's it gathered and will
	* signal us.
	*/
	sio = avl_find(&queue->q_sios_by_addr, srch_sio, &idx);
	sio_free(srch_sio);

	if (sio != NULL) {
	blkptr_t tmpbp;

	/* Got it while it was cold in the queue */
	ASSERT3U(start, ==, SIO_GET_OFFSET(sio));
	ASSERT3U(size, ==, SIO_GET_ASIZE(sio));
	avl_remove(&queue->q_sios_by_addr, sio);
	if (avl_is_empty(&queue->q_sios_by_addr))
	atomic_add_64(&scn->scn_queues_pending, -1);
	queue->q_sio_memused -= SIO_GET_MUSED(sio);

	ASSERT(range_tree_contains(queue->q_exts_by_addr, start, size));
	range_tree_remove_fill(queue->q_exts_by_addr, start, size);

	/* count the block as though we issued it */
	sio2bp(sio, &tmpbp);
	count_block_issued(spa, &tmpbp, B_FALSE);

	sio_free(sio);
	}
	mutex_exit(q_lock);
	}

	/*
	* Callback invoked when a zio_free() zio is executing. This needs to be
	* intercepted to prevent the zio from deallocating a particular portion
	* of disk space and it then getting reallocated and written to, while we
	* still have it queued up for processing.
	*/
	void
	dsl_scan_freed(spa_t spa, const blkptr_t bp)
	{
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dsl_scan_t *scn = dp->dp_scan;

	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(scn != NULL);
	if (!dsl_scan_is_running(scn))
	return;

	for (int i = 0; i < BP_GET_NDVAS(bp); i++)
	dsl_scan_freed_dva(spa, bp, i);
	}

	/*
	* Check if a vdev needs resilvering (non-empty DTL), if so, and resilver has
	* not started, start it. Otherwise, only restart if max txg in DTL range is
	* greater than the max txg in the current scan. If the DTL max is less than
	* the scan max, then the vdev has not missed any new data since the resilver
	* started, so a restart is not needed.
	*/
	void
	dsl_scan_assess_vdev(dsl_pool_t dp, vdev_t vd)
	{
	uint64_t min, max;

	if (!vdev_resilver_needed(vd, &min, &max))
	return;

	if (!dsl_scan_resilvering(dp)) {
	spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER);
	return;
	}

	if (max <= dp->dp_scan->scn_phys.scn_max_txg)
	return;

	/* restart is needed, check if it can be deferred */
	if (spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER))
	vdev_defer_resilver(vd);
	else
	spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER);
	}

	ZFS_MODULE_PARAM(zfs, zfs_, scan_vdev_limit, ULONG, ZMOD_RW,
	"Max bytes in flight per leaf vdev for scrubs and resilvers");

	-ZFS_MODULE_PARAM(zfs, zfs_, scrub_min_time_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scrub_min_time_ms, UINT, ZMOD_RW,
	"Min millisecs to scrub per txg");

	-ZFS_MODULE_PARAM(zfs, zfs_, obsolete_min_time_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, obsolete_min_time_ms, UINT, ZMOD_RW,
	"Min millisecs to obsolete per txg");

	-ZFS_MODULE_PARAM(zfs, zfs_, free_min_time_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, free_min_time_ms, UINT, ZMOD_RW,
	"Min millisecs to free per txg");

	-ZFS_MODULE_PARAM(zfs, zfs_, resilver_min_time_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, resilver_min_time_ms, UINT, ZMOD_RW,
	"Min millisecs to resilver per txg");

	ZFS_MODULE_PARAM(zfs, zfs_, scan_suspend_progress, INT, ZMOD_RW,
	"Set to prevent scans from progressing");

	ZFS_MODULE_PARAM(zfs, zfs_, no_scrub_io, INT, ZMOD_RW,
	"Set to disable scrub I/O");

	ZFS_MODULE_PARAM(zfs, zfs_, no_scrub_prefetch, INT, ZMOD_RW,
	"Set to disable scrub prefetching");

	ZFS_MODULE_PARAM(zfs, zfs_, async_block_max_blocks, ULONG, ZMOD_RW,
	"Max number of blocks freed in one txg");

	ZFS_MODULE_PARAM(zfs, zfs_, max_async_dedup_frees, ULONG, ZMOD_RW,
	"Max number of dedup blocks freed in one txg");

	ZFS_MODULE_PARAM(zfs, zfs_, free_bpobj_enabled, INT, ZMOD_RW,
	"Enable processing of the free_bpobj");

	ZFS_MODULE_PARAM(zfs, zfs_, scan_blkstats, INT, ZMOD_RW,
	"Enable block statistics calculation during scrub");

	-ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_fact, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_fact, UINT, ZMOD_RW,
	"Fraction of RAM for scan hard limit");

	-ZFS_MODULE_PARAM(zfs, zfs_, scan_issue_strategy, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scan_issue_strategy, UINT, ZMOD_RW,
	"IO issuing strategy during scrubbing. 0 = default, 1 = LBA, 2 = size");

	ZFS_MODULE_PARAM(zfs, zfs_, scan_legacy, INT, ZMOD_RW,
	"Scrub using legacy non-sequential method");

	-ZFS_MODULE_PARAM(zfs, zfs_, scan_checkpoint_intval, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scan_checkpoint_intval, UINT, ZMOD_RW,
	"Scan progress on-disk checkpointing interval");

	ZFS_MODULE_PARAM(zfs, zfs_, scan_max_ext_gap, ULONG, ZMOD_RW,
	"Max gap in bytes between sequential scrub / resilver I/Os");

	-ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_soft_fact, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scan_mem_lim_soft_fact, UINT, ZMOD_RW,
	"Fraction of hard limit used as soft limit");

	ZFS_MODULE_PARAM(zfs, zfs_, scan_strict_mem_lim, INT, ZMOD_RW,
	"Tunable to attempt to reduce lock contention");

	-ZFS_MODULE_PARAM(zfs, zfs_, scan_fill_weight, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, scan_fill_weight, UINT, ZMOD_RW,
	"Tunable to adjust bias towards more filled segments during scans");

	ZFS_MODULE_PARAM(zfs, zfs_, resilver_disable_defer, INT, ZMOD_RW,
	"Process all resilvers immediately");
	diff --git a/sys/contrib/openzfs/module/zfs/fm.c b/sys/contrib/openzfs/module/zfs/fm.c
	index bc13b5517c4e..32b5cf8facd1 100644
	--- a/sys/contrib/openzfs/module/zfs/fm.c
	+++ b/sys/contrib/openzfs/module/zfs/fm.c
	@@ -1,1374 +1,1374 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
	*/

	/*
	* Fault Management Architecture (FMA) Resource and Protocol Support
	*
	* The routines contained herein provide services to support kernel subsystems
	* in publishing fault management telemetry (see PSARC 2002/412 and 2003/089).
	*
	* Name-Value Pair Lists
	*
	* The embodiment of an FMA protocol element (event, fmri or authority) is a
	* name-value pair list (nvlist_t). FMA-specific nvlist constructor and
	* destructor functions, fm_nvlist_create() and fm_nvlist_destroy(), are used
	* to create an nvpair list using custom allocators. Callers may choose to
	* allocate either from the kernel memory allocator, or from a preallocated
	* buffer, useful in constrained contexts like high-level interrupt routines.
	*
	* Protocol Event and FMRI Construction
	*
	* Convenience routines are provided to construct nvlist events according to
	* the FMA Event Protocol and Naming Schema specification for ereports and
	* FMRIs for the dev, cpu, hc, mem, legacy hc and de schemes.
	*
	* ENA Manipulation
	*
	* Routines to generate ENA formats 0, 1 and 2 are available as well as
	* routines to increment formats 1 and 2. Individual fields within the
	* ENA are extractable via fm_ena_time_get(), fm_ena_id_get(),
	* fm_ena_format_get() and fm_ena_gen_get().
	*/

	#include <sys/types.h>
	#include <sys/time.h>
	#include <sys/list.h>
	#include <sys/nvpair.h>
	#include <sys/cmn_err.h>
	#include <sys/sysmacros.h>
	#include <sys/sunddi.h>
	#include <sys/systeminfo.h>
	#include <sys/fm/util.h>
	#include <sys/fm/protocol.h>
	#include <sys/kstat.h>
	#include <sys/zfs_context.h>
	#ifdef _KERNEL
	#include <sys/atomic.h>
	#include <sys/condvar.h>
	#include <sys/zfs_ioctl.h>

	-static int zfs_zevent_len_max = 512;
	+static uint_t zfs_zevent_len_max = 512;

	-static int zevent_len_cur = 0;
	+static uint_t zevent_len_cur = 0;
	static int zevent_waiters = 0;
	static int zevent_flags = 0;

	/* Num events rate limited since the last time zfs_zevent_next() was called */
	static uint64_t ratelimit_dropped = 0;

	/*
	* The EID (Event IDentifier) is used to uniquely tag a zevent when it is
	* posted. The posted EIDs are monotonically increasing but not persistent.
	* They will be reset to the initial value (1) each time the kernel module is
	* loaded.
	*/
	static uint64_t zevent_eid = 0;

	static kmutex_t zevent_lock;
	static list_t zevent_list;
	static kcondvar_t zevent_cv;
	#endif /* _KERNEL */


	/*
	* Common fault management kstats to record event generation failures
	*/

	struct erpt_kstat {
	kstat_named_t erpt_dropped; /* num erpts dropped on post */
	kstat_named_t erpt_set_failed; /* num erpt set failures */
	kstat_named_t fmri_set_failed; /* num fmri set failures */
	kstat_named_t payload_set_failed; /* num payload set failures */
	kstat_named_t erpt_duplicates; /* num duplicate erpts */
	};

	static struct erpt_kstat erpt_kstat_data = {
	{ "erpt-dropped", KSTAT_DATA_UINT64 },
	{ "erpt-set-failed", KSTAT_DATA_UINT64 },
	{ "fmri-set-failed", KSTAT_DATA_UINT64 },
	{ "payload-set-failed", KSTAT_DATA_UINT64 },
	{ "erpt-duplicates", KSTAT_DATA_UINT64 }
	};

	kstat_t *fm_ksp;

	#ifdef _KERNEL

	static zevent_t *
	zfs_zevent_alloc(void)
	{
	zevent_t *ev;

	ev = kmem_zalloc(sizeof (zevent_t), KM_SLEEP);

	list_create(&ev->ev_ze_list, sizeof (zfs_zevent_t),
	offsetof(zfs_zevent_t, ze_node));
	list_link_init(&ev->ev_node);

	return (ev);
	}

	static void
	zfs_zevent_free(zevent_t *ev)
	{
	/* Run provided cleanup callback */
	ev->ev_cb(ev->ev_nvl, ev->ev_detector);

	list_destroy(&ev->ev_ze_list);
	kmem_free(ev, sizeof (zevent_t));
	}

	static void
	zfs_zevent_drain(zevent_t *ev)
	{
	zfs_zevent_t *ze;

	ASSERT(MUTEX_HELD(&zevent_lock));
	list_remove(&zevent_list, ev);

	/* Remove references to this event in all private file data */
	while ((ze = list_head(&ev->ev_ze_list)) != NULL) {
	list_remove(&ev->ev_ze_list, ze);
	ze->ze_zevent = NULL;
	ze->ze_dropped++;
	}

	zfs_zevent_free(ev);
	}

	void
	-zfs_zevent_drain_all(int *count)
	+zfs_zevent_drain_all(uint_t *count)
	{
	zevent_t *ev;

	mutex_enter(&zevent_lock);
	while ((ev = list_head(&zevent_list)) != NULL)
	zfs_zevent_drain(ev);

	*count = zevent_len_cur;
	zevent_len_cur = 0;
	mutex_exit(&zevent_lock);
	}

	/*
	* New zevents are inserted at the head. If the maximum queue
	* length is exceeded a zevent will be drained from the tail.
	* As part of this any user space processes which currently have
	* a reference to this zevent_t in their private data will have
	* this reference set to NULL.
	*/
	static void
	zfs_zevent_insert(zevent_t *ev)
	{
	ASSERT(MUTEX_HELD(&zevent_lock));
	list_insert_head(&zevent_list, ev);

	if (zevent_len_cur >= zfs_zevent_len_max)
	zfs_zevent_drain(list_tail(&zevent_list));
	else
	zevent_len_cur++;
	}

	/*
	* Post a zevent. The cb will be called when nvl and detector are no longer
	* needed, i.e.:
	* - An error happened and a zevent can't be posted. In this case, cb is called
	* before zfs_zevent_post() returns.
	* - The event is being drained and freed.
	*/
	int
	zfs_zevent_post(nvlist_t nvl, nvlist_t detector, zevent_cb_t *cb)
	{
	inode_timespec_t tv;
	int64_t tv_array[2];
	uint64_t eid;
	size_t nvl_size = 0;
	zevent_t *ev;
	int error;

	ASSERT(cb != NULL);

	gethrestime(&tv);
	tv_array[0] = tv.tv_sec;
	tv_array[1] = tv.tv_nsec;

	error = nvlist_add_int64_array(nvl, FM_EREPORT_TIME, tv_array, 2);
	if (error) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	goto out;
	}

	eid = atomic_inc_64_nv(&zevent_eid);
	error = nvlist_add_uint64(nvl, FM_EREPORT_EID, eid);
	if (error) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	goto out;
	}

	error = nvlist_size(nvl, &nvl_size, NV_ENCODE_NATIVE);
	if (error) {
	atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
	goto out;
	}

	if (nvl_size > ERPT_DATA_SZ \|\| nvl_size == 0) {
	atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
	error = EOVERFLOW;
	goto out;
	}

	ev = zfs_zevent_alloc();
	if (ev == NULL) {
	atomic_inc_64(&erpt_kstat_data.erpt_dropped.value.ui64);
	error = ENOMEM;
	goto out;
	}

	ev->ev_nvl = nvl;
	ev->ev_detector = detector;
	ev->ev_cb = cb;
	ev->ev_eid = eid;

	mutex_enter(&zevent_lock);
	zfs_zevent_insert(ev);
	cv_broadcast(&zevent_cv);
	mutex_exit(&zevent_lock);

	out:
	if (error)
	cb(nvl, detector);

	return (error);
	}

	void
	zfs_zevent_track_duplicate(void)
	{
	atomic_inc_64(&erpt_kstat_data.erpt_duplicates.value.ui64);
	}

	static int
	zfs_zevent_minor_to_state(minor_t minor, zfs_zevent_t **ze)
	{
	*ze = zfsdev_get_state(minor, ZST_ZEVENT);
	if (*ze == NULL)
	return (SET_ERROR(EBADF));

	return (0);
	}

	zfs_file_t *
	zfs_zevent_fd_hold(int fd, minor_t minorp, zfs_zevent_t *ze)
	{
	zfs_file_t *fp = zfs_file_get(fd);
	if (fp == NULL)
	return (NULL);

	int error = zfsdev_getminor(fp, minorp);
	if (error == 0)
	error = zfs_zevent_minor_to_state(*minorp, ze);

	if (error) {
	zfs_zevent_fd_rele(fp);
	fp = NULL;
	}

	return (fp);
	}

	void
	zfs_zevent_fd_rele(zfs_file_t *fp)
	{
	zfs_file_put(fp);
	}

	/*
	* Get the next zevent in the stream and place a copy in 'event'. This
	* may fail with ENOMEM if the encoded nvlist size exceeds the passed
	* 'event_size'. In this case the stream pointer is not advanced and
	* and 'event_size' is set to the minimum required buffer size.
	*/
	int
	zfs_zevent_next(zfs_zevent_t ze, nvlist_t event, uint64_t event_size,
	uint64_t *dropped)
	{
	zevent_t *ev;
	size_t size;
	int error = 0;

	mutex_enter(&zevent_lock);
	if (ze->ze_zevent == NULL) {
	/* New stream start at the beginning/tail */
	ev = list_tail(&zevent_list);
	if (ev == NULL) {
	error = ENOENT;
	goto out;
	}
	} else {
	/*
	* Existing stream continue with the next element and remove
	* ourselves from the wait queue for the previous element
	*/
	ev = list_prev(&zevent_list, ze->ze_zevent);
	if (ev == NULL) {
	error = ENOENT;
	goto out;
	}
	}

	VERIFY(nvlist_size(ev->ev_nvl, &size, NV_ENCODE_NATIVE) == 0);
	if (size > *event_size) {
	*event_size = size;
	error = ENOMEM;
	goto out;
	}

	if (ze->ze_zevent)
	list_remove(&ze->ze_zevent->ev_ze_list, ze);

	ze->ze_zevent = ev;
	list_insert_head(&ev->ev_ze_list, ze);
	(void) nvlist_dup(ev->ev_nvl, event, KM_SLEEP);
	*dropped = ze->ze_dropped;

	#ifdef _KERNEL
	/* Include events dropped due to rate limiting */
	*dropped += atomic_swap_64(&ratelimit_dropped, 0);
	#endif
	ze->ze_dropped = 0;
	out:
	mutex_exit(&zevent_lock);

	return (error);
	}

	/*
	* Wait in an interruptible state for any new events.
	*/
	int
	zfs_zevent_wait(zfs_zevent_t *ze)
	{
	int error = EAGAIN;

	mutex_enter(&zevent_lock);
	zevent_waiters++;

	while (error == EAGAIN) {
	if (zevent_flags & ZEVENT_SHUTDOWN) {
	error = SET_ERROR(ESHUTDOWN);
	break;
	}

	error = cv_wait_sig(&zevent_cv, &zevent_lock);
	if (signal_pending(current)) {
	error = SET_ERROR(EINTR);
	break;
	} else if (!list_is_empty(&zevent_list)) {
	error = 0;
	continue;
	} else {
	error = EAGAIN;
	}
	}

	zevent_waiters--;
	mutex_exit(&zevent_lock);

	return (error);
	}

	/*
	* The caller may seek to a specific EID by passing that EID. If the EID
	* is still available in the posted list of events the cursor is positioned
	* there. Otherwise ENOENT is returned and the cursor is not moved.
	*
	* There are two reserved EIDs which may be passed and will never fail.
	* ZEVENT_SEEK_START positions the cursor at the start of the list, and
	* ZEVENT_SEEK_END positions the cursor at the end of the list.
	*/
	int
	zfs_zevent_seek(zfs_zevent_t *ze, uint64_t eid)
	{
	zevent_t *ev;
	int error = 0;

	mutex_enter(&zevent_lock);

	if (eid == ZEVENT_SEEK_START) {
	if (ze->ze_zevent)
	list_remove(&ze->ze_zevent->ev_ze_list, ze);

	ze->ze_zevent = NULL;
	goto out;
	}

	if (eid == ZEVENT_SEEK_END) {
	if (ze->ze_zevent)
	list_remove(&ze->ze_zevent->ev_ze_list, ze);

	ev = list_head(&zevent_list);
	if (ev) {
	ze->ze_zevent = ev;
	list_insert_head(&ev->ev_ze_list, ze);
	} else {
	ze->ze_zevent = NULL;
	}

	goto out;
	}

	for (ev = list_tail(&zevent_list); ev != NULL;
	ev = list_prev(&zevent_list, ev)) {
	if (ev->ev_eid == eid) {
	if (ze->ze_zevent)
	list_remove(&ze->ze_zevent->ev_ze_list, ze);

	ze->ze_zevent = ev;
	list_insert_head(&ev->ev_ze_list, ze);
	break;
	}
	}

	if (ev == NULL)
	error = ENOENT;

	out:
	mutex_exit(&zevent_lock);

	return (error);
	}

	void
	zfs_zevent_init(zfs_zevent_t **zep)
	{
	zfs_zevent_t *ze;

	ze = *zep = kmem_zalloc(sizeof (zfs_zevent_t), KM_SLEEP);
	list_link_init(&ze->ze_node);
	}

	void
	zfs_zevent_destroy(zfs_zevent_t *ze)
	{
	mutex_enter(&zevent_lock);
	if (ze->ze_zevent)
	list_remove(&ze->ze_zevent->ev_ze_list, ze);
	mutex_exit(&zevent_lock);

	kmem_free(ze, sizeof (zfs_zevent_t));
	}
	#endif /* _KERNEL */

	/*
	* Wrappers for FM nvlist allocators
	*/
	static void *
	i_fm_alloc(nv_alloc_t *nva, size_t size)
	{
	(void) nva;
	return (kmem_alloc(size, KM_SLEEP));
	}

	static void
	i_fm_free(nv_alloc_t nva, void buf, size_t size)
	{
	(void) nva;
	kmem_free(buf, size);
	}

	static const nv_alloc_ops_t fm_mem_alloc_ops = {
	.nv_ao_init = NULL,
	.nv_ao_fini = NULL,
	.nv_ao_alloc = i_fm_alloc,
	.nv_ao_free = i_fm_free,
	.nv_ao_reset = NULL
	};

	/*
	* Create and initialize a new nv_alloc_t for a fixed buffer, buf. A pointer
	* to the newly allocated nv_alloc_t structure is returned upon success or NULL
	* is returned to indicate that the nv_alloc structure could not be created.
	*/
	nv_alloc_t *
	fm_nva_xcreate(char *buf, size_t bufsz)
	{
	nv_alloc_t *nvhdl = kmem_zalloc(sizeof (nv_alloc_t), KM_SLEEP);

	if (bufsz == 0 \|\| nv_alloc_init(nvhdl, nv_fixed_ops, buf, bufsz) != 0) {
	kmem_free(nvhdl, sizeof (nv_alloc_t));
	return (NULL);
	}

	return (nvhdl);
	}

	/*
	* Destroy a previously allocated nv_alloc structure. The fixed buffer
	* associated with nva must be freed by the caller.
	*/
	void
	fm_nva_xdestroy(nv_alloc_t *nva)
	{
	nv_alloc_fini(nva);
	kmem_free(nva, sizeof (nv_alloc_t));
	}

	/*
	* Create a new nv list. A pointer to a new nv list structure is returned
	* upon success or NULL is returned to indicate that the structure could
	* not be created. The newly created nv list is created and managed by the
	* operations installed in nva. If nva is NULL, the default FMA nva
	* operations are installed and used.
	*
	* When called from the kernel and nva == NULL, this function must be called
	* from passive kernel context with no locks held that can prevent a
	* sleeping memory allocation from occurring. Otherwise, this function may
	* be called from other kernel contexts as long a valid nva created via
	* fm_nva_create() is supplied.
	*/
	nvlist_t *
	fm_nvlist_create(nv_alloc_t *nva)
	{
	int hdl_alloced = 0;
	nvlist_t *nvl;
	nv_alloc_t *nvhdl;

	if (nva == NULL) {
	nvhdl = kmem_zalloc(sizeof (nv_alloc_t), KM_SLEEP);

	if (nv_alloc_init(nvhdl, &fm_mem_alloc_ops, NULL, 0) != 0) {
	kmem_free(nvhdl, sizeof (nv_alloc_t));
	return (NULL);
	}
	hdl_alloced = 1;
	} else {
	nvhdl = nva;
	}

	if (nvlist_xalloc(&nvl, NV_UNIQUE_NAME, nvhdl) != 0) {
	if (hdl_alloced) {
	nv_alloc_fini(nvhdl);
	kmem_free(nvhdl, sizeof (nv_alloc_t));
	}
	return (NULL);
	}

	return (nvl);
	}

	/*
	* Destroy a previously allocated nvlist structure. flag indicates whether
	* or not the associated nva structure should be freed (FM_NVA_FREE) or
	* retained (FM_NVA_RETAIN). Retaining the nv alloc structure allows
	* it to be re-used for future nvlist creation operations.
	*/
	void
	fm_nvlist_destroy(nvlist_t *nvl, int flag)
	{
	nv_alloc_t *nva = nvlist_lookup_nv_alloc(nvl);

	nvlist_free(nvl);

	if (nva != NULL) {
	if (flag == FM_NVA_FREE)
	fm_nva_xdestroy(nva);
	}
	}

	int
	i_fm_payload_set(nvlist_t payload, const char name, va_list ap)
	{
	int nelem, ret = 0;
	data_type_t type;

	while (ret == 0 && name != NULL) {
	type = va_arg(ap, data_type_t);
	switch (type) {
	case DATA_TYPE_BYTE:
	ret = nvlist_add_byte(payload, name,
	va_arg(ap, uint_t));
	break;
	case DATA_TYPE_BYTE_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_byte_array(payload, name,
	va_arg(ap, uchar_t *), nelem);
	break;
	case DATA_TYPE_BOOLEAN_VALUE:
	ret = nvlist_add_boolean_value(payload, name,
	va_arg(ap, boolean_t));
	break;
	case DATA_TYPE_BOOLEAN_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_boolean_array(payload, name,
	va_arg(ap, boolean_t *), nelem);
	break;
	case DATA_TYPE_INT8:
	ret = nvlist_add_int8(payload, name,
	va_arg(ap, int));
	break;
	case DATA_TYPE_INT8_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_int8_array(payload, name,
	va_arg(ap, int8_t *), nelem);
	break;
	case DATA_TYPE_UINT8:
	ret = nvlist_add_uint8(payload, name,
	va_arg(ap, uint_t));
	break;
	case DATA_TYPE_UINT8_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_uint8_array(payload, name,
	va_arg(ap, uint8_t *), nelem);
	break;
	case DATA_TYPE_INT16:
	ret = nvlist_add_int16(payload, name,
	va_arg(ap, int));
	break;
	case DATA_TYPE_INT16_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_int16_array(payload, name,
	va_arg(ap, int16_t *), nelem);
	break;
	case DATA_TYPE_UINT16:
	ret = nvlist_add_uint16(payload, name,
	va_arg(ap, uint_t));
	break;
	case DATA_TYPE_UINT16_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_uint16_array(payload, name,
	va_arg(ap, uint16_t *), nelem);
	break;
	case DATA_TYPE_INT32:
	ret = nvlist_add_int32(payload, name,
	va_arg(ap, int32_t));
	break;
	case DATA_TYPE_INT32_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_int32_array(payload, name,
	va_arg(ap, int32_t *), nelem);
	break;
	case DATA_TYPE_UINT32:
	ret = nvlist_add_uint32(payload, name,
	va_arg(ap, uint32_t));
	break;
	case DATA_TYPE_UINT32_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_uint32_array(payload, name,
	va_arg(ap, uint32_t *), nelem);
	break;
	case DATA_TYPE_INT64:
	ret = nvlist_add_int64(payload, name,
	va_arg(ap, int64_t));
	break;
	case DATA_TYPE_INT64_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_int64_array(payload, name,
	va_arg(ap, int64_t *), nelem);
	break;
	case DATA_TYPE_UINT64:
	ret = nvlist_add_uint64(payload, name,
	va_arg(ap, uint64_t));
	break;
	case DATA_TYPE_UINT64_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_uint64_array(payload, name,
	va_arg(ap, uint64_t *), nelem);
	break;
	case DATA_TYPE_STRING:
	ret = nvlist_add_string(payload, name,
	va_arg(ap, char *));
	break;
	case DATA_TYPE_STRING_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_string_array(payload, name,
	va_arg(ap, const char **), nelem);
	break;
	case DATA_TYPE_NVLIST:
	ret = nvlist_add_nvlist(payload, name,
	va_arg(ap, nvlist_t *));
	break;
	case DATA_TYPE_NVLIST_ARRAY:
	nelem = va_arg(ap, int);
	ret = nvlist_add_nvlist_array(payload, name,
	va_arg(ap, const nvlist_t **), nelem);
	break;
	default:
	ret = EINVAL;
	}

	name = va_arg(ap, char *);
	}
	return (ret);
	}

	void
	fm_payload_set(nvlist_t *payload, ...)
	{
	int ret;
	const char *name;
	va_list ap;

	va_start(ap, payload);
	name = va_arg(ap, char *);
	ret = i_fm_payload_set(payload, name, ap);
	va_end(ap);

	if (ret)
	atomic_inc_64(&erpt_kstat_data.payload_set_failed.value.ui64);
	}

	/*
	* Set-up and validate the members of an ereport event according to:
	*
	* Member name Type Value
	* ====================================================
	* class string ereport
	* version uint8_t 0
	* ena uint64_t <ena>
	* detector nvlist_t <detector>
	* ereport-payload nvlist_t <var args>
	*
	* We don't actually add a 'version' member to the payload. Really,
	* the version quoted to us by our caller is that of the category 1
	* "ereport" event class (and we require FM_EREPORT_VERS0) but
	* the payload version of the actual leaf class event under construction
	* may be something else. Callers should supply a version in the varargs,
	* or (better) we could take two version arguments - one for the
	* ereport category 1 classification (expect FM_EREPORT_VERS0) and one
	* for the leaf class.
	*/
	void
	fm_ereport_set(nvlist_t ereport, int version, const char erpt_class,
	uint64_t ena, const nvlist_t *detector, ...)
	{
	char ereport_class[FM_MAX_CLASS];
	const char *name;
	va_list ap;
	int ret;

	if (version != FM_EREPORT_VERS0) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	return;
	}

	(void) snprintf(ereport_class, FM_MAX_CLASS, "%s.%s",
	FM_EREPORT_CLASS, erpt_class);
	if (nvlist_add_string(ereport, FM_CLASS, ereport_class) != 0) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_uint64(ereport, FM_EREPORT_ENA, ena)) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	}

	if (nvlist_add_nvlist(ereport, FM_EREPORT_DETECTOR,
	(nvlist_t *)detector) != 0) {
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	}

	va_start(ap, detector);
	name = va_arg(ap, const char *);
	ret = i_fm_payload_set(ereport, name, ap);
	va_end(ap);

	if (ret)
	atomic_inc_64(&erpt_kstat_data.erpt_set_failed.value.ui64);
	}

	/*
	* Set-up and validate the members of an hc fmri according to;
	*
	* Member name Type Value
	* ===================================================
	* version uint8_t 0
	* auth nvlist_t <auth>
	* hc-name string <name>
	* hc-id string <id>
	*
	* Note that auth and hc-id are optional members.
	*/

	#define HC_MAXPAIRS 20
	#define HC_MAXNAMELEN 50

	static int
	fm_fmri_hc_set_common(nvlist_t fmri, int version, const nvlist_t auth)
	{
	if (version != FM_HC_SCHEME_VERSION) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return (0);
	}

	if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0 \|\|
	nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_HC) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return (0);
	}

	if (auth != NULL && nvlist_add_nvlist(fmri, FM_FMRI_AUTHORITY,
	(nvlist_t *)auth) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return (0);
	}

	return (1);
	}

	void
	fm_fmri_hc_set(nvlist_t fmri, int version, const nvlist_t auth,
	nvlist_t *snvl, int npairs, ...)
	{
	nv_alloc_t *nva = nvlist_lookup_nv_alloc(fmri);
	nvlist_t *pairs[HC_MAXPAIRS];
	va_list ap;
	int i;

	if (!fm_fmri_hc_set_common(fmri, version, auth))
	return;

	npairs = MIN(npairs, HC_MAXPAIRS);

	va_start(ap, npairs);
	for (i = 0; i < npairs; i++) {
	const char name = va_arg(ap, const char );
	uint32_t id = va_arg(ap, uint32_t);
	char idstr[11];

	(void) snprintf(idstr, sizeof (idstr), "%u", id);

	pairs[i] = fm_nvlist_create(nva);
	if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, name) != 0 \|\|
	nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	}
	va_end(ap);

	if (nvlist_add_nvlist_array(fmri, FM_FMRI_HC_LIST,
	(const nvlist_t **)pairs, npairs) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	}

	for (i = 0; i < npairs; i++)
	fm_nvlist_destroy(pairs[i], FM_NVA_RETAIN);

	if (snvl != NULL) {
	if (nvlist_add_nvlist(fmri, FM_FMRI_HC_SPECIFIC, snvl) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	}
	}

	void
	fm_fmri_hc_create(nvlist_t fmri, int version, const nvlist_t auth,
	nvlist_t snvl, nvlist_t bboard, int npairs, ...)
	{
	nv_alloc_t *nva = nvlist_lookup_nv_alloc(fmri);
	nvlist_t *pairs[HC_MAXPAIRS];
	nvlist_t **hcl;
	uint_t n;
	int i, j;
	va_list ap;
	char hcname, hcid;

	if (!fm_fmri_hc_set_common(fmri, version, auth))
	return;

	/*
	* copy the bboard nvpairs to the pairs array
	*/
	if (nvlist_lookup_nvlist_array(bboard, FM_FMRI_HC_LIST, &hcl, &n)
	!= 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	for (i = 0; i < n; i++) {
	if (nvlist_lookup_string(hcl[i], FM_FMRI_HC_NAME,
	&hcname) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}
	if (nvlist_lookup_string(hcl[i], FM_FMRI_HC_ID, &hcid) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	pairs[i] = fm_nvlist_create(nva);
	if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, hcname) != 0 \|\|
	nvlist_add_string(pairs[i], FM_FMRI_HC_ID, hcid) != 0) {
	for (j = 0; j <= i; j++) {
	if (pairs[j] != NULL)
	fm_nvlist_destroy(pairs[j],
	FM_NVA_RETAIN);
	}
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}
	}

	/*
	* create the pairs from passed in pairs
	*/
	npairs = MIN(npairs, HC_MAXPAIRS);

	va_start(ap, npairs);
	for (i = n; i < npairs + n; i++) {
	const char name = va_arg(ap, const char );
	uint32_t id = va_arg(ap, uint32_t);
	char idstr[11];
	(void) snprintf(idstr, sizeof (idstr), "%u", id);
	pairs[i] = fm_nvlist_create(nva);
	if (nvlist_add_string(pairs[i], FM_FMRI_HC_NAME, name) != 0 \|\|
	nvlist_add_string(pairs[i], FM_FMRI_HC_ID, idstr) != 0) {
	for (j = 0; j <= i; j++) {
	if (pairs[j] != NULL)
	fm_nvlist_destroy(pairs[j],
	FM_NVA_RETAIN);
	}
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}
	}
	va_end(ap);

	/*
	* Create the fmri hc list
	*/
	if (nvlist_add_nvlist_array(fmri, FM_FMRI_HC_LIST,
	(const nvlist_t **)pairs, npairs + n) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	for (i = 0; i < npairs + n; i++) {
	fm_nvlist_destroy(pairs[i], FM_NVA_RETAIN);
	}

	if (snvl != NULL) {
	if (nvlist_add_nvlist(fmri, FM_FMRI_HC_SPECIFIC, snvl) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}
	}
	}

	/*
	* Set-up and validate the members of an dev fmri according to:
	*
	* Member name Type Value
	* ====================================================
	* version uint8_t 0
	* auth nvlist_t <auth>
	* devpath string <devpath>
	* [devid] string <devid>
	* [target-port-l0id] string <target-port-lun0-id>
	*
	* Note that auth and devid are optional members.
	*/
	void
	fm_fmri_dev_set(nvlist_t fmri_dev, int version, const nvlist_t auth,
	const char devpath, const char devid, const char *tpl0)
	{
	int err = 0;

	if (version != DEV_SCHEME_VERSION0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	err \|= nvlist_add_uint8(fmri_dev, FM_VERSION, version);
	err \|= nvlist_add_string(fmri_dev, FM_FMRI_SCHEME, FM_FMRI_SCHEME_DEV);

	if (auth != NULL) {
	err \|= nvlist_add_nvlist(fmri_dev, FM_FMRI_AUTHORITY,
	(nvlist_t *)auth);
	}

	err \|= nvlist_add_string(fmri_dev, FM_FMRI_DEV_PATH, devpath);

	if (devid != NULL)
	err \|= nvlist_add_string(fmri_dev, FM_FMRI_DEV_ID, devid);

	if (tpl0 != NULL)
	err \|= nvlist_add_string(fmri_dev, FM_FMRI_DEV_TGTPTLUN0, tpl0);

	if (err)
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);

	}

	/*
	* Set-up and validate the members of an cpu fmri according to:
	*
	* Member name Type Value
	* ====================================================
	* version uint8_t 0
	* auth nvlist_t <auth>
	* cpuid uint32_t <cpu_id>
	* cpumask uint8_t <cpu_mask>
	* serial uint64_t <serial_id>
	*
	* Note that auth, cpumask, serial are optional members.
	*
	*/
	void
	fm_fmri_cpu_set(nvlist_t fmri_cpu, int version, const nvlist_t auth,
	uint32_t cpu_id, uint8_t cpu_maskp, const char serial_idp)
	{
	uint64_t *failedp = &erpt_kstat_data.fmri_set_failed.value.ui64;

	if (version < CPU_SCHEME_VERSION1) {
	atomic_inc_64(failedp);
	return;
	}

	if (nvlist_add_uint8(fmri_cpu, FM_VERSION, version) != 0) {
	atomic_inc_64(failedp);
	return;
	}

	if (nvlist_add_string(fmri_cpu, FM_FMRI_SCHEME,
	FM_FMRI_SCHEME_CPU) != 0) {
	atomic_inc_64(failedp);
	return;
	}

	if (auth != NULL && nvlist_add_nvlist(fmri_cpu, FM_FMRI_AUTHORITY,
	(nvlist_t *)auth) != 0)
	atomic_inc_64(failedp);

	if (nvlist_add_uint32(fmri_cpu, FM_FMRI_CPU_ID, cpu_id) != 0)
	atomic_inc_64(failedp);

	if (cpu_maskp != NULL && nvlist_add_uint8(fmri_cpu, FM_FMRI_CPU_MASK,
	*cpu_maskp) != 0)
	atomic_inc_64(failedp);

	if (serial_idp == NULL \|\| nvlist_add_string(fmri_cpu,
	FM_FMRI_CPU_SERIAL_ID, (char *)serial_idp) != 0)
	atomic_inc_64(failedp);
	}

	/*
	* Set-up and validate the members of a mem according to:
	*
	* Member name Type Value
	* ====================================================
	* version uint8_t 0
	* auth nvlist_t <auth> [optional]
	* unum string <unum>
	* serial string <serial> [optional*]
	* offset uint64_t <offset> [optional]
	*
	* * serial is required if offset is present
	*/
	void
	fm_fmri_mem_set(nvlist_t fmri, int version, const nvlist_t auth,
	const char unum, const char serial, uint64_t offset)
	{
	if (version != MEM_SCHEME_VERSION0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (!serial && (offset != (uint64_t)-1)) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_MEM) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (auth != NULL) {
	if (nvlist_add_nvlist(fmri, FM_FMRI_AUTHORITY,
	(nvlist_t *)auth) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	}

	if (nvlist_add_string(fmri, FM_FMRI_MEM_UNUM, unum) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	}

	if (serial != NULL) {
	if (nvlist_add_string_array(fmri, FM_FMRI_MEM_SERIAL_ID,
	(const char **)&serial, 1) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	if (offset != (uint64_t)-1 && nvlist_add_uint64(fmri,
	FM_FMRI_MEM_OFFSET, offset) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	}
	}

	void
	fm_fmri_zfs_set(nvlist_t *fmri, int version, uint64_t pool_guid,
	uint64_t vdev_guid)
	{
	if (version != ZFS_SCHEME_VERSION0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_uint8(fmri, FM_VERSION, version) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_string(fmri, FM_FMRI_SCHEME, FM_FMRI_SCHEME_ZFS) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	return;
	}

	if (nvlist_add_uint64(fmri, FM_FMRI_ZFS_POOL, pool_guid) != 0) {
	atomic_inc_64(&erpt_kstat_data.fmri_set_failed.value.ui64);
	}

	if (vdev_guid != 0) {
	if (nvlist_add_uint64(fmri, FM_FMRI_ZFS_VDEV, vdev_guid) != 0) {
	atomic_inc_64(
	&erpt_kstat_data.fmri_set_failed.value.ui64);
	}
	}
	}

	uint64_t
	fm_ena_increment(uint64_t ena)
	{
	uint64_t new_ena;

	switch (ENA_FORMAT(ena)) {
	case FM_ENA_FMT1:
	new_ena = ena + (1 << ENA_FMT1_GEN_SHFT);
	break;
	case FM_ENA_FMT2:
	new_ena = ena + (1 << ENA_FMT2_GEN_SHFT);
	break;
	default:
	new_ena = 0;
	}

	return (new_ena);
	}

	uint64_t
	fm_ena_generate_cpu(uint64_t timestamp, processorid_t cpuid, uchar_t format)
	{
	uint64_t ena = 0;

	switch (format) {
	case FM_ENA_FMT1:
	if (timestamp) {
	ena = (uint64_t)((format & ENA_FORMAT_MASK) \|
	((cpuid << ENA_FMT1_CPUID_SHFT) &
	ENA_FMT1_CPUID_MASK) \|
	((timestamp << ENA_FMT1_TIME_SHFT) &
	ENA_FMT1_TIME_MASK));
	} else {
	ena = (uint64_t)((format & ENA_FORMAT_MASK) \|
	((cpuid << ENA_FMT1_CPUID_SHFT) &
	ENA_FMT1_CPUID_MASK) \|
	((gethrtime() << ENA_FMT1_TIME_SHFT) &
	ENA_FMT1_TIME_MASK));
	}
	break;
	case FM_ENA_FMT2:
	ena = (uint64_t)((format & ENA_FORMAT_MASK) \|
	((timestamp << ENA_FMT2_TIME_SHFT) & ENA_FMT2_TIME_MASK));
	break;
	default:
	break;
	}

	return (ena);
	}

	uint64_t
	fm_ena_generate(uint64_t timestamp, uchar_t format)
	{
	uint64_t ena;

	kpreempt_disable();
	ena = fm_ena_generate_cpu(timestamp, getcpuid(), format);
	kpreempt_enable();

	return (ena);
	}

	uint64_t
	fm_ena_generation_get(uint64_t ena)
	{
	uint64_t gen;

	switch (ENA_FORMAT(ena)) {
	case FM_ENA_FMT1:
	gen = (ena & ENA_FMT1_GEN_MASK) >> ENA_FMT1_GEN_SHFT;
	break;
	case FM_ENA_FMT2:
	gen = (ena & ENA_FMT2_GEN_MASK) >> ENA_FMT2_GEN_SHFT;
	break;
	default:
	gen = 0;
	break;
	}

	return (gen);
	}

	uchar_t
	fm_ena_format_get(uint64_t ena)
	{

	return (ENA_FORMAT(ena));
	}

	uint64_t
	fm_ena_id_get(uint64_t ena)
	{
	uint64_t id;

	switch (ENA_FORMAT(ena)) {
	case FM_ENA_FMT1:
	id = (ena & ENA_FMT1_ID_MASK) >> ENA_FMT1_ID_SHFT;
	break;
	case FM_ENA_FMT2:
	id = (ena & ENA_FMT2_ID_MASK) >> ENA_FMT2_ID_SHFT;
	break;
	default:
	id = 0;
	}

	return (id);
	}

	uint64_t
	fm_ena_time_get(uint64_t ena)
	{
	uint64_t time;

	switch (ENA_FORMAT(ena)) {
	case FM_ENA_FMT1:
	time = (ena & ENA_FMT1_TIME_MASK) >> ENA_FMT1_TIME_SHFT;
	break;
	case FM_ENA_FMT2:
	time = (ena & ENA_FMT2_TIME_MASK) >> ENA_FMT2_TIME_SHFT;
	break;
	default:
	time = 0;
	}

	return (time);
	}

	#ifdef _KERNEL
	/*
	* Helper function to increment ereport dropped count. Used by the event
	* rate limiting code to give feedback to the user about how many events were
	* rate limited by including them in the 'dropped' count.
	*/
	void
	fm_erpt_dropped_increment(void)
	{
	atomic_inc_64(&ratelimit_dropped);
	}

	void
	fm_init(void)
	{
	zevent_len_cur = 0;
	zevent_flags = 0;

	/* Initialize zevent allocation and generation kstats */
	fm_ksp = kstat_create("zfs", 0, "fm", "misc", KSTAT_TYPE_NAMED,
	sizeof (struct erpt_kstat) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);

	if (fm_ksp != NULL) {
	fm_ksp->ks_data = &erpt_kstat_data;
	kstat_install(fm_ksp);
	} else {
	cmn_err(CE_NOTE, "failed to create fm/misc kstat\n");
	}

	mutex_init(&zevent_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&zevent_list, sizeof (zevent_t),
	offsetof(zevent_t, ev_node));
	cv_init(&zevent_cv, NULL, CV_DEFAULT, NULL);

	zfs_ereport_init();
	}

	void
	fm_fini(void)
	{
	- int count;
	+ uint_t count;

	zfs_ereport_fini();

	zfs_zevent_drain_all(&count);

	mutex_enter(&zevent_lock);
	cv_broadcast(&zevent_cv);

	zevent_flags \|= ZEVENT_SHUTDOWN;
	while (zevent_waiters > 0) {
	mutex_exit(&zevent_lock);
	kpreempt(KPREEMPT_SYNC);
	mutex_enter(&zevent_lock);
	}
	mutex_exit(&zevent_lock);

	cv_destroy(&zevent_cv);
	list_destroy(&zevent_list);
	mutex_destroy(&zevent_lock);

	if (fm_ksp != NULL) {
	kstat_delete(fm_ksp);
	fm_ksp = NULL;
	}
	}
	#endif /* _KERNEL */

	-ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, len_max, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, len_max, UINT, ZMOD_RW,
	"Max event queue length");
	diff --git a/sys/contrib/openzfs/module/zfs/metaslab.c b/sys/contrib/openzfs/module/zfs/metaslab.c
	index 02cf121d83d7..efcfeecd778e 100644
	--- a/sys/contrib/openzfs/module/zfs/metaslab.c
	+++ b/sys/contrib/openzfs/module/zfs/metaslab.c
	@@ -1,6273 +1,6273 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	*/

	#include <sys/zfs_context.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/space_map.h>
	#include <sys/metaslab_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/zfeature.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/zap.h>
	#include <sys/btree.h>

	#define WITH_DF_BLOCK_ALLOCATOR

	#define GANG_ALLOCATION(flags) \
	((flags) & (METASLAB_GANG_CHILD \| METASLAB_GANG_HEADER))

	/*
	* Metaslab granularity, in bytes. This is roughly similar to what would be
	* referred to as the "stripe size" in traditional RAID arrays. In normal
	* operation, we will try to write this amount of data to each disk before
	* moving on to the next top-level vdev.
	*/
	static unsigned long metaslab_aliquot = 1024 * 1024;

	/*
	* For testing, make some blocks above a certain size be gang blocks.
	*/
	unsigned long metaslab_force_ganging = SPA_MAXBLOCKSIZE + 1;

	/*
	* In pools where the log space map feature is not enabled we touch
	* multiple metaslabs (and their respective space maps) with each
	* transaction group. Thus, we benefit from having a small space map
	* block size since it allows us to issue more I/O operations scattered
	* around the disk. So a sane default for the space map block size
	* is 8~16K.
	*/
	int zfs_metaslab_sm_blksz_no_log = (1 << 14);

	/*
	* When the log space map feature is enabled, we accumulate a lot of
	* changes per metaslab that are flushed once in a while so we benefit
	* from a bigger block size like 128K for the metaslab space maps.
	*/
	int zfs_metaslab_sm_blksz_with_log = (1 << 17);

	/*
	* The in-core space map representation is more compact than its on-disk form.
	* The zfs_condense_pct determines how much more compact the in-core
	* space map representation must be before we compact it on-disk.
	* Values should be greater than or equal to 100.
	*/
	-int zfs_condense_pct = 200;
	+uint_t zfs_condense_pct = 200;

	/*
	* Condensing a metaslab is not guaranteed to actually reduce the amount of
	* space used on disk. In particular, a space map uses data in increments of
	* MAX(1 << ashift, space_map_blksz), so a metaslab might use the
	* same number of blocks after condensing. Since the goal of condensing is to
	* reduce the number of IOPs required to read the space map, we only want to
	* condense when we can be sure we will reduce the number of blocks used by the
	* space map. Unfortunately, we cannot precisely compute whether or not this is
	* the case in metaslab_should_condense since we are holding ms_lock. Instead,
	* we apply the following heuristic: do not condense a spacemap unless the
	* uncondensed size consumes greater than zfs_metaslab_condense_block_threshold
	* blocks.
	*/
	static const int zfs_metaslab_condense_block_threshold = 4;

	/*
	* The zfs_mg_noalloc_threshold defines which metaslab groups should
	* be eligible for allocation. The value is defined as a percentage of
	* free space. Metaslab groups that have more free space than
	* zfs_mg_noalloc_threshold are always eligible for allocations. Once
	* a metaslab group's free space is less than or equal to the
	* zfs_mg_noalloc_threshold the allocator will avoid allocating to that
	* group unless all groups in the pool have reached zfs_mg_noalloc_threshold.
	* Once all groups in the pool reach zfs_mg_noalloc_threshold then all
	* groups are allowed to accept allocations. Gang blocks are always
	* eligible to allocate on any metaslab group. The default value of 0 means
	* no metaslab group will be excluded based on this criterion.
	*/
	-static int zfs_mg_noalloc_threshold = 0;
	+static uint_t zfs_mg_noalloc_threshold = 0;

	/*
	* Metaslab groups are considered eligible for allocations if their
	* fragmentation metric (measured as a percentage) is less than or
	* equal to zfs_mg_fragmentation_threshold. If a metaslab group
	* exceeds this threshold then it will be skipped unless all metaslab
	* groups within the metaslab class have also crossed this threshold.
	*
	* This tunable was introduced to avoid edge cases where we continue
	* allocating from very fragmented disks in our pool while other, less
	* fragmented disks, exists. On the other hand, if all disks in the
	* pool are uniformly approaching the threshold, the threshold can
	* be a speed bump in performance, where we keep switching the disks
	* that we allocate from (e.g. we allocate some segments from disk A
	* making it bypassing the threshold while freeing segments from disk
	* B getting its fragmentation below the threshold).
	*
	* Empirically, we've seen that our vdev selection for allocations is
	* good enough that fragmentation increases uniformly across all vdevs
	* the majority of the time. Thus we set the threshold percentage high
	* enough to avoid hitting the speed bump on pools that are being pushed
	* to the edge.
	*/
	-static int zfs_mg_fragmentation_threshold = 95;
	+static uint_t zfs_mg_fragmentation_threshold = 95;

	/*
	* Allow metaslabs to keep their active state as long as their fragmentation
	* percentage is less than or equal to zfs_metaslab_fragmentation_threshold. An
	* active metaslab that exceeds this threshold will no longer keep its active
	* status allowing better metaslabs to be selected.
	*/
	-static int zfs_metaslab_fragmentation_threshold = 70;
	+static uint_t zfs_metaslab_fragmentation_threshold = 70;

	/*
	* When set will load all metaslabs when pool is first opened.
	*/
	int metaslab_debug_load = B_FALSE;

	/*
	* When set will prevent metaslabs from being unloaded.
	*/
	static int metaslab_debug_unload = B_FALSE;

	/*
	* Minimum size which forces the dynamic allocator to change
	* it's allocation strategy. Once the space map cannot satisfy
	* an allocation of this size then it switches to using more
	* aggressive strategy (i.e search by size rather than offset).
	*/
	uint64_t metaslab_df_alloc_threshold = SPA_OLD_MAXBLOCKSIZE;

	/*
	* The minimum free space, in percent, which must be available
	* in a space map to continue allocations in a first-fit fashion.
	* Once the space map's free space drops below this level we dynamically
	* switch to using best-fit allocations.
	*/
	-int metaslab_df_free_pct = 4;
	+uint_t metaslab_df_free_pct = 4;

	/*
	* Maximum distance to search forward from the last offset. Without this
	* limit, fragmented pools can see >100,000 iterations and
	* metaslab_block_picker() becomes the performance limiting factor on
	* high-performance storage.
	*
	* With the default setting of 16MB, we typically see less than 500
	* iterations, even with very fragmented, ashift=9 pools. The maximum number
	* of iterations possible is:
	* metaslab_df_max_search / (2 * (1<<ashift))
	* With the default setting of 16MB this is 16*1024 (with ashift=9) or
	* 2048 (with ashift=12).
	*/
	-static int metaslab_df_max_search = 16 * 1024 * 1024;
	+static uint_t metaslab_df_max_search = 16 * 1024 * 1024;

	/*
	* Forces the metaslab_block_picker function to search for at least this many
	* segments forwards until giving up on finding a segment that the allocation
	* will fit into.
	*/
	static const uint32_t metaslab_min_search_count = 100;

	/*
	* If we are not searching forward (due to metaslab_df_max_search,
	* metaslab_df_free_pct, or metaslab_df_alloc_threshold), this tunable
	* controls what segment is used. If it is set, we will use the largest free
	* segment. If it is not set, we will use a segment of exactly the requested
	* size (or larger).
	*/
	static int metaslab_df_use_largest_segment = B_FALSE;

	/*
	* Percentage of all cpus that can be used by the metaslab taskq.
	*/
	int metaslab_load_pct = 50;

	/*
	* These tunables control how long a metaslab will remain loaded after the
	* last allocation from it. A metaslab can't be unloaded until at least
	* metaslab_unload_delay TXG's and metaslab_unload_delay_ms milliseconds
	* have elapsed. However, zfs_metaslab_mem_limit may cause it to be
	* unloaded sooner. These settings are intended to be generous -- to keep
	* metaslabs loaded for a long time, reducing the rate of metaslab loading.
	*/
	-static int metaslab_unload_delay = 32;
	-static int metaslab_unload_delay_ms = 10 * 60 * 1000; /* ten minutes */
	+static uint_t metaslab_unload_delay = 32;
	+static uint_t metaslab_unload_delay_ms = 10 * 60 * 1000; /* ten minutes */

	/*
	* Max number of metaslabs per group to preload.
	*/
	-int metaslab_preload_limit = 10;
	+uint_t metaslab_preload_limit = 10;

	/*
	* Enable/disable preloading of metaslab.
	*/
	static int metaslab_preload_enabled = B_TRUE;

	/*
	* Enable/disable fragmentation weighting on metaslabs.
	*/
	static int metaslab_fragmentation_factor_enabled = B_TRUE;

	/*
	* Enable/disable lba weighting (i.e. outer tracks are given preference).
	*/
	static int metaslab_lba_weighting_enabled = B_TRUE;

	/*
	* Enable/disable metaslab group biasing.
	*/
	static int metaslab_bias_enabled = B_TRUE;

	/*
	* Enable/disable remapping of indirect DVAs to their concrete vdevs.
	*/
	static const boolean_t zfs_remap_blkptr_enable = B_TRUE;

	/*
	* Enable/disable segment-based metaslab selection.
	*/
	static int zfs_metaslab_segment_weight_enabled = B_TRUE;

	/*
	* When using segment-based metaslab selection, we will continue
	* allocating from the active metaslab until we have exhausted
	* zfs_metaslab_switch_threshold of its buckets.
	*/
	static int zfs_metaslab_switch_threshold = 2;

	/*
	* Internal switch to enable/disable the metaslab allocation tracing
	* facility.
	*/
	static const boolean_t metaslab_trace_enabled = B_FALSE;

	/*
	* Maximum entries that the metaslab allocation tracing facility will keep
	* in a given list when running in non-debug mode. We limit the number
	* of entries in non-debug mode to prevent us from using up too much memory.
	* The limit should be sufficiently large that we don't expect any allocation
	* to every exceed this value. In debug mode, the system will panic if this
	* limit is ever reached allowing for further investigation.
	*/
	static const uint64_t metaslab_trace_max_entries = 5000;

	/*
	* Maximum number of metaslabs per group that can be disabled
	* simultaneously.
	*/
	static const int max_disabled_ms = 3;

	/*
	* Time (in seconds) to respect ms_max_size when the metaslab is not loaded.
	* To avoid 64-bit overflow, don't set above UINT32_MAX.
	*/
	static unsigned long zfs_metaslab_max_size_cache_sec = 1 * 60 * 60; /* 1 hour */

	/*
	* Maximum percentage of memory to use on storing loaded metaslabs. If loading
	* a metaslab would take it over this percentage, the oldest selected metaslab
	* is automatically unloaded.
	*/
	-static int zfs_metaslab_mem_limit = 25;
	+static uint_t zfs_metaslab_mem_limit = 25;

	/*
	* Force the per-metaslab range trees to use 64-bit integers to store
	* segments. Used for debugging purposes.
	*/
	static const boolean_t zfs_metaslab_force_large_segs = B_FALSE;

	/*
	* By default we only store segments over a certain size in the size-sorted
	* metaslab trees (ms_allocatable_by_size and
	* ms_unflushed_frees_by_size). This dramatically reduces memory usage and
	* improves load and unload times at the cost of causing us to use slightly
	* larger segments than we would otherwise in some cases.
	*/
	static const uint32_t metaslab_by_size_min_shift = 14;

	/*
	* If not set, we will first try normal allocation. If that fails then
	* we will do a gang allocation. If that fails then we will do a "try hard"
	* gang allocation. If that fails then we will have a multi-layer gang
	* block.
	*
	* If set, we will first try normal allocation. If that fails then
	* we will do a "try hard" allocation. If that fails we will do a gang
	* allocation. If that fails we will do a "try hard" gang allocation. If
	* that fails then we will have a multi-layer gang block.
	*/
	static int zfs_metaslab_try_hard_before_gang = B_FALSE;

	/*
	* When not trying hard, we only consider the best zfs_metaslab_find_max_tries
	* metaslabs. This improves performance, especially when there are many
	* metaslabs per vdev and the allocation can't actually be satisfied (so we
	* would otherwise iterate all the metaslabs). If there is a metaslab with a
	* worse weight but it can actually satisfy the allocation, we won't find it
	* until trying hard. This may happen if the worse metaslab is not loaded
	* (and the true weight is better than we have calculated), or due to weight
	* bucketization. E.g. we are looking for a 60K segment, and the best
	* metaslabs all have free segments in the 32-63K bucket, but the best
	* zfs_metaslab_find_max_tries metaslabs have ms_max_size <60KB, and a
	* subsequent metaslab has ms_max_size >60KB (but fewer segments in this
	* bucket, and therefore a lower weight).
	*/
	-static int zfs_metaslab_find_max_tries = 100;
	+static uint_t zfs_metaslab_find_max_tries = 100;

	static uint64_t metaslab_weight(metaslab_t *, boolean_t);
	static void metaslab_set_fragmentation(metaslab_t *, boolean_t);
	static void metaslab_free_impl(vdev_t *, uint64_t, uint64_t, boolean_t);
	static void metaslab_check_free_impl(vdev_t *, uint64_t, uint64_t);

	static void metaslab_passivate(metaslab_t *msp, uint64_t weight);
	static uint64_t metaslab_weight_from_range_tree(metaslab_t *msp);
	static void metaslab_flush_update(metaslab_t , dmu_tx_t );
	static unsigned int metaslab_idx_func(multilist_t , void );
	static void metaslab_evict(metaslab_t *, uint64_t);
	static void metaslab_rt_add(range_tree_t rt, range_seg_t rs, void *arg);
	kmem_cache_t *metaslab_alloc_trace_cache;

	typedef struct metaslab_stats {
	kstat_named_t metaslabstat_trace_over_limit;
	kstat_named_t metaslabstat_reload_tree;
	kstat_named_t metaslabstat_too_many_tries;
	kstat_named_t metaslabstat_try_hard;
	} metaslab_stats_t;

	static metaslab_stats_t metaslab_stats = {
	{ "trace_over_limit", KSTAT_DATA_UINT64 },
	{ "reload_tree", KSTAT_DATA_UINT64 },
	{ "too_many_tries", KSTAT_DATA_UINT64 },
	{ "try_hard", KSTAT_DATA_UINT64 },
	};

	#define METASLABSTAT_BUMP(stat) \
	atomic_inc_64(&metaslab_stats.stat.value.ui64);


	static kstat_t *metaslab_ksp;

	void
	metaslab_stat_init(void)
	{
	ASSERT(metaslab_alloc_trace_cache == NULL);
	metaslab_alloc_trace_cache = kmem_cache_create(
	"metaslab_alloc_trace_cache", sizeof (metaslab_alloc_trace_t),
	0, NULL, NULL, NULL, NULL, NULL, 0);
	metaslab_ksp = kstat_create("zfs", 0, "metaslab_stats",
	"misc", KSTAT_TYPE_NAMED, sizeof (metaslab_stats) /
	sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
	if (metaslab_ksp != NULL) {
	metaslab_ksp->ks_data = &metaslab_stats;
	kstat_install(metaslab_ksp);
	}
	}

	void
	metaslab_stat_fini(void)
	{
	if (metaslab_ksp != NULL) {
	kstat_delete(metaslab_ksp);
	metaslab_ksp = NULL;
	}

	kmem_cache_destroy(metaslab_alloc_trace_cache);
	metaslab_alloc_trace_cache = NULL;
	}

	/*
	* ==========================================================================
	* Metaslab classes
	* ==========================================================================
	*/
	metaslab_class_t *
	metaslab_class_create(spa_t spa, const metaslab_ops_t ops)
	{
	metaslab_class_t *mc;

	mc = kmem_zalloc(offsetof(metaslab_class_t,
	mc_allocator[spa->spa_alloc_count]), KM_SLEEP);

	mc->mc_spa = spa;
	mc->mc_ops = ops;
	mutex_init(&mc->mc_lock, NULL, MUTEX_DEFAULT, NULL);
	multilist_create(&mc->mc_metaslab_txg_list, sizeof (metaslab_t),
	offsetof(metaslab_t, ms_class_txg_node), metaslab_idx_func);
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
	mca->mca_rotor = NULL;
	zfs_refcount_create_tracked(&mca->mca_alloc_slots);
	}

	return (mc);
	}

	void
	metaslab_class_destroy(metaslab_class_t *mc)
	{
	spa_t *spa = mc->mc_spa;

	ASSERT(mc->mc_alloc == 0);
	ASSERT(mc->mc_deferred == 0);
	ASSERT(mc->mc_space == 0);
	ASSERT(mc->mc_dspace == 0);

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
	ASSERT(mca->mca_rotor == NULL);
	zfs_refcount_destroy(&mca->mca_alloc_slots);
	}
	mutex_destroy(&mc->mc_lock);
	multilist_destroy(&mc->mc_metaslab_txg_list);
	kmem_free(mc, offsetof(metaslab_class_t,
	mc_allocator[spa->spa_alloc_count]));
	}

	int
	metaslab_class_validate(metaslab_class_t *mc)
	{
	metaslab_group_t *mg;
	vdev_t *vd;

	/*
	* Must hold one of the spa_config locks.
	*/
	ASSERT(spa_config_held(mc->mc_spa, SCL_ALL, RW_READER) \|\|
	spa_config_held(mc->mc_spa, SCL_ALL, RW_WRITER));

	if ((mg = mc->mc_allocator[0].mca_rotor) == NULL)
	return (0);

	do {
	vd = mg->mg_vd;
	ASSERT(vd->vdev_mg != NULL);
	ASSERT3P(vd->vdev_top, ==, vd);
	ASSERT3P(mg->mg_class, ==, mc);
	ASSERT3P(vd->vdev_ops, !=, &vdev_hole_ops);
	} while ((mg = mg->mg_next) != mc->mc_allocator[0].mca_rotor);

	return (0);
	}

	static void
	metaslab_class_space_update(metaslab_class_t *mc, int64_t alloc_delta,
	int64_t defer_delta, int64_t space_delta, int64_t dspace_delta)
	{
	atomic_add_64(&mc->mc_alloc, alloc_delta);
	atomic_add_64(&mc->mc_deferred, defer_delta);
	atomic_add_64(&mc->mc_space, space_delta);
	atomic_add_64(&mc->mc_dspace, dspace_delta);
	}

	uint64_t
	metaslab_class_get_alloc(metaslab_class_t *mc)
	{
	return (mc->mc_alloc);
	}

	uint64_t
	metaslab_class_get_deferred(metaslab_class_t *mc)
	{
	return (mc->mc_deferred);
	}

	uint64_t
	metaslab_class_get_space(metaslab_class_t *mc)
	{
	return (mc->mc_space);
	}

	uint64_t
	metaslab_class_get_dspace(metaslab_class_t *mc)
	{
	return (spa_deflate(mc->mc_spa) ? mc->mc_dspace : mc->mc_space);
	}

	void
	metaslab_class_histogram_verify(metaslab_class_t *mc)
	{
	spa_t *spa = mc->mc_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t *mc_hist;
	int i;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
	return;

	mc_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	KM_SLEEP);

	mutex_enter(&mc->mc_lock);
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = vdev_get_mg(tvd, mc);

	/*
	* Skip any holes, uninitialized top-levels, or
	* vdevs that are not in this metalab class.
	*/
	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
	mc_hist[i] += mg->mg_histogram[i];
	}

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	VERIFY3U(mc_hist[i], ==, mc->mc_histogram[i]);
	}

	mutex_exit(&mc->mc_lock);
	kmem_free(mc_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
	}

	/*
	* Calculate the metaslab class's fragmentation metric. The metric
	* is weighted based on the space contribution of each metaslab group.
	* The return value will be a number between 0 and 100 (inclusive), or
	* ZFS_FRAG_INVALID if the metric has not been set. See comment above the
	* zfs_frag_table for more information about the metric.
	*/
	uint64_t
	metaslab_class_fragmentation(metaslab_class_t *mc)
	{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t fragmentation = 0;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);

	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = tvd->vdev_mg;

	/*
	* Skip any holes, uninitialized top-levels,
	* or vdevs that are not in this metalab class.
	*/
	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	/*
	* If a metaslab group does not contain a fragmentation
	* metric then just bail out.
	*/
	if (mg->mg_fragmentation == ZFS_FRAG_INVALID) {
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (ZFS_FRAG_INVALID);
	}

	/*
	* Determine how much this metaslab_group is contributing
	* to the overall pool fragmentation metric.
	*/
	fragmentation += mg->mg_fragmentation *
	metaslab_group_get_space(mg);
	}
	fragmentation /= metaslab_class_get_space(mc);

	ASSERT3U(fragmentation, <=, 100);
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (fragmentation);
	}

	/*
	* Calculate the amount of expandable space that is available in
	* this metaslab class. If a device is expanded then its expandable
	* space will be the amount of allocatable space that is currently not
	* part of this metaslab class.
	*/
	uint64_t
	metaslab_class_expandable_space(metaslab_class_t *mc)
	{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t space = 0;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = tvd->vdev_mg;

	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	/*
	* Calculate if we have enough space to add additional
	* metaslabs. We report the expandable space in terms
	* of the metaslab size since that's the unit of expansion.
	*/
	space += P2ALIGN(tvd->vdev_max_asize - tvd->vdev_asize,
	1ULL << tvd->vdev_ms_shift);
	}
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (space);
	}

	void
	metaslab_class_evict_old(metaslab_class_t *mc, uint64_t txg)
	{
	multilist_t *ml = &mc->mc_metaslab_txg_list;
	for (int i = 0; i < multilist_get_num_sublists(ml); i++) {
	multilist_sublist_t *mls = multilist_sublist_lock(ml, i);
	metaslab_t *msp = multilist_sublist_head(mls);
	multilist_sublist_unlock(mls);
	while (msp != NULL) {
	mutex_enter(&msp->ms_lock);

	/*
	* If the metaslab has been removed from the list
	* (which could happen if we were at the memory limit
	* and it was evicted during this loop), then we can't
	* proceed and we should restart the sublist.
	*/
	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	mutex_exit(&msp->ms_lock);
	i--;
	break;
	}
	mls = multilist_sublist_lock(ml, i);
	metaslab_t *next_msp = multilist_sublist_next(mls, msp);
	multilist_sublist_unlock(mls);
	if (txg >
	msp->ms_selected_txg + metaslab_unload_delay &&
	gethrtime() > msp->ms_selected_time +
	(uint64_t)MSEC2NSEC(metaslab_unload_delay_ms)) {
	metaslab_evict(msp, txg);
	} else {
	/*
	* Once we've hit a metaslab selected too
	* recently to evict, we're done evicting for
	* now.
	*/
	mutex_exit(&msp->ms_lock);
	break;
	}
	mutex_exit(&msp->ms_lock);
	msp = next_msp;
	}
	}
	}

	static int
	metaslab_compare(const void x1, const void x2)
	{
	const metaslab_t m1 = (const metaslab_t )x1;
	const metaslab_t m2 = (const metaslab_t )x2;

	int sort1 = 0;
	int sort2 = 0;
	if (m1->ms_allocator != -1 && m1->ms_primary)
	sort1 = 1;
	else if (m1->ms_allocator != -1 && !m1->ms_primary)
	sort1 = 2;
	if (m2->ms_allocator != -1 && m2->ms_primary)
	sort2 = 1;
	else if (m2->ms_allocator != -1 && !m2->ms_primary)
	sort2 = 2;

	/*
	* Sort inactive metaslabs first, then primaries, then secondaries. When
	* selecting a metaslab to allocate from, an allocator first tries its
	* primary, then secondary active metaslab. If it doesn't have active
	* metaslabs, or can't allocate from them, it searches for an inactive
	* metaslab to activate. If it can't find a suitable one, it will steal
	* a primary or secondary metaslab from another allocator.
	*/
	if (sort1 < sort2)
	return (-1);
	if (sort1 > sort2)
	return (1);

	int cmp = TREE_CMP(m2->ms_weight, m1->ms_weight);
	if (likely(cmp))
	return (cmp);

	IMPLY(TREE_CMP(m1->ms_start, m2->ms_start) == 0, m1 == m2);

	return (TREE_CMP(m1->ms_start, m2->ms_start));
	}

	/*
	* ==========================================================================
	* Metaslab groups
	* ==========================================================================
	*/
	/*
	* Update the allocatable flag and the metaslab group's capacity.
	* The allocatable flag is set to true if the capacity is below
	* the zfs_mg_noalloc_threshold or has a fragmentation value that is
	* greater than zfs_mg_fragmentation_threshold. If a metaslab group
	* transitions from allocatable to non-allocatable or vice versa then the
	* metaslab group's class is updated to reflect the transition.
	*/
	static void
	metaslab_group_alloc_update(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	metaslab_class_t *mc = mg->mg_class;
	vdev_stat_t *vs = &vd->vdev_stat;
	boolean_t was_allocatable;
	boolean_t was_initialized;

	ASSERT(vd == vd->vdev_top);
	ASSERT3U(spa_config_held(mc->mc_spa, SCL_ALLOC, RW_READER), ==,
	SCL_ALLOC);

	mutex_enter(&mg->mg_lock);
	was_allocatable = mg->mg_allocatable;
	was_initialized = mg->mg_initialized;

	mg->mg_free_capacity = ((vs->vs_space - vs->vs_alloc) * 100) /
	(vs->vs_space + 1);

	mutex_enter(&mc->mc_lock);

	/*
	* If the metaslab group was just added then it won't
	* have any space until we finish syncing out this txg.
	* At that point we will consider it initialized and available
	* for allocations. We also don't consider non-activated
	* metaslab groups (e.g. vdevs that are in the middle of being removed)
	* to be initialized, because they can't be used for allocation.
	*/
	mg->mg_initialized = metaslab_group_initialized(mg);
	if (!was_initialized && mg->mg_initialized) {
	mc->mc_groups++;
	} else if (was_initialized && !mg->mg_initialized) {
	ASSERT3U(mc->mc_groups, >, 0);
	mc->mc_groups--;
	}
	if (mg->mg_initialized)
	mg->mg_no_free_space = B_FALSE;

	/*
	* A metaslab group is considered allocatable if it has plenty
	* of free space or is not heavily fragmented. We only take
	* fragmentation into account if the metaslab group has a valid
	* fragmentation metric (i.e. a value between 0 and 100).
	*/
	mg->mg_allocatable = (mg->mg_activation_count > 0 &&
	mg->mg_free_capacity > zfs_mg_noalloc_threshold &&
	(mg->mg_fragmentation == ZFS_FRAG_INVALID \|\|
	mg->mg_fragmentation <= zfs_mg_fragmentation_threshold));

	/*
	* The mc_alloc_groups maintains a count of the number of
	* groups in this metaslab class that are still above the
	* zfs_mg_noalloc_threshold. This is used by the allocating
	* threads to determine if they should avoid allocations to
	* a given group. The allocator will avoid allocations to a group
	* if that group has reached or is below the zfs_mg_noalloc_threshold
	* and there are still other groups that are above the threshold.
	* When a group transitions from allocatable to non-allocatable or
	* vice versa we update the metaslab class to reflect that change.
	* When the mc_alloc_groups value drops to 0 that means that all
	* groups have reached the zfs_mg_noalloc_threshold making all groups
	* eligible for allocations. This effectively means that all devices
	* are balanced again.
	*/
	if (was_allocatable && !mg->mg_allocatable)
	mc->mc_alloc_groups--;
	else if (!was_allocatable && mg->mg_allocatable)
	mc->mc_alloc_groups++;
	mutex_exit(&mc->mc_lock);

	mutex_exit(&mg->mg_lock);
	}

	int
	metaslab_sort_by_flushed(const void va, const void vb)
	{
	const metaslab_t *a = va;
	const metaslab_t *b = vb;

	int cmp = TREE_CMP(a->ms_unflushed_txg, b->ms_unflushed_txg);
	if (likely(cmp))
	return (cmp);

	uint64_t a_vdev_id = a->ms_group->mg_vd->vdev_id;
	uint64_t b_vdev_id = b->ms_group->mg_vd->vdev_id;
	cmp = TREE_CMP(a_vdev_id, b_vdev_id);
	if (cmp)
	return (cmp);

	return (TREE_CMP(a->ms_id, b->ms_id));
	}

	metaslab_group_t *
	metaslab_group_create(metaslab_class_t mc, vdev_t vd, int allocators)
	{
	metaslab_group_t *mg;

	mg = kmem_zalloc(offsetof(metaslab_group_t,
	mg_allocator[allocators]), KM_SLEEP);
	mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&mg->mg_ms_disabled_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&mg->mg_ms_disabled_cv, NULL, CV_DEFAULT, NULL);
	avl_create(&mg->mg_metaslab_tree, metaslab_compare,
	sizeof (metaslab_t), offsetof(metaslab_t, ms_group_node));
	mg->mg_vd = vd;
	mg->mg_class = mc;
	mg->mg_activation_count = 0;
	mg->mg_initialized = B_FALSE;
	mg->mg_no_free_space = B_TRUE;
	mg->mg_allocators = allocators;

	for (int i = 0; i < allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	zfs_refcount_create_tracked(&mga->mga_alloc_queue_depth);
	}

	mg->mg_taskq = taskq_create("metaslab_group_taskq", metaslab_load_pct,
	maxclsyspri, 10, INT_MAX, TASKQ_THREADS_CPU_PCT \| TASKQ_DYNAMIC);

	return (mg);
	}

	void
	metaslab_group_destroy(metaslab_group_t *mg)
	{
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	/*
	* We may have gone below zero with the activation count
	* either because we never activated in the first place or
	* because we're done, and possibly removing the vdev.
	*/
	ASSERT(mg->mg_activation_count <= 0);

	taskq_destroy(mg->mg_taskq);
	avl_destroy(&mg->mg_metaslab_tree);
	mutex_destroy(&mg->mg_lock);
	mutex_destroy(&mg->mg_ms_disabled_lock);
	cv_destroy(&mg->mg_ms_disabled_cv);

	for (int i = 0; i < mg->mg_allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	zfs_refcount_destroy(&mga->mga_alloc_queue_depth);
	}
	kmem_free(mg, offsetof(metaslab_group_t,
	mg_allocator[mg->mg_allocators]));
	}

	void
	metaslab_group_activate(metaslab_group_t *mg)
	{
	metaslab_class_t *mc = mg->mg_class;
	spa_t *spa = mc->mc_spa;
	metaslab_group_t mgprev, mgnext;

	ASSERT3U(spa_config_held(spa, SCL_ALLOC, RW_WRITER), !=, 0);

	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	ASSERT(mg->mg_activation_count <= 0);

	if (++mg->mg_activation_count <= 0)
	return;

	mg->mg_aliquot = metaslab_aliquot * MAX(1,
	vdev_get_ndisks(mg->mg_vd) - vdev_get_nparity(mg->mg_vd));
	metaslab_group_alloc_update(mg);

	if ((mgprev = mc->mc_allocator[0].mca_rotor) == NULL) {
	mg->mg_prev = mg;
	mg->mg_next = mg;
	} else {
	mgnext = mgprev->mg_next;
	mg->mg_prev = mgprev;
	mg->mg_next = mgnext;
	mgprev->mg_next = mg;
	mgnext->mg_prev = mg;
	}
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mc->mc_allocator[i].mca_rotor = mg;
	mg = mg->mg_next;
	}
	}

	/*
	* Passivate a metaslab group and remove it from the allocation rotor.
	* Callers must hold both the SCL_ALLOC and SCL_ZIO lock prior to passivating
	* a metaslab group. This function will momentarily drop spa_config_locks
	* that are lower than the SCL_ALLOC lock (see comment below).
	*/
	void
	metaslab_group_passivate(metaslab_group_t *mg)
	{
	metaslab_class_t *mc = mg->mg_class;
	spa_t *spa = mc->mc_spa;
	metaslab_group_t mgprev, mgnext;
	int locks = spa_config_held(spa, SCL_ALL, RW_WRITER);

	ASSERT3U(spa_config_held(spa, SCL_ALLOC \| SCL_ZIO, RW_WRITER), ==,
	(SCL_ALLOC \| SCL_ZIO));

	if (--mg->mg_activation_count != 0) {
	for (int i = 0; i < spa->spa_alloc_count; i++)
	ASSERT(mc->mc_allocator[i].mca_rotor != mg);
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	ASSERT(mg->mg_activation_count < 0);
	return;
	}

	/*
	* The spa_config_lock is an array of rwlocks, ordered as
	* follows (from highest to lowest):
	* SCL_CONFIG > SCL_STATE > SCL_L2ARC > SCL_ALLOC >
	* SCL_ZIO > SCL_FREE > SCL_VDEV
	* (For more information about the spa_config_lock see spa_misc.c)
	* The higher the lock, the broader its coverage. When we passivate
	* a metaslab group, we must hold both the SCL_ALLOC and the SCL_ZIO
	* config locks. However, the metaslab group's taskq might be trying
	* to preload metaslabs so we must drop the SCL_ZIO lock and any
	* lower locks to allow the I/O to complete. At a minimum,
	* we continue to hold the SCL_ALLOC lock, which prevents any future
	* allocations from taking place and any changes to the vdev tree.
	*/
	spa_config_exit(spa, locks & ~(SCL_ZIO - 1), spa);
	taskq_wait_outstanding(mg->mg_taskq, 0);
	spa_config_enter(spa, locks & ~(SCL_ZIO - 1), spa, RW_WRITER);
	metaslab_group_alloc_update(mg);
	for (int i = 0; i < mg->mg_allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	metaslab_t *msp = mga->mga_primary;
	if (msp != NULL) {
	mutex_enter(&msp->ms_lock);
	metaslab_passivate(msp,
	metaslab_weight_from_range_tree(msp));
	mutex_exit(&msp->ms_lock);
	}
	msp = mga->mga_secondary;
	if (msp != NULL) {
	mutex_enter(&msp->ms_lock);
	metaslab_passivate(msp,
	metaslab_weight_from_range_tree(msp));
	mutex_exit(&msp->ms_lock);
	}
	}

	mgprev = mg->mg_prev;
	mgnext = mg->mg_next;

	if (mg == mgnext) {
	mgnext = NULL;
	} else {
	mgprev->mg_next = mgnext;
	mgnext->mg_prev = mgprev;
	}
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	if (mc->mc_allocator[i].mca_rotor == mg)
	mc->mc_allocator[i].mca_rotor = mgnext;
	}

	mg->mg_prev = NULL;
	mg->mg_next = NULL;
	}

	boolean_t
	metaslab_group_initialized(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	vdev_stat_t *vs = &vd->vdev_stat;

	return (vs->vs_space != 0 && mg->mg_activation_count > 0);
	}

	uint64_t
	metaslab_group_get_space(metaslab_group_t *mg)
	{
	/*
	* Note that the number of nodes in mg_metaslab_tree may be one less
	* than vdev_ms_count, due to the embedded log metaslab.
	*/
	mutex_enter(&mg->mg_lock);
	uint64_t ms_count = avl_numnodes(&mg->mg_metaslab_tree);
	mutex_exit(&mg->mg_lock);
	return ((1ULL << mg->mg_vd->vdev_ms_shift) * ms_count);
	}

	void
	metaslab_group_histogram_verify(metaslab_group_t *mg)
	{
	uint64_t *mg_hist;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
	return;

	mg_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	KM_SLEEP);

	ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE, >=,
	SPACE_MAP_HISTOGRAM_SIZE + ashift);

	mutex_enter(&mg->mg_lock);
	for (metaslab_t *msp = avl_first(t);
	msp != NULL; msp = AVL_NEXT(t, msp)) {
	VERIFY3P(msp->ms_group, ==, mg);
	/* skip if not active */
	if (msp->ms_sm == NULL)
	continue;

	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	mg_hist[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	}

	for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
	VERIFY3U(mg_hist[i], ==, mg->mg_histogram[i]);

	mutex_exit(&mg->mg_lock);

	kmem_free(mg_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
	}

	static void
	metaslab_group_histogram_add(metaslab_group_t mg, metaslab_t msp)
	{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
	return;

	mutex_enter(&mg->mg_lock);
	mutex_enter(&mc->mc_lock);
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
	mg->mg_histogram[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	mc->mc_histogram[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mc->mc_lock);
	mutex_exit(&mg->mg_lock);
	}

	void
	metaslab_group_histogram_remove(metaslab_group_t mg, metaslab_t msp)
	{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
	return;

	mutex_enter(&mg->mg_lock);
	mutex_enter(&mc->mc_lock);
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	ASSERT3U(mg->mg_histogram[i + ashift], >=,
	msp->ms_sm->sm_phys->smp_histogram[i]);
	ASSERT3U(mc->mc_histogram[i + ashift], >=,
	msp->ms_sm->sm_phys->smp_histogram[i]);
	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

	mg->mg_histogram[i + ashift] -=
	msp->ms_sm->sm_phys->smp_histogram[i];
	mc->mc_histogram[i + ashift] -=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mc->mc_lock);
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_group_add(metaslab_group_t mg, metaslab_t msp)
	{
	ASSERT(msp->ms_group == NULL);
	mutex_enter(&mg->mg_lock);
	msp->ms_group = mg;
	msp->ms_weight = 0;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);

	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_add(mg, msp);
	mutex_exit(&msp->ms_lock);
	}

	static void
	metaslab_group_remove(metaslab_group_t mg, metaslab_t msp)
	{
	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_remove(mg, msp);
	mutex_exit(&msp->ms_lock);

	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);

	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	multilist_sublist_unlock(mls);

	msp->ms_group = NULL;
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_group_sort_impl(metaslab_group_t mg, metaslab_t msp, uint64_t weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(MUTEX_HELD(&mg->mg_lock));
	ASSERT(msp->ms_group == mg);

	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_weight = weight;
	avl_add(&mg->mg_metaslab_tree, msp);

	}

	static void
	metaslab_group_sort(metaslab_group_t mg, metaslab_t msp, uint64_t weight)
	{
	/*
	* Although in principle the weight can be any value, in
	* practice we do not use values in the range [1, 511].
	*/
	ASSERT(weight >= SPA_MINBLOCKSIZE \|\| weight == 0);
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	mutex_enter(&mg->mg_lock);
	metaslab_group_sort_impl(mg, msp, weight);
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Calculate the fragmentation for a given metaslab group. We can use
	* a simple average here since all metaslabs within the group must have
	* the same size. The return value will be a value between 0 and 100
	* (inclusive), or ZFS_FRAG_INVALID if less than half of the metaslab in this
	* group have a fragmentation metric.
	*/
	uint64_t
	metaslab_group_fragmentation(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	uint64_t fragmentation = 0;
	uint64_t valid_ms = 0;

	for (int m = 0; m < vd->vdev_ms_count; m++) {
	metaslab_t *msp = vd->vdev_ms[m];

	if (msp->ms_fragmentation == ZFS_FRAG_INVALID)
	continue;
	if (msp->ms_group != mg)
	continue;

	valid_ms++;
	fragmentation += msp->ms_fragmentation;
	}

	if (valid_ms <= mg->mg_vd->vdev_ms_count / 2)
	return (ZFS_FRAG_INVALID);

	fragmentation /= valid_ms;
	ASSERT3U(fragmentation, <=, 100);
	return (fragmentation);
	}

	/*
	* Determine if a given metaslab group should skip allocations. A metaslab
	* group should avoid allocations if its free capacity is less than the
	* zfs_mg_noalloc_threshold or its fragmentation metric is greater than
	* zfs_mg_fragmentation_threshold and there is at least one metaslab group
	* that can still handle allocations. If the allocation throttle is enabled
	* then we skip allocations to devices that have reached their maximum
	* allocation queue depth unless the selected metaslab group is the only
	* eligible group remaining.
	*/
	static boolean_t
	metaslab_group_allocatable(metaslab_group_t mg, metaslab_group_t rotor,
	uint64_t psize, int allocator, int d)
	{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_class_t *mc = mg->mg_class;

	/*
	* We can only consider skipping this metaslab group if it's
	* in the normal metaslab class and there are other metaslab
	* groups to select from. Otherwise, we always consider it eligible
	* for allocations.
	*/
	if ((mc != spa_normal_class(spa) &&
	mc != spa_special_class(spa) &&
	mc != spa_dedup_class(spa)) \|\|
	mc->mc_groups <= 1)
	return (B_TRUE);

	/*
	* If the metaslab group's mg_allocatable flag is set (see comments
	* in metaslab_group_alloc_update() for more information) and
	* the allocation throttle is disabled then allow allocations to this
	* device. However, if the allocation throttle is enabled then
	* check if we have reached our allocation limit (mga_alloc_queue_depth)
	* to determine if we should allow allocations to this metaslab group.
	* If all metaslab groups are no longer considered allocatable
	* (mc_alloc_groups == 0) or we're trying to allocate the smallest
	* gang block size then we allow allocations on this metaslab group
	* regardless of the mg_allocatable or throttle settings.
	*/
	if (mg->mg_allocatable) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	int64_t qdepth;
	uint64_t qmax = mga->mga_cur_max_alloc_queue_depth;

	if (!mc->mc_alloc_throttle_enabled)
	return (B_TRUE);

	/*
	* If this metaslab group does not have any free space, then
	* there is no point in looking further.
	*/
	if (mg->mg_no_free_space)
	return (B_FALSE);

	/*
	* Relax allocation throttling for ditto blocks. Due to
	* random imbalances in allocation it tends to push copies
	* to one vdev, that looks a bit better at the moment.
	*/
	qmax = qmax * (4 + d) / 4;

	qdepth = zfs_refcount_count(&mga->mga_alloc_queue_depth);

	/*
	* If this metaslab group is below its qmax or it's
	* the only allocatable metasable group, then attempt
	* to allocate from it.
	*/
	if (qdepth < qmax \|\| mc->mc_alloc_groups == 1)
	return (B_TRUE);
	ASSERT3U(mc->mc_alloc_groups, >, 1);

	/*
	* Since this metaslab group is at or over its qmax, we
	* need to determine if there are metaslab groups after this
	* one that might be able to handle this allocation. This is
	* racy since we can't hold the locks for all metaslab
	* groups at the same time when we make this check.
	*/
	for (metaslab_group_t *mgp = mg->mg_next;
	mgp != rotor; mgp = mgp->mg_next) {
	metaslab_group_allocator_t *mgap =
	&mgp->mg_allocator[allocator];
	qmax = mgap->mga_cur_max_alloc_queue_depth;
	qmax = qmax * (4 + d) / 4;
	qdepth =
	zfs_refcount_count(&mgap->mga_alloc_queue_depth);

	/*
	* If there is another metaslab group that
	* might be able to handle the allocation, then
	* we return false so that we skip this group.
	*/
	if (qdepth < qmax && !mgp->mg_no_free_space)
	return (B_FALSE);
	}

	/*
	* We didn't find another group to handle the allocation
	* so we can't skip this metaslab group even though
	* we are at or over our qmax.
	*/
	return (B_TRUE);

	} else if (mc->mc_alloc_groups == 0 \|\| psize == SPA_MINBLOCKSIZE) {
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* ==========================================================================
	* Range tree callbacks
	* ==========================================================================
	*/

	/*
	* Comparison function for the private size-ordered tree using 32-bit
	* ranges. Tree is sorted by size, larger sizes at the end of the tree.
	*/
	static int
	metaslab_rangesize32_compare(const void x1, const void x2)
	{
	const range_seg32_t *r1 = x1;
	const range_seg32_t *r2 = x2;

	uint64_t rs_size1 = r1->rs_end - r1->rs_start;
	uint64_t rs_size2 = r2->rs_end - r2->rs_start;

	int cmp = TREE_CMP(rs_size1, rs_size2);
	if (likely(cmp))
	return (cmp);

	return (TREE_CMP(r1->rs_start, r2->rs_start));
	}

	/*
	* Comparison function for the private size-ordered tree using 64-bit
	* ranges. Tree is sorted by size, larger sizes at the end of the tree.
	*/
	static int
	metaslab_rangesize64_compare(const void x1, const void x2)
	{
	const range_seg64_t *r1 = x1;
	const range_seg64_t *r2 = x2;

	uint64_t rs_size1 = r1->rs_end - r1->rs_start;
	uint64_t rs_size2 = r2->rs_end - r2->rs_start;

	int cmp = TREE_CMP(rs_size1, rs_size2);
	if (likely(cmp))
	return (cmp);

	return (TREE_CMP(r1->rs_start, r2->rs_start));
	}
	typedef struct metaslab_rt_arg {
	zfs_btree_t *mra_bt;
	uint32_t mra_floor_shift;
	} metaslab_rt_arg_t;

	struct mssa_arg {
	range_tree_t *rt;
	metaslab_rt_arg_t *mra;
	};

	static void
	metaslab_size_sorted_add(void *arg, uint64_t start, uint64_t size)
	{
	struct mssa_arg *mssap = arg;
	range_tree_t *rt = mssap->rt;
	metaslab_rt_arg_t *mrap = mssap->mra;
	range_seg_max_t seg = {0};
	rs_set_start(&seg, rt, start);
	rs_set_end(&seg, rt, start + size);
	metaslab_rt_add(rt, &seg, mrap);
	}

	static void
	metaslab_size_tree_full_load(range_tree_t *rt)
	{
	metaslab_rt_arg_t *mrap = rt->rt_arg;
	METASLABSTAT_BUMP(metaslabstat_reload_tree);
	ASSERT0(zfs_btree_numnodes(mrap->mra_bt));
	mrap->mra_floor_shift = 0;
	struct mssa_arg arg = {0};
	arg.rt = rt;
	arg.mra = mrap;
	range_tree_walk(rt, metaslab_size_sorted_add, &arg);
	}

	/*
	* Create any block allocator specific components. The current allocators
	* rely on using both a size-ordered range_tree_t and an array of uint64_t's.
	*/
	static void
	metaslab_rt_create(range_tree_t rt, void arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	size_t size;
	int (compare) (const void , const void *);
	switch (rt->rt_type) {
	case RANGE_SEG32:
	size = sizeof (range_seg32_t);
	compare = metaslab_rangesize32_compare;
	break;
	case RANGE_SEG64:
	size = sizeof (range_seg64_t);
	compare = metaslab_rangesize64_compare;
	break;
	default:
	panic("Invalid range seg type %d", rt->rt_type);
	}
	zfs_btree_create(size_tree, compare, size);
	mrap->mra_floor_shift = metaslab_by_size_min_shift;
	}

	static void
	metaslab_rt_destroy(range_tree_t rt, void arg)
	{
	(void) rt;
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	zfs_btree_destroy(size_tree);
	kmem_free(mrap, sizeof (*mrap));
	}

	static void
	metaslab_rt_add(range_tree_t rt, range_seg_t rs, void *arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	if (rs_get_end(rs, rt) - rs_get_start(rs, rt) <
	- (1 << mrap->mra_floor_shift))
	+ (1ULL << mrap->mra_floor_shift))
	return;

	zfs_btree_add(size_tree, rs);
	}

	static void
	metaslab_rt_remove(range_tree_t rt, range_seg_t rs, void *arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	- if (rs_get_end(rs, rt) - rs_get_start(rs, rt) < (1 <<
	+ if (rs_get_end(rs, rt) - rs_get_start(rs, rt) < (1ULL <<
	mrap->mra_floor_shift))
	return;

	zfs_btree_remove(size_tree, rs);
	}

	static void
	metaslab_rt_vacate(range_tree_t rt, void arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;
	zfs_btree_clear(size_tree);
	zfs_btree_destroy(size_tree);

	metaslab_rt_create(rt, arg);
	}

	static const range_tree_ops_t metaslab_rt_ops = {
	.rtop_create = metaslab_rt_create,
	.rtop_destroy = metaslab_rt_destroy,
	.rtop_add = metaslab_rt_add,
	.rtop_remove = metaslab_rt_remove,
	.rtop_vacate = metaslab_rt_vacate
	};

	/*
	* ==========================================================================
	* Common allocator routines
	* ==========================================================================
	*/

	/*
	* Return the maximum contiguous segment within the metaslab.
	*/
	uint64_t
	metaslab_largest_allocatable(metaslab_t *msp)
	{
	zfs_btree_t *t = &msp->ms_allocatable_by_size;
	range_seg_t *rs;

	if (t == NULL)
	return (0);
	if (zfs_btree_numnodes(t) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);

	rs = zfs_btree_last(t, NULL);
	if (rs == NULL)
	return (0);

	return (rs_get_end(rs, msp->ms_allocatable) - rs_get_start(rs,
	msp->ms_allocatable));
	}

	/*
	* Return the maximum contiguous segment within the unflushed frees of this
	* metaslab.
	*/
	static uint64_t
	metaslab_largest_unflushed_free(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if (msp->ms_unflushed_frees == NULL)
	return (0);

	if (zfs_btree_numnodes(&msp->ms_unflushed_frees_by_size) == 0)
	metaslab_size_tree_full_load(msp->ms_unflushed_frees);
	range_seg_t *rs = zfs_btree_last(&msp->ms_unflushed_frees_by_size,
	NULL);
	if (rs == NULL)
	return (0);

	/*
	* When a range is freed from the metaslab, that range is added to
	* both the unflushed frees and the deferred frees. While the block
	* will eventually be usable, if the metaslab were loaded the range
	* would not be added to the ms_allocatable tree until TXG_DEFER_SIZE
	* txgs had passed. As a result, when attempting to estimate an upper
	* bound for the largest currently-usable free segment in the
	* metaslab, we need to not consider any ranges currently in the defer
	* trees. This algorithm approximates the largest available chunk in
	* the largest range in the unflushed_frees tree by taking the first
	* chunk. While this may be a poor estimate, it should only remain so
	* briefly and should eventually self-correct as frees are no longer
	* deferred. Similar logic applies to the ms_freed tree. See
	* metaslab_load() for more details.
	*
	* There are two primary sources of inaccuracy in this estimate. Both
	* are tolerated for performance reasons. The first source is that we
	* only check the largest segment for overlaps. Smaller segments may
	* have more favorable overlaps with the other trees, resulting in
	* larger usable chunks. Second, we only look at the first chunk in
	* the largest segment; there may be other usable chunks in the
	* largest segment, but we ignore them.
	*/
	uint64_t rstart = rs_get_start(rs, msp->ms_unflushed_frees);
	uint64_t rsize = rs_get_end(rs, msp->ms_unflushed_frees) - rstart;
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	uint64_t start = 0;
	uint64_t size = 0;
	boolean_t found = range_tree_find_in(msp->ms_defer[t], rstart,
	rsize, &start, &size);
	if (found) {
	if (rstart == start)
	return (0);
	rsize = start - rstart;
	}
	}

	uint64_t start = 0;
	uint64_t size = 0;
	boolean_t found = range_tree_find_in(msp->ms_freed, rstart,
	rsize, &start, &size);
	if (found)
	rsize = start - rstart;

	return (rsize);
	}

	static range_seg_t *
	metaslab_block_find(zfs_btree_t t, range_tree_t rt, uint64_t start,
	uint64_t size, zfs_btree_index_t *where)
	{
	range_seg_t *rs;
	range_seg_max_t rsearch;

	rs_set_start(&rsearch, rt, start);
	rs_set_end(&rsearch, rt, start + size);

	rs = zfs_btree_find(t, &rsearch, where);
	if (rs == NULL) {
	rs = zfs_btree_next(t, where, where);
	}

	return (rs);
	}

	#if defined(WITH_DF_BLOCK_ALLOCATOR) \|\| \
	defined(WITH_CF_BLOCK_ALLOCATOR)

	/*
	* This is a helper function that can be used by the allocator to find a
	* suitable block to allocate. This will search the specified B-tree looking
	* for a block that matches the specified criteria.
	*/
	static uint64_t
	metaslab_block_picker(range_tree_t rt, uint64_t cursor, uint64_t size,
	uint64_t max_search)
	{
	if (*cursor == 0)
	*cursor = rt->rt_start;
	zfs_btree_t *bt = &rt->rt_root;
	zfs_btree_index_t where;
	range_seg_t rs = metaslab_block_find(bt, rt, cursor, size, &where);
	uint64_t first_found;
	int count_searched = 0;

	if (rs != NULL)
	first_found = rs_get_start(rs, rt);

	while (rs != NULL && (rs_get_start(rs, rt) - first_found <=
	max_search \|\| count_searched < metaslab_min_search_count)) {
	uint64_t offset = rs_get_start(rs, rt);
	if (offset + size <= rs_get_end(rs, rt)) {
	*cursor = offset + size;
	return (offset);
	}
	rs = zfs_btree_next(bt, &where, &where);
	count_searched++;
	}

	*cursor = 0;
	return (-1ULL);
	}
	#endif /* WITH_DF/CF_BLOCK_ALLOCATOR */

	#if defined(WITH_DF_BLOCK_ALLOCATOR)
	/*
	* ==========================================================================
	* Dynamic Fit (df) block allocator
	*
	* Search for a free chunk of at least this size, starting from the last
	* offset (for this alignment of block) looking for up to
	* metaslab_df_max_search bytes (16MB). If a large enough free chunk is not
	* found within 16MB, then return a free chunk of exactly the requested size (or
	* larger).
	*
	* If it seems like searching from the last offset will be unproductive, skip
	* that and just return a free chunk of exactly the requested size (or larger).
	* This is based on metaslab_df_alloc_threshold and metaslab_df_free_pct. This
	* mechanism is probably not very useful and may be removed in the future.
	*
	* The behavior when not searching can be changed to return the largest free
	* chunk, instead of a free chunk of exactly the requested size, by setting
	* metaslab_df_use_largest_segment.
	* ==========================================================================
	*/
	static uint64_t
	metaslab_df_alloc(metaslab_t *msp, uint64_t size)
	{
	/*
	* Find the largest power of 2 block size that evenly divides the
	* requested size. This is used to try to allocate blocks with similar
	* alignment from the same area of the metaslab (i.e. same cursor
	* bucket) but it does not guarantee that other allocations sizes
	* may exist in the same region.
	*/
	uint64_t align = size & -size;
	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
	range_tree_t *rt = msp->ms_allocatable;
	- int free_pct = range_tree_space(rt) * 100 / msp->ms_size;
	+ uint_t free_pct = range_tree_space(rt) * 100 / msp->ms_size;
	uint64_t offset;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* If we're running low on space, find a segment based on size,
	* rather than iterating based on offset.
	*/
	if (metaslab_largest_allocatable(msp) < metaslab_df_alloc_threshold \|\|
	free_pct < metaslab_df_free_pct) {
	offset = -1;
	} else {
	offset = metaslab_block_picker(rt,
	cursor, size, metaslab_df_max_search);
	}

	if (offset == -1) {
	range_seg_t *rs;
	if (zfs_btree_numnodes(&msp->ms_allocatable_by_size) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);

	if (metaslab_df_use_largest_segment) {
	/* use largest free segment */
	rs = zfs_btree_last(&msp->ms_allocatable_by_size, NULL);
	} else {
	zfs_btree_index_t where;
	/* use segment of this size, or next largest */
	rs = metaslab_block_find(&msp->ms_allocatable_by_size,
	rt, msp->ms_start, size, &where);
	}
	if (rs != NULL && rs_get_start(rs, rt) + size <= rs_get_end(rs,
	rt)) {
	offset = rs_get_start(rs, rt);
	*cursor = offset + size;
	}
	}

	return (offset);
	}

	const metaslab_ops_t zfs_metaslab_ops = {
	metaslab_df_alloc
	};
	#endif /* WITH_DF_BLOCK_ALLOCATOR */

	#if defined(WITH_CF_BLOCK_ALLOCATOR)
	/*
	* ==========================================================================
	* Cursor fit block allocator -
	* Select the largest region in the metaslab, set the cursor to the beginning
	* of the range and the cursor_end to the end of the range. As allocations
	* are made advance the cursor. Continue allocating from the cursor until
	* the range is exhausted and then find a new range.
	* ==========================================================================
	*/
	static uint64_t
	metaslab_cf_alloc(metaslab_t *msp, uint64_t size)
	{
	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_t *t = &msp->ms_allocatable_by_size;
	uint64_t *cursor = &msp->ms_lbas[0];
	uint64_t *cursor_end = &msp->ms_lbas[1];
	uint64_t offset = 0;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	ASSERT3U(cursor_end, >=, cursor);

	if ((cursor + size) > cursor_end) {
	range_seg_t *rs;

	if (zfs_btree_numnodes(t) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);
	rs = zfs_btree_last(t, NULL);
	if (rs == NULL \|\| (rs_get_end(rs, rt) - rs_get_start(rs, rt)) <
	size)
	return (-1ULL);

	*cursor = rs_get_start(rs, rt);
	*cursor_end = rs_get_end(rs, rt);
	}

	offset = *cursor;
	*cursor += size;

	return (offset);
	}

	const metaslab_ops_t zfs_metaslab_ops = {
	metaslab_cf_alloc
	};
	#endif /* WITH_CF_BLOCK_ALLOCATOR */

	#if defined(WITH_NDF_BLOCK_ALLOCATOR)
	/*
	* ==========================================================================
	* New dynamic fit allocator -
	* Select a region that is large enough to allocate 2^metaslab_ndf_clump_shift
	* contiguous blocks. If no region is found then just use the largest segment
	* that remains.
	* ==========================================================================
	*/

	/*
	* Determines desired number of contiguous blocks (2^metaslab_ndf_clump_shift)
	* to request from the allocator.
	*/
	uint64_t metaslab_ndf_clump_shift = 4;

	static uint64_t
	metaslab_ndf_alloc(metaslab_t *msp, uint64_t size)
	{
	zfs_btree_t *t = &msp->ms_allocatable->rt_root;
	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_index_t where;
	range_seg_t *rs;
	range_seg_max_t rsearch;
	uint64_t hbit = highbit64(size);
	uint64_t *cursor = &msp->ms_lbas[hbit - 1];
	uint64_t max_size = metaslab_largest_allocatable(msp);

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if (max_size < size)
	return (-1ULL);

	rs_set_start(&rsearch, rt, *cursor);
	rs_set_end(&rsearch, rt, *cursor + size);

	rs = zfs_btree_find(t, &rsearch, &where);
	if (rs == NULL \|\| (rs_get_end(rs, rt) - rs_get_start(rs, rt)) < size) {
	t = &msp->ms_allocatable_by_size;

	rs_set_start(&rsearch, rt, 0);
	rs_set_end(&rsearch, rt, MIN(max_size, 1ULL << (hbit +
	metaslab_ndf_clump_shift)));

	rs = zfs_btree_find(t, &rsearch, &where);
	if (rs == NULL)
	rs = zfs_btree_next(t, &where, &where);
	ASSERT(rs != NULL);
	}

	if ((rs_get_end(rs, rt) - rs_get_start(rs, rt)) >= size) {
	*cursor = rs_get_start(rs, rt) + size;
	return (rs_get_start(rs, rt));
	}
	return (-1ULL);
	}

	const metaslab_ops_t zfs_metaslab_ops = {
	metaslab_ndf_alloc
	};
	#endif /* WITH_NDF_BLOCK_ALLOCATOR */


	/*
	* ==========================================================================
	* Metaslabs
	* ==========================================================================
	*/

	/*
	* Wait for any in-progress metaslab loads to complete.
	*/
	static void
	metaslab_load_wait(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	while (msp->ms_loading) {
	ASSERT(!msp->ms_loaded);
	cv_wait(&msp->ms_load_cv, &msp->ms_lock);
	}
	}

	/*
	* Wait for any in-progress flushing to complete.
	*/
	static void
	metaslab_flush_wait(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	while (msp->ms_flushing)
	cv_wait(&msp->ms_flush_cv, &msp->ms_lock);
	}

	static unsigned int
	metaslab_idx_func(multilist_t ml, void arg)
	{
	metaslab_t *msp = arg;

	/*
	* ms_id values are allocated sequentially, so full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)msp->ms_id % multilist_get_num_sublists(ml));
	}

	uint64_t
	metaslab_allocated_space(metaslab_t *msp)
	{
	return (msp->ms_allocated_space);
	}

	/*
	* Verify that the space accounting on disk matches the in-core range_trees.
	*/
	static void
	metaslab_verify_space(metaslab_t *msp, uint64_t txg)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	uint64_t allocating = 0;
	uint64_t sm_free_space, msp_free_space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(!msp->ms_condensing);

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	/*
	* We can only verify the metaslab space when we're called
	* from syncing context with a loaded metaslab that has an
	* allocated space map. Calling this in non-syncing context
	* does not provide a consistent view of the metaslab since
	* we're performing allocations in the future.
	*/
	if (txg != spa_syncing_txg(spa) \|\| msp->ms_sm == NULL \|\|
	!msp->ms_loaded)
	return;

	/*
	* Even though the smp_alloc field can get negative,
	* when it comes to a metaslab's space map, that should
	* never be the case.
	*/
	ASSERT3S(space_map_allocated(msp->ms_sm), >=, 0);

	ASSERT3U(space_map_allocated(msp->ms_sm), >=,
	range_tree_space(msp->ms_unflushed_frees));

	ASSERT3U(metaslab_allocated_space(msp), ==,
	space_map_allocated(msp->ms_sm) +
	range_tree_space(msp->ms_unflushed_allocs) -
	range_tree_space(msp->ms_unflushed_frees));

	sm_free_space = msp->ms_size - metaslab_allocated_space(msp);

	/*
	* Account for future allocations since we would have
	* already deducted that space from the ms_allocatable.
	*/
	for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
	allocating +=
	range_tree_space(msp->ms_allocating[(txg + t) & TXG_MASK]);
	}
	ASSERT3U(allocating + msp->ms_allocated_this_txg, ==,
	msp->ms_allocating_total);

	ASSERT3U(msp->ms_deferspace, ==,
	range_tree_space(msp->ms_defer[0]) +
	range_tree_space(msp->ms_defer[1]));

	msp_free_space = range_tree_space(msp->ms_allocatable) + allocating +
	msp->ms_deferspace + range_tree_space(msp->ms_freed);

	VERIFY3U(sm_free_space, ==, msp_free_space);
	}

	static void
	metaslab_aux_histograms_clear(metaslab_t *msp)
	{
	/*
	* Auxiliary histograms are only cleared when resetting them,
	* which can only happen while the metaslab is loaded.
	*/
	ASSERT(msp->ms_loaded);

	memset(msp->ms_synchist, 0, sizeof (msp->ms_synchist));
	for (int t = 0; t < TXG_DEFER_SIZE; t++)
	memset(msp->ms_deferhist[t], 0, sizeof (msp->ms_deferhist[t]));
	}

	static void
	metaslab_aux_histogram_add(uint64_t *histogram, uint64_t shift,
	range_tree_t *rt)
	{
	/*
	* This is modeled after space_map_histogram_add(), so refer to that
	* function for implementation details. We want this to work like
	* the space map histogram, and not the range tree histogram, as we
	* are essentially constructing a delta that will be later subtracted
	* from the space map histogram.
	*/
	int idx = 0;
	for (int i = shift; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	ASSERT3U(i, >=, idx + shift);
	histogram[idx] += rt->rt_histogram[i] << (i - idx - shift);

	if (idx < SPACE_MAP_HISTOGRAM_SIZE - 1) {
	ASSERT3U(idx + shift, ==, i);
	idx++;
	ASSERT3U(idx, <, SPACE_MAP_HISTOGRAM_SIZE);
	}
	}
	}

	/*
	* Called at every sync pass that the metaslab gets synced.
	*
	* The reason is that we want our auxiliary histograms to be updated
	* wherever the metaslab's space map histogram is updated. This way
	* we stay consistent on which parts of the metaslab space map's
	* histogram are currently not available for allocations (e.g because
	* they are in the defer, freed, and freeing trees).
	*/
	static void
	metaslab_aux_histograms_update(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	ASSERT(sm != NULL);

	/*
	* This is similar to the metaslab's space map histogram updates
	* that take place in metaslab_sync(). The only difference is that
	* we only care about segments that haven't made it into the
	* ms_allocatable tree yet.
	*/
	if (msp->ms_loaded) {
	metaslab_aux_histograms_clear(msp);

	metaslab_aux_histogram_add(msp->ms_synchist,
	sm->sm_shift, msp->ms_freed);

	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	metaslab_aux_histogram_add(msp->ms_deferhist[t],
	sm->sm_shift, msp->ms_defer[t]);
	}
	}

	metaslab_aux_histogram_add(msp->ms_synchist,
	sm->sm_shift, msp->ms_freeing);
	}

	/*
	* Called every time we are done syncing (writing to) the metaslab,
	* i.e. at the end of each sync pass.
	* [see the comment in metaslab_impl.h for ms_synchist, ms_deferhist]
	*/
	static void
	metaslab_aux_histograms_update_done(metaslab_t *msp, boolean_t defer_allowed)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	space_map_t *sm = msp->ms_sm;

	if (sm == NULL) {
	/*
	* We came here from metaslab_init() when creating/opening a
	* pool, looking at a metaslab that hasn't had any allocations
	* yet.
	*/
	return;
	}

	/*
	* This is similar to the actions that we take for the ms_freed
	* and ms_defer trees in metaslab_sync_done().
	*/
	uint64_t hist_index = spa_syncing_txg(spa) % TXG_DEFER_SIZE;
	if (defer_allowed) {
	memcpy(msp->ms_deferhist[hist_index], msp->ms_synchist,
	sizeof (msp->ms_synchist));
	} else {
	memset(msp->ms_deferhist[hist_index], 0,
	sizeof (msp->ms_deferhist[hist_index]));
	}
	memset(msp->ms_synchist, 0, sizeof (msp->ms_synchist));
	}

	/*
	* Ensure that the metaslab's weight and fragmentation are consistent
	* with the contents of the histogram (either the range tree's histogram
	* or the space map's depending whether the metaslab is loaded).
	*/
	static void
	metaslab_verify_weight_and_frag(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	/*
	* We can end up here from vdev_remove_complete(), in which case we
	* cannot do these assertions because we hold spa config locks and
	* thus we are not allowed to read from the DMU.
	*
	* We check if the metaslab group has been removed and if that's
	* the case we return immediately as that would mean that we are
	* here from the aforementioned code path.
	*/
	if (msp->ms_group == NULL)
	return;

	/*
	* Devices being removed always return a weight of 0 and leave
	* fragmentation and ms_max_size as is - there is nothing for
	* us to verify here.
	*/
	vdev_t *vd = msp->ms_group->mg_vd;
	if (vd->vdev_removing)
	return;

	/*
	* If the metaslab is dirty it probably means that we've done
	* some allocations or frees that have changed our histograms
	* and thus the weight.
	*/
	for (int t = 0; t < TXG_SIZE; t++) {
	if (txg_list_member(&vd->vdev_ms_list, msp, t))
	return;
	}

	/*
	* This verification checks that our in-memory state is consistent
	* with what's on disk. If the pool is read-only then there aren't
	* any changes and we just have the initially-loaded state.
	*/
	if (!spa_writeable(msp->ms_group->mg_vd->vdev_spa))
	return;

	/* some extra verification for in-core tree if you can */
	if (msp->ms_loaded) {
	range_tree_stat_verify(msp->ms_allocatable);
	VERIFY(space_map_histogram_verify(msp->ms_sm,
	msp->ms_allocatable));
	}

	uint64_t weight = msp->ms_weight;
	uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
	boolean_t space_based = WEIGHT_IS_SPACEBASED(msp->ms_weight);
	uint64_t frag = msp->ms_fragmentation;
	uint64_t max_segsize = msp->ms_max_size;

	msp->ms_weight = 0;
	msp->ms_fragmentation = 0;

	/*
	* This function is used for verification purposes and thus should
	* not introduce any side-effects/mutations on the system's state.
	*
	* Regardless of whether metaslab_weight() thinks this metaslab
	* should be active or not, we want to ensure that the actual weight
	* (and therefore the value of ms_weight) would be the same if it
	* was to be recalculated at this point.
	*
	* In addition we set the nodirty flag so metaslab_weight() does
	* not dirty the metaslab for future TXGs (e.g. when trying to
	* force condensing to upgrade the metaslab spacemaps).
	*/
	msp->ms_weight = metaslab_weight(msp, B_TRUE) \| was_active;

	VERIFY3U(max_segsize, ==, msp->ms_max_size);

	/*
	* If the weight type changed then there is no point in doing
	* verification. Revert fields to their original values.
	*/
	if ((space_based && !WEIGHT_IS_SPACEBASED(msp->ms_weight)) \|\|
	(!space_based && WEIGHT_IS_SPACEBASED(msp->ms_weight))) {
	msp->ms_fragmentation = frag;
	msp->ms_weight = weight;
	return;
	}

	VERIFY3U(msp->ms_fragmentation, ==, frag);
	VERIFY3U(msp->ms_weight, ==, weight);
	}

	/*
	* If we're over the zfs_metaslab_mem_limit, select the loaded metaslab from
	* this class that was used longest ago, and attempt to unload it. We don't
	* want to spend too much time in this loop to prevent performance
	* degradation, and we expect that most of the time this operation will
	* succeed. Between that and the normal unloading processing during txg sync,
	* we expect this to keep the metaslab memory usage under control.
	*/
	static void
	metaslab_potentially_evict(metaslab_class_t *mc)
	{
	#ifdef _KERNEL
	uint64_t allmem = arc_all_memory();
	uint64_t inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	uint64_t size = spl_kmem_cache_entry_size(zfs_btree_leaf_cache);
	- int tries = 0;
	+ uint_t tries = 0;
	for (; allmem * zfs_metaslab_mem_limit / 100 < inuse * size &&
	tries < multilist_get_num_sublists(&mc->mc_metaslab_txg_list) * 2;
	tries++) {
	unsigned int idx = multilist_get_random_index(
	&mc->mc_metaslab_txg_list);
	multilist_sublist_t *mls =
	multilist_sublist_lock(&mc->mc_metaslab_txg_list, idx);
	metaslab_t *msp = multilist_sublist_head(mls);
	multilist_sublist_unlock(mls);
	while (msp != NULL && allmem * zfs_metaslab_mem_limit / 100 <
	inuse * size) {
	VERIFY3P(mls, ==, multilist_sublist_lock(
	&mc->mc_metaslab_txg_list, idx));
	ASSERT3U(idx, ==,
	metaslab_idx_func(&mc->mc_metaslab_txg_list, msp));

	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	multilist_sublist_unlock(mls);
	break;
	}
	metaslab_t *next_msp = multilist_sublist_next(mls, msp);
	multilist_sublist_unlock(mls);
	/*
	* If the metaslab is currently loading there are two
	* cases. If it's the metaslab we're evicting, we
	* can't continue on or we'll panic when we attempt to
	* recursively lock the mutex. If it's another
	* metaslab that's loading, it can be safely skipped,
	* since we know it's very new and therefore not a
	* good eviction candidate. We check later once the
	* lock is held that the metaslab is fully loaded
	* before actually unloading it.
	*/
	if (msp->ms_loading) {
	msp = next_msp;
	inuse =
	spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	continue;
	}
	/*
	* We can't unload metaslabs with no spacemap because
	* they're not ready to be unloaded yet. We can't
	* unload metaslabs with outstanding allocations
	* because doing so could cause the metaslab's weight
	* to decrease while it's unloaded, which violates an
	* invariant that we use to prevent unnecessary
	* loading. We also don't unload metaslabs that are
	* currently active because they are high-weight
	* metaslabs that are likely to be used in the near
	* future.
	*/
	mutex_enter(&msp->ms_lock);
	if (msp->ms_allocator == -1 && msp->ms_sm != NULL &&
	msp->ms_allocating_total == 0) {
	metaslab_unload(msp);
	}
	mutex_exit(&msp->ms_lock);
	msp = next_msp;
	inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	}
	}
	#else
	(void) mc, (void) zfs_metaslab_mem_limit;
	#endif
	}

	static int
	metaslab_load_impl(metaslab_t *msp)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loading);
	ASSERT(!msp->ms_condensing);

	/*
	* We temporarily drop the lock to unblock other operations while we
	* are reading the space map. Therefore, metaslab_sync() and
	* metaslab_sync_done() can run at the same time as we do.
	*
	* If we are using the log space maps, metaslab_sync() can't write to
	* the metaslab's space map while we are loading as we only write to
	* it when we are flushing the metaslab, and that can't happen while
	* we are loading it.
	*
	* If we are not using log space maps though, metaslab_sync() can
	* append to the space map while we are loading. Therefore we load
	* only entries that existed when we started the load. Additionally,
	* metaslab_sync_done() has to wait for the load to complete because
	* there are potential races like metaslab_load() loading parts of the
	* space map that are currently being appended by metaslab_sync(). If
	* we didn't, the ms_allocatable would have entries that
	* metaslab_sync_done() would try to re-add later.
	*
	* That's why before dropping the lock we remember the synced length
	* of the metaslab and read up to that point of the space map,
	* ignoring entries appended by metaslab_sync() that happen after we
	* drop the lock.
	*/
	uint64_t length = msp->ms_synced_length;
	mutex_exit(&msp->ms_lock);

	hrtime_t load_start = gethrtime();
	metaslab_rt_arg_t *mrap;
	if (msp->ms_allocatable->rt_arg == NULL) {
	mrap = kmem_zalloc(sizeof (*mrap), KM_SLEEP);
	} else {
	mrap = msp->ms_allocatable->rt_arg;
	msp->ms_allocatable->rt_ops = NULL;
	msp->ms_allocatable->rt_arg = NULL;
	}
	mrap->mra_bt = &msp->ms_allocatable_by_size;
	mrap->mra_floor_shift = metaslab_by_size_min_shift;

	if (msp->ms_sm != NULL) {
	error = space_map_load_length(msp->ms_sm, msp->ms_allocatable,
	SM_FREE, length);

	/* Now, populate the size-sorted tree. */
	metaslab_rt_create(msp->ms_allocatable, mrap);
	msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
	msp->ms_allocatable->rt_arg = mrap;

	struct mssa_arg arg = {0};
	arg.rt = msp->ms_allocatable;
	arg.mra = mrap;
	range_tree_walk(msp->ms_allocatable, metaslab_size_sorted_add,
	&arg);
	} else {
	/*
	* Add the size-sorted tree first, since we don't need to load
	* the metaslab from the spacemap.
	*/
	metaslab_rt_create(msp->ms_allocatable, mrap);
	msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
	msp->ms_allocatable->rt_arg = mrap;
	/*
	* The space map has not been allocated yet, so treat
	* all the space in the metaslab as free and add it to the
	* ms_allocatable tree.
	*/
	range_tree_add(msp->ms_allocatable,
	msp->ms_start, msp->ms_size);

	if (msp->ms_new) {
	/*
	* If the ms_sm doesn't exist, this means that this
	* metaslab hasn't gone through metaslab_sync() and
	* thus has never been dirtied. So we shouldn't
	* expect any unflushed allocs or frees from previous
	* TXGs.
	*/
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
	}
	}

	/*
	* We need to grab the ms_sync_lock to prevent metaslab_sync() from
	* changing the ms_sm (or log_sm) and the metaslab's range trees
	* while we are about to use them and populate the ms_allocatable.
	* The ms_lock is insufficient for this because metaslab_sync() doesn't
	* hold the ms_lock while writing the ms_checkpointing tree to disk.
	*/
	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	ASSERT(!msp->ms_condensing);
	ASSERT(!msp->ms_flushing);

	if (error != 0) {
	mutex_exit(&msp->ms_sync_lock);
	return (error);
	}

	ASSERT3P(msp->ms_group, !=, NULL);
	msp->ms_loaded = B_TRUE;

	/*
	* Apply all the unflushed changes to ms_allocatable right
	* away so any manipulations we do below have a clear view
	* of what is allocated and what is free.
	*/
	range_tree_walk(msp->ms_unflushed_allocs,
	range_tree_remove, msp->ms_allocatable);
	range_tree_walk(msp->ms_unflushed_frees,
	range_tree_add, msp->ms_allocatable);

	ASSERT3P(msp->ms_group, !=, NULL);
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	if (spa_syncing_log_sm(spa) != NULL) {
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LOG_SPACEMAP));

	/*
	* If we use a log space map we add all the segments
	* that are in ms_unflushed_frees so they are available
	* for allocation.
	*
	* ms_allocatable needs to contain all free segments
	* that are ready for allocations (thus not segments
	* from ms_freeing, ms_freed, and the ms_defer trees).
	* But if we grab the lock in this code path at a sync
	* pass later that 1, then it also contains the
	* segments of ms_freed (they were added to it earlier
	* in this path through ms_unflushed_frees). So we
	* need to remove all the segments that exist in
	* ms_freed from ms_allocatable as they will be added
	* later in metaslab_sync_done().
	*
	* When there's no log space map, the ms_allocatable
	* correctly doesn't contain any segments that exist
	* in ms_freed [see ms_synced_length].
	*/
	range_tree_walk(msp->ms_freed,
	range_tree_remove, msp->ms_allocatable);
	}

	/*
	* If we are not using the log space map, ms_allocatable
	* contains the segments that exist in the ms_defer trees
	* [see ms_synced_length]. Thus we need to remove them
	* from ms_allocatable as they will be added again in
	* metaslab_sync_done().
	*
	* If we are using the log space map, ms_allocatable still
	* contains the segments that exist in the ms_defer trees.
	* Not because it read them through the ms_sm though. But
	* because these segments are part of ms_unflushed_frees
	* whose segments we add to ms_allocatable earlier in this
	* code path.
	*/
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_walk(msp->ms_defer[t],
	range_tree_remove, msp->ms_allocatable);
	}

	/*
	* Call metaslab_recalculate_weight_and_sort() now that the
	* metaslab is loaded so we get the metaslab's real weight.
	*
	* Unless this metaslab was created with older software and
	* has not yet been converted to use segment-based weight, we
	* expect the new weight to be better or equal to the weight
	* that the metaslab had while it was not loaded. This is
	* because the old weight does not take into account the
	* consolidation of adjacent segments between TXGs. [see
	* comment for ms_synchist and ms_deferhist[] for more info]
	*/
	uint64_t weight = msp->ms_weight;
	uint64_t max_size = msp->ms_max_size;
	metaslab_recalculate_weight_and_sort(msp);
	if (!WEIGHT_IS_SPACEBASED(weight))
	ASSERT3U(weight, <=, msp->ms_weight);
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	ASSERT3U(max_size, <=, msp->ms_max_size);
	hrtime_t load_end = gethrtime();
	msp->ms_load_time = load_end;
	zfs_dbgmsg("metaslab_load: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, smp_length %llu, "
	"unflushed_allocs %llu, unflushed_frees %llu, "
	"freed %llu, defer %llu + %llu, unloaded time %llu ms, "
	"loading_time %lld ms, ms_max_size %llu, "
	"max size error %lld, "
	"old_weight %llx, new_weight %llx",
	(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)space_map_length(msp->ms_sm),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
	(u_longlong_t)range_tree_space(msp->ms_freed),
	(u_longlong_t)range_tree_space(msp->ms_defer[0]),
	(u_longlong_t)range_tree_space(msp->ms_defer[1]),
	(longlong_t)((load_start - msp->ms_unload_time) / 1000000),
	(longlong_t)((load_end - load_start) / 1000000),
	(u_longlong_t)msp->ms_max_size,
	(u_longlong_t)msp->ms_max_size - max_size,
	(u_longlong_t)weight, (u_longlong_t)msp->ms_weight);

	metaslab_verify_space(msp, spa_syncing_txg(spa));
	mutex_exit(&msp->ms_sync_lock);
	return (0);
	}

	int
	metaslab_load(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* There may be another thread loading the same metaslab, if that's
	* the case just wait until the other thread is done and return.
	*/
	metaslab_load_wait(msp);
	if (msp->ms_loaded)
	return (0);
	VERIFY(!msp->ms_loading);
	ASSERT(!msp->ms_condensing);

	/*
	* We set the loading flag BEFORE potentially dropping the lock to
	* wait for an ongoing flush (see ms_flushing below). This way other
	* threads know that there is already a thread that is loading this
	* metaslab.
	*/
	msp->ms_loading = B_TRUE;

	/*
	* Wait for any in-progress flushing to finish as we drop the ms_lock
	* both here (during space_map_load()) and in metaslab_flush() (when
	* we flush our changes to the ms_sm).
	*/
	if (msp->ms_flushing)
	metaslab_flush_wait(msp);

	/*
	* In the possibility that we were waiting for the metaslab to be
	* flushed (where we temporarily dropped the ms_lock), ensure that
	* no one else loaded the metaslab somehow.
	*/
	ASSERT(!msp->ms_loaded);

	/*
	* If we're loading a metaslab in the normal class, consider evicting
	* another one to keep our memory usage under the limit defined by the
	* zfs_metaslab_mem_limit tunable.
	*/
	if (spa_normal_class(msp->ms_group->mg_class->mc_spa) ==
	msp->ms_group->mg_class) {
	metaslab_potentially_evict(msp->ms_group->mg_class);
	}

	int error = metaslab_load_impl(msp);

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	msp->ms_loading = B_FALSE;
	cv_broadcast(&msp->ms_load_cv);

	return (error);
	}

	void
	metaslab_unload(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* This can happen if a metaslab is selected for eviction (in
	* metaslab_potentially_evict) and then unloaded during spa_sync (via
	* metaslab_class_evict_old).
	*/
	if (!msp->ms_loaded)
	return;

	range_tree_vacate(msp->ms_allocatable, NULL, NULL);
	msp->ms_loaded = B_FALSE;
	msp->ms_unload_time = gethrtime();

	msp->ms_activation_weight = 0;
	msp->ms_weight &= ~METASLAB_ACTIVE_MASK;

	if (msp->ms_group != NULL) {
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	multilist_sublist_unlock(mls);

	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	zfs_dbgmsg("metaslab_unload: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, weight %llx, "
	"selected txg %llu (%llu ms ago), alloc_txg %llu, "
	"loaded %llu ms ago, max_size %llu",
	(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)msp->ms_weight,
	(u_longlong_t)msp->ms_selected_txg,
	(u_longlong_t)(msp->ms_unload_time -
	msp->ms_selected_time) / 1000 / 1000,
	(u_longlong_t)msp->ms_alloc_txg,
	(u_longlong_t)(msp->ms_unload_time -
	msp->ms_load_time) / 1000 / 1000,
	(u_longlong_t)msp->ms_max_size);
	}

	/*
	* We explicitly recalculate the metaslab's weight based on its space
	* map (as it is now not loaded). We want unload metaslabs to always
	* have their weights calculated from the space map histograms, while
	* loaded ones have it calculated from their in-core range tree
	* [see metaslab_load()]. This way, the weight reflects the information
	* available in-core, whether it is loaded or not.
	*
	* If ms_group == NULL means that we came here from metaslab_fini(),
	* at which point it doesn't make sense for us to do the recalculation
	* and the sorting.
	*/
	if (msp->ms_group != NULL)
	metaslab_recalculate_weight_and_sort(msp);
	}

	/*
	* We want to optimize the memory use of the per-metaslab range
	* trees. To do this, we store the segments in the range trees in
	* units of sectors, zero-indexing from the start of the metaslab. If
	* the vdev_ms_shift - the vdev_ashift is less than 32, we can store
	* the ranges using two uint32_ts, rather than two uint64_ts.
	*/
	range_seg_type_t
	metaslab_calculate_range_tree_type(vdev_t vdev, metaslab_t msp,
	uint64_t start, uint64_t shift)
	{
	if (vdev->vdev_ms_shift - vdev->vdev_ashift < 32 &&
	!zfs_metaslab_force_large_segs) {
	*shift = vdev->vdev_ashift;
	*start = msp->ms_start;
	return (RANGE_SEG32);
	} else {
	*shift = 0;
	*start = 0;
	return (RANGE_SEG64);
	}
	}

	void
	metaslab_set_selected_txg(metaslab_t *msp, uint64_t txg)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	msp->ms_selected_txg = txg;
	msp->ms_selected_time = gethrtime();
	multilist_sublist_insert_tail(mls, msp);
	multilist_sublist_unlock(mls);
	}

	void
	metaslab_space_update(vdev_t vd, metaslab_class_t mc, int64_t alloc_delta,
	int64_t defer_delta, int64_t space_delta)
	{
	vdev_space_update(vd, alloc_delta, defer_delta, space_delta);

	ASSERT3P(vd->vdev_spa->spa_root_vdev, ==, vd->vdev_parent);
	ASSERT(vd->vdev_ms_count != 0);

	metaslab_class_space_update(mc, alloc_delta, defer_delta, space_delta,
	vdev_deflated_space(vd, space_delta));
	}

	int
	metaslab_init(metaslab_group_t *mg, uint64_t id, uint64_t object,
	uint64_t txg, metaslab_t **msp)
	{
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	metaslab_t *ms;
	int error;

	ms = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
	mutex_init(&ms->ms_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ms->ms_sync_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&ms->ms_load_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&ms->ms_flush_cv, NULL, CV_DEFAULT, NULL);
	multilist_link_init(&ms->ms_class_txg_node);

	ms->ms_id = id;
	ms->ms_start = id << vd->vdev_ms_shift;
	ms->ms_size = 1ULL << vd->vdev_ms_shift;
	ms->ms_allocator = -1;
	ms->ms_new = B_TRUE;

	vdev_ops_t *ops = vd->vdev_ops;
	if (ops->vdev_op_metaslab_init != NULL)
	ops->vdev_op_metaslab_init(vd, &ms->ms_start, &ms->ms_size);

	/*
	* We only open space map objects that already exist. All others
	* will be opened when we finally allocate an object for it. For
	* readonly pools there is no need to open the space map object.
	*
	* Note:
	* When called from vdev_expand(), we can't call into the DMU as
	* we are holding the spa_config_lock as a writer and we would
	* deadlock [see relevant comment in vdev_metaslab_init()]. in
	* that case, the object parameter is zero though, so we won't
	* call into the DMU.
	*/
	if (object != 0 && !(spa->spa_mode == SPA_MODE_READ &&
	!spa->spa_read_spacemaps)) {
	error = space_map_open(&ms->ms_sm, mos, object, ms->ms_start,
	ms->ms_size, vd->vdev_ashift);

	if (error != 0) {
	kmem_free(ms, sizeof (metaslab_t));
	return (error);
	}

	ASSERT(ms->ms_sm != NULL);
	ms->ms_allocated_space = space_map_allocated(ms->ms_sm);
	}

	uint64_t shift, start;
	range_seg_type_t type =
	metaslab_calculate_range_tree_type(vd, ms, &start, &shift);

	ms->ms_allocatable = range_tree_create(NULL, type, NULL, start, shift);
	for (int t = 0; t < TXG_SIZE; t++) {
	ms->ms_allocating[t] = range_tree_create(NULL, type,
	NULL, start, shift);
	}
	ms->ms_freeing = range_tree_create(NULL, type, NULL, start, shift);
	ms->ms_freed = range_tree_create(NULL, type, NULL, start, shift);
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	ms->ms_defer[t] = range_tree_create(NULL, type, NULL,
	start, shift);
	}
	ms->ms_checkpointing =
	range_tree_create(NULL, type, NULL, start, shift);
	ms->ms_unflushed_allocs =
	range_tree_create(NULL, type, NULL, start, shift);

	metaslab_rt_arg_t mrap = kmem_zalloc(sizeof (mrap), KM_SLEEP);
	mrap->mra_bt = &ms->ms_unflushed_frees_by_size;
	mrap->mra_floor_shift = metaslab_by_size_min_shift;
	ms->ms_unflushed_frees = range_tree_create(&metaslab_rt_ops,
	type, mrap, start, shift);

	ms->ms_trim = range_tree_create(NULL, type, NULL, start, shift);

	metaslab_group_add(mg, ms);
	metaslab_set_fragmentation(ms, B_FALSE);

	/*
	* If we're opening an existing pool (txg == 0) or creating
	* a new one (txg == TXG_INITIAL), all space is available now.
	* If we're adding space to an existing pool, the new space
	* does not become available until after this txg has synced.
	* The metaslab's weight will also be initialized when we sync
	* out this txg. This ensures that we don't attempt to allocate
	* from it before we have initialized it completely.
	*/
	if (txg <= TXG_INITIAL) {
	metaslab_sync_done(ms, 0);
	metaslab_space_update(vd, mg->mg_class,
	metaslab_allocated_space(ms), 0, 0);
	}

	if (txg != 0) {
	vdev_dirty(vd, 0, NULL, txg);
	vdev_dirty(vd, VDD_METASLAB, ms, txg);
	}

	*msp = ms;

	return (0);
	}

	static void
	metaslab_fini_flush_data(metaslab_t *msp)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	if (metaslab_unflushed_txg(msp) == 0) {
	ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL),
	==, NULL);
	return;
	}
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	mutex_enter(&spa->spa_flushed_ms_lock);
	avl_remove(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	spa_log_sm_decrement_mscount(spa, metaslab_unflushed_txg(msp));
	spa_log_summary_decrement_mscount(spa, metaslab_unflushed_txg(msp),
	metaslab_unflushed_dirty(msp));
	}

	uint64_t
	metaslab_unflushed_changes_memused(metaslab_t *ms)
	{
	return ((range_tree_numsegs(ms->ms_unflushed_allocs) +
	range_tree_numsegs(ms->ms_unflushed_frees)) *
	ms->ms_unflushed_allocs->rt_root.bt_elem_size);
	}

	void
	metaslab_fini(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;

	metaslab_fini_flush_data(msp);

	metaslab_group_remove(mg, msp);

	mutex_enter(&msp->ms_lock);
	VERIFY(msp->ms_group == NULL);

	/*
	* If this metaslab hasn't been through metaslab_sync_done() yet its
	* space hasn't been accounted for in its vdev and doesn't need to be
	* subtracted.
	*/
	if (!msp->ms_new) {
	metaslab_space_update(vd, mg->mg_class,
	-metaslab_allocated_space(msp), 0, -msp->ms_size);

	}
	space_map_close(msp->ms_sm);
	msp->ms_sm = NULL;

	metaslab_unload(msp);

	range_tree_destroy(msp->ms_allocatable);
	range_tree_destroy(msp->ms_freeing);
	range_tree_destroy(msp->ms_freed);

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_destroy(msp->ms_unflushed_allocs);
	range_tree_destroy(msp->ms_checkpointing);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
	range_tree_destroy(msp->ms_unflushed_frees);

	for (int t = 0; t < TXG_SIZE; t++) {
	range_tree_destroy(msp->ms_allocating[t]);
	}
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_destroy(msp->ms_defer[t]);
	}
	ASSERT0(msp->ms_deferspace);

	for (int t = 0; t < TXG_SIZE; t++)
	ASSERT(!txg_list_member(&vd->vdev_ms_list, msp, t));

	range_tree_vacate(msp->ms_trim, NULL, NULL);
	range_tree_destroy(msp->ms_trim);

	mutex_exit(&msp->ms_lock);
	cv_destroy(&msp->ms_load_cv);
	cv_destroy(&msp->ms_flush_cv);
	mutex_destroy(&msp->ms_lock);
	mutex_destroy(&msp->ms_sync_lock);
	ASSERT3U(msp->ms_allocator, ==, -1);

	kmem_free(msp, sizeof (metaslab_t));
	}

	#define FRAGMENTATION_TABLE_SIZE 17

	/*
	* This table defines a segment size based fragmentation metric that will
	* allow each metaslab to derive its own fragmentation value. This is done
	* by calculating the space in each bucket of the spacemap histogram and
	* multiplying that by the fragmentation metric in this table. Doing
	* this for all buckets and dividing it by the total amount of free
	* space in this metaslab (i.e. the total free space in all buckets) gives
	* us the fragmentation metric. This means that a high fragmentation metric
	* equates to most of the free space being comprised of small segments.
	* Conversely, if the metric is low, then most of the free space is in
	* large segments. A 10% change in fragmentation equates to approximately
	* double the number of segments.
	*
	* This table defines 0% fragmented space using 16MB segments. Testing has
	* shown that segments that are greater than or equal to 16MB do not suffer
	* from drastic performance problems. Using this value, we derive the rest
	* of the table. Since the fragmentation value is never stored on disk, it
	* is possible to change these calculations in the future.
	*/
	static const int zfs_frag_table[FRAGMENTATION_TABLE_SIZE] = {
	100, /* 512B */
	100, /* 1K */
	98, /* 2K */
	95, /* 4K */
	90, /* 8K */
	80, /* 16K */
	70, /* 32K */
	60, /* 64K */
	50, /* 128K */
	40, /* 256K */
	30, /* 512K */
	20, /* 1M */
	15, /* 2M */
	10, /* 4M */
	5, /* 8M */
	0 /* 16M */
	};

	/*
	* Calculate the metaslab's fragmentation metric and set ms_fragmentation.
	* Setting this value to ZFS_FRAG_INVALID means that the metaslab has not
	* been upgraded and does not support this metric. Otherwise, the return
	* value should be in the range [0, 100].
	*/
	static void
	metaslab_set_fragmentation(metaslab_t *msp, boolean_t nodirty)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	uint64_t fragmentation = 0;
	uint64_t total = 0;
	boolean_t feature_enabled = spa_feature_is_enabled(spa,
	SPA_FEATURE_SPACEMAP_HISTOGRAM);

	if (!feature_enabled) {
	msp->ms_fragmentation = ZFS_FRAG_INVALID;
	return;
	}

	/*
	* A null space map means that the entire metaslab is free
	* and thus is not fragmented.
	*/
	if (msp->ms_sm == NULL) {
	msp->ms_fragmentation = 0;
	return;
	}

	/*
	* If this metaslab's space map has not been upgraded, flag it
	* so that we upgrade next time we encounter it.
	*/
	if (msp->ms_sm->sm_dbuf->db_size != sizeof (space_map_phys_t)) {
	uint64_t txg = spa_syncing_txg(spa);
	vdev_t *vd = msp->ms_group->mg_vd;

	/*
	* If we've reached the final dirty txg, then we must
	* be shutting down the pool. We don't want to dirty
	* any data past this point so skip setting the condense
	* flag. We can retry this action the next time the pool
	* is imported. We also skip marking this metaslab for
	* condensing if the caller has explicitly set nodirty.
	*/
	if (!nodirty &&
	spa_writeable(spa) && txg < spa_final_dirty_txg(spa)) {
	msp->ms_condense_wanted = B_TRUE;
	vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
	zfs_dbgmsg("txg %llu, requesting force condense: "
	"ms_id %llu, vdev_id %llu", (u_longlong_t)txg,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)vd->vdev_id);
	}
	msp->ms_fragmentation = ZFS_FRAG_INVALID;
	return;
	}

	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	uint64_t space = 0;
	uint8_t shift = msp->ms_sm->sm_shift;

	int idx = MIN(shift - SPA_MINBLOCKSHIFT + i,
	FRAGMENTATION_TABLE_SIZE - 1);

	if (msp->ms_sm->sm_phys->smp_histogram[i] == 0)
	continue;

	space = msp->ms_sm->sm_phys->smp_histogram[i] << (i + shift);
	total += space;

	ASSERT3U(idx, <, FRAGMENTATION_TABLE_SIZE);
	fragmentation += space * zfs_frag_table[idx];
	}

	if (total > 0)
	fragmentation /= total;
	ASSERT3U(fragmentation, <=, 100);

	msp->ms_fragmentation = fragmentation;
	}

	/*
	* Compute a weight -- a selection preference value -- for the given metaslab.
	* This is based on the amount of free space, the level of fragmentation,
	* the LBA range, and whether the metaslab is loaded.
	*/
	static uint64_t
	metaslab_space_weight(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	uint64_t weight, space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The baseline weight is the metaslab's free space.
	*/
	space = msp->ms_size - metaslab_allocated_space(msp);

	if (metaslab_fragmentation_factor_enabled &&
	msp->ms_fragmentation != ZFS_FRAG_INVALID) {
	/*
	* Use the fragmentation information to inversely scale
	* down the baseline weight. We need to ensure that we
	* don't exclude this metaslab completely when it's 100%
	* fragmented. To avoid this we reduce the fragmented value
	* by 1.
	*/
	space = (space * (100 - (msp->ms_fragmentation - 1))) / 100;

	/*
	* If space < SPA_MINBLOCKSIZE, then we will not allocate from
	* this metaslab again. The fragmentation metric may have
	* decreased the space to something smaller than
	* SPA_MINBLOCKSIZE, so reset the space to SPA_MINBLOCKSIZE
	* so that we can consume any remaining space.
	*/
	if (space > 0 && space < SPA_MINBLOCKSIZE)
	space = SPA_MINBLOCKSIZE;
	}
	weight = space;

	/*
	* Modern disks have uniform bit density and constant angular velocity.
	* Therefore, the outer recording zones are faster (higher bandwidth)
	* than the inner zones by the ratio of outer to inner track diameter,
	* which is typically around 2:1. We account for this by assigning
	* higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
	* In effect, this means that we'll select the metaslab with the most
	* free bandwidth rather than simply the one with the most free space.
	*/
	if (!vd->vdev_nonrot && metaslab_lba_weighting_enabled) {
	weight = 2 * weight - (msp->ms_id * weight) / vd->vdev_ms_count;
	ASSERT(weight >= space && weight <= 2 * space);
	}

	/*
	* If this metaslab is one we're actively using, adjust its
	* weight to make it preferable to any inactive metaslab so
	* we'll polish it off. If the fragmentation on this metaslab
	* has exceed our threshold, then don't mark it active.
	*/
	if (msp->ms_loaded && msp->ms_fragmentation != ZFS_FRAG_INVALID &&
	msp->ms_fragmentation <= zfs_metaslab_fragmentation_threshold) {
	weight \|= (msp->ms_weight & METASLAB_ACTIVE_MASK);
	}

	WEIGHT_SET_SPACEBASED(weight);
	return (weight);
	}

	/*
	* Return the weight of the specified metaslab, according to the segment-based
	* weighting algorithm. The metaslab must be loaded. This function can
	* be called within a sync pass since it relies only on the metaslab's
	* range tree which is always accurate when the metaslab is loaded.
	*/
	static uint64_t
	metaslab_weight_from_range_tree(metaslab_t *msp)
	{
	uint64_t weight = 0;
	uint32_t segments = 0;

	ASSERT(msp->ms_loaded);

	for (int i = RANGE_TREE_HISTOGRAM_SIZE - 1; i >= SPA_MINBLOCKSHIFT;
	i--) {
	uint8_t shift = msp->ms_group->mg_vd->vdev_ashift;
	int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;

	segments <<= 1;
	segments += msp->ms_allocatable->rt_histogram[i];

	/*
	* The range tree provides more precision than the space map
	* and must be downgraded so that all values fit within the
	* space map's histogram. This allows us to compare loaded
	* vs. unloaded metaslabs to determine which metaslab is
	* considered "best".
	*/
	if (i > max_idx)
	continue;

	if (segments != 0) {
	WEIGHT_SET_COUNT(weight, segments);
	WEIGHT_SET_INDEX(weight, i);
	WEIGHT_SET_ACTIVE(weight, 0);
	break;
	}
	}
	return (weight);
	}

	/*
	* Calculate the weight based on the on-disk histogram. Should be applied
	* only to unloaded metaslabs (i.e no incoming allocations) in-order to
	* give results consistent with the on-disk state
	*/
	static uint64_t
	metaslab_weight_from_spacemap(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	ASSERT(!msp->ms_loaded);
	ASSERT(sm != NULL);
	ASSERT3U(space_map_object(sm), !=, 0);
	ASSERT3U(sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));

	/*
	* Create a joint histogram from all the segments that have made
	* it to the metaslab's space map histogram, that are not yet
	* available for allocation because they are still in the freeing
	* pipeline (e.g. freeing, freed, and defer trees). Then subtract
	* these segments from the space map's histogram to get a more
	* accurate weight.
	*/
	uint64_t deferspace_histogram[SPACE_MAP_HISTOGRAM_SIZE] = {0};
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++)
	deferspace_histogram[i] += msp->ms_synchist[i];
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	deferspace_histogram[i] += msp->ms_deferhist[t][i];
	}
	}

	uint64_t weight = 0;
	for (int i = SPACE_MAP_HISTOGRAM_SIZE - 1; i >= 0; i--) {
	ASSERT3U(sm->sm_phys->smp_histogram[i], >=,
	deferspace_histogram[i]);
	uint64_t count =
	sm->sm_phys->smp_histogram[i] - deferspace_histogram[i];
	if (count != 0) {
	WEIGHT_SET_COUNT(weight, count);
	WEIGHT_SET_INDEX(weight, i + sm->sm_shift);
	WEIGHT_SET_ACTIVE(weight, 0);
	break;
	}
	}
	return (weight);
	}

	/*
	* Compute a segment-based weight for the specified metaslab. The weight
	* is determined by highest bucket in the histogram. The information
	* for the highest bucket is encoded into the weight value.
	*/
	static uint64_t
	metaslab_segment_weight(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	uint64_t weight = 0;
	uint8_t shift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The metaslab is completely free.
	*/
	if (metaslab_allocated_space(msp) == 0) {
	int idx = highbit64(msp->ms_size) - 1;
	int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;

	if (idx < max_idx) {
	WEIGHT_SET_COUNT(weight, 1ULL);
	WEIGHT_SET_INDEX(weight, idx);
	} else {
	WEIGHT_SET_COUNT(weight, 1ULL << (idx - max_idx));
	WEIGHT_SET_INDEX(weight, max_idx);
	}
	WEIGHT_SET_ACTIVE(weight, 0);
	ASSERT(!WEIGHT_IS_SPACEBASED(weight));
	return (weight);
	}

	ASSERT3U(msp->ms_sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));

	/*
	* If the metaslab is fully allocated then just make the weight 0.
	*/
	if (metaslab_allocated_space(msp) == msp->ms_size)
	return (0);
	/*
	* If the metaslab is already loaded, then use the range tree to
	* determine the weight. Otherwise, we rely on the space map information
	* to generate the weight.
	*/
	if (msp->ms_loaded) {
	weight = metaslab_weight_from_range_tree(msp);
	} else {
	weight = metaslab_weight_from_spacemap(msp);
	}

	/*
	* If the metaslab was active the last time we calculated its weight
	* then keep it active. We want to consume the entire region that
	* is associated with this weight.
	*/
	if (msp->ms_activation_weight != 0 && weight != 0)
	WEIGHT_SET_ACTIVE(weight, WEIGHT_GET_ACTIVE(msp->ms_weight));
	return (weight);
	}

	/*
	* Determine if we should attempt to allocate from this metaslab. If the
	* metaslab is loaded, then we can determine if the desired allocation
	* can be satisfied by looking at the size of the maximum free segment
	* on that metaslab. Otherwise, we make our decision based on the metaslab's
	* weight. For segment-based weighting we can determine the maximum
	* allocation based on the index encoded in its value. For space-based
	* weights we rely on the entire weight (excluding the weight-type bit).
	*/
	static boolean_t
	metaslab_should_allocate(metaslab_t *msp, uint64_t asize, boolean_t try_hard)
	{
	/*
	* If the metaslab is loaded, ms_max_size is definitive and we can use
	* the fast check. If it's not, the ms_max_size is a lower bound (once
	* set), and we should use the fast check as long as we're not in
	* try_hard and it's been less than zfs_metaslab_max_size_cache_sec
	* seconds since the metaslab was unloaded.
	*/
	if (msp->ms_loaded \|\|
	(msp->ms_max_size != 0 && !try_hard && gethrtime() <
	msp->ms_unload_time + SEC2NSEC(zfs_metaslab_max_size_cache_sec)))
	return (msp->ms_max_size >= asize);

	boolean_t should_allocate;
	if (!WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
	/*
	* The metaslab segment weight indicates segments in the
	* range [2^i, 2^(i+1)), where i is the index in the weight.
	* Since the asize might be in the middle of the range, we
	* should attempt the allocation if asize < 2^(i+1).
	*/
	should_allocate = (asize <
	1ULL << (WEIGHT_GET_INDEX(msp->ms_weight) + 1));
	} else {
	should_allocate = (asize <=
	(msp->ms_weight & ~METASLAB_WEIGHT_TYPE));
	}

	return (should_allocate);
	}

	static uint64_t
	metaslab_weight(metaslab_t *msp, boolean_t nodirty)
	{
	vdev_t *vd = msp->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	uint64_t weight;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	metaslab_set_fragmentation(msp, nodirty);

	/*
	* Update the maximum size. If the metaslab is loaded, this will
	* ensure that we get an accurate maximum size if newly freed space
	* has been added back into the free tree. If the metaslab is
	* unloaded, we check if there's a larger free segment in the
	* unflushed frees. This is a lower bound on the largest allocatable
	* segment size. Coalescing of adjacent entries may reveal larger
	* allocatable segments, but we aren't aware of those until loading
	* the space map into a range tree.
	*/
	if (msp->ms_loaded) {
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	} else {
	msp->ms_max_size = MAX(msp->ms_max_size,
	metaslab_largest_unflushed_free(msp));
	}

	/*
	* Segment-based weighting requires space map histogram support.
	*/
	if (zfs_metaslab_segment_weight_enabled &&
	spa_feature_is_enabled(spa, SPA_FEATURE_SPACEMAP_HISTOGRAM) &&
	(msp->ms_sm == NULL \|\| msp->ms_sm->sm_dbuf->db_size ==
	sizeof (space_map_phys_t))) {
	weight = metaslab_segment_weight(msp);
	} else {
	weight = metaslab_space_weight(msp);
	}
	return (weight);
	}

	void
	metaslab_recalculate_weight_and_sort(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/* note: we preserve the mask (e.g. indication of primary, etc..) */
	uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
	metaslab_group_sort(msp->ms_group, msp,
	metaslab_weight(msp, B_FALSE) \| was_active);
	}

	static int
	metaslab_activate_allocator(metaslab_group_t mg, metaslab_t msp,
	int allocator, uint64_t activation_weight)
	{
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* If we're activating for the claim code, we don't want to actually
	* set the metaslab up for a specific allocator.
	*/
	if (activation_weight == METASLAB_WEIGHT_CLAIM) {
	ASSERT0(msp->ms_activation_weight);
	msp->ms_activation_weight = msp->ms_weight;
	metaslab_group_sort(mg, msp, msp->ms_weight \|
	activation_weight);
	return (0);
	}

	metaslab_t **mspp = (activation_weight == METASLAB_WEIGHT_PRIMARY ?
	&mga->mga_primary : &mga->mga_secondary);

	mutex_enter(&mg->mg_lock);
	if (*mspp != NULL) {
	mutex_exit(&mg->mg_lock);
	return (EEXIST);
	}

	*mspp = msp;
	ASSERT3S(msp->ms_allocator, ==, -1);
	msp->ms_allocator = allocator;
	msp->ms_primary = (activation_weight == METASLAB_WEIGHT_PRIMARY);

	ASSERT0(msp->ms_activation_weight);
	msp->ms_activation_weight = msp->ms_weight;
	metaslab_group_sort_impl(mg, msp,
	msp->ms_weight \| activation_weight);
	mutex_exit(&mg->mg_lock);

	return (0);
	}

	static int
	metaslab_activate(metaslab_t *msp, int allocator, uint64_t activation_weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The current metaslab is already activated for us so there
	* is nothing to do. Already activated though, doesn't mean
	* that this metaslab is activated for our allocator nor our
	* requested activation weight. The metaslab could have started
	* as an active one for our allocator but changed allocators
	* while we were waiting to grab its ms_lock or we stole it
	* [see find_valid_metaslab()]. This means that there is a
	* possibility of passivating a metaslab of another allocator
	* or from a different activation mask, from this thread.
	*/
	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
	ASSERT(msp->ms_loaded);
	return (0);
	}

	int error = metaslab_load(msp);
	if (error != 0) {
	metaslab_group_sort(msp->ms_group, msp, 0);
	return (error);
	}

	/*
	* When entering metaslab_load() we may have dropped the
	* ms_lock because we were loading this metaslab, or we
	* were waiting for another thread to load it for us. In
	* that scenario, we recheck the weight of the metaslab
	* to see if it was activated by another thread.
	*
	* If the metaslab was activated for another allocator or
	* it was activated with a different activation weight (e.g.
	* we wanted to make it a primary but it was activated as
	* secondary) we return error (EBUSY).
	*
	* If the metaslab was activated for the same allocator
	* and requested activation mask, skip activating it.
	*/
	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
	if (msp->ms_allocator != allocator)
	return (EBUSY);

	if ((msp->ms_weight & activation_weight) == 0)
	return (SET_ERROR(EBUSY));

	EQUIV((activation_weight == METASLAB_WEIGHT_PRIMARY),
	msp->ms_primary);
	return (0);
	}

	/*
	* If the metaslab has literally 0 space, it will have weight 0. In
	* that case, don't bother activating it. This can happen if the
	* metaslab had space during find_valid_metaslab, but another thread
	* loaded it and used all that space while we were waiting to grab the
	* lock.
	*/
	if (msp->ms_weight == 0) {
	ASSERT0(range_tree_space(msp->ms_allocatable));
	return (SET_ERROR(ENOSPC));
	}

	if ((error = metaslab_activate_allocator(msp->ms_group, msp,
	allocator, activation_weight)) != 0) {
	return (error);
	}

	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);

	return (0);
	}

	static void
	metaslab_passivate_allocator(metaslab_group_t mg, metaslab_t msp,
	uint64_t weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);

	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
	metaslab_group_sort(mg, msp, weight);
	return;
	}

	mutex_enter(&mg->mg_lock);
	ASSERT3P(msp->ms_group, ==, mg);
	ASSERT3S(0, <=, msp->ms_allocator);
	ASSERT3U(msp->ms_allocator, <, mg->mg_allocators);

	metaslab_group_allocator_t *mga = &mg->mg_allocator[msp->ms_allocator];
	if (msp->ms_primary) {
	ASSERT3P(mga->mga_primary, ==, msp);
	ASSERT(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	mga->mga_primary = NULL;
	} else {
	ASSERT3P(mga->mga_secondary, ==, msp);
	ASSERT(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	mga->mga_secondary = NULL;
	}
	msp->ms_allocator = -1;
	metaslab_group_sort_impl(mg, msp, weight);
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_passivate(metaslab_t *msp, uint64_t weight)
	{
	uint64_t size __maybe_unused = weight & ~METASLAB_WEIGHT_TYPE;

	/*
	* If size < SPA_MINBLOCKSIZE, then we will not allocate from
	* this metaslab again. In that case, it had better be empty,
	* or we would be leaving space on the table.
	*/
	ASSERT(!WEIGHT_IS_SPACEBASED(msp->ms_weight) \|\|
	size >= SPA_MINBLOCKSIZE \|\|
	range_tree_space(msp->ms_allocatable) == 0);
	ASSERT0(weight & METASLAB_ACTIVE_MASK);

	ASSERT(msp->ms_activation_weight != 0);
	msp->ms_activation_weight = 0;
	metaslab_passivate_allocator(msp->ms_group, msp, weight);
	ASSERT0(msp->ms_weight & METASLAB_ACTIVE_MASK);
	}

	/*
	* Segment-based metaslabs are activated once and remain active until
	* we either fail an allocation attempt (similar to space-based metaslabs)
	* or have exhausted the free space in zfs_metaslab_switch_threshold
	* buckets since the metaslab was activated. This function checks to see
	* if we've exhausted the zfs_metaslab_switch_threshold buckets in the
	* metaslab and passivates it proactively. This will allow us to select a
	* metaslab with a larger contiguous region, if any, remaining within this
	* metaslab group. If we're in sync pass > 1, then we continue using this
	* metaslab so that we don't dirty more block and cause more sync passes.
	*/
	static void
	metaslab_segment_may_passivate(metaslab_t *msp)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	if (WEIGHT_IS_SPACEBASED(msp->ms_weight) \|\| spa_sync_pass(spa) > 1)
	return;

	/*
	* Since we are in the middle of a sync pass, the most accurate
	* information that is accessible to us is the in-core range tree
	* histogram; calculate the new weight based on that information.
	*/
	uint64_t weight = metaslab_weight_from_range_tree(msp);
	int activation_idx = WEIGHT_GET_INDEX(msp->ms_activation_weight);
	int current_idx = WEIGHT_GET_INDEX(weight);

	if (current_idx <= activation_idx - zfs_metaslab_switch_threshold)
	metaslab_passivate(msp, weight);
	}

	static void
	metaslab_preload(void *arg)
	{
	metaslab_t *msp = arg;
	metaslab_class_t *mc = msp->ms_group->mg_class;
	spa_t *spa = mc->mc_spa;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	ASSERT(!MUTEX_HELD(&msp->ms_group->mg_lock));

	mutex_enter(&msp->ms_lock);
	(void) metaslab_load(msp);
	metaslab_set_selected_txg(msp, spa_syncing_txg(spa));
	mutex_exit(&msp->ms_lock);
	spl_fstrans_unmark(cookie);
	}

	static void
	metaslab_group_preload(metaslab_group_t *mg)
	{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_t *msp;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	int m = 0;

	if (spa_shutting_down(spa) \|\| !metaslab_preload_enabled) {
	taskq_wait_outstanding(mg->mg_taskq, 0);
	return;
	}

	mutex_enter(&mg->mg_lock);

	/*
	* Load the next potential metaslabs
	*/
	for (msp = avl_first(t); msp != NULL; msp = AVL_NEXT(t, msp)) {
	ASSERT3P(msp->ms_group, ==, mg);

	/*
	* We preload only the maximum number of metaslabs specified
	* by metaslab_preload_limit. If a metaslab is being forced
	* to condense then we preload it too. This will ensure
	* that force condensing happens in the next txg.
	*/
	if (++m > metaslab_preload_limit && !msp->ms_condense_wanted) {
	continue;
	}

	VERIFY(taskq_dispatch(mg->mg_taskq, metaslab_preload,
	msp, TQ_SLEEP) != TASKQID_INVALID);
	}
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Determine if the space map's on-disk footprint is past our tolerance for
	* inefficiency. We would like to use the following criteria to make our
	* decision:
	*
	* 1. Do not condense if the size of the space map object would dramatically
	* increase as a result of writing out the free space range tree.
	*
	* 2. Condense if the on on-disk space map representation is at least
	* zfs_condense_pct/100 times the size of the optimal representation
	* (i.e. zfs_condense_pct = 110 and in-core = 1MB, optimal = 1.1MB).
	*
	* 3. Do not condense if the on-disk size of the space map does not actually
	* decrease.
	*
	* Unfortunately, we cannot compute the on-disk size of the space map in this
	* context because we cannot accurately compute the effects of compression, etc.
	* Instead, we apply the heuristic described in the block comment for
	* zfs_metaslab_condense_block_threshold - we only condense if the space used
	* is greater than a threshold number of blocks.
	*/
	static boolean_t
	metaslab_should_condense(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	vdev_t *vd = msp->ms_group->mg_vd;
	- uint64_t vdev_blocksize = 1 << vd->vdev_ashift;
	+ uint64_t vdev_blocksize = 1ULL << vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);
	ASSERT(sm != NULL);
	ASSERT3U(spa_sync_pass(vd->vdev_spa), ==, 1);

	/*
	* We always condense metaslabs that are empty and metaslabs for
	* which a condense request has been made.
	*/
	if (range_tree_numsegs(msp->ms_allocatable) == 0 \|\|
	msp->ms_condense_wanted)
	return (B_TRUE);

	uint64_t record_size = MAX(sm->sm_blksz, vdev_blocksize);
	uint64_t object_size = space_map_length(sm);
	uint64_t optimal_size = space_map_estimate_optimal_size(sm,
	msp->ms_allocatable, SM_NO_VDEVID);

	return (object_size >= (optimal_size * zfs_condense_pct / 100) &&
	object_size > zfs_metaslab_condense_block_threshold * record_size);
	}

	/*
	* Condense the on-disk space map representation to its minimized form.
	* The minimized form consists of a small number of allocations followed
	* by the entries of the free range tree (ms_allocatable). The condensed
	* spacemap contains all the entries of previous TXGs (including those in
	* the pool-wide log spacemaps; thus this is effectively a superset of
	* metaslab_flush()), but this TXG's entries still need to be written.
	*/
	static void
	metaslab_condense(metaslab_t msp, dmu_tx_t tx)
	{
	range_tree_t *condense_tree;
	space_map_t *sm = msp->ms_sm;
	uint64_t txg = dmu_tx_get_txg(tx);
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_sm != NULL);

	/*
	* In order to condense the space map, we need to change it so it
	* only describes which segments are currently allocated and free.
	*
	* All the current free space resides in the ms_allocatable, all
	* the ms_defer trees, and all the ms_allocating trees. We ignore
	* ms_freed because it is empty because we're in sync pass 1. We
	* ignore ms_freeing because these changes are not yet reflected
	* in the spacemap (they will be written later this txg).
	*
	* So to truncate the space map to represent all the entries of
	* previous TXGs we do the following:
	*
	* 1] We create a range tree (condense tree) that is 100% empty.
	* 2] We add to it all segments found in the ms_defer trees
	* as those segments are marked as free in the original space
	* map. We do the same with the ms_allocating trees for the same
	* reason. Adding these segments should be a relatively
	* inexpensive operation since we expect these trees to have a
	* small number of nodes.
	* 3] We vacate any unflushed allocs, since they are not frees we
	* need to add to the condense tree. Then we vacate any
	* unflushed frees as they should already be part of ms_allocatable.
	* 4] At this point, we would ideally like to add all segments
	* in the ms_allocatable tree from the condense tree. This way
	* we would write all the entries of the condense tree as the
	* condensed space map, which would only contain freed
	* segments with everything else assumed to be allocated.
	*
	* Doing so can be prohibitively expensive as ms_allocatable can
	* be large, and therefore computationally expensive to add to
	* the condense_tree. Instead we first sync out an entry marking
	* everything as allocated, then the condense_tree and then the
	* ms_allocatable, in the condensed space map. While this is not
	* optimal, it is typically close to optimal and more importantly
	* much cheaper to compute.
	*
	* 5] Finally, as both of the unflushed trees were written to our
	* new and condensed metaslab space map, we basically flushed
	* all the unflushed changes to disk, thus we call
	* metaslab_flush_update().
	*/
	ASSERT3U(spa_sync_pass(spa), ==, 1);
	ASSERT(range_tree_is_empty(msp->ms_freed)); /* since it is pass 1 */

	zfs_dbgmsg("condensing: txg %llu, msp[%llu] %px, vdev id %llu, "
	"spa %s, smp size %llu, segments %llu, forcing condense=%s",
	(u_longlong_t)txg, (u_longlong_t)msp->ms_id, msp,
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	spa->spa_name, (u_longlong_t)space_map_length(msp->ms_sm),
	(u_longlong_t)range_tree_numsegs(msp->ms_allocatable),
	msp->ms_condense_wanted ? "TRUE" : "FALSE");

	msp->ms_condense_wanted = B_FALSE;

	range_seg_type_t type;
	uint64_t shift, start;
	type = metaslab_calculate_range_tree_type(msp->ms_group->mg_vd, msp,
	&start, &shift);

	condense_tree = range_tree_create(NULL, type, NULL, start, shift);

	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_walk(msp->ms_defer[t],
	range_tree_add, condense_tree);
	}

	for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
	range_tree_walk(msp->ms_allocating[(txg + t) & TXG_MASK],
	range_tree_add, condense_tree);
	}

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);

	/*
	* We're about to drop the metaslab's lock thus allowing other
	* consumers to change it's content. Set the metaslab's ms_condensing
	* flag to ensure that allocations on this metaslab do not occur
	* while we're in the middle of committing it to disk. This is only
	* critical for ms_allocatable as all other range trees use per TXG
	* views of their content.
	*/
	msp->ms_condensing = B_TRUE;

	mutex_exit(&msp->ms_lock);
	uint64_t object = space_map_object(msp->ms_sm);
	space_map_truncate(sm,
	spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
	zfs_metaslab_sm_blksz_with_log : zfs_metaslab_sm_blksz_no_log, tx);

	/*
	* space_map_truncate() may have reallocated the spacemap object.
	* If so, update the vdev_ms_array.
	*/
	if (space_map_object(msp->ms_sm) != object) {
	object = space_map_object(msp->ms_sm);
	dmu_write(spa->spa_meta_objset,
	msp->ms_group->mg_vd->vdev_ms_array, sizeof (uint64_t) *
	msp->ms_id, sizeof (uint64_t), &object, tx);
	}

	/*
	* Note:
	* When the log space map feature is enabled, each space map will
	* always have ALLOCS followed by FREES for each sync pass. This is
	* typically true even when the log space map feature is disabled,
	* except from the case where a metaslab goes through metaslab_sync()
	* and gets condensed. In that case the metaslab's space map will have
	* ALLOCS followed by FREES (due to condensing) followed by ALLOCS
	* followed by FREES (due to space_map_write() in metaslab_sync()) for
	* sync pass 1.
	*/
	range_tree_t *tmp_tree = range_tree_create(NULL, type, NULL, start,
	shift);
	range_tree_add(tmp_tree, msp->ms_start, msp->ms_size);
	space_map_write(sm, tmp_tree, SM_ALLOC, SM_NO_VDEVID, tx);
	space_map_write(sm, msp->ms_allocatable, SM_FREE, SM_NO_VDEVID, tx);
	space_map_write(sm, condense_tree, SM_FREE, SM_NO_VDEVID, tx);

	range_tree_vacate(condense_tree, NULL, NULL);
	range_tree_destroy(condense_tree);
	range_tree_vacate(tmp_tree, NULL, NULL);
	range_tree_destroy(tmp_tree);
	mutex_enter(&msp->ms_lock);

	msp->ms_condensing = B_FALSE;
	metaslab_flush_update(msp, tx);
	}

	static void
	metaslab_unflushed_add(metaslab_t msp, dmu_tx_t tx)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	ASSERT(spa_syncing_log_sm(spa) != NULL);
	ASSERT(msp->ms_sm != NULL);
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));

	mutex_enter(&spa->spa_flushed_ms_lock);
	metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
	metaslab_set_unflushed_dirty(msp, B_TRUE);
	avl_add(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	spa_log_sm_increment_current_mscount(spa);
	spa_log_summary_add_flushed_metaslab(spa, B_TRUE);
	}

	void
	metaslab_unflushed_bump(metaslab_t msp, dmu_tx_t tx, boolean_t dirty)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	ASSERT(spa_syncing_log_sm(spa) != NULL);
	ASSERT(msp->ms_sm != NULL);
	ASSERT(metaslab_unflushed_txg(msp) != 0);
	ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL), ==, msp);
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));

	VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(spa));

	/* update metaslab's position in our flushing tree */
	uint64_t ms_prev_flushed_txg = metaslab_unflushed_txg(msp);
	boolean_t ms_prev_flushed_dirty = metaslab_unflushed_dirty(msp);
	mutex_enter(&spa->spa_flushed_ms_lock);
	avl_remove(&spa->spa_metaslabs_by_flushed, msp);
	metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
	metaslab_set_unflushed_dirty(msp, dirty);
	avl_add(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	/* update metaslab counts of spa_log_sm_t nodes */
	spa_log_sm_decrement_mscount(spa, ms_prev_flushed_txg);
	spa_log_sm_increment_current_mscount(spa);

	/* update log space map summary */
	spa_log_summary_decrement_mscount(spa, ms_prev_flushed_txg,
	ms_prev_flushed_dirty);
	spa_log_summary_add_flushed_metaslab(spa, dirty);

	/* cleanup obsolete logs if any */
	spa_cleanup_old_sm_logs(spa, tx);
	}

	/*
	* Called when the metaslab has been flushed (its own spacemap now reflects
	* all the contents of the pool-wide spacemap log). Updates the metaslab's
	* metadata and any pool-wide related log space map data (e.g. summary,
	* obsolete logs, etc..) to reflect that.
	*/
	static void
	metaslab_flush_update(metaslab_t msp, dmu_tx_t tx)
	{
	metaslab_group_t *mg = msp->ms_group;
	spa_t *spa = mg->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	ASSERT3U(spa_sync_pass(spa), ==, 1);

	/*
	* Just because a metaslab got flushed, that doesn't mean that
	* it will pass through metaslab_sync_done(). Thus, make sure to
	* update ms_synced_length here in case it doesn't.
	*/
	msp->ms_synced_length = space_map_length(msp->ms_sm);

	/*
	* We may end up here from metaslab_condense() without the
	* feature being active. In that case this is a no-op.
	*/
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP) \|\|
	metaslab_unflushed_txg(msp) == 0)
	return;

	metaslab_unflushed_bump(msp, tx, B_FALSE);
	}

	boolean_t
	metaslab_flush(metaslab_t msp, dmu_tx_t tx)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(spa_sync_pass(spa), ==, 1);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	ASSERT(msp->ms_sm != NULL);
	ASSERT(metaslab_unflushed_txg(msp) != 0);
	ASSERT(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL) != NULL);

	/*
	* There is nothing wrong with flushing the same metaslab twice, as
	* this codepath should work on that case. However, the current
	* flushing scheme makes sure to avoid this situation as we would be
	* making all these calls without having anything meaningful to write
	* to disk. We assert this behavior here.
	*/
	ASSERT3U(metaslab_unflushed_txg(msp), <, dmu_tx_get_txg(tx));

	/*
	* We can not flush while loading, because then we would
	* not load the ms_unflushed_{allocs,frees}.
	*/
	if (msp->ms_loading)
	return (B_FALSE);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	/*
	* Metaslab condensing is effectively flushing. Therefore if the
	* metaslab can be condensed we can just condense it instead of
	* flushing it.
	*
	* Note that metaslab_condense() does call metaslab_flush_update()
	* so we can just return immediately after condensing. We also
	* don't need to care about setting ms_flushing or broadcasting
	* ms_flush_cv, even if we temporarily drop the ms_lock in
	* metaslab_condense(), as the metaslab is already loaded.
	*/
	if (msp->ms_loaded && metaslab_should_condense(msp)) {
	metaslab_group_t *mg = msp->ms_group;

	/*
	* For all histogram operations below refer to the
	* comments of metaslab_sync() where we follow a
	* similar procedure.
	*/
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);
	metaslab_group_histogram_remove(mg, msp);

	metaslab_condense(msp, tx);

	space_map_histogram_clear(msp->ms_sm);
	space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);
	ASSERT(range_tree_is_empty(msp->ms_freed));
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	space_map_histogram_add(msp->ms_sm,
	msp->ms_defer[t], tx);
	}
	metaslab_aux_histograms_update(msp);

	metaslab_group_histogram_add(mg, msp);
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));

	/*
	* Since we recreated the histogram (and potentially
	* the ms_sm too while condensing) ensure that the
	* weight is updated too because we are not guaranteed
	* that this metaslab is dirty and will go through
	* metaslab_sync_done().
	*/
	metaslab_recalculate_weight_and_sort(msp);
	return (B_TRUE);
	}

	msp->ms_flushing = B_TRUE;
	uint64_t sm_len_before = space_map_length(msp->ms_sm);

	mutex_exit(&msp->ms_lock);
	space_map_write(msp->ms_sm, msp->ms_unflushed_allocs, SM_ALLOC,
	SM_NO_VDEVID, tx);
	space_map_write(msp->ms_sm, msp->ms_unflushed_frees, SM_FREE,
	SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);

	uint64_t sm_len_after = space_map_length(msp->ms_sm);
	if (zfs_flags & ZFS_DEBUG_LOG_SPACEMAP) {
	zfs_dbgmsg("flushing: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, unflushed_allocs %llu, unflushed_frees %llu, "
	"appended %llu bytes", (u_longlong_t)dmu_tx_get_txg(tx),
	spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
	(u_longlong_t)(sm_len_after - sm_len_before));
	}

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	metaslab_flush_update(msp, tx);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	msp->ms_flushing = B_FALSE;
	cv_broadcast(&msp->ms_flush_cv);
	return (B_TRUE);
	}

	/*
	* Write a metaslab to disk in the context of the specified transaction group.
	*/
	void
	metaslab_sync(metaslab_t *msp, uint64_t txg)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa_meta_objset(spa);
	range_tree_t *alloctree = msp->ms_allocating[txg & TXG_MASK];
	dmu_tx_t *tx;

	ASSERT(!vd->vdev_ishole);

	/*
	* This metaslab has just been added so there's no work to do now.
	*/
	if (msp->ms_new) {
	ASSERT0(range_tree_space(alloctree));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_freed));
	ASSERT0(range_tree_space(msp->ms_checkpointing));
	ASSERT0(range_tree_space(msp->ms_trim));
	return;
	}

	/*
	* Normally, we don't want to process a metaslab if there are no
	* allocations or frees to perform. However, if the metaslab is being
	* forced to condense, it's loaded and we're not beyond the final
	* dirty txg, we need to let it through. Not condensing beyond the
	* final dirty txg prevents an issue where metaslabs that need to be
	* condensed but were loaded for other reasons could cause a panic
	* here. By only checking the txg in that branch of the conditional,
	* we preserve the utility of the VERIFY statements in all other
	* cases.
	*/
	if (range_tree_is_empty(alloctree) &&
	range_tree_is_empty(msp->ms_freeing) &&
	range_tree_is_empty(msp->ms_checkpointing) &&
	!(msp->ms_loaded && msp->ms_condense_wanted &&
	txg <= spa_final_dirty_txg(spa)))
	return;


	VERIFY3U(txg, <=, spa_final_dirty_txg(spa));

	/*
	* The only state that can actually be changing concurrently
	* with metaslab_sync() is the metaslab's ms_allocatable. No
	* other thread can be modifying this txg's alloc, freeing,
	* freed, or space_map_phys_t. We drop ms_lock whenever we
	* could call into the DMU, because the DMU can call down to
	* us (e.g. via zio_free()) at any time.
	*
	* The spa_vdev_remove_thread() can be reading metaslab state
	* concurrently, and it is locked out by the ms_sync_lock.
	* Note that the ms_lock is insufficient for this, because it
	* is dropped by space_map_write().
	*/
	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	/*
	* Generate a log space map if one doesn't exist already.
	*/
	spa_generate_syncing_log_sm(spa, tx);

	if (msp->ms_sm == NULL) {
	uint64_t new_object = space_map_alloc(mos,
	spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
	zfs_metaslab_sm_blksz_with_log :
	zfs_metaslab_sm_blksz_no_log, tx);
	VERIFY3U(new_object, !=, 0);

	dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
	msp->ms_id, sizeof (uint64_t), &new_object, tx);

	VERIFY0(space_map_open(&msp->ms_sm, mos, new_object,
	msp->ms_start, msp->ms_size, vd->vdev_ashift));
	ASSERT(msp->ms_sm != NULL);

	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
	ASSERT0(metaslab_allocated_space(msp));
	}

	if (!range_tree_is_empty(msp->ms_checkpointing) &&
	vd->vdev_checkpoint_sm == NULL) {
	ASSERT(spa_has_checkpoint(spa));

	uint64_t new_object = space_map_alloc(mos,
	zfs_vdev_standard_sm_blksz, tx);
	VERIFY3U(new_object, !=, 0);

	VERIFY0(space_map_open(&vd->vdev_checkpoint_sm,
	mos, new_object, 0, vd->vdev_asize, vd->vdev_ashift));
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* We save the space map object as an entry in vdev_top_zap
	* so it can be retrieved when the pool is reopened after an
	* export or through zdb.
	*/
	VERIFY0(zap_add(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM,
	sizeof (new_object), 1, &new_object, tx));
	}

	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	/*
	* Note: metaslab_condense() clears the space map's histogram.
	* Therefore we must verify and remove this histogram before
	* condensing.
	*/
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);
	metaslab_group_histogram_remove(mg, msp);

	if (spa->spa_sync_pass == 1 && msp->ms_loaded &&
	metaslab_should_condense(msp))
	metaslab_condense(msp, tx);

	/*
	* We'll be going to disk to sync our space accounting, thus we
	* drop the ms_lock during that time so allocations coming from
	* open-context (ZIL) for future TXGs do not block.
	*/
	mutex_exit(&msp->ms_lock);
	space_map_t *log_sm = spa_syncing_log_sm(spa);
	if (log_sm != NULL) {
	ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));
	if (metaslab_unflushed_txg(msp) == 0)
	metaslab_unflushed_add(msp, tx);
	else if (!metaslab_unflushed_dirty(msp))
	metaslab_unflushed_bump(msp, tx, B_TRUE);

	space_map_write(log_sm, alloctree, SM_ALLOC,
	vd->vdev_id, tx);
	space_map_write(log_sm, msp->ms_freeing, SM_FREE,
	vd->vdev_id, tx);
	mutex_enter(&msp->ms_lock);

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_remove_xor_add(alloctree,
	msp->ms_unflushed_frees, msp->ms_unflushed_allocs);
	range_tree_remove_xor_add(msp->ms_freeing,
	msp->ms_unflushed_allocs, msp->ms_unflushed_frees);
	spa->spa_unflushed_stats.sus_memused +=
	metaslab_unflushed_changes_memused(msp);
	} else {
	ASSERT(!spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));

	space_map_write(msp->ms_sm, alloctree, SM_ALLOC,
	SM_NO_VDEVID, tx);
	space_map_write(msp->ms_sm, msp->ms_freeing, SM_FREE,
	SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);
	}

	msp->ms_allocated_space += range_tree_space(alloctree);
	ASSERT3U(msp->ms_allocated_space, >=,
	range_tree_space(msp->ms_freeing));
	msp->ms_allocated_space -= range_tree_space(msp->ms_freeing);

	if (!range_tree_is_empty(msp->ms_checkpointing)) {
	ASSERT(spa_has_checkpoint(spa));
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* Since we are doing writes to disk and the ms_checkpointing
	* tree won't be changing during that time, we drop the
	* ms_lock while writing to the checkpoint space map, for the
	* same reason mentioned above.
	*/
	mutex_exit(&msp->ms_lock);
	space_map_write(vd->vdev_checkpoint_sm,
	msp->ms_checkpointing, SM_FREE, SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);

	spa->spa_checkpoint_info.sci_dspace +=
	range_tree_space(msp->ms_checkpointing);
	vd->vdev_stat.vs_checkpoint_space +=
	range_tree_space(msp->ms_checkpointing);
	ASSERT3U(vd->vdev_stat.vs_checkpoint_space, ==,
	-space_map_allocated(vd->vdev_checkpoint_sm));

	range_tree_vacate(msp->ms_checkpointing, NULL, NULL);
	}

	if (msp->ms_loaded) {
	/*
	* When the space map is loaded, we have an accurate
	* histogram in the range tree. This gives us an opportunity
	* to bring the space map's histogram up-to-date so we clear
	* it first before updating it.
	*/
	space_map_histogram_clear(msp->ms_sm);
	space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);

	/*
	* Since we've cleared the histogram we need to add back
	* any free space that has already been processed, plus
	* any deferred space. This allows the on-disk histogram
	* to accurately reflect all free space even if some space
	* is not yet available for allocation (i.e. deferred).
	*/
	space_map_histogram_add(msp->ms_sm, msp->ms_freed, tx);

	/*
	* Add back any deferred free space that has not been
	* added back into the in-core free tree yet. This will
	* ensure that we don't end up with a space map histogram
	* that is completely empty unless the metaslab is fully
	* allocated.
	*/
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	space_map_histogram_add(msp->ms_sm,
	msp->ms_defer[t], tx);
	}
	}

	/*
	* Always add the free space from this sync pass to the space
	* map histogram. We want to make sure that the on-disk histogram
	* accounts for all free space. If the space map is not loaded,
	* then we will lose some accuracy but will correct it the next
	* time we load the space map.
	*/
	space_map_histogram_add(msp->ms_sm, msp->ms_freeing, tx);
	metaslab_aux_histograms_update(msp);

	metaslab_group_histogram_add(mg, msp);
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);

	/*
	* For sync pass 1, we avoid traversing this txg's free range tree
	* and instead will just swap the pointers for freeing and freed.
	* We can safely do this since the freed_tree is guaranteed to be
	* empty on the initial pass.
	*
	* Keep in mind that even if we are currently using a log spacemap
	* we want current frees to end up in the ms_allocatable (but not
	* get appended to the ms_sm) so their ranges can be reused as usual.
	*/
	if (spa_sync_pass(spa) == 1) {
	range_tree_swap(&msp->ms_freeing, &msp->ms_freed);
	ASSERT0(msp->ms_allocated_this_txg);
	} else {
	range_tree_vacate(msp->ms_freeing,
	range_tree_add, msp->ms_freed);
	}
	msp->ms_allocated_this_txg += range_tree_space(alloctree);
	range_tree_vacate(alloctree, NULL, NULL);

	ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_allocating[TXG_CLEAN(txg)
	& TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_checkpointing));

	mutex_exit(&msp->ms_lock);

	/*
	* Verify that the space map object ID has been recorded in the
	* vdev_ms_array.
	*/
	uint64_t object;
	VERIFY0(dmu_read(mos, vd->vdev_ms_array,
	msp->ms_id * sizeof (uint64_t), sizeof (uint64_t), &object, 0));
	VERIFY3U(object, ==, space_map_object(msp->ms_sm));

	mutex_exit(&msp->ms_sync_lock);
	dmu_tx_commit(tx);
	}

	static void
	metaslab_evict(metaslab_t *msp, uint64_t txg)
	{
	if (!msp->ms_loaded \|\| msp->ms_disabled != 0)
	return;

	for (int t = 1; t < TXG_CONCURRENT_STATES; t++) {
	VERIFY0(range_tree_space(
	msp->ms_allocating[(txg + t) & TXG_MASK]));
	}
	if (msp->ms_allocator != -1)
	metaslab_passivate(msp, msp->ms_weight & ~METASLAB_ACTIVE_MASK);

	if (!metaslab_debug_unload)
	metaslab_unload(msp);
	}

	/*
	* Called after a transaction group has completely synced to mark
	* all of the metaslab's free space as usable.
	*/
	void
	metaslab_sync_done(metaslab_t *msp, uint64_t txg)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	range_tree_t **defer_tree;
	int64_t alloc_delta, defer_delta;
	boolean_t defer_allowed = B_TRUE;

	ASSERT(!vd->vdev_ishole);

	mutex_enter(&msp->ms_lock);

	if (msp->ms_new) {
	/* this is a new metaslab, add its capacity to the vdev */
	metaslab_space_update(vd, mg->mg_class, 0, 0, msp->ms_size);

	/* there should be no allocations nor frees at this point */
	VERIFY0(msp->ms_allocated_this_txg);
	VERIFY0(range_tree_space(msp->ms_freed));
	}

	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_checkpointing));

	defer_tree = &msp->ms_defer[txg % TXG_DEFER_SIZE];

	uint64_t free_space = metaslab_class_get_space(spa_normal_class(spa)) -
	metaslab_class_get_alloc(spa_normal_class(spa));
	if (free_space <= spa_get_slop_space(spa) \|\| vd->vdev_removing) {
	defer_allowed = B_FALSE;
	}

	defer_delta = 0;
	alloc_delta = msp->ms_allocated_this_txg -
	range_tree_space(msp->ms_freed);

	if (defer_allowed) {
	defer_delta = range_tree_space(msp->ms_freed) -
	range_tree_space(*defer_tree);
	} else {
	defer_delta -= range_tree_space(*defer_tree);
	}
	metaslab_space_update(vd, mg->mg_class, alloc_delta + defer_delta,
	defer_delta, 0);

	if (spa_syncing_log_sm(spa) == NULL) {
	/*
	* If there's a metaslab_load() in progress and we don't have
	* a log space map, it means that we probably wrote to the
	* metaslab's space map. If this is the case, we need to
	* make sure that we wait for the load to complete so that we
	* have a consistent view at the in-core side of the metaslab.
	*/
	metaslab_load_wait(msp);
	} else {
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
	}

	/*
	* When auto-trimming is enabled, free ranges which are added to
	* ms_allocatable are also be added to ms_trim. The ms_trim tree is
	* periodically consumed by the vdev_autotrim_thread() which issues
	* trims for all ranges and then vacates the tree. The ms_trim tree
	* can be discarded at any time with the sole consequence of recent
	* frees not being trimmed.
	*/
	if (spa_get_autotrim(spa) == SPA_AUTOTRIM_ON) {
	range_tree_walk(*defer_tree, range_tree_add, msp->ms_trim);
	if (!defer_allowed) {
	range_tree_walk(msp->ms_freed, range_tree_add,
	msp->ms_trim);
	}
	} else {
	range_tree_vacate(msp->ms_trim, NULL, NULL);
	}

	/*
	* Move the frees from the defer_tree back to the free
	* range tree (if it's loaded). Swap the freed_tree and
	* the defer_tree -- this is safe to do because we've
	* just emptied out the defer_tree.
	*/
	range_tree_vacate(*defer_tree,
	msp->ms_loaded ? range_tree_add : NULL, msp->ms_allocatable);
	if (defer_allowed) {
	range_tree_swap(&msp->ms_freed, defer_tree);
	} else {
	range_tree_vacate(msp->ms_freed,
	msp->ms_loaded ? range_tree_add : NULL,
	msp->ms_allocatable);
	}

	msp->ms_synced_length = space_map_length(msp->ms_sm);

	msp->ms_deferspace += defer_delta;
	ASSERT3S(msp->ms_deferspace, >=, 0);
	ASSERT3S(msp->ms_deferspace, <=, msp->ms_size);
	if (msp->ms_deferspace != 0) {
	/*
	* Keep syncing this metaslab until all deferred frees
	* are back in circulation.
	*/
	vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
	}
	metaslab_aux_histograms_update_done(msp, defer_allowed);

	if (msp->ms_new) {
	msp->ms_new = B_FALSE;
	mutex_enter(&mg->mg_lock);
	mg->mg_ms_ready++;
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Re-sort metaslab within its group now that we've adjusted
	* its allocatable space.
	*/
	metaslab_recalculate_weight_and_sort(msp);

	ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_freed));
	ASSERT0(range_tree_space(msp->ms_checkpointing));
	msp->ms_allocating_total -= msp->ms_allocated_this_txg;
	msp->ms_allocated_this_txg = 0;
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_sync_reassess(metaslab_group_t *mg)
	{
	spa_t *spa = mg->mg_class->mc_spa;

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
	metaslab_group_alloc_update(mg);
	mg->mg_fragmentation = metaslab_group_fragmentation(mg);

	/*
	* Preload the next potential metaslabs but only on active
	* metaslab groups. We can get into a state where the metaslab
	* is no longer active since we dirty metaslabs as we remove a
	* a device, thus potentially making the metaslab group eligible
	* for preloading.
	*/
	if (mg->mg_activation_count > 0) {
	metaslab_group_preload(mg);
	}
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	}

	/*
	* When writing a ditto block (i.e. more than one DVA for a given BP) on
	* the same vdev as an existing DVA of this BP, then try to allocate it
	* on a different metaslab than existing DVAs (i.e. a unique metaslab).
	*/
	static boolean_t
	metaslab_is_unique(metaslab_t msp, dva_t dva)
	{
	uint64_t dva_ms_id;

	if (DVA_GET_ASIZE(dva) == 0)
	return (B_TRUE);

	if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
	return (B_TRUE);

	dva_ms_id = DVA_GET_OFFSET(dva) >> msp->ms_group->mg_vd->vdev_ms_shift;

	return (msp->ms_id != dva_ms_id);
	}

	/*
	* ==========================================================================
	* Metaslab allocation tracing facility
	* ==========================================================================
	*/

	/*
	* Add an allocation trace element to the allocation tracing list.
	*/
	static void
	metaslab_trace_add(zio_alloc_list_t zal, metaslab_group_t mg,
	metaslab_t *msp, uint64_t psize, uint32_t dva_id, uint64_t offset,
	int allocator)
	{
	metaslab_alloc_trace_t *mat;

	if (!metaslab_trace_enabled)
	return;

	/*
	* When the tracing list reaches its maximum we remove
	* the second element in the list before adding a new one.
	* By removing the second element we preserve the original
	* entry as a clue to what allocations steps have already been
	* performed.
	*/
	if (zal->zal_size == metaslab_trace_max_entries) {
	metaslab_alloc_trace_t *mat_next;
	#ifdef ZFS_DEBUG
	panic("too many entries in allocation list");
	#endif
	METASLABSTAT_BUMP(metaslabstat_trace_over_limit);
	zal->zal_size--;
	mat_next = list_next(&zal->zal_list, list_head(&zal->zal_list));
	list_remove(&zal->zal_list, mat_next);
	kmem_cache_free(metaslab_alloc_trace_cache, mat_next);
	}

	mat = kmem_cache_alloc(metaslab_alloc_trace_cache, KM_SLEEP);
	list_link_init(&mat->mat_list_node);
	mat->mat_mg = mg;
	mat->mat_msp = msp;
	mat->mat_size = psize;
	mat->mat_dva_id = dva_id;
	mat->mat_offset = offset;
	mat->mat_weight = 0;
	mat->mat_allocator = allocator;

	if (msp != NULL)
	mat->mat_weight = msp->ms_weight;

	/*
	* The list is part of the zio so locking is not required. Only
	* a single thread will perform allocations for a given zio.
	*/
	list_insert_tail(&zal->zal_list, mat);
	zal->zal_size++;

	ASSERT3U(zal->zal_size, <=, metaslab_trace_max_entries);
	}

	void
	metaslab_trace_init(zio_alloc_list_t *zal)
	{
	list_create(&zal->zal_list, sizeof (metaslab_alloc_trace_t),
	offsetof(metaslab_alloc_trace_t, mat_list_node));
	zal->zal_size = 0;
	}

	void
	metaslab_trace_fini(zio_alloc_list_t *zal)
	{
	metaslab_alloc_trace_t *mat;

	while ((mat = list_remove_head(&zal->zal_list)) != NULL)
	kmem_cache_free(metaslab_alloc_trace_cache, mat);
	list_destroy(&zal->zal_list);
	zal->zal_size = 0;
	}

	/*
	* ==========================================================================
	* Metaslab block operations
	* ==========================================================================
	*/

	static void
	metaslab_group_alloc_increment(spa_t spa, uint64_t vdev, const void tag,
	int flags, int allocator)
	{
	if (!(flags & METASLAB_ASYNC_ALLOC) \|\|
	(flags & METASLAB_DONT_THROTTLE))
	return;

	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
	return;

	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	(void) zfs_refcount_add(&mga->mga_alloc_queue_depth, tag);
	}

	static void
	metaslab_group_increment_qdepth(metaslab_group_t *mg, int allocator)
	{
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	metaslab_class_allocator_t *mca =
	&mg->mg_class->mc_allocator[allocator];
	uint64_t max = mg->mg_max_alloc_queue_depth;
	uint64_t cur = mga->mga_cur_max_alloc_queue_depth;
	while (cur < max) {
	if (atomic_cas_64(&mga->mga_cur_max_alloc_queue_depth,
	cur, cur + 1) == cur) {
	atomic_inc_64(&mca->mca_alloc_max_slots);
	return;
	}
	cur = mga->mga_cur_max_alloc_queue_depth;
	}
	}

	void
	metaslab_group_alloc_decrement(spa_t spa, uint64_t vdev, const void tag,
	int flags, int allocator, boolean_t io_complete)
	{
	if (!(flags & METASLAB_ASYNC_ALLOC) \|\|
	(flags & METASLAB_DONT_THROTTLE))
	return;

	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
	return;

	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	(void) zfs_refcount_remove(&mga->mga_alloc_queue_depth, tag);
	if (io_complete)
	metaslab_group_increment_qdepth(mg, allocator);
	}

	void
	metaslab_group_alloc_verify(spa_t spa, const blkptr_t bp, const void *tag,
	int allocator)
	{
	#ifdef ZFS_DEBUG
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);

	for (int d = 0; d < ndvas; d++) {
	uint64_t vdev = DVA_GET_VDEV(&dva[d]);
	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	VERIFY(zfs_refcount_not_held(&mga->mga_alloc_queue_depth, tag));
	}
	#endif
	}

	static uint64_t
	metaslab_block_alloc(metaslab_t *msp, uint64_t size, uint64_t txg)
	{
	uint64_t start;
	range_tree_t *rt = msp->ms_allocatable;
	metaslab_class_t *mc = msp->ms_group->mg_class;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	VERIFY(!msp->ms_condensing);
	VERIFY0(msp->ms_disabled);

	start = mc->mc_ops->msop_alloc(msp, size);
	if (start != -1ULL) {
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;

	VERIFY0(P2PHASE(start, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	VERIFY3U(range_tree_space(rt) - size, <=, msp->ms_size);
	range_tree_remove(rt, start, size);
	range_tree_clear(msp->ms_trim, start, size);

	if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
	vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);

	range_tree_add(msp->ms_allocating[txg & TXG_MASK], start, size);
	msp->ms_allocating_total += size;

	/* Track the last successful allocation */
	msp->ms_alloc_txg = txg;
	metaslab_verify_space(msp, txg);
	}

	/*
	* Now that we've attempted the allocation we need to update the
	* metaslab's maximum block size since it may have changed.
	*/
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	return (start);
	}

	/*
	* Find the metaslab with the highest weight that is less than what we've
	* already tried. In the common case, this means that we will examine each
	* metaslab at most once. Note that concurrent callers could reorder metaslabs
	* by activation/passivation once we have dropped the mg_lock. If a metaslab is
	* activated by another thread, and we fail to allocate from the metaslab we
	* have selected, we may not try the newly-activated metaslab, and instead
	* activate another metaslab. This is not optimal, but generally does not cause
	* any problems (a possible exception being if every metaslab is completely full
	* except for the newly-activated metaslab which we fail to examine).
	*/
	static metaslab_t *
	find_valid_metaslab(metaslab_group_t *mg, uint64_t activation_weight,
	dva_t *dva, int d, boolean_t want_unique, uint64_t asize, int allocator,
	boolean_t try_hard, zio_alloc_list_t zal, metaslab_t search,
	boolean_t *was_active)
	{
	avl_index_t idx;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	metaslab_t *msp = avl_find(t, search, &idx);
	if (msp == NULL)
	msp = avl_nearest(t, idx, AVL_AFTER);

	- int tries = 0;
	+ uint_t tries = 0;
	for (; msp != NULL; msp = AVL_NEXT(t, msp)) {
	int i;

	if (!try_hard && tries > zfs_metaslab_find_max_tries) {
	METASLABSTAT_BUMP(metaslabstat_too_many_tries);
	return (NULL);
	}
	tries++;

	if (!metaslab_should_allocate(msp, asize, try_hard)) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_TOO_SMALL, allocator);
	continue;
	}

	/*
	* If the selected metaslab is condensing or disabled,
	* skip it.
	*/
	if (msp->ms_condensing \|\| msp->ms_disabled > 0)
	continue;

	*was_active = msp->ms_allocator != -1;
	/*
	* If we're activating as primary, this is our first allocation
	* from this disk, so we don't need to check how close we are.
	* If the metaslab under consideration was already active,
	* we're getting desperate enough to steal another allocator's
	* metaslab, so we still don't care about distances.
	*/
	if (activation_weight == METASLAB_WEIGHT_PRIMARY \|\| *was_active)
	break;

	for (i = 0; i < d; i++) {
	if (want_unique &&
	!metaslab_is_unique(msp, &dva[i]))
	break; /* try another metaslab */
	}
	if (i == d)
	break;
	}

	if (msp != NULL) {
	search->ms_weight = msp->ms_weight;
	search->ms_start = msp->ms_start + 1;
	search->ms_allocator = msp->ms_allocator;
	search->ms_primary = msp->ms_primary;
	}
	return (msp);
	}

	static void
	metaslab_active_mask_verify(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0)
	return;

	if (msp->ms_weight & METASLAB_WEIGHT_PRIMARY) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
	VERIFY3S(msp->ms_allocator, !=, -1);
	VERIFY(msp->ms_primary);
	return;
	}

	if (msp->ms_weight & METASLAB_WEIGHT_SECONDARY) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
	VERIFY3S(msp->ms_allocator, !=, -1);
	VERIFY(!msp->ms_primary);
	return;
	}

	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	VERIFY3S(msp->ms_allocator, ==, -1);
	return;
	}
	}

	static uint64_t
	metaslab_group_alloc_normal(metaslab_group_t mg, zio_alloc_list_t zal,
	uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
	int allocator, boolean_t try_hard)
	{
	metaslab_t *msp = NULL;
	uint64_t offset = -1ULL;

	uint64_t activation_weight = METASLAB_WEIGHT_PRIMARY;
	for (int i = 0; i < d; i++) {
	if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
	DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
	activation_weight = METASLAB_WEIGHT_SECONDARY;
	} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
	DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
	activation_weight = METASLAB_WEIGHT_CLAIM;
	break;
	}
	}

	/*
	* If we don't have enough metaslabs active to fill the entire array, we
	* just use the 0th slot.
	*/
	if (mg->mg_ms_ready < mg->mg_allocators * 3)
	allocator = 0;
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];

	ASSERT3U(mg->mg_vd->vdev_ms_count, >=, 2);

	metaslab_t search = kmem_alloc(sizeof (search), KM_SLEEP);
	search->ms_weight = UINT64_MAX;
	search->ms_start = 0;
	/*
	* At the end of the metaslab tree are the already-active metaslabs,
	* first the primaries, then the secondaries. When we resume searching
	* through the tree, we need to consider ms_allocator and ms_primary so
	* we start in the location right after where we left off, and don't
	* accidentally loop forever considering the same metaslabs.
	*/
	search->ms_allocator = -1;
	search->ms_primary = B_TRUE;
	for (;;) {
	boolean_t was_active = B_FALSE;

	mutex_enter(&mg->mg_lock);

	if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
	mga->mga_primary != NULL) {
	msp = mga->mga_primary;

	/*
	* Even though we don't hold the ms_lock for the
	* primary metaslab, those fields should not
	* change while we hold the mg_lock. Thus it is
	* safe to make assertions on them.
	*/
	ASSERT(msp->ms_primary);
	ASSERT3S(msp->ms_allocator, ==, allocator);
	ASSERT(msp->ms_loaded);

	was_active = B_TRUE;
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
	mga->mga_secondary != NULL) {
	msp = mga->mga_secondary;

	/*
	* See comment above about the similar assertions
	* for the primary metaslab.
	*/
	ASSERT(!msp->ms_primary);
	ASSERT3S(msp->ms_allocator, ==, allocator);
	ASSERT(msp->ms_loaded);

	was_active = B_TRUE;
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	} else {
	msp = find_valid_metaslab(mg, activation_weight, dva, d,
	want_unique, asize, allocator, try_hard, zal,
	search, &was_active);
	}

	mutex_exit(&mg->mg_lock);
	if (msp == NULL) {
	kmem_free(search, sizeof (*search));
	return (-1ULL);
	}
	mutex_enter(&msp->ms_lock);

	metaslab_active_mask_verify(msp);

	/*
	* This code is disabled out because of issues with
	* tracepoints in non-gpl kernel modules.
	*/
	#if 0
	DTRACE_PROBE3(ms__activation__attempt,
	metaslab_t *, msp, uint64_t, activation_weight,
	boolean_t, was_active);
	#endif

	/*
	* Ensure that the metaslab we have selected is still
	* capable of handling our request. It's possible that
	* another thread may have changed the weight while we
	* were blocked on the metaslab lock. We check the
	* active status first to see if we need to set_selected_txg
	* a new metaslab.
	*/
	if (was_active && !(msp->ms_weight & METASLAB_ACTIVE_MASK)) {
	ASSERT3S(msp->ms_allocator, ==, -1);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* If the metaslab was activated for another allocator
	* while we were waiting in the ms_lock above, or it's
	* a primary and we're seeking a secondary (or vice versa),
	* we go back and select a new metaslab.
	*/
	if (!was_active && (msp->ms_weight & METASLAB_ACTIVE_MASK) &&
	(msp->ms_allocator != -1) &&
	(msp->ms_allocator != allocator \|\| ((activation_weight ==
	METASLAB_WEIGHT_PRIMARY) != msp->ms_primary))) {
	ASSERT(msp->ms_loaded);
	ASSERT((msp->ms_weight & METASLAB_WEIGHT_CLAIM) \|\|
	msp->ms_allocator != -1);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* This metaslab was used for claiming regions allocated
	* by the ZIL during pool import. Once these regions are
	* claimed we don't need to keep the CLAIM bit set
	* anymore. Passivate this metaslab to zero its activation
	* mask.
	*/
	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM &&
	activation_weight != METASLAB_WEIGHT_CLAIM) {
	ASSERT(msp->ms_loaded);
	ASSERT3S(msp->ms_allocator, ==, -1);
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_WEIGHT_CLAIM);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	metaslab_set_selected_txg(msp, txg);

	int activation_error =
	metaslab_activate(msp, allocator, activation_weight);
	metaslab_active_mask_verify(msp);

	/*
	* If the metaslab was activated by another thread for
	* another allocator or activation_weight (EBUSY), or it
	* failed because another metaslab was assigned as primary
	* for this allocator (EEXIST) we continue using this
	* metaslab for our allocation, rather than going on to a
	* worse metaslab (we waited for that metaslab to be loaded
	* after all).
	*
	* If the activation failed due to an I/O error or ENOSPC we
	* skip to the next metaslab.
	*/
	boolean_t activated;
	if (activation_error == 0) {
	activated = B_TRUE;
	} else if (activation_error == EBUSY \|\|
	activation_error == EEXIST) {
	activated = B_FALSE;
	} else {
	mutex_exit(&msp->ms_lock);
	continue;
	}
	ASSERT(msp->ms_loaded);

	/*
	* Now that we have the lock, recheck to see if we should
	* continue to use this metaslab for this allocation. The
	* the metaslab is now loaded so metaslab_should_allocate()
	* can accurately determine if the allocation attempt should
	* proceed.
	*/
	if (!metaslab_should_allocate(msp, asize, try_hard)) {
	/* Passivate this metaslab and select a new one. */
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_TOO_SMALL, allocator);
	goto next;
	}

	/*
	* If this metaslab is currently condensing then pick again
	* as we can't manipulate this metaslab until it's committed
	* to disk. If this metaslab is being initialized, we shouldn't
	* allocate from it since the allocated region might be
	* overwritten after allocation.
	*/
	if (msp->ms_condensing) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_CONDENSING, allocator);
	if (activated) {
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_ACTIVE_MASK);
	}
	mutex_exit(&msp->ms_lock);
	continue;
	} else if (msp->ms_disabled > 0) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_DISABLED, allocator);
	if (activated) {
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_ACTIVE_MASK);
	}
	mutex_exit(&msp->ms_lock);
	continue;
	}

	offset = metaslab_block_alloc(msp, asize, txg);
	metaslab_trace_add(zal, mg, msp, asize, d, offset, allocator);

	if (offset != -1ULL) {
	/* Proactively passivate the metaslab, if needed */
	if (activated)
	metaslab_segment_may_passivate(msp);
	break;
	}
	next:
	ASSERT(msp->ms_loaded);

	/*
	* This code is disabled out because of issues with
	* tracepoints in non-gpl kernel modules.
	*/
	#if 0
	DTRACE_PROBE2(ms__alloc__failure, metaslab_t *, msp,
	uint64_t, asize);
	#endif

	/*
	* We were unable to allocate from this metaslab so determine
	* a new weight for this metaslab. Now that we have loaded
	* the metaslab we can provide a better hint to the metaslab
	* selector.
	*
	* For space-based metaslabs, we use the maximum block size.
	* This information is only available when the metaslab
	* is loaded and is more accurate than the generic free
	* space weight that was calculated by metaslab_weight().
	* This information allows us to quickly compare the maximum
	* available allocation in the metaslab to the allocation
	* size being requested.
	*
	* For segment-based metaslabs, determine the new weight
	* based on the highest bucket in the range tree. We
	* explicitly use the loaded segment weight (i.e. the range
	* tree histogram) since it contains the space that is
	* currently available for allocation and is accurate
	* even within a sync pass.
	*/
	uint64_t weight;
	if (WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
	weight = metaslab_largest_allocatable(msp);
	WEIGHT_SET_SPACEBASED(weight);
	} else {
	weight = metaslab_weight_from_range_tree(msp);
	}

	if (activated) {
	metaslab_passivate(msp, weight);
	} else {
	/*
	* For the case where we use the metaslab that is
	* active for another allocator we want to make
	* sure that we retain the activation mask.
	*
	* Note that we could attempt to use something like
	* metaslab_recalculate_weight_and_sort() that
	* retains the activation mask here. That function
	* uses metaslab_weight() to set the weight though
	* which is not as accurate as the calculations
	* above.
	*/
	weight \|= msp->ms_weight & METASLAB_ACTIVE_MASK;
	metaslab_group_sort(mg, msp, weight);
	}
	metaslab_active_mask_verify(msp);

	/*
	* We have just failed an allocation attempt, check
	* that metaslab_should_allocate() agrees. Otherwise,
	* we may end up in an infinite loop retrying the same
	* metaslab.
	*/
	ASSERT(!metaslab_should_allocate(msp, asize, try_hard));

	mutex_exit(&msp->ms_lock);
	}
	mutex_exit(&msp->ms_lock);
	kmem_free(search, sizeof (*search));
	return (offset);
	}

	static uint64_t
	metaslab_group_alloc(metaslab_group_t mg, zio_alloc_list_t zal,
	uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
	int allocator, boolean_t try_hard)
	{
	uint64_t offset;
	ASSERT(mg->mg_initialized);

	offset = metaslab_group_alloc_normal(mg, zal, asize, txg, want_unique,
	dva, d, allocator, try_hard);

	mutex_enter(&mg->mg_lock);
	if (offset == -1ULL) {
	mg->mg_failed_allocations++;
	metaslab_trace_add(zal, mg, NULL, asize, d,
	TRACE_GROUP_FAILURE, allocator);
	if (asize == SPA_GANGBLOCKSIZE) {
	/*
	* This metaslab group was unable to allocate
	* the minimum gang block size so it must be out of
	* space. We must notify the allocation throttle
	* to start skipping allocation attempts to this
	* metaslab group until more space becomes available.
	* Note: this failure cannot be caused by the
	* allocation throttle since the allocation throttle
	* is only responsible for skipping devices and
	* not failing block allocations.
	*/
	mg->mg_no_free_space = B_TRUE;
	}
	}
	mg->mg_allocations++;
	mutex_exit(&mg->mg_lock);
	return (offset);
	}

	/*
	* Allocate a block for the specified i/o.
	*/
	int
	metaslab_alloc_dva(spa_t spa, metaslab_class_t mc, uint64_t psize,
	dva_t dva, int d, dva_t hintdva, uint64_t txg, int flags,
	zio_alloc_list_t *zal, int allocator)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
	metaslab_group_t mg, fast_mg, *rotor;
	vdev_t *vd;
	boolean_t try_hard = B_FALSE;

	ASSERT(!DVA_IS_VALID(&dva[d]));

	/*
	* For testing, make some blocks above a certain size be gang blocks.
	* This will result in more split blocks when using device removal,
	* and a large number of split blocks coupled with ztest-induced
	* damage can result in extremely long reconstruction times. This
	* will also test spilling from special to normal.
	*/
	if (psize >= metaslab_force_ganging && (random_in_range(100) < 3)) {
	metaslab_trace_add(zal, NULL, NULL, psize, d, TRACE_FORCE_GANG,
	allocator);
	return (SET_ERROR(ENOSPC));
	}

	/*
	* Start at the rotor and loop through all mgs until we find something.
	* Note that there's no locking on mca_rotor or mca_aliquot because
	* nothing actually breaks if we miss a few updates -- we just won't
	* allocate quite as evenly. It all balances out over time.
	*
	* If we are doing ditto or log blocks, try to spread them across
	* consecutive vdevs. If we're forced to reuse a vdev before we've
	* allocated all of our ditto blocks, then try and spread them out on
	* that vdev as much as possible. If it turns out to not be possible,
	* gradually lower our standards until anything becomes acceptable.
	* Also, allocating on consecutive vdevs (as opposed to random vdevs)
	* gives us hope of containing our fault domains to something we're
	* able to reason about. Otherwise, any two top-level vdev failures
	* will guarantee the loss of data. With consecutive allocation,
	* only two adjacent top-level vdev failures will result in data loss.
	*
	* If we are doing gang blocks (hintdva is non-NULL), try to keep
	* ourselves on the same vdev as our gang block header. That
	* way, we can hope for locality in vdev_cache, plus it makes our
	* fault domains something tractable.
	*/
	if (hintdva) {
	vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));

	/*
	* It's possible the vdev we're using as the hint no
	* longer exists or its mg has been closed (e.g. by
	* device removal). Consult the rotor when
	* all else fails.
	*/
	if (vd != NULL && vd->vdev_mg != NULL) {
	mg = vdev_get_mg(vd, mc);

	if (flags & METASLAB_HINTBP_AVOID &&
	mg->mg_next != NULL)
	mg = mg->mg_next;
	} else {
	mg = mca->mca_rotor;
	}
	} else if (d != 0) {
	vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
	mg = vd->vdev_mg->mg_next;
	} else if (flags & METASLAB_FASTWRITE) {
	mg = fast_mg = mca->mca_rotor;

	do {
	if (fast_mg->mg_vd->vdev_pending_fastwrite <
	mg->mg_vd->vdev_pending_fastwrite)
	mg = fast_mg;
	} while ((fast_mg = fast_mg->mg_next) != mca->mca_rotor);

	} else {
	ASSERT(mca->mca_rotor != NULL);
	mg = mca->mca_rotor;
	}

	/*
	* If the hint put us into the wrong metaslab class, or into a
	* metaslab group that has been passivated, just follow the rotor.
	*/
	if (mg->mg_class != mc \|\| mg->mg_activation_count <= 0)
	mg = mca->mca_rotor;

	rotor = mg;
	top:
	do {
	boolean_t allocatable;

	ASSERT(mg->mg_activation_count == 1);
	vd = mg->mg_vd;

	/*
	* Don't allocate from faulted devices.
	*/
	if (try_hard) {
	spa_config_enter(spa, SCL_ZIO, FTAG, RW_READER);
	allocatable = vdev_allocatable(vd);
	spa_config_exit(spa, SCL_ZIO, FTAG);
	} else {
	allocatable = vdev_allocatable(vd);
	}

	/*
	* Determine if the selected metaslab group is eligible
	* for allocations. If we're ganging then don't allow
	* this metaslab group to skip allocations since that would
	* inadvertently return ENOSPC and suspend the pool
	* even though space is still available.
	*/
	if (allocatable && !GANG_ALLOCATION(flags) && !try_hard) {
	allocatable = metaslab_group_allocatable(mg, rotor,
	psize, allocator, d);
	}

	if (!allocatable) {
	metaslab_trace_add(zal, mg, NULL, psize, d,
	TRACE_NOT_ALLOCATABLE, allocator);
	goto next;
	}

	ASSERT(mg->mg_initialized);

	/*
	* Avoid writing single-copy data to a failing,
	* non-redundant vdev, unless we've already tried all
	* other vdevs.
	*/
	if ((vd->vdev_stat.vs_write_errors > 0 \|\|
	vd->vdev_state < VDEV_STATE_HEALTHY) &&
	d == 0 && !try_hard && vd->vdev_children == 0) {
	metaslab_trace_add(zal, mg, NULL, psize, d,
	TRACE_VDEV_ERROR, allocator);
	goto next;
	}

	ASSERT(mg->mg_class == mc);

	uint64_t asize = vdev_psize_to_asize(vd, psize);
	ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);

	/*
	* If we don't need to try hard, then require that the
	* block be on a different metaslab from any other DVAs
	* in this BP (unique=true). If we are trying hard, then
	* allow any metaslab to be used (unique=false).
	*/
	uint64_t offset = metaslab_group_alloc(mg, zal, asize, txg,
	!try_hard, dva, d, allocator, try_hard);

	if (offset != -1ULL) {
	/*
	* If we've just selected this metaslab group,
	* figure out whether the corresponding vdev is
	* over- or under-used relative to the pool,
	* and set an allocation bias to even it out.
	*
	* Bias is also used to compensate for unequally
	* sized vdevs so that space is allocated fairly.
	*/
	if (mca->mca_aliquot == 0 && metaslab_bias_enabled) {
	vdev_stat_t *vs = &vd->vdev_stat;
	int64_t vs_free = vs->vs_space - vs->vs_alloc;
	int64_t mc_free = mc->mc_space - mc->mc_alloc;
	int64_t ratio;

	/*
	* Calculate how much more or less we should
	* try to allocate from this device during
	* this iteration around the rotor.
	*
	* This basically introduces a zero-centered
	* bias towards the devices with the most
	* free space, while compensating for vdev
	* size differences.
	*
	* Examples:
	* vdev V1 = 16M/128M
	* vdev V2 = 16M/128M
	* ratio(V1) = 100% ratio(V2) = 100%
	*
	* vdev V1 = 16M/128M
	* vdev V2 = 64M/128M
	* ratio(V1) = 127% ratio(V2) = 72%
	*
	* vdev V1 = 16M/128M
	* vdev V2 = 64M/512M
	* ratio(V1) = 40% ratio(V2) = 160%
	*/
	ratio = (vs_free * mc->mc_alloc_groups * 100) /
	(mc_free + 1);
	mg->mg_bias = ((ratio - 100) *
	(int64_t)mg->mg_aliquot) / 100;
	} else if (!metaslab_bias_enabled) {
	mg->mg_bias = 0;
	}

	if ((flags & METASLAB_FASTWRITE) \|\|
	atomic_add_64_nv(&mca->mca_aliquot, asize) >=
	mg->mg_aliquot + mg->mg_bias) {
	mca->mca_rotor = mg->mg_next;
	mca->mca_aliquot = 0;
	}

	DVA_SET_VDEV(&dva[d], vd->vdev_id);
	DVA_SET_OFFSET(&dva[d], offset);
	DVA_SET_GANG(&dva[d],
	((flags & METASLAB_GANG_HEADER) ? 1 : 0));
	DVA_SET_ASIZE(&dva[d], asize);

	if (flags & METASLAB_FASTWRITE) {
	atomic_add_64(&vd->vdev_pending_fastwrite,
	psize);
	}

	return (0);
	}
	next:
	mca->mca_rotor = mg->mg_next;
	mca->mca_aliquot = 0;
	} while ((mg = mg->mg_next) != rotor);

	/*
	* If we haven't tried hard, perhaps do so now.
	*/
	if (!try_hard && (zfs_metaslab_try_hard_before_gang \|\|
	GANG_ALLOCATION(flags) \|\| (flags & METASLAB_ZIL) != 0 \|\|
	psize <= 1 << spa->spa_min_ashift)) {
	METASLABSTAT_BUMP(metaslabstat_try_hard);
	try_hard = B_TRUE;
	goto top;
	}

	memset(&dva[d], 0, sizeof (dva_t));

	metaslab_trace_add(zal, rotor, NULL, psize, d, TRACE_ENOSPC, allocator);
	return (SET_ERROR(ENOSPC));
	}

	void
	metaslab_free_concrete(vdev_t *vd, uint64_t offset, uint64_t asize,
	boolean_t checkpoint)
	{
	metaslab_t *msp;
	spa_t *spa = vd->vdev_spa;

	ASSERT(vdev_is_concrete(vd));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
	ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	VERIFY(!msp->ms_condensing);
	VERIFY3U(offset, >=, msp->ms_start);
	VERIFY3U(offset + asize, <=, msp->ms_start + msp->ms_size);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(asize, 1ULL << vd->vdev_ashift));

	metaslab_check_free_impl(vd, offset, asize);

	mutex_enter(&msp->ms_lock);
	if (range_tree_is_empty(msp->ms_freeing) &&
	range_tree_is_empty(msp->ms_checkpointing)) {
	vdev_dirty(vd, VDD_METASLAB, msp, spa_syncing_txg(spa));
	}

	if (checkpoint) {
	ASSERT(spa_has_checkpoint(spa));
	range_tree_add(msp->ms_checkpointing, offset, asize);
	} else {
	range_tree_add(msp->ms_freeing, offset, asize);
	}
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_free_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner_offset;
	boolean_t *checkpoint = arg;

	ASSERT3P(checkpoint, !=, NULL);

	if (vd->vdev_ops->vdev_op_remap != NULL)
	vdev_indirect_mark_obsolete(vd, offset, size);
	else
	metaslab_free_impl(vd, offset, size, *checkpoint);
	}

	static void
	metaslab_free_impl(vdev_t *vd, uint64_t offset, uint64_t size,
	boolean_t checkpoint)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (spa_syncing_txg(spa) > spa_freeze_txg(spa))
	return;

	if (spa->spa_vdev_removal != NULL &&
	spa->spa_vdev_removal->svr_vdev_id == vd->vdev_id &&
	vdev_is_concrete(vd)) {
	/*
	* Note: we check if the vdev is concrete because when
	* we complete the removal, we first change the vdev to be
	* an indirect vdev (in open context), and then (in syncing
	* context) clear spa_vdev_removal.
	*/
	free_from_removing_vdev(vd, offset, size);
	} else if (vd->vdev_ops->vdev_op_remap != NULL) {
	vdev_indirect_mark_obsolete(vd, offset, size);
	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_free_impl_cb, &checkpoint);
	} else {
	metaslab_free_concrete(vd, offset, size, checkpoint);
	}
	}

	typedef struct remap_blkptr_cb_arg {
	blkptr_t *rbca_bp;
	spa_remap_cb_t rbca_cb;
	vdev_t *rbca_remap_vd;
	uint64_t rbca_remap_offset;
	void *rbca_cb_arg;
	} remap_blkptr_cb_arg_t;

	static void
	remap_blkptr_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	remap_blkptr_cb_arg_t *rbca = arg;
	blkptr_t *bp = rbca->rbca_bp;

	/* We can not remap split blocks. */
	if (size != DVA_GET_ASIZE(&bp->blk_dva[0]))
	return;
	ASSERT0(inner_offset);

	if (rbca->rbca_cb != NULL) {
	/*
	* At this point we know that we are not handling split
	* blocks and we invoke the callback on the previous
	* vdev which must be indirect.
	*/
	ASSERT3P(rbca->rbca_remap_vd->vdev_ops, ==, &vdev_indirect_ops);

	rbca->rbca_cb(rbca->rbca_remap_vd->vdev_id,
	rbca->rbca_remap_offset, size, rbca->rbca_cb_arg);

	/* set up remap_blkptr_cb_arg for the next call */
	rbca->rbca_remap_vd = vd;
	rbca->rbca_remap_offset = offset;
	}

	/*
	* The phys birth time is that of dva[0]. This ensures that we know
	* when each dva was written, so that resilver can determine which
	* blocks need to be scrubbed (i.e. those written during the time
	* the vdev was offline). It also ensures that the key used in
	* the ARC hash table is unique (i.e. dva[0] + phys_birth). If
	* we didn't change the phys_birth, a lookup in the ARC for a
	* remapped BP could find the data that was previously stored at
	* this vdev + offset.
	*/
	vdev_t *oldvd = vdev_lookup_top(vd->vdev_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]));
	vdev_indirect_births_t *vib = oldvd->vdev_indirect_births;
	bp->blk_phys_birth = vdev_indirect_births_physbirth(vib,
	DVA_GET_OFFSET(&bp->blk_dva[0]), DVA_GET_ASIZE(&bp->blk_dva[0]));

	DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id);
	DVA_SET_OFFSET(&bp->blk_dva[0], offset);
	}

	/*
	* If the block pointer contains any indirect DVAs, modify them to refer to
	* concrete DVAs. Note that this will sometimes not be possible, leaving
	* the indirect DVA in place. This happens if the indirect DVA spans multiple
	* segments in the mapping (i.e. it is a "split block").
	*
	* If the BP was remapped, calls the callback on the original dva (note the
	* callback can be called multiple times if the original indirect DVA refers
	* to another indirect DVA, etc).
	*
	* Returns TRUE if the BP was remapped.
	*/
	boolean_t
	spa_remap_blkptr(spa_t spa, blkptr_t bp, spa_remap_cb_t callback, void *arg)
	{
	remap_blkptr_cb_arg_t rbca;

	if (!zfs_remap_blkptr_enable)
	return (B_FALSE);

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS))
	return (B_FALSE);

	/*
	* Dedup BP's can not be remapped, because ddt_phys_select() depends
	* on DVA[0] being the same in the BP as in the DDT (dedup table).
	*/
	if (BP_GET_DEDUP(bp))
	return (B_FALSE);

	/*
	* Gang blocks can not be remapped, because
	* zio_checksum_gang_verifier() depends on the DVA[0] that's in
	* the BP used to read the gang block header (GBH) being the same
	* as the DVA[0] that we allocated for the GBH.
	*/
	if (BP_IS_GANG(bp))
	return (B_FALSE);

	/*
	* Embedded BP's have no DVA to remap.
	*/
	if (BP_GET_NDVAS(bp) < 1)
	return (B_FALSE);

	/*
	* Note: we only remap dva[0]. If we remapped other dvas, we
	* would no longer know what their phys birth txg is.
	*/
	dva_t *dva = &bp->blk_dva[0];

	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));

	if (vd->vdev_ops->vdev_op_remap == NULL)
	return (B_FALSE);

	rbca.rbca_bp = bp;
	rbca.rbca_cb = callback;
	rbca.rbca_remap_vd = vd;
	rbca.rbca_remap_offset = offset;
	rbca.rbca_cb_arg = arg;

	/*
	* remap_blkptr_cb() will be called in order for each level of
	* indirection, until a concrete vdev is reached or a split block is
	* encountered. old_vd and old_offset are updated within the callback
	* as we go from the one indirect vdev to the next one (either concrete
	* or indirect again) in that order.
	*/
	vd->vdev_ops->vdev_op_remap(vd, offset, size, remap_blkptr_cb, &rbca);

	/* Check if the DVA wasn't remapped because it is a split block */
	if (DVA_GET_VDEV(&rbca.rbca_bp->blk_dva[0]) == vd->vdev_id)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Undo the allocation of a DVA which happened in the given transaction group.
	*/
	void
	metaslab_unalloc_dva(spa_t spa, const dva_t dva, uint64_t txg)
	{
	metaslab_t *msp;
	vdev_t *vd;
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);

	ASSERT(DVA_IS_VALID(dva));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (txg > spa_freeze_txg(spa))
	return;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL \|\| !DVA_IS_VALID(dva) \|\|
	(offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
	zfs_panic_recover("metaslab_free_dva(): bad DVA %llu:%llu:%llu",
	(u_longlong_t)vdev, (u_longlong_t)offset,
	(u_longlong_t)size);
	return;
	}

	ASSERT(!vd->vdev_removing);
	ASSERT(vdev_is_concrete(vd));
	ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
	ASSERT3P(vd->vdev_indirect_mapping, ==, NULL);

	if (DVA_GET_GANG(dva))
	size = vdev_gang_header_asize(vd);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);
	range_tree_remove(msp->ms_allocating[txg & TXG_MASK],
	offset, size);
	msp->ms_allocating_total -= size;

	VERIFY(!msp->ms_condensing);
	VERIFY3U(offset, >=, msp->ms_start);
	VERIFY3U(offset + size, <=, msp->ms_start + msp->ms_size);
	VERIFY3U(range_tree_space(msp->ms_allocatable) + size, <=,
	msp->ms_size);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	range_tree_add(msp->ms_allocatable, offset, size);
	mutex_exit(&msp->ms_lock);
	}

	/*
	* Free the block represented by the given DVA.
	*/
	void
	metaslab_free_dva(spa_t spa, const dva_t dva, boolean_t checkpoint)
	{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd = vdev_lookup_top(spa, vdev);

	ASSERT(DVA_IS_VALID(dva));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (DVA_GET_GANG(dva)) {
	size = vdev_gang_header_asize(vd);
	}

	metaslab_free_impl(vd, offset, size, checkpoint);
	}

	/*
	* Reserve some allocation slots. The reservation system must be called
	* before we call into the allocator. If there aren't any available slots
	* then the I/O will be throttled until an I/O completes and its slots are
	* freed up. The function returns true if it was successful in placing
	* the reservation.
	*/
	boolean_t
	metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, int allocator,
	zio_t *zio, int flags)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
	uint64_t max = mca->mca_alloc_max_slots;

	ASSERT(mc->mc_alloc_throttle_enabled);
	if (GANG_ALLOCATION(flags) \|\| (flags & METASLAB_MUST_RESERVE) \|\|
	zfs_refcount_count(&mca->mca_alloc_slots) + slots <= max) {
	/*
	* The potential race between _count() and _add() is covered
	* by the allocator lock in most cases, or irrelevant due to
	* GANG_ALLOCATION() or METASLAB_MUST_RESERVE set in others.
	* But even if we assume some other non-existing scenario, the
	* worst that can happen is few more I/Os get to allocation
	* earlier, that is not a problem.
	*
	* We reserve the slots individually so that we can unreserve
	* them individually when an I/O completes.
	*/
	for (int d = 0; d < slots; d++)
	zfs_refcount_add(&mca->mca_alloc_slots, zio);
	zio->io_flags \|= ZIO_FLAG_IO_ALLOCATING;
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	void
	metaslab_class_throttle_unreserve(metaslab_class_t *mc, int slots,
	int allocator, zio_t *zio)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];

	ASSERT(mc->mc_alloc_throttle_enabled);
	for (int d = 0; d < slots; d++)
	zfs_refcount_remove(&mca->mca_alloc_slots, zio);
	}

	static int
	metaslab_claim_concrete(vdev_t *vd, uint64_t offset, uint64_t size,
	uint64_t txg)
	{
	metaslab_t *msp;
	spa_t *spa = vd->vdev_spa;
	int error = 0;

	if (offset >> vd->vdev_ms_shift >= vd->vdev_ms_count)
	return (SET_ERROR(ENXIO));

	ASSERT3P(vd->vdev_ms, !=, NULL);
	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);

	if ((txg != 0 && spa_writeable(spa)) \|\| !msp->ms_loaded) {
	error = metaslab_activate(msp, 0, METASLAB_WEIGHT_CLAIM);
	if (error == EBUSY) {
	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	error = 0;
	}
	}

	if (error == 0 &&
	!range_tree_contains(msp->ms_allocatable, offset, size))
	error = SET_ERROR(ENOENT);

	if (error \|\| txg == 0) { /* txg == 0 indicates dry run */
	mutex_exit(&msp->ms_lock);
	return (error);
	}

	VERIFY(!msp->ms_condensing);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	VERIFY3U(range_tree_space(msp->ms_allocatable) - size, <=,
	msp->ms_size);
	range_tree_remove(msp->ms_allocatable, offset, size);
	range_tree_clear(msp->ms_trim, offset, size);

	if (spa_writeable(spa)) { /* don't dirty if we're zdb(8) */
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	msp->ms_selected_txg = txg;
	multilist_sublist_insert_head(mls, msp);
	}
	multilist_sublist_unlock(mls);

	if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
	vdev_dirty(vd, VDD_METASLAB, msp, txg);
	range_tree_add(msp->ms_allocating[txg & TXG_MASK],
	offset, size);
	msp->ms_allocating_total += size;
	}

	mutex_exit(&msp->ms_lock);

	return (0);
	}

	typedef struct metaslab_claim_cb_arg_t {
	uint64_t mcca_txg;
	int mcca_error;
	} metaslab_claim_cb_arg_t;

	static void
	metaslab_claim_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner_offset;
	metaslab_claim_cb_arg_t *mcca_arg = arg;

	if (mcca_arg->mcca_error == 0) {
	mcca_arg->mcca_error = metaslab_claim_concrete(vd, offset,
	size, mcca_arg->mcca_txg);
	}
	}

	int
	metaslab_claim_impl(vdev_t *vd, uint64_t offset, uint64_t size, uint64_t txg)
	{
	if (vd->vdev_ops->vdev_op_remap != NULL) {
	metaslab_claim_cb_arg_t arg;

	/*
	* Only zdb(8) can claim on indirect vdevs. This is used
	* to detect leaks of mapped space (that are not accounted
	* for in the obsolete counts, spacemap, or bpobj).
	*/
	ASSERT(!spa_writeable(vd->vdev_spa));
	arg.mcca_error = 0;
	arg.mcca_txg = txg;

	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_claim_impl_cb, &arg);

	if (arg.mcca_error == 0) {
	arg.mcca_error = metaslab_claim_concrete(vd,
	offset, size, txg);
	}
	return (arg.mcca_error);
	} else {
	return (metaslab_claim_concrete(vd, offset, size, txg));
	}
	}

	/*
	* Intent log support: upon opening the pool after a crash, notify the SPA
	* of blocks that the intent log has allocated for immediate write, but
	* which are still considered free by the SPA because the last transaction
	* group didn't commit yet.
	*/
	static int
	metaslab_claim_dva(spa_t spa, const dva_t dva, uint64_t txg)
	{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL) {
	return (SET_ERROR(ENXIO));
	}

	ASSERT(DVA_IS_VALID(dva));

	if (DVA_GET_GANG(dva))
	size = vdev_gang_header_asize(vd);

	return (metaslab_claim_impl(vd, offset, size, txg));
	}

	int
	metaslab_alloc(spa_t spa, metaslab_class_t mc, uint64_t psize, blkptr_t *bp,
	int ndvas, uint64_t txg, blkptr_t *hintbp, int flags,
	zio_alloc_list_t zal, zio_t zio, int allocator)
	{
	dva_t *dva = bp->blk_dva;
	dva_t *hintdva = (hintbp != NULL) ? hintbp->blk_dva : NULL;
	int error = 0;

	ASSERT(bp->blk_birth == 0);
	ASSERT(BP_PHYSICAL_BIRTH(bp) == 0);

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	if (mc->mc_allocator[allocator].mca_rotor == NULL) {
	/* no vdevs in this class */
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	return (SET_ERROR(ENOSPC));
	}

	ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
	ASSERT(BP_GET_NDVAS(bp) == 0);
	ASSERT(hintbp == NULL \|\| ndvas <= BP_GET_NDVAS(hintbp));
	ASSERT3P(zal, !=, NULL);

	for (int d = 0; d < ndvas; d++) {
	error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
	txg, flags, zal, allocator);
	if (error != 0) {
	for (d--; d >= 0; d--) {
	metaslab_unalloc_dva(spa, &dva[d], txg);
	metaslab_group_alloc_decrement(spa,
	DVA_GET_VDEV(&dva[d]), zio, flags,
	allocator, B_FALSE);
	memset(&dva[d], 0, sizeof (dva_t));
	}
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	return (error);
	} else {
	/*
	* Update the metaslab group's queue depth
	* based on the newly allocated dva.
	*/
	metaslab_group_alloc_increment(spa,
	DVA_GET_VDEV(&dva[d]), zio, flags, allocator);
	}
	}
	ASSERT(error == 0);
	ASSERT(BP_GET_NDVAS(bp) == ndvas);

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	BP_SET_BIRTH(bp, txg, 0);

	return (0);
	}

	void
	metaslab_free(spa_t spa, const blkptr_t bp, uint64_t txg, boolean_t now)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!now \|\| bp->blk_birth >= spa_syncing_txg(spa));

	/*
	* If we have a checkpoint for the pool we need to make sure that
	* the blocks that we free that are part of the checkpoint won't be
	* reused until the checkpoint is discarded or we revert to it.
	*
	* The checkpoint flag is passed down the metaslab_free code path
	* and is set whenever we want to add a block to the checkpoint's
	* accounting. That is, we "checkpoint" blocks that existed at the
	* time the checkpoint was created and are therefore referenced by
	* the checkpointed uberblock.
	*
	* Note that, we don't checkpoint any blocks if the current
	* syncing txg <= spa_checkpoint_txg. We want these frees to sync
	* normally as they will be referenced by the checkpointed uberblock.
	*/
	boolean_t checkpoint = B_FALSE;
	if (bp->blk_birth <= spa->spa_checkpoint_txg &&
	spa_syncing_txg(spa) > spa->spa_checkpoint_txg) {
	/*
	* At this point, if the block is part of the checkpoint
	* there is no way it was created in the current txg.
	*/
	ASSERT(!now);
	ASSERT3U(spa_syncing_txg(spa), ==, txg);
	checkpoint = B_TRUE;
	}

	spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++) {
	if (now) {
	metaslab_unalloc_dva(spa, &dva[d], txg);
	} else {
	ASSERT3U(txg, ==, spa_syncing_txg(spa));
	metaslab_free_dva(spa, &dva[d], checkpoint);
	}
	}

	spa_config_exit(spa, SCL_FREE, FTAG);
	}

	int
	metaslab_claim(spa_t spa, const blkptr_t bp, uint64_t txg)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	int error = 0;

	ASSERT(!BP_IS_HOLE(bp));

	if (txg != 0) {
	/*
	* First do a dry run to make sure all DVAs are claimable,
	* so we don't have to unwind from partial failures below.
	*/
	if ((error = metaslab_claim(spa, bp, 0)) != 0)
	return (error);
	}

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++) {
	error = metaslab_claim_dva(spa, &dva[d], txg);
	if (error != 0)
	break;
	}

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	ASSERT(error == 0 \|\| txg == 0);

	return (error);
	}

	void
	metaslab_fastwrite_mark(spa_t spa, const blkptr_t bp)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	int d;
	vdev_t *vd;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(psize > 0);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++) {
	if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
	continue;
	atomic_add_64(&vd->vdev_pending_fastwrite, psize);
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);
	}

	void
	metaslab_fastwrite_unmark(spa_t spa, const blkptr_t bp)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	int d;
	vdev_t *vd;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(psize > 0);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++) {
	if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
	continue;
	ASSERT3U(vd->vdev_pending_fastwrite, >=, psize);
	atomic_sub_64(&vd->vdev_pending_fastwrite, psize);
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);
	}

	static void
	metaslab_check_free_impl_cb(uint64_t inner, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner, (void) arg;

	if (vd->vdev_ops == &vdev_indirect_ops)
	return;

	metaslab_check_free_impl(vd, offset, size);
	}

	static void
	metaslab_check_free_impl(vdev_t *vd, uint64_t offset, uint64_t size)
	{
	metaslab_t *msp;
	spa_t *spa __maybe_unused = vd->vdev_spa;

	if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
	return;

	if (vd->vdev_ops->vdev_op_remap != NULL) {
	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_check_free_impl_cb, NULL);
	return;
	}

	ASSERT(vdev_is_concrete(vd));
	ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);
	if (msp->ms_loaded) {
	range_tree_verify_not_present(msp->ms_allocatable,
	offset, size);
	}

	/*
	* Check all segments that currently exist in the freeing pipeline.
	*
	* It would intuitively make sense to also check the current allocating
	* tree since metaslab_unalloc_dva() exists for extents that are
	* allocated and freed in the same sync pass within the same txg.
	* Unfortunately there are places (e.g. the ZIL) where we allocate a
	* segment but then we free part of it within the same txg
	* [see zil_sync()]. Thus, we don't call range_tree_verify() in the
	* current allocating tree.
	*/
	range_tree_verify_not_present(msp->ms_freeing, offset, size);
	range_tree_verify_not_present(msp->ms_checkpointing, offset, size);
	range_tree_verify_not_present(msp->ms_freed, offset, size);
	for (int j = 0; j < TXG_DEFER_SIZE; j++)
	range_tree_verify_not_present(msp->ms_defer[j], offset, size);
	range_tree_verify_not_present(msp->ms_trim, offset, size);
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_check_free(spa_t spa, const blkptr_t bp)
	{
	if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
	return;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
	uint64_t vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
	vdev_t *vd = vdev_lookup_top(spa, vdev);
	uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]);
	uint64_t size = DVA_GET_ASIZE(&bp->blk_dva[i]);

	if (DVA_GET_GANG(&bp->blk_dva[i]))
	size = vdev_gang_header_asize(vd);

	ASSERT3P(vd, !=, NULL);

	metaslab_check_free_impl(vd, offset, size);
	}
	spa_config_exit(spa, SCL_VDEV, FTAG);
	}

	static void
	metaslab_group_disable_wait(metaslab_group_t *mg)
	{
	ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
	while (mg->mg_disabled_updating) {
	cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
	}
	}

	static void
	metaslab_group_disabled_increment(metaslab_group_t *mg)
	{
	ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
	ASSERT(mg->mg_disabled_updating);

	while (mg->mg_ms_disabled >= max_disabled_ms) {
	cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
	}
	mg->mg_ms_disabled++;
	ASSERT3U(mg->mg_ms_disabled, <=, max_disabled_ms);
	}

	/*
	* Mark the metaslab as disabled to prevent any allocations on this metaslab.
	* We must also track how many metaslabs are currently disabled within a
	* metaslab group and limit them to prevent allocation failures from
	* occurring because all metaslabs are disabled.
	*/
	void
	metaslab_disable(metaslab_t *msp)
	{
	ASSERT(!MUTEX_HELD(&msp->ms_lock));
	metaslab_group_t *mg = msp->ms_group;

	mutex_enter(&mg->mg_ms_disabled_lock);

	/*
	* To keep an accurate count of how many threads have disabled
	* a specific metaslab group, we only allow one thread to mark
	* the metaslab group at a time. This ensures that the value of
	* ms_disabled will be accurate when we decide to mark a metaslab
	* group as disabled. To do this we force all other threads
	* to wait till the metaslab's mg_disabled_updating flag is no
	* longer set.
	*/
	metaslab_group_disable_wait(mg);
	mg->mg_disabled_updating = B_TRUE;
	if (msp->ms_disabled == 0) {
	metaslab_group_disabled_increment(mg);
	}
	mutex_enter(&msp->ms_lock);
	msp->ms_disabled++;
	mutex_exit(&msp->ms_lock);

	mg->mg_disabled_updating = B_FALSE;
	cv_broadcast(&mg->mg_ms_disabled_cv);
	mutex_exit(&mg->mg_ms_disabled_lock);
	}

	void
	metaslab_enable(metaslab_t *msp, boolean_t sync, boolean_t unload)
	{
	metaslab_group_t *mg = msp->ms_group;
	spa_t *spa = mg->mg_vd->vdev_spa;

	/*
	* Wait for the outstanding IO to be synced to prevent newly
	* allocated blocks from being overwritten. This used by
	* initialize and TRIM which are modifying unallocated space.
	*/
	if (sync)
	txg_wait_synced(spa_get_dsl(spa), 0);

	mutex_enter(&mg->mg_ms_disabled_lock);
	mutex_enter(&msp->ms_lock);
	if (--msp->ms_disabled == 0) {
	mg->mg_ms_disabled--;
	cv_broadcast(&mg->mg_ms_disabled_cv);
	if (unload)
	metaslab_unload(msp);
	}
	mutex_exit(&msp->ms_lock);
	mutex_exit(&mg->mg_ms_disabled_lock);
	}

	void
	metaslab_set_unflushed_dirty(metaslab_t *ms, boolean_t dirty)
	{
	ms->ms_unflushed_dirty = dirty;
	}

	static void
	metaslab_update_ondisk_flush_data(metaslab_t ms, dmu_tx_t tx)
	{
	vdev_t *vd = ms->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa_meta_objset(spa);

	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	metaslab_unflushed_phys_t entry = {
	.msp_unflushed_txg = metaslab_unflushed_txg(ms),
	};
	uint64_t entry_size = sizeof (entry);
	uint64_t entry_offset = ms->ms_id * entry_size;

	uint64_t object = 0;
	int err = zap_lookup(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
	&object);
	if (err == ENOENT) {
	object = dmu_object_alloc(mos, DMU_OTN_UINT64_METADATA,
	SPA_OLD_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
	VERIFY0(zap_add(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
	&object, tx));
	} else {
	VERIFY0(err);
	}

	dmu_write(spa_meta_objset(spa), object, entry_offset, entry_size,
	&entry, tx);
	}

	void
	metaslab_set_unflushed_txg(metaslab_t ms, uint64_t txg, dmu_tx_t tx)
	{
	ms->ms_unflushed_txg = txg;
	metaslab_update_ondisk_flush_data(ms, tx);
	}

	boolean_t
	metaslab_unflushed_dirty(metaslab_t *ms)
	{
	return (ms->ms_unflushed_dirty);
	}

	uint64_t
	metaslab_unflushed_txg(metaslab_t *ms)
	{
	return (ms->ms_unflushed_txg);
	}

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, aliquot, ULONG, ZMOD_RW,
	"Allocation granularity (a.k.a. stripe size)");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_load, INT, ZMOD_RW,
	"Load all metaslabs when pool is first opened");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_unload, INT, ZMOD_RW,
	"Prevent metaslabs from being unloaded");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_enabled, INT, ZMOD_RW,
	"Preload potential metaslabs during reassessment");

	-ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay, UINT, ZMOD_RW,
	"Delay in txgs after metaslab was last used before unloading");

	-ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay_ms, UINT, ZMOD_RW,
	"Delay in milliseconds after metaslab was last used before unloading");

	/* BEGIN CSTYLED */
	-ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, noalloc_threshold, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, noalloc_threshold, UINT, ZMOD_RW,
	"Percentage of metaslab group size that should be free to make it "
	"eligible for allocation");

	-ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, fragmentation_threshold, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, fragmentation_threshold, UINT, ZMOD_RW,
	"Percentage of metaslab group size that should be considered eligible "
	"for allocations unless all metaslab groups within the metaslab class "
	"have also crossed this threshold");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, fragmentation_factor_enabled, INT,
	ZMOD_RW,
	"Use the fragmentation metric to prefer less fragmented metaslabs");
	/* END CSTYLED */

	-ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, fragmentation_threshold, INT,
	+ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, fragmentation_threshold, UINT,
	ZMOD_RW, "Fragmentation for metaslab to allow allocation");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, lba_weighting_enabled, INT, ZMOD_RW,
	"Prefer metaslabs with lower LBAs");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, bias_enabled, INT, ZMOD_RW,
	"Enable metaslab group biasing");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, segment_weight_enabled, INT,
	ZMOD_RW, "Enable segment-based metaslab selection");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, switch_threshold, INT, ZMOD_RW,
	"Segment-based metaslab selection maximum buckets before switching");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging, ULONG, ZMOD_RW,
	"Blocks larger than this size are forced to be gang blocks");

	-ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_max_search, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_max_search, UINT, ZMOD_RW,
	"Max distance (bytes) to search forward before using size tree");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_use_largest_segment, INT, ZMOD_RW,
	"When looking in size tree, use largest segment instead of exact fit");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, max_size_cache_sec, ULONG,
	ZMOD_RW, "How long to trust the cached max chunk size of a metaslab");

	-ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, mem_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, mem_limit, UINT, ZMOD_RW,
	"Percentage of memory that can be used to store metaslab range trees");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, try_hard_before_gang, INT,
	ZMOD_RW, "Try hard to allocate before ganging");

	-ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, find_max_tries, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, find_max_tries, UINT, ZMOD_RW,
	"Normally only consider this many of the best metaslabs in each vdev");
	diff --git a/sys/contrib/openzfs/module/zfs/multilist.c b/sys/contrib/openzfs/module/zfs/multilist.c
	index fdc5f07c47ea..b1cdf1c5c5f4 100644
	--- a/sys/contrib/openzfs/module/zfs/multilist.c
	+++ b/sys/contrib/openzfs/module/zfs/multilist.c
	@@ -1,429 +1,429 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/multilist.h>
	#include <sys/trace_zfs.h>

	/*
	* This overrides the number of sublists in each multilist_t, which defaults
	* to the number of CPUs in the system (see multilist_create()).
	*/
	-int zfs_multilist_num_sublists = 0;
	+uint_t zfs_multilist_num_sublists = 0;

	/*
	* Given the object contained on the list, return a pointer to the
	* object's multilist_node_t structure it contains.
	*/
	#ifdef ZFS_DEBUG
	static multilist_node_t *
	multilist_d2l(multilist_t ml, void obj)
	{
	return ((multilist_node_t )((char )obj + ml->ml_offset));
	}
	#else
	#define multilist_d2l(ml, obj) ((void) sizeof (ml), (void) sizeof (obj), NULL)
	#endif

	/*
	* Initialize a new mutlilist using the parameters specified.
	*
	* - 'size' denotes the size of the structure containing the
	* multilist_node_t.
	* - 'offset' denotes the byte offset of the mutlilist_node_t within
	* the structure that contains it.
	* - 'num' specifies the number of internal sublists to create.
	* - 'index_func' is used to determine which sublist to insert into
	* when the multilist_insert() function is called; as well as which
	* sublist to remove from when multilist_remove() is called. The
	* requirements this function must meet, are the following:
	*
	* - It must always return the same value when called on the same
	* object (to ensure the object is removed from the list it was
	* inserted into).
	*
	* - It must return a value in the range [0, number of sublists).
	* The multilist_get_num_sublists() function may be used to
	* determine the number of sublists in the multilist.
	*
	* Also, in order to reduce internal contention between the sublists
	* during insertion and removal, this function should choose evenly
	* between all available sublists when inserting. This isn't a hard
	* requirement, but a general rule of thumb in order to garner the
	* best multi-threaded performance out of the data structure.
	*/
	static void
	multilist_create_impl(multilist_t *ml, size_t size, size_t offset,
	- unsigned int num, multilist_sublist_index_func_t *index_func)
	+ uint_t num, multilist_sublist_index_func_t *index_func)
	{
	ASSERT3U(size, >, 0);
	ASSERT3U(size, >=, offset + sizeof (multilist_node_t));
	ASSERT3U(num, >, 0);
	ASSERT3P(index_func, !=, NULL);

	ml->ml_offset = offset;
	ml->ml_num_sublists = num;
	ml->ml_index_func = index_func;

	ml->ml_sublists = kmem_zalloc(sizeof (multilist_sublist_t) *
	ml->ml_num_sublists, KM_SLEEP);

	ASSERT3P(ml->ml_sublists, !=, NULL);

	for (int i = 0; i < ml->ml_num_sublists; i++) {
	multilist_sublist_t *mls = &ml->ml_sublists[i];
	mutex_init(&mls->mls_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	list_create(&mls->mls_list, size, offset);
	}
	}

	/*
	* Allocate a new multilist, using the default number of sublists (the number
	* of CPUs, or at least 4, or the tunable zfs_multilist_num_sublists). Note
	* that the multilists do not expand if more CPUs are hot-added. In that case,
	* we will have less fanout than boot_ncpus, but we don't want to always
	* reserve the RAM necessary to create the extra slots for additional CPUs up
	* front, and dynamically adding them is a complex task.
	*/
	void
	multilist_create(multilist_t *ml, size_t size, size_t offset,
	multilist_sublist_index_func_t *index_func)
	{
	- int num_sublists;
	+ uint_t num_sublists;

	if (zfs_multilist_num_sublists > 0) {
	num_sublists = zfs_multilist_num_sublists;
	} else {
	num_sublists = MAX(boot_ncpus, 4);
	}

	multilist_create_impl(ml, size, offset, num_sublists, index_func);
	}

	/*
	* Destroy the given multilist object, and free up any memory it holds.
	*/
	void
	multilist_destroy(multilist_t *ml)
	{
	ASSERT(multilist_is_empty(ml));

	for (int i = 0; i < ml->ml_num_sublists; i++) {
	multilist_sublist_t *mls = &ml->ml_sublists[i];

	ASSERT(list_is_empty(&mls->mls_list));

	list_destroy(&mls->mls_list);
	mutex_destroy(&mls->mls_lock);
	}

	ASSERT3P(ml->ml_sublists, !=, NULL);
	kmem_free(ml->ml_sublists,
	sizeof (multilist_sublist_t) * ml->ml_num_sublists);

	ml->ml_num_sublists = 0;
	ml->ml_offset = 0;
	ml->ml_sublists = NULL;
	}

	/*
	* Insert the given object into the multilist.
	*
	* This function will insert the object specified into the sublist
	* determined using the function given at multilist creation time.
	*
	* The sublist locks are automatically acquired if not already held, to
	* ensure consistency when inserting and removing from multiple threads.
	*/
	void
	multilist_insert(multilist_t ml, void obj)
	{
	unsigned int sublist_idx = ml->ml_index_func(ml, obj);
	multilist_sublist_t *mls;
	boolean_t need_lock;

	DTRACE_PROBE3(multilist__insert, multilist_t *, ml,
	unsigned int, sublist_idx, void *, obj);

	ASSERT3U(sublist_idx, <, ml->ml_num_sublists);

	mls = &ml->ml_sublists[sublist_idx];

	/*
	* Note: Callers may already hold the sublist lock by calling
	* multilist_sublist_lock(). Here we rely on MUTEX_HELD()
	* returning TRUE if and only if the current thread holds the
	* lock. While it's a little ugly to make the lock recursive in
	* this way, it works and allows the calling code to be much
	* simpler -- otherwise it would have to pass around a flag
	* indicating that it already has the lock.
	*/
	need_lock = !MUTEX_HELD(&mls->mls_lock);

	if (need_lock)
	mutex_enter(&mls->mls_lock);

	ASSERT(!multilist_link_active(multilist_d2l(ml, obj)));

	multilist_sublist_insert_head(mls, obj);

	if (need_lock)
	mutex_exit(&mls->mls_lock);
	}

	/*
	* Remove the given object from the multilist.
	*
	* This function will remove the object specified from the sublist
	* determined using the function given at multilist creation time.
	*
	* The necessary sublist locks are automatically acquired, to ensure
	* consistency when inserting and removing from multiple threads.
	*/
	void
	multilist_remove(multilist_t ml, void obj)
	{
	unsigned int sublist_idx = ml->ml_index_func(ml, obj);
	multilist_sublist_t *mls;
	boolean_t need_lock;

	DTRACE_PROBE3(multilist__remove, multilist_t *, ml,
	unsigned int, sublist_idx, void *, obj);

	ASSERT3U(sublist_idx, <, ml->ml_num_sublists);

	mls = &ml->ml_sublists[sublist_idx];
	/* See comment in multilist_insert(). */
	need_lock = !MUTEX_HELD(&mls->mls_lock);

	if (need_lock)
	mutex_enter(&mls->mls_lock);

	ASSERT(multilist_link_active(multilist_d2l(ml, obj)));

	multilist_sublist_remove(mls, obj);

	if (need_lock)
	mutex_exit(&mls->mls_lock);
	}

	/*
	* Check to see if this multilist object is empty.
	*
	* This will return TRUE if it finds all of the sublists of this
	* multilist to be empty, and FALSE otherwise. Each sublist lock will be
	* automatically acquired as necessary.
	*
	* If concurrent insertions and removals are occurring, the semantics
	* of this function become a little fuzzy. Instead of locking all
	* sublists for the entire call time of the function, each sublist is
	* only locked as it is individually checked for emptiness. Thus, it's
	* possible for this function to return TRUE with non-empty sublists at
	* the time the function returns. This would be due to another thread
	* inserting into a given sublist, after that specific sublist was check
	* and deemed empty, but before all sublists have been checked.
	*/
	int
	multilist_is_empty(multilist_t *ml)
	{
	for (int i = 0; i < ml->ml_num_sublists; i++) {
	multilist_sublist_t *mls = &ml->ml_sublists[i];
	/* See comment in multilist_insert(). */
	boolean_t need_lock = !MUTEX_HELD(&mls->mls_lock);

	if (need_lock)
	mutex_enter(&mls->mls_lock);

	if (!list_is_empty(&mls->mls_list)) {
	if (need_lock)
	mutex_exit(&mls->mls_lock);

	return (FALSE);
	}

	if (need_lock)
	mutex_exit(&mls->mls_lock);
	}

	return (TRUE);
	}

	/* Return the number of sublists composing this multilist */
	unsigned int
	multilist_get_num_sublists(multilist_t *ml)
	{
	return (ml->ml_num_sublists);
	}

	/* Return a randomly selected, valid sublist index for this multilist */
	unsigned int
	multilist_get_random_index(multilist_t *ml)
	{
	return (random_in_range(ml->ml_num_sublists));
	}

	/* Lock and return the sublist specified at the given index */
	multilist_sublist_t *
	multilist_sublist_lock(multilist_t *ml, unsigned int sublist_idx)
	{
	multilist_sublist_t *mls;

	ASSERT3U(sublist_idx, <, ml->ml_num_sublists);
	mls = &ml->ml_sublists[sublist_idx];
	mutex_enter(&mls->mls_lock);

	return (mls);
	}

	/* Lock and return the sublist that would be used to store the specified obj */
	multilist_sublist_t *
	multilist_sublist_lock_obj(multilist_t ml, void obj)
	{
	return (multilist_sublist_lock(ml, ml->ml_index_func(ml, obj)));
	}

	void
	multilist_sublist_unlock(multilist_sublist_t *mls)
	{
	mutex_exit(&mls->mls_lock);
	}

	/*
	* We're allowing any object to be inserted into this specific sublist,
	* but this can lead to trouble if multilist_remove() is called to
	* remove this object. Specifically, if calling ml_index_func on this
	* object returns an index for sublist different than what is passed as
	* a parameter here, any call to multilist_remove() with this newly
	* inserted object is undefined! (the call to multilist_remove() will
	* remove the object from a list that it isn't contained in)
	*/
	void
	multilist_sublist_insert_head(multilist_sublist_t mls, void obj)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	list_insert_head(&mls->mls_list, obj);
	}

	/* please see comment above multilist_sublist_insert_head */
	void
	multilist_sublist_insert_tail(multilist_sublist_t mls, void obj)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	list_insert_tail(&mls->mls_list, obj);
	}

	/*
	* Move the object one element forward in the list.
	*
	* This function will move the given object forward in the list (towards
	* the head) by one object. So, in essence, it will swap its position in
	* the list with its "prev" pointer. If the given object is already at the
	* head of the list, it cannot be moved forward any more than it already
	* is, so no action is taken.
	*
	* NOTE: This function must not remove any object from the list other
	* than the object given as the parameter. This is relied upon in
	* arc_evict_state_impl().
	*/
	void
	multilist_sublist_move_forward(multilist_sublist_t mls, void obj)
	{
	void *prev = list_prev(&mls->mls_list, obj);

	ASSERT(MUTEX_HELD(&mls->mls_lock));
	ASSERT(!list_is_empty(&mls->mls_list));

	/* 'obj' must be at the head of the list, nothing to do */
	if (prev == NULL)
	return;

	list_remove(&mls->mls_list, obj);
	list_insert_before(&mls->mls_list, prev, obj);
	}

	void
	multilist_sublist_remove(multilist_sublist_t mls, void obj)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	list_remove(&mls->mls_list, obj);
	}

	int
	multilist_sublist_is_empty(multilist_sublist_t *mls)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	return (list_is_empty(&mls->mls_list));
	}

	int
	multilist_sublist_is_empty_idx(multilist_t *ml, unsigned int sublist_idx)
	{
	multilist_sublist_t *mls;
	int empty;

	ASSERT3U(sublist_idx, <, ml->ml_num_sublists);
	mls = &ml->ml_sublists[sublist_idx];
	ASSERT(!MUTEX_HELD(&mls->mls_lock));
	mutex_enter(&mls->mls_lock);
	empty = list_is_empty(&mls->mls_list);
	mutex_exit(&mls->mls_lock);
	return (empty);
	}

	void *
	multilist_sublist_head(multilist_sublist_t *mls)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	return (list_head(&mls->mls_list));
	}

	void *
	multilist_sublist_tail(multilist_sublist_t *mls)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	return (list_tail(&mls->mls_list));
	}

	void *
	multilist_sublist_next(multilist_sublist_t mls, void obj)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	return (list_next(&mls->mls_list, obj));
	}

	void *
	multilist_sublist_prev(multilist_sublist_t mls, void obj)
	{
	ASSERT(MUTEX_HELD(&mls->mls_lock));
	return (list_prev(&mls->mls_list, obj));
	}

	void
	multilist_link_init(multilist_node_t *link)
	{
	list_link_init(link);
	}

	int
	multilist_link_active(multilist_node_t *link)
	{
	return (list_link_active(link));
	}

	-ZFS_MODULE_PARAM(zfs, zfs_, multilist_num_sublists, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, multilist_num_sublists, UINT, ZMOD_RW,
	"Number of sublists used in each multilist");
	diff --git a/sys/contrib/openzfs/module/zfs/refcount.c b/sys/contrib/openzfs/module/zfs/refcount.c
	index b215df98d7a8..62ec03e1035a 100644
	--- a/sys/contrib/openzfs/module/zfs/refcount.c
	+++ b/sys/contrib/openzfs/module/zfs/refcount.c
	@@ -1,336 +1,335 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2021 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/zfs_refcount.h>

	#ifdef ZFS_DEBUG
	/*
	* Reference count tracking is disabled by default. It's memory requirements
	* are reasonable, however as implemented it consumes a significant amount of
	* cpu time. Until its performance is improved it should be manually enabled.
	*/
	int reference_tracking_enable = B_FALSE;
	-static int reference_history = 3; /* tunable */
	+static uint_t reference_history = 3; /* tunable */

	static kmem_cache_t *reference_cache;
	static kmem_cache_t *reference_history_cache;

	void
	zfs_refcount_init(void)
	{
	reference_cache = kmem_cache_create("reference_cache",
	sizeof (reference_t), 0, NULL, NULL, NULL, NULL, NULL, 0);

	reference_history_cache = kmem_cache_create("reference_history_cache",
	sizeof (uint64_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
	}

	void
	zfs_refcount_fini(void)
	{
	kmem_cache_destroy(reference_cache);
	kmem_cache_destroy(reference_history_cache);
	}

	void
	zfs_refcount_create(zfs_refcount_t *rc)
	{
	mutex_init(&rc->rc_mtx, NULL, MUTEX_DEFAULT, NULL);
	list_create(&rc->rc_list, sizeof (reference_t),
	offsetof(reference_t, ref_link));
	list_create(&rc->rc_removed, sizeof (reference_t),
	offsetof(reference_t, ref_link));
	rc->rc_count = 0;
	rc->rc_removed_count = 0;
	rc->rc_tracked = reference_tracking_enable;
	}

	void
	zfs_refcount_create_tracked(zfs_refcount_t *rc)
	{
	zfs_refcount_create(rc);
	rc->rc_tracked = B_TRUE;
	}

	void
	zfs_refcount_create_untracked(zfs_refcount_t *rc)
	{
	zfs_refcount_create(rc);
	rc->rc_tracked = B_FALSE;
	}

	void
	zfs_refcount_destroy_many(zfs_refcount_t *rc, uint64_t number)
	{
	reference_t *ref;

	ASSERT3U(rc->rc_count, ==, number);
	while ((ref = list_head(&rc->rc_list))) {
	list_remove(&rc->rc_list, ref);
	kmem_cache_free(reference_cache, ref);
	}
	list_destroy(&rc->rc_list);

	while ((ref = list_head(&rc->rc_removed))) {
	list_remove(&rc->rc_removed, ref);
	kmem_cache_free(reference_history_cache, ref->ref_removed);
	kmem_cache_free(reference_cache, ref);
	}
	list_destroy(&rc->rc_removed);
	mutex_destroy(&rc->rc_mtx);
	}

	void
	zfs_refcount_destroy(zfs_refcount_t *rc)
	{
	zfs_refcount_destroy_many(rc, 0);
	}

	int
	zfs_refcount_is_zero(zfs_refcount_t *rc)
	{
	return (zfs_refcount_count(rc) == 0);
	}

	int64_t
	zfs_refcount_count(zfs_refcount_t *rc)
	{
	return (atomic_load_64(&rc->rc_count));
	}

	int64_t
	zfs_refcount_add_many(zfs_refcount_t rc, uint64_t number, const void holder)
	{
	reference_t *ref = NULL;
	int64_t count;

	if (!rc->rc_tracked) {
	count = atomic_add_64_nv(&(rc)->rc_count, number);
	ASSERT3U(count, >=, number);
	return (count);
	}

	ref = kmem_cache_alloc(reference_cache, KM_SLEEP);
	ref->ref_holder = holder;
	ref->ref_number = number;
	mutex_enter(&rc->rc_mtx);
	- ASSERT3U(rc->rc_count, >=, 0);
	list_insert_head(&rc->rc_list, ref);
	rc->rc_count += number;
	count = rc->rc_count;
	mutex_exit(&rc->rc_mtx);

	return (count);
	}

	int64_t
	zfs_refcount_add(zfs_refcount_t rc, const void holder)
	{
	return (zfs_refcount_add_many(rc, 1, holder));
	}

	int64_t
	zfs_refcount_remove_many(zfs_refcount_t *rc, uint64_t number,
	const void *holder)
	{
	reference_t *ref;
	int64_t count;

	if (!rc->rc_tracked) {
	count = atomic_add_64_nv(&(rc)->rc_count, -number);
	ASSERT3S(count, >=, 0);
	return (count);
	}

	mutex_enter(&rc->rc_mtx);
	ASSERT3U(rc->rc_count, >=, number);
	for (ref = list_head(&rc->rc_list); ref;
	ref = list_next(&rc->rc_list, ref)) {
	if (ref->ref_holder == holder && ref->ref_number == number) {
	list_remove(&rc->rc_list, ref);
	if (reference_history > 0) {
	ref->ref_removed =
	kmem_cache_alloc(reference_history_cache,
	KM_SLEEP);
	list_insert_head(&rc->rc_removed, ref);
	rc->rc_removed_count++;
	if (rc->rc_removed_count > reference_history) {
	ref = list_tail(&rc->rc_removed);
	list_remove(&rc->rc_removed, ref);
	kmem_cache_free(reference_history_cache,
	ref->ref_removed);
	kmem_cache_free(reference_cache, ref);
	rc->rc_removed_count--;
	}
	} else {
	kmem_cache_free(reference_cache, ref);
	}
	rc->rc_count -= number;
	count = rc->rc_count;
	mutex_exit(&rc->rc_mtx);
	return (count);
	}
	}
	panic("No such hold %p on refcount %llx", holder,
	(u_longlong_t)(uintptr_t)rc);
	return (-1);
	}

	int64_t
	zfs_refcount_remove(zfs_refcount_t rc, const void holder)
	{
	return (zfs_refcount_remove_many(rc, 1, holder));
	}

	void
	zfs_refcount_transfer(zfs_refcount_t dst, zfs_refcount_t src)
	{
	int64_t count, removed_count;
	list_t list, removed;

	list_create(&list, sizeof (reference_t),
	offsetof(reference_t, ref_link));
	list_create(&removed, sizeof (reference_t),
	offsetof(reference_t, ref_link));

	mutex_enter(&src->rc_mtx);
	count = src->rc_count;
	removed_count = src->rc_removed_count;
	src->rc_count = 0;
	src->rc_removed_count = 0;
	list_move_tail(&list, &src->rc_list);
	list_move_tail(&removed, &src->rc_removed);
	mutex_exit(&src->rc_mtx);

	mutex_enter(&dst->rc_mtx);
	dst->rc_count += count;
	dst->rc_removed_count += removed_count;
	list_move_tail(&dst->rc_list, &list);
	list_move_tail(&dst->rc_removed, &removed);
	mutex_exit(&dst->rc_mtx);

	list_destroy(&list);
	list_destroy(&removed);
	}

	void
	zfs_refcount_transfer_ownership_many(zfs_refcount_t *rc, uint64_t number,
	const void current_holder, const void new_holder)
	{
	reference_t *ref;
	boolean_t found = B_FALSE;

	if (!rc->rc_tracked)
	return;

	mutex_enter(&rc->rc_mtx);
	for (ref = list_head(&rc->rc_list); ref;
	ref = list_next(&rc->rc_list, ref)) {
	if (ref->ref_holder == current_holder &&
	ref->ref_number == number) {
	ref->ref_holder = new_holder;
	found = B_TRUE;
	break;
	}
	}
	ASSERT(found);
	mutex_exit(&rc->rc_mtx);
	}

	void
	zfs_refcount_transfer_ownership(zfs_refcount_t rc, const void current_holder,
	const void *new_holder)
	{
	return (zfs_refcount_transfer_ownership_many(rc, 1, current_holder,
	new_holder));
	}

	/*
	* If tracking is enabled, return true if a reference exists that matches
	* the "holder" tag. If tracking is disabled, then return true if a reference
	* might be held.
	*/
	boolean_t
	zfs_refcount_held(zfs_refcount_t rc, const void holder)
	{
	reference_t *ref;

	if (!rc->rc_tracked)
	return (zfs_refcount_count(rc) > 0);

	mutex_enter(&rc->rc_mtx);
	for (ref = list_head(&rc->rc_list); ref;
	ref = list_next(&rc->rc_list, ref)) {
	if (ref->ref_holder == holder) {
	mutex_exit(&rc->rc_mtx);
	return (B_TRUE);
	}
	}
	mutex_exit(&rc->rc_mtx);
	return (B_FALSE);
	}

	/*
	* If tracking is enabled, return true if a reference does not exist that
	* matches the "holder" tag. If tracking is disabled, always return true
	* since the reference might not be held.
	*/
	boolean_t
	zfs_refcount_not_held(zfs_refcount_t rc, const void holder)
	{
	reference_t *ref;

	if (!rc->rc_tracked)
	return (B_TRUE);

	mutex_enter(&rc->rc_mtx);
	for (ref = list_head(&rc->rc_list); ref;
	ref = list_next(&rc->rc_list, ref)) {
	if (ref->ref_holder == holder) {
	mutex_exit(&rc->rc_mtx);
	return (B_FALSE);
	}
	}
	mutex_exit(&rc->rc_mtx);
	return (B_TRUE);
	}

	EXPORT_SYMBOL(zfs_refcount_create);
	EXPORT_SYMBOL(zfs_refcount_destroy);
	EXPORT_SYMBOL(zfs_refcount_is_zero);
	EXPORT_SYMBOL(zfs_refcount_count);
	EXPORT_SYMBOL(zfs_refcount_add);
	EXPORT_SYMBOL(zfs_refcount_remove);
	EXPORT_SYMBOL(zfs_refcount_held);

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, , reference_tracking_enable, INT, ZMOD_RW,
	"Track reference holders to refcount_t objects");

	-ZFS_MODULE_PARAM(zfs, , reference_history, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, , reference_history, UINT, ZMOD_RW,
	"Maximum reference holders being tracked");
	/* END CSTYLED */
	#endif /* ZFS_DEBUG */
	diff --git a/sys/contrib/openzfs/module/zfs/sa.c b/sys/contrib/openzfs/module/zfs/sa.c
	index c094a8f0730a..5568a24b84fe 100644
	--- a/sys/contrib/openzfs/module/zfs/sa.c
	+++ b/sys/contrib/openzfs/module/zfs/sa.c
	@@ -1,2257 +1,2257 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/sysmacros.h>
	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zap.h>
	#include <sys/sa.h>
	#include <sys/sunddi.h>
	#include <sys/sa_impl.h>
	#include <sys/errno.h>
	#include <sys/zfs_context.h>

	#ifdef _KERNEL
	#include <sys/zfs_znode.h>
	#endif

	/*
	* ZFS System attributes:
	*
	* A generic mechanism to allow for arbitrary attributes
	* to be stored in a dnode. The data will be stored in the bonus buffer of
	* the dnode and if necessary a special "spill" block will be used to handle
	* overflow situations. The spill block will be sized to fit the data
	* from 512 - 128K. When a spill block is used the BP (blkptr_t) for the
	* spill block is stored at the end of the current bonus buffer. Any
	* attributes that would be in the way of the blkptr_t will be relocated
	* into the spill block.
	*
	* Attribute registration:
	*
	* Stored persistently on a per dataset basis
	* a mapping between attribute "string" names and their actual attribute
	* numeric values, length, and byteswap function. The names are only used
	* during registration. All attributes are known by their unique attribute
	* id value. If an attribute can have a variable size then the value
	* 0 will be used to indicate this.
	*
	* Attribute Layout:
	*
	* Attribute layouts are a way to compactly store multiple attributes, but
	* without taking the overhead associated with managing each attribute
	* individually. Since you will typically have the same set of attributes
	* stored in the same order a single table will be used to represent that
	* layout. The ZPL for example will usually have only about 10 different
	* layouts (regular files, device files, symlinks,
	* regular files + scanstamp, files/dir with extended attributes, and then
	* you have the possibility of all of those minus ACL, because it would
	* be kicked out into the spill block)
	*
	* Layouts are simply an array of the attributes and their
	* ordering i.e. [0, 1, 4, 5, 2]
	*
	* Each distinct layout is given a unique layout number and that is what's
	* stored in the header at the beginning of the SA data buffer.
	*
	* A layout only covers a single dbuf (bonus or spill). If a set of
	* attributes is split up between the bonus buffer and a spill buffer then
	* two different layouts will be used. This allows us to byteswap the
	* spill without looking at the bonus buffer and keeps the on disk format of
	* the bonus and spill buffer the same.
	*
	* Adding a single attribute will cause the entire set of attributes to
	* be rewritten and could result in a new layout number being constructed
	* as part of the rewrite if no such layout exists for the new set of
	* attributes. The new attribute will be appended to the end of the already
	* existing attributes.
	*
	* Both the attribute registration and attribute layout information are
	* stored in normal ZAP attributes. Their should be a small number of
	* known layouts and the set of attributes is assumed to typically be quite
	* small.
	*
	* The registered attributes and layout "table" information is maintained
	* in core and a special "sa_os_t" is attached to the objset_t.
	*
	* A special interface is provided to allow for quickly applying
	* a large set of attributes at once. sa_replace_all_by_template() is
	* used to set an array of attributes. This is used by the ZPL when
	* creating a brand new file. The template that is passed into the function
	* specifies the attribute, size for variable length attributes, location of
	* data and special "data locator" function if the data isn't in a contiguous
	* location.
	*
	* Byteswap implications:
	*
	* Since the SA attributes are not entirely self describing we can't do
	* the normal byteswap processing. The special ZAP layout attribute and
	* attribute registration attributes define the byteswap function and the
	* size of the attributes, unless it is variable sized.
	* The normal ZFS byteswapping infrastructure assumes you don't need
	* to read any objects in order to do the necessary byteswapping. Whereas
	* SA attributes can only be properly byteswapped if the dataset is opened
	* and the layout/attribute ZAP attributes are available. Because of this
	* the SA attributes will be byteswapped when they are first accessed by
	* the SA code that will read the SA data.
	*/

	typedef void (sa_iterfunc_t)(void hdr, void addr, sa_attr_type_t,
	uint16_t length, int length_idx, boolean_t, void *userp);

	static int sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype);
	static void sa_idx_tab_hold(objset_t os, sa_idx_tab_t idx_tab);
	static sa_idx_tab_t sa_find_idx_tab(objset_t os, dmu_object_type_t bonustype,
	sa_hdr_phys_t *hdr);
	static void sa_idx_tab_rele(objset_t os, void arg);
	static void sa_copy_data(sa_data_locator_t func, void start, void *target,
	int buflen);
	static int sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
	sa_data_op_t action, sa_data_locator_t locator, void datastart,
	uint16_t buflen, dmu_tx_t *tx);

	-static const arc_byteswap_func_t sa_bswap_table[] = {
	+static arc_byteswap_func_t sa_bswap_table[] = {
	byteswap_uint64_array,
	byteswap_uint32_array,
	byteswap_uint16_array,
	byteswap_uint8_array,
	zfs_acl_byteswap,
	};

	#ifdef HAVE_EFFICIENT_UNALIGNED_ACCESS
	#define SA_COPY_DATA(f, s, t, l) \
	do { \
	if (f == NULL) { \
	if (l == 8) { \
	(uint64_t )t = (uint64_t )s; \
	} else if (l == 16) { \
	(uint64_t )t = (uint64_t )s; \
	(uint64_t )((uintptr_t)t + 8) = \
	(uint64_t )((uintptr_t)s + 8); \
	} else { \
	memcpy(t, s, l); \
	} \
	} else { \
	sa_copy_data(f, s, t, l); \
	} \
	} while (0)
	#else
	#define SA_COPY_DATA(f, s, t, l) sa_copy_data(f, s, t, l)
	#endif

	/*
	* This table is fixed and cannot be changed. Its purpose is to
	* allow the SA code to work with both old/new ZPL file systems.
	* It contains the list of legacy attributes. These attributes aren't
	* stored in the "attribute" registry zap objects, since older ZPL file systems
	* won't have the registry. Only objsets of type ZFS_TYPE_FILESYSTEM will
	* use this static table.
	*/
	static const sa_attr_reg_t sa_legacy_attrs[] = {
	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
	};

	/*
	* This is only used for objects of type DMU_OT_ZNODE
	*/
	static const sa_attr_type_t sa_legacy_zpl_layout[] = {
	0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
	};

	/*
	* Special dummy layout used for buffers with no attributes.
	*/
	static const sa_attr_type_t sa_dummy_zpl_layout[] = { 0 };

	static const size_t sa_legacy_attr_count = ARRAY_SIZE(sa_legacy_attrs);
	static kmem_cache_t *sa_cache = NULL;

	static int
	sa_cache_constructor(void buf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	sa_handle_t *hdl = buf;

	mutex_init(&hdl->sa_lock, NULL, MUTEX_DEFAULT, NULL);
	return (0);
	}

	static void
	sa_cache_destructor(void buf, void unused)
	{
	(void) unused;
	sa_handle_t *hdl = buf;
	mutex_destroy(&hdl->sa_lock);
	}

	void
	sa_cache_init(void)
	{
	sa_cache = kmem_cache_create("sa_cache",
	sizeof (sa_handle_t), 0, sa_cache_constructor,
	sa_cache_destructor, NULL, NULL, NULL, 0);
	}

	void
	sa_cache_fini(void)
	{
	if (sa_cache)
	kmem_cache_destroy(sa_cache);
	}

	static int
	layout_num_compare(const void arg1, const void arg2)
	{
	const sa_lot_t node1 = (const sa_lot_t )arg1;
	const sa_lot_t node2 = (const sa_lot_t )arg2;

	return (TREE_CMP(node1->lot_num, node2->lot_num));
	}

	static int
	layout_hash_compare(const void arg1, const void arg2)
	{
	const sa_lot_t node1 = (const sa_lot_t )arg1;
	const sa_lot_t node2 = (const sa_lot_t )arg2;

	int cmp = TREE_CMP(node1->lot_hash, node2->lot_hash);
	if (likely(cmp))
	return (cmp);

	return (TREE_CMP(node1->lot_instance, node2->lot_instance));
	}

	static boolean_t
	sa_layout_equal(sa_lot_t tbf, sa_attr_type_t attrs, int count)
	{
	int i;

	if (count != tbf->lot_attr_count)
	return (1);

	for (i = 0; i != count; i++) {
	if (attrs[i] != tbf->lot_attrs[i])
	return (1);
	}
	return (0);
	}

	#define SA_ATTR_HASH(attr) (zfs_crc64_table[(-1ULL ^ attr) & 0xFF])

	static uint64_t
	sa_layout_info_hash(const sa_attr_type_t *attrs, int attr_count)
	{
	uint64_t crc = -1ULL;

	for (int i = 0; i != attr_count; i++)
	crc ^= SA_ATTR_HASH(attrs[i]);

	return (crc);
	}

	static int
	sa_get_spill(sa_handle_t *hdl)
	{
	int rc;
	if (hdl->sa_spill == NULL) {
	if ((rc = dmu_spill_hold_existing(hdl->sa_bonus, NULL,
	&hdl->sa_spill)) == 0)
	VERIFY(0 == sa_build_index(hdl, SA_SPILL));
	} else {
	rc = 0;
	}

	return (rc);
	}

	/*
	* Main attribute lookup/update function
	* returns 0 for success or non zero for failures
	*
	* Operates on bulk array, first failure will abort further processing
	*/
	static int
	sa_attr_op(sa_handle_t hdl, sa_bulk_attr_t bulk, int count,
	sa_data_op_t data_op, dmu_tx_t *tx)
	{
	sa_os_t *sa = hdl->sa_os->os_sa;
	int i;
	int error = 0;
	sa_buf_type_t buftypes;

	buftypes = 0;

	ASSERT(count > 0);
	for (i = 0; i != count; i++) {
	ASSERT(bulk[i].sa_attr <= hdl->sa_os->os_sa->sa_num_attrs);

	bulk[i].sa_addr = NULL;
	/* First check the bonus buffer */

	if (hdl->sa_bonus_tab && TOC_ATTR_PRESENT(
	hdl->sa_bonus_tab->sa_idx_tab[bulk[i].sa_attr])) {
	SA_ATTR_INFO(sa, hdl->sa_bonus_tab,
	SA_GET_HDR(hdl, SA_BONUS),
	bulk[i].sa_attr, bulk[i], SA_BONUS, hdl);
	if (tx && !(buftypes & SA_BONUS)) {
	dmu_buf_will_dirty(hdl->sa_bonus, tx);
	buftypes \|= SA_BONUS;
	}
	}
	if (bulk[i].sa_addr == NULL &&
	((error = sa_get_spill(hdl)) == 0)) {
	if (TOC_ATTR_PRESENT(
	hdl->sa_spill_tab->sa_idx_tab[bulk[i].sa_attr])) {
	SA_ATTR_INFO(sa, hdl->sa_spill_tab,
	SA_GET_HDR(hdl, SA_SPILL),
	bulk[i].sa_attr, bulk[i], SA_SPILL, hdl);
	if (tx && !(buftypes & SA_SPILL) &&
	bulk[i].sa_size == bulk[i].sa_length) {
	dmu_buf_will_dirty(hdl->sa_spill, tx);
	buftypes \|= SA_SPILL;
	}
	}
	}
	if (error && error != ENOENT) {
	return ((error == ECKSUM) ? EIO : error);
	}

	switch (data_op) {
	case SA_LOOKUP:
	if (bulk[i].sa_addr == NULL)
	return (SET_ERROR(ENOENT));
	if (bulk[i].sa_data) {
	SA_COPY_DATA(bulk[i].sa_data_func,
	bulk[i].sa_addr, bulk[i].sa_data,
	bulk[i].sa_size);
	}
	continue;

	case SA_UPDATE:
	/* existing rewrite of attr */
	if (bulk[i].sa_addr &&
	bulk[i].sa_size == bulk[i].sa_length) {
	SA_COPY_DATA(bulk[i].sa_data_func,
	bulk[i].sa_data, bulk[i].sa_addr,
	bulk[i].sa_length);
	continue;
	} else if (bulk[i].sa_addr) { /* attr size change */
	error = sa_modify_attrs(hdl, bulk[i].sa_attr,
	SA_REPLACE, bulk[i].sa_data_func,
	bulk[i].sa_data, bulk[i].sa_length, tx);
	} else { /* adding new attribute */
	error = sa_modify_attrs(hdl, bulk[i].sa_attr,
	SA_ADD, bulk[i].sa_data_func,
	bulk[i].sa_data, bulk[i].sa_length, tx);
	}
	if (error)
	return (error);
	break;
	default:
	break;
	}
	}
	return (error);
	}

	static sa_lot_t *
	sa_add_layout_entry(objset_t os, const sa_attr_type_t attrs, int attr_count,
	uint64_t lot_num, uint64_t hash, boolean_t zapadd, dmu_tx_t *tx)
	{
	sa_os_t *sa = os->os_sa;
	sa_lot_t tb, findtb;
	int i;
	avl_index_t loc;

	ASSERT(MUTEX_HELD(&sa->sa_lock));
	tb = kmem_zalloc(sizeof (sa_lot_t), KM_SLEEP);
	tb->lot_attr_count = attr_count;
	tb->lot_attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
	KM_SLEEP);
	memcpy(tb->lot_attrs, attrs, sizeof (sa_attr_type_t) * attr_count);
	tb->lot_num = lot_num;
	tb->lot_hash = hash;
	tb->lot_instance = 0;

	if (zapadd) {
	char attr_name[8];

	if (sa->sa_layout_attr_obj == 0) {
	sa->sa_layout_attr_obj = zap_create_link(os,
	DMU_OT_SA_ATTR_LAYOUTS,
	sa->sa_master_obj, SA_LAYOUTS, tx);
	}

	(void) snprintf(attr_name, sizeof (attr_name),
	"%d", (int)lot_num);
	VERIFY(0 == zap_update(os, os->os_sa->sa_layout_attr_obj,
	attr_name, 2, attr_count, attrs, tx));
	}

	list_create(&tb->lot_idx_tab, sizeof (sa_idx_tab_t),
	offsetof(sa_idx_tab_t, sa_next));

	for (i = 0; i != attr_count; i++) {
	if (sa->sa_attr_table[tb->lot_attrs[i]].sa_length == 0)
	tb->lot_var_sizes++;
	}

	avl_add(&sa->sa_layout_num_tree, tb);

	/* verify we don't have a hash collision */
	if ((findtb = avl_find(&sa->sa_layout_hash_tree, tb, &loc)) != NULL) {
	for (; findtb && findtb->lot_hash == hash;
	findtb = AVL_NEXT(&sa->sa_layout_hash_tree, findtb)) {
	if (findtb->lot_instance != tb->lot_instance)
	break;
	tb->lot_instance++;
	}
	}
	avl_add(&sa->sa_layout_hash_tree, tb);
	return (tb);
	}

	static void
	sa_find_layout(objset_t os, uint64_t hash, sa_attr_type_t attrs,
	int count, dmu_tx_t tx, sa_lot_t *lot)
	{
	sa_lot_t *tb, tbsearch;
	avl_index_t loc;
	sa_os_t *sa = os->os_sa;
	boolean_t found = B_FALSE;

	mutex_enter(&sa->sa_lock);
	tbsearch.lot_hash = hash;
	tbsearch.lot_instance = 0;
	tb = avl_find(&sa->sa_layout_hash_tree, &tbsearch, &loc);
	if (tb) {
	for (; tb && tb->lot_hash == hash;
	tb = AVL_NEXT(&sa->sa_layout_hash_tree, tb)) {
	if (sa_layout_equal(tb, attrs, count) == 0) {
	found = B_TRUE;
	break;
	}
	}
	}
	if (!found) {
	tb = sa_add_layout_entry(os, attrs, count,
	avl_numnodes(&sa->sa_layout_num_tree), hash, B_TRUE, tx);
	}
	mutex_exit(&sa->sa_lock);
	*lot = tb;
	}

	static int
	sa_resize_spill(sa_handle_t hdl, uint32_t size, dmu_tx_t tx)
	{
	int error;
	uint32_t blocksize;

	if (size == 0) {
	blocksize = SPA_MINBLOCKSIZE;
	} else if (size > SPA_OLD_MAXBLOCKSIZE) {
	ASSERT(0);
	return (SET_ERROR(EFBIG));
	} else {
	blocksize = P2ROUNDUP_TYPED(size, SPA_MINBLOCKSIZE, uint32_t);
	}

	error = dbuf_spill_set_blksz(hdl->sa_spill, blocksize, tx);
	ASSERT(error == 0);
	return (error);
	}

	static void
	sa_copy_data(sa_data_locator_t func, void datastart, void *target, int buflen)
	{
	if (func == NULL) {
	memcpy(target, datastart, buflen);
	} else {
	boolean_t start;
	int bytes;
	void *dataptr;
	void *saptr = target;
	uint32_t length;

	start = B_TRUE;
	bytes = 0;
	while (bytes < buflen) {
	func(&dataptr, &length, buflen, start, datastart);
	memcpy(saptr, dataptr, length);
	saptr = (void *)((caddr_t)saptr + length);
	bytes += length;
	start = B_FALSE;
	}
	}
	}

	/*
	* Determine several different values pertaining to system attribute
	* buffers.
	*
	* Return the size of the sa_hdr_phys_t header for the buffer. Each
	* variable length attribute except the first contributes two bytes to
	* the header size, which is then rounded up to an 8-byte boundary.
	*
	* The following output parameters are also computed.
	*
	* index - The index of the first attribute in attr_desc that will
	* spill over. Only valid if will_spill is set.
	*
	* total - The total number of bytes of all system attributes described
	* in attr_desc.
	*
	* will_spill - Set when spilling is necessary. It is only set when
	* the buftype is SA_BONUS.
	*/
	static int
	sa_find_sizes(sa_os_t sa, sa_bulk_attr_t attr_desc, int attr_count,
	dmu_buf_t db, sa_buf_type_t buftype, int full_space, int index,
	int total, boolean_t will_spill)
	{
	int var_size_count = 0;
	int i;
	int hdrsize;
	int extra_hdrsize;

	if (buftype == SA_BONUS && sa->sa_force_spill) {
	*total = 0;
	*index = 0;
	*will_spill = B_TRUE;
	return (0);
	}

	*index = -1;
	*total = 0;
	*will_spill = B_FALSE;

	extra_hdrsize = 0;
	hdrsize = (SA_BONUSTYPE_FROM_DB(db) == DMU_OT_ZNODE) ? 0 :
	sizeof (sa_hdr_phys_t);

	ASSERT(IS_P2ALIGNED(full_space, 8));

	for (i = 0; i != attr_count; i++) {
	boolean_t is_var_sz, might_spill_here;
	int tmp_hdrsize;

	total = P2ROUNDUP(total, 8);
	*total += attr_desc[i].sa_length;
	if (*will_spill)
	continue;

	is_var_sz = (SA_REGISTERED_LEN(sa, attr_desc[i].sa_attr) == 0);
	if (is_var_sz)
	var_size_count++;

	/*
	* Calculate what the SA header size would be if this
	* attribute doesn't spill.
	*/
	tmp_hdrsize = hdrsize + ((is_var_sz && var_size_count > 1) ?
	sizeof (uint16_t) : 0);

	/*
	* Check whether this attribute spans into the space
	* that would be used by the spill block pointer should
	* a spill block be needed.
	*/
	might_spill_here =
	buftype == SA_BONUS && *index == -1 &&
	(*total + P2ROUNDUP(tmp_hdrsize, 8)) >
	(full_space - sizeof (blkptr_t));

	if (is_var_sz && var_size_count > 1) {
	if (buftype == SA_SPILL \|\|
	tmp_hdrsize + *total < full_space) {
	/*
	* Record the extra header size in case this
	* increase needs to be reversed due to
	* spill-over.
	*/
	hdrsize = tmp_hdrsize;
	if (*index != -1 \|\| might_spill_here)
	extra_hdrsize += sizeof (uint16_t);
	} else {
	ASSERT(buftype == SA_BONUS);
	if (*index == -1)
	*index = i;
	*will_spill = B_TRUE;
	continue;
	}
	}

	/*
	* Store index of where spill could occur. Then
	* continue to count the remaining attribute sizes. The
	* sum is used later for sizing bonus and spill buffer.
	*/
	if (might_spill_here)
	*index = i;

	if ((*total + P2ROUNDUP(hdrsize, 8)) > full_space &&
	buftype == SA_BONUS)
	*will_spill = B_TRUE;
	}

	if (*will_spill)
	hdrsize -= extra_hdrsize;

	hdrsize = P2ROUNDUP(hdrsize, 8);
	return (hdrsize);
	}

	#define BUF_SPACE_NEEDED(total, header) (total + header)

	/*
	* Find layout that corresponds to ordering of attributes
	* If not found a new layout number is created and added to
	* persistent layout tables.
	*/
	static int
	sa_build_layouts(sa_handle_t hdl, sa_bulk_attr_t attr_desc, int attr_count,
	dmu_tx_t *tx)
	{
	sa_os_t *sa = hdl->sa_os->os_sa;
	uint64_t hash;
	sa_buf_type_t buftype;
	sa_hdr_phys_t *sahdr;
	void *data_start;
	sa_attr_type_t attrs, attrs_start;
	int i, lot_count;
	int dnodesize;
	int spill_idx;
	int hdrsize;
	int spillhdrsize = 0;
	int used;
	dmu_object_type_t bonustype;
	sa_lot_t *lot;
	int len_idx;
	int spill_used;
	int bonuslen;
	boolean_t spilling;

	dmu_buf_will_dirty(hdl->sa_bonus, tx);
	bonustype = SA_BONUSTYPE_FROM_DB(hdl->sa_bonus);
	dmu_object_dnsize_from_db(hdl->sa_bonus, &dnodesize);
	bonuslen = DN_BONUS_SIZE(dnodesize);

	/* first determine bonus header size and sum of all attributes */
	hdrsize = sa_find_sizes(sa, attr_desc, attr_count, hdl->sa_bonus,
	SA_BONUS, bonuslen, &spill_idx, &used, &spilling);

	if (used > SPA_OLD_MAXBLOCKSIZE)
	return (SET_ERROR(EFBIG));

	VERIFY0(dmu_set_bonus(hdl->sa_bonus, spilling ?
	MIN(bonuslen - sizeof (blkptr_t), used + hdrsize) :
	used + hdrsize, tx));

	ASSERT((bonustype == DMU_OT_ZNODE && spilling == 0) \|\|
	bonustype == DMU_OT_SA);

	/* setup and size spill buffer when needed */
	if (spilling) {
	boolean_t dummy;

	if (hdl->sa_spill == NULL) {
	VERIFY(dmu_spill_hold_by_bonus(hdl->sa_bonus, 0, NULL,
	&hdl->sa_spill) == 0);
	}
	dmu_buf_will_dirty(hdl->sa_spill, tx);

	spillhdrsize = sa_find_sizes(sa, &attr_desc[spill_idx],
	attr_count - spill_idx, hdl->sa_spill, SA_SPILL,
	hdl->sa_spill->db_size, &i, &spill_used, &dummy);

	if (spill_used > SPA_OLD_MAXBLOCKSIZE)
	return (SET_ERROR(EFBIG));

	if (BUF_SPACE_NEEDED(spill_used, spillhdrsize) >
	hdl->sa_spill->db_size)
	VERIFY(0 == sa_resize_spill(hdl,
	BUF_SPACE_NEEDED(spill_used, spillhdrsize), tx));
	}

	/* setup starting pointers to lay down data */
	data_start = (void *)((uintptr_t)hdl->sa_bonus->db_data + hdrsize);
	sahdr = (sa_hdr_phys_t *)hdl->sa_bonus->db_data;
	buftype = SA_BONUS;

	attrs_start = attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
	KM_SLEEP);
	lot_count = 0;

	for (i = 0, len_idx = 0, hash = -1ULL; i != attr_count; i++) {
	uint16_t length;

	ASSERT(IS_P2ALIGNED(data_start, 8));
	attrs[i] = attr_desc[i].sa_attr;
	length = SA_REGISTERED_LEN(sa, attrs[i]);
	if (length == 0)
	length = attr_desc[i].sa_length;

	if (spilling && i == spill_idx) { /* switch to spill buffer */
	VERIFY(bonustype == DMU_OT_SA);
	if (buftype == SA_BONUS && !sa->sa_force_spill) {
	sa_find_layout(hdl->sa_os, hash, attrs_start,
	lot_count, tx, &lot);
	SA_SET_HDR(sahdr, lot->lot_num, hdrsize);
	}

	buftype = SA_SPILL;
	hash = -1ULL;
	len_idx = 0;

	sahdr = (sa_hdr_phys_t *)hdl->sa_spill->db_data;
	sahdr->sa_magic = SA_MAGIC;
	data_start = (void *)((uintptr_t)sahdr +
	spillhdrsize);
	attrs_start = &attrs[i];
	lot_count = 0;
	}
	hash ^= SA_ATTR_HASH(attrs[i]);
	attr_desc[i].sa_addr = data_start;
	attr_desc[i].sa_size = length;
	SA_COPY_DATA(attr_desc[i].sa_data_func, attr_desc[i].sa_data,
	data_start, length);
	if (sa->sa_attr_table[attrs[i]].sa_length == 0) {
	sahdr->sa_lengths[len_idx++] = length;
	}
	data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
	length), 8);
	lot_count++;
	}

	sa_find_layout(hdl->sa_os, hash, attrs_start, lot_count, tx, &lot);

	/*
	* Verify that old znodes always have layout number 0.
	* Must be DMU_OT_SA for arbitrary layouts
	*/
	VERIFY((bonustype == DMU_OT_ZNODE && lot->lot_num == 0) \|\|
	(bonustype == DMU_OT_SA && lot->lot_num > 1));

	if (bonustype == DMU_OT_SA) {
	SA_SET_HDR(sahdr, lot->lot_num,
	buftype == SA_BONUS ? hdrsize : spillhdrsize);
	}

	kmem_free(attrs, sizeof (sa_attr_type_t) * attr_count);
	if (hdl->sa_bonus_tab) {
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);
	hdl->sa_bonus_tab = NULL;
	}
	if (!sa->sa_force_spill)
	VERIFY(0 == sa_build_index(hdl, SA_BONUS));
	if (hdl->sa_spill) {
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
	if (!spilling) {
	/*
	* remove spill block that is no longer needed.
	*/
	dmu_buf_rele(hdl->sa_spill, NULL);
	hdl->sa_spill = NULL;
	hdl->sa_spill_tab = NULL;
	VERIFY(0 == dmu_rm_spill(hdl->sa_os,
	sa_handle_object(hdl), tx));
	} else {
	VERIFY(0 == sa_build_index(hdl, SA_SPILL));
	}
	}

	return (0);
	}

	static void
	sa_free_attr_table(sa_os_t *sa)
	{
	int i;

	if (sa->sa_attr_table == NULL)
	return;

	for (i = 0; i != sa->sa_num_attrs; i++) {
	if (sa->sa_attr_table[i].sa_name)
	kmem_free(sa->sa_attr_table[i].sa_name,
	strlen(sa->sa_attr_table[i].sa_name) + 1);
	}

	kmem_free(sa->sa_attr_table,
	sizeof (sa_attr_table_t) * sa->sa_num_attrs);

	sa->sa_attr_table = NULL;
	}

	static int
	sa_attr_table_setup(objset_t os, const sa_attr_reg_t reg_attrs, int count)
	{
	sa_os_t *sa = os->os_sa;
	uint64_t sa_attr_count = 0;
	uint64_t sa_reg_count = 0;
	int error = 0;
	uint64_t attr_value;
	sa_attr_table_t *tb;
	zap_cursor_t zc;
	zap_attribute_t za;
	int registered_count = 0;
	int i;
	dmu_objset_type_t ostype = dmu_objset_type(os);

	sa->sa_user_table =
	kmem_zalloc(count * sizeof (sa_attr_type_t), KM_SLEEP);
	sa->sa_user_table_sz = count * sizeof (sa_attr_type_t);

	if (sa->sa_reg_attr_obj != 0) {
	error = zap_count(os, sa->sa_reg_attr_obj,
	&sa_attr_count);

	/*
	* Make sure we retrieved a count and that it isn't zero
	*/
	if (error \|\| (error == 0 && sa_attr_count == 0)) {
	if (error == 0)
	error = SET_ERROR(EINVAL);
	goto bail;
	}
	sa_reg_count = sa_attr_count;
	}

	if (ostype == DMU_OST_ZFS && sa_attr_count == 0)
	sa_attr_count += sa_legacy_attr_count;

	/* Allocate attribute numbers for attributes that aren't registered */
	for (i = 0; i != count; i++) {
	boolean_t found = B_FALSE;
	int j;

	if (ostype == DMU_OST_ZFS) {
	for (j = 0; j != sa_legacy_attr_count; j++) {
	if (strcmp(reg_attrs[i].sa_name,
	sa_legacy_attrs[j].sa_name) == 0) {
	sa->sa_user_table[i] =
	sa_legacy_attrs[j].sa_attr;
	found = B_TRUE;
	}
	}
	}
	if (found)
	continue;

	if (sa->sa_reg_attr_obj)
	error = zap_lookup(os, sa->sa_reg_attr_obj,
	reg_attrs[i].sa_name, 8, 1, &attr_value);
	else
	error = SET_ERROR(ENOENT);
	switch (error) {
	case ENOENT:
	sa->sa_user_table[i] = (sa_attr_type_t)sa_attr_count;
	sa_attr_count++;
	break;
	case 0:
	sa->sa_user_table[i] = ATTR_NUM(attr_value);
	break;
	default:
	goto bail;
	}
	}

	sa->sa_num_attrs = sa_attr_count;
	tb = sa->sa_attr_table =
	kmem_zalloc(sizeof (sa_attr_table_t) * sa_attr_count, KM_SLEEP);

	/*
	* Attribute table is constructed from requested attribute list,
	* previously foreign registered attributes, and also the legacy
	* ZPL set of attributes.
	*/

	if (sa->sa_reg_attr_obj) {
	for (zap_cursor_init(&zc, os, sa->sa_reg_attr_obj);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t value;
	value = za.za_first_integer;

	registered_count++;
	tb[ATTR_NUM(value)].sa_attr = ATTR_NUM(value);
	tb[ATTR_NUM(value)].sa_length = ATTR_LENGTH(value);
	tb[ATTR_NUM(value)].sa_byteswap = ATTR_BSWAP(value);
	tb[ATTR_NUM(value)].sa_registered = B_TRUE;

	if (tb[ATTR_NUM(value)].sa_name) {
	continue;
	}
	tb[ATTR_NUM(value)].sa_name =
	kmem_zalloc(strlen(za.za_name) +1, KM_SLEEP);
	(void) strlcpy(tb[ATTR_NUM(value)].sa_name, za.za_name,
	strlen(za.za_name) +1);
	}
	zap_cursor_fini(&zc);
	/*
	* Make sure we processed the correct number of registered
	* attributes
	*/
	if (registered_count != sa_reg_count) {
	ASSERT(error != 0);
	goto bail;
	}

	}

	if (ostype == DMU_OST_ZFS) {
	for (i = 0; i != sa_legacy_attr_count; i++) {
	if (tb[i].sa_name)
	continue;
	tb[i].sa_attr = sa_legacy_attrs[i].sa_attr;
	tb[i].sa_length = sa_legacy_attrs[i].sa_length;
	tb[i].sa_byteswap = sa_legacy_attrs[i].sa_byteswap;
	tb[i].sa_registered = B_FALSE;
	tb[i].sa_name =
	kmem_zalloc(strlen(sa_legacy_attrs[i].sa_name) +1,
	KM_SLEEP);
	(void) strlcpy(tb[i].sa_name,
	sa_legacy_attrs[i].sa_name,
	strlen(sa_legacy_attrs[i].sa_name) + 1);
	}
	}

	for (i = 0; i != count; i++) {
	sa_attr_type_t attr_id;

	attr_id = sa->sa_user_table[i];
	if (tb[attr_id].sa_name)
	continue;

	tb[attr_id].sa_length = reg_attrs[i].sa_length;
	tb[attr_id].sa_byteswap = reg_attrs[i].sa_byteswap;
	tb[attr_id].sa_attr = attr_id;
	tb[attr_id].sa_name =
	kmem_zalloc(strlen(reg_attrs[i].sa_name) + 1, KM_SLEEP);
	(void) strlcpy(tb[attr_id].sa_name, reg_attrs[i].sa_name,
	strlen(reg_attrs[i].sa_name) + 1);
	}

	sa->sa_need_attr_registration =
	(sa_attr_count != registered_count);

	return (0);
	bail:
	kmem_free(sa->sa_user_table, count * sizeof (sa_attr_type_t));
	sa->sa_user_table = NULL;
	sa_free_attr_table(sa);
	ASSERT(error != 0);
	return (error);
	}

	int
	sa_setup(objset_t os, uint64_t sa_obj, const sa_attr_reg_t reg_attrs,
	int count, sa_attr_type_t **user_table)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	sa_os_t *sa;
	dmu_objset_type_t ostype = dmu_objset_type(os);
	sa_attr_type_t *tb;
	int error;

	mutex_enter(&os->os_user_ptr_lock);
	if (os->os_sa) {
	mutex_enter(&os->os_sa->sa_lock);
	mutex_exit(&os->os_user_ptr_lock);
	tb = os->os_sa->sa_user_table;
	mutex_exit(&os->os_sa->sa_lock);
	*user_table = tb;
	return (0);
	}

	sa = kmem_zalloc(sizeof (sa_os_t), KM_SLEEP);
	mutex_init(&sa->sa_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	sa->sa_master_obj = sa_obj;

	os->os_sa = sa;
	mutex_enter(&sa->sa_lock);
	mutex_exit(&os->os_user_ptr_lock);
	avl_create(&sa->sa_layout_num_tree, layout_num_compare,
	sizeof (sa_lot_t), offsetof(sa_lot_t, lot_num_node));
	avl_create(&sa->sa_layout_hash_tree, layout_hash_compare,
	sizeof (sa_lot_t), offsetof(sa_lot_t, lot_hash_node));

	if (sa_obj) {
	error = zap_lookup(os, sa_obj, SA_LAYOUTS,
	8, 1, &sa->sa_layout_attr_obj);
	if (error != 0 && error != ENOENT)
	goto fail;
	error = zap_lookup(os, sa_obj, SA_REGISTRY,
	8, 1, &sa->sa_reg_attr_obj);
	if (error != 0 && error != ENOENT)
	goto fail;
	}

	if ((error = sa_attr_table_setup(os, reg_attrs, count)) != 0)
	goto fail;

	if (sa->sa_layout_attr_obj != 0) {
	uint64_t layout_count;

	error = zap_count(os, sa->sa_layout_attr_obj,
	&layout_count);

	/*
	* Layout number count should be > 0
	*/
	if (error \|\| (error == 0 && layout_count == 0)) {
	if (error == 0)
	error = SET_ERROR(EINVAL);
	goto fail;
	}

	for (zap_cursor_init(&zc, os, sa->sa_layout_attr_obj);
	(error = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	sa_attr_type_t *lot_attrs;
	uint64_t lot_num;

	lot_attrs = kmem_zalloc(sizeof (sa_attr_type_t) *
	za.za_num_integers, KM_SLEEP);

	if ((error = (zap_lookup(os, sa->sa_layout_attr_obj,
	za.za_name, 2, za.za_num_integers,
	lot_attrs))) != 0) {
	kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
	za.za_num_integers);
	break;
	}
	VERIFY0(ddi_strtoull(za.za_name, NULL, 10,
	(unsigned long long *)&lot_num));

	(void) sa_add_layout_entry(os, lot_attrs,
	za.za_num_integers, lot_num,
	sa_layout_info_hash(lot_attrs,
	za.za_num_integers), B_FALSE, NULL);
	kmem_free(lot_attrs, sizeof (sa_attr_type_t) *
	za.za_num_integers);
	}
	zap_cursor_fini(&zc);

	/*
	* Make sure layout count matches number of entries added
	* to AVL tree
	*/
	if (avl_numnodes(&sa->sa_layout_num_tree) != layout_count) {
	ASSERT(error != 0);
	goto fail;
	}
	}

	/* Add special layout number for old ZNODES */
	if (ostype == DMU_OST_ZFS) {
	(void) sa_add_layout_entry(os, sa_legacy_zpl_layout,
	sa_legacy_attr_count, 0,
	sa_layout_info_hash(sa_legacy_zpl_layout,
	sa_legacy_attr_count), B_FALSE, NULL);

	(void) sa_add_layout_entry(os, sa_dummy_zpl_layout, 0, 1,
	0, B_FALSE, NULL);
	}
	*user_table = os->os_sa->sa_user_table;
	mutex_exit(&sa->sa_lock);
	return (0);
	fail:
	os->os_sa = NULL;
	sa_free_attr_table(sa);
	if (sa->sa_user_table)
	kmem_free(sa->sa_user_table, sa->sa_user_table_sz);
	mutex_exit(&sa->sa_lock);
	avl_destroy(&sa->sa_layout_hash_tree);
	avl_destroy(&sa->sa_layout_num_tree);
	mutex_destroy(&sa->sa_lock);
	kmem_free(sa, sizeof (sa_os_t));
	return ((error == ECKSUM) ? EIO : error);
	}

	void
	sa_tear_down(objset_t *os)
	{
	sa_os_t *sa = os->os_sa;
	sa_lot_t *layout;
	void *cookie;

	kmem_free(sa->sa_user_table, sa->sa_user_table_sz);

	/* Free up attr table */

	sa_free_attr_table(sa);

	cookie = NULL;
	while ((layout =
	avl_destroy_nodes(&sa->sa_layout_hash_tree, &cookie))) {
	sa_idx_tab_t *tab;
	while ((tab = list_head(&layout->lot_idx_tab))) {
	ASSERT(zfs_refcount_count(&tab->sa_refcount));
	sa_idx_tab_rele(os, tab);
	}
	}

	cookie = NULL;
	while ((layout = avl_destroy_nodes(&sa->sa_layout_num_tree, &cookie))) {
	kmem_free(layout->lot_attrs,
	sizeof (sa_attr_type_t) * layout->lot_attr_count);
	kmem_free(layout, sizeof (sa_lot_t));
	}

	avl_destroy(&sa->sa_layout_hash_tree);
	avl_destroy(&sa->sa_layout_num_tree);
	mutex_destroy(&sa->sa_lock);

	kmem_free(sa, sizeof (sa_os_t));
	os->os_sa = NULL;
	}

	static void
	sa_build_idx_tab(void hdr, void attr_addr, sa_attr_type_t attr,
	uint16_t length, int length_idx, boolean_t var_length, void *userp)
	{
	sa_idx_tab_t *idx_tab = userp;

	if (var_length) {
	ASSERT(idx_tab->sa_variable_lengths);
	idx_tab->sa_variable_lengths[length_idx] = length;
	}
	TOC_ATTR_ENCODE(idx_tab->sa_idx_tab[attr], length_idx,
	(uint32_t)((uintptr_t)attr_addr - (uintptr_t)hdr));
	}

	static void
	sa_attr_iter(objset_t os, sa_hdr_phys_t hdr, dmu_object_type_t type,
	sa_iterfunc_t func, sa_lot_t tab, void userp)
	{
	void *data_start;
	sa_lot_t *tb = tab;
	sa_lot_t search;
	avl_index_t loc;
	sa_os_t *sa = os->os_sa;
	int i;
	uint16_t *length_start = NULL;
	uint8_t length_idx = 0;

	if (tab == NULL) {
	search.lot_num = SA_LAYOUT_NUM(hdr, type);
	tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);
	ASSERT(tb);
	}

	if (IS_SA_BONUSTYPE(type)) {
	data_start = (void *)P2ROUNDUP(((uintptr_t)hdr +
	offsetof(sa_hdr_phys_t, sa_lengths) +
	(sizeof (uint16_t) * tb->lot_var_sizes)), 8);
	length_start = hdr->sa_lengths;
	} else {
	data_start = hdr;
	}

	for (i = 0; i != tb->lot_attr_count; i++) {
	int attr_length, reg_length;
	uint8_t idx_len;

	reg_length = sa->sa_attr_table[tb->lot_attrs[i]].sa_length;
	if (reg_length) {
	attr_length = reg_length;
	idx_len = 0;
	} else {
	attr_length = length_start[length_idx];
	idx_len = length_idx++;
	}

	func(hdr, data_start, tb->lot_attrs[i], attr_length,
	idx_len, reg_length == 0 ? B_TRUE : B_FALSE, userp);

	data_start = (void *)P2ROUNDUP(((uintptr_t)data_start +
	attr_length), 8);
	}
	}

	static void
	sa_byteswap_cb(void hdr, void attr_addr, sa_attr_type_t attr,
	uint16_t length, int length_idx, boolean_t variable_length, void *userp)
	{
	(void) hdr, (void) length_idx, (void) variable_length;
	sa_handle_t *hdl = userp;
	sa_os_t *sa = hdl->sa_os->os_sa;

	sa_bswap_table[sa->sa_attr_table[attr].sa_byteswap](attr_addr, length);
	}

	static void
	sa_byteswap(sa_handle_t *hdl, sa_buf_type_t buftype)
	{
	sa_hdr_phys_t *sa_hdr_phys = SA_GET_HDR(hdl, buftype);
	dmu_buf_impl_t *db;
	int num_lengths = 1;
	int i;
	sa_os_t *sa __maybe_unused = hdl->sa_os->os_sa;

	ASSERT(MUTEX_HELD(&sa->sa_lock));
	if (sa_hdr_phys->sa_magic == SA_MAGIC)
	return;

	db = SA_GET_DB(hdl, buftype);

	if (buftype == SA_SPILL) {
	arc_release(db->db_buf, NULL);
	arc_buf_thaw(db->db_buf);
	}

	sa_hdr_phys->sa_magic = BSWAP_32(sa_hdr_phys->sa_magic);
	sa_hdr_phys->sa_layout_info = BSWAP_16(sa_hdr_phys->sa_layout_info);

	/*
	* Determine number of variable lengths in header
	* The standard 8 byte header has one for free and a
	* 16 byte header would have 4 + 1;
	*/
	if (SA_HDR_SIZE(sa_hdr_phys) > 8)
	num_lengths += (SA_HDR_SIZE(sa_hdr_phys) - 8) >> 1;
	for (i = 0; i != num_lengths; i++)
	sa_hdr_phys->sa_lengths[i] =
	BSWAP_16(sa_hdr_phys->sa_lengths[i]);

	sa_attr_iter(hdl->sa_os, sa_hdr_phys, DMU_OT_SA,
	sa_byteswap_cb, NULL, hdl);

	if (buftype == SA_SPILL)
	arc_buf_freeze(((dmu_buf_impl_t *)hdl->sa_spill)->db_buf);
	}

	static int
	sa_build_index(sa_handle_t *hdl, sa_buf_type_t buftype)
	{
	sa_hdr_phys_t *sa_hdr_phys;
	dmu_buf_impl_t *db = SA_GET_DB(hdl, buftype);
	dmu_object_type_t bonustype = SA_BONUSTYPE_FROM_DB(db);
	sa_os_t *sa = hdl->sa_os->os_sa;
	sa_idx_tab_t *idx_tab;

	sa_hdr_phys = SA_GET_HDR(hdl, buftype);

	mutex_enter(&sa->sa_lock);

	/* Do we need to byteswap? */

	/* only check if not old znode */
	if (IS_SA_BONUSTYPE(bonustype) && sa_hdr_phys->sa_magic != SA_MAGIC &&
	sa_hdr_phys->sa_magic != 0) {
	if (BSWAP_32(sa_hdr_phys->sa_magic) != SA_MAGIC) {
	mutex_exit(&sa->sa_lock);
	zfs_dbgmsg("Buffer Header: %x != SA_MAGIC:%x "
	"object=%#llx\n", sa_hdr_phys->sa_magic, SA_MAGIC,
	(u_longlong_t)db->db.db_object);
	return (SET_ERROR(EIO));
	}
	sa_byteswap(hdl, buftype);
	}

	idx_tab = sa_find_idx_tab(hdl->sa_os, bonustype, sa_hdr_phys);

	if (buftype == SA_BONUS)
	hdl->sa_bonus_tab = idx_tab;
	else
	hdl->sa_spill_tab = idx_tab;

	mutex_exit(&sa->sa_lock);
	return (0);
	}

	static void
	sa_evict_sync(void *dbu)
	{
	(void) dbu;
	panic("evicting sa dbuf\n");
	}

	static void
	sa_idx_tab_rele(objset_t os, void arg)
	{
	sa_os_t *sa = os->os_sa;
	sa_idx_tab_t *idx_tab = arg;

	if (idx_tab == NULL)
	return;

	mutex_enter(&sa->sa_lock);
	if (zfs_refcount_remove(&idx_tab->sa_refcount, NULL) == 0) {
	list_remove(&idx_tab->sa_layout->lot_idx_tab, idx_tab);
	if (idx_tab->sa_variable_lengths)
	kmem_free(idx_tab->sa_variable_lengths,
	sizeof (uint16_t) *
	idx_tab->sa_layout->lot_var_sizes);
	zfs_refcount_destroy(&idx_tab->sa_refcount);
	kmem_free(idx_tab->sa_idx_tab,
	sizeof (uint32_t) * sa->sa_num_attrs);
	kmem_free(idx_tab, sizeof (sa_idx_tab_t));
	}
	mutex_exit(&sa->sa_lock);
	}

	static void
	sa_idx_tab_hold(objset_t os, sa_idx_tab_t idx_tab)
	{
	sa_os_t *sa __maybe_unused = os->os_sa;

	ASSERT(MUTEX_HELD(&sa->sa_lock));
	(void) zfs_refcount_add(&idx_tab->sa_refcount, NULL);
	}

	void
	sa_spill_rele(sa_handle_t *hdl)
	{
	mutex_enter(&hdl->sa_lock);
	if (hdl->sa_spill) {
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
	dmu_buf_rele(hdl->sa_spill, NULL);
	hdl->sa_spill = NULL;
	hdl->sa_spill_tab = NULL;
	}
	mutex_exit(&hdl->sa_lock);
	}

	void
	sa_handle_destroy(sa_handle_t *hdl)
	{
	dmu_buf_t *db = hdl->sa_bonus;

	mutex_enter(&hdl->sa_lock);
	(void) dmu_buf_remove_user(db, &hdl->sa_dbu);

	if (hdl->sa_bonus_tab)
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_bonus_tab);

	if (hdl->sa_spill_tab)
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);

	dmu_buf_rele(hdl->sa_bonus, NULL);

	if (hdl->sa_spill)
	dmu_buf_rele(hdl->sa_spill, NULL);
	mutex_exit(&hdl->sa_lock);

	kmem_cache_free(sa_cache, hdl);
	}

	int
	sa_handle_get_from_db(objset_t os, dmu_buf_t db, void *userp,
	sa_handle_type_t hdl_type, sa_handle_t **handlepp)
	{
	int error = 0;
	sa_handle_t *handle = NULL;
	#ifdef ZFS_DEBUG
	dmu_object_info_t doi;

	dmu_object_info_from_db(db, &doi);
	ASSERT(doi.doi_bonus_type == DMU_OT_SA \|\|
	doi.doi_bonus_type == DMU_OT_ZNODE);
	#endif
	/* find handle, if it exists */
	/* if one doesn't exist then create a new one, and initialize it */

	if (hdl_type == SA_HDL_SHARED)
	handle = dmu_buf_get_user(db);

	if (handle == NULL) {
	sa_handle_t *winner = NULL;

	handle = kmem_cache_alloc(sa_cache, KM_SLEEP);
	handle->sa_dbu.dbu_evict_func_sync = NULL;
	handle->sa_dbu.dbu_evict_func_async = NULL;
	handle->sa_userp = userp;
	handle->sa_bonus = db;
	handle->sa_os = os;
	handle->sa_spill = NULL;
	handle->sa_bonus_tab = NULL;
	handle->sa_spill_tab = NULL;

	error = sa_build_index(handle, SA_BONUS);

	if (hdl_type == SA_HDL_SHARED) {
	dmu_buf_init_user(&handle->sa_dbu, sa_evict_sync, NULL,
	NULL);
	winner = dmu_buf_set_user_ie(db, &handle->sa_dbu);
	}

	if (winner != NULL) {
	kmem_cache_free(sa_cache, handle);
	handle = winner;
	}
	}
	*handlepp = handle;

	return (error);
	}

	int
	sa_handle_get(objset_t objset, uint64_t objid, void userp,
	sa_handle_type_t hdl_type, sa_handle_t **handlepp)
	{
	dmu_buf_t *db;
	int error;

	if ((error = dmu_bonus_hold(objset, objid, NULL, &db)))
	return (error);

	return (sa_handle_get_from_db(objset, db, userp, hdl_type,
	handlepp));
	}

	int
	sa_buf_hold(objset_t objset, uint64_t obj_num, const void tag, dmu_buf_t **db)
	{
	return (dmu_bonus_hold(objset, obj_num, tag, db));
	}

	void
	sa_buf_rele(dmu_buf_t db, const void tag)
	{
	dmu_buf_rele(db, tag);
	}

	static int
	sa_lookup_impl(sa_handle_t hdl, sa_bulk_attr_t bulk, int count)
	{
	ASSERT(hdl);
	ASSERT(MUTEX_HELD(&hdl->sa_lock));
	return (sa_attr_op(hdl, bulk, count, SA_LOOKUP, NULL));
	}

	static int
	sa_lookup_locked(sa_handle_t hdl, sa_attr_type_t attr, void buf,
	uint32_t buflen)
	{
	int error;
	sa_bulk_attr_t bulk;

	VERIFY3U(buflen, <=, SA_ATTR_MAX_LEN);

	bulk.sa_attr = attr;
	bulk.sa_data = buf;
	bulk.sa_length = buflen;
	bulk.sa_data_func = NULL;

	ASSERT(hdl);
	error = sa_lookup_impl(hdl, &bulk, 1);
	return (error);
	}

	int
	sa_lookup(sa_handle_t hdl, sa_attr_type_t attr, void buf, uint32_t buflen)
	{
	int error;

	mutex_enter(&hdl->sa_lock);
	error = sa_lookup_locked(hdl, attr, buf, buflen);
	mutex_exit(&hdl->sa_lock);

	return (error);
	}

	#ifdef _KERNEL
	int
	sa_lookup_uio(sa_handle_t hdl, sa_attr_type_t attr, zfs_uio_t uio)
	{
	int error;
	sa_bulk_attr_t bulk;

	bulk.sa_data = NULL;
	bulk.sa_attr = attr;
	bulk.sa_data_func = NULL;

	ASSERT(hdl);

	mutex_enter(&hdl->sa_lock);
	if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) == 0) {
	error = zfs_uiomove((void *)bulk.sa_addr, MIN(bulk.sa_size,
	zfs_uio_resid(uio)), UIO_READ, uio);
	}
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	/*
	* For the existed object that is upgraded from old system, its ondisk layout
	* has no slot for the project ID attribute. But quota accounting logic needs
	* to access related slots by offset directly. So we need to adjust these old
	* objects' layout to make the project ID to some unified and fixed offset.
	*/
	int
	sa_add_projid(sa_handle_t hdl, dmu_tx_t tx, uint64_t projid)
	{
	znode_t *zp = sa_get_userdata(hdl);
	dmu_buf_t *db = sa_get_db(hdl);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int count = 0, err = 0;
	sa_bulk_attr_t bulk, attrs;
	zfs_acl_locator_cb_t locate = { 0 };
	uint64_t uid, gid, mode, rdev, xattr = 0, parent, gen, links;
	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
	zfs_acl_phys_t znode_acl = { 0 };
	char scanstamp[AV_SCANSTAMP_SZ];

	if (zp->z_acl_cached == NULL) {
	zfs_acl_t *aclp;

	mutex_enter(&zp->z_acl_lock);
	err = zfs_acl_node_read(zp, B_FALSE, &aclp, B_FALSE);
	mutex_exit(&zp->z_acl_lock);
	if (err != 0 && err != ENOENT)
	return (err);
	}

	bulk = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
	attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
	mutex_enter(&hdl->sa_lock);
	mutex_enter(&zp->z_lock);

	err = sa_lookup_locked(hdl, SA_ZPL_PROJID(zfsvfs), &projid,
	sizeof (uint64_t));
	if (unlikely(err == 0))
	/* Someone has added project ID attr by race. */
	err = EEXIST;
	if (err != ENOENT)
	goto out;

	/* First do a bulk query of the attributes that aren't cached */
	if (zp->z_is_sa) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&mode, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
	&gen, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
	&uid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
	&gid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
	&parent, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
	&atime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	&mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
	&crtime, 16);
	if (Z_ISBLK(ZTOTYPE(zp)) \|\| Z_ISCHR(ZTOTYPE(zp)))
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
	&rdev, 8);
	} else {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
	&atime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	&mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL,
	&crtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
	&gen, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&mode, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
	&parent, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL,
	&xattr, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
	&rdev, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
	&uid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
	&gid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
	&znode_acl, 88);
	}
	err = sa_bulk_lookup_locked(hdl, bulk, count);
	if (err != 0)
	goto out;

	err = sa_lookup_locked(hdl, SA_ZPL_XATTR(zfsvfs), &xattr, 8);
	if (err != 0 && err != ENOENT)
	goto out;

	zp->z_projid = projid;
	zp->z_pflags \|= ZFS_PROJID;
	links = ZTONLNK(zp);
	count = 0;
	err = 0;

	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&zp->z_size, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GEN(zfsvfs), NULL, &gen, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
	&crtime, 16);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_PROJID(zfsvfs), NULL, &projid, 8);

	if (Z_ISBLK(ZTOTYPE(zp)) \|\| Z_ISCHR(ZTOTYPE(zp)))
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
	&rdev, 8);

	if (zp->z_acl_cached != NULL) {
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
	&zp->z_acl_cached->z_acl_count, 8);
	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
	zfs_acl_xform(zp, zp->z_acl_cached, CRED());
	locate.cb_aclp = zp->z_acl_cached;
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
	zfs_acl_data_locator, &locate,
	zp->z_acl_cached->z_acl_bytes);
	}

	if (xattr)
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_XATTR(zfsvfs), NULL,
	&xattr, 8);

	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
	memcpy(scanstamp,
	(caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	AV_SCANSTAMP_SZ);
	SA_ADD_BULK_ATTR(attrs, count, SA_ZPL_SCANSTAMP(zfsvfs), NULL,
	scanstamp, AV_SCANSTAMP_SZ);
	zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
	}

	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
	VERIFY(sa_replace_all_by_template_locked(hdl, attrs, count, tx) == 0);
	if (znode_acl.z_acl_extern_obj) {
	VERIFY(0 == dmu_object_free(zfsvfs->z_os,
	znode_acl.z_acl_extern_obj, tx));
	}

	zp->z_is_sa = B_TRUE;

	out:
	mutex_exit(&zp->z_lock);
	mutex_exit(&hdl->sa_lock);
	kmem_free(attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
	kmem_free(bulk, sizeof (sa_bulk_attr_t) * ZPL_END);
	return (err);
	}
	#endif

	static sa_idx_tab_t *
	sa_find_idx_tab(objset_t os, dmu_object_type_t bonustype, sa_hdr_phys_t hdr)
	{
	sa_idx_tab_t *idx_tab;
	sa_os_t *sa = os->os_sa;
	sa_lot_t *tb, search;
	avl_index_t loc;

	/*
	* Deterimine layout number. If SA node and header == 0 then
	* force the index table to the dummy "1" empty layout.
	*
	* The layout number would only be zero for a newly created file
	* that has not added any attributes yet, or with crypto enabled which
	* doesn't write any attributes to the bonus buffer.
	*/

	search.lot_num = SA_LAYOUT_NUM(hdr, bonustype);

	tb = avl_find(&sa->sa_layout_num_tree, &search, &loc);

	/* Verify header size is consistent with layout information */
	ASSERT(tb);
	ASSERT((IS_SA_BONUSTYPE(bonustype) &&
	SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb)) \|\| !IS_SA_BONUSTYPE(bonustype) \|\|
	(IS_SA_BONUSTYPE(bonustype) && hdr->sa_layout_info == 0));

	/*
	* See if any of the already existing TOC entries can be reused?
	*/

	for (idx_tab = list_head(&tb->lot_idx_tab); idx_tab;
	idx_tab = list_next(&tb->lot_idx_tab, idx_tab)) {
	boolean_t valid_idx = B_TRUE;
	int i;

	if (tb->lot_var_sizes != 0 &&
	idx_tab->sa_variable_lengths != NULL) {
	for (i = 0; i != tb->lot_var_sizes; i++) {
	if (hdr->sa_lengths[i] !=
	idx_tab->sa_variable_lengths[i]) {
	valid_idx = B_FALSE;
	break;
	}
	}
	}
	if (valid_idx) {
	sa_idx_tab_hold(os, idx_tab);
	return (idx_tab);
	}
	}

	/* No such luck, create a new entry */
	idx_tab = kmem_zalloc(sizeof (sa_idx_tab_t), KM_SLEEP);
	idx_tab->sa_idx_tab =
	kmem_zalloc(sizeof (uint32_t) * sa->sa_num_attrs, KM_SLEEP);
	idx_tab->sa_layout = tb;
	zfs_refcount_create(&idx_tab->sa_refcount);
	if (tb->lot_var_sizes)
	idx_tab->sa_variable_lengths = kmem_alloc(sizeof (uint16_t) *
	tb->lot_var_sizes, KM_SLEEP);

	sa_attr_iter(os, hdr, bonustype, sa_build_idx_tab,
	tb, idx_tab);
	sa_idx_tab_hold(os, idx_tab); /* one hold for consumer */
	sa_idx_tab_hold(os, idx_tab); /* one for layout */
	list_insert_tail(&tb->lot_idx_tab, idx_tab);
	return (idx_tab);
	}

	void
	sa_default_locator(void *dataptr, uint32_t len, uint32_t total_len,
	boolean_t start, void *userdata)
	{
	ASSERT(start);

	*dataptr = userdata;
	*len = total_len;
	}

	static void
	sa_attr_register_sync(sa_handle_t hdl, dmu_tx_t tx)
	{
	uint64_t attr_value = 0;
	sa_os_t *sa = hdl->sa_os->os_sa;
	sa_attr_table_t *tb = sa->sa_attr_table;
	int i;

	mutex_enter(&sa->sa_lock);

	if (!sa->sa_need_attr_registration \|\| sa->sa_master_obj == 0) {
	mutex_exit(&sa->sa_lock);
	return;
	}

	if (sa->sa_reg_attr_obj == 0) {
	sa->sa_reg_attr_obj = zap_create_link(hdl->sa_os,
	DMU_OT_SA_ATTR_REGISTRATION,
	sa->sa_master_obj, SA_REGISTRY, tx);
	}
	for (i = 0; i != sa->sa_num_attrs; i++) {
	if (sa->sa_attr_table[i].sa_registered)
	continue;
	ATTR_ENCODE(attr_value, tb[i].sa_attr, tb[i].sa_length,
	tb[i].sa_byteswap);
	VERIFY(0 == zap_update(hdl->sa_os, sa->sa_reg_attr_obj,
	tb[i].sa_name, 8, 1, &attr_value, tx));
	tb[i].sa_registered = B_TRUE;
	}
	sa->sa_need_attr_registration = B_FALSE;
	mutex_exit(&sa->sa_lock);
	}

	/*
	* Replace all attributes with attributes specified in template.
	* If dnode had a spill buffer then those attributes will be
	* also be replaced, possibly with just an empty spill block
	*
	* This interface is intended to only be used for bulk adding of
	* attributes for a new file. It will also be used by the ZPL
	* when converting and old formatted znode to native SA support.
	*/
	int
	sa_replace_all_by_template_locked(sa_handle_t hdl, sa_bulk_attr_t attr_desc,
	int attr_count, dmu_tx_t *tx)
	{
	sa_os_t *sa = hdl->sa_os->os_sa;

	if (sa->sa_need_attr_registration)
	sa_attr_register_sync(hdl, tx);
	return (sa_build_layouts(hdl, attr_desc, attr_count, tx));
	}

	int
	sa_replace_all_by_template(sa_handle_t hdl, sa_bulk_attr_t attr_desc,
	int attr_count, dmu_tx_t *tx)
	{
	int error;

	mutex_enter(&hdl->sa_lock);
	error = sa_replace_all_by_template_locked(hdl, attr_desc,
	attr_count, tx);
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	/*
	* Add/remove a single attribute or replace a variable-sized attribute value
	* with a value of a different size, and then rewrite the entire set
	* of attributes.
	* Same-length attribute value replacement (including fixed-length attributes)
	* is handled more efficiently by the upper layers.
	*/
	static int
	sa_modify_attrs(sa_handle_t *hdl, sa_attr_type_t newattr,
	sa_data_op_t action, sa_data_locator_t locator, void datastart,
	uint16_t buflen, dmu_tx_t *tx)
	{
	sa_os_t *sa = hdl->sa_os->os_sa;
	dmu_buf_impl_t db = (dmu_buf_impl_t )hdl->sa_bonus;
	dnode_t *dn;
	sa_bulk_attr_t *attr_desc;
	void *old_data[2];
	int bonus_attr_count = 0;
	int bonus_data_size = 0;
	int spill_data_size = 0;
	int spill_attr_count = 0;
	int error;
	uint16_t length, reg_length;
	int i, j, k, length_idx;
	sa_hdr_phys_t *hdr;
	sa_idx_tab_t *idx_tab;
	int attr_count;
	int count;

	ASSERT(MUTEX_HELD(&hdl->sa_lock));

	/* First make of copy of the old data */

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	if (dn->dn_bonuslen != 0) {
	bonus_data_size = hdl->sa_bonus->db_size;
	old_data[0] = kmem_alloc(bonus_data_size, KM_SLEEP);
	memcpy(old_data[0], hdl->sa_bonus->db_data,
	hdl->sa_bonus->db_size);
	bonus_attr_count = hdl->sa_bonus_tab->sa_layout->lot_attr_count;
	} else {
	old_data[0] = NULL;
	}
	DB_DNODE_EXIT(db);

	/* Bring spill buffer online if it isn't currently */

	if ((error = sa_get_spill(hdl)) == 0) {
	spill_data_size = hdl->sa_spill->db_size;
	old_data[1] = vmem_alloc(spill_data_size, KM_SLEEP);
	memcpy(old_data[1], hdl->sa_spill->db_data,
	hdl->sa_spill->db_size);
	spill_attr_count =
	hdl->sa_spill_tab->sa_layout->lot_attr_count;
	} else if (error && error != ENOENT) {
	if (old_data[0])
	kmem_free(old_data[0], bonus_data_size);
	return (error);
	} else {
	old_data[1] = NULL;
	}

	/* build descriptor of all attributes */

	attr_count = bonus_attr_count + spill_attr_count;
	if (action == SA_ADD)
	attr_count++;
	else if (action == SA_REMOVE)
	attr_count--;

	attr_desc = kmem_zalloc(sizeof (sa_bulk_attr_t) * attr_count, KM_SLEEP);

	/*
	* loop through bonus and spill buffer if it exists, and
	* build up new attr_descriptor to reset the attributes
	*/
	k = j = 0;
	count = bonus_attr_count;
	hdr = SA_GET_HDR(hdl, SA_BONUS);
	idx_tab = SA_IDX_TAB_GET(hdl, SA_BONUS);
	for (; k != 2; k++) {
	/*
	* Iterate over each attribute in layout. Fetch the
	* size of variable-length attributes needing rewrite
	* from sa_lengths[].
	*/
	for (i = 0, length_idx = 0; i != count; i++) {
	sa_attr_type_t attr;

	attr = idx_tab->sa_layout->lot_attrs[i];
	reg_length = SA_REGISTERED_LEN(sa, attr);
	if (reg_length == 0) {
	length = hdr->sa_lengths[length_idx];
	length_idx++;
	} else {
	length = reg_length;
	}
	if (attr == newattr) {
	/*
	* There is nothing to do for SA_REMOVE,
	* so it is just skipped.
	*/
	if (action == SA_REMOVE)
	continue;

	/*
	* Duplicate attributes are not allowed, so the
	* action can not be SA_ADD here.
	*/
	ASSERT3S(action, ==, SA_REPLACE);

	/*
	* Only a variable-sized attribute can be
	* replaced here, and its size must be changing.
	*/
	ASSERT3U(reg_length, ==, 0);
	ASSERT3U(length, !=, buflen);
	SA_ADD_BULK_ATTR(attr_desc, j, attr,
	locator, datastart, buflen);
	} else {
	SA_ADD_BULK_ATTR(attr_desc, j, attr,
	NULL, (void *)
	(TOC_OFF(idx_tab->sa_idx_tab[attr]) +
	(uintptr_t)old_data[k]), length);
	}
	}
	if (k == 0 && hdl->sa_spill) {
	hdr = SA_GET_HDR(hdl, SA_SPILL);
	idx_tab = SA_IDX_TAB_GET(hdl, SA_SPILL);
	count = spill_attr_count;
	} else {
	break;
	}
	}
	if (action == SA_ADD) {
	reg_length = SA_REGISTERED_LEN(sa, newattr);
	IMPLY(reg_length != 0, reg_length == buflen);
	SA_ADD_BULK_ATTR(attr_desc, j, newattr, locator,
	datastart, buflen);
	}
	ASSERT3U(j, ==, attr_count);

	error = sa_build_layouts(hdl, attr_desc, attr_count, tx);

	if (old_data[0])
	kmem_free(old_data[0], bonus_data_size);
	if (old_data[1])
	vmem_free(old_data[1], spill_data_size);
	kmem_free(attr_desc, sizeof (sa_bulk_attr_t) * attr_count);

	return (error);
	}

	static int
	sa_bulk_update_impl(sa_handle_t hdl, sa_bulk_attr_t bulk, int count,
	dmu_tx_t *tx)
	{
	int error;
	sa_os_t *sa = hdl->sa_os->os_sa;
	dmu_object_type_t bonustype;
	dmu_buf_t *saved_spill;

	ASSERT(hdl);
	ASSERT(MUTEX_HELD(&hdl->sa_lock));

	bonustype = SA_BONUSTYPE_FROM_DB(SA_GET_DB(hdl, SA_BONUS));
	saved_spill = hdl->sa_spill;

	/* sync out registration table if necessary */
	if (sa->sa_need_attr_registration)
	sa_attr_register_sync(hdl, tx);

	error = sa_attr_op(hdl, bulk, count, SA_UPDATE, tx);
	if (error == 0 && !IS_SA_BONUSTYPE(bonustype) && sa->sa_update_cb)
	sa->sa_update_cb(hdl, tx);

	/*
	* If saved_spill is NULL and current sa_spill is not NULL that
	* means we increased the refcount of the spill buffer through
	* sa_get_spill() or dmu_spill_hold_by_dnode(). Therefore we
	* must release the hold before calling dmu_tx_commit() to avoid
	* making a copy of this buffer in dbuf_sync_leaf() due to the
	* reference count now being greater than 1.
	*/
	if (!saved_spill && hdl->sa_spill) {
	if (hdl->sa_spill_tab) {
	sa_idx_tab_rele(hdl->sa_os, hdl->sa_spill_tab);
	hdl->sa_spill_tab = NULL;
	}

	dmu_buf_rele(hdl->sa_spill, NULL);
	hdl->sa_spill = NULL;
	}

	return (error);
	}

	/*
	* update or add new attribute
	*/
	int
	sa_update(sa_handle_t *hdl, sa_attr_type_t type,
	void buf, uint32_t buflen, dmu_tx_t tx)
	{
	int error;
	sa_bulk_attr_t bulk;

	VERIFY3U(buflen, <=, SA_ATTR_MAX_LEN);

	bulk.sa_attr = type;
	bulk.sa_data_func = NULL;
	bulk.sa_length = buflen;
	bulk.sa_data = buf;

	mutex_enter(&hdl->sa_lock);
	error = sa_bulk_update_impl(hdl, &bulk, 1, tx);
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	/*
	* Return size of an attribute
	*/

	int
	sa_size(sa_handle_t hdl, sa_attr_type_t attr, int size)
	{
	sa_bulk_attr_t bulk;
	int error;

	bulk.sa_data = NULL;
	bulk.sa_attr = attr;
	bulk.sa_data_func = NULL;

	ASSERT(hdl);
	mutex_enter(&hdl->sa_lock);
	if ((error = sa_attr_op(hdl, &bulk, 1, SA_LOOKUP, NULL)) != 0) {
	mutex_exit(&hdl->sa_lock);
	return (error);
	}
	*size = bulk.sa_size;

	mutex_exit(&hdl->sa_lock);
	return (0);
	}

	int
	sa_bulk_lookup_locked(sa_handle_t hdl, sa_bulk_attr_t attrs, int count)
	{
	ASSERT(hdl);
	ASSERT(MUTEX_HELD(&hdl->sa_lock));
	return (sa_lookup_impl(hdl, attrs, count));
	}

	int
	sa_bulk_lookup(sa_handle_t hdl, sa_bulk_attr_t attrs, int count)
	{
	int error;

	ASSERT(hdl);
	mutex_enter(&hdl->sa_lock);
	error = sa_bulk_lookup_locked(hdl, attrs, count);
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	int
	sa_bulk_update(sa_handle_t hdl, sa_bulk_attr_t attrs, int count, dmu_tx_t *tx)
	{
	int error;

	ASSERT(hdl);
	mutex_enter(&hdl->sa_lock);
	error = sa_bulk_update_impl(hdl, attrs, count, tx);
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	int
	sa_remove(sa_handle_t hdl, sa_attr_type_t attr, dmu_tx_t tx)
	{
	int error;

	mutex_enter(&hdl->sa_lock);
	error = sa_modify_attrs(hdl, attr, SA_REMOVE, NULL,
	NULL, 0, tx);
	mutex_exit(&hdl->sa_lock);
	return (error);
	}

	void
	sa_object_info(sa_handle_t hdl, dmu_object_info_t doi)
	{
	dmu_object_info_from_db(hdl->sa_bonus, doi);
	}

	void
	sa_object_size(sa_handle_t hdl, uint32_t blksize, u_longlong_t *nblocks)
	{
	dmu_object_size_from_db(hdl->sa_bonus,
	blksize, nblocks);
	}

	void
	sa_set_userp(sa_handle_t hdl, void ptr)
	{
	hdl->sa_userp = ptr;
	}

	dmu_buf_t *
	sa_get_db(sa_handle_t *hdl)
	{
	return (hdl->sa_bonus);
	}

	void *
	sa_get_userdata(sa_handle_t *hdl)
	{
	return (hdl->sa_userp);
	}

	void
	sa_register_update_callback_locked(objset_t os, sa_update_cb_t func)
	{
	ASSERT(MUTEX_HELD(&os->os_sa->sa_lock));
	os->os_sa->sa_update_cb = func;
	}

	void
	sa_register_update_callback(objset_t os, sa_update_cb_t func)
	{

	mutex_enter(&os->os_sa->sa_lock);
	sa_register_update_callback_locked(os, func);
	mutex_exit(&os->os_sa->sa_lock);
	}

	uint64_t
	sa_handle_object(sa_handle_t *hdl)
	{
	return (hdl->sa_bonus->db_object);
	}

	boolean_t
	sa_enabled(objset_t *os)
	{
	return (os->os_sa == NULL);
	}

	int
	sa_set_sa_object(objset_t *os, uint64_t sa_object)
	{
	sa_os_t *sa = os->os_sa;

	if (sa->sa_master_obj)
	return (1);

	sa->sa_master_obj = sa_object;

	return (0);
	}

	int
	sa_hdrsize(void *arg)
	{
	sa_hdr_phys_t *hdr = arg;

	return (SA_HDR_SIZE(hdr));
	}

	void
	sa_handle_lock(sa_handle_t *hdl)
	{
	ASSERT(hdl);
	mutex_enter(&hdl->sa_lock);
	}

	void
	sa_handle_unlock(sa_handle_t *hdl)
	{
	ASSERT(hdl);
	mutex_exit(&hdl->sa_lock);
	}

	#ifdef _KERNEL
	EXPORT_SYMBOL(sa_handle_get);
	EXPORT_SYMBOL(sa_handle_get_from_db);
	EXPORT_SYMBOL(sa_handle_destroy);
	EXPORT_SYMBOL(sa_buf_hold);
	EXPORT_SYMBOL(sa_buf_rele);
	EXPORT_SYMBOL(sa_spill_rele);
	EXPORT_SYMBOL(sa_lookup);
	EXPORT_SYMBOL(sa_update);
	EXPORT_SYMBOL(sa_remove);
	EXPORT_SYMBOL(sa_bulk_lookup);
	EXPORT_SYMBOL(sa_bulk_lookup_locked);
	EXPORT_SYMBOL(sa_bulk_update);
	EXPORT_SYMBOL(sa_size);
	EXPORT_SYMBOL(sa_object_info);
	EXPORT_SYMBOL(sa_object_size);
	EXPORT_SYMBOL(sa_get_userdata);
	EXPORT_SYMBOL(sa_set_userp);
	EXPORT_SYMBOL(sa_get_db);
	EXPORT_SYMBOL(sa_handle_object);
	EXPORT_SYMBOL(sa_register_update_callback);
	EXPORT_SYMBOL(sa_setup);
	EXPORT_SYMBOL(sa_replace_all_by_template);
	EXPORT_SYMBOL(sa_replace_all_by_template_locked);
	EXPORT_SYMBOL(sa_enabled);
	EXPORT_SYMBOL(sa_cache_init);
	EXPORT_SYMBOL(sa_cache_fini);
	EXPORT_SYMBOL(sa_set_sa_object);
	EXPORT_SYMBOL(sa_hdrsize);
	EXPORT_SYMBOL(sa_handle_lock);
	EXPORT_SYMBOL(sa_handle_unlock);
	EXPORT_SYMBOL(sa_lookup_uio);
	EXPORT_SYMBOL(sa_add_projid);
	#endif /* _KERNEL */
	diff --git a/sys/contrib/openzfs/module/zfs/spa.c b/sys/contrib/openzfs/module/zfs/spa.c
	index eeec3b6be9ca..cc367745e486 100644
	--- a/sys/contrib/openzfs/module/zfs/spa.c
	+++ b/sys/contrib/openzfs/module/zfs/spa.c
	@@ -1,10036 +1,10043 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2018, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright 2018 Joyent, Inc.
	* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	/*
	* SPA: Storage Pool Allocator
	*
	* This file contains all the routines used when modifying on-disk SPA state.
	* This includes opening, importing, destroying, exporting a pool, and syncing a
	* pool.
	*/

	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/zil.h>
	#include <sys/ddt.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_removal.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/vdev_indirect_births.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_disk.h>
	#include <sys/vdev_draid.h>
	#include <sys/metaslab.h>
	#include <sys/metaslab_impl.h>
	#include <sys/mmp.h>
	#include <sys/uberblock_impl.h>
	#include <sys/txg.h>
	#include <sys/avl.h>
	#include <sys/bpobj.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dmu_objset.h>
	#include <sys/unique.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/fs/zfs.h>
	#include <sys/arc.h>
	#include <sys/callb.h>
	#include <sys/systeminfo.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/dsl_scan.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_destroy.h>
	#include <sys/zvol.h>

	#ifdef _KERNEL
	#include <sys/fm/protocol.h>
	#include <sys/fm/util.h>
	#include <sys/callb.h>
	#include <sys/zone.h>
	#include <sys/vmsystm.h>
	#endif /* _KERNEL */

	#include "zfs_prop.h"
	#include "zfs_comutil.h"

	/*
	* The interval, in seconds, at which failed configuration cache file writes
	* should be retried.
	*/
	int zfs_ccw_retry_interval = 300;

	typedef enum zti_modes {
	ZTI_MODE_FIXED, /* value is # of threads (min 1) */
	ZTI_MODE_BATCH, /* cpu-intensive; value is ignored */
	ZTI_MODE_SCALE, /* Taskqs scale with CPUs. */
	ZTI_MODE_NULL, /* don't create a taskq */
	ZTI_NMODES
	} zti_modes_t;

	#define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) }
	#define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 }
	#define ZTI_BATCH { ZTI_MODE_BATCH, 0, 1 }
	#define ZTI_SCALE { ZTI_MODE_SCALE, 0, 1 }
	#define ZTI_NULL { ZTI_MODE_NULL, 0, 0 }

	#define ZTI_N(n) ZTI_P(n, 1)
	#define ZTI_ONE ZTI_N(1)

	typedef struct zio_taskq_info {
	zti_modes_t zti_mode;
	uint_t zti_value;
	uint_t zti_count;
	} zio_taskq_info_t;

	static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
	"iss", "iss_h", "int", "int_h"
	};

	/*
	* This table defines the taskq settings for each ZFS I/O type. When
	* initializing a pool, we use this table to create an appropriately sized
	* taskq. Some operations are low volume and therefore have a small, static
	* number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
	* macros. Other operations process a large amount of data; the ZTI_BATCH
	* macro causes us to create a taskq oriented for throughput. Some operations
	* are so high frequency and short-lived that the taskq itself can become a
	* point of lock contention. The ZTI_P(#, #) macro indicates that we need an
	* additional degree of parallelism specified by the number of threads per-
	* taskq and the number of taskqs; when dispatching an event in this case, the
	* particular taskq is chosen at random. ZTI_SCALE is similar to ZTI_BATCH,
	* but with number of taskqs also scaling with number of CPUs.
	*
	* The different taskq priorities are to handle the different contexts (issue
	* and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
	* need to be handled with minimum delay.
	*/
	static const zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
	/* ISSUE ISSUE_HIGH INTR INTR_HIGH */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */
	{ ZTI_N(8), ZTI_NULL, ZTI_SCALE, ZTI_NULL }, /* READ */
	{ ZTI_BATCH, ZTI_N(5), ZTI_SCALE, ZTI_N(5) }, /* WRITE */
	{ ZTI_SCALE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* IOCTL */
	{ ZTI_N(4), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* TRIM */
	};

	static void spa_sync_version(void arg, dmu_tx_t tx);
	static void spa_sync_props(void arg, dmu_tx_t tx);
	static boolean_t spa_has_active_shared_spare(spa_t *spa);
	static int spa_load_impl(spa_t *spa, spa_import_type_t type,
	const char **ereport);
	static void spa_vdev_resilver_done(spa_t *spa);

	static uint_t zio_taskq_batch_pct = 80; /* 1 thread per cpu in pset */
	static uint_t zio_taskq_batch_tpq; /* threads per taskq */
	static const boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
	static const uint_t zio_taskq_basedc = 80; /* base duty cycle */

	static const boolean_t spa_create_process = B_TRUE; /* no process => no sysdc */

	/*
	* Report any spa_load_verify errors found, but do not fail spa_load.
	* This is used by zdb to analyze non-idle pools.
	*/
	boolean_t spa_load_verify_dryrun = B_FALSE;

	/*
	* Allow read spacemaps in case of readonly import (spa_mode == SPA_MODE_READ).
	* This is used by zdb for spacemaps verification.
	*/
	boolean_t spa_mode_readable_spacemaps = B_FALSE;

	/*
	* This (illegal) pool name is used when temporarily importing a spa_t in order
	* to get the vdev stats associated with the imported devices.
	*/
	#define TRYIMPORT_NAME "$import"

	/*
	* For debugging purposes: print out vdev tree during pool import.
	*/
	static int spa_load_print_vdev_tree = B_FALSE;

	/*
	* A non-zero value for zfs_max_missing_tvds means that we allow importing
	* pools with missing top-level vdevs. This is strictly intended for advanced
	* pool recovery cases since missing data is almost inevitable. Pools with
	* missing devices can only be imported read-only for safety reasons, and their
	* fail-mode will be automatically set to "continue".
	*
	* With 1 missing vdev we should be able to import the pool and mount all
	* datasets. User data that was not modified after the missing device has been
	* added should be recoverable. This means that snapshots created prior to the
	* addition of that device should be completely intact.
	*
	* With 2 missing vdevs, some datasets may fail to mount since there are
	* dataset statistics that are stored as regular metadata. Some data might be
	* recoverable if those vdevs were added recently.
	*
	* With 3 or more missing vdevs, the pool is severely damaged and MOS entries
	* may be missing entirely. Chances of data recovery are very low. Note that
	* there are also risks of performing an inadvertent rewind as we might be
	* missing all the vdevs with the latest uberblocks.
	*/
	unsigned long zfs_max_missing_tvds = 0;

	/*
	* The parameters below are similar to zfs_max_missing_tvds but are only
	* intended for a preliminary open of the pool with an untrusted config which
	* might be incomplete or out-dated.
	*
	* We are more tolerant for pools opened from a cachefile since we could have
	* an out-dated cachefile where a device removal was not registered.
	* We could have set the limit arbitrarily high but in the case where devices
	* are really missing we would want to return the proper error codes; we chose
	* SPA_DVAS_PER_BP - 1 so that some copies of the MOS would still be available
	* and we get a chance to retrieve the trusted config.
	*/
	uint64_t zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1;

	/*
	* In the case where config was assembled by scanning device paths (/dev/dsks
	* by default) we are less tolerant since all the existing devices should have
	* been detected and we want spa_load to return the right error codes.
	*/
	uint64_t zfs_max_missing_tvds_scan = 0;

	/*
	* Debugging aid that pauses spa_sync() towards the end.
	*/
	static const boolean_t zfs_pause_spa_sync = B_FALSE;

	/*
	* Variables to indicate the livelist condense zthr func should wait at certain
	* points for the livelist to be removed - used to test condense/destroy races
	*/
	static int zfs_livelist_condense_zthr_pause = 0;
	static int zfs_livelist_condense_sync_pause = 0;

	/*
	* Variables to track whether or not condense cancellation has been
	* triggered in testing.
	*/
	static int zfs_livelist_condense_sync_cancel = 0;
	static int zfs_livelist_condense_zthr_cancel = 0;

	/*
	* Variable to track whether or not extra ALLOC blkptrs were added to a
	* livelist entry while it was being condensed (caused by the way we track
	* remapped blkptrs in dbuf_remap_impl)
	*/
	static int zfs_livelist_condense_new_alloc = 0;

	/*
	* ==========================================================================
	* SPA properties routines
	* ==========================================================================
	*/

	/*
	* Add a (source=src, propname=propval) list to an nvlist.
	*/
	static void
	spa_prop_add_list(nvlist_t nvl, zpool_prop_t prop, const char strval,
	uint64_t intval, zprop_source_t src)
	{
	const char *propname = zpool_prop_to_name(prop);
	nvlist_t *propval;

	propval = fnvlist_alloc();
	fnvlist_add_uint64(propval, ZPROP_SOURCE, src);

	if (strval != NULL)
	fnvlist_add_string(propval, ZPROP_VALUE, strval);
	else
	fnvlist_add_uint64(propval, ZPROP_VALUE, intval);

	fnvlist_add_nvlist(nvl, propname, propval);
	nvlist_free(propval);
	}

	/*
	* Get property values from the spa configuration.
	*/
	static void
	spa_prop_get_config(spa_t spa, nvlist_t *nvp)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	dsl_pool_t *pool = spa->spa_dsl_pool;
	uint64_t size, alloc, cap, version;
	const zprop_source_t src = ZPROP_SRC_NONE;
	spa_config_dirent_t *dp;
	metaslab_class_t *mc = spa_normal_class(spa);

	ASSERT(MUTEX_HELD(&spa->spa_props_lock));

	if (rvd != NULL) {
	alloc = metaslab_class_get_alloc(mc);
	alloc += metaslab_class_get_alloc(spa_special_class(spa));
	alloc += metaslab_class_get_alloc(spa_dedup_class(spa));
	alloc += metaslab_class_get_alloc(spa_embedded_log_class(spa));

	size = metaslab_class_get_space(mc);
	size += metaslab_class_get_space(spa_special_class(spa));
	size += metaslab_class_get_space(spa_dedup_class(spa));
	size += metaslab_class_get_space(spa_embedded_log_class(spa));

	spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL,
	size - alloc, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_CHECKPOINT, NULL,
	spa->spa_checkpoint_info.sci_dspace, src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL,
	metaslab_class_fragmentation(mc), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL,
	metaslab_class_expandable_space(mc), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL,
	(spa_mode(spa) == SPA_MODE_READ), src);

	cap = (size == 0) ? 0 : (alloc * 100 / size);
	spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL,
	ddt_get_pool_dedup_ratio(spa), src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
	rvd->vdev_state, src);

	version = spa_version(spa);
	if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
	version, ZPROP_SRC_DEFAULT);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
	version, ZPROP_SRC_LOCAL);
	}
	spa_prop_add_list(*nvp, ZPOOL_PROP_LOAD_GUID,
	NULL, spa_load_guid(spa), src);
	}

	if (pool != NULL) {
	/*
	* The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
	* when opening pools before this version freedir will be NULL.
	*/
	if (pool->dp_free_dir != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
	dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes,
	src);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
	NULL, 0, src);
	}

	if (pool->dp_leak_dir != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL,
	dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes,
	src);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED,
	NULL, 0, src);
	}
	}

	spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);

	if (spa->spa_comment != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment,
	0, ZPROP_SRC_LOCAL);
	}

	if (spa->spa_compatibility != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_COMPATIBILITY,
	spa->spa_compatibility, 0, ZPROP_SRC_LOCAL);
	}

	if (spa->spa_root != NULL)
	spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
	0, ZPROP_SRC_LOCAL);

	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
	MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
	SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
	}

	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
	DNODE_MAX_SIZE, ZPROP_SRC_NONE);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
	DNODE_MIN_SIZE, ZPROP_SRC_NONE);
	}

	if ((dp = list_head(&spa->spa_config_list)) != NULL) {
	if (dp->scd_path == NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
	"none", 0, ZPROP_SRC_LOCAL);
	} else if (strcmp(dp->scd_path, spa_config_path) != 0) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
	dp->scd_path, 0, ZPROP_SRC_LOCAL);
	}
	}
	}

	/*
	* Get zpool property values.
	*/
	int
	spa_prop_get(spa_t spa, nvlist_t *nvp)
	{
	objset_t *mos = spa->spa_meta_objset;
	zap_cursor_t zc;
	zap_attribute_t za;
	dsl_pool_t *dp;
	int err;

	err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP);
	if (err)
	return (err);

	dp = spa_get_dsl(spa);
	dsl_pool_config_enter(dp, FTAG);
	mutex_enter(&spa->spa_props_lock);

	/*
	* Get properties from the spa config.
	*/
	spa_prop_get_config(spa, nvp);

	/* If no pool property object, no more prop to get. */
	if (mos == NULL \|\| spa->spa_pool_props_object == 0)
	goto out;

	/*
	* Get properties from the MOS pool property object.
	*/
	for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object);
	(err = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t intval = 0;
	char *strval = NULL;
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	zpool_prop_t prop;

	if ((prop = zpool_name_to_prop(za.za_name)) == ZPOOL_PROP_INVAL)
	continue;

	switch (za.za_integer_length) {
	case 8:
	/* integer property */
	if (za.za_first_integer !=
	zpool_prop_default_numeric(prop))
	src = ZPROP_SRC_LOCAL;

	if (prop == ZPOOL_PROP_BOOTFS) {
	dsl_dataset_t *ds = NULL;

	err = dsl_dataset_hold_obj(dp,
	za.za_first_integer, FTAG, &ds);
	if (err != 0)
	break;

	strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN,
	KM_SLEEP);
	dsl_dataset_name(ds, strval);
	dsl_dataset_rele(ds, FTAG);
	} else {
	strval = NULL;
	intval = za.za_first_integer;
	}

	spa_prop_add_list(*nvp, prop, strval, intval, src);

	if (strval != NULL)
	kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN);

	break;

	case 1:
	/* string property */
	strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
	err = zap_lookup(mos, spa->spa_pool_props_object,
	za.za_name, 1, za.za_num_integers, strval);
	if (err) {
	kmem_free(strval, za.za_num_integers);
	break;
	}
	spa_prop_add_list(*nvp, prop, strval, 0, src);
	kmem_free(strval, za.za_num_integers);
	break;

	default:
	break;
	}
	}
	zap_cursor_fini(&zc);
	out:
	mutex_exit(&spa->spa_props_lock);
	dsl_pool_config_exit(dp, FTAG);
	if (err && err != ENOENT) {
	nvlist_free(*nvp);
	*nvp = NULL;
	return (err);
	}

	return (0);
	}

	/*
	* Validate the given pool properties nvlist and modify the list
	* for the property values to be set.
	*/
	static int
	spa_prop_validate(spa_t spa, nvlist_t props)
	{
	nvpair_t *elem;
	int error = 0, reset_bootfs = 0;
	uint64_t objnum = 0;
	boolean_t has_feature = B_FALSE;

	elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	uint64_t intval;
	char strval, slash, check, fname;
	const char *propname = nvpair_name(elem);
	zpool_prop_t prop = zpool_name_to_prop(propname);

	switch (prop) {
	case ZPOOL_PROP_INVAL:
	if (!zpool_prop_feature(propname)) {
	error = SET_ERROR(EINVAL);
	break;
	}

	/*
	* Sanitize the input.
	*/
	if (nvpair_type(elem) != DATA_TYPE_UINT64) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (intval != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	fname = strchr(propname, '@') + 1;
	if (zfeature_lookup_name(fname, NULL) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	has_feature = B_TRUE;
	break;

	case ZPOOL_PROP_VERSION:
	error = nvpair_value_uint64(elem, &intval);
	if (!error &&
	(intval < spa_version(spa) \|\|
	intval > SPA_VERSION_BEFORE_FEATURES \|\|
	has_feature))
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_DELEGATION:
	case ZPOOL_PROP_AUTOREPLACE:
	case ZPOOL_PROP_LISTSNAPS:
	case ZPOOL_PROP_AUTOEXPAND:
	case ZPOOL_PROP_AUTOTRIM:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > 1)
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_MULTIHOST:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > 1)
	error = SET_ERROR(EINVAL);

	if (!error) {
	uint32_t hostid = zone_get_hostid(NULL);
	if (hostid)
	spa->spa_hostid = hostid;
	else
	error = SET_ERROR(ENOTSUP);
	}

	break;

	case ZPOOL_PROP_BOOTFS:
	/*
	* If the pool version is less than SPA_VERSION_BOOTFS,
	* or the pool is still being created (version == 0),
	* the bootfs property cannot be set.
	*/
	if (spa_version(spa) < SPA_VERSION_BOOTFS) {
	error = SET_ERROR(ENOTSUP);
	break;
	}

	/*
	* Make sure the vdev config is bootable
	*/
	if (!vdev_is_bootable(spa->spa_root_vdev)) {
	error = SET_ERROR(ENOTSUP);
	break;
	}

	reset_bootfs = 1;

	error = nvpair_value_string(elem, &strval);

	if (!error) {
	objset_t *os;

	if (strval == NULL \|\| strval[0] == '\0') {
	objnum = zpool_prop_default_numeric(
	ZPOOL_PROP_BOOTFS);
	break;
	}

	error = dmu_objset_hold(strval, FTAG, &os);
	if (error != 0)
	break;

	/* Must be ZPL. */
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	error = SET_ERROR(ENOTSUP);
	} else {
	objnum = dmu_objset_id(os);
	}
	dmu_objset_rele(os, FTAG);
	}
	break;

	case ZPOOL_PROP_FAILUREMODE:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > ZIO_FAILURE_MODE_PANIC)
	error = SET_ERROR(EINVAL);

	/*
	* This is a special case which only occurs when
	* the pool has completely failed. This allows
	* the user to change the in-core failmode property
	* without syncing it out to disk (I/Os might
	* currently be blocked). We do this by returning
	* EIO to the caller (spa_prop_set) to trick it
	* into thinking we encountered a property validation
	* error.
	*/
	if (!error && spa_suspended(spa)) {
	spa->spa_failmode = intval;
	error = SET_ERROR(EIO);
	}
	break;

	case ZPOOL_PROP_CACHEFILE:
	if ((error = nvpair_value_string(elem, &strval)) != 0)
	break;

	if (strval[0] == '\0')
	break;

	if (strcmp(strval, "none") == 0)
	break;

	if (strval[0] != '/') {
	error = SET_ERROR(EINVAL);
	break;
	}

	slash = strrchr(strval, '/');
	ASSERT(slash != NULL);

	if (slash[1] == '\0' \|\| strcmp(slash, "/.") == 0 \|\|
	strcmp(slash, "/..") == 0)
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_COMMENT:
	if ((error = nvpair_value_string(elem, &strval)) != 0)
	break;
	for (check = strval; *check != '\0'; check++) {
	if (!isprint(*check)) {
	error = SET_ERROR(EINVAL);
	break;
	}
	}
	if (strlen(strval) > ZPROP_MAX_COMMENT)
	error = SET_ERROR(E2BIG);
	break;

	default:
	break;
	}

	if (error)
	break;
	}

	(void) nvlist_remove_all(props,
	zpool_prop_to_name(ZPOOL_PROP_DEDUPDITTO));

	if (!error && reset_bootfs) {
	error = nvlist_remove(props,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING);

	if (!error) {
	error = nvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum);
	}
	}

	return (error);
	}

	void
	spa_configfile_set(spa_t spa, nvlist_t nvp, boolean_t need_sync)
	{
	char *cachefile;
	spa_config_dirent_t *dp;

	if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE),
	&cachefile) != 0)
	return;

	dp = kmem_alloc(sizeof (spa_config_dirent_t),
	KM_SLEEP);

	if (cachefile[0] == '\0')
	dp->scd_path = spa_strdup(spa_config_path);
	else if (strcmp(cachefile, "none") == 0)
	dp->scd_path = NULL;
	else
	dp->scd_path = spa_strdup(cachefile);

	list_insert_head(&spa->spa_config_list, dp);
	if (need_sync)
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	int
	spa_prop_set(spa_t spa, nvlist_t nvp)
	{
	int error;
	nvpair_t *elem = NULL;
	boolean_t need_sync = B_FALSE;

	if ((error = spa_prop_validate(spa, nvp)) != 0)
	return (error);

	while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
	zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));

	if (prop == ZPOOL_PROP_CACHEFILE \|\|
	prop == ZPOOL_PROP_ALTROOT \|\|
	prop == ZPOOL_PROP_READONLY)
	continue;

	if (prop == ZPOOL_PROP_VERSION \|\| prop == ZPOOL_PROP_INVAL) {
	uint64_t ver = 0;

	if (prop == ZPOOL_PROP_VERSION) {
	VERIFY(nvpair_value_uint64(elem, &ver) == 0);
	} else {
	ASSERT(zpool_prop_feature(nvpair_name(elem)));
	ver = SPA_VERSION_FEATURES;
	need_sync = B_TRUE;
	}

	/* Save time if the version is already set. */
	if (ver == spa_version(spa))
	continue;

	/*
	* In addition to the pool directory object, we might
	* create the pool properties object, the features for
	* read object, the features for write object, or the
	* feature descriptions object.
	*/
	error = dsl_sync_task(spa->spa_name, NULL,
	spa_sync_version, &ver,
	6, ZFS_SPACE_CHECK_RESERVED);
	if (error)
	return (error);
	continue;
	}

	need_sync = B_TRUE;
	break;
	}

	if (need_sync) {
	return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
	nvp, 6, ZFS_SPACE_CHECK_RESERVED));
	}

	return (0);
	}

	/*
	* If the bootfs property value is dsobj, clear it.
	*/
	void
	spa_prop_clear_bootfs(spa_t spa, uint64_t dsobj, dmu_tx_t tx)
	{
	if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) {
	VERIFY(zap_remove(spa->spa_meta_objset,
	spa->spa_pool_props_object,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0);
	spa->spa_bootfs = 0;
	}
	}

	static int
	spa_change_guid_check(void arg, dmu_tx_t tx)
	{
	uint64_t *newguid __maybe_unused = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t vdev_state;

	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	int error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (SET_ERROR(error));
	}

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	vdev_state = rvd->vdev_state;
	spa_config_exit(spa, SCL_STATE, FTAG);

	if (vdev_state != VDEV_STATE_HEALTHY)
	return (SET_ERROR(ENXIO));

	ASSERT3U(spa_guid(spa), !=, *newguid);

	return (0);
	}

	static void
	spa_change_guid_sync(void arg, dmu_tx_t tx)
	{
	uint64_t *newguid = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	uint64_t oldguid;
	vdev_t *rvd = spa->spa_root_vdev;

	oldguid = spa_guid(spa);

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	rvd->vdev_guid = *newguid;
	rvd->vdev_guid_sum += (*newguid - oldguid);
	vdev_config_dirty(rvd);
	spa_config_exit(spa, SCL_STATE, FTAG);

	spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu",
	(u_longlong_t)oldguid, (u_longlong_t)*newguid);
	}

	/*
	* Change the GUID for the pool. This is done so that we can later
	* re-import a pool built from a clone of our own vdevs. We will modify
	* the root vdev's guid, our own pool guid, and then mark all of our
	* vdevs dirty. Note that we must make sure that all our vdevs are
	* online when we do this, or else any vdevs that weren't present
	* would be orphaned from our pool. We are also going to issue a
	* sysevent to update any watchers.
	*/
	int
	spa_change_guid(spa_t *spa)
	{
	int error;
	uint64_t guid;

	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);
	guid = spa_generate_guid(NULL);

	error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
	spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED);

	if (error == 0) {
	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ /*
	+ * Clear the kobj flag from all the vdevs to allow
	+ * vdev_cache_process_kobj_evt() to post events to all the
	+ * vdevs since GUID is updated.
	+ */
	+ vdev_clear_kobj_evt(spa->spa_root_vdev);
	+ for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
	+ vdev_clear_kobj_evt(spa->spa_l2cache.sav_vdevs[i]);
	+
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID);
	}

	mutex_exit(&spa_namespace_lock);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* ==========================================================================
	* SPA state manipulation (open/create/destroy/import/export)
	* ==========================================================================
	*/

	static int
	spa_error_entry_compare(const void a, const void b)
	{
	const spa_error_entry_t sa = (const spa_error_entry_t )a;
	const spa_error_entry_t sb = (const spa_error_entry_t )b;
	int ret;

	ret = memcmp(&sa->se_bookmark, &sb->se_bookmark,
	sizeof (zbookmark_phys_t));

	return (TREE_ISIGN(ret));
	}

	/*
	* Utility function which retrieves copies of the current logs and
	* re-initializes them in the process.
	*/
	void
	spa_get_errlists(spa_t spa, avl_tree_t last, avl_tree_t *scrub)
	{
	ASSERT(MUTEX_HELD(&spa->spa_errlist_lock));

	memcpy(last, &spa->spa_errlist_last, sizeof (avl_tree_t));
	memcpy(scrub, &spa->spa_errlist_scrub, sizeof (avl_tree_t));

	avl_create(&spa->spa_errlist_scrub,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_last,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	}

	static void
	spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
	{
	const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
	enum zti_modes mode = ztip->zti_mode;
	uint_t value = ztip->zti_value;
	uint_t count = ztip->zti_count;
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	uint_t cpus, flags = TASKQ_DYNAMIC;
	boolean_t batch = B_FALSE;

	switch (mode) {
	case ZTI_MODE_FIXED:
	ASSERT3U(value, >, 0);
	break;

	case ZTI_MODE_BATCH:
	batch = B_TRUE;
	flags \|= TASKQ_THREADS_CPU_PCT;
	value = MIN(zio_taskq_batch_pct, 100);
	break;

	case ZTI_MODE_SCALE:
	flags \|= TASKQ_THREADS_CPU_PCT;
	/*
	* We want more taskqs to reduce lock contention, but we want
	* less for better request ordering and CPU utilization.
	*/
	cpus = MAX(1, boot_ncpus * zio_taskq_batch_pct / 100);
	if (zio_taskq_batch_tpq > 0) {
	count = MAX(1, (cpus + zio_taskq_batch_tpq / 2) /
	zio_taskq_batch_tpq);
	} else {
	/*
	* Prefer 6 threads per taskq, but no more taskqs
	* than threads in them on large systems. For 80%:
	*
	* taskq taskq total
	* cpus taskqs percent threads threads
	* ------- ------- ------- ------- -------
	* 1 1 80% 1 1
	* 2 1 80% 1 1
	* 4 1 80% 3 3
	* 8 2 40% 3 6
	* 16 3 27% 4 12
	* 32 5 16% 5 25
	* 64 7 11% 7 49
	* 128 10 8% 10 100
	* 256 14 6% 15 210
	*/
	count = 1 + cpus / 6;
	while (count * count > cpus)
	count--;
	}
	/* Limit each taskq within 100% to not trigger assertion. */
	count = MAX(count, (zio_taskq_batch_pct + 99) / 100);
	value = (zio_taskq_batch_pct + count / 2) / count;
	break;

	case ZTI_MODE_NULL:
	tqs->stqs_count = 0;
	tqs->stqs_taskq = NULL;
	return;

	default:
	panic("unrecognized mode for %s_%s taskq (%u:%u) in "
	"spa_activate()",
	zio_type_name[t], zio_taskq_types[q], mode, value);
	break;
	}

	ASSERT3U(count, >, 0);
	tqs->stqs_count = count;
	tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);

	for (uint_t i = 0; i < count; i++) {
	taskq_t *tq;
	char name[32];

	if (count > 1)
	(void) snprintf(name, sizeof (name), "%s_%s_%u",
	zio_type_name[t], zio_taskq_types[q], i);
	else
	(void) snprintf(name, sizeof (name), "%s_%s",
	zio_type_name[t], zio_taskq_types[q]);

	if (zio_taskq_sysdc && spa->spa_proc != &p0) {
	if (batch)
	flags \|= TASKQ_DC_BATCH;

	(void) zio_taskq_basedc;
	tq = taskq_create_sysdc(name, value, 50, INT_MAX,
	spa->spa_proc, zio_taskq_basedc, flags);
	} else {
	pri_t pri = maxclsyspri;
	/*
	* The write issue taskq can be extremely CPU
	* intensive. Run it at slightly less important
	* priority than the other taskqs.
	*
	* Under Linux and FreeBSD this means incrementing
	* the priority value as opposed to platforms like
	* illumos where it should be decremented.
	*
	* On FreeBSD, if priorities divided by four (RQ_PPQ)
	* are equal then a difference between them is
	* insignificant.
	*/
	if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE) {
	#if defined(__linux__)
	pri++;
	#elif defined(__FreeBSD__)
	pri += 4;
	#else
	#error "unknown OS"
	#endif
	}
	tq = taskq_create_proc(name, value, pri, 50,
	INT_MAX, spa->spa_proc, flags);
	}

	tqs->stqs_taskq[i] = tq;
	}
	}

	static void
	spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
	{
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];

	if (tqs->stqs_taskq == NULL) {
	ASSERT3U(tqs->stqs_count, ==, 0);
	return;
	}

	for (uint_t i = 0; i < tqs->stqs_count; i++) {
	ASSERT3P(tqs->stqs_taskq[i], !=, NULL);
	taskq_destroy(tqs->stqs_taskq[i]);
	}

	kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *));
	tqs->stqs_taskq = NULL;
	}

	/*
	* Dispatch a task to the appropriate taskq for the ZFS I/O type and priority.
	* Note that a type may have multiple discrete taskqs to avoid lock contention
	* on the taskq itself. In that case we choose which taskq at random by using
	* the low bits of gethrtime().
	*/
	void
	spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	task_func_t func, void arg, uint_t flags, taskq_ent_t *ent)
	{
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	taskq_t *tq;

	ASSERT3P(tqs->stqs_taskq, !=, NULL);
	ASSERT3U(tqs->stqs_count, !=, 0);

	if (tqs->stqs_count == 1) {
	tq = tqs->stqs_taskq[0];
	} else {
	tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
	}

	taskq_dispatch_ent(tq, func, arg, flags, ent);
	}

	/*
	* Same as spa_taskq_dispatch_ent() but block on the task until completion.
	*/
	void
	spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	task_func_t func, void arg, uint_t flags)
	{
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	taskq_t *tq;
	taskqid_t id;

	ASSERT3P(tqs->stqs_taskq, !=, NULL);
	ASSERT3U(tqs->stqs_count, !=, 0);

	if (tqs->stqs_count == 1) {
	tq = tqs->stqs_taskq[0];
	} else {
	tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
	}

	id = taskq_dispatch(tq, func, arg, flags);
	if (id)
	taskq_wait_id(tq, id);
	}

	static void
	spa_create_zio_taskqs(spa_t *spa)
	{
	for (int t = 0; t < ZIO_TYPES; t++) {
	for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
	spa_taskqs_init(spa, t, q);
	}
	}
	}

	/*
	* Disabled until spa_thread() can be adapted for Linux.
	*/
	#undef HAVE_SPA_THREAD

	#if defined(_KERNEL) && defined(HAVE_SPA_THREAD)
	static void
	spa_thread(void *arg)
	{
	psetid_t zio_taskq_psrset_bind = PS_NONE;
	callb_cpr_t cprinfo;

	spa_t *spa = arg;
	user_t *pu = PTOU(curproc);

	CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr,
	spa->spa_name);

	ASSERT(curproc != &p0);
	(void) snprintf(pu->u_psargs, sizeof (pu->u_psargs),
	"zpool-%s", spa->spa_name);
	(void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm));

	/* bind this thread to the requested psrset */
	if (zio_taskq_psrset_bind != PS_NONE) {
	pool_lock();
	mutex_enter(&cpu_lock);
	mutex_enter(&pidlock);
	mutex_enter(&curproc->p_lock);

	if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind,
	0, NULL, NULL) == 0) {
	curthread->t_bind_pset = zio_taskq_psrset_bind;
	} else {
	cmn_err(CE_WARN,
	"Couldn't bind process for zfs pool \"%s\" to "
	"pset %d\n", spa->spa_name, zio_taskq_psrset_bind);
	}

	mutex_exit(&curproc->p_lock);
	mutex_exit(&pidlock);
	mutex_exit(&cpu_lock);
	pool_unlock();
	}

	if (zio_taskq_sysdc) {
	sysdc_thread_enter(curthread, 100, 0);
	}

	spa->spa_proc = curproc;
	spa->spa_did = curthread->t_did;

	spa_create_zio_taskqs(spa);

	mutex_enter(&spa->spa_proc_lock);
	ASSERT(spa->spa_proc_state == SPA_PROC_CREATED);

	spa->spa_proc_state = SPA_PROC_ACTIVE;
	cv_broadcast(&spa->spa_proc_cv);

	CALLB_CPR_SAFE_BEGIN(&cprinfo);
	while (spa->spa_proc_state == SPA_PROC_ACTIVE)
	cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
	CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock);

	ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE);
	spa->spa_proc_state = SPA_PROC_GONE;
	spa->spa_proc = &p0;
	cv_broadcast(&spa->spa_proc_cv);
	CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */

	mutex_enter(&curproc->p_lock);
	lwp_exit();
	}
	#endif

	/*
	* Activate an uninitialized pool.
	*/
	static void
	spa_activate(spa_t *spa, spa_mode_t mode)
	{
	ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);

	spa->spa_state = POOL_STATE_ACTIVE;
	spa->spa_mode = mode;
	spa->spa_read_spacemaps = spa_mode_readable_spacemaps;

	spa->spa_normal_class = metaslab_class_create(spa, &zfs_metaslab_ops);
	spa->spa_log_class = metaslab_class_create(spa, &zfs_metaslab_ops);
	spa->spa_embedded_log_class =
	metaslab_class_create(spa, &zfs_metaslab_ops);
	spa->spa_special_class = metaslab_class_create(spa, &zfs_metaslab_ops);
	spa->spa_dedup_class = metaslab_class_create(spa, &zfs_metaslab_ops);

	/* Try to create a covering process */
	mutex_enter(&spa->spa_proc_lock);
	ASSERT(spa->spa_proc_state == SPA_PROC_NONE);
	ASSERT(spa->spa_proc == &p0);
	spa->spa_did = 0;

	(void) spa_create_process;
	#ifdef HAVE_SPA_THREAD
	/* Only create a process if we're going to be around a while. */
	if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) {
	if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri,
	NULL, 0) == 0) {
	spa->spa_proc_state = SPA_PROC_CREATED;
	while (spa->spa_proc_state == SPA_PROC_CREATED) {
	cv_wait(&spa->spa_proc_cv,
	&spa->spa_proc_lock);
	}
	ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
	ASSERT(spa->spa_proc != &p0);
	ASSERT(spa->spa_did != 0);
	} else {
	#ifdef _KERNEL
	cmn_err(CE_WARN,
	"Couldn't create process for zfs pool \"%s\"\n",
	spa->spa_name);
	#endif
	}
	}
	#endif /* HAVE_SPA_THREAD */
	mutex_exit(&spa->spa_proc_lock);

	/* If we didn't create a process, we need to create our taskqs. */
	if (spa->spa_proc == &p0) {
	spa_create_zio_taskqs(spa);
	}

	for (size_t i = 0; i < TXG_SIZE; i++) {
	spa->spa_txg_zio[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);
	}

	list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_config_dirty_node));
	list_create(&spa->spa_evicting_os_list, sizeof (objset_t),
	offsetof(objset_t, os_evicting_node));
	list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_state_dirty_node));

	txg_list_create(&spa->spa_vdev_txg_list, spa,
	offsetof(struct vdev, vdev_txg_node));

	avl_create(&spa->spa_errlist_scrub,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_last,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_healed,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));

	spa_activate_os(spa);

	spa_keystore_init(&spa->spa_keystore);

	/*
	* This taskq is used to perform zvol-minor-related tasks
	* asynchronously. This has several advantages, including easy
	* resolution of various deadlocks.
	*
	* The taskq must be single threaded to ensure tasks are always
	* processed in the order in which they were dispatched.
	*
	* A taskq per pool allows one to keep the pools independent.
	* This way if one pool is suspended, it will not impact another.
	*
	* The preferred location to dispatch a zvol minor task is a sync
	* task. In this context, there is easy access to the spa_t and minimal
	* error handling is required because the sync task must succeed.
	*/
	spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri,
	1, INT_MAX, 0);

	/*
	* Taskq dedicated to prefetcher threads: this is used to prevent the
	* pool traverse code from monopolizing the global (and limited)
	* system_taskq by inappropriately scheduling long running tasks on it.
	*/
	spa->spa_prefetch_taskq = taskq_create("z_prefetch", 100,
	defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);

	/*
	* The taskq to upgrade datasets in this pool. Currently used by
	* feature SPA_FEATURE_USEROBJ_ACCOUNTING/SPA_FEATURE_PROJECT_QUOTA.
	*/
	spa->spa_upgrade_taskq = taskq_create("z_upgrade", 100,
	defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);
	}

	/*
	* Opposite of spa_activate().
	*/
	static void
	spa_deactivate(spa_t *spa)
	{
	ASSERT(spa->spa_sync_on == B_FALSE);
	ASSERT(spa->spa_dsl_pool == NULL);
	ASSERT(spa->spa_root_vdev == NULL);
	ASSERT(spa->spa_async_zio_root == NULL);
	ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED);

	spa_evicting_os_wait(spa);

	if (spa->spa_zvol_taskq) {
	taskq_destroy(spa->spa_zvol_taskq);
	spa->spa_zvol_taskq = NULL;
	}

	if (spa->spa_prefetch_taskq) {
	taskq_destroy(spa->spa_prefetch_taskq);
	spa->spa_prefetch_taskq = NULL;
	}

	if (spa->spa_upgrade_taskq) {
	taskq_destroy(spa->spa_upgrade_taskq);
	spa->spa_upgrade_taskq = NULL;
	}

	txg_list_destroy(&spa->spa_vdev_txg_list);

	list_destroy(&spa->spa_config_dirty_list);
	list_destroy(&spa->spa_evicting_os_list);
	list_destroy(&spa->spa_state_dirty_list);

	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);

	for (int t = 0; t < ZIO_TYPES; t++) {
	for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
	spa_taskqs_fini(spa, t, q);
	}
	}

	for (size_t i = 0; i < TXG_SIZE; i++) {
	ASSERT3P(spa->spa_txg_zio[i], !=, NULL);
	VERIFY0(zio_wait(spa->spa_txg_zio[i]));
	spa->spa_txg_zio[i] = NULL;
	}

	metaslab_class_destroy(spa->spa_normal_class);
	spa->spa_normal_class = NULL;

	metaslab_class_destroy(spa->spa_log_class);
	spa->spa_log_class = NULL;

	metaslab_class_destroy(spa->spa_embedded_log_class);
	spa->spa_embedded_log_class = NULL;

	metaslab_class_destroy(spa->spa_special_class);
	spa->spa_special_class = NULL;

	metaslab_class_destroy(spa->spa_dedup_class);
	spa->spa_dedup_class = NULL;

	/*
	* If this was part of an import or the open otherwise failed, we may
	* still have errors left in the queues. Empty them just in case.
	*/
	spa_errlog_drain(spa);
	avl_destroy(&spa->spa_errlist_scrub);
	avl_destroy(&spa->spa_errlist_last);
	avl_destroy(&spa->spa_errlist_healed);

	spa_keystore_fini(&spa->spa_keystore);

	spa->spa_state = POOL_STATE_UNINITIALIZED;

	mutex_enter(&spa->spa_proc_lock);
	if (spa->spa_proc_state != SPA_PROC_NONE) {
	ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
	spa->spa_proc_state = SPA_PROC_DEACTIVATE;
	cv_broadcast(&spa->spa_proc_cv);
	while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) {
	ASSERT(spa->spa_proc != &p0);
	cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
	}
	ASSERT(spa->spa_proc_state == SPA_PROC_GONE);
	spa->spa_proc_state = SPA_PROC_NONE;
	}
	ASSERT(spa->spa_proc == &p0);
	mutex_exit(&spa->spa_proc_lock);

	/*
	* We want to make sure spa_thread() has actually exited the ZFS
	* module, so that the module can't be unloaded out from underneath
	* it.
	*/
	if (spa->spa_did != 0) {
	thread_join(spa->spa_did);
	spa->spa_did = 0;
	}

	spa_deactivate_os(spa);

	}

	/*
	* Verify a pool configuration, and construct the vdev tree appropriately. This
	* will create all the necessary vdevs in the appropriate layout, with each vdev
	* in the CLOSED state. This will prep the pool before open/creation/import.
	* All vdev validation is done by the vdev_alloc() routine.
	*/
	int
	spa_config_parse(spa_t spa, vdev_t vdp, nvlist_t nv, vdev_t *parent,
	uint_t id, int atype)
	{
	nvlist_t **child;
	uint_t children;
	int error;

	if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0)
	return (error);

	if ((*vdp)->vdev_ops->vdev_op_leaf)
	return (0);

	error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children);

	if (error == ENOENT)
	return (0);

	if (error) {
	vdev_free(*vdp);
	*vdp = NULL;
	return (SET_ERROR(EINVAL));
	}

	for (int c = 0; c < children; c++) {
	vdev_t *vd;
	if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c,
	atype)) != 0) {
	vdev_free(*vdp);
	*vdp = NULL;
	return (error);
	}
	}

	ASSERT(*vdp != NULL);

	return (0);
	}

	static boolean_t
	spa_should_flush_logs_on_unload(spa_t *spa)
	{
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
	return (B_FALSE);

	if (!spa_writeable(spa))
	return (B_FALSE);

	if (!spa->spa_sync_on)
	return (B_FALSE);

	if (spa_state(spa) != POOL_STATE_EXPORTED)
	return (B_FALSE);

	if (zfs_keep_log_spacemaps_at_export)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Opens a transaction that will set the flag that will instruct
	* spa_sync to attempt to flush all the metaslabs for that txg.
	*/
	static void
	spa_unload_log_sm_flush_all(spa_t *spa)
	{
	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	ASSERT3U(spa->spa_log_flushall_txg, ==, 0);
	spa->spa_log_flushall_txg = dmu_tx_get_txg(tx);

	dmu_tx_commit(tx);
	txg_wait_synced(spa_get_dsl(spa), spa->spa_log_flushall_txg);
	}

	static void
	spa_unload_log_sm_metadata(spa_t *spa)
	{
	void *cookie = NULL;
	spa_log_sm_t *sls;
	while ((sls = avl_destroy_nodes(&spa->spa_sm_logs_by_txg,
	&cookie)) != NULL) {
	VERIFY0(sls->sls_mscount);
	kmem_free(sls, sizeof (spa_log_sm_t));
	}

	for (log_summary_entry_t *e = list_head(&spa->spa_log_summary);
	e != NULL; e = list_head(&spa->spa_log_summary)) {
	VERIFY0(e->lse_mscount);
	list_remove(&spa->spa_log_summary, e);
	kmem_free(e, sizeof (log_summary_entry_t));
	}

	spa->spa_unflushed_stats.sus_nblocks = 0;
	spa->spa_unflushed_stats.sus_memused = 0;
	spa->spa_unflushed_stats.sus_blocklimit = 0;
	}

	static void
	spa_destroy_aux_threads(spa_t *spa)
	{
	if (spa->spa_condense_zthr != NULL) {
	zthr_destroy(spa->spa_condense_zthr);
	spa->spa_condense_zthr = NULL;
	}
	if (spa->spa_checkpoint_discard_zthr != NULL) {
	zthr_destroy(spa->spa_checkpoint_discard_zthr);
	spa->spa_checkpoint_discard_zthr = NULL;
	}
	if (spa->spa_livelist_delete_zthr != NULL) {
	zthr_destroy(spa->spa_livelist_delete_zthr);
	spa->spa_livelist_delete_zthr = NULL;
	}
	if (spa->spa_livelist_condense_zthr != NULL) {
	zthr_destroy(spa->spa_livelist_condense_zthr);
	spa->spa_livelist_condense_zthr = NULL;
	}
	}

	/*
	* Opposite of spa_load().
	*/
	static void
	spa_unload(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_state(spa) != POOL_STATE_UNINITIALIZED);

	spa_import_progress_remove(spa_guid(spa));
	spa_load_note(spa, "UNLOADING");

	spa_wake_waiters(spa);

	/*
	* If we have set the spa_final_txg, we have already performed the
	* tasks below in spa_export_common(). We should not redo it here since
	* we delay the final TXGs beyond what spa_final_txg is set at.
	*/
	if (spa->spa_final_txg == UINT64_MAX) {
	/*
	* If the log space map feature is enabled and the pool is
	* getting exported (but not destroyed), we want to spend some
	* time flushing as many metaslabs as we can in an attempt to
	* destroy log space maps and save import time.
	*/
	if (spa_should_flush_logs_on_unload(spa))
	spa_unload_log_sm_flush_all(spa);

	/*
	* Stop async tasks.
	*/
	spa_async_suspend(spa);

	if (spa->spa_root_vdev) {
	vdev_t *root_vdev = spa->spa_root_vdev;
	vdev_initialize_stop_all(root_vdev,
	VDEV_INITIALIZE_ACTIVE);
	vdev_trim_stop_all(root_vdev, VDEV_TRIM_ACTIVE);
	vdev_autotrim_stop_all(spa);
	vdev_rebuild_stop_all(spa);
	}
	}

	/*
	* Stop syncing.
	*/
	if (spa->spa_sync_on) {
	txg_sync_stop(spa->spa_dsl_pool);
	spa->spa_sync_on = B_FALSE;
	}

	/*
	* This ensures that there is no async metaslab prefetching
	* while we attempt to unload the spa.
	*/
	if (spa->spa_root_vdev != NULL) {
	for (int c = 0; c < spa->spa_root_vdev->vdev_children; c++) {
	vdev_t *vc = spa->spa_root_vdev->vdev_child[c];
	if (vc->vdev_mg != NULL)
	taskq_wait(vc->vdev_mg->mg_taskq);
	}
	}

	if (spa->spa_mmp.mmp_thread)
	mmp_thread_stop(spa);

	/*
	* Wait for any outstanding async I/O to complete.
	*/
	if (spa->spa_async_zio_root != NULL) {
	for (int i = 0; i < max_ncpus; i++)
	(void) zio_wait(spa->spa_async_zio_root[i]);
	kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *));
	spa->spa_async_zio_root = NULL;
	}

	if (spa->spa_vdev_removal != NULL) {
	spa_vdev_removal_destroy(spa->spa_vdev_removal);
	spa->spa_vdev_removal = NULL;
	}

	spa_destroy_aux_threads(spa);

	spa_condense_fini(spa);

	bpobj_close(&spa->spa_deferred_bpobj);

	spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);

	/*
	* Close all vdevs.
	*/
	if (spa->spa_root_vdev)
	vdev_free(spa->spa_root_vdev);
	ASSERT(spa->spa_root_vdev == NULL);

	/*
	* Close the dsl pool.
	*/
	if (spa->spa_dsl_pool) {
	dsl_pool_close(spa->spa_dsl_pool);
	spa->spa_dsl_pool = NULL;
	spa->spa_meta_objset = NULL;
	}

	ddt_unload(spa);
	spa_unload_log_sm_metadata(spa);

	/*
	* Drop and purge level 2 cache
	*/
	spa_l2cache_drop(spa);

	for (int i = 0; i < spa->spa_spares.sav_count; i++)
	vdev_free(spa->spa_spares.sav_vdevs[i]);
	if (spa->spa_spares.sav_vdevs) {
	kmem_free(spa->spa_spares.sav_vdevs,
	spa->spa_spares.sav_count * sizeof (void *));
	spa->spa_spares.sav_vdevs = NULL;
	}
	if (spa->spa_spares.sav_config) {
	nvlist_free(spa->spa_spares.sav_config);
	spa->spa_spares.sav_config = NULL;
	}
	spa->spa_spares.sav_count = 0;

	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]);
	vdev_free(spa->spa_l2cache.sav_vdevs[i]);
	}
	if (spa->spa_l2cache.sav_vdevs) {
	kmem_free(spa->spa_l2cache.sav_vdevs,
	spa->spa_l2cache.sav_count * sizeof (void *));
	spa->spa_l2cache.sav_vdevs = NULL;
	}
	if (spa->spa_l2cache.sav_config) {
	nvlist_free(spa->spa_l2cache.sav_config);
	spa->spa_l2cache.sav_config = NULL;
	}
	spa->spa_l2cache.sav_count = 0;

	spa->spa_async_suspended = 0;

	spa->spa_indirect_vdevs_loaded = B_FALSE;

	if (spa->spa_comment != NULL) {
	spa_strfree(spa->spa_comment);
	spa->spa_comment = NULL;
	}
	if (spa->spa_compatibility != NULL) {
	spa_strfree(spa->spa_compatibility);
	spa->spa_compatibility = NULL;
	}

	spa_config_exit(spa, SCL_ALL, spa);
	}

	/*
	* Load (or re-load) the current list of vdevs describing the active spares for
	* this pool. When this is called, we have some form of basic information in
	* 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and
	* then re-generate a more complete list including status information.
	*/
	void
	spa_load_spares(spa_t *spa)
	{
	nvlist_t **spares;
	uint_t nspares;
	int i;
	vdev_t vd, tvd;

	#ifndef _KERNEL
	/*
	* zdb opens both the current state of the pool and the
	* checkpointed state (if present), with a different spa_t.
	*
	* As spare vdevs are shared among open pools, we skip loading
	* them when we load the checkpointed state of the pool.
	*/
	if (!spa_writeable(spa))
	return;
	#endif

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	/*
	* First, close and free any existing spare vdevs.
	*/
	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	vd = spa->spa_spares.sav_vdevs[i];

	/* Undo the call to spa_activate() below */
	if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
	B_FALSE)) != NULL && tvd->vdev_isspare)
	spa_spare_remove(tvd);
	vdev_close(vd);
	vdev_free(vd);
	}

	if (spa->spa_spares.sav_vdevs)
	kmem_free(spa->spa_spares.sav_vdevs,
	spa->spa_spares.sav_count * sizeof (void *));

	if (spa->spa_spares.sav_config == NULL)
	nspares = 0;
	else
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, &spares, &nspares));

	spa->spa_spares.sav_count = (int)nspares;
	spa->spa_spares.sav_vdevs = NULL;

	if (nspares == 0)
	return;

	/*
	* Construct the array of vdevs, opening them to get status in the
	* process. For each spare, there is potentially two different vdev_t
	* structures associated with it: one in the list of spares (used only
	* for basic validation purposes) and one in the active vdev
	* configuration (if it's spared in). During this phase we open and
	* validate each vdev on the spare list. If the vdev also exists in the
	* active configuration, then we also mark this vdev as an active spare.
	*/
	spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0,
	VDEV_ALLOC_SPARE) == 0);
	ASSERT(vd != NULL);

	spa->spa_spares.sav_vdevs[i] = vd;

	if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
	B_FALSE)) != NULL) {
	if (!tvd->vdev_isspare)
	spa_spare_add(tvd);

	/*
	* We only mark the spare active if we were successfully
	* able to load the vdev. Otherwise, importing a pool
	* with a bad active spare would result in strange
	* behavior, because multiple pool would think the spare
	* is actively in use.
	*
	* There is a vulnerability here to an equally bizarre
	* circumstance, where a dead active spare is later
	* brought back to life (onlined or otherwise). Given
	* the rarity of this scenario, and the extra complexity
	* it adds, we ignore the possibility.
	*/
	if (!vdev_is_dead(tvd))
	spa_spare_activate(tvd);
	}

	vd->vdev_top = vd;
	vd->vdev_aux = &spa->spa_spares;

	if (vdev_open(vd) != 0)
	continue;

	if (vdev_validate_aux(vd) == 0)
	spa_spare_add(vd);
	}

	/*
	* Recompute the stashed list of spares, with status information
	* this time.
	*/
	fnvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES);

	spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < spa->spa_spares.sav_count; i++)
	spares[i] = vdev_config_generate(spa,
	spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE);
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	spa->spa_spares.sav_count);
	for (i = 0; i < spa->spa_spares.sav_count; i++)
	nvlist_free(spares[i]);
	kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *));
	}

	/*
	* Load (or re-load) the current list of vdevs describing the active l2cache for
	* this pool. When this is called, we have some form of basic information in
	* 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and
	* then re-generate a more complete list including status information.
	* Devices which are already active have their details maintained, and are
	* not re-opened.
	*/
	void
	spa_load_l2cache(spa_t *spa)
	{
	nvlist_t **l2cache = NULL;
	uint_t nl2cache;
	int i, j, oldnvdevs;
	uint64_t guid;
	vdev_t vd, oldvdevs, *newvdevs;
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	#ifndef _KERNEL
	/*
	* zdb opens both the current state of the pool and the
	* checkpointed state (if present), with a different spa_t.
	*
	* As L2 caches are part of the ARC which is shared among open
	* pools, we skip loading them when we load the checkpointed
	* state of the pool.
	*/
	if (!spa_writeable(spa))
	return;
	#endif

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	oldvdevs = sav->sav_vdevs;
	oldnvdevs = sav->sav_count;
	sav->sav_vdevs = NULL;
	sav->sav_count = 0;

	if (sav->sav_config == NULL) {
	nl2cache = 0;
	newvdevs = NULL;
	goto out;
	}

	VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
	newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);

	/*
	* Process new nvlist of vdevs.
	*/
	for (i = 0; i < nl2cache; i++) {
	guid = fnvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID);

	newvdevs[i] = NULL;
	for (j = 0; j < oldnvdevs; j++) {
	vd = oldvdevs[j];
	if (vd != NULL && guid == vd->vdev_guid) {
	/*
	* Retain previous vdev for add/remove ops.
	*/
	newvdevs[i] = vd;
	oldvdevs[j] = NULL;
	break;
	}
	}

	if (newvdevs[i] == NULL) {
	/*
	* Create new vdev
	*/
	VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0,
	VDEV_ALLOC_L2CACHE) == 0);
	ASSERT(vd != NULL);
	newvdevs[i] = vd;

	/*
	* Commit this vdev as an l2cache device,
	* even if it fails to open.
	*/
	spa_l2cache_add(vd);

	vd->vdev_top = vd;
	vd->vdev_aux = sav;

	spa_l2cache_activate(vd);

	if (vdev_open(vd) != 0)
	continue;

	(void) vdev_validate_aux(vd);

	if (!vdev_is_dead(vd))
	l2arc_add_vdev(spa, vd);

	/*
	* Upon cache device addition to a pool or pool
	* creation with a cache device or if the header
	* of the device is invalid we issue an async
	* TRIM command for the whole device which will
	* execute if l2arc_trim_ahead > 0.
	*/
	spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
	}
	}

	sav->sav_vdevs = newvdevs;
	sav->sav_count = (int)nl2cache;

	/*
	* Recompute the stashed list of l2cache devices, with status
	* information this time.
	*/
	fnvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE);

	if (sav->sav_count > 0)
	l2cache = kmem_alloc(sav->sav_count * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < sav->sav_count; i++)
	l2cache[i] = vdev_config_generate(spa,
	sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE);
	fnvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
	(const nvlist_t * const *)l2cache, sav->sav_count);

	out:
	/*
	* Purge vdevs that were dropped
	*/
	for (i = 0; i < oldnvdevs; i++) {
	uint64_t pool;

	vd = oldvdevs[i];
	if (vd != NULL) {
	ASSERT(vd->vdev_isl2cache);

	if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
	pool != 0ULL && l2arc_vdev_present(vd))
	l2arc_remove_vdev(vd);
	vdev_clear_stats(vd);
	vdev_free(vd);
	}
	}

	if (oldvdevs)
	kmem_free(oldvdevs, oldnvdevs * sizeof (void *));

	for (i = 0; i < sav->sav_count; i++)
	nvlist_free(l2cache[i]);
	if (sav->sav_count)
	kmem_free(l2cache, sav->sav_count * sizeof (void *));
	}

	static int
	load_nvlist(spa_t spa, uint64_t obj, nvlist_t *value)
	{
	dmu_buf_t *db;
	char *packed = NULL;
	size_t nvsize = 0;
	int error;
	*value = NULL;

	error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db);
	if (error)
	return (error);

	nvsize = (uint64_t )db->db_data;
	dmu_buf_rele(db, FTAG);

	packed = vmem_alloc(nvsize, KM_SLEEP);
	error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed,
	DMU_READ_PREFETCH);
	if (error == 0)
	error = nvlist_unpack(packed, nvsize, value, 0);
	vmem_free(packed, nvsize);

	return (error);
	}

	/*
	* Concrete top-level vdevs that are not missing and are not logs. At every
	* spa_sync we write new uberblocks to at least SPA_SYNC_MIN_VDEVS core tvds.
	*/
	static uint64_t
	spa_healthy_core_tvds(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t tvds = 0;

	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];
	if (vd->vdev_islog)
	continue;
	if (vdev_is_concrete(vd) && !vdev_is_dead(vd))
	tvds++;
	}

	return (tvds);
	}

	/*
	* Checks to see if the given vdev could not be opened, in which case we post a
	* sysevent to notify the autoreplace code that the device has been removed.
	*/
	static void
	spa_check_removed(vdev_t *vd)
	{
	for (uint64_t c = 0; c < vd->vdev_children; c++)
	spa_check_removed(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
	vdev_is_concrete(vd)) {
	zfs_post_autoreplace(vd->vdev_spa, vd);
	spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK);
	}
	}

	static int
	spa_check_for_missing_logs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If we're doing a normal import, then build up any additional
	* diagnostic information about missing log devices.
	* We'll pass this up to the user for further processing.
	*/
	if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) {
	nvlist_t *child, nv;
	uint64_t idx = 0;

	child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t *),
	KM_SLEEP);
	nv = fnvlist_alloc();

	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	/*
	* We consider a device as missing only if it failed
	* to open (i.e. offline or faulted is not considered
	* as missing).
	*/
	if (tvd->vdev_islog &&
	tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
	child[idx++] = vdev_config_generate(spa, tvd,
	B_FALSE, VDEV_CONFIG_MISSING);
	}
	}

	if (idx > 0) {
	fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t * const *)child, idx);
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_MISSING_DEVICES, nv);

	for (uint64_t i = 0; i < idx; i++)
	nvlist_free(child[i]);
	}
	nvlist_free(nv);
	kmem_free(child, rvd->vdev_children * sizeof (char **));

	if (idx > 0) {
	spa_load_failed(spa, "some log devices are missing");
	vdev_dbgmsg_print_tree(rvd, 2);
	return (SET_ERROR(ENXIO));
	}
	} else {
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog &&
	tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	spa_load_note(spa, "some log devices are "
	"missing, ZIL is dropped.");
	vdev_dbgmsg_print_tree(rvd, 2);
	break;
	}
	}
	}

	return (0);
	}

	/*
	* Check for missing log devices
	*/
	static boolean_t
	spa_check_logs(spa_t *spa)
	{
	boolean_t rv = B_FALSE;
	dsl_pool_t *dp = spa_get_dsl(spa);

	switch (spa->spa_log_state) {
	default:
	break;
	case SPA_LOG_MISSING:
	/* need to recheck in case slog has been restored */
	case SPA_LOG_UNKNOWN:
	rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0);
	if (rv)
	spa_set_log_state(spa, SPA_LOG_MISSING);
	break;
	}
	return (rv);
	}

	/*
	* Passivate any log vdevs (note, does not apply to embedded log metaslabs).
	*/
	static boolean_t
	spa_passivate_log(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	boolean_t slog_found = B_FALSE;

	ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog) {
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_passivate(tvd->vdev_mg);
	slog_found = B_TRUE;
	}
	}

	return (slog_found);
	}

	/*
	* Activate any log vdevs (note, does not apply to embedded log metaslabs).
	*/
	static void
	spa_activate_log(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog) {
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_activate(tvd->vdev_mg);
	}
	}
	}

	int
	spa_reset_logs(spa_t *spa)
	{
	int error;

	error = dmu_objset_find(spa_name(spa), zil_reset,
	NULL, DS_FIND_CHILDREN);
	if (error == 0) {
	/*
	* We successfully offlined the log device, sync out the
	* current txg so that the "stubby" block can be removed
	* by zil_sync().
	*/
	txg_wait_synced(spa->spa_dsl_pool, 0);
	}
	return (error);
	}

	static void
	spa_aux_check_removed(spa_aux_vdev_t *sav)
	{
	for (int i = 0; i < sav->sav_count; i++)
	spa_check_removed(sav->sav_vdevs[i]);
	}

	void
	spa_claim_notify(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;

	if (zio->io_error)
	return;

	mutex_enter(&spa->spa_props_lock); /* any mutex will do */
	if (spa->spa_claim_max_txg < zio->io_bp->blk_birth)
	spa->spa_claim_max_txg = zio->io_bp->blk_birth;
	mutex_exit(&spa->spa_props_lock);
	}

	typedef struct spa_load_error {
	boolean_t sle_verify_data;
	uint64_t sle_meta_count;
	uint64_t sle_data_count;
	} spa_load_error_t;

	static void
	spa_load_verify_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	spa_load_error_t *sle = zio->io_private;
	dmu_object_type_t type = BP_GET_TYPE(bp);
	int error = zio->io_error;
	spa_t *spa = zio->io_spa;

	abd_free(zio->io_abd);
	if (error) {
	if ((BP_GET_LEVEL(bp) != 0 \|\| DMU_OT_IS_METADATA(type)) &&
	type != DMU_OT_INTENT_LOG)
	atomic_inc_64(&sle->sle_meta_count);
	else
	atomic_inc_64(&sle->sle_data_count);
	}

	mutex_enter(&spa->spa_scrub_lock);
	spa->spa_load_verify_bytes -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);
	}

	/*
	* Maximum number of inflight bytes is the log2 fraction of the arc size.
	* By default, we set it to 1/16th of the arc.
	*/
	-static int spa_load_verify_shift = 4;
	+static uint_t spa_load_verify_shift = 4;
	static int spa_load_verify_metadata = B_TRUE;
	static int spa_load_verify_data = B_TRUE;

	static int
	spa_load_verify_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	zio_t *rio = arg;
	spa_load_error_t *sle = rio->io_private;

	(void) zilog, (void) dnp;

	/*
	* Note: normally this routine will not be called if
	* spa_load_verify_metadata is not set. However, it may be useful
	* to manually set the flag after the traversal has begun.
	*/
	if (!spa_load_verify_metadata)
	return (0);

	/*
	* Sanity check the block pointer in order to detect obvious damage
	* before using the contents in subsequent checks or in zio_read().
	* When damaged consider it to be a metadata error since we cannot
	* trust the BP_GET_TYPE and BP_GET_LEVEL values.
	*/
	if (!zfs_blkptr_verify(spa, bp, B_FALSE, BLK_VERIFY_LOG)) {
	atomic_inc_64(&sle->sle_meta_count);
	return (0);
	}

	if (zb->zb_level == ZB_DNODE_LEVEL \|\| BP_IS_HOLE(bp) \|\|
	BP_IS_EMBEDDED(bp) \|\| BP_IS_REDACTED(bp))
	return (0);

	if (!BP_IS_METADATA(bp) &&
	(!spa_load_verify_data \|\| !sle->sle_verify_data))
	return (0);

	uint64_t maxinflight_bytes =
	arc_target_bytes() >> spa_load_verify_shift;
	size_t size = BP_GET_PSIZE(bp);

	mutex_enter(&spa->spa_scrub_lock);
	while (spa->spa_load_verify_bytes >= maxinflight_bytes)
	cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
	spa->spa_load_verify_bytes += size;
	mutex_exit(&spa->spa_scrub_lock);

	zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size,
	spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
	ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_SCRUB \| ZIO_FLAG_RAW, zb));
	return (0);
	}

	static int
	verify_dataset_name_len(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) dp, (void) arg;

	if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));

	return (0);
	}

	static int
	spa_load_verify(spa_t *spa)
	{
	zio_t *rio;
	spa_load_error_t sle = { 0 };
	zpool_load_policy_t policy;
	boolean_t verify_ok = B_FALSE;
	int error = 0;

	zpool_get_load_policy(spa->spa_config, &policy);

	if (policy.zlp_rewind & ZPOOL_NEVER_REWIND \|\|
	policy.zlp_maxmeta == UINT64_MAX)
	return (0);

	dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
	error = dmu_objset_find_dp(spa->spa_dsl_pool,
	spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL,
	DS_FIND_CHILDREN);
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	if (error != 0)
	return (error);

	/*
	* Verify data only if we are rewinding or error limit was set.
	* Otherwise nothing except dbgmsg care about it to waste time.
	*/
	sle.sle_verify_data = (policy.zlp_rewind & ZPOOL_REWIND_MASK) \|\|
	(policy.zlp_maxdata < UINT64_MAX);

	rio = zio_root(spa, NULL, &sle,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE);

	if (spa_load_verify_metadata) {
	if (spa->spa_extreme_rewind) {
	spa_load_note(spa, "performing a complete scan of the "
	"pool since extreme rewind is on. This may take "
	"a very long time.\n (spa_load_verify_data=%u, "
	"spa_load_verify_metadata=%u)",
	spa_load_verify_data, spa_load_verify_metadata);
	}

	error = traverse_pool(spa, spa->spa_verify_min_txg,
	TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \|
	TRAVERSE_NO_DECRYPT, spa_load_verify_cb, rio);
	}

	(void) zio_wait(rio);
	ASSERT0(spa->spa_load_verify_bytes);

	spa->spa_load_meta_errors = sle.sle_meta_count;
	spa->spa_load_data_errors = sle.sle_data_count;

	if (sle.sle_meta_count != 0 \|\| sle.sle_data_count != 0) {
	spa_load_note(spa, "spa_load_verify found %llu metadata errors "
	"and %llu data errors", (u_longlong_t)sle.sle_meta_count,
	(u_longlong_t)sle.sle_data_count);
	}

	if (spa_load_verify_dryrun \|\|
	(!error && sle.sle_meta_count <= policy.zlp_maxmeta &&
	sle.sle_data_count <= policy.zlp_maxdata)) {
	int64_t loss = 0;

	verify_ok = B_TRUE;
	spa->spa_load_txg = spa->spa_uberblock.ub_txg;
	spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp;

	loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts;
	fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_TIME,
	spa->spa_load_txg_ts);
	fnvlist_add_int64(spa->spa_load_info, ZPOOL_CONFIG_REWIND_TIME,
	loss);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_LOAD_META_ERRORS, sle.sle_meta_count);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count);
	} else {
	spa->spa_load_max_txg = spa->spa_uberblock.ub_txg;
	}

	if (spa_load_verify_dryrun)
	return (0);

	if (error) {
	if (error != ENXIO && error != EIO)
	error = SET_ERROR(EIO);
	return (error);
	}

	return (verify_ok ? 0 : EIO);
	}

	/*
	* Find a value in the pool props object.
	*/
	static void
	spa_prop_find(spa_t spa, zpool_prop_t prop, uint64_t val)
	{
	(void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object,
	zpool_prop_to_name(prop), sizeof (uint64_t), 1, val);
	}

	/*
	* Find a value in the pool directory object.
	*/
	static int
	spa_dir_prop(spa_t spa, const char name, uint64_t *val, boolean_t log_enoent)
	{
	int error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	name, sizeof (uint64_t), 1, val);

	if (error != 0 && (error != ENOENT \|\| log_enoent)) {
	spa_load_failed(spa, "couldn't get '%s' value in MOS directory "
	"[error=%d]", name, error);
	}

	return (error);
	}

	static int
	spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err)
	{
	vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux);
	return (SET_ERROR(err));
	}

	boolean_t
	spa_livelist_delete_check(spa_t *spa)
	{
	return (spa->spa_livelists_to_delete != 0);
	}

	static boolean_t
	spa_livelist_delete_cb_check(void arg, zthr_t z)
	{
	(void) z;
	spa_t *spa = arg;
	return (spa_livelist_delete_check(spa));
	}

	static int
	delete_blkptr_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	spa_t *spa = arg;
	zio_free(spa, tx->tx_txg, bp);
	dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
	-bp_get_dsize_sync(spa, bp),
	-BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
	return (0);
	}

	static int
	dsl_get_next_livelist_obj(objset_t os, uint64_t zap_obj, uint64_t llp)
	{
	int err;
	zap_cursor_t zc;
	zap_attribute_t za;
	zap_cursor_init(&zc, os, zap_obj);
	err = zap_cursor_retrieve(&zc, &za);
	zap_cursor_fini(&zc);
	if (err == 0)
	*llp = za.za_first_integer;
	return (err);
	}

	/*
	* Components of livelist deletion that must be performed in syncing
	* context: freeing block pointers and updating the pool-wide data
	* structures to indicate how much work is left to do
	*/
	typedef struct sublist_delete_arg {
	spa_t *spa;
	dsl_deadlist_t *ll;
	uint64_t key;
	bplist_t *to_free;
	} sublist_delete_arg_t;

	static void
	sublist_delete_sync(void arg, dmu_tx_t tx)
	{
	sublist_delete_arg_t *sda = arg;
	spa_t *spa = sda->spa;
	dsl_deadlist_t *ll = sda->ll;
	uint64_t key = sda->key;
	bplist_t *to_free = sda->to_free;

	bplist_iterate(to_free, delete_blkptr_cb, spa, tx);
	dsl_deadlist_remove_entry(ll, key, tx);
	}

	typedef struct livelist_delete_arg {
	spa_t *spa;
	uint64_t ll_obj;
	uint64_t zap_obj;
	} livelist_delete_arg_t;

	static void
	livelist_delete_sync(void arg, dmu_tx_t tx)
	{
	livelist_delete_arg_t *lda = arg;
	spa_t *spa = lda->spa;
	uint64_t ll_obj = lda->ll_obj;
	uint64_t zap_obj = lda->zap_obj;
	objset_t *mos = spa->spa_meta_objset;
	uint64_t count;

	/* free the livelist and decrement the feature count */
	VERIFY0(zap_remove_int(mos, zap_obj, ll_obj, tx));
	dsl_deadlist_free(mos, ll_obj, tx);
	spa_feature_decr(spa, SPA_FEATURE_LIVELIST, tx);
	VERIFY0(zap_count(mos, zap_obj, &count));
	if (count == 0) {
	/* no more livelists to delete */
	VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_DELETED_CLONES, tx));
	VERIFY0(zap_destroy(mos, zap_obj, tx));
	spa->spa_livelists_to_delete = 0;
	spa_notify_waiters(spa);
	}
	}

	/*
	* Load in the value for the livelist to be removed and open it. Then,
	* load its first sublist and determine which block pointers should actually
	* be freed. Then, call a synctask which performs the actual frees and updates
	* the pool-wide livelist data.
	*/
	static void
	spa_livelist_delete_cb(void arg, zthr_t z)
	{
	spa_t *spa = arg;
	uint64_t ll_obj = 0, count;
	objset_t *mos = spa->spa_meta_objset;
	uint64_t zap_obj = spa->spa_livelists_to_delete;
	/*
	* Determine the next livelist to delete. This function should only
	* be called if there is at least one deleted clone.
	*/
	VERIFY0(dsl_get_next_livelist_obj(mos, zap_obj, &ll_obj));
	VERIFY0(zap_count(mos, ll_obj, &count));
	if (count > 0) {
	dsl_deadlist_t *ll;
	dsl_deadlist_entry_t *dle;
	bplist_t to_free;
	ll = kmem_zalloc(sizeof (dsl_deadlist_t), KM_SLEEP);
	dsl_deadlist_open(ll, mos, ll_obj);
	dle = dsl_deadlist_first(ll);
	ASSERT3P(dle, !=, NULL);
	bplist_create(&to_free);
	int err = dsl_process_sub_livelist(&dle->dle_bpobj, &to_free,
	z, NULL);
	if (err == 0) {
	sublist_delete_arg_t sync_arg = {
	.spa = spa,
	.ll = ll,
	.key = dle->dle_mintxg,
	.to_free = &to_free
	};
	zfs_dbgmsg("deleting sublist (id %llu) from"
	" livelist %llu, %lld remaining",
	(u_longlong_t)dle->dle_bpobj.bpo_object,
	(u_longlong_t)ll_obj, (longlong_t)count - 1);
	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	sublist_delete_sync, &sync_arg, 0,
	ZFS_SPACE_CHECK_DESTROY));
	} else {
	VERIFY3U(err, ==, EINTR);
	}
	bplist_clear(&to_free);
	bplist_destroy(&to_free);
	dsl_deadlist_close(ll);
	kmem_free(ll, sizeof (dsl_deadlist_t));
	} else {
	livelist_delete_arg_t sync_arg = {
	.spa = spa,
	.ll_obj = ll_obj,
	.zap_obj = zap_obj
	};
	zfs_dbgmsg("deletion of livelist %llu completed",
	(u_longlong_t)ll_obj);
	VERIFY0(dsl_sync_task(spa_name(spa), NULL, livelist_delete_sync,
	&sync_arg, 0, ZFS_SPACE_CHECK_DESTROY));
	}
	}

	static void
	spa_start_livelist_destroy_thread(spa_t *spa)
	{
	ASSERT3P(spa->spa_livelist_delete_zthr, ==, NULL);
	spa->spa_livelist_delete_zthr =
	zthr_create("z_livelist_destroy",
	spa_livelist_delete_cb_check, spa_livelist_delete_cb, spa,
	minclsyspri);
	}

	typedef struct livelist_new_arg {
	bplist_t *allocs;
	bplist_t *frees;
	} livelist_new_arg_t;

	static int
	livelist_track_new_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(tx == NULL);
	livelist_new_arg_t *lna = arg;
	if (bp_freed) {
	bplist_append(lna->frees, bp);
	} else {
	bplist_append(lna->allocs, bp);
	zfs_livelist_condense_new_alloc++;
	}
	return (0);
	}

	typedef struct livelist_condense_arg {
	spa_t *spa;
	bplist_t to_keep;
	uint64_t first_size;
	uint64_t next_size;
	} livelist_condense_arg_t;

	static void
	spa_livelist_condense_sync(void arg, dmu_tx_t tx)
	{
	livelist_condense_arg_t *lca = arg;
	spa_t *spa = lca->spa;
	bplist_t new_frees;
	dsl_dataset_t *ds = spa->spa_to_condense.ds;

	/* Have we been cancelled? */
	if (spa->spa_to_condense.cancelled) {
	zfs_livelist_condense_sync_cancel++;
	goto out;
	}

	dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
	dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
	dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist;

	/*
	* It's possible that the livelist was changed while the zthr was
	* running. Therefore, we need to check for new blkptrs in the two
	* entries being condensed and continue to track them in the livelist.
	* Because of the way we handle remapped blkptrs (see dbuf_remap_impl),
	* it's possible that the newly added blkptrs are FREEs or ALLOCs so
	* we need to sort them into two different bplists.
	*/
	uint64_t first_obj = first->dle_bpobj.bpo_object;
	uint64_t next_obj = next->dle_bpobj.bpo_object;
	uint64_t cur_first_size = first->dle_bpobj.bpo_phys->bpo_num_blkptrs;
	uint64_t cur_next_size = next->dle_bpobj.bpo_phys->bpo_num_blkptrs;

	bplist_create(&new_frees);
	livelist_new_arg_t new_bps = {
	.allocs = &lca->to_keep,
	.frees = &new_frees,
	};

	if (cur_first_size > lca->first_size) {
	VERIFY0(livelist_bpobj_iterate_from_nofree(&first->dle_bpobj,
	livelist_track_new_cb, &new_bps, lca->first_size));
	}
	if (cur_next_size > lca->next_size) {
	VERIFY0(livelist_bpobj_iterate_from_nofree(&next->dle_bpobj,
	livelist_track_new_cb, &new_bps, lca->next_size));
	}

	dsl_deadlist_clear_entry(first, ll, tx);
	ASSERT(bpobj_is_empty(&first->dle_bpobj));
	dsl_deadlist_remove_entry(ll, next->dle_mintxg, tx);

	bplist_iterate(&lca->to_keep, dsl_deadlist_insert_alloc_cb, ll, tx);
	bplist_iterate(&new_frees, dsl_deadlist_insert_free_cb, ll, tx);
	bplist_destroy(&new_frees);

	char dsname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(ds, dsname);
	zfs_dbgmsg("txg %llu condensing livelist of %s (id %llu), bpobj %llu "
	"(%llu blkptrs) and bpobj %llu (%llu blkptrs) -> bpobj %llu "
	"(%llu blkptrs)", (u_longlong_t)tx->tx_txg, dsname,
	(u_longlong_t)ds->ds_object, (u_longlong_t)first_obj,
	(u_longlong_t)cur_first_size, (u_longlong_t)next_obj,
	(u_longlong_t)cur_next_size,
	(u_longlong_t)first->dle_bpobj.bpo_object,
	(u_longlong_t)first->dle_bpobj.bpo_phys->bpo_num_blkptrs);
	out:
	dmu_buf_rele(ds->ds_dbuf, spa);
	spa->spa_to_condense.ds = NULL;
	bplist_clear(&lca->to_keep);
	bplist_destroy(&lca->to_keep);
	kmem_free(lca, sizeof (livelist_condense_arg_t));
	spa->spa_to_condense.syncing = B_FALSE;
	}

	static void
	spa_livelist_condense_cb(void arg, zthr_t t)
	{
	while (zfs_livelist_condense_zthr_pause &&
	!(zthr_has_waiters(t) \|\| zthr_iscancelled(t)))
	delay(1);

	spa_t *spa = arg;
	dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
	dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
	uint64_t first_size, next_size;

	livelist_condense_arg_t *lca =
	kmem_alloc(sizeof (livelist_condense_arg_t), KM_SLEEP);
	bplist_create(&lca->to_keep);

	/*
	* Process the livelists (matching FREEs and ALLOCs) in open context
	* so we have minimal work in syncing context to condense.
	*
	* We save bpobj sizes (first_size and next_size) to use later in
	* syncing context to determine if entries were added to these sublists
	* while in open context. This is possible because the clone is still
	* active and open for normal writes and we want to make sure the new,
	* unprocessed blockpointers are inserted into the livelist normally.
	*
	* Note that dsl_process_sub_livelist() both stores the size number of
	* blockpointers and iterates over them while the bpobj's lock held, so
	* the sizes returned to us are consistent which what was actually
	* processed.
	*/
	int err = dsl_process_sub_livelist(&first->dle_bpobj, &lca->to_keep, t,
	&first_size);
	if (err == 0)
	err = dsl_process_sub_livelist(&next->dle_bpobj, &lca->to_keep,
	t, &next_size);

	if (err == 0) {
	while (zfs_livelist_condense_sync_pause &&
	!(zthr_has_waiters(t) \|\| zthr_iscancelled(t)))
	delay(1);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	dmu_tx_mark_netfree(tx);
	dmu_tx_hold_space(tx, 1);
	err = dmu_tx_assign(tx, TXG_NOWAIT \| TXG_NOTHROTTLE);
	if (err == 0) {
	/*
	* Prevent the condense zthr restarting before
	* the synctask completes.
	*/
	spa->spa_to_condense.syncing = B_TRUE;
	lca->spa = spa;
	lca->first_size = first_size;
	lca->next_size = next_size;
	dsl_sync_task_nowait(spa_get_dsl(spa),
	spa_livelist_condense_sync, lca, tx);
	dmu_tx_commit(tx);
	return;
	}
	}
	/*
	* Condensing can not continue: either it was externally stopped or
	* we were unable to assign to a tx because the pool has run out of
	* space. In the second case, we'll just end up trying to condense
	* again in a later txg.
	*/
	ASSERT(err != 0);
	bplist_clear(&lca->to_keep);
	bplist_destroy(&lca->to_keep);
	kmem_free(lca, sizeof (livelist_condense_arg_t));
	dmu_buf_rele(spa->spa_to_condense.ds->ds_dbuf, spa);
	spa->spa_to_condense.ds = NULL;
	if (err == EINTR)
	zfs_livelist_condense_zthr_cancel++;
	}

	/*
	* Check that there is something to condense but that a condense is not
	* already in progress and that condensing has not been cancelled.
	*/
	static boolean_t
	spa_livelist_condense_cb_check(void arg, zthr_t z)
	{
	(void) z;
	spa_t *spa = arg;
	if ((spa->spa_to_condense.ds != NULL) &&
	(spa->spa_to_condense.syncing == B_FALSE) &&
	(spa->spa_to_condense.cancelled == B_FALSE)) {
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static void
	spa_start_livelist_condensing_thread(spa_t *spa)
	{
	spa->spa_to_condense.ds = NULL;
	spa->spa_to_condense.first = NULL;
	spa->spa_to_condense.next = NULL;
	spa->spa_to_condense.syncing = B_FALSE;
	spa->spa_to_condense.cancelled = B_FALSE;

	ASSERT3P(spa->spa_livelist_condense_zthr, ==, NULL);
	spa->spa_livelist_condense_zthr =
	zthr_create("z_livelist_condense",
	spa_livelist_condense_cb_check,
	spa_livelist_condense_cb, spa, minclsyspri);
	}

	static void
	spa_spawn_aux_threads(spa_t *spa)
	{
	ASSERT(spa_writeable(spa));

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_start_indirect_condensing_thread(spa);
	spa_start_livelist_destroy_thread(spa);
	spa_start_livelist_condensing_thread(spa);

	ASSERT3P(spa->spa_checkpoint_discard_zthr, ==, NULL);
	spa->spa_checkpoint_discard_zthr =
	zthr_create("z_checkpoint_discard",
	spa_checkpoint_discard_thread_check,
	spa_checkpoint_discard_thread, spa, minclsyspri);
	}

	/*
	* Fix up config after a partly-completed split. This is done with the
	* ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off
	* pool have that entry in their config, but only the splitting one contains
	* a list of all the guids of the vdevs that are being split off.
	*
	* This function determines what to do with that list: either rejoin
	* all the disks to the pool, or complete the splitting process. To attempt
	* the rejoin, each disk that is offlined is marked online again, and
	* we do a reopen() call. If the vdev label for every disk that was
	* marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL)
	* then we call vdev_split() on each disk, and complete the split.
	*
	* Otherwise we leave the config alone, with all the vdevs in place in
	* the original pool.
	*/
	static void
	spa_try_repair(spa_t spa, nvlist_t config)
	{
	uint_t extracted;
	uint64_t *glist;
	uint_t i, gcount;
	nvlist_t *nvl;
	vdev_t **vd;
	boolean_t attempt_reopen;

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0)
	return;

	/* check that the config is complete */
	if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
	&glist, &gcount) != 0)
	return;

	vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP);

	/* attempt to online all the vdevs & validate */
	attempt_reopen = B_TRUE;
	for (i = 0; i < gcount; i++) {
	if (glist[i] == 0) /* vdev is hole */
	continue;

	vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE);
	if (vd[i] == NULL) {
	/*
	* Don't bother attempting to reopen the disks;
	* just do the split.
	*/
	attempt_reopen = B_FALSE;
	} else {
	/* attempt to re-online it */
	vd[i]->vdev_offline = B_FALSE;
	}
	}

	if (attempt_reopen) {
	vdev_reopen(spa->spa_root_vdev);

	/* check each device to see what state it's in */
	for (extracted = 0, i = 0; i < gcount; i++) {
	if (vd[i] != NULL &&
	vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL)
	break;
	++extracted;
	}
	}

	/*
	* If every disk has been moved to the new pool, or if we never
	* even attempted to look at them, then we split them off for
	* good.
	*/
	if (!attempt_reopen \|\| gcount == extracted) {
	for (i = 0; i < gcount; i++)
	if (vd[i] != NULL)
	vdev_split(vd[i]);
	vdev_reopen(spa->spa_root_vdev);
	}

	kmem_free(vd, gcount * sizeof (vdev_t *));
	}

	static int
	spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type)
	{
	const char *ereport = FM_EREPORT_ZFS_POOL;
	int error;

	spa->spa_load_state = state;
	(void) spa_import_progress_set_state(spa_guid(spa),
	spa_load_state(spa));

	gethrestime(&spa->spa_loaded_ts);
	error = spa_load_impl(spa, type, &ereport);

	/*
	* Don't count references from objsets that are already closed
	* and are making their way through the eviction process.
	*/
	spa_evicting_os_wait(spa);
	spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
	if (error) {
	if (error != EEXIST) {
	spa->spa_loaded_ts.tv_sec = 0;
	spa->spa_loaded_ts.tv_nsec = 0;
	}
	if (error != EBADF) {
	(void) zfs_ereport_post(ereport, spa,
	NULL, NULL, NULL, 0);
	}
	}
	spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE;
	spa->spa_ena = 0;

	(void) spa_import_progress_set_state(spa_guid(spa),
	spa_load_state(spa));

	return (error);
	}

	#ifdef ZFS_DEBUG
	/*
	* Count the number of per-vdev ZAPs associated with all of the vdevs in the
	* vdev tree rooted in the given vd, and ensure that each ZAP is present in the
	* spa's per-vdev ZAP list.
	*/
	static uint64_t
	vdev_count_verify_zaps(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t total = 0;

	if (vd->vdev_top_zap != 0) {
	total++;
	ASSERT0(zap_lookup_int(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, vd->vdev_top_zap));
	}
	if (vd->vdev_leaf_zap != 0) {
	total++;
	ASSERT0(zap_lookup_int(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, vd->vdev_leaf_zap));
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	total += vdev_count_verify_zaps(vd->vdev_child[i]);
	}

	return (total);
	}
	#else
	#define vdev_count_verify_zaps(vd) ((void) sizeof (vd), 0)
	#endif

	/*
	* Determine whether the activity check is required.
	*/
	static boolean_t
	spa_activity_check_required(spa_t spa, uberblock_t ub, nvlist_t *label,
	nvlist_t *config)
	{
	uint64_t state = 0;
	uint64_t hostid = 0;
	uint64_t tryconfig_txg = 0;
	uint64_t tryconfig_timestamp = 0;
	uint16_t tryconfig_mmp_seq = 0;
	nvlist_t *nvinfo;

	if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	(void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG,
	&tryconfig_txg);
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
	&tryconfig_timestamp);
	(void) nvlist_lookup_uint16(nvinfo, ZPOOL_CONFIG_MMP_SEQ,
	&tryconfig_mmp_seq);
	}

	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state);

	/*
	* Disable the MMP activity check - This is used by zdb which
	* is intended to be used on potentially active pools.
	*/
	if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP)
	return (B_FALSE);

	/*
	* Skip the activity check when the MMP feature is disabled.
	*/
	if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0)
	return (B_FALSE);

	/*
	* If the tryconfig_ values are nonzero, they are the results of an
	* earlier tryimport. If they all match the uberblock we just found,
	* then the pool has not changed and we return false so we do not test
	* a second time.
	*/
	if (tryconfig_txg && tryconfig_txg == ub->ub_txg &&
	tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp &&
	tryconfig_mmp_seq && tryconfig_mmp_seq ==
	(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0))
	return (B_FALSE);

	/*
	* Allow the activity check to be skipped when importing the pool
	* on the same host which last imported it. Since the hostid from
	* configuration may be stale use the one read from the label.
	*/
	if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID))
	hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID);

	if (hostid == spa_get_hostid(spa))
	return (B_FALSE);

	/*
	* Skip the activity test when the pool was cleanly exported.
	*/
	if (state != POOL_STATE_ACTIVE)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Nanoseconds the activity check must watch for changes on-disk.
	*/
	static uint64_t
	spa_activity_check_duration(spa_t spa, uberblock_t ub)
	{
	uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1);
	uint64_t multihost_interval = MSEC2NSEC(
	MMP_INTERVAL_OK(zfs_multihost_interval));
	uint64_t import_delay = MAX(NANOSEC, import_intervals *
	multihost_interval);

	/*
	* Local tunables determine a minimum duration except for the case
	* where we know when the remote host will suspend the pool if MMP
	* writes do not land.
	*
	* See Big Theory comment at the top of mmp.c for the reasoning behind
	* these cases and times.
	*/

	ASSERT(MMP_IMPORT_SAFETY_FACTOR >= 100);

	if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
	MMP_FAIL_INT(ub) > 0) {

	/* MMP on remote host will suspend pool after failed writes */
	import_delay = MMP_FAIL_INT(ub) * MSEC2NSEC(MMP_INTERVAL(ub)) *
	MMP_IMPORT_SAFETY_FACTOR / 100;

	zfs_dbgmsg("fail_intvals>0 import_delay=%llu ub_mmp "
	"mmp_fails=%llu ub_mmp mmp_interval=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)MMP_FAIL_INT(ub),
	(u_longlong_t)MMP_INTERVAL(ub),
	(u_longlong_t)import_intervals);

	} else if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
	MMP_FAIL_INT(ub) == 0) {

	/* MMP on remote host will never suspend pool */
	import_delay = MAX(import_delay, (MSEC2NSEC(MMP_INTERVAL(ub)) +
	ub->ub_mmp_delay) * import_intervals);

	zfs_dbgmsg("fail_intvals=0 import_delay=%llu ub_mmp "
	"mmp_interval=%llu ub_mmp_delay=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)MMP_INTERVAL(ub),
	(u_longlong_t)ub->ub_mmp_delay,
	(u_longlong_t)import_intervals);

	} else if (MMP_VALID(ub)) {
	/*
	* zfs-0.7 compatibility case
	*/

	import_delay = MAX(import_delay, (multihost_interval +
	ub->ub_mmp_delay) * import_intervals);

	zfs_dbgmsg("import_delay=%llu ub_mmp_delay=%llu "
	"import_intervals=%llu leaves=%u",
	(u_longlong_t)import_delay,
	(u_longlong_t)ub->ub_mmp_delay,
	(u_longlong_t)import_intervals,
	vdev_count_leaves(spa));
	} else {
	/* Using local tunings is the only reasonable option */
	zfs_dbgmsg("pool last imported on non-MMP aware "
	"host using import_delay=%llu multihost_interval=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)multihost_interval,
	(u_longlong_t)import_intervals);
	}

	return (import_delay);
	}

	/*
	* Perform the import activity check. If the user canceled the import or
	* we detected activity then fail.
	*/
	static int
	spa_activity_check(spa_t spa, uberblock_t ub, nvlist_t *config)
	{
	uint64_t txg = ub->ub_txg;
	uint64_t timestamp = ub->ub_timestamp;
	uint64_t mmp_config = ub->ub_mmp_config;
	uint16_t mmp_seq = MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0;
	uint64_t import_delay;
	hrtime_t import_expire;
	nvlist_t *mmp_label = NULL;
	vdev_t *rvd = spa->spa_root_vdev;
	kcondvar_t cv;
	kmutex_t mtx;
	int error = 0;

	cv_init(&cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL);
	mutex_enter(&mtx);

	/*
	* If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed
	* during the earlier tryimport. If the txg recorded there is 0 then
	* the pool is known to be active on another host.
	*
	* Otherwise, the pool might be in use on another host. Check for
	* changes in the uberblocks on disk if necessary.
	*/
	if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
	nvlist_t *nvinfo = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_LOAD_INFO);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) &&
	fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) {
	vdev_uberblock_load(rvd, ub, &mmp_label);
	error = SET_ERROR(EREMOTEIO);
	goto out;
	}
	}

	import_delay = spa_activity_check_duration(spa, ub);

	/* Add a small random factor in case of simultaneous imports (0-25%) */
	import_delay += import_delay * random_in_range(250) / 1000;

	import_expire = gethrtime() + import_delay;

	while (gethrtime() < import_expire) {
	(void) spa_import_progress_set_mmp_check(spa_guid(spa),
	NSEC2SEC(import_expire - gethrtime()));

	vdev_uberblock_load(rvd, ub, &mmp_label);

	if (txg != ub->ub_txg \|\| timestamp != ub->ub_timestamp \|\|
	mmp_seq != (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)) {
	zfs_dbgmsg("multihost activity detected "
	"txg %llu ub_txg %llu "
	"timestamp %llu ub_timestamp %llu "
	"mmp_config %#llx ub_mmp_config %#llx",
	(u_longlong_t)txg, (u_longlong_t)ub->ub_txg,
	(u_longlong_t)timestamp,
	(u_longlong_t)ub->ub_timestamp,
	(u_longlong_t)mmp_config,
	(u_longlong_t)ub->ub_mmp_config);

	error = SET_ERROR(EREMOTEIO);
	break;
	}

	if (mmp_label) {
	nvlist_free(mmp_label);
	mmp_label = NULL;
	}

	error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz);
	if (error != -1) {
	error = SET_ERROR(EINTR);
	break;
	}
	error = 0;
	}

	out:
	mutex_exit(&mtx);
	mutex_destroy(&mtx);
	cv_destroy(&cv);

	/*
	* If the pool is determined to be active store the status in the
	* spa->spa_load_info nvlist. If the remote hostname or hostid are
	* available from configuration read from disk store them as well.
	* This allows 'zpool import' to generate a more useful message.
	*
	* ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory)
	* ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool
	* ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool
	*/
	if (error == EREMOTEIO) {
	const char *hostname = "<unknown>";
	uint64_t hostid = 0;

	if (mmp_label) {
	if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) {
	hostname = fnvlist_lookup_string(mmp_label,
	ZPOOL_CONFIG_HOSTNAME);
	fnvlist_add_string(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_HOSTNAME, hostname);
	}

	if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) {
	hostid = fnvlist_lookup_uint64(mmp_label,
	ZPOOL_CONFIG_HOSTID);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_HOSTID, hostid);
	}
	}

	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_TXG, 0);

	error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO);
	}

	if (mmp_label)
	nvlist_free(mmp_label);

	return (error);
	}

	static int
	spa_verify_host(spa_t spa, nvlist_t mos_config)
	{
	uint64_t hostid;
	char *hostname;
	uint64_t myhostid = 0;

	if (!spa_is_root(spa) && nvlist_lookup_uint64(mos_config,
	ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
	hostname = fnvlist_lookup_string(mos_config,
	ZPOOL_CONFIG_HOSTNAME);

	myhostid = zone_get_hostid(NULL);

	if (hostid != 0 && myhostid != 0 && hostid != myhostid) {
	cmn_err(CE_WARN, "pool '%s' could not be "
	"loaded as it was last accessed by "
	"another system (host: %s hostid: 0x%llx). "
	"See: https://openzfs.github.io/openzfs-docs/msg/"
	"ZFS-8000-EY",
	spa_name(spa), hostname, (u_longlong_t)hostid);
	spa_load_failed(spa, "hostid verification failed: pool "
	"last accessed by host: %s (hostid: 0x%llx)",
	hostname, (u_longlong_t)hostid);
	return (SET_ERROR(EBADF));
	}
	}

	return (0);
	}

	static int
	spa_ld_parse_config(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;
	nvlist_t nvtree, nvl, *config = spa->spa_config;
	int parse;
	vdev_t *rvd;
	uint64_t pool_guid;
	char *comment;
	char *compatibility;

	/*
	* Versioning wasn't explicitly added to the label until later, so if
	* it's not present treat it as the initial version.
	*/
	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&spa->spa_ubsync.ub_version) != 0)
	spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL;

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) {
	spa_load_failed(spa, "invalid config provided: '%s' missing",
	ZPOOL_CONFIG_POOL_GUID);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If we are doing an import, ensure that the pool is not already
	* imported by checking if its pool guid already exists in the
	* spa namespace.
	*
	* The only case that we allow an already imported pool to be
	* imported again, is when the pool is checkpointed and we want to
	* look at its checkpointed state from userland tools like zdb.
	*/
	#ifdef _KERNEL
	if ((spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
	spa_guid_exists(pool_guid, 0)) {
	#else
	if ((spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
	spa_guid_exists(pool_guid, 0) &&
	!spa_importing_readonly_checkpoint(spa)) {
	#endif
	spa_load_failed(spa, "a pool with guid %llu is already open",
	(u_longlong_t)pool_guid);
	return (SET_ERROR(EEXIST));
	}

	spa->spa_config_guid = pool_guid;

	nvlist_free(spa->spa_load_info);
	spa->spa_load_info = fnvlist_alloc();

	ASSERT(spa->spa_comment == NULL);
	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
	spa->spa_comment = spa_strdup(comment);

	ASSERT(spa->spa_compatibility == NULL);
	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMPATIBILITY,
	&compatibility) == 0)
	spa->spa_compatibility = spa_strdup(compatibility);

	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
	&spa->spa_config_txg);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) == 0)
	spa->spa_config_splitting = fnvlist_dup(nvl);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtree)) {
	spa_load_failed(spa, "invalid config provided: '%s' missing",
	ZPOOL_CONFIG_VDEV_TREE);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Create "The Godfather" zio to hold all async IOs
	*/
	spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
	KM_SLEEP);
	for (int i = 0; i < max_ncpus; i++) {
	spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);
	}

	/*
	* Parse the configuration into a vdev tree. We explicitly set the
	* value that will be returned by spa_version() since parsing the
	* configuration requires knowing the version number.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	parse = (type == SPA_IMPORT_EXISTING ?
	VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
	error = spa_config_parse(spa, &rvd, nvtree, NULL, 0, parse);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "unable to parse config [error=%d]",
	error);
	return (error);
	}

	ASSERT(spa->spa_root_vdev == rvd);
	ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT);
	ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT);

	if (type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_guid(spa) == pool_guid);
	}

	return (0);
	}

	/*
	* Recursively open all vdevs in the vdev tree. This function is called twice:
	* first with the untrusted config, then with the trusted config.
	*/
	static int
	spa_ld_open_vdevs(spa_t *spa)
	{
	int error = 0;

	/*
	* spa_missing_tvds_allowed defines how many top-level vdevs can be
	* missing/unopenable for the root vdev to be still considered openable.
	*/
	if (spa->spa_trust_config) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds;
	} else if (spa->spa_config_source == SPA_CONFIG_SRC_CACHEFILE) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_cachefile;
	} else if (spa->spa_config_source == SPA_CONFIG_SRC_SCAN) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_scan;
	} else {
	spa->spa_missing_tvds_allowed = 0;
	}

	spa->spa_missing_tvds_allowed =
	MAX(zfs_max_missing_tvds, spa->spa_missing_tvds_allowed);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	error = vdev_open(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (spa->spa_missing_tvds != 0) {
	spa_load_note(spa, "vdev tree has %lld missing top-level "
	"vdevs.", (u_longlong_t)spa->spa_missing_tvds);
	if (spa->spa_trust_config && (spa->spa_mode & SPA_MODE_WRITE)) {
	/*
	* Although theoretically we could allow users to open
	* incomplete pools in RW mode, we'd need to add a lot
	* of extra logic (e.g. adjust pool space to account
	* for missing vdevs).
	* This limitation also prevents users from accidentally
	* opening the pool in RW mode during data recovery and
	* damaging it further.
	*/
	spa_load_note(spa, "pools with missing top-level "
	"vdevs can only be opened in read-only mode.");
	error = SET_ERROR(ENXIO);
	} else {
	spa_load_note(spa, "current settings allow for maximum "
	"%lld missing top-level vdevs at this stage.",
	(u_longlong_t)spa->spa_missing_tvds_allowed);
	}
	}
	if (error != 0) {
	spa_load_failed(spa, "unable to open vdev tree [error=%d]",
	error);
	}
	if (spa->spa_missing_tvds != 0 \|\| error != 0)
	vdev_dbgmsg_print_tree(spa->spa_root_vdev, 2);

	return (error);
	}

	/*
	* We need to validate the vdev labels against the configuration that
	* we have in hand. This function is called twice: first with an untrusted
	* config, then with a trusted config. The validation is more strict when the
	* config is trusted.
	*/
	static int
	spa_ld_validate_vdevs(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	error = vdev_validate(rvd);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "vdev_validate failed [error=%d]", error);
	return (error);
	}

	if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
	spa_load_failed(spa, "cannot open vdev tree after invalidating "
	"some vdevs");
	vdev_dbgmsg_print_tree(rvd, 2);
	return (SET_ERROR(ENXIO));
	}

	return (0);
	}

	static void
	spa_ld_select_uberblock_done(spa_t spa, uberblock_t ub)
	{
	spa->spa_state = POOL_STATE_ACTIVE;
	spa->spa_ubsync = spa->spa_uberblock;
	spa->spa_verify_min_txg = spa->spa_extreme_rewind ?
	TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1;
	spa->spa_first_txg = spa->spa_last_ubsync_txg ?
	spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1;
	spa->spa_claim_max_txg = spa->spa_first_txg;
	spa->spa_prev_software_version = ub->ub_software_version;
	}

	static int
	spa_ld_select_uberblock(spa_t *spa, spa_import_type_t type)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	nvlist_t *label;
	uberblock_t *ub = &spa->spa_uberblock;
	boolean_t activity_check = B_FALSE;

	/*
	* If we are opening the checkpointed state of the pool by
	* rewinding to it, at this point we will have written the
	* checkpointed uberblock to the vdev labels, so searching
	* the labels will find the right uberblock. However, if
	* we are opening the checkpointed state read-only, we have
	* not modified the labels. Therefore, we must ignore the
	* labels and continue using the spa_uberblock that was set
	* by spa_ld_checkpoint_rewind.
	*
	* Note that it would be fine to ignore the labels when
	* rewinding (opening writeable) as well. However, if we
	* crash just after writing the labels, we will end up
	* searching the labels. Doing so in the common case means
	* that this code path gets exercised normally, rather than
	* just in the edge case.
	*/
	if (ub->ub_checkpoint_txg != 0 &&
	spa_importing_readonly_checkpoint(spa)) {
	spa_ld_select_uberblock_done(spa, ub);
	return (0);
	}

	/*
	* Find the best uberblock.
	*/
	vdev_uberblock_load(rvd, ub, &label);

	/*
	* If we weren't able to find a single valid uberblock, return failure.
	*/
	if (ub->ub_txg == 0) {
	nvlist_free(label);
	spa_load_failed(spa, "no valid uberblock found");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
	}

	if (spa->spa_load_max_txg != UINT64_MAX) {
	(void) spa_import_progress_set_max_txg(spa_guid(spa),
	(u_longlong_t)spa->spa_load_max_txg);
	}
	spa_load_note(spa, "using uberblock with txg=%llu",
	(u_longlong_t)ub->ub_txg);


	/*
	* For pools which have the multihost property on determine if the
	* pool is truly inactive and can be safely imported. Prevent
	* hosts which don't have a hostid set from importing the pool.
	*/
	activity_check = spa_activity_check_required(spa, ub, label,
	spa->spa_config);
	if (activity_check) {
	if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay &&
	spa_get_hostid(spa) == 0) {
	nvlist_free(label);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
	return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
	}

	int error = spa_activity_check(spa, ub, spa->spa_config);
	if (error) {
	nvlist_free(label);
	return (error);
	}

	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_TXG, ub->ub_txg);
	fnvlist_add_uint16(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_SEQ,
	(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0));
	}

	/*
	* If the pool has an unsupported version we can't open it.
	*/
	if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) {
	nvlist_free(label);
	spa_load_failed(spa, "version %llu is not supported",
	(u_longlong_t)ub->ub_version);
	return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP));
	}

	if (ub->ub_version >= SPA_VERSION_FEATURES) {
	nvlist_t *features;

	/*
	* If we weren't able to find what's necessary for reading the
	* MOS in the label, return failure.
	*/
	if (label == NULL) {
	spa_load_failed(spa, "label config unavailable");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	ENXIO));
	}

	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_FEATURES_FOR_READ,
	&features) != 0) {
	nvlist_free(label);
	spa_load_failed(spa, "invalid label: '%s' missing",
	ZPOOL_CONFIG_FEATURES_FOR_READ);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	ENXIO));
	}

	/*
	* Update our in-core representation with the definitive values
	* from the label.
	*/
	nvlist_free(spa->spa_label_features);
	spa->spa_label_features = fnvlist_dup(features);
	}

	nvlist_free(label);

	/*
	* Look through entries in the label nvlist's features_for_read. If
	* there is a feature listed there which we don't understand then we
	* cannot open a pool.
	*/
	if (ub->ub_version >= SPA_VERSION_FEATURES) {
	nvlist_t *unsup_feat;

	unsup_feat = fnvlist_alloc();

	for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features,
	NULL); nvp != NULL;
	nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) {
	if (!zfeature_is_supported(nvpair_name(nvp))) {
	fnvlist_add_string(unsup_feat,
	nvpair_name(nvp), "");
	}
	}

	if (!nvlist_empty(unsup_feat)) {
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
	nvlist_free(unsup_feat);
	spa_load_failed(spa, "some features are unsupported");
	return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP));
	}

	nvlist_free(unsup_feat);
	}

	if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_try_repair(spa, spa->spa_config);
	spa_config_exit(spa, SCL_ALL, FTAG);
	nvlist_free(spa->spa_config_splitting);
	spa->spa_config_splitting = NULL;
	}

	/*
	* Initialize internal SPA structures.
	*/
	spa_ld_select_uberblock_done(spa, ub);

	return (0);
	}

	static int
	spa_ld_open_rootbp(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
	if (error != 0) {
	spa_load_failed(spa, "unable to open rootbp in dsl_pool_init "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}
	spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;

	return (0);
	}

	static int
	spa_ld_trusted_config(spa_t *spa, spa_import_type_t type,
	boolean_t reloading)
	{
	vdev_t mrvd, rvd = spa->spa_root_vdev;
	nvlist_t nv, mos_config, *policy;
	int error = 0, copy_error;
	uint64_t healthy_tvds, healthy_tvds_mos;
	uint64_t mos_config_txg;

	if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object, B_TRUE)
	!= 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* If we're assembling a pool from a split, the config provided is
	* already trusted so there is nothing to do.
	*/
	if (type == SPA_IMPORT_ASSEMBLE)
	return (0);

	healthy_tvds = spa_healthy_core_tvds(spa);

	if (load_nvlist(spa, spa->spa_config_object, &mos_config)
	!= 0) {
	spa_load_failed(spa, "unable to retrieve MOS config");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* If we are doing an open, pool owner wasn't verified yet, thus do
	* the verification here.
	*/
	if (spa->spa_load_state == SPA_LOAD_OPEN) {
	error = spa_verify_host(spa, mos_config);
	if (error != 0) {
	nvlist_free(mos_config);
	return (error);
	}
	}

	nv = fnvlist_lookup_nvlist(mos_config, ZPOOL_CONFIG_VDEV_TREE);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* Build a new vdev tree from the trusted config
	*/
	error = spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD);
	if (error != 0) {
	nvlist_free(mos_config);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa_load_failed(spa, "spa_config_parse failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	/*
	* Vdev paths in the MOS may be obsolete. If the untrusted config was
	* obtained by scanning /dev/dsk, then it will have the right vdev
	* paths. We update the trusted MOS config with this information.
	* We first try to copy the paths with vdev_copy_path_strict, which
	* succeeds only when both configs have exactly the same vdev tree.
	* If that fails, we fall back to a more flexible method that has a
	* best effort policy.
	*/
	copy_error = vdev_copy_path_strict(rvd, mrvd);
	if (copy_error != 0 \|\| spa_load_print_vdev_tree) {
	spa_load_note(spa, "provided vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	spa_load_note(spa, "MOS vdev tree:");
	vdev_dbgmsg_print_tree(mrvd, 2);
	}
	if (copy_error != 0) {
	spa_load_note(spa, "vdev_copy_path_strict failed, falling "
	"back to vdev_copy_path_relaxed");
	vdev_copy_path_relaxed(rvd, mrvd);
	}

	vdev_close(rvd);
	vdev_free(rvd);
	spa->spa_root_vdev = mrvd;
	rvd = mrvd;
	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* We will use spa_config if we decide to reload the spa or if spa_load
	* fails and we rewind. We must thus regenerate the config using the
	* MOS information with the updated paths. ZPOOL_LOAD_POLICY is used to
	* pass settings on how to load the pool and is not stored in the MOS.
	* We copy it over to our new, trusted config.
	*/
	mos_config_txg = fnvlist_lookup_uint64(mos_config,
	ZPOOL_CONFIG_POOL_TXG);
	nvlist_free(mos_config);
	mos_config = spa_config_generate(spa, NULL, mos_config_txg, B_FALSE);
	if (nvlist_lookup_nvlist(spa->spa_config, ZPOOL_LOAD_POLICY,
	&policy) == 0)
	fnvlist_add_nvlist(mos_config, ZPOOL_LOAD_POLICY, policy);
	spa_config_set(spa, mos_config);
	spa->spa_config_source = SPA_CONFIG_SRC_MOS;

	/*
	* Now that we got the config from the MOS, we should be more strict
	* in checking blkptrs and can make assumptions about the consistency
	* of the vdev tree. spa_trust_config must be set to true before opening
	* vdevs in order for them to be writeable.
	*/
	spa->spa_trust_config = B_TRUE;

	/*
	* Open and validate the new vdev tree
	*/
	error = spa_ld_open_vdevs(spa);
	if (error != 0)
	return (error);

	error = spa_ld_validate_vdevs(spa);
	if (error != 0)
	return (error);

	if (copy_error != 0 \|\| spa_load_print_vdev_tree) {
	spa_load_note(spa, "final vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	}

	if (spa->spa_load_state != SPA_LOAD_TRYIMPORT &&
	!spa->spa_extreme_rewind && zfs_max_missing_tvds == 0) {
	/*
	* Sanity check to make sure that we are indeed loading the
	* latest uberblock. If we missed SPA_SYNC_MIN_VDEVS tvds
	* in the config provided and they happened to be the only ones
	* to have the latest uberblock, we could involuntarily perform
	* an extreme rewind.
	*/
	healthy_tvds_mos = spa_healthy_core_tvds(spa);
	if (healthy_tvds_mos - healthy_tvds >=
	SPA_SYNC_MIN_VDEVS) {
	spa_load_note(spa, "config provided misses too many "
	"top-level vdevs compared to MOS (%lld vs %lld). ",
	(u_longlong_t)healthy_tvds,
	(u_longlong_t)healthy_tvds_mos);
	spa_load_note(spa, "vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	if (reloading) {
	spa_load_failed(spa, "config was already "
	"provided from MOS. Aborting.");
	return (spa_vdev_err(rvd,
	VDEV_AUX_CORRUPT_DATA, EIO));
	}
	spa_load_note(spa, "spa must be reloaded using MOS "
	"config");
	return (SET_ERROR(EAGAIN));
	}
	}

	error = spa_check_for_missing_logs(spa);
	if (error != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO));

	if (rvd->vdev_guid_sum != spa->spa_uberblock.ub_guid_sum) {
	spa_load_failed(spa, "uberblock guid sum doesn't match MOS "
	"guid sum (%llu != %llu)",
	(u_longlong_t)spa->spa_uberblock.ub_guid_sum,
	(u_longlong_t)rvd->vdev_guid_sum);
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM,
	ENXIO));
	}

	return (0);
	}

	static int
	spa_ld_open_indirect_vdev_metadata(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* Everything that we read before spa_remove_init() must be stored
	* on concreted vdevs. Therefore we do this as early as possible.
	*/
	error = spa_remove_init(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_remove_init failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Retrieve information needed to condense indirect vdev mappings.
	*/
	error = spa_condense_init(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_condense_init failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	return (0);
	}

	static int
	spa_ld_check_features(spa_t spa, boolean_t missing_feat_writep)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	if (spa_version(spa) >= SPA_VERSION_FEATURES) {
	boolean_t missing_feat_read = B_FALSE;
	nvlist_t unsup_feat, enabled_feat;

	if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
	&spa->spa_feat_for_read_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE,
	&spa->spa_feat_for_write_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS,
	&spa->spa_feat_desc_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	enabled_feat = fnvlist_alloc();
	unsup_feat = fnvlist_alloc();

	if (!spa_features_check(spa, B_FALSE,
	unsup_feat, enabled_feat))
	missing_feat_read = B_TRUE;

	if (spa_writeable(spa) \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) {
	if (!spa_features_check(spa, B_TRUE,
	unsup_feat, enabled_feat)) {
	*missing_feat_writep = B_TRUE;
	}
	}

	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat);

	if (!nvlist_empty(unsup_feat)) {
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
	}

	fnvlist_free(enabled_feat);
	fnvlist_free(unsup_feat);

	if (!missing_feat_read) {
	fnvlist_add_boolean(spa->spa_load_info,
	ZPOOL_CONFIG_CAN_RDONLY);
	}

	/*
	* If the state is SPA_LOAD_TRYIMPORT, our objective is
	* twofold: to determine whether the pool is available for
	* import in read-write mode and (if it is not) whether the
	* pool is available for import in read-only mode. If the pool
	* is available for import in read-write mode, it is displayed
	* as available in userland; if it is not available for import
	* in read-only mode, it is displayed as unavailable in
	* userland. If the pool is available for import in read-only
	* mode but not read-write mode, it is displayed as unavailable
	* in userland with a special note that the pool is actually
	* available for open in read-only mode.
	*
	* As a result, if the state is SPA_LOAD_TRYIMPORT and we are
	* missing a feature for write, we must first determine whether
	* the pool can be opened read-only before returning to
	* userland in order to know whether to display the
	* abovementioned note.
	*/
	if (missing_feat_read \|\| (*missing_feat_writep &&
	spa_writeable(spa))) {
	spa_load_failed(spa, "pool uses unsupported features");
	return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP));
	}

	/*
	* Load refcounts for ZFS features from disk into an in-memory
	* cache during SPA initialization.
	*/
	for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
	uint64_t refcount;

	error = feature_get_refcount_from_disk(spa,
	&spa_feature_table[i], &refcount);
	if (error == 0) {
	spa->spa_feat_refcount_cache[i] = refcount;
	} else if (error == ENOTSUP) {
	spa->spa_feat_refcount_cache[i] =
	SPA_FEATURE_DISABLED;
	} else {
	spa_load_failed(spa, "error getting refcount "
	"for feature %s [error=%d]",
	spa_feature_table[i].fi_guid, error);
	return (spa_vdev_err(rvd,
	VDEV_AUX_CORRUPT_DATA, EIO));
	}
	}
	}

	if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
	if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG,
	&spa->spa_feat_enabled_txg_obj, B_TRUE) != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Encryption was added before bookmark_v2, even though bookmark_v2
	* is now a dependency. If this pool has encryption enabled without
	* bookmark_v2, trigger an errata message.
	*/
	if (spa_feature_is_enabled(spa, SPA_FEATURE_ENCRYPTION) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_BOOKMARK_V2)) {
	spa->spa_errata = ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
	}

	return (0);
	}

	static int
	spa_ld_load_special_directories(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	spa->spa_is_initializing = B_TRUE;
	error = dsl_pool_open(spa->spa_dsl_pool);
	spa->spa_is_initializing = B_FALSE;
	if (error != 0) {
	spa_load_failed(spa, "dsl_pool_open failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	return (0);
	}

	static int
	spa_ld_get_props(spa_t *spa)
	{
	int error = 0;
	uint64_t obj;
	vdev_t *rvd = spa->spa_root_vdev;

	/* Grab the checksum salt from the MOS. */
	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CHECKSUM_SALT, 1,
	sizeof (spa->spa_cksum_salt.zcs_bytes),
	spa->spa_cksum_salt.zcs_bytes);
	if (error == ENOENT) {
	/* Generate a new salt for subsequent use */
	(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
	sizeof (spa->spa_cksum_salt.zcs_bytes));
	} else if (error != 0) {
	spa_load_failed(spa, "unable to retrieve checksum salt from "
	"MOS [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj, B_TRUE) != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj);
	if (error != 0) {
	spa_load_failed(spa, "error opening deferred-frees bpobj "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Load the bit that tells us to use the new accounting function
	* (raid-z deflation). If we have an older pool, this will not
	* be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION,
	&spa->spa_creation_version, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the persistent error log. If we have an older pool, this will
	* not be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last,
	B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB,
	&spa->spa_errlog_scrub, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the livelist deletion field. If a livelist is queued for
	* deletion, indicate that in the spa
	*/
	error = spa_dir_prop(spa, DMU_POOL_DELETED_CLONES,
	&spa->spa_livelists_to_delete, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the history object. If we have an older pool, this
	* will not be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the per-vdev ZAP map. If we have an older pool, this will not
	* be present; in this case, defer its creation to a later time to
	* avoid dirtying the MOS this early / out of sync context. See
	* spa_sync_config_object.
	*/

	/* The sentinel is only available in the MOS config. */
	nvlist_t *mos_config;
	if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0) {
	spa_load_failed(spa, "unable to retrieve MOS config");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
	&spa->spa_all_vdev_zaps, B_FALSE);

	if (error == ENOENT) {
	VERIFY(!nvlist_exists(mos_config,
	ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
	spa->spa_avz_action = AVZ_ACTION_INITIALIZE;
	ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
	} else if (error != 0) {
	nvlist_free(mos_config);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	} else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
	/*
	* An older version of ZFS overwrote the sentinel value, so
	* we have orphaned per-vdev ZAPs in the MOS. Defer their
	* destruction to later; see spa_sync_config_object.
	*/
	spa->spa_avz_action = AVZ_ACTION_DESTROY;
	/*
	* We're assuming that no vdevs have had their ZAPs created
	* before this. Better be sure of it.
	*/
	ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
	}
	nvlist_free(mos_config);

	spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);

	error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object,
	B_FALSE);
	if (error && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	if (error == 0) {
	uint64_t autoreplace = 0;

	spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs);
	spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace);
	spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation);
	spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode);
	spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand);
	spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost);
	spa_prop_find(spa, ZPOOL_PROP_AUTOTRIM, &spa->spa_autotrim);
	spa->spa_autoreplace = (autoreplace != 0);
	}

	/*
	* If we are importing a pool with missing top-level vdevs,
	* we enforce that the pool doesn't panic or get suspended on
	* error since the likelihood of missing data is extremely high.
	*/
	if (spa->spa_missing_tvds > 0 &&
	spa->spa_failmode != ZIO_FAILURE_MODE_CONTINUE &&
	spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	spa_load_note(spa, "forcing failmode to 'continue' "
	"as some top level vdevs are missing");
	spa->spa_failmode = ZIO_FAILURE_MODE_CONTINUE;
	}

	return (0);
	}

	static int
	spa_ld_open_aux_vdevs(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If we're assembling the pool from the split-off vdevs of
	* an existing pool, we don't want to attach the spares & cache
	* devices.
	*/

	/*
	* Load any hot spares for this pool.
	*/
	error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object,
	B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_version(spa) >= SPA_VERSION_SPARES);
	if (load_nvlist(spa, spa->spa_spares.sav_object,
	&spa->spa_spares.sav_config) != 0) {
	spa_load_failed(spa, "error loading spares nvlist");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	} else if (error == 0) {
	spa->spa_spares.sav_sync = B_TRUE;
	}

	/*
	* Load any level 2 ARC devices for this pool.
	*/
	error = spa_dir_prop(spa, DMU_POOL_L2CACHE,
	&spa->spa_l2cache.sav_object, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE);
	if (load_nvlist(spa, spa->spa_l2cache.sav_object,
	&spa->spa_l2cache.sav_config) != 0) {
	spa_load_failed(spa, "error loading l2cache nvlist");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	} else if (error == 0) {
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	return (0);
	}

	static int
	spa_ld_load_vdev_metadata(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If the 'multihost' property is set, then never allow a pool to
	* be imported when the system hostid is zero. The exception to
	* this rule is zdb which is always allowed to access pools.
	*/
	if (spa_multihost(spa) && spa_get_hostid(spa) == 0 &&
	(spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) {
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
	return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
	}

	/*
	* If the 'autoreplace' property is set, then post a resource notifying
	* the ZFS DE that it should not issue any faults for unopenable
	* devices. We also iterate over the vdevs, and post a sysevent for any
	* unopenable vdevs so that the normal autoreplace handler can take
	* over.
	*/
	if (spa->spa_autoreplace && spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	spa_check_removed(spa->spa_root_vdev);
	/*
	* For the import case, this is done in spa_import(), because
	* at this point we're using the spare definitions from
	* the MOS config, not necessarily from the userland config.
	*/
	if (spa->spa_load_state != SPA_LOAD_IMPORT) {
	spa_aux_check_removed(&spa->spa_spares);
	spa_aux_check_removed(&spa->spa_l2cache);
	}
	}

	/*
	* Load the vdev metadata such as metaslabs, DTLs, spacemap object, etc.
	*/
	error = vdev_load(rvd);
	if (error != 0) {
	spa_load_failed(spa, "vdev_load failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	error = spa_ld_log_spacemaps(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_ld_log_spacemaps failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	/*
	* Propagate the leaf DTLs we just loaded all the way up the vdev tree.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	vdev_dtl_reassess(rvd, 0, 0, B_FALSE, B_FALSE);
	spa_config_exit(spa, SCL_ALL, FTAG);

	return (0);
	}

	static int
	spa_ld_load_dedup_tables(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	error = ddt_load(spa);
	if (error != 0) {
	spa_load_failed(spa, "ddt_load failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	return (0);
	}

	static int
	spa_ld_verify_logs(spa_t spa, spa_import_type_t type, const char *ereport)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	if (type != SPA_IMPORT_ASSEMBLE && spa_writeable(spa)) {
	boolean_t missing = spa_check_logs(spa);
	if (missing) {
	if (spa->spa_missing_tvds != 0) {
	spa_load_note(spa, "spa_check_logs failed "
	"so dropping the logs");
	} else {
	*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
	spa_load_failed(spa, "spa_check_logs failed");
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG,
	ENXIO));
	}
	}
	}

	return (0);
	}

	static int
	spa_ld_verify_pool_data(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* We've successfully opened the pool, verify that we're ready
	* to start pushing transactions.
	*/
	if (spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	error = spa_load_verify(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_load_verify failed "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	error));
	}
	}

	return (0);
	}

	static void
	spa_ld_claim_log_blocks(spa_t *spa)
	{
	dmu_tx_t *tx;
	dsl_pool_t *dp = spa_get_dsl(spa);

	/*
	* Claim log blocks that haven't been committed yet.
	* This must all happen in a single txg.
	* Note: spa_claim_max_txg is updated by spa_claim_notify(),
	* invoked from zil_claim_log_block()'s i/o done callback.
	* Price of rollback is that we abandon the log.
	*/
	spa->spa_claiming = B_TRUE;

	tx = dmu_tx_create_assigned(dp, spa_first_txg(spa));
	(void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	zil_claim, tx, DS_FIND_CHILDREN);
	dmu_tx_commit(tx);

	spa->spa_claiming = B_FALSE;

	spa_set_log_state(spa, SPA_LOG_GOOD);
	}

	static void
	spa_ld_check_for_config_update(spa_t *spa, uint64_t config_cache_txg,
	boolean_t update_config_cache)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	int need_update = B_FALSE;

	/*
	* If the config cache is stale, or we have uninitialized
	* metaslabs (see spa_vdev_add()), then update the config.
	*
	* If this is a verbatim import, trust the current
	* in-core spa_config and update the disk labels.
	*/
	if (update_config_cache \|\| config_cache_txg != spa->spa_config_txg \|\|
	spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_RECOVER \|\|
	(spa->spa_import_flags & ZFS_IMPORT_VERBATIM))
	need_update = B_TRUE;

	for (int c = 0; c < rvd->vdev_children; c++)
	if (rvd->vdev_child[c]->vdev_ms_array == 0)
	need_update = B_TRUE;

	/*
	* Update the config cache asynchronously in case we're the
	* root pool, in which case the config cache isn't writable yet.
	*/
	if (need_update)
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	static void
	spa_ld_prepare_for_reload(spa_t *spa)
	{
	spa_mode_t mode = spa->spa_mode;
	int async_suspended = spa->spa_async_suspended;

	spa_unload(spa);
	spa_deactivate(spa);
	spa_activate(spa, mode);

	/*
	* We save the value of spa_async_suspended as it gets reset to 0 by
	* spa_unload(). We want to restore it back to the original value before
	* returning as we might be calling spa_async_resume() later.
	*/
	spa->spa_async_suspended = async_suspended;
	}

	static int
	spa_ld_read_checkpoint_txg(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error = 0;

	ASSERT0(spa->spa_checkpoint_txg);
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error == ENOENT)
	return (0);

	if (error != 0)
	return (error);

	ASSERT3U(checkpoint.ub_txg, !=, 0);
	ASSERT3U(checkpoint.ub_checkpoint_txg, !=, 0);
	ASSERT3U(checkpoint.ub_timestamp, !=, 0);
	spa->spa_checkpoint_txg = checkpoint.ub_txg;
	spa->spa_checkpoint_info.sci_timestamp = checkpoint.ub_timestamp;

	return (0);
	}

	static int
	spa_ld_mos_init(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);

	/*
	* Never trust the config that is provided unless we are assembling
	* a pool following a split.
	* This means don't trust blkptrs and the vdev tree in general. This
	* also effectively puts the spa in read-only mode since
	* spa_writeable() checks for spa_trust_config to be true.
	* We will later load a trusted config from the MOS.
	*/
	if (type != SPA_IMPORT_ASSEMBLE)
	spa->spa_trust_config = B_FALSE;

	/*
	* Parse the config provided to create a vdev tree.
	*/
	error = spa_ld_parse_config(spa, type);
	if (error != 0)
	return (error);

	spa_import_progress_add(spa);

	/*
	* Now that we have the vdev tree, try to open each vdev. This involves
	* opening the underlying physical device, retrieving its geometry and
	* probing the vdev with a dummy I/O. The state of each vdev will be set
	* based on the success of those operations. After this we'll be ready
	* to read from the vdevs.
	*/
	error = spa_ld_open_vdevs(spa);
	if (error != 0)
	return (error);

	/*
	* Read the label of each vdev and make sure that the GUIDs stored
	* there match the GUIDs in the config provided.
	* If we're assembling a new pool that's been split off from an
	* existing pool, the labels haven't yet been updated so we skip
	* validation for now.
	*/
	if (type != SPA_IMPORT_ASSEMBLE) {
	error = spa_ld_validate_vdevs(spa);
	if (error != 0)
	return (error);
	}

	/*
	* Read all vdev labels to find the best uberblock (i.e. latest,
	* unless spa_load_max_txg is set) and store it in spa_uberblock. We
	* get the list of features required to read blkptrs in the MOS from
	* the vdev label with the best uberblock and verify that our version
	* of zfs supports them all.
	*/
	error = spa_ld_select_uberblock(spa, type);
	if (error != 0)
	return (error);

	/*
	* Pass that uberblock to the dsl_pool layer which will open the root
	* blkptr. This blkptr points to the latest version of the MOS and will
	* allow us to read its contents.
	*/
	error = spa_ld_open_rootbp(spa);
	if (error != 0)
	return (error);

	return (0);
	}

	static int
	spa_ld_checkpoint_rewind(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error = 0;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error != 0) {
	spa_load_failed(spa, "unable to retrieve checkpointed "
	"uberblock from the MOS config [error=%d]", error);

	if (error == ENOENT)
	error = ZFS_ERR_NO_CHECKPOINT;

	return (error);
	}

	ASSERT3U(checkpoint.ub_txg, <, spa->spa_uberblock.ub_txg);
	ASSERT3U(checkpoint.ub_txg, ==, checkpoint.ub_checkpoint_txg);

	/*
	* We need to update the txg and timestamp of the checkpointed
	* uberblock to be higher than the latest one. This ensures that
	* the checkpointed uberblock is selected if we were to close and
	* reopen the pool right after we've written it in the vdev labels.
	* (also see block comment in vdev_uberblock_compare)
	*/
	checkpoint.ub_txg = spa->spa_uberblock.ub_txg + 1;
	checkpoint.ub_timestamp = gethrestime_sec();

	/*
	* Set current uberblock to be the checkpointed uberblock.
	*/
	spa->spa_uberblock = checkpoint;

	/*
	* If we are doing a normal rewind, then the pool is open for
	* writing and we sync the "updated" checkpointed uberblock to
	* disk. Once this is done, we've basically rewound the whole
	* pool and there is no way back.
	*
	* There are cases when we don't want to attempt and sync the
	* checkpointed uberblock to disk because we are opening a
	* pool as read-only. Specifically, verifying the checkpointed
	* state with zdb, and importing the checkpointed state to get
	* a "preview" of its content.
	*/
	if (spa_writeable(spa)) {
	vdev_t *rvd = spa->spa_root_vdev;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
	int svdcount = 0;
	int children = rvd->vdev_children;
	int c0 = random_in_range(children);

	for (int c = 0; c < children; c++) {
	vdev_t *vd = rvd->vdev_child[(c0 + c) % children];

	/* Stop when revisiting the first vdev */
	if (c > 0 && svd[0] == vd)
	break;

	if (vd->vdev_ms_array == 0 \|\| vd->vdev_islog \|\|
	!vdev_is_concrete(vd))
	continue;

	svd[svdcount++] = vd;
	if (svdcount == SPA_SYNC_MIN_VDEVS)
	break;
	}
	error = vdev_config_sync(svd, svdcount, spa->spa_first_txg);
	if (error == 0)
	spa->spa_last_synced_guid = rvd->vdev_guid;
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "failed to write checkpointed "
	"uberblock to the vdev labels [error=%d]", error);
	return (error);
	}
	}

	return (0);
	}

	static int
	spa_ld_mos_with_trusted_config(spa_t *spa, spa_import_type_t type,
	boolean_t *update_config_cache)
	{
	int error;

	/*
	* Parse the config for pool, open and validate vdevs,
	* select an uberblock, and use that uberblock to open
	* the MOS.
	*/
	error = spa_ld_mos_init(spa, type);
	if (error != 0)
	return (error);

	/*
	* Retrieve the trusted config stored in the MOS and use it to create
	* a new, exact version of the vdev tree, then reopen all vdevs.
	*/
	error = spa_ld_trusted_config(spa, type, B_FALSE);
	if (error == EAGAIN) {
	if (update_config_cache != NULL)
	*update_config_cache = B_TRUE;

	/*
	* Redo the loading process with the trusted config if it is
	* too different from the untrusted config.
	*/
	spa_ld_prepare_for_reload(spa);
	spa_load_note(spa, "RELOADING");
	error = spa_ld_mos_init(spa, type);
	if (error != 0)
	return (error);

	error = spa_ld_trusted_config(spa, type, B_TRUE);
	if (error != 0)
	return (error);

	} else if (error != 0) {
	return (error);
	}

	return (0);
	}

	/*
	* Load an existing storage pool, using the config provided. This config
	* describes which vdevs are part of the pool and is later validated against
	* partial configs present in each vdev's label and an entire copy of the
	* config stored in the MOS.
	*/
	static int
	spa_load_impl(spa_t spa, spa_import_type_t type, const char *ereport)
	{
	int error = 0;
	boolean_t missing_feat_write = B_FALSE;
	boolean_t checkpoint_rewind =
	(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
	boolean_t update_config_cache = B_FALSE;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);

	spa_load_note(spa, "LOADING");

	error = spa_ld_mos_with_trusted_config(spa, type, &update_config_cache);
	if (error != 0)
	return (error);

	/*
	* If we are rewinding to the checkpoint then we need to repeat
	* everything we've done so far in this function but this time
	* selecting the checkpointed uberblock and using that to open
	* the MOS.
	*/
	if (checkpoint_rewind) {
	/*
	* If we are rewinding to the checkpoint update config cache
	* anyway.
	*/
	update_config_cache = B_TRUE;

	/*
	* Extract the checkpointed uberblock from the current MOS
	* and use this as the pool's uberblock from now on. If the
	* pool is imported as writeable we also write the checkpoint
	* uberblock to the labels, making the rewind permanent.
	*/
	error = spa_ld_checkpoint_rewind(spa);
	if (error != 0)
	return (error);

	/*
	* Redo the loading process again with the
	* checkpointed uberblock.
	*/
	spa_ld_prepare_for_reload(spa);
	spa_load_note(spa, "LOADING checkpointed uberblock");
	error = spa_ld_mos_with_trusted_config(spa, type, NULL);
	if (error != 0)
	return (error);
	}

	/*
	* Retrieve the checkpoint txg if the pool has a checkpoint.
	*/
	error = spa_ld_read_checkpoint_txg(spa);
	if (error != 0)
	return (error);

	/*
	* Retrieve the mapping of indirect vdevs. Those vdevs were removed
	* from the pool and their contents were re-mapped to other vdevs. Note
	* that everything that we read before this step must have been
	* rewritten on concrete vdevs after the last device removal was
	* initiated. Otherwise we could be reading from indirect vdevs before
	* we have loaded their mappings.
	*/
	error = spa_ld_open_indirect_vdev_metadata(spa);
	if (error != 0)
	return (error);

	/*
	* Retrieve the full list of active features from the MOS and check if
	* they are all supported.
	*/
	error = spa_ld_check_features(spa, &missing_feat_write);
	if (error != 0)
	return (error);

	/*
	* Load several special directories from the MOS needed by the dsl_pool
	* layer.
	*/
	error = spa_ld_load_special_directories(spa);
	if (error != 0)
	return (error);

	/*
	* Retrieve pool properties from the MOS.
	*/
	error = spa_ld_get_props(spa);
	if (error != 0)
	return (error);

	/*
	* Retrieve the list of auxiliary devices - cache devices and spares -
	* and open them.
	*/
	error = spa_ld_open_aux_vdevs(spa, type);
	if (error != 0)
	return (error);

	/*
	* Load the metadata for all vdevs. Also check if unopenable devices
	* should be autoreplaced.
	*/
	error = spa_ld_load_vdev_metadata(spa);
	if (error != 0)
	return (error);

	error = spa_ld_load_dedup_tables(spa);
	if (error != 0)
	return (error);

	/*
	* Verify the logs now to make sure we don't have any unexpected errors
	* when we claim log blocks later.
	*/
	error = spa_ld_verify_logs(spa, type, ereport);
	if (error != 0)
	return (error);

	if (missing_feat_write) {
	ASSERT(spa->spa_load_state == SPA_LOAD_TRYIMPORT);

	/*
	* At this point, we know that we can open the pool in
	* read-only mode but not read-write mode. We now have enough
	* information and can return to userland.
	*/
	return (spa_vdev_err(spa->spa_root_vdev, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP));
	}

	/*
	* Traverse the last txgs to make sure the pool was left off in a safe
	* state. When performing an extreme rewind, we verify the whole pool,
	* which can take a very long time.
	*/
	error = spa_ld_verify_pool_data(spa);
	if (error != 0)
	return (error);

	/*
	* Calculate the deflated space for the pool. This must be done before
	* we write anything to the pool because we'd need to update the space
	* accounting using the deflated sizes.
	*/
	spa_update_dspace(spa);

	/*
	* We have now retrieved all the information we needed to open the
	* pool. If we are importing the pool in read-write mode, a few
	* additional steps must be performed to finish the import.
	*/
	if (spa_writeable(spa) && (spa->spa_load_state == SPA_LOAD_RECOVER \|\|
	spa->spa_load_max_txg == UINT64_MAX)) {
	uint64_t config_cache_txg = spa->spa_config_txg;

	ASSERT(spa->spa_load_state != SPA_LOAD_TRYIMPORT);

	/*
	* In case of a checkpoint rewind, log the original txg
	* of the checkpointed uberblock.
	*/
	if (checkpoint_rewind) {
	spa_history_log_internal(spa, "checkpoint rewind",
	NULL, "rewound state to txg=%llu",
	(u_longlong_t)spa->spa_uberblock.ub_checkpoint_txg);
	}

	/*
	* Traverse the ZIL and claim all blocks.
	*/
	spa_ld_claim_log_blocks(spa);

	/*
	* Kick-off the syncing thread.
	*/
	spa->spa_sync_on = B_TRUE;
	txg_sync_start(spa->spa_dsl_pool);
	mmp_thread_start(spa);

	/*
	* Wait for all claims to sync. We sync up to the highest
	* claimed log block birth time so that claimed log blocks
	* don't appear to be from the future. spa_claim_max_txg
	* will have been set for us by ZIL traversal operations
	* performed above.
	*/
	txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg);

	/*
	* Check if we need to request an update of the config. On the
	* next sync, we would update the config stored in vdev labels
	* and the cachefile (by default /etc/zfs/zpool.cache).
	*/
	spa_ld_check_for_config_update(spa, config_cache_txg,
	update_config_cache);

	/*
	* Check if a rebuild was in progress and if so resume it.
	* Then check all DTLs to see if anything needs resilvering.
	* The resilver will be deferred if a rebuild was started.
	*/
	if (vdev_rebuild_active(spa->spa_root_vdev)) {
	vdev_rebuild_restart(spa);
	} else if (!dsl_scan_resilvering(spa->spa_dsl_pool) &&
	vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
	spa_async_request(spa, SPA_ASYNC_RESILVER);
	}

	/*
	* Log the fact that we booted up (so that we can detect if
	* we rebooted in the middle of an operation).
	*/
	spa_history_log_version(spa, "open", NULL);

	spa_restart_removal(spa);
	spa_spawn_aux_threads(spa);

	/*
	* Delete any inconsistent datasets.
	*
	* Note:
	* Since we may be issuing deletes for clones here,
	* we make sure to do so after we've spawned all the
	* auxiliary threads above (from which the livelist
	* deletion zthr is part of).
	*/
	(void) dmu_objset_find(spa_name(spa),
	dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN);

	/*
	* Clean up any stale temporary dataset userrefs.
	*/
	dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_initialize_restart(spa->spa_root_vdev);
	vdev_trim_restart(spa->spa_root_vdev);
	vdev_autotrim_restart(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	spa_import_progress_remove(spa_guid(spa));
	spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);

	spa_load_note(spa, "LOADED");

	return (0);
	}

	static int
	spa_load_retry(spa_t *spa, spa_load_state_t state)
	{
	spa_mode_t mode = spa->spa_mode;

	spa_unload(spa);
	spa_deactivate(spa);

	spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1;

	spa_activate(spa, mode);
	spa_async_suspend(spa);

	spa_load_note(spa, "spa_load_retry: rewind, max txg: %llu",
	(u_longlong_t)spa->spa_load_max_txg);

	return (spa_load(spa, state, SPA_IMPORT_EXISTING));
	}

	/*
	* If spa_load() fails this function will try loading prior txg's. If
	* 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool
	* will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this
	* function will not rewind the pool and will return the same error as
	* spa_load().
	*/
	static int
	spa_load_best(spa_t *spa, spa_load_state_t state, uint64_t max_request,
	int rewind_flags)
	{
	nvlist_t *loadinfo = NULL;
	nvlist_t *config = NULL;
	int load_error, rewind_error;
	uint64_t safe_rewind_txg;
	uint64_t min_txg;

	if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) {
	spa->spa_load_max_txg = spa->spa_load_txg;
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	} else {
	spa->spa_load_max_txg = max_request;
	if (max_request != UINT64_MAX)
	spa->spa_extreme_rewind = B_TRUE;
	}

	load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING);
	if (load_error == 0)
	return (0);
	if (load_error == ZFS_ERR_NO_CHECKPOINT) {
	/*
	* When attempting checkpoint-rewind on a pool with no
	* checkpoint, we should not attempt to load uberblocks
	* from previous txgs when spa_load fails.
	*/
	ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}

	if (spa->spa_root_vdev != NULL)
	config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);

	spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg;
	spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp;

	if (rewind_flags & ZPOOL_NEVER_REWIND) {
	nvlist_free(config);
	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}

	if (state == SPA_LOAD_RECOVER) {
	/* Price of rolling back is discarding txgs, including log */
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	} else {
	/*
	* If we aren't rolling back save the load info from our first
	* import attempt so that we can restore it after attempting
	* to rewind.
	*/
	loadinfo = spa->spa_load_info;
	spa->spa_load_info = fnvlist_alloc();
	}

	spa->spa_load_max_txg = spa->spa_last_ubsync_txg;
	safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE;
	min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ?
	TXG_INITIAL : safe_rewind_txg;

	/*
	* Continue as long as we're finding errors, we're still within
	* the acceptable rewind range, and we're still finding uberblocks
	*/
	while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg &&
	spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) {
	if (spa->spa_load_max_txg < safe_rewind_txg)
	spa->spa_extreme_rewind = B_TRUE;
	rewind_error = spa_load_retry(spa, state);
	}

	spa->spa_extreme_rewind = B_FALSE;
	spa->spa_load_max_txg = UINT64_MAX;

	if (config && (rewind_error \|\| state != SPA_LOAD_RECOVER))
	spa_config_set(spa, config);
	else
	nvlist_free(config);

	if (state == SPA_LOAD_RECOVER) {
	ASSERT3P(loadinfo, ==, NULL);
	spa_import_progress_remove(spa_guid(spa));
	return (rewind_error);
	} else {
	/* Store the rewind info as part of the initial load info */
	fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO,
	spa->spa_load_info);

	/* Restore the initial load info */
	fnvlist_free(spa->spa_load_info);
	spa->spa_load_info = loadinfo;

	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}
	}

	/*
	* Pool Open/Import
	*
	* The import case is identical to an open except that the configuration is sent
	* down from userland, instead of grabbed from the configuration cache. For the
	* case of an open, the pool configuration will exist in the
	* POOL_STATE_UNINITIALIZED state.
	*
	* The stats information (gen/count/ustats) is used to gather vdev statistics at
	* the same time open the pool, without having to keep around the spa_t in some
	* ambiguous state.
	*/
	static int
	spa_open_common(const char pool, spa_t spapp, const void tag,
	nvlist_t nvpolicy, nvlist_t *config)
	{
	spa_t *spa;
	spa_load_state_t state = SPA_LOAD_OPEN;
	int error;
	int locked = B_FALSE;
	int firstopen = B_FALSE;

	*spapp = NULL;

	/*
	* As disgusting as this is, we need to support recursive calls to this
	* function because dsl_dir_open() is called during spa_load(), and ends
	* up calling spa_open() again. The real fix is to figure out how to
	* avoid dsl_dir_open() calling this in the first place.
	*/
	if (MUTEX_NOT_HELD(&spa_namespace_lock)) {
	mutex_enter(&spa_namespace_lock);
	locked = B_TRUE;
	}

	if ((spa = spa_lookup(pool)) == NULL) {
	if (locked)
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
	zpool_load_policy_t policy;

	firstopen = B_TRUE;

	zpool_get_load_policy(nvpolicy ? nvpolicy : spa->spa_config,
	&policy);
	if (policy.zlp_rewind & ZPOOL_DO_REWIND)
	state = SPA_LOAD_RECOVER;

	spa_activate(spa, spa_mode_global);

	if (state != SPA_LOAD_RECOVER)
	spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
	spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;

	zfs_dbgmsg("spa_open_common: opening %s", pool);
	error = spa_load_best(spa, state, policy.zlp_txg,
	policy.zlp_rewind);

	if (error == EBADF) {
	/*
	* If vdev_validate() returns failure (indicated by
	* EBADF), it indicates that one of the vdevs indicates
	* that the pool has been exported or destroyed. If
	* this is the case, the config cache is out of sync and
	* we should remove the pool from the namespace.
	*/
	spa_unload(spa);
	spa_deactivate(spa);
	- spa_write_cachefile(spa, B_TRUE, B_TRUE);
	+ spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE);
	spa_remove(spa);
	if (locked)
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (error) {
	/*
	* We can't open the pool, but we still have useful
	* information: the state of each vdev after the
	* attempted vdev_open(). Return this to the user.
	*/
	if (config != NULL && spa->spa_config) {
	*config = fnvlist_dup(spa->spa_config);
	fnvlist_add_nvlist(*config,
	ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	}
	spa_unload(spa);
	spa_deactivate(spa);
	spa->spa_last_open_failed = error;
	if (locked)
	mutex_exit(&spa_namespace_lock);
	*spapp = NULL;
	return (error);
	}
	}

	spa_open_ref(spa, tag);

	if (config != NULL)
	*config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);

	/*
	* If we've recovered the pool, pass back any information we
	* gathered while doing the load.
	*/
	if (state == SPA_LOAD_RECOVER) {
	fnvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	}

	if (locked) {
	spa->spa_last_open_failed = 0;
	spa->spa_last_ubsync_txg = 0;
	spa->spa_load_txg = 0;
	mutex_exit(&spa_namespace_lock);
	}

	if (firstopen)
	zvol_create_minors_recursive(spa_name(spa));

	*spapp = spa;

	return (0);
	}

	int
	spa_open_rewind(const char name, spa_t spapp, const void tag,
	nvlist_t policy, nvlist_t *config)
	{
	return (spa_open_common(name, spapp, tag, policy, config));
	}

	int
	spa_open(const char name, spa_t spapp, const void tag)
	{
	return (spa_open_common(name, spapp, tag, NULL, NULL));
	}

	/*
	* Lookup the given spa_t, incrementing the inject count in the process,
	* preventing it from being exported or destroyed.
	*/
	spa_t *
	spa_inject_addref(char *name)
	{
	spa_t *spa;

	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(name)) == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (NULL);
	}
	spa->spa_inject_ref++;
	mutex_exit(&spa_namespace_lock);

	return (spa);
	}

	void
	spa_inject_delref(spa_t *spa)
	{
	mutex_enter(&spa_namespace_lock);
	spa->spa_inject_ref--;
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Add spares device information to the nvlist.
	*/
	static void
	spa_add_spares(spa_t spa, nvlist_t config)
	{
	nvlist_t **spares;
	uint_t i, nspares;
	nvlist_t *nvroot;
	uint64_t guid;
	vdev_stat_t *vs;
	uint_t vsc;
	uint64_t pool;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (spa->spa_spares.sav_count == 0)
	return;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, &spares, &nspares));
	if (nspares != 0) {
	fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	(const nvlist_t * const *)spares, nspares);
	VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares));

	/*
	* Go through and find any spares which have since been
	* repurposed as an active spare. If this is the case, update
	* their status appropriately.
	*/
	for (i = 0; i < nspares; i++) {
	guid = fnvlist_lookup_uint64(spares[i],
	ZPOOL_CONFIG_GUID);
	if (spa_spare_exists(guid, &pool, NULL) &&
	pool != 0ULL) {
	VERIFY0(nvlist_lookup_uint64_array(spares[i],
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs,
	&vsc));
	vs->vs_state = VDEV_STATE_CANT_OPEN;
	vs->vs_aux = VDEV_AUX_SPARED;
	}
	}
	}
	}

	/*
	* Add l2cache device information to the nvlist, including vdev stats.
	*/
	static void
	spa_add_l2cache(spa_t spa, nvlist_t config)
	{
	nvlist_t **l2cache;
	uint_t i, j, nl2cache;
	nvlist_t *nvroot;
	uint64_t guid;
	vdev_t *vd;
	vdev_stat_t *vs;
	uint_t vsc;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (spa->spa_l2cache.sav_count == 0)
	return;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
	if (nl2cache != 0) {
	fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	(const nvlist_t * const *)l2cache, nl2cache);
	VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache));

	/*
	* Update level 2 cache device stats.
	*/

	for (i = 0; i < nl2cache; i++) {
	guid = fnvlist_lookup_uint64(l2cache[i],
	ZPOOL_CONFIG_GUID);

	vd = NULL;
	for (j = 0; j < spa->spa_l2cache.sav_count; j++) {
	if (guid ==
	spa->spa_l2cache.sav_vdevs[j]->vdev_guid) {
	vd = spa->spa_l2cache.sav_vdevs[j];
	break;
	}
	}
	ASSERT(vd != NULL);

	VERIFY0(nvlist_lookup_uint64_array(l2cache[i],
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc));
	vdev_get_stats(vd, vs);
	vdev_config_generate_stats(vd, l2cache[i]);

	}
	}
	}

	static void
	spa_feature_stats_from_disk(spa_t spa, nvlist_t features)
	{
	zap_cursor_t zc;
	zap_attribute_t za;

	if (spa->spa_feat_for_read_obj != 0) {
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_feat_for_read_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	ASSERT(za.za_integer_length == sizeof (uint64_t) &&
	za.za_num_integers == 1);
	VERIFY0(nvlist_add_uint64(features, za.za_name,
	za.za_first_integer));
	}
	zap_cursor_fini(&zc);
	}

	if (spa->spa_feat_for_write_obj != 0) {
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_feat_for_write_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	ASSERT(za.za_integer_length == sizeof (uint64_t) &&
	za.za_num_integers == 1);
	VERIFY0(nvlist_add_uint64(features, za.za_name,
	za.za_first_integer));
	}
	zap_cursor_fini(&zc);
	}
	}

	static void
	spa_feature_stats_from_cache(spa_t spa, nvlist_t features)
	{
	int i;

	for (i = 0; i < SPA_FEATURES; i++) {
	zfeature_info_t feature = spa_feature_table[i];
	uint64_t refcount;

	if (feature_get_refcount(spa, &feature, &refcount) != 0)
	continue;

	VERIFY0(nvlist_add_uint64(features, feature.fi_guid, refcount));
	}
	}

	/*
	* Store a list of pool features and their reference counts in the
	* config.
	*
	* The first time this is called on a spa, allocate a new nvlist, fetch
	* the pool features and reference counts from disk, then save the list
	* in the spa. In subsequent calls on the same spa use the saved nvlist
	* and refresh its values from the cached reference counts. This
	* ensures we don't block here on I/O on a suspended pool so 'zpool
	* clear' can resume the pool.
	*/
	static void
	spa_add_feature_stats(spa_t spa, nvlist_t config)
	{
	nvlist_t *features;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	mutex_enter(&spa->spa_feat_stats_lock);
	features = spa->spa_feat_stats;

	if (features != NULL) {
	spa_feature_stats_from_cache(spa, features);
	} else {
	VERIFY0(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP));
	spa->spa_feat_stats = features;
	spa_feature_stats_from_disk(spa, features);
	}

	VERIFY0(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
	features));

	mutex_exit(&spa->spa_feat_stats_lock);
	}

	int
	spa_get_stats(const char name, nvlist_t *config,
	char *altroot, size_t buflen)
	{
	int error;
	spa_t *spa;

	*config = NULL;
	error = spa_open_common(name, &spa, FTAG, NULL, config);

	if (spa != NULL) {
	/*
	* This still leaves a window of inconsistency where the spares
	* or l2cache devices could change and the config would be
	* self-inconsistent.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	if (*config != NULL) {
	uint64_t loadtimes[2];

	loadtimes[0] = spa->spa_loaded_ts.tv_sec;
	loadtimes[1] = spa->spa_loaded_ts.tv_nsec;
	fnvlist_add_uint64_array(*config,
	ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2);

	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_ERRCOUNT,
	spa_get_errlog_size(spa));

	if (spa_suspended(spa)) {
	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_SUSPENDED,
	spa->spa_failmode);
	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_SUSPENDED_REASON,
	spa->spa_suspended);
	}

	spa_add_spares(spa, *config);
	spa_add_l2cache(spa, *config);
	spa_add_feature_stats(spa, *config);
	}
	}

	/*
	* We want to get the alternate root even for faulted pools, so we cheat
	* and call spa_lookup() directly.
	*/
	if (altroot) {
	if (spa == NULL) {
	mutex_enter(&spa_namespace_lock);
	spa = spa_lookup(name);
	if (spa)
	spa_altroot(spa, altroot, buflen);
	else
	altroot[0] = '\0';
	spa = NULL;
	mutex_exit(&spa_namespace_lock);
	} else {
	spa_altroot(spa, altroot, buflen);
	}
	}

	if (spa != NULL) {
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	spa_close(spa, FTAG);
	}

	return (error);
	}

	/*
	* Validate that the auxiliary device array is well formed. We must have an
	* array of nvlists, each which describes a valid leaf vdev. If this is an
	* import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be
	* specified, as long as they are well-formed.
	*/
	static int
	spa_validate_aux_devs(spa_t spa, nvlist_t nvroot, uint64_t crtxg, int mode,
	spa_aux_vdev_t sav, const char config, uint64_t version,
	vdev_labeltype_t label)
	{
	nvlist_t **dev;
	uint_t i, ndev;
	vdev_t *vd;
	int error;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	/*
	* It's acceptable to have no devs specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0)
	return (0);

	if (ndev == 0)
	return (SET_ERROR(EINVAL));

	/*
	* Make sure the pool is formatted with a version that supports this
	* device type.
	*/
	if (spa_version(spa) < version)
	return (SET_ERROR(ENOTSUP));

	/*
	* Set the pending device list so we correctly handle device in-use
	* checking.
	*/
	sav->sav_pending = dev;
	sav->sav_npending = ndev;

	for (i = 0; i < ndev; i++) {
	if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0,
	mode)) != 0)
	goto out;

	if (!vd->vdev_ops->vdev_op_leaf) {
	vdev_free(vd);
	error = SET_ERROR(EINVAL);
	goto out;
	}

	vd->vdev_top = vd;

	if ((error = vdev_open(vd)) == 0 &&
	(error = vdev_label_init(vd, crtxg, label)) == 0) {
	fnvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID,
	vd->vdev_guid);
	}

	vdev_free(vd);

	if (error &&
	(mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE))
	goto out;
	else
	error = 0;
	}

	out:
	sav->sav_pending = NULL;
	sav->sav_npending = 0;
	return (error);
	}

	static int
	spa_validate_aux(spa_t spa, nvlist_t nvroot, uint64_t crtxg, int mode)
	{
	int error;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode,
	&spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES,
	VDEV_LABEL_SPARE)) != 0) {
	return (error);
	}

	return (spa_validate_aux_devs(spa, nvroot, crtxg, mode,
	&spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE,
	VDEV_LABEL_L2CACHE));
	}

	static void
	spa_set_aux_vdevs(spa_aux_vdev_t sav, nvlist_t *devs, int ndevs,
	const char *config)
	{
	int i;

	if (sav->sav_config != NULL) {
	nvlist_t **olddevs;
	uint_t oldndevs;
	nvlist_t **newdevs;

	/*
	* Generate new dev list by concatenating with the
	* current dev list.
	*/
	VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, config,
	&olddevs, &oldndevs));

	newdevs = kmem_alloc(sizeof (void )
	(ndevs + oldndevs), KM_SLEEP);
	for (i = 0; i < oldndevs; i++)
	newdevs[i] = fnvlist_dup(olddevs[i]);
	for (i = 0; i < ndevs; i++)
	newdevs[i + oldndevs] = fnvlist_dup(devs[i]);

	fnvlist_remove(sav->sav_config, config);

	fnvlist_add_nvlist_array(sav->sav_config, config,
	(const nvlist_t * const *)newdevs, ndevs + oldndevs);
	for (i = 0; i < oldndevs + ndevs; i++)
	nvlist_free(newdevs[i]);
	kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *));
	} else {
	/*
	* Generate a new dev list.
	*/
	sav->sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(sav->sav_config, config,
	(const nvlist_t * const *)devs, ndevs);
	}
	}

	/*
	* Stop and drop level 2 ARC devices
	*/
	void
	spa_l2cache_drop(spa_t *spa)
	{
	vdev_t *vd;
	int i;
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	for (i = 0; i < sav->sav_count; i++) {
	uint64_t pool;

	vd = sav->sav_vdevs[i];
	ASSERT(vd != NULL);

	if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
	pool != 0ULL && l2arc_vdev_present(vd))
	l2arc_remove_vdev(vd);
	}
	}

	/*
	* Verify encryption parameters for spa creation. If we are encrypting, we must
	* have the encryption feature flag enabled.
	*/
	static int
	spa_create_check_encryption_params(dsl_crypto_params_t *dcp,
	boolean_t has_encryption)
	{
	if (dcp->cp_crypt != ZIO_CRYPT_OFF &&
	dcp->cp_crypt != ZIO_CRYPT_INHERIT &&
	!has_encryption)
	return (SET_ERROR(ENOTSUP));

	return (dmu_objset_create_crypt_check(NULL, dcp, NULL));
	}

	/*
	* Pool Creation
	*/
	int
	spa_create(const char pool, nvlist_t nvroot, nvlist_t *props,
	nvlist_t zplprops, dsl_crypto_params_t dcp)
	{
	spa_t *spa;
	char *altroot = NULL;
	vdev_t *rvd;
	dsl_pool_t *dp;
	dmu_tx_t *tx;
	int error = 0;
	uint64_t txg = TXG_INITIAL;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;
	uint64_t version, obj, ndraid = 0;
	boolean_t has_features;
	boolean_t has_encryption;
	boolean_t has_allocclass;
	spa_feature_t feat;
	char *feat_name;
	char *poolname;
	nvlist_t *nvl;

	if (props == NULL \|\|
	nvlist_lookup_string(props, "tname", &poolname) != 0)
	poolname = (char *)pool;

	/*
	* If this pool already exists, return failure.
	*/
	mutex_enter(&spa_namespace_lock);
	if (spa_lookup(poolname) != NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EEXIST));
	}

	/*
	* Allocate a new spa_t structure.
	*/
	nvl = fnvlist_alloc();
	fnvlist_add_string(nvl, ZPOOL_CONFIG_POOL_NAME, pool);
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
	spa = spa_add(poolname, nvl, altroot);
	fnvlist_free(nvl);
	spa_activate(spa, spa_mode_global);

	if (props && (error = spa_prop_validate(spa, props))) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Temporary pool names should never be written to disk.
	*/
	if (poolname != pool)
	spa->spa_import_flags \|= ZFS_IMPORT_TEMP_NAME;

	has_features = B_FALSE;
	has_encryption = B_FALSE;
	has_allocclass = B_FALSE;
	for (nvpair_t *elem = nvlist_next_nvpair(props, NULL);
	elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
	if (zpool_prop_feature(nvpair_name(elem))) {
	has_features = B_TRUE;

	feat_name = strchr(nvpair_name(elem), '@') + 1;
	VERIFY0(zfeature_lookup_name(feat_name, &feat));
	if (feat == SPA_FEATURE_ENCRYPTION)
	has_encryption = B_TRUE;
	if (feat == SPA_FEATURE_ALLOCATION_CLASSES)
	has_allocclass = B_TRUE;
	}
	}

	/* verify encryption params, if they were provided */
	if (dcp != NULL) {
	error = spa_create_check_encryption_params(dcp, has_encryption);
	if (error != 0) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}
	}
	if (!has_allocclass && zfs_special_devs(nvroot, NULL)) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (ENOTSUP);
	}

	if (has_features \|\| nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) {
	version = SPA_VERSION;
	}
	ASSERT(SPA_VERSION_IS_SUPPORTED(version));

	spa->spa_first_txg = txg;
	spa->spa_uberblock.ub_txg = txg - 1;
	spa->spa_uberblock.ub_version = version;
	spa->spa_ubsync = spa->spa_uberblock;
	spa->spa_load_state = SPA_LOAD_CREATE;
	spa->spa_removing_phys.sr_state = DSS_NONE;
	spa->spa_removing_phys.sr_removing_vdev = -1;
	spa->spa_removing_phys.sr_prev_indirect_vdev = -1;
	spa->spa_indirect_vdevs_loaded = B_TRUE;

	/*
	* Create "The Godfather" zio to hold all async IOs
	*/
	spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
	KM_SLEEP);
	for (int i = 0; i < max_ncpus; i++) {
	spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);
	}

	/*
	* Create the root vdev.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD);

	ASSERT(error != 0 \|\| rvd != NULL);
	ASSERT(error != 0 \|\| spa->spa_root_vdev == rvd);

	if (error == 0 && !zfs_allocatable_devs(nvroot))
	error = SET_ERROR(EINVAL);

	if (error == 0 &&
	(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
	(error = vdev_draid_spare_create(nvroot, rvd, &ndraid, 0)) == 0 &&
	(error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) == 0) {
	/*
	* instantiate the metaslab groups (this will dirty the vdevs)
	* we can no longer error exit past this point
	*/
	for (int c = 0; error == 0 && c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	vdev_metaslab_set_size(vd);
	vdev_expand(vd, txg);
	}
	}

	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Get the list of spares, if specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	spa->spa_spares.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	nspares);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_spares.sav_sync = B_TRUE;
	}

	/*
	* Get the list of level 2 cache devices, if specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	VERIFY0(nvlist_alloc(&spa->spa_l2cache.sav_config,
	NV_UNIQUE_NAME, KM_SLEEP));
	fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
	nl2cache);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	spa->spa_is_initializing = B_TRUE;
	spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, dcp, txg);
	spa->spa_is_initializing = B_FALSE;

	/*
	* Create DDTs (dedup tables).
	*/
	ddt_create(spa);

	spa_update_dspace(spa);

	tx = dmu_tx_create_assigned(dp, txg);

	/*
	* Create the pool's history object.
	*/
	if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history)
	spa_history_create_obj(spa, tx);

	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE);
	spa_history_log_version(spa, "create", tx);

	/*
	* Create the pool config object.
	*/
	spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE,
	DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx);

	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
	sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add pool config");
	}

	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
	sizeof (uint64_t), 1, &version, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add pool version");
	}

	/* Newly created pools with the right version are always deflated. */
	if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
	spa->spa_deflate = TRUE;
	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
	sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add deflate");
	}
	}

	/*
	* Create the deferred-free bpobj. Turn off compression
	* because sync-to-convergence takes longer if the blocksize
	* keeps changing.
	*/
	obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx);
	dmu_object_set_compress(spa->spa_meta_objset, obj,
	ZIO_COMPRESS_OFF, tx);
	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ,
	sizeof (uint64_t), 1, &obj, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add bpobj");
	}
	VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
	spa->spa_meta_objset, obj));

	/*
	* Generate some random noise for salted checksums to operate on.
	*/
	(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
	sizeof (spa->spa_cksum_salt.zcs_bytes));

	/*
	* Set pool properties.
	*/
	spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
	spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
	spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
	spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
	spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST);
	spa->spa_autotrim = zpool_prop_default_numeric(ZPOOL_PROP_AUTOTRIM);

	if (props != NULL) {
	spa_configfile_set(spa, props, B_FALSE);
	spa_sync_props(props, tx);
	}

	for (int i = 0; i < ndraid; i++)
	spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);

	dmu_tx_commit(tx);

	spa->spa_sync_on = B_TRUE;
	txg_sync_start(dp);
	mmp_thread_start(spa);
	txg_wait_synced(dp, txg);

	spa_spawn_aux_threads(spa);

	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE);

	/*
	* Don't count references from objsets that are already closed
	* and are making their way through the eviction process.
	*/
	spa_evicting_os_wait(spa);
	spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
	spa->spa_load_state = SPA_LOAD_NONE;

	spa_import_os(spa);

	mutex_exit(&spa_namespace_lock);

	return (0);
	}

	/*
	* Import a non-root pool into the system.
	*/
	int
	spa_import(char pool, nvlist_t config, nvlist_t *props, uint64_t flags)
	{
	spa_t *spa;
	char *altroot = NULL;
	spa_load_state_t state = SPA_LOAD_IMPORT;
	zpool_load_policy_t policy;
	spa_mode_t mode = spa_mode_global;
	uint64_t readonly = B_FALSE;
	int error;
	nvlist_t *nvroot;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	/*
	* If a pool with this name exists, return failure.
	*/
	mutex_enter(&spa_namespace_lock);
	if (spa_lookup(pool) != NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EEXIST));
	}

	/*
	* Create and initialize the spa structure.
	*/
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
	(void) nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
	if (readonly)
	mode = SPA_MODE_READ;
	spa = spa_add(pool, config, altroot);
	spa->spa_import_flags = flags;

	/*
	* Verbatim import - Take a pool and insert it into the namespace
	* as if it had been loaded at boot.
	*/
	if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) {
	if (props != NULL)
	spa_configfile_set(spa, props, B_FALSE);

	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
	zfs_dbgmsg("spa_import: verbatim import of %s", pool);
	mutex_exit(&spa_namespace_lock);
	return (0);
	}

	spa_activate(spa, mode);

	/*
	* Don't start async tasks until we know everything is healthy.
	*/
	spa_async_suspend(spa);

	zpool_get_load_policy(config, &policy);
	if (policy.zlp_rewind & ZPOOL_DO_REWIND)
	state = SPA_LOAD_RECOVER;

	spa->spa_config_source = SPA_CONFIG_SRC_TRYIMPORT;

	if (state != SPA_LOAD_RECOVER) {
	spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
	zfs_dbgmsg("spa_import: importing %s", pool);
	} else {
	zfs_dbgmsg("spa_import: importing %s, max_txg=%lld "
	"(RECOVERY MODE)", pool, (longlong_t)policy.zlp_txg);
	}
	error = spa_load_best(spa, state, policy.zlp_txg, policy.zlp_rewind);

	/*
	* Propagate anything learned while loading the pool and pass it
	* back to caller (i.e. rewind info, missing devices, etc).
	*/
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	/*
	* Toss any existing sparelist, as it doesn't have any validity
	* anymore, and conflicts with spa_has_spare().
	*/
	if (spa->spa_spares.sav_config) {
	nvlist_free(spa->spa_spares.sav_config);
	spa->spa_spares.sav_config = NULL;
	spa_load_spares(spa);
	}
	if (spa->spa_l2cache.sav_config) {
	nvlist_free(spa->spa_l2cache.sav_config);
	spa->spa_l2cache.sav_config = NULL;
	spa_load_l2cache(spa);
	}

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (props != NULL)
	spa_configfile_set(spa, props, B_FALSE);

	if (error != 0 \|\| (props && spa_writeable(spa) &&
	(error = spa_prop_set(spa, props)))) {
	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	spa_async_resume(spa);

	/*
	* Override any spares and level 2 cache devices as specified by
	* the user, as these may have correct device names/devids, etc.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	if (spa->spa_spares.sav_config)
	fnvlist_remove(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES);
	else
	spa->spa_spares.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	nspares);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_spares.sav_sync = B_TRUE;
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	if (spa->spa_l2cache.sav_config)
	fnvlist_remove(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE);
	else
	spa->spa_l2cache.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
	nl2cache);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	/*
	* Check for any removed devices.
	*/
	if (spa->spa_autoreplace) {
	spa_aux_check_removed(&spa->spa_spares);
	spa_aux_check_removed(&spa->spa_l2cache);
	}

	if (spa_writeable(spa)) {
	/*
	* Update the config cache to include the newly-imported pool.
	*/
	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
	}

	/*
	* It's possible that the pool was expanded while it was exported.
	* We kick off an async task to handle this for us.
	*/
	spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);

	spa_history_log_version(spa, "import", NULL);

	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);

	mutex_exit(&spa_namespace_lock);

	zvol_create_minors_recursive(pool);

	spa_import_os(spa);

	return (0);
	}

	nvlist_t *
	spa_tryimport(nvlist_t *tryconfig)
	{
	nvlist_t *config = NULL;
	char poolname, cachefile;
	spa_t *spa;
	uint64_t state;
	int error;
	zpool_load_policy_t policy;

	if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname))
	return (NULL);

	if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state))
	return (NULL);

	/*
	* Create and initialize the spa structure.
	*/
	mutex_enter(&spa_namespace_lock);
	spa = spa_add(TRYIMPORT_NAME, tryconfig, NULL);
	spa_activate(spa, SPA_MODE_READ);

	/*
	* Rewind pool if a max txg was provided.
	*/
	zpool_get_load_policy(spa->spa_config, &policy);
	if (policy.zlp_txg != UINT64_MAX) {
	spa->spa_load_max_txg = policy.zlp_txg;
	spa->spa_extreme_rewind = B_TRUE;
	zfs_dbgmsg("spa_tryimport: importing %s, max_txg=%lld",
	poolname, (longlong_t)policy.zlp_txg);
	} else {
	zfs_dbgmsg("spa_tryimport: importing %s", poolname);
	}

	if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_CACHEFILE, &cachefile)
	== 0) {
	zfs_dbgmsg("spa_tryimport: using cachefile '%s'", cachefile);
	spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;
	} else {
	spa->spa_config_source = SPA_CONFIG_SRC_SCAN;
	}

	error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING);

	/*
	* If 'tryconfig' was at least parsable, return the current config.
	*/
	if (spa->spa_root_vdev != NULL) {
	config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, poolname);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, state);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
	spa->spa_uberblock.ub_timestamp);
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA,
	spa->spa_errata);

	/*
	* If the bootfs property exists on this pool then we
	* copy it out so that external consumers can tell which
	* pools are bootable.
	*/
	if ((!error \|\| error == EEXIST) && spa->spa_bootfs) {
	char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	/*
	* We have to play games with the name since the
	* pool was opened as TRYIMPORT_NAME.
	*/
	if (dsl_dsobj_to_dsname(spa_name(spa),
	spa->spa_bootfs, tmpname) == 0) {
	char *cp;
	char *dsname;

	dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	cp = strchr(tmpname, '/');
	if (cp == NULL) {
	(void) strlcpy(dsname, tmpname,
	MAXPATHLEN);
	} else {
	(void) snprintf(dsname, MAXPATHLEN,
	"%s/%s", poolname, ++cp);
	}
	fnvlist_add_string(config, ZPOOL_CONFIG_BOOTFS,
	dsname);
	kmem_free(dsname, MAXPATHLEN);
	}
	kmem_free(tmpname, MAXPATHLEN);
	}

	/*
	* Add the list of hot spares and level 2 cache devices.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	spa_add_spares(spa, config);
	spa_add_l2cache(spa, config);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);

	return (config);
	}

	/*
	* Pool export/destroy
	*
	* The act of destroying or exporting a pool is very simple. We make sure there
	* is no more pending I/O and any references to the pool are gone. Then, we
	* update the pool state and sync all the labels to disk, removing the
	* configuration from the cache afterwards. If the 'hardforce' flag is set, then
	* we don't sync the labels or remove the configuration cache.
	*/
	static int
	spa_export_common(const char pool, int new_state, nvlist_t *oldconfig,
	boolean_t force, boolean_t hardforce)
	{
	int error;
	spa_t *spa;

	if (oldconfig)
	*oldconfig = NULL;

	if (!(spa_mode_global & SPA_MODE_WRITE))
	return (SET_ERROR(EROFS));

	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(pool)) == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (spa->spa_is_exporting) {
	/* the pool is being exported by another thread */
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ZFS_ERR_EXPORT_IN_PROGRESS));
	}
	spa->spa_is_exporting = B_TRUE;

	/*
	* Put a hold on the pool, drop the namespace lock, stop async tasks,
	* reacquire the namespace lock, and see if we can export.
	*/
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	spa_async_suspend(spa);
	if (spa->spa_zvol_taskq) {
	zvol_remove_minors(spa, spa_name(spa), B_TRUE);
	taskq_wait(spa->spa_zvol_taskq);
	}
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);

	if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	goto export_spa;
	/*
	* The pool will be in core if it's openable, in which case we can
	* modify its state. Objsets may be open only because they're dirty,
	* so we have to force it to sync before checking spa_refcnt.
	*/
	if (spa->spa_sync_on) {
	txg_wait_synced(spa->spa_dsl_pool, 0);
	spa_evicting_os_wait(spa);
	}

	/*
	* A pool cannot be exported or destroyed if there are active
	* references. If we are resetting a pool, allow references by
	* fault injection handlers.
	*/
	if (!spa_refcount_zero(spa) \|\| (spa->spa_inject_ref != 0)) {
	error = SET_ERROR(EBUSY);
	goto fail;
	}

	if (spa->spa_sync_on) {
	+ vdev_t *rvd = spa->spa_root_vdev;
	/*
	* A pool cannot be exported if it has an active shared spare.
	* This is to prevent other pools stealing the active spare
	* from an exported pool. At user's own will, such pool can
	* be forcedly exported.
	*/
	if (!force && new_state == POOL_STATE_EXPORTED &&
	spa_has_active_shared_spare(spa)) {
	error = SET_ERROR(EXDEV);
	goto fail;
	}

	/*
	* We're about to export or destroy this pool. Make sure
	* we stop all initialization and trim activity here before
	* we set the spa_final_txg. This will ensure that all
	* dirty data resulting from the initialization is
	* committed to disk before we unload the pool.
	*/
	- if (spa->spa_root_vdev != NULL) {
	- vdev_t *rvd = spa->spa_root_vdev;
	- vdev_initialize_stop_all(rvd, VDEV_INITIALIZE_ACTIVE);
	- vdev_trim_stop_all(rvd, VDEV_TRIM_ACTIVE);
	- vdev_autotrim_stop_all(spa);
	- vdev_rebuild_stop_all(spa);
	- }
	+ vdev_initialize_stop_all(rvd, VDEV_INITIALIZE_ACTIVE);
	+ vdev_trim_stop_all(rvd, VDEV_TRIM_ACTIVE);
	+ vdev_autotrim_stop_all(spa);
	+ vdev_rebuild_stop_all(spa);

	/*
	* We want this to be reflected on every label,
	* so mark them all dirty. spa_unload() will do the
	* final sync that pushes these changes out.
	*/
	if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa->spa_state = new_state;
	- vdev_config_dirty(spa->spa_root_vdev);
	+ vdev_config_dirty(rvd);
	spa_config_exit(spa, SCL_ALL, FTAG);
	}

	/*
	* If the log space map feature is enabled and the pool is
	* getting exported (but not destroyed), we want to spend some
	* time flushing as many metaslabs as we can in an attempt to
	* destroy log space maps and save import time. This has to be
	* done before we set the spa_final_txg, otherwise
	* spa_sync() -> spa_flush_metaslabs() may dirty the final TXGs.
	* spa_should_flush_logs_on_unload() should be called after
	* spa_state has been set to the new_state.
	*/
	if (spa_should_flush_logs_on_unload(spa))
	spa_unload_log_sm_flush_all(spa);

	if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa->spa_final_txg = spa_last_synced_txg(spa) +
	TXG_DEFER_SIZE + 1;
	spa_config_exit(spa, SCL_ALL, FTAG);
	}
	}

	export_spa:
	spa_export_os(spa);

	if (new_state == POOL_STATE_DESTROYED)
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY);
	else if (new_state == POOL_STATE_EXPORTED)
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT);

	if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
	spa_unload(spa);
	spa_deactivate(spa);
	}

	if (oldconfig && spa->spa_config)
	*oldconfig = fnvlist_dup(spa->spa_config);

	if (new_state != POOL_STATE_UNINITIALIZED) {
	if (!hardforce)
	- spa_write_cachefile(spa, B_TRUE, B_TRUE);
	+ spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE);
	spa_remove(spa);
	} else {
	/*
	* If spa_remove() is not called for this spa_t and
	* there is any possibility that it can be reused,
	* we make sure to reset the exporting flag.
	*/
	spa->spa_is_exporting = B_FALSE;
	}

	mutex_exit(&spa_namespace_lock);
	return (0);

	fail:
	spa->spa_is_exporting = B_FALSE;
	spa_async_resume(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Destroy a storage pool.
	*/
	int
	spa_destroy(const char *pool)
	{
	return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL,
	B_FALSE, B_FALSE));
	}

	/*
	* Export a storage pool.
	*/
	int
	spa_export(const char pool, nvlist_t *oldconfig, boolean_t force,
	boolean_t hardforce)
	{
	return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig,
	force, hardforce));
	}

	/*
	* Similar to spa_export(), this unloads the spa_t without actually removing it
	* from the namespace in any way.
	*/
	int
	spa_reset(const char *pool)
	{
	return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL,
	B_FALSE, B_FALSE));
	}

	/*
	* ==========================================================================
	* Device manipulation
	* ==========================================================================
	*/

	/*
	* This is called as a synctask to increment the draid feature flag
	*/
	static void
	spa_draid_feature_incr(void arg, dmu_tx_t tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	int draid = (int)(uintptr_t)arg;

	for (int c = 0; c < draid; c++)
	spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);
	}

	/*
	* Add a device to a storage pool.
	*/
	int
	spa_vdev_add(spa_t spa, nvlist_t nvroot)
	{
	uint64_t txg, ndraid = 0;
	int error;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t vd, tvd;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0,
	VDEV_ALLOC_ADD)) != 0)
	return (spa_vdev_exit(spa, NULL, txg, error));

	spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares,
	&nspares) != 0)
	nspares = 0;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache,
	&nl2cache) != 0)
	nl2cache = 0;

	if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0)
	return (spa_vdev_exit(spa, vd, txg, EINVAL));

	if (vd->vdev_children != 0 &&
	(error = vdev_create(vd, txg, B_FALSE)) != 0) {
	return (spa_vdev_exit(spa, vd, txg, error));
	}

	/*
	* The virtual dRAID spares must be added after vdev tree is created
	* and the vdev guids are generated. The guid of their associated
	* dRAID is stored in the config and used when opening the spare.
	*/
	if ((error = vdev_draid_spare_create(nvroot, vd, &ndraid,
	rvd->vdev_children)) == 0) {
	if (ndraid > 0 && nvlist_lookup_nvlist_array(nvroot,
	ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0)
	nspares = 0;
	} else {
	return (spa_vdev_exit(spa, vd, txg, error));
	}

	/*
	* We must validate the spares and l2cache devices after checking the
	* children. Otherwise, vdev_inuse() will blindly overwrite the spare.
	*/
	if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0)
	return (spa_vdev_exit(spa, vd, txg, error));

	/*
	* If we are in the middle of a device removal, we can only add
	* devices which match the existing devices in the pool.
	* If we are in the middle of a removal, or have some indirect
	* vdevs, we can not add raidz or dRAID top levels.
	*/
	if (spa->spa_vdev_removal != NULL \|\|
	spa->spa_removing_phys.sr_prev_indirect_vdev != -1) {
	for (int c = 0; c < vd->vdev_children; c++) {
	tvd = vd->vdev_child[c];
	if (spa->spa_vdev_removal != NULL &&
	tvd->vdev_ashift != spa->spa_max_ashift) {
	return (spa_vdev_exit(spa, vd, txg, EINVAL));
	}
	/* Fail if top level vdev is raidz or a dRAID */
	if (vdev_get_nparity(tvd) != 0)
	return (spa_vdev_exit(spa, vd, txg, EINVAL));

	/*
	* Need the top level mirror to be
	* a mirror of leaf vdevs only
	*/
	if (tvd->vdev_ops == &vdev_mirror_ops) {
	for (uint64_t cid = 0;
	cid < tvd->vdev_children; cid++) {
	vdev_t *cvd = tvd->vdev_child[cid];
	if (!cvd->vdev_ops->vdev_op_leaf) {
	return (spa_vdev_exit(spa, vd,
	txg, EINVAL));
	}
	}
	}
	}
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	tvd = vd->vdev_child[c];
	vdev_remove_child(vd, tvd);
	tvd->vdev_id = rvd->vdev_children;
	vdev_add_child(rvd, tvd);
	vdev_config_dirty(tvd);
	}

	if (nspares != 0) {
	spa_set_aux_vdevs(&spa->spa_spares, spares, nspares,
	ZPOOL_CONFIG_SPARES);
	spa_load_spares(spa);
	spa->spa_spares.sav_sync = B_TRUE;
	}

	if (nl2cache != 0) {
	spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache,
	ZPOOL_CONFIG_L2CACHE);
	spa_load_l2cache(spa);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	/*
	* We can't increment a feature while holding spa_vdev so we
	* have to do it in a synctask.
	*/
	if (ndraid != 0) {
	dmu_tx_t *tx;

	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	dsl_sync_task_nowait(spa->spa_dsl_pool, spa_draid_feature_incr,
	(void *)(uintptr_t)ndraid, tx);
	dmu_tx_commit(tx);
	}

	/*
	* We have to be careful when adding new vdevs to an existing pool.
	* If other threads start allocating from these vdevs before we
	* sync the config cache, and we lose power, then upon reboot we may
	* fail to open the pool because there are DVAs that the config cache
	* can't translate. Therefore, we first add the vdevs without
	* initializing metaslabs; sync the config cache (via spa_vdev_exit());
	* and then let spa_config_update() initialize the new metaslabs.
	*
	* spa_load() checks for added-but-not-initialized vdevs, so that
	* if we lose power at any point in this sequence, the remaining
	* steps will be completed the next time we load the pool.
	*/
	(void) spa_vdev_exit(spa, vd, txg, 0);

	mutex_enter(&spa_namespace_lock);
	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD);
	mutex_exit(&spa_namespace_lock);

	return (0);
	}

	/*
	* Attach a device to a mirror. The arguments are the path to any device
	* in the mirror, and the nvroot for the new device. If the path specifies
	* a device that is not mirrored, we automatically insert the mirror vdev.
	*
	* If 'replacing' is specified, the new device is intended to replace the
	* existing device; in this case the two devices are made into their own
	* mirror using the 'replacing' vdev, which is functionally identical to
	* the mirror vdev (it actually reuses all the same ops) but has a few
	* extra rules: you can't attach to it after it's been created, and upon
	* completion of resilvering, the first disk (the one being replaced)
	* is automatically detached.
	*
	* If 'rebuild' is specified, then sequential reconstruction (a.ka. rebuild)
	* should be performed instead of traditional healing reconstruction. From
	* an administrators perspective these are both resilver operations.
	*/
	int
	spa_vdev_attach(spa_t spa, uint64_t guid, nvlist_t nvroot, int replacing,
	int rebuild)
	{
	uint64_t txg, dtl_max_txg;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t oldvd, newvd, newrootvd, pvd, *tvd;
	vdev_ops_t *pvops;
	char oldvdpath, newvdpath;
	int newvd_isspare;
	int error;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	oldvd = spa_lookup_by_guid(spa, guid, B_FALSE);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	if (rebuild) {
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD))
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	if (dsl_scan_resilvering(spa_get_dsl(spa)))
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_RESILVER_IN_PROGRESS));
	} else {
	if (vdev_rebuild_active(rvd))
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_REBUILD_IN_PROGRESS));
	}

	if (spa->spa_vdev_removal != NULL)
	return (spa_vdev_exit(spa, NULL, txg, EBUSY));

	if (oldvd == NULL)
	return (spa_vdev_exit(spa, NULL, txg, ENODEV));

	if (!oldvd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	pvd = oldvd->vdev_parent;

	if ((error = spa_config_parse(spa, &newrootvd, nvroot, NULL, 0,
	VDEV_ALLOC_ATTACH)) != 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	if (newrootvd->vdev_children != 1)
	return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));

	newvd = newrootvd->vdev_child[0];

	if (!newvd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));

	if ((error = vdev_create(newrootvd, txg, replacing)) != 0)
	return (spa_vdev_exit(spa, newrootvd, txg, error));

	/*
	* Spares can't replace logs
	*/
	if (oldvd->vdev_top->vdev_islog && newvd->vdev_isspare)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* A dRAID spare can only replace a child of its parent dRAID vdev.
	*/
	if (newvd->vdev_ops == &vdev_draid_spare_ops &&
	oldvd->vdev_top != vdev_draid_spare_get_parent(newvd)) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}

	if (rebuild) {
	/*
	* For rebuilds, the top vdev must support reconstruction
	* using only space maps. This means the only allowable
	* vdevs types are the root vdev, a mirror, or dRAID.
	*/
	tvd = pvd;
	if (pvd->vdev_top != NULL)
	tvd = pvd->vdev_top;

	if (tvd->vdev_ops != &vdev_mirror_ops &&
	tvd->vdev_ops != &vdev_root_ops &&
	tvd->vdev_ops != &vdev_draid_ops) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}
	}

	if (!replacing) {
	/*
	* For attach, the only allowable parent is a mirror or the root
	* vdev.
	*/
	if (pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_root_ops)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	pvops = &vdev_mirror_ops;
	} else {
	/*
	* Active hot spares can only be replaced by inactive hot
	* spares.
	*/
	if (pvd->vdev_ops == &vdev_spare_ops &&
	oldvd->vdev_isspare &&
	!spa_has_spare(spa, newvd->vdev_guid))
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* If the source is a hot spare, and the parent isn't already a
	* spare, then we want to create a new hot spare. Otherwise, we
	* want to create a replacing vdev. The user is not allowed to
	* attach to a spared vdev child unless the 'isspare' state is
	* the same (spare replaces spare, non-spare replaces
	* non-spare).
	*/
	if (pvd->vdev_ops == &vdev_replacing_ops &&
	spa_version(spa) < SPA_VERSION_MULTI_REPLACE) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	} else if (pvd->vdev_ops == &vdev_spare_ops &&
	newvd->vdev_isspare != oldvd->vdev_isspare) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}

	if (newvd->vdev_isspare)
	pvops = &vdev_spare_ops;
	else
	pvops = &vdev_replacing_ops;
	}

	/*
	* Make sure the new device is big enough.
	*/
	if (newvd->vdev_asize < vdev_get_min_asize(oldvd))
	return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW));

	/*
	* The new device cannot have a higher alignment requirement
	* than the top-level vdev.
	*/
	if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* If this is an in-place replacement, update oldvd's path and devid
	* to make it distinguishable from newvd, and unopenable from now on.
	*/
	if (strcmp(oldvd->vdev_path, newvd->vdev_path) == 0) {
	spa_strfree(oldvd->vdev_path);
	oldvd->vdev_path = kmem_alloc(strlen(newvd->vdev_path) + 5,
	KM_SLEEP);
	(void) snprintf(oldvd->vdev_path, strlen(newvd->vdev_path) + 5,
	"%s/%s", newvd->vdev_path, "old");
	if (oldvd->vdev_devid != NULL) {
	spa_strfree(oldvd->vdev_devid);
	oldvd->vdev_devid = NULL;
	}
	}

	/*
	* If the parent is not a mirror, or if we're replacing, insert the new
	* mirror/replacing/spare vdev above oldvd.
	*/
	if (pvd->vdev_ops != pvops)
	pvd = vdev_add_parent(oldvd, pvops);

	ASSERT(pvd->vdev_top->vdev_parent == rvd);
	ASSERT(pvd->vdev_ops == pvops);
	ASSERT(oldvd->vdev_parent == pvd);

	/*
	* Extract the new device from its root and add it to pvd.
	*/
	vdev_remove_child(newrootvd, newvd);
	newvd->vdev_id = pvd->vdev_children;
	newvd->vdev_crtxg = oldvd->vdev_crtxg;
	vdev_add_child(pvd, newvd);

	/*
	* Reevaluate the parent vdev state.
	*/
	vdev_propagate_state(pvd);

	tvd = newvd->vdev_top;
	ASSERT(pvd->vdev_top == tvd);
	ASSERT(tvd->vdev_parent == rvd);

	vdev_config_dirty(tvd);

	/*
	* Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account
	* for any dmu_sync-ed blocks. It will propagate upward when
	* spa_vdev_exit() calls vdev_dtl_reassess().
	*/
	dtl_max_txg = txg + TXG_CONCURRENT_STATES;

	vdev_dtl_dirty(newvd, DTL_MISSING,
	TXG_INITIAL, dtl_max_txg - TXG_INITIAL);

	if (newvd->vdev_isspare) {
	spa_spare_activate(newvd);
	spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE);
	}

	oldvdpath = spa_strdup(oldvd->vdev_path);
	newvdpath = spa_strdup(newvd->vdev_path);
	newvd_isspare = newvd->vdev_isspare;

	/*
	* Mark newvd's DTL dirty in this txg.
	*/
	vdev_dirty(tvd, VDD_DTL, newvd, txg);

	/*
	* Schedule the resilver or rebuild to restart in the future. We do
	* this to ensure that dmu_sync-ed blocks have been stitched into the
	* respective datasets.
	*/
	if (rebuild) {
	newvd->vdev_rebuild_txg = txg;

	vdev_rebuild(tvd);
	} else {
	newvd->vdev_resilver_txg = txg;

	if (dsl_scan_resilvering(spa_get_dsl(spa)) &&
	spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER)) {
	vdev_defer_resilver(newvd);
	} else {
	dsl_scan_restart_resilver(spa->spa_dsl_pool,
	dtl_max_txg);
	}
	}

	if (spa->spa_bootfs)
	spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH);

	spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH);

	/*
	* Commit the config
	*/
	(void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);

	spa_history_log_internal(spa, "vdev attach", NULL,
	"%s vdev=%s %s vdev=%s",
	replacing && newvd_isspare ? "spare in" :
	replacing ? "replace" : "attach", newvdpath,
	replacing ? "for" : "to", oldvdpath);

	spa_strfree(oldvdpath);
	spa_strfree(newvdpath);

	return (0);
	}

	/*
	* Detach a device from a mirror or replacing vdev.
	*
	* If 'replace_done' is specified, only detach if the parent
	* is a replacing vdev.
	*/
	int
	spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done)
	{
	uint64_t txg;
	int error;
	vdev_t *rvd __maybe_unused = spa->spa_root_vdev;
	vdev_t vd, pvd, cvd, tvd;
	boolean_t unspare = B_FALSE;
	uint64_t unspare_guid = 0;
	char *vdpath;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_detach_enter(spa, guid);

	vd = spa_lookup_by_guid(spa, guid, B_FALSE);

	/*
	* Besides being called directly from the userland through the
	* ioctl interface, spa_vdev_detach() can be potentially called
	* at the end of spa_vdev_resilver_done().
	*
	* In the regular case, when we have a checkpoint this shouldn't
	* happen as we never empty the DTLs of a vdev during the scrub
	* [see comment in dsl_scan_done()]. Thus spa_vdev_resilvering_done()
	* should never get here when we have a checkpoint.
	*
	* That said, even in a case when we checkpoint the pool exactly
	* as spa_vdev_resilver_done() calls this function everything
	* should be fine as the resilver will return right away.
	*/
	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	if (vd == NULL)
	return (spa_vdev_exit(spa, NULL, txg, ENODEV));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	pvd = vd->vdev_parent;

	/*
	* If the parent/child relationship is not as expected, don't do it.
	* Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing
	* vdev that's replacing B with C. The user's intent in replacing
	* is to go from M(A,B) to M(A,C). If the user decides to cancel
	* the replace by detaching C, the expected behavior is to end up
	* M(A,B). But suppose that right after deciding to detach C,
	* the replacement of B completes. We would have M(A,C), and then
	* ask to detach C, which would leave us with just A -- not what
	* the user wanted. To prevent this, we make sure that the
	* parent/child relationship hasn't changed -- in this example,
	* that C's parent is still the replacing vdev R.
	*/
	if (pvd->vdev_guid != pguid && pguid != 0)
	return (spa_vdev_exit(spa, NULL, txg, EBUSY));

	/*
	* Only 'replacing' or 'spare' vdevs can be replaced.
	*/
	if (replace_done && pvd->vdev_ops != &vdev_replacing_ops &&
	pvd->vdev_ops != &vdev_spare_ops)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	ASSERT(pvd->vdev_ops != &vdev_spare_ops \|\|
	spa_version(spa) >= SPA_VERSION_SPARES);

	/*
	* Only mirror, replacing, and spare vdevs support detach.
	*/
	if (pvd->vdev_ops != &vdev_replacing_ops &&
	pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_spare_ops)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	/*
	* If this device has the only valid copy of some data,
	* we cannot safely detach it.
	*/
	if (vdev_dtl_required(vd))
	return (spa_vdev_exit(spa, NULL, txg, EBUSY));

	ASSERT(pvd->vdev_children >= 2);

	/*
	* If we are detaching the second disk from a replacing vdev, then
	* check to see if we changed the original vdev's path to have "/old"
	* at the end in spa_vdev_attach(). If so, undo that change now.
	*/
	if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 &&
	vd->vdev_path != NULL) {
	size_t len = strlen(vd->vdev_path);

	for (int c = 0; c < pvd->vdev_children; c++) {
	cvd = pvd->vdev_child[c];

	if (cvd == vd \|\| cvd->vdev_path == NULL)
	continue;

	if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
	strcmp(cvd->vdev_path + len, "/old") == 0) {
	spa_strfree(cvd->vdev_path);
	cvd->vdev_path = spa_strdup(vd->vdev_path);
	break;
	}
	}
	}

	/*
	* If we are detaching the original disk from a normal spare, then it
	* implies that the spare should become a real disk, and be removed
	* from the active spare list for the pool. dRAID spares on the
	* other hand are coupled to the pool and thus should never be removed
	* from the spares list.
	*/
	if (pvd->vdev_ops == &vdev_spare_ops && vd->vdev_id == 0) {
	vdev_t *last_cvd = pvd->vdev_child[pvd->vdev_children - 1];

	if (last_cvd->vdev_isspare &&
	last_cvd->vdev_ops != &vdev_draid_spare_ops) {
	unspare = B_TRUE;
	}
	}

	/*
	* Erase the disk labels so the disk can be used for other things.
	* This must be done after all other error cases are handled,
	* but before we disembowel vd (so we can still do I/O to it).
	* But if we can't do it, don't treat the error as fatal --
	* it may be that the unwritability of the disk is the reason
	* it's being detached!
	*/
	error = vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);

	/*
	* Remove vd from its parent and compact the parent's children.
	*/
	vdev_remove_child(pvd, vd);
	vdev_compact_children(pvd);

	/*
	* Remember one of the remaining children so we can get tvd below.
	*/
	cvd = pvd->vdev_child[pvd->vdev_children - 1];

	/*
	* If we need to remove the remaining child from the list of hot spares,
	* do it now, marking the vdev as no longer a spare in the process.
	* We must do this before vdev_remove_parent(), because that can
	* change the GUID if it creates a new toplevel GUID. For a similar
	* reason, we must remove the spare now, in the same txg as the detach;
	* otherwise someone could attach a new sibling, change the GUID, and
	* the subsequent attempt to spa_vdev_remove(unspare_guid) would fail.
	*/
	if (unspare) {
	ASSERT(cvd->vdev_isspare);
	spa_spare_remove(cvd);
	unspare_guid = cvd->vdev_guid;
	(void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
	cvd->vdev_unspare = B_TRUE;
	}

	/*
	* If the parent mirror/replacing vdev only has one child,
	* the parent is no longer needed. Remove it from the tree.
	*/
	if (pvd->vdev_children == 1) {
	if (pvd->vdev_ops == &vdev_spare_ops)
	cvd->vdev_unspare = B_FALSE;
	vdev_remove_parent(cvd);
	}

	/*
	* We don't set tvd until now because the parent we just removed
	* may have been the previous top-level vdev.
	*/
	tvd = cvd->vdev_top;
	ASSERT(tvd->vdev_parent == rvd);

	/*
	* Reevaluate the parent vdev state.
	*/
	vdev_propagate_state(cvd);

	/*
	* If the 'autoexpand' property is set on the pool then automatically
	* try to expand the size of the pool. For example if the device we
	* just detached was smaller than the others, it may be possible to
	* add metaslabs (i.e. grow the pool). We need to reopen the vdev
	* first so that we can obtain the updated sizes of the leaf vdevs.
	*/
	if (spa->spa_autoexpand) {
	vdev_reopen(tvd);
	vdev_expand(tvd, txg);
	}

	vdev_config_dirty(tvd);

	/*
	* Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that
	* vd->vdev_detached is set and free vd's DTL object in syncing context.
	* But first make sure we're not on any other txg's DTL list, to
	* prevent vd from being accessed after it's freed.
	*/
	vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none");
	for (int t = 0; t < TXG_SIZE; t++)
	(void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
	vd->vdev_detached = B_TRUE;
	vdev_dirty(tvd, VDD_DTL, vd, txg);

	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE);
	spa_notify_waiters(spa);

	/* hang on to the spa before we release the lock */
	spa_open_ref(spa, FTAG);

	error = spa_vdev_exit(spa, vd, txg, 0);

	spa_history_log_internal(spa, "detach", NULL,
	"vdev=%s", vdpath);
	spa_strfree(vdpath);

	/*
	* If this was the removal of the original device in a hot spare vdev,
	* then we want to go through and remove the device from the hot spare
	* list of every other pool.
	*/
	if (unspare) {
	spa_t *altspa = NULL;

	mutex_enter(&spa_namespace_lock);
	while ((altspa = spa_next(altspa)) != NULL) {
	if (altspa->spa_state != POOL_STATE_ACTIVE \|\|
	altspa == spa)
	continue;

	spa_open_ref(altspa, FTAG);
	mutex_exit(&spa_namespace_lock);
	(void) spa_vdev_remove(altspa, unspare_guid, B_TRUE);
	mutex_enter(&spa_namespace_lock);
	spa_close(altspa, FTAG);
	}
	mutex_exit(&spa_namespace_lock);

	/* search the rest of the vdevs for spares to remove */
	spa_vdev_resilver_done(spa);
	}

	/* all done with the spa; OK to release */
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);
	mutex_exit(&spa_namespace_lock);

	return (error);
	}

	static int
	spa_vdev_initialize_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
	list_t *vd_list)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);

	/* Look up vdev and ensure it's a leaf. */
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_detached) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(ENODEV));
	} else if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EINVAL));
	} else if (!vdev_writeable(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EROFS));
	}
	mutex_enter(&vd->vdev_initialize_lock);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);

	/*
	* When we activate an initialize action we check to see
	* if the vdev_initialize_thread is NULL. We do this instead
	* of using the vdev_initialize_state since there might be
	* a previous initialization process which has completed but
	* the thread is not exited.
	*/
	if (cmd_type == POOL_INITIALIZE_START &&
	(vd->vdev_initialize_thread != NULL \|\|
	vd->vdev_top->vdev_removing)) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(EBUSY));
	} else if (cmd_type == POOL_INITIALIZE_CANCEL &&
	(vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE &&
	vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED)) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(ESRCH));
	} else if (cmd_type == POOL_INITIALIZE_SUSPEND &&
	vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(ESRCH));
	}

	switch (cmd_type) {
	case POOL_INITIALIZE_START:
	vdev_initialize(vd);
	break;
	case POOL_INITIALIZE_CANCEL:
	vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, vd_list);
	break;
	case POOL_INITIALIZE_SUSPEND:
	vdev_initialize_stop(vd, VDEV_INITIALIZE_SUSPENDED, vd_list);
	break;
	default:
	panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
	}
	mutex_exit(&vd->vdev_initialize_lock);

	return (0);
	}

	int
	spa_vdev_initialize(spa_t spa, nvlist_t nv, uint64_t cmd_type,
	nvlist_t *vdev_errlist)
	{
	int total_errors = 0;
	list_t vd_list;

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_initialize_node));

	/*
	* We hold the namespace lock through the whole function
	* to prevent any changes to the pool while we're starting or
	* stopping initialization. The config and state locks are held so that
	* we can properly assess the vdev state before we commit to
	* the initializing operation.
	*/
	mutex_enter(&spa_namespace_lock);

	for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
	uint64_t vdev_guid = fnvpair_value_uint64(pair);

	int error = spa_vdev_initialize_impl(spa, vdev_guid, cmd_type,
	&vd_list);
	if (error != 0) {
	char guid_as_str[MAXNAMELEN];

	(void) snprintf(guid_as_str, sizeof (guid_as_str),
	"%llu", (unsigned long long)vdev_guid);
	fnvlist_add_int64(vdev_errlist, guid_as_str, error);
	total_errors++;
	}
	}

	/* Wait for all initialize threads to stop. */
	vdev_initialize_stop_wait(spa, &vd_list);

	/* Sync out the initializing state */
	txg_wait_synced(spa->spa_dsl_pool, 0);
	mutex_exit(&spa_namespace_lock);

	list_destroy(&vd_list);

	return (total_errors);
	}

	static int
	spa_vdev_trim_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
	uint64_t rate, boolean_t partial, boolean_t secure, list_t *vd_list)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);

	/* Look up vdev and ensure it's a leaf. */
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_detached) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(ENODEV));
	} else if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EINVAL));
	} else if (!vdev_writeable(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EROFS));
	} else if (!vd->vdev_has_trim) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	} else if (secure && !vd->vdev_has_securetrim) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}
	mutex_enter(&vd->vdev_trim_lock);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);

	/*
	* When we activate a TRIM action we check to see if the
	* vdev_trim_thread is NULL. We do this instead of using the
	* vdev_trim_state since there might be a previous TRIM process
	* which has completed but the thread is not exited.
	*/
	if (cmd_type == POOL_TRIM_START &&
	(vd->vdev_trim_thread != NULL \|\| vd->vdev_top->vdev_removing)) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(EBUSY));
	} else if (cmd_type == POOL_TRIM_CANCEL &&
	(vd->vdev_trim_state != VDEV_TRIM_ACTIVE &&
	vd->vdev_trim_state != VDEV_TRIM_SUSPENDED)) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(ESRCH));
	} else if (cmd_type == POOL_TRIM_SUSPEND &&
	vd->vdev_trim_state != VDEV_TRIM_ACTIVE) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(ESRCH));
	}

	switch (cmd_type) {
	case POOL_TRIM_START:
	vdev_trim(vd, rate, partial, secure);
	break;
	case POOL_TRIM_CANCEL:
	vdev_trim_stop(vd, VDEV_TRIM_CANCELED, vd_list);
	break;
	case POOL_TRIM_SUSPEND:
	vdev_trim_stop(vd, VDEV_TRIM_SUSPENDED, vd_list);
	break;
	default:
	panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
	}
	mutex_exit(&vd->vdev_trim_lock);

	return (0);
	}

	/*
	* Initiates a manual TRIM for the requested vdevs. This kicks off individual
	* TRIM threads for each child vdev. These threads pass over all of the free
	* space in the vdev's metaslabs and issues TRIM commands for that space.
	*/
	int
	spa_vdev_trim(spa_t spa, nvlist_t nv, uint64_t cmd_type, uint64_t rate,
	boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist)
	{
	int total_errors = 0;
	list_t vd_list;

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_trim_node));

	/*
	* We hold the namespace lock through the whole function
	* to prevent any changes to the pool while we're starting or
	* stopping TRIM. The config and state locks are held so that
	* we can properly assess the vdev state before we commit to
	* the TRIM operation.
	*/
	mutex_enter(&spa_namespace_lock);

	for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
	uint64_t vdev_guid = fnvpair_value_uint64(pair);

	int error = spa_vdev_trim_impl(spa, vdev_guid, cmd_type,
	rate, partial, secure, &vd_list);
	if (error != 0) {
	char guid_as_str[MAXNAMELEN];

	(void) snprintf(guid_as_str, sizeof (guid_as_str),
	"%llu", (unsigned long long)vdev_guid);
	fnvlist_add_int64(vdev_errlist, guid_as_str, error);
	total_errors++;
	}
	}

	/* Wait for all TRIM threads to stop. */
	vdev_trim_stop_wait(spa, &vd_list);

	/* Sync out the TRIM state */
	txg_wait_synced(spa->spa_dsl_pool, 0);
	mutex_exit(&spa_namespace_lock);

	list_destroy(&vd_list);

	return (total_errors);
	}

	/*
	* Split a set of devices from their mirrors, and create a new pool from them.
	*/
	int
	spa_vdev_split_mirror(spa_t spa, const char newname, nvlist_t *config,
	nvlist_t *props, boolean_t exp)
	{
	int error = 0;
	uint64_t txg, *glist;
	spa_t *newspa;
	uint_t c, children, lastlog;
	nvlist_t *child, nvl, *tmp;
	dmu_tx_t *tx;
	char *altroot = NULL;
	vdev_t rvd, vml = NULL; / vdev modify list */
	boolean_t activate_slog;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	/* clear the log and flush everything up to now */
	activate_slog = spa_passivate_log(spa);
	(void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
	error = spa_reset_logs(spa);
	txg = spa_vdev_config_enter(spa);

	if (activate_slog)
	spa_activate_log(spa);

	if (error != 0)
	return (spa_vdev_exit(spa, NULL, txg, error));

	/* check new spa name before going any further */
	if (spa_lookup(newname) != NULL)
	return (spa_vdev_exit(spa, NULL, txg, EEXIST));

	/*
	* scan through all the children to ensure they're all mirrors
	*/
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 \|\|
	nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child,
	&children) != 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	/* first, check to ensure we've got the right child count */
	rvd = spa->spa_root_vdev;
	lastlog = 0;
	for (c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	/* don't count the holes & logs as children */
	if (vd->vdev_islog \|\| (vd->vdev_ops != &vdev_indirect_ops &&
	!vdev_is_concrete(vd))) {
	if (lastlog == 0)
	lastlog = c;
	continue;
	}

	lastlog = 0;
	}
	if (children != (lastlog != 0 ? lastlog : rvd->vdev_children))
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	/* next, ensure no spare or cache devices are part of the split */
	if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 \|\|
	nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP);
	glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP);

	/* then, loop over each vdev and validate it */
	for (c = 0; c < children; c++) {
	uint64_t is_hole = 0;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&is_hole);

	if (is_hole != 0) {
	if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole \|\|
	spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
	continue;
	} else {
	error = SET_ERROR(EINVAL);
	break;
	}
	}

	/* deal with indirect vdevs */
	if (spa->spa_root_vdev->vdev_child[c]->vdev_ops ==
	&vdev_indirect_ops)
	continue;

	/* which disk is going to be split? */
	if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
	&glist[c]) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	/* look it up in the spa */
	vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE);
	if (vml[c] == NULL) {
	error = SET_ERROR(ENODEV);
	break;
	}

	/* make sure there's nothing stopping the split */
	if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops \|\|
	vml[c]->vdev_islog \|\|
	!vdev_is_concrete(vml[c]) \|\|
	vml[c]->vdev_isspare \|\|
	vml[c]->vdev_isl2cache \|\|
	!vdev_writeable(vml[c]) \|\|
	vml[c]->vdev_children != 0 \|\|
	vml[c]->vdev_state != VDEV_STATE_HEALTHY \|\|
	c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (vdev_dtl_required(vml[c]) \|\|
	vdev_resilver_needed(vml[c], NULL, NULL)) {
	error = SET_ERROR(EBUSY);
	break;
	}

	/* we need certain info from the top level */
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY,
	vml[c]->vdev_top->vdev_ms_array);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT,
	vml[c]->vdev_top->vdev_ms_shift);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE,
	vml[c]->vdev_top->vdev_asize);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
	vml[c]->vdev_top->vdev_ashift);

	/* transfer per-vdev ZAPs */
	ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0);
	VERIFY0(nvlist_add_uint64(child[c],
	ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap));

	ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0);
	VERIFY0(nvlist_add_uint64(child[c],
	ZPOOL_CONFIG_VDEV_TOP_ZAP,
	vml[c]->vdev_parent->vdev_top_zap));
	}

	if (error != 0) {
	kmem_free(vml, children * sizeof (vdev_t *));
	kmem_free(glist, children * sizeof (uint64_t));
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	/* stop writers from using the disks */
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL)
	vml[c]->vdev_offline = B_TRUE;
	}
	vdev_reopen(spa->spa_root_vdev);

	/*
	* Temporarily record the splitting vdevs in the spa config. This
	* will disappear once the config is regenerated.
	*/
	nvl = fnvlist_alloc();
	fnvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, glist, children);
	kmem_free(glist, children * sizeof (uint64_t));

	mutex_enter(&spa->spa_props_lock);
	fnvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, nvl);
	mutex_exit(&spa->spa_props_lock);
	spa->spa_config_splitting = nvl;
	vdev_config_dirty(spa->spa_root_vdev);

	/* configure and create the new pool */
	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, spa->spa_config_txg);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	spa_generate_guid(NULL));
	VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);

	/* add the new pool to the namespace */
	newspa = spa_add(newname, config, altroot);
	newspa->spa_avz_action = AVZ_ACTION_REBUILD;
	newspa->spa_config_txg = spa->spa_config_txg;
	spa_set_log_state(newspa, SPA_LOG_CLEAR);

	/* release the spa config lock, retaining the namespace lock */
	spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 1);

	spa_activate(newspa, spa_mode_global);
	spa_async_suspend(newspa);

	/*
	* Temporarily stop the initializing and TRIM activity. We set the
	* state to ACTIVE so that we know to resume initializing or TRIM
	* once the split has completed.
	*/
	list_t vd_initialize_list;
	list_create(&vd_initialize_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_initialize_node));

	list_t vd_trim_list;
	list_create(&vd_trim_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_trim_node));

	for (c = 0; c < children; c++) {
	if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
	mutex_enter(&vml[c]->vdev_initialize_lock);
	vdev_initialize_stop(vml[c],
	VDEV_INITIALIZE_ACTIVE, &vd_initialize_list);
	mutex_exit(&vml[c]->vdev_initialize_lock);

	mutex_enter(&vml[c]->vdev_trim_lock);
	vdev_trim_stop(vml[c], VDEV_TRIM_ACTIVE, &vd_trim_list);
	mutex_exit(&vml[c]->vdev_trim_lock);
	}
	}

	vdev_initialize_stop_wait(spa, &vd_initialize_list);
	vdev_trim_stop_wait(spa, &vd_trim_list);

	list_destroy(&vd_initialize_list);
	list_destroy(&vd_trim_list);

	newspa->spa_config_source = SPA_CONFIG_SRC_SPLIT;
	newspa->spa_is_splitting = B_TRUE;

	/* create the new pool from the disks of the original pool */
	error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE);
	if (error)
	goto out;

	/* if that worked, generate a real config for the new pool */
	if (newspa->spa_root_vdev != NULL) {
	newspa->spa_config_splitting = fnvlist_alloc();
	fnvlist_add_uint64(newspa->spa_config_splitting,
	ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa));
	spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL,
	B_TRUE));
	}

	/* set the props */
	if (props != NULL) {
	spa_configfile_set(newspa, props, B_FALSE);
	error = spa_prop_set(newspa, props);
	if (error)
	goto out;
	}

	/* flush everything */
	txg = spa_vdev_config_enter(newspa);
	vdev_config_dirty(newspa->spa_root_vdev);
	(void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 2);

	spa_async_resume(newspa);

	/* finally, update the original pool's config */
	txg = spa_vdev_config_enter(spa);
	tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0)
	dmu_tx_abort(tx);
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
	vdev_t *tvd = vml[c]->vdev_top;

	/*
	* Need to be sure the detachable VDEV is not
	* on any other txg's DTL list to prevent it
	* from being accessed after it's freed.
	*/
	for (int t = 0; t < TXG_SIZE; t++) {
	(void) txg_list_remove_this(
	&tvd->vdev_dtl_list, vml[c], t);
	}

	vdev_split(vml[c]);
	if (error == 0)
	spa_history_log_internal(spa, "detach", tx,
	"vdev=%s", vml[c]->vdev_path);

	vdev_free(vml[c]);
	}
	}
	spa->spa_avz_action = AVZ_ACTION_REBUILD;
	vdev_config_dirty(spa->spa_root_vdev);
	spa->spa_config_splitting = NULL;
	nvlist_free(nvl);
	if (error == 0)
	dmu_tx_commit(tx);
	(void) spa_vdev_exit(spa, NULL, txg, 0);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 3);

	/* split is complete; log a history record */
	spa_history_log_internal(newspa, "split", NULL,
	"from pool %s", spa_name(spa));

	newspa->spa_is_splitting = B_FALSE;
	kmem_free(vml, children * sizeof (vdev_t *));

	/* if we're not going to mount the filesystems in userland, export */
	if (exp)
	error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL,
	B_FALSE, B_FALSE);

	return (error);

	out:
	spa_unload(newspa);
	spa_deactivate(newspa);
	spa_remove(newspa);

	txg = spa_vdev_config_enter(spa);

	/* re-online all offlined disks */
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL)
	vml[c]->vdev_offline = B_FALSE;
	}

	/* restart initializing or trimming disks as necessary */
	spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
	spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
	spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);

	vdev_reopen(spa->spa_root_vdev);

	nvlist_free(spa->spa_config_splitting);
	spa->spa_config_splitting = NULL;
	(void) spa_vdev_exit(spa, NULL, txg, error);

	kmem_free(vml, children * sizeof (vdev_t *));
	return (error);
	}

	/*
	* Find any device that's done replacing, or a vdev marked 'unspare' that's
	* currently spared, so we can detach it.
	*/
	static vdev_t *
	spa_vdev_resilver_done_hunt(vdev_t *vd)
	{
	vdev_t newvd, oldvd;

	for (int c = 0; c < vd->vdev_children; c++) {
	oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]);
	if (oldvd != NULL)
	return (oldvd);
	}

	/*
	* Check for a completed replacement. We always consider the first
	* vdev in the list to be the oldest vdev, and the last one to be
	* the newest (see spa_vdev_attach() for how that works). In
	* the case where the newest vdev is faulted, we will not automatically
	* remove it after a resilver completes. This is OK as it will require
	* user intervention to determine which disk the admin wishes to keep.
	*/
	if (vd->vdev_ops == &vdev_replacing_ops) {
	ASSERT(vd->vdev_children > 1);

	newvd = vd->vdev_child[vd->vdev_children - 1];
	oldvd = vd->vdev_child[0];

	if (vdev_dtl_empty(newvd, DTL_MISSING) &&
	vdev_dtl_empty(newvd, DTL_OUTAGE) &&
	!vdev_dtl_required(oldvd))
	return (oldvd);
	}

	/*
	* Check for a completed resilver with the 'unspare' flag set.
	* Also potentially update faulted state.
	*/
	if (vd->vdev_ops == &vdev_spare_ops) {
	vdev_t *first = vd->vdev_child[0];
	vdev_t *last = vd->vdev_child[vd->vdev_children - 1];

	if (last->vdev_unspare) {
	oldvd = first;
	newvd = last;
	} else if (first->vdev_unspare) {
	oldvd = last;
	newvd = first;
	} else {
	oldvd = NULL;
	}

	if (oldvd != NULL &&
	vdev_dtl_empty(newvd, DTL_MISSING) &&
	vdev_dtl_empty(newvd, DTL_OUTAGE) &&
	!vdev_dtl_required(oldvd))
	return (oldvd);

	vdev_propagate_state(vd);

	/*
	* If there are more than two spares attached to a disk,
	* and those spares are not required, then we want to
	* attempt to free them up now so that they can be used
	* by other pools. Once we're back down to a single
	* disk+spare, we stop removing them.
	*/
	if (vd->vdev_children > 2) {
	newvd = vd->vdev_child[1];

	if (newvd->vdev_isspare && last->vdev_isspare &&
	vdev_dtl_empty(last, DTL_MISSING) &&
	vdev_dtl_empty(last, DTL_OUTAGE) &&
	!vdev_dtl_required(newvd))
	return (newvd);
	}
	}

	return (NULL);
	}

	static void
	spa_vdev_resilver_done(spa_t *spa)
	{
	vdev_t vd, pvd, *ppvd;
	uint64_t guid, sguid, pguid, ppguid;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) {
	pvd = vd->vdev_parent;
	ppvd = pvd->vdev_parent;
	guid = vd->vdev_guid;
	pguid = pvd->vdev_guid;
	ppguid = ppvd->vdev_guid;
	sguid = 0;
	/*
	* If we have just finished replacing a hot spared device, then
	* we need to detach the parent's first child (the original hot
	* spare) as well.
	*/
	if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 &&
	ppvd->vdev_children == 2) {
	ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
	sguid = ppvd->vdev_child[1]->vdev_guid;
	}
	ASSERT(vd->vdev_resilver_txg == 0 \|\| !vdev_dtl_required(vd));

	spa_config_exit(spa, SCL_ALL, FTAG);
	if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0)
	return;
	if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0)
	return;
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	}

	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* If a detach was not performed above replace waiters will not have
	* been notified. In which case we must do so now.
	*/
	spa_notify_waiters(spa);
	}

	/*
	* Update the stored path or FRU for this vdev.
	*/
	static int
	spa_vdev_set_common(spa_t spa, uint64_t guid, const char value,
	boolean_t ispath)
	{
	vdev_t *vd;
	boolean_t sync = B_FALSE;

	ASSERT(spa_writeable(spa));

	spa_vdev_state_enter(spa, SCL_ALL);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, ENOENT));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, ENOTSUP));

	if (ispath) {
	if (strcmp(value, vd->vdev_path) != 0) {
	spa_strfree(vd->vdev_path);
	vd->vdev_path = spa_strdup(value);
	sync = B_TRUE;
	}
	} else {
	if (vd->vdev_fru == NULL) {
	vd->vdev_fru = spa_strdup(value);
	sync = B_TRUE;
	} else if (strcmp(value, vd->vdev_fru) != 0) {
	spa_strfree(vd->vdev_fru);
	vd->vdev_fru = spa_strdup(value);
	sync = B_TRUE;
	}
	}

	return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0));
	}

	int
	spa_vdev_setpath(spa_t spa, uint64_t guid, const char newpath)
	{
	return (spa_vdev_set_common(spa, guid, newpath, B_TRUE));
	}

	int
	spa_vdev_setfru(spa_t spa, uint64_t guid, const char newfru)
	{
	return (spa_vdev_set_common(spa, guid, newfru, B_FALSE));
	}

	/*
	* ==========================================================================
	* SPA Scanning
	* ==========================================================================
	*/
	int
	spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);

	if (dsl_scan_resilvering(spa->spa_dsl_pool))
	return (SET_ERROR(EBUSY));

	return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd));
	}

	int
	spa_scan_stop(spa_t *spa)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
	if (dsl_scan_resilvering(spa->spa_dsl_pool))
	return (SET_ERROR(EBUSY));
	return (dsl_scan_cancel(spa->spa_dsl_pool));
	}

	int
	spa_scan(spa_t *spa, pool_scan_func_t func)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);

	if (func >= POOL_SCAN_FUNCS \|\| func == POOL_SCAN_NONE)
	return (SET_ERROR(ENOTSUP));

	if (func == POOL_SCAN_RESILVER &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER))
	return (SET_ERROR(ENOTSUP));

	/*
	* If a resilver was requested, but there is no DTL on a
	* writeable leaf device, we have nothing to do.
	*/
	if (func == POOL_SCAN_RESILVER &&
	!vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
	spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
	return (0);
	}

	return (dsl_scan(spa->spa_dsl_pool, func));
	}

	/*
	* ==========================================================================
	* SPA async task processing
	* ==========================================================================
	*/

	static void
	spa_async_remove(spa_t spa, vdev_t vd)
	{
	if (vd->vdev_remove_wanted) {
	vd->vdev_remove_wanted = B_FALSE;
	vd->vdev_delayed_close = B_FALSE;
	vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE);

	/*
	* We want to clear the stats, but we don't want to do a full
	* vdev_clear() as that will cause us to throw away
	* degraded/faulted state as well as attempt to reopen the
	* device, all of which is a waste.
	*/
	vd->vdev_stat.vs_read_errors = 0;
	vd->vdev_stat.vs_write_errors = 0;
	vd->vdev_stat.vs_checksum_errors = 0;

	vdev_state_dirty(vd->vdev_top);

	/* Tell userspace that the vdev is gone. */
	zfs_post_remove(spa, vd);
	}

	for (int c = 0; c < vd->vdev_children; c++)
	spa_async_remove(spa, vd->vdev_child[c]);
	}

	static void
	spa_async_probe(spa_t spa, vdev_t vd)
	{
	if (vd->vdev_probe_wanted) {
	vd->vdev_probe_wanted = B_FALSE;
	vdev_reopen(vd); /* vdev_open() does the actual probe */
	}

	for (int c = 0; c < vd->vdev_children; c++)
	spa_async_probe(spa, vd->vdev_child[c]);
	}

	static void
	spa_async_autoexpand(spa_t spa, vdev_t vd)
	{
	if (!spa->spa_autoexpand)
	return;

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	spa_async_autoexpand(spa, cvd);
	}

	if (!vd->vdev_ops->vdev_op_leaf \|\| vd->vdev_physpath == NULL)
	return;

	spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
	}

	static __attribute__((noreturn)) void
	spa_async_thread(void *arg)
	{
	spa_t spa = (spa_t )arg;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	int tasks;

	ASSERT(spa->spa_sync_on);

	mutex_enter(&spa->spa_async_lock);
	tasks = spa->spa_async_tasks;
	spa->spa_async_tasks = 0;
	mutex_exit(&spa->spa_async_lock);

	/*
	* See if the config needs to be updated.
	*/
	if (tasks & SPA_ASYNC_CONFIG_UPDATE) {
	uint64_t old_space, new_space;

	mutex_enter(&spa_namespace_lock);
	old_space = metaslab_class_get_space(spa_normal_class(spa));
	old_space += metaslab_class_get_space(spa_special_class(spa));
	old_space += metaslab_class_get_space(spa_dedup_class(spa));
	old_space += metaslab_class_get_space(
	spa_embedded_log_class(spa));

	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);

	new_space = metaslab_class_get_space(spa_normal_class(spa));
	new_space += metaslab_class_get_space(spa_special_class(spa));
	new_space += metaslab_class_get_space(spa_dedup_class(spa));
	new_space += metaslab_class_get_space(
	spa_embedded_log_class(spa));
	mutex_exit(&spa_namespace_lock);

	/*
	* If the pool grew as a result of the config update,
	* then log an internal history event.
	*/
	if (new_space != old_space) {
	spa_history_log_internal(spa, "vdev online", NULL,
	"pool '%s' size: %llu(+%llu)",
	spa_name(spa), (u_longlong_t)new_space,
	(u_longlong_t)(new_space - old_space));
	}
	}

	/*
	* See if any devices need to be marked REMOVED.
	*/
	if (tasks & SPA_ASYNC_REMOVE) {
	spa_vdev_state_enter(spa, SCL_NONE);
	spa_async_remove(spa, spa->spa_root_vdev);
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
	spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]);
	for (int i = 0; i < spa->spa_spares.sav_count; i++)
	spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	}

	if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) {
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	spa_async_autoexpand(spa, spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	/*
	* See if any devices need to be probed.
	*/
	if (tasks & SPA_ASYNC_PROBE) {
	spa_vdev_state_enter(spa, SCL_NONE);
	spa_async_probe(spa, spa->spa_root_vdev);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	}

	/*
	* If any devices are done replacing, detach them.
	*/
	if (tasks & SPA_ASYNC_RESILVER_DONE \|\|
	tasks & SPA_ASYNC_REBUILD_DONE) {
	spa_vdev_resilver_done(spa);
	}

	/*
	* Kick off a resilver.
	*/
	if (tasks & SPA_ASYNC_RESILVER &&
	!vdev_rebuild_active(spa->spa_root_vdev) &&
	(!dsl_scan_resilvering(dp) \|\|
	!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER)))
	dsl_scan_restart_resilver(dp, 0);

	if (tasks & SPA_ASYNC_INITIALIZE_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_initialize_restart(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	if (tasks & SPA_ASYNC_TRIM_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_trim_restart(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	if (tasks & SPA_ASYNC_AUTOTRIM_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_autotrim_restart(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Kick off L2 cache whole device TRIM.
	*/
	if (tasks & SPA_ASYNC_L2CACHE_TRIM) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_trim_l2arc(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Kick off L2 cache rebuilding.
	*/
	if (tasks & SPA_ASYNC_L2CACHE_REBUILD) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_L2ARC, FTAG, RW_READER);
	l2arc_spa_rebuild_start(spa);
	spa_config_exit(spa, SCL_L2ARC, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Let the world know that we're done.
	*/
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_thread = NULL;
	cv_broadcast(&spa->spa_async_cv);
	mutex_exit(&spa->spa_async_lock);
	thread_exit();
	}

	void
	spa_async_suspend(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_suspended++;
	while (spa->spa_async_thread != NULL)
	cv_wait(&spa->spa_async_cv, &spa->spa_async_lock);
	mutex_exit(&spa->spa_async_lock);

	spa_vdev_remove_suspend(spa);

	zthr_t *condense_thread = spa->spa_condense_zthr;
	if (condense_thread != NULL)
	zthr_cancel(condense_thread);

	zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
	if (discard_thread != NULL)
	zthr_cancel(discard_thread);

	zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
	if (ll_delete_thread != NULL)
	zthr_cancel(ll_delete_thread);

	zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
	if (ll_condense_thread != NULL)
	zthr_cancel(ll_condense_thread);
	}

	void
	spa_async_resume(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	ASSERT(spa->spa_async_suspended != 0);
	spa->spa_async_suspended--;
	mutex_exit(&spa->spa_async_lock);
	spa_restart_removal(spa);

	zthr_t *condense_thread = spa->spa_condense_zthr;
	if (condense_thread != NULL)
	zthr_resume(condense_thread);

	zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
	if (discard_thread != NULL)
	zthr_resume(discard_thread);

	zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
	if (ll_delete_thread != NULL)
	zthr_resume(ll_delete_thread);

	zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
	if (ll_condense_thread != NULL)
	zthr_resume(ll_condense_thread);
	}

	static boolean_t
	spa_async_tasks_pending(spa_t *spa)
	{
	uint_t non_config_tasks;
	uint_t config_task;
	boolean_t config_task_suspended;

	non_config_tasks = spa->spa_async_tasks & ~SPA_ASYNC_CONFIG_UPDATE;
	config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE;
	if (spa->spa_ccw_fail_time == 0) {
	config_task_suspended = B_FALSE;
	} else {
	config_task_suspended =
	(gethrtime() - spa->spa_ccw_fail_time) <
	((hrtime_t)zfs_ccw_retry_interval * NANOSEC);
	}

	return (non_config_tasks \|\| (config_task && !config_task_suspended));
	}

	static void
	spa_async_dispatch(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	if (spa_async_tasks_pending(spa) &&
	!spa->spa_async_suspended &&
	spa->spa_async_thread == NULL)
	spa->spa_async_thread = thread_create(NULL, 0,
	spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri);
	mutex_exit(&spa->spa_async_lock);
	}

	void
	spa_async_request(spa_t *spa, int task)
	{
	zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task);
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_tasks \|= task;
	mutex_exit(&spa->spa_async_lock);
	}

	int
	spa_async_tasks(spa_t *spa)
	{
	return (spa->spa_async_tasks);
	}

	/*
	* ==========================================================================
	* SPA syncing routines
	* ==========================================================================
	*/


	static int
	bpobj_enqueue_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	bpobj_t *bpo = arg;
	bpobj_enqueue(bpo, bp, bp_freed, tx);
	return (0);
	}

	int
	bpobj_enqueue_alloc_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	return (bpobj_enqueue_cb(arg, bp, B_FALSE, tx));
	}

	int
	bpobj_enqueue_free_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	return (bpobj_enqueue_cb(arg, bp, B_TRUE, tx));
	}

	static int
	spa_free_sync_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	zio_t *pio = arg;

	zio_nowait(zio_free_sync(pio, pio->io_spa, dmu_tx_get_txg(tx), bp,
	pio->io_flags));
	return (0);
	}

	static int
	bpobj_spa_free_sync_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(!bp_freed);
	return (spa_free_sync_cb(arg, bp, tx));
	}

	/*
	* Note: this simple function is not inlined to make it easier to dtrace the
	* amount of time spent syncing frees.
	*/
	static void
	spa_sync_frees(spa_t spa, bplist_t bpl, dmu_tx_t *tx)
	{
	zio_t *zio = zio_root(spa, NULL, NULL, 0);
	bplist_iterate(bpl, spa_free_sync_cb, zio, tx);
	VERIFY(zio_wait(zio) == 0);
	}

	/*
	* Note: this simple function is not inlined to make it easier to dtrace the
	* amount of time spent syncing deferred frees.
	*/
	static void
	spa_sync_deferred_frees(spa_t spa, dmu_tx_t tx)
	{
	if (spa_sync_pass(spa) != 1)
	return;

	/*
	* Note:
	* If the log space map feature is active, we stop deferring
	* frees to the next TXG and therefore running this function
	* would be considered a no-op as spa_deferred_bpobj should
	* not have any entries.
	*
	* That said we run this function anyway (instead of returning
	* immediately) for the edge-case scenario where we just
	* activated the log space map feature in this TXG but we have
	* deferred frees from the previous TXG.
	*/
	zio_t *zio = zio_root(spa, NULL, NULL, 0);
	VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj,
	bpobj_spa_free_sync_cb, zio, tx), ==, 0);
	VERIFY0(zio_wait(zio));
	}

	static void
	spa_sync_nvlist(spa_t spa, uint64_t obj, nvlist_t nv, dmu_tx_t *tx)
	{
	char *packed = NULL;
	size_t bufsize;
	size_t nvsize = 0;
	dmu_buf_t *db;

	VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0);

	/*
	* Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
	* information. This avoids the dmu_buf_will_dirty() path and
	* saves us a pre-read to get data we don't actually care about.
	*/
	bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
	packed = vmem_alloc(bufsize, KM_SLEEP);

	VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
	KM_SLEEP) == 0);
	memset(packed + nvsize, 0, bufsize - nvsize);

	dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);

	vmem_free(packed, bufsize);

	VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
	dmu_buf_will_dirty(db, tx);
	(uint64_t )db->db_data = nvsize;
	dmu_buf_rele(db, FTAG);
	}

	static void
	spa_sync_aux_dev(spa_t spa, spa_aux_vdev_t sav, dmu_tx_t *tx,
	const char config, const char entry)
	{
	nvlist_t *nvroot;
	nvlist_t **list;
	int i;

	if (!sav->sav_sync)
	return;

	/*
	* Update the MOS nvlist describing the list of available devices.
	* spa_validate_aux() will have already made sure this nvlist is
	* valid and the vdevs are labeled appropriately.
	*/
	if (sav->sav_object == 0) {
	sav->sav_object = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE,
	sizeof (uint64_t), tx);
	VERIFY(zap_update(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1,
	&sav->sav_object, tx) == 0);
	}

	nvroot = fnvlist_alloc();
	if (sav->sav_count == 0) {
	fnvlist_add_nvlist_array(nvroot, config,
	(const nvlist_t * const *)NULL, 0);
	} else {
	list = kmem_alloc(sav->sav_countsizeof (void ), KM_SLEEP);
	for (i = 0; i < sav->sav_count; i++)
	list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
	B_FALSE, VDEV_CONFIG_L2CACHE);
	fnvlist_add_nvlist_array(nvroot, config,
	(const nvlist_t * const *)list, sav->sav_count);
	for (i = 0; i < sav->sav_count; i++)
	nvlist_free(list[i]);
	kmem_free(list, sav->sav_count * sizeof (void *));
	}

	spa_sync_nvlist(spa, sav->sav_object, nvroot, tx);
	nvlist_free(nvroot);

	sav->sav_sync = B_FALSE;
	}

	/*
	* Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t.
	* The all-vdev ZAP must be empty.
	*/
	static void
	spa_avz_build(vdev_t vd, uint64_t avz, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;

	if (vd->vdev_top_zap != 0) {
	VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
	vd->vdev_top_zap, tx));
	}
	if (vd->vdev_leaf_zap != 0) {
	VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
	vd->vdev_leaf_zap, tx));
	}
	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	spa_avz_build(vd->vdev_child[i], avz, tx);
	}
	}

	static void
	spa_sync_config_object(spa_t spa, dmu_tx_t tx)
	{
	nvlist_t *config;

	/*
	* If the pool is being imported from a pre-per-vdev-ZAP version of ZFS,
	* its config may not be dirty but we still need to build per-vdev ZAPs.
	* Similarly, if the pool is being assembled (e.g. after a split), we
	* need to rebuild the AVZ although the config may not be dirty.
	*/
	if (list_is_empty(&spa->spa_config_dirty_list) &&
	spa->spa_avz_action == AVZ_ACTION_NONE)
	return;

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE \|\|
	spa->spa_avz_action == AVZ_ACTION_INITIALIZE \|\|
	spa->spa_all_vdev_zaps != 0);

	if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
	/* Make and build the new AVZ */
	uint64_t new_avz = zap_create(spa->spa_meta_objset,
	DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
	spa_avz_build(spa->spa_root_vdev, new_avz, tx);

	/* Diff old AVZ with new one */
	zap_cursor_t zc;
	zap_attribute_t za;

	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_all_vdev_zaps);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t vdzap = za.za_first_integer;
	if (zap_lookup_int(spa->spa_meta_objset, new_avz,
	vdzap) == ENOENT) {
	/*
	* ZAP is listed in old AVZ but not in new one;
	* destroy it
	*/
	VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap,
	tx));
	}
	}

	zap_cursor_fini(&zc);

	/* Destroy the old AVZ */
	VERIFY0(zap_destroy(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, tx));

	/* Replace the old AVZ in the dir obj with the new one */
	VERIFY0(zap_update(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP,
	sizeof (new_avz), 1, &new_avz, tx));

	spa->spa_all_vdev_zaps = new_avz;
	} else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) {
	zap_cursor_t zc;
	zap_attribute_t za;

	/* Walk through the AVZ and destroy all listed ZAPs */
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_all_vdev_zaps);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t zap = za.za_first_integer;
	VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx));
	}

	zap_cursor_fini(&zc);

	/* Destroy and unlink the AVZ itself */
	VERIFY0(zap_destroy(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, tx));
	VERIFY0(zap_remove(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx));
	spa->spa_all_vdev_zaps = 0;
	}

	if (spa->spa_all_vdev_zaps == 0) {
	spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset,
	DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_VDEV_ZAP_MAP, tx);
	}
	spa->spa_avz_action = AVZ_ACTION_NONE;

	/* Create ZAPs for vdevs that don't have them. */
	vdev_construct_zaps(spa->spa_root_vdev, tx);

	config = spa_config_generate(spa, spa->spa_root_vdev,
	dmu_tx_get_txg(tx), B_FALSE);

	/*
	* If we're upgrading the spa version then make sure that
	* the config object gets updated with the correct version.
	*/
	if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version)
	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
	spa->spa_uberblock.ub_version);

	spa_config_exit(spa, SCL_STATE, FTAG);

	nvlist_free(spa->spa_config_syncing);
	spa->spa_config_syncing = config;

	spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
	}

	static void
	spa_sync_version(void arg, dmu_tx_t tx)
	{
	uint64_t *versionp = arg;
	uint64_t version = *versionp;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	/*
	* Setting the version is special cased when first creating the pool.
	*/
	ASSERT(tx->tx_txg != TXG_INITIAL);

	ASSERT(SPA_VERSION_IS_SUPPORTED(version));
	ASSERT(version >= spa_version(spa));

	spa->spa_uberblock.ub_version = version;
	vdev_config_dirty(spa->spa_root_vdev);
	spa_history_log_internal(spa, "set", tx, "version=%lld",
	(longlong_t)version);
	}

	/*
	* Set zpool properties.
	*/
	static void
	spa_sync_props(void arg, dmu_tx_t tx)
	{
	nvlist_t *nvp = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = spa->spa_meta_objset;
	nvpair_t *elem = NULL;

	mutex_enter(&spa->spa_props_lock);

	while ((elem = nvlist_next_nvpair(nvp, elem))) {
	uint64_t intval;
	char strval, fname;
	zpool_prop_t prop;
	const char *propname;
	zprop_type_t proptype;
	spa_feature_t fid;

	switch (prop = zpool_name_to_prop(nvpair_name(elem))) {
	case ZPOOL_PROP_INVAL:
	/*
	* We checked this earlier in spa_prop_validate().
	*/
	ASSERT(zpool_prop_feature(nvpair_name(elem)));

	fname = strchr(nvpair_name(elem), '@') + 1;
	VERIFY0(zfeature_lookup_name(fname, &fid));

	spa_feature_enable(spa, fid, tx);
	spa_history_log_internal(spa, "set", tx,
	"%s=enabled", nvpair_name(elem));
	break;

	case ZPOOL_PROP_VERSION:
	intval = fnvpair_value_uint64(elem);
	/*
	* The version is synced separately before other
	* properties and should be correct by now.
	*/
	ASSERT3U(spa_version(spa), >=, intval);
	break;

	case ZPOOL_PROP_ALTROOT:
	/*
	* 'altroot' is a non-persistent property. It should
	* have been set temporarily at creation or import time.
	*/
	ASSERT(spa->spa_root != NULL);
	break;

	case ZPOOL_PROP_READONLY:
	case ZPOOL_PROP_CACHEFILE:
	/*
	* 'readonly' and 'cachefile' are also non-persistent
	* properties.
	*/
	break;
	case ZPOOL_PROP_COMMENT:
	strval = fnvpair_value_string(elem);
	if (spa->spa_comment != NULL)
	spa_strfree(spa->spa_comment);
	spa->spa_comment = spa_strdup(strval);
	/*
	* We need to dirty the configuration on all the vdevs
	* so that their labels get updated. We also need to
	* update the cache file to keep it in sync with the
	* MOS version. It's unnecessary to do this for pool
	* creation since the vdev's configuration has already
	* been dirtied.
	*/
	if (tx->tx_txg != TXG_INITIAL) {
	vdev_config_dirty(spa->spa_root_vdev);
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}
	spa_history_log_internal(spa, "set", tx,
	"%s=%s", nvpair_name(elem), strval);
	break;
	case ZPOOL_PROP_COMPATIBILITY:
	strval = fnvpair_value_string(elem);
	if (spa->spa_compatibility != NULL)
	spa_strfree(spa->spa_compatibility);
	spa->spa_compatibility = spa_strdup(strval);
	/*
	* Dirty the configuration on vdevs as above.
	*/
	if (tx->tx_txg != TXG_INITIAL) {
	vdev_config_dirty(spa->spa_root_vdev);
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	spa_history_log_internal(spa, "set", tx,
	"%s=%s", nvpair_name(elem), strval);
	break;

	default:
	/*
	* Set pool property values in the poolprops mos object.
	*/
	if (spa->spa_pool_props_object == 0) {
	spa->spa_pool_props_object =
	zap_create_link(mos, DMU_OT_POOL_PROPS,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
	tx);
	}

	/* normalize the property name */
	propname = zpool_prop_to_name(prop);
	proptype = zpool_prop_get_type(prop);

	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	ASSERT(proptype == PROP_TYPE_STRING);
	strval = fnvpair_value_string(elem);
	VERIFY0(zap_update(mos,
	spa->spa_pool_props_object, propname,
	1, strlen(strval) + 1, strval, tx));
	spa_history_log_internal(spa, "set", tx,
	"%s=%s", nvpair_name(elem), strval);
	} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
	intval = fnvpair_value_uint64(elem);

	if (proptype == PROP_TYPE_INDEX) {
	const char *unused;
	VERIFY0(zpool_prop_index_to_string(
	prop, intval, &unused));
	}
	VERIFY0(zap_update(mos,
	spa->spa_pool_props_object, propname,
	8, 1, &intval, tx));
	spa_history_log_internal(spa, "set", tx,
	"%s=%lld", nvpair_name(elem),
	(longlong_t)intval);
	} else {
	ASSERT(0); /* not allowed */
	}

	switch (prop) {
	case ZPOOL_PROP_DELEGATION:
	spa->spa_delegation = intval;
	break;
	case ZPOOL_PROP_BOOTFS:
	spa->spa_bootfs = intval;
	break;
	case ZPOOL_PROP_FAILUREMODE:
	spa->spa_failmode = intval;
	break;
	case ZPOOL_PROP_AUTOTRIM:
	spa->spa_autotrim = intval;
	spa_async_request(spa,
	SPA_ASYNC_AUTOTRIM_RESTART);
	break;
	case ZPOOL_PROP_AUTOEXPAND:
	spa->spa_autoexpand = intval;
	if (tx->tx_txg != TXG_INITIAL)
	spa_async_request(spa,
	SPA_ASYNC_AUTOEXPAND);
	break;
	case ZPOOL_PROP_MULTIHOST:
	spa->spa_multihost = intval;
	break;
	default:
	break;
	}
	}

	}

	mutex_exit(&spa->spa_props_lock);
	}

	/*
	* Perform one-time upgrade on-disk changes. spa_version() does not
	* reflect the new version this txg, so there must be no changes this
	* txg to anything that the upgrade code depends on after it executes.
	* Therefore this must be called after dsl_pool_sync() does the sync
	* tasks.
	*/
	static void
	spa_sync_upgrades(spa_t spa, dmu_tx_t tx)
	{
	if (spa_sync_pass(spa) != 1)
	return;

	dsl_pool_t *dp = spa->spa_dsl_pool;
	rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);

	if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
	dsl_pool_create_origin(dp, tx);

	/* Keeping the origin open increases spa_minref */
	spa->spa_minref += 3;
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
	dsl_pool_upgrade_clones(dp, tx);
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) {
	dsl_pool_upgrade_dir_clones(dp, tx);

	/* Keeping the freedir open increases spa_minref */
	spa->spa_minref += 3;
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
	spa_feature_create_zap_objects(spa, tx);
	}

	/*
	* LZ4_COMPRESS feature's behaviour was changed to activate_on_enable
	* when possibility to use lz4 compression for metadata was added
	* Old pools that have this feature enabled must be upgraded to have
	* this feature active
	*/
	if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
	boolean_t lz4_en = spa_feature_is_enabled(spa,
	SPA_FEATURE_LZ4_COMPRESS);
	boolean_t lz4_ac = spa_feature_is_active(spa,
	SPA_FEATURE_LZ4_COMPRESS);

	if (lz4_en && !lz4_ac)
	spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx);
	}

	/*
	* If we haven't written the salt, do so now. Note that the
	* feature may not be activated yet, but that's fine since
	* the presence of this ZAP entry is backwards compatible.
	*/
	if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CHECKSUM_SALT) == ENOENT) {
	VERIFY0(zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1,
	sizeof (spa->spa_cksum_salt.zcs_bytes),
	spa->spa_cksum_salt.zcs_bytes, tx));
	}

	rrw_exit(&dp->dp_config_rwlock, FTAG);
	}

	static void
	vdev_indirect_state_sync_verify(vdev_t *vd)
	{
	vdev_indirect_mapping_t *vim __maybe_unused = vd->vdev_indirect_mapping;
	vdev_indirect_births_t *vib __maybe_unused = vd->vdev_indirect_births;

	if (vd->vdev_ops == &vdev_indirect_ops) {
	ASSERT(vim != NULL);
	ASSERT(vib != NULL);
	}

	uint64_t obsolete_sm_object = 0;
	ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	ASSERT(vd->vdev_obsolete_sm != NULL);
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);
	ASSERT(vdev_indirect_mapping_num_entries(vim) > 0);
	ASSERT(vdev_indirect_mapping_bytes_mapped(vim) > 0);
	ASSERT3U(obsolete_sm_object, ==,
	space_map_object(vd->vdev_obsolete_sm));
	ASSERT3U(vdev_indirect_mapping_bytes_mapped(vim), >=,
	space_map_allocated(vd->vdev_obsolete_sm));
	}
	ASSERT(vd->vdev_obsolete_segments != NULL);

	/*
	* Since frees / remaps to an indirect vdev can only
	* happen in syncing context, the obsolete segments
	* tree must be empty when we start syncing.
	*/
	ASSERT0(range_tree_space(vd->vdev_obsolete_segments));
	}

	/*
	* Set the top-level vdev's max queue depth. Evaluate each top-level's
	* async write queue depth in case it changed. The max queue depth will
	* not change in the middle of syncing out this txg.
	*/
	static void
	spa_sync_adjust_vdev_max_queue_depth(spa_t *spa)
	{
	ASSERT(spa_writeable(spa));

	vdev_t *rvd = spa->spa_root_vdev;
	uint32_t max_queue_depth = zfs_vdev_async_write_max_active *
	zfs_vdev_queue_depth_pct / 100;
	metaslab_class_t *normal = spa_normal_class(spa);
	metaslab_class_t *special = spa_special_class(spa);
	metaslab_class_t *dedup = spa_dedup_class(spa);

	uint64_t slots_per_allocator = 0;
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	metaslab_group_t *mg = tvd->vdev_mg;
	if (mg == NULL \|\| !metaslab_group_initialized(mg))
	continue;

	metaslab_class_t *mc = mg->mg_class;
	if (mc != normal && mc != special && mc != dedup)
	continue;

	/*
	* It is safe to do a lock-free check here because only async
	* allocations look at mg_max_alloc_queue_depth, and async
	* allocations all happen from spa_sync().
	*/
	for (int i = 0; i < mg->mg_allocators; i++) {
	ASSERT0(zfs_refcount_count(
	&(mg->mg_allocator[i].mga_alloc_queue_depth)));
	}
	mg->mg_max_alloc_queue_depth = max_queue_depth;

	for (int i = 0; i < mg->mg_allocators; i++) {
	mg->mg_allocator[i].mga_cur_max_alloc_queue_depth =
	zfs_vdev_def_queue_depth;
	}
	slots_per_allocator += zfs_vdev_def_queue_depth;
	}

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	ASSERT0(zfs_refcount_count(&normal->mc_allocator[i].
	mca_alloc_slots));
	ASSERT0(zfs_refcount_count(&special->mc_allocator[i].
	mca_alloc_slots));
	ASSERT0(zfs_refcount_count(&dedup->mc_allocator[i].
	mca_alloc_slots));
	normal->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	special->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	dedup->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	}
	normal->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	special->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	dedup->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	}

	static void
	spa_sync_condense_indirect(spa_t spa, dmu_tx_t tx)
	{
	ASSERT(spa_writeable(spa));

	vdev_t *rvd = spa->spa_root_vdev;
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];
	vdev_indirect_state_sync_verify(vd);

	if (vdev_indirect_should_condense(vd)) {
	spa_condense_indirect_start_sync(vd, tx);
	break;
	}
	}
	}

	static void
	spa_sync_iterate_to_convergence(spa_t spa, dmu_tx_t tx)
	{
	objset_t *mos = spa->spa_meta_objset;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	uint64_t txg = tx->tx_txg;
	bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];

	do {
	int pass = ++spa->spa_sync_pass;

	spa_sync_config_object(spa, tx);
	spa_sync_aux_dev(spa, &spa->spa_spares, tx,
	ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES);
	spa_sync_aux_dev(spa, &spa->spa_l2cache, tx,
	ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE);
	spa_errlog_sync(spa, txg);
	dsl_pool_sync(dp, txg);

	if (pass < zfs_sync_pass_deferred_free \|\|
	spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
	/*
	* If the log space map feature is active we don't
	* care about deferred frees and the deferred bpobj
	* as the log space map should effectively have the
	* same results (i.e. appending only to one object).
	*/
	spa_sync_frees(spa, free_bpl, tx);
	} else {
	/*
	* We can not defer frees in pass 1, because
	* we sync the deferred frees later in pass 1.
	*/
	ASSERT3U(pass, >, 1);
	bplist_iterate(free_bpl, bpobj_enqueue_alloc_cb,
	&spa->spa_deferred_bpobj, tx);
	}

	ddt_sync(spa, txg);
	dsl_scan_sync(dp, tx);
	svr_sync(spa, tx);
	spa_sync_upgrades(spa, tx);

	spa_flush_metaslabs(spa, tx);

	vdev_t *vd = NULL;
	while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg))
	!= NULL)
	vdev_sync(vd, txg);

	/*
	* Note: We need to check if the MOS is dirty because we could
	* have marked the MOS dirty without updating the uberblock
	* (e.g. if we have sync tasks but no dirty user data). We need
	* to check the uberblock's rootbp because it is updated if we
	* have synced out dirty data (though in this case the MOS will
	* most likely also be dirty due to second order effects, we
	* don't want to rely on that here).
	*/
	if (pass == 1 &&
	spa->spa_uberblock.ub_rootbp.blk_birth < txg &&
	!dmu_objset_is_dirty(mos, txg)) {
	/*
	* Nothing changed on the first pass, therefore this
	* TXG is a no-op. Avoid syncing deferred frees, so
	* that we can keep this TXG as a no-op.
	*/
	ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
	ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
	ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg));
	ASSERT(txg_list_empty(&dp->dp_early_sync_tasks, txg));
	break;
	}

	spa_sync_deferred_frees(spa, tx);
	} while (dmu_objset_is_dirty(mos, txg));
	}

	/*
	* Rewrite the vdev configuration (which includes the uberblock) to
	* commit the transaction group.
	*
	* If there are no dirty vdevs, we sync the uberblock to a few random
	* top-level vdevs that are known to be visible in the config cache
	* (see spa_vdev_add() for a complete description). If there are dirty
	* vdevs, sync the uberblock to all vdevs.
	*/
	static void
	spa_sync_rewrite_vdev_config(spa_t spa, dmu_tx_t tx)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t txg = tx->tx_txg;

	for (;;) {
	int error = 0;

	/*
	* We hold SCL_STATE to prevent vdev open/close/etc.
	* while we're attempting to write the vdev labels.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	if (list_is_empty(&spa->spa_config_dirty_list)) {
	vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
	int svdcount = 0;
	int children = rvd->vdev_children;
	int c0 = random_in_range(children);

	for (int c = 0; c < children; c++) {
	vdev_t *vd =
	rvd->vdev_child[(c0 + c) % children];

	/* Stop when revisiting the first vdev */
	if (c > 0 && svd[0] == vd)
	break;

	if (vd->vdev_ms_array == 0 \|\|
	vd->vdev_islog \|\|
	!vdev_is_concrete(vd))
	continue;

	svd[svdcount++] = vd;
	if (svdcount == SPA_SYNC_MIN_VDEVS)
	break;
	}
	error = vdev_config_sync(svd, svdcount, txg);
	} else {
	error = vdev_config_sync(rvd->vdev_child,
	rvd->vdev_children, txg);
	}

	if (error == 0)
	spa->spa_last_synced_guid = rvd->vdev_guid;

	spa_config_exit(spa, SCL_STATE, FTAG);

	if (error == 0)
	break;
	zio_suspend(spa, NULL, ZIO_SUSPEND_IOERR);
	zio_resume_wait(spa);
	}
	}

	/*
	* Sync the specified transaction group. New blocks may be dirtied as
	* part of the process, so we iterate until it converges.
	*/
	void
	spa_sync(spa_t *spa, uint64_t txg)
	{
	vdev_t *vd = NULL;

	VERIFY(spa_writeable(spa));

	/*
	* Wait for i/os issued in open context that need to complete
	* before this txg syncs.
	*/
	(void) zio_wait(spa->spa_txg_zio[txg & TXG_MASK]);
	spa->spa_txg_zio[txg & TXG_MASK] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);

	/*
	* Lock out configuration changes.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	spa->spa_syncing_txg = txg;
	spa->spa_sync_pass = 0;

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_enter(&spa->spa_allocs[i].spaa_lock);
	VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
	mutex_exit(&spa->spa_allocs[i].spaa_lock);
	}

	/*
	* If there are any pending vdev state changes, convert them
	* into config changes that go out with this transaction group.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	while (list_head(&spa->spa_state_dirty_list) != NULL) {
	/*
	* We need the write lock here because, for aux vdevs,
	* calling vdev_config_dirty() modifies sav_config.
	* This is ugly and will become unnecessary when we
	* eliminate the aux vdev wart by integrating all vdevs
	* into the root vdev tree.
	*/
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_WRITER);
	while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) {
	vdev_state_clean(vd);
	vdev_config_dirty(vd);
	}
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	}
	spa_config_exit(spa, SCL_STATE, FTAG);

	dsl_pool_t *dp = spa->spa_dsl_pool;
	dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);

	spa->spa_sync_starttime = gethrtime();
	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
	spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
	spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
	NSEC_TO_TICK(spa->spa_deadman_synctime));

	/*
	* If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
	* set spa_deflate if we have no raid-z vdevs.
	*/
	if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) {
	vdev_t *rvd = spa->spa_root_vdev;

	int i;
	for (i = 0; i < rvd->vdev_children; i++) {
	vd = rvd->vdev_child[i];
	if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE)
	break;
	}
	if (i == rvd->vdev_children) {
	spa->spa_deflate = TRUE;
	VERIFY0(zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
	sizeof (uint64_t), 1, &spa->spa_deflate, tx));
	}
	}

	spa_sync_adjust_vdev_max_queue_depth(spa);

	spa_sync_condense_indirect(spa, tx);

	spa_sync_iterate_to_convergence(spa, tx);

	#ifdef ZFS_DEBUG
	if (!list_is_empty(&spa->spa_config_dirty_list)) {
	/*
	* Make sure that the number of ZAPs for all the vdevs matches
	* the number of ZAPs in the per-vdev ZAP list. This only gets
	* called if the config is dirty; otherwise there may be
	* outstanding AVZ operations that weren't completed in
	* spa_sync_config_object.
	*/
	uint64_t all_vdev_zap_entry_count;
	ASSERT0(zap_count(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count));
	ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==,
	all_vdev_zap_entry_count);
	}
	#endif

	if (spa->spa_vdev_removal != NULL) {
	ASSERT0(spa->spa_vdev_removal->svr_bytes_done[txg & TXG_MASK]);
	}

	spa_sync_rewrite_vdev_config(spa, tx);
	dmu_tx_commit(tx);

	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
	spa->spa_deadman_tqid = 0;

	/*
	* Clear the dirty config list.
	*/
	while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL)
	vdev_config_clean(vd);

	/*
	* Now that the new config has synced transactionally,
	* let it become visible to the config cache.
	*/
	if (spa->spa_config_syncing != NULL) {
	spa_config_set(spa, spa->spa_config_syncing);
	spa->spa_config_txg = txg;
	spa->spa_config_syncing = NULL;
	}

	dsl_pool_sync_done(dp, txg);

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_enter(&spa->spa_allocs[i].spaa_lock);
	VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
	mutex_exit(&spa->spa_allocs[i].spaa_lock);
	}

	/*
	* Update usable space statistics.
	*/
	while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))
	!= NULL)
	vdev_sync_done(vd, txg);

	metaslab_class_evict_old(spa->spa_normal_class, txg);
	metaslab_class_evict_old(spa->spa_log_class, txg);

	spa_sync_close_syncing_log_sm(spa);

	spa_update_dspace(spa);

	/*
	* It had better be the case that we didn't dirty anything
	* since vdev_config_sync().
	*/
	ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
	ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
	ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg));

	while (zfs_pause_spa_sync)
	delay(1);

	spa->spa_sync_pass = 0;

	/*
	* Update the last synced uberblock here. We want to do this at
	* the end of spa_sync() so that consumers of spa_last_synced_txg()
	* will be guaranteed that all the processing associated with
	* that txg has been completed.
	*/
	spa->spa_ubsync = spa->spa_uberblock;
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	spa_handle_ignored_writes(spa);

	/*
	* If any async tasks have been requested, kick them off.
	*/
	spa_async_dispatch(spa);
	}

	/*
	* Sync all pools. We don't want to hold the namespace lock across these
	* operations, so we take a reference on the spa_t and drop the lock during the
	* sync.
	*/
	void
	spa_sync_allpools(void)
	{
	spa_t *spa = NULL;
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL) {
	if (spa_state(spa) != POOL_STATE_ACTIVE \|\|
	!spa_writeable(spa) \|\| spa_suspended(spa))
	continue;
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	txg_wait_synced(spa_get_dsl(spa), 0);
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);
	}
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* ==========================================================================
	* Miscellaneous routines
	* ==========================================================================
	*/

	/*
	* Remove all pools in the system.
	*/
	void
	spa_evict_all(void)
	{
	spa_t *spa;

	/*
	* Remove all cached state. All pools should be closed now,
	* so every spa in the AVL tree should be unreferenced.
	*/
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(NULL)) != NULL) {
	/*
	* Stop async tasks. The async thread may need to detach
	* a device that's been replaced, which requires grabbing
	* spa_namespace_lock, so we must drop it here.
	*/
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	spa_async_suspend(spa);
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);

	if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
	spa_unload(spa);
	spa_deactivate(spa);
	}
	spa_remove(spa);
	}
	mutex_exit(&spa_namespace_lock);
	}

	vdev_t *
	spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux)
	{
	vdev_t *vd;
	int i;

	if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL)
	return (vd);

	if (aux) {
	for (i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vd = spa->spa_l2cache.sav_vdevs[i];
	if (vd->vdev_guid == guid)
	return (vd);
	}

	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	vd = spa->spa_spares.sav_vdevs[i];
	if (vd->vdev_guid == guid)
	return (vd);
	}
	}

	return (NULL);
	}

	void
	spa_upgrade(spa_t *spa, uint64_t version)
	{
	ASSERT(spa_writeable(spa));

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* This should only be called for a non-faulted pool, and since a
	* future version would result in an unopenable pool, this shouldn't be
	* possible.
	*/
	ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version));
	ASSERT3U(version, >=, spa->spa_uberblock.ub_version);

	spa->spa_uberblock.ub_version = version;
	vdev_config_dirty(spa->spa_root_vdev);

	spa_config_exit(spa, SCL_ALL, FTAG);

	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	static boolean_t
	spa_has_aux_vdev(spa_t spa, uint64_t guid, spa_aux_vdev_t sav)
	{
	(void) spa;
	int i;
	uint64_t vdev_guid;

	for (i = 0; i < sav->sav_count; i++)
	if (sav->sav_vdevs[i]->vdev_guid == guid)
	return (B_TRUE);

	for (i = 0; i < sav->sav_npending; i++) {
	if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID,
	&vdev_guid) == 0 && vdev_guid == guid)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	boolean_t
	spa_has_l2cache(spa_t *spa, uint64_t guid)
	{
	return (spa_has_aux_vdev(spa, guid, &spa->spa_l2cache));
	}

	boolean_t
	spa_has_spare(spa_t *spa, uint64_t guid)
	{
	return (spa_has_aux_vdev(spa, guid, &spa->spa_spares));
	}

	/*
	* Check if a pool has an active shared spare device.
	* Note: reference count of an active spare is 2, as a spare and as a replace
	*/
	static boolean_t
	spa_has_active_shared_spare(spa_t *spa)
	{
	int i, refcnt;
	uint64_t pool;
	spa_aux_vdev_t *sav = &spa->spa_spares;

	for (i = 0; i < sav->sav_count; i++) {
	if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool,
	&refcnt) && pool != 0ULL && pool == spa_guid(spa) &&
	refcnt > 2)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	uint64_t
	spa_total_metaslabs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	uint64_t m = 0;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];
	if (!vdev_is_concrete(vd))
	continue;
	m += vd->vdev_ms_count;
	}
	return (m);
	}

	/*
	* Notify any waiting threads that some activity has switched from being in-
	* progress to not-in-progress so that the thread can wake up and determine
	* whether it is finished waiting.
	*/
	void
	spa_notify_waiters(spa_t *spa)
	{
	/*
	* Acquiring spa_activities_lock here prevents the cv_broadcast from
	* happening between the waiting thread's check and cv_wait.
	*/
	mutex_enter(&spa->spa_activities_lock);
	cv_broadcast(&spa->spa_activities_cv);
	mutex_exit(&spa->spa_activities_lock);
	}

	/*
	* Notify any waiting threads that the pool is exporting, and then block until
	* they are finished using the spa_t.
	*/
	void
	spa_wake_waiters(spa_t *spa)
	{
	mutex_enter(&spa->spa_activities_lock);
	spa->spa_waiters_cancel = B_TRUE;
	cv_broadcast(&spa->spa_activities_cv);
	while (spa->spa_waiters != 0)
	cv_wait(&spa->spa_waiters_cv, &spa->spa_activities_lock);
	spa->spa_waiters_cancel = B_FALSE;
	mutex_exit(&spa->spa_activities_lock);
	}

	/* Whether the vdev or any of its descendants are being initialized/trimmed. */
	static boolean_t
	spa_vdev_activity_in_progress_impl(vdev_t *vd, zpool_wait_activity_t activity)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_CONFIG \| SCL_STATE, RW_READER));
	ASSERT(MUTEX_HELD(&spa->spa_activities_lock));
	ASSERT(activity == ZPOOL_WAIT_INITIALIZE \|\|
	activity == ZPOOL_WAIT_TRIM);

	kmutex_t *lock = activity == ZPOOL_WAIT_INITIALIZE ?
	&vd->vdev_initialize_lock : &vd->vdev_trim_lock;

	mutex_exit(&spa->spa_activities_lock);
	mutex_enter(lock);
	mutex_enter(&spa->spa_activities_lock);

	boolean_t in_progress = (activity == ZPOOL_WAIT_INITIALIZE) ?
	(vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) :
	(vd->vdev_trim_state == VDEV_TRIM_ACTIVE);
	mutex_exit(lock);

	if (in_progress)
	return (B_TRUE);

	for (int i = 0; i < vd->vdev_children; i++) {
	if (spa_vdev_activity_in_progress_impl(vd->vdev_child[i],
	activity))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* If use_guid is true, this checks whether the vdev specified by guid is
	* being initialized/trimmed. Otherwise, it checks whether any vdev in the pool
	* is being initialized/trimmed. The caller must hold the config lock and
	* spa_activities_lock.
	*/
	static int
	spa_vdev_activity_in_progress(spa_t *spa, boolean_t use_guid, uint64_t guid,
	zpool_wait_activity_t activity, boolean_t *in_progress)
	{
	mutex_exit(&spa->spa_activities_lock);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	mutex_enter(&spa->spa_activities_lock);

	vdev_t *vd;
	if (use_guid) {
	vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| !vd->vdev_ops->vdev_op_leaf) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (EINVAL);
	}
	} else {
	vd = spa->spa_root_vdev;
	}

	*in_progress = spa_vdev_activity_in_progress_impl(vd, activity);

	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (0);
	}

	/*
	* Locking for waiting threads
	* ---------------------------
	*
	* Waiting threads need a way to check whether a given activity is in progress,
	* and then, if it is, wait for it to complete. Each activity will have some
	* in-memory representation of the relevant on-disk state which can be used to
	* determine whether or not the activity is in progress. The in-memory state and
	* the locking used to protect it will be different for each activity, and may
	* not be suitable for use with a cvar (e.g., some state is protected by the
	* config lock). To allow waiting threads to wait without any races, another
	* lock, spa_activities_lock, is used.
	*
	* When the state is checked, both the activity-specific lock (if there is one)
	* and spa_activities_lock are held. In some cases, the activity-specific lock
	* is acquired explicitly (e.g. the config lock). In others, the locking is
	* internal to some check (e.g. bpobj_is_empty). After checking, the waiting
	* thread releases the activity-specific lock and, if the activity is in
	* progress, then cv_waits using spa_activities_lock.
	*
	* The waiting thread is woken when another thread, one completing some
	* activity, updates the state of the activity and then calls
	* spa_notify_waiters, which will cv_broadcast. This 'completing' thread only
	* needs to hold its activity-specific lock when updating the state, and this
	* lock can (but doesn't have to) be dropped before calling spa_notify_waiters.
	*
	* Because spa_notify_waiters acquires spa_activities_lock before broadcasting,
	* and because it is held when the waiting thread checks the state of the
	* activity, it can never be the case that the completing thread both updates
	* the activity state and cv_broadcasts in between the waiting thread's check
	* and cv_wait. Thus, a waiting thread can never miss a wakeup.
	*
	* In order to prevent deadlock, when the waiting thread does its check, in some
	* cases it will temporarily drop spa_activities_lock in order to acquire the
	* activity-specific lock. The order in which spa_activities_lock and the
	* activity specific lock are acquired in the waiting thread is determined by
	* the order in which they are acquired in the completing thread; if the
	* completing thread calls spa_notify_waiters with the activity-specific lock
	* held, then the waiting thread must also acquire the activity-specific lock
	* first.
	*/

	static int
	spa_activity_in_progress(spa_t *spa, zpool_wait_activity_t activity,
	boolean_t use_tag, uint64_t tag, boolean_t *in_progress)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&spa->spa_activities_lock));

	switch (activity) {
	case ZPOOL_WAIT_CKPT_DISCARD:
	*in_progress =
	(spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT) &&
	zap_contains(spa_meta_objset(spa),
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT) ==
	ENOENT);
	break;
	case ZPOOL_WAIT_FREE:
	*in_progress = ((spa_version(spa) >= SPA_VERSION_DEADLISTS &&
	!bpobj_is_empty(&spa->spa_dsl_pool->dp_free_bpobj)) \|\|
	spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY) \|\|
	spa_livelist_delete_check(spa));
	break;
	case ZPOOL_WAIT_INITIALIZE:
	case ZPOOL_WAIT_TRIM:
	error = spa_vdev_activity_in_progress(spa, use_tag, tag,
	activity, in_progress);
	break;
	case ZPOOL_WAIT_REPLACE:
	mutex_exit(&spa->spa_activities_lock);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	mutex_enter(&spa->spa_activities_lock);

	*in_progress = vdev_replace_in_progress(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	break;
	case ZPOOL_WAIT_REMOVE:
	*in_progress = (spa->spa_removing_phys.sr_state ==
	DSS_SCANNING);
	break;
	case ZPOOL_WAIT_RESILVER:
	if ((*in_progress = vdev_rebuild_active(spa->spa_root_vdev)))
	break;
	zfs_fallthrough;
	case ZPOOL_WAIT_SCRUB:
	{
	boolean_t scanning, paused, is_scrub;
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;

	is_scrub = (scn->scn_phys.scn_func == POOL_SCAN_SCRUB);
	scanning = (scn->scn_phys.scn_state == DSS_SCANNING);
	paused = dsl_scan_is_paused_scrub(scn);
	*in_progress = (scanning && !paused &&
	is_scrub == (activity == ZPOOL_WAIT_SCRUB));
	break;
	}
	default:
	panic("unrecognized value for activity %d", activity);
	}

	return (error);
	}

	static int
	spa_wait_common(const char *pool, zpool_wait_activity_t activity,
	boolean_t use_tag, uint64_t tag, boolean_t *waited)
	{
	/*
	* The tag is used to distinguish between instances of an activity.
	* 'initialize' and 'trim' are the only activities that we use this for.
	* The other activities can only have a single instance in progress in a
	* pool at one time, making the tag unnecessary.
	*
	* There can be multiple devices being replaced at once, but since they
	* all finish once resilvering finishes, we don't bother keeping track
	* of them individually, we just wait for them all to finish.
	*/
	if (use_tag && activity != ZPOOL_WAIT_INITIALIZE &&
	activity != ZPOOL_WAIT_TRIM)
	return (EINVAL);

	if (activity < 0 \|\| activity >= ZPOOL_WAIT_NUM_ACTIVITIES)
	return (EINVAL);

	spa_t *spa;
	int error = spa_open(pool, &spa, FTAG);
	if (error != 0)
	return (error);

	/*
	* Increment the spa's waiter count so that we can call spa_close and
	* still ensure that the spa_t doesn't get freed before this thread is
	* finished with it when the pool is exported. We want to call spa_close
	* before we start waiting because otherwise the additional ref would
	* prevent the pool from being exported or destroyed throughout the
	* potentially long wait.
	*/
	mutex_enter(&spa->spa_activities_lock);
	spa->spa_waiters++;
	spa_close(spa, FTAG);

	*waited = B_FALSE;
	for (;;) {
	boolean_t in_progress;
	error = spa_activity_in_progress(spa, activity, use_tag, tag,
	&in_progress);

	if (error \|\| !in_progress \|\| spa->spa_waiters_cancel)
	break;

	*waited = B_TRUE;

	if (cv_wait_sig(&spa->spa_activities_cv,
	&spa->spa_activities_lock) == 0) {
	error = EINTR;
	break;
	}
	}

	spa->spa_waiters--;
	cv_signal(&spa->spa_waiters_cv);
	mutex_exit(&spa->spa_activities_lock);

	return (error);
	}

	/*
	* Wait for a particular instance of the specified activity to complete, where
	* the instance is identified by 'tag'
	*/
	int
	spa_wait_tag(const char *pool, zpool_wait_activity_t activity, uint64_t tag,
	boolean_t *waited)
	{
	return (spa_wait_common(pool, activity, B_TRUE, tag, waited));
	}

	/*
	* Wait for all instances of the specified activity complete
	*/
	int
	spa_wait(const char pool, zpool_wait_activity_t activity, boolean_t waited)
	{

	return (spa_wait_common(pool, activity, B_FALSE, 0, waited));
	}

	sysevent_t *
	spa_event_create(spa_t spa, vdev_t vd, nvlist_t hist_nvl, const char name)
	{
	sysevent_t *ev = NULL;
	#ifdef _KERNEL
	nvlist_t *resource;

	resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl);
	if (resource) {
	ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP);
	ev->resource = resource;
	}
	#else
	(void) spa, (void) vd, (void) hist_nvl, (void) name;
	#endif
	return (ev);
	}

	void
	spa_event_post(sysevent_t *ev)
	{
	#ifdef _KERNEL
	if (ev) {
	zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb);
	kmem_free(ev, sizeof (*ev));
	}
	#else
	(void) ev;
	#endif
	}

	/*
	* Post a zevent corresponding to the given sysevent. The 'name' must be one
	* of the event definitions in sys/sysevent/eventdefs.h. The payload will be
	* filled in from the spa and (optionally) the vdev. This doesn't do anything
	* in the userland libzpool, as we don't want consumers to misinterpret ztest
	* or zdb as real changes.
	*/
	void
	spa_event_notify(spa_t spa, vdev_t vd, nvlist_t hist_nvl, const char name)
	{
	spa_event_post(spa_event_create(spa, vd, hist_nvl, name));
	}

	/* state manipulation functions */
	EXPORT_SYMBOL(spa_open);
	EXPORT_SYMBOL(spa_open_rewind);
	EXPORT_SYMBOL(spa_get_stats);
	EXPORT_SYMBOL(spa_create);
	EXPORT_SYMBOL(spa_import);
	EXPORT_SYMBOL(spa_tryimport);
	EXPORT_SYMBOL(spa_destroy);
	EXPORT_SYMBOL(spa_export);
	EXPORT_SYMBOL(spa_reset);
	EXPORT_SYMBOL(spa_async_request);
	EXPORT_SYMBOL(spa_async_suspend);
	EXPORT_SYMBOL(spa_async_resume);
	EXPORT_SYMBOL(spa_inject_addref);
	EXPORT_SYMBOL(spa_inject_delref);
	EXPORT_SYMBOL(spa_scan_stat_init);
	EXPORT_SYMBOL(spa_scan_get_stats);

	/* device manipulation */
	EXPORT_SYMBOL(spa_vdev_add);
	EXPORT_SYMBOL(spa_vdev_attach);
	EXPORT_SYMBOL(spa_vdev_detach);
	EXPORT_SYMBOL(spa_vdev_setpath);
	EXPORT_SYMBOL(spa_vdev_setfru);
	EXPORT_SYMBOL(spa_vdev_split_mirror);

	/* spare statech is global across all pools) */
	EXPORT_SYMBOL(spa_spare_add);
	EXPORT_SYMBOL(spa_spare_remove);
	EXPORT_SYMBOL(spa_spare_exists);
	EXPORT_SYMBOL(spa_spare_activate);

	/* L2ARC statech is global across all pools) */
	EXPORT_SYMBOL(spa_l2cache_add);
	EXPORT_SYMBOL(spa_l2cache_remove);
	EXPORT_SYMBOL(spa_l2cache_exists);
	EXPORT_SYMBOL(spa_l2cache_activate);
	EXPORT_SYMBOL(spa_l2cache_drop);

	/* scanning */
	EXPORT_SYMBOL(spa_scan);
	EXPORT_SYMBOL(spa_scan_stop);

	/* spa syncing */
	EXPORT_SYMBOL(spa_sync); /* only for DMU use */
	EXPORT_SYMBOL(spa_sync_allpools);

	/* properties */
	EXPORT_SYMBOL(spa_prop_set);
	EXPORT_SYMBOL(spa_prop_get);
	EXPORT_SYMBOL(spa_prop_clear_bootfs);

	/* asynchronous event notification */
	EXPORT_SYMBOL(spa_event_notify);

	/* BEGIN CSTYLED */
	-ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_shift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_shift, UINT, ZMOD_RW,
	"log2 fraction of arc that can be used by inflight I/Os when "
	"verifying pool during import");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_metadata, INT, ZMOD_RW,
	"Set to traverse metadata on pool import");

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_data, INT, ZMOD_RW,
	"Set to traverse data on pool import");

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_print_vdev_tree, INT, ZMOD_RW,
	"Print vdev tree to zfs_dbgmsg during pool import");

	ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_pct, UINT, ZMOD_RD,
	"Percentage of CPUs to run an IO worker thread");

	ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_tpq, UINT, ZMOD_RD,
	"Number of threads per IO worker taskqueue");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, zfs_, max_missing_tvds, ULONG, ZMOD_RW,
	"Allow importing pool with up to this number of missing top-level "
	"vdevs (in read-only mode)");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_pause, INT,
	ZMOD_RW, "Set the livelist condense zthr to pause");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_pause, INT,
	ZMOD_RW, "Set the livelist condense synctask to pause");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_cancel,
	INT, ZMOD_RW,
	"Whether livelist condensing was canceled in the synctask");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_cancel,
	INT, ZMOD_RW,
	"Whether livelist condensing was canceled in the zthr function");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, new_alloc, INT,
	ZMOD_RW,
	"Whether extra ALLOC blkptrs were added to a livelist entry while it "
	"was being condensed");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/spa_config.c b/sys/contrib/openzfs/module/zfs/spa_config.c
	index 91ac5c05e8af..5165c370403b 100644
	--- a/sys/contrib/openzfs/module/zfs/spa_config.c
	+++ b/sys/contrib/openzfs/module/zfs/spa_config.c
	@@ -1,623 +1,635 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	#include <sys/spa.h>
	#include <sys/file.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa_impl.h>
	#include <sys/nvpair.h>
	#include <sys/fs/zfs.h>
	#include <sys/vdev_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/systeminfo.h>
	#include <sys/sunddi.h>
	#include <sys/zfeature.h>
	#include <sys/zfs_file.h>
	#include <sys/zfs_context.h>
	#ifdef _KERNEL
	#include <sys/zone.h>
	#endif

	/*
	* Pool configuration repository.
	*
	* Pool configuration is stored as a packed nvlist on the filesystem. By
	* default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
	* (when the ZFS module is loaded). Pools can also have the 'cachefile'
	* property set that allows them to be stored in an alternate location until
	* the control of external software.
	*
	* For each cache file, we have a single nvlist which holds all the
	* configuration information. When the module loads, we read this information
	* from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
	* maintained independently in spa.c. Whenever the namespace is modified, or
	* the configuration of a pool is changed, we call spa_write_cachefile(), which
	* walks through all the active pools and writes the configuration to disk.
	*/

	static uint64_t spa_config_generation = 1;

	/*
	* This can be overridden in userland to preserve an alternate namespace for
	* userland pools when doing testing.
	*/
	char spa_config_path = (char )ZPOOL_CACHE;
	#ifdef _KERNEL
	static int zfs_autoimport_disable = B_TRUE;
	#endif

	/*
	* Called when the module is first loaded, this routine loads the configuration
	* file into the SPA namespace. It does not actually open or load the pools; it
	* only populates the namespace.
	*/
	void
	spa_config_load(void)
	{
	void *buf = NULL;
	nvlist_t nvlist, child;
	nvpair_t *nvpair;
	char *pathname;
	zfs_file_t *fp;
	zfs_file_attr_t zfa;
	uint64_t fsize;
	int err;

	#ifdef _KERNEL
	if (zfs_autoimport_disable)
	return;
	#endif

	/*
	* Open the configuration file.
	*/
	pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	(void) snprintf(pathname, MAXPATHLEN, "%s", spa_config_path);

	err = zfs_file_open(pathname, O_RDONLY, 0, &fp);

	#ifdef __FreeBSD__
	if (err)
	err = zfs_file_open(ZPOOL_CACHE_BOOT, O_RDONLY, 0, &fp);
	#endif
	kmem_free(pathname, MAXPATHLEN);

	if (err)
	return;

	if (zfs_file_getattr(fp, &zfa))
	goto out;

	fsize = zfa.zfa_size;
	buf = kmem_alloc(fsize, KM_SLEEP);

	/*
	* Read the nvlist from the file.
	*/
	if (zfs_file_read(fp, buf, fsize, NULL) < 0)
	goto out;

	/*
	* Unpack the nvlist.
	*/
	if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
	goto out;

	/*
	* Iterate over all elements in the nvlist, creating a new spa_t for
	* each one with the specified configuration.
	*/
	mutex_enter(&spa_namespace_lock);
	nvpair = NULL;
	while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
	if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
	continue;

	child = fnvpair_value_nvlist(nvpair);

	if (spa_lookup(nvpair_name(nvpair)) != NULL)
	continue;
	(void) spa_add(nvpair_name(nvpair), child, NULL);
	}
	mutex_exit(&spa_namespace_lock);

	nvlist_free(nvlist);

	out:
	if (buf != NULL)
	kmem_free(buf, fsize);

	zfs_file_close(fp);
	}

	static int
	spa_config_remove(spa_config_dirent_t *dp)
	{
	int error = 0;

	/*
	* Remove the cache file. If zfs_file_unlink() in not supported by the
	* platform fallback to truncating the file which is functionally
	* equivalent.
	*/
	error = zfs_file_unlink(dp->scd_path);
	if (error == EOPNOTSUPP) {
	int flags = O_RDWR \| O_TRUNC;
	zfs_file_t *fp;

	error = zfs_file_open(dp->scd_path, flags, 0644, &fp);
	if (error == 0) {
	(void) zfs_file_fsync(fp, O_SYNC);
	(void) zfs_file_close(fp);
	}
	}

	return (error);
	}

	static int
	spa_config_write(spa_config_dirent_t dp, nvlist_t nvl)
	{
	size_t buflen;
	char *buf;
	int oflags = O_RDWR \| O_TRUNC \| O_CREAT \| O_LARGEFILE;
	char *temp;
	int err;
	zfs_file_t *fp;

	/*
	* If the nvlist is empty (NULL), then remove the old cachefile.
	*/
	if (nvl == NULL) {
	err = spa_config_remove(dp);
	if (err == ENOENT)
	err = 0;

	return (err);
	}

	/*
	* Pack the configuration into a buffer.
	*/
	buf = fnvlist_pack(nvl, &buflen);
	temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);

	/*
	* Write the configuration to disk. Due to the complexity involved
	* in performing a rename and remove from within the kernel the file
	* is instead truncated and overwritten in place. This way we always
	* have a consistent view of the data or a zero length file.
	*/
	err = zfs_file_open(dp->scd_path, oflags, 0644, &fp);
	if (err == 0) {
	err = zfs_file_write(fp, buf, buflen, NULL);
	if (err == 0)
	err = zfs_file_fsync(fp, O_SYNC);

	zfs_file_close(fp);
	if (err)
	(void) spa_config_remove(dp);
	}
	fnvlist_pack_free(buf, buflen);
	kmem_free(temp, MAXPATHLEN);
	return (err);
	}

	/*
	* Synchronize pool configuration to disk. This must be called with the
	* namespace lock held. Synchronizing the pool cache is typically done after
	* the configuration has been synced to the MOS. This exposes a window where
	* the MOS config will have been updated but the cache file has not. If
	* the system were to crash at that instant then the cached config may not
	* contain the correct information to open the pool and an explicit import
	* would be required.
	*/
	void
	-spa_write_cachefile(spa_t *target, boolean_t removing, boolean_t postsysevent)
	+spa_write_cachefile(spa_t *target, boolean_t removing, boolean_t postsysevent,
	+ boolean_t postblkidevent)
	{
	spa_config_dirent_t dp, tdp;
	nvlist_t *nvl;
	char *pool_name;
	boolean_t ccw_failure;
	int error = 0;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	if (!(spa_mode_global & SPA_MODE_WRITE))
	return;

	/*
	* Iterate over all cachefiles for the pool, past or present. When the
	* cachefile is changed, the new one is pushed onto this list, allowing
	* us to update previous cachefiles that no longer contain this pool.
	*/
	ccw_failure = B_FALSE;
	for (dp = list_head(&target->spa_config_list); dp != NULL;
	dp = list_next(&target->spa_config_list, dp)) {
	spa_t *spa = NULL;
	if (dp->scd_path == NULL)
	continue;

	/*
	* Iterate over all pools, adding any matching pools to 'nvl'.
	*/
	nvl = NULL;
	while ((spa = spa_next(spa)) != NULL) {
	/*
	* Skip over our own pool if we're about to remove
	* ourselves from the spa namespace or any pool that
	* is readonly. Since we cannot guarantee that a
	* readonly pool would successfully import upon reboot,
	* we don't allow them to be written to the cache file.
	*/
	if ((spa == target && removing) \|\|
	!spa_writeable(spa))
	continue;

	mutex_enter(&spa->spa_props_lock);
	tdp = list_head(&spa->spa_config_list);
	if (spa->spa_config == NULL \|\|
	tdp == NULL \|\|
	tdp->scd_path == NULL \|\|
	strcmp(tdp->scd_path, dp->scd_path) != 0) {
	mutex_exit(&spa->spa_props_lock);
	continue;
	}

	if (nvl == NULL)
	nvl = fnvlist_alloc();

	if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME)
	pool_name = fnvlist_lookup_string(
	spa->spa_config, ZPOOL_CONFIG_POOL_NAME);
	else
	pool_name = spa_name(spa);

	fnvlist_add_nvlist(nvl, pool_name, spa->spa_config);
	mutex_exit(&spa->spa_props_lock);
	}

	error = spa_config_write(dp, nvl);
	if (error != 0)
	ccw_failure = B_TRUE;
	nvlist_free(nvl);
	}

	if (ccw_failure) {
	/*
	* Keep trying so that configuration data is
	* written if/when any temporary filesystem
	* resource issues are resolved.
	*/
	if (target->spa_ccw_fail_time == 0) {
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_CONFIG_CACHE_WRITE,
	target, NULL, NULL, NULL, 0);
	}
	target->spa_ccw_fail_time = gethrtime();
	spa_async_request(target, SPA_ASYNC_CONFIG_UPDATE);
	} else {
	/*
	* Do not rate limit future attempts to update
	* the config cache.
	*/
	target->spa_ccw_fail_time = 0;
	}

	/*
	* Remove any config entries older than the current one.
	*/
	dp = list_head(&target->spa_config_list);
	while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
	list_remove(&target->spa_config_list, tdp);
	if (tdp->scd_path != NULL)
	spa_strfree(tdp->scd_path);
	kmem_free(tdp, sizeof (spa_config_dirent_t));
	}

	spa_config_generation++;

	if (postsysevent)
	spa_event_notify(target, NULL, NULL, ESC_ZFS_CONFIG_SYNC);
	+
	+ /*
	+ * Post udev event to sync blkid information if the pool is created
	+ * or a new vdev is added to the pool.
	+ */
	+ if ((target->spa_root_vdev) && postblkidevent) {
	+ vdev_post_kobj_evt(target->spa_root_vdev);
	+ for (int i = 0; i < target->spa_l2cache.sav_count; i++)
	+ vdev_post_kobj_evt(target->spa_l2cache.sav_vdevs[i]);
	+ }
	}

	/*
	* Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
	* and we don't want to allow the local zone to see all the pools anyway.
	* So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
	* information for all pool visible within the zone.
	*/
	nvlist_t *
	spa_all_configs(uint64_t *generation)
	{
	nvlist_t *pools;
	spa_t *spa = NULL;

	if (*generation == spa_config_generation)
	return (NULL);

	pools = fnvlist_alloc();

	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL) {
	if (INGLOBALZONE(curproc) \|\|
	zone_dataset_visible(spa_name(spa), NULL)) {
	mutex_enter(&spa->spa_props_lock);
	fnvlist_add_nvlist(pools, spa_name(spa),
	spa->spa_config);
	mutex_exit(&spa->spa_props_lock);
	}
	}
	*generation = spa_config_generation;
	mutex_exit(&spa_namespace_lock);

	return (pools);
	}

	void
	spa_config_set(spa_t spa, nvlist_t config)
	{
	mutex_enter(&spa->spa_props_lock);
	if (spa->spa_config != NULL && spa->spa_config != config)
	nvlist_free(spa->spa_config);
	spa->spa_config = config;
	mutex_exit(&spa->spa_props_lock);
	}

	/*
	* Generate the pool's configuration based on the current in-core state.
	*
	* We infer whether to generate a complete config or just one top-level config
	* based on whether vd is the root vdev.
	*/
	nvlist_t *
	spa_config_generate(spa_t spa, vdev_t vd, uint64_t txg, int getstats)
	{
	nvlist_t config, nvroot;
	vdev_t *rvd = spa->spa_root_vdev;
	unsigned long hostid = 0;
	boolean_t locked = B_FALSE;
	uint64_t split_guid;
	char *pool_name;

	if (vd == NULL) {
	vd = rvd;
	locked = B_TRUE;
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	}

	ASSERT(spa_config_held(spa, SCL_CONFIG \| SCL_STATE, RW_READER) ==
	(SCL_CONFIG \| SCL_STATE));

	/*
	* If txg is -1, report the current value of spa->spa_config_txg.
	*/
	if (txg == -1ULL)
	txg = spa->spa_config_txg;

	/*
	* Originally, users had to handle spa namespace collisions by either
	* exporting the already imported pool or by specifying a new name for
	* the pool with a conflicting name. In the case of root pools from
	* virtual guests, neither approach to collision resolution is
	* reasonable. This is addressed by extending the new name syntax with
	* an option to specify that the new name is temporary. When specified,
	* ZFS_IMPORT_TEMP_NAME will be set in spa->spa_import_flags to tell us
	* to use the previous name, which we do below.
	*/
	if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) {
	VERIFY0(nvlist_lookup_string(spa->spa_config,
	ZPOOL_CONFIG_POOL_NAME, &pool_name));
	} else
	pool_name = spa_name(spa);

	config = fnvlist_alloc();

	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, pool_name);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, spa_state(spa));
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, txg);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID, spa_guid(spa));
	fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA, spa->spa_errata);
	if (spa->spa_comment != NULL)
	fnvlist_add_string(config, ZPOOL_CONFIG_COMMENT,
	spa->spa_comment);
	if (spa->spa_compatibility != NULL)
	fnvlist_add_string(config, ZPOOL_CONFIG_COMPATIBILITY,
	spa->spa_compatibility);

	hostid = spa_get_hostid(spa);
	if (hostid != 0)
	fnvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, hostid);
	fnvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, utsname()->nodename);

	int config_gen_flags = 0;
	if (vd != rvd) {
	fnvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
	vd->vdev_top->vdev_guid);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
	vd->vdev_guid);
	if (vd->vdev_isspare)
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_IS_SPARE, 1ULL);
	if (vd->vdev_islog)
	fnvlist_add_uint64(config,
	ZPOOL_CONFIG_IS_LOG, 1ULL);
	vd = vd->vdev_top; /* label contains top config */
	} else {
	/*
	* Only add the (potentially large) split information
	* in the mos config, and not in the vdev labels
	*/
	if (spa->spa_config_splitting != NULL)
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
	spa->spa_config_splitting);

	fnvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS);

	config_gen_flags \|= VDEV_CONFIG_MOS;
	}

	/*
	* Add the top-level config. We even add this on pools which
	* don't support holes in the namespace.
	*/
	vdev_top_config_generate(spa, config);

	/*
	* If we're splitting, record the original pool's guid.
	*/
	if (spa->spa_config_splitting != NULL &&
	nvlist_lookup_uint64(spa->spa_config_splitting,
	ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
	fnvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID, split_guid);
	}

	nvroot = vdev_config_generate(spa, vd, getstats, config_gen_flags);
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot);
	nvlist_free(nvroot);

	/*
	* Store what's necessary for reading the MOS in the label.
	*/
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
	spa->spa_label_features);

	if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
	ddt_histogram_t *ddh;
	ddt_stat_t *dds;
	ddt_object_t *ddo;

	ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
	ddt_get_dedup_histogram(spa, ddh);
	fnvlist_add_uint64_array(config,
	ZPOOL_CONFIG_DDT_HISTOGRAM,
	(uint64_t )ddh, sizeof (ddh) / sizeof (uint64_t));
	kmem_free(ddh, sizeof (ddt_histogram_t));

	ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
	ddt_get_dedup_object_stats(spa, ddo);
	fnvlist_add_uint64_array(config,
	ZPOOL_CONFIG_DDT_OBJ_STATS,
	(uint64_t )ddo, sizeof (ddo) / sizeof (uint64_t));
	kmem_free(ddo, sizeof (ddt_object_t));

	dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
	ddt_get_dedup_stats(spa, dds);
	fnvlist_add_uint64_array(config,
	ZPOOL_CONFIG_DDT_STATS,
	(uint64_t )dds, sizeof (dds) / sizeof (uint64_t));
	kmem_free(dds, sizeof (ddt_stat_t));
	}

	if (locked)
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);

	return (config);
	}

	/*
	* Update all disk labels, generate a fresh config based on the current
	* in-core state, and sync the global config cache (do not sync the config
	* cache if this is a booting rootpool).
	*/
	void
	spa_config_update(spa_t *spa, int what)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t txg;
	int c;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	txg = spa_last_synced_txg(spa) + 1;
	if (what == SPA_CONFIG_UPDATE_POOL) {
	vdev_config_dirty(rvd);
	} else {
	/*
	* If we have top-level vdevs that were added but have
	* not yet been prepared for allocation, do that now.
	* (It's safe now because the config cache is up to date,
	* so it will be able to translate the new DVAs.)
	* See comments in spa_vdev_add() for full details.
	*/
	for (c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	/*
	* Explicitly skip vdevs that are indirect or
	* log vdevs that are being removed. The reason
	* is that both of those can have vdev_ms_array
	* set to 0 and we wouldn't want to change their
	* metaslab size nor call vdev_expand() on them.
	*/
	if (!vdev_is_concrete(tvd) \|\|
	(tvd->vdev_islog && tvd->vdev_removing))
	continue;

	if (tvd->vdev_ms_array == 0)
	vdev_metaslab_set_size(tvd);
	vdev_expand(tvd, txg);
	}
	}
	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* Wait for the mosconfig to be regenerated and synced.
	*/
	txg_wait_synced(spa->spa_dsl_pool, txg);

	/*
	* Update the global config cache to reflect the new mosconfig.
	*/
	if (!spa->spa_is_root) {
	spa_write_cachefile(spa, B_FALSE,
	+ what != SPA_CONFIG_UPDATE_POOL,
	what != SPA_CONFIG_UPDATE_POOL);
	}

	if (what == SPA_CONFIG_UPDATE_POOL)
	spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
	}

	EXPORT_SYMBOL(spa_config_load);
	EXPORT_SYMBOL(spa_all_configs);
	EXPORT_SYMBOL(spa_config_set);
	EXPORT_SYMBOL(spa_config_generate);
	EXPORT_SYMBOL(spa_config_update);

	#ifdef __linux__
	/* string sysctls require a char array on FreeBSD */
	ZFS_MODULE_PARAM(zfs_spa, spa_, config_path, STRING, ZMOD_RD,
	"SPA config file (/etc/zfs/zpool.cache)");
	#endif

	ZFS_MODULE_PARAM(zfs, zfs_, autoimport_disable, INT, ZMOD_RW,
	"Disable pool import at module load");
	diff --git a/sys/contrib/openzfs/module/zfs/spa_errlog.c b/sys/contrib/openzfs/module/zfs/spa_errlog.c
	index e682f6c69402..30e1249dd3b0 100644
	--- a/sys/contrib/openzfs/module/zfs/spa_errlog.c
	+++ b/sys/contrib/openzfs/module/zfs/spa_errlog.c
	@@ -1,1373 +1,1373 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2014, Delphix. All rights reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2021, 2022, George Amanakis. All rights reserved.
	*/

	/*
	* Routines to manage the on-disk persistent error log.
	*
	* Each pool stores a log of all logical data errors seen during normal
	* operation. This is actually the union of two distinct logs: the last log,
	* and the current log. All errors seen are logged to the current log. When a
	* scrub completes, the current log becomes the last log, the last log is thrown
	* out, and the current log is reinitialized. This way, if an error is somehow
	* corrected, a new scrub will show that it no longer exists, and will be
	* deleted from the log when the scrub completes.
	*
	* The log is stored using a ZAP object whose key is a string form of the
	* zbookmark_phys tuple (objset, object, level, blkid), and whose contents is an
	* optional 'objset:object' human-readable string describing the data. When an
	* error is first logged, this string will be empty, indicating that no name is
	* known. This prevents us from having to issue a potentially large amount of
	* I/O to discover the object name during an error path. Instead, we do the
	* calculation when the data is requested, storing the result so future queries
	* will be faster.
	*
	* If the head_errlog feature is enabled, a different on-disk format is used.
	* The error log of each head dataset is stored separately in the zap object
	* and keyed by the head id. This enables listing every dataset affected in
	* userland. In order to be able to track whether an error block has been
	* modified or added to snapshots since it was marked as an error, a new tuple
	* is introduced: zbookmark_err_phys_t. It allows the storage of the birth
	* transaction group of an error block on-disk. The birth transaction group is
	* used by check_filesystem() to assess whether this block was freed,
	* re-written or added to a snapshot since its marking as an error.
	*
	* This log is then shipped into an nvlist where the key is the dataset name and
	* the value is the object name. Userland is then responsible for uniquifying
	* this list and displaying it to the user.
	*/

	#include <sys/dmu_tx.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/zap.h>
	#include <sys/zio.h>
	#include <sys/dsl_dir.h>
	#include <sys/dmu_objset.h>
	#include <sys/dbuf.h>
	#include <sys/zfs_znode.h>

	#define NAME_MAX_LEN 64

	/*
	* spa_upgrade_errlog_limit : A zfs module parameter that controls the number
	* of on-disk error log entries that will be converted to the new
	* format when enabling head_errlog. Defaults to 0 which converts
	* all log entries.
	*/
	-static uint32_t spa_upgrade_errlog_limit = 0;
	+static uint_t spa_upgrade_errlog_limit = 0;

	/*
	* Convert a bookmark to a string.
	*/
	static void
	bookmark_to_name(zbookmark_phys_t zb, char buf, size_t len)
	{
	(void) snprintf(buf, len, "%llx:%llx:%llx:%llx",
	(u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object,
	(u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid);
	}

	/*
	* Convert an err_phys to a string.
	*/
	static void
	errphys_to_name(zbookmark_err_phys_t zep, char buf, size_t len)
	{
	(void) snprintf(buf, len, "%llx:%llx:%llx:%llx",
	(u_longlong_t)zep->zb_object, (u_longlong_t)zep->zb_level,
	(u_longlong_t)zep->zb_blkid, (u_longlong_t)zep->zb_birth);
	}

	/*
	* Convert a string to a err_phys.
	*/
	static void
	name_to_errphys(char buf, zbookmark_err_phys_t zep)
	{
	zep->zb_object = zfs_strtonum(buf, &buf);
	ASSERT(*buf == ':');
	zep->zb_level = (int)zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == ':');
	zep->zb_blkid = zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == ':');
	zep->zb_birth = zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == '\0');
	}

	/*
	* Convert a string to a bookmark.
	*/
	static void
	name_to_bookmark(char buf, zbookmark_phys_t zb)
	{
	zb->zb_objset = zfs_strtonum(buf, &buf);
	ASSERT(*buf == ':');
	zb->zb_object = zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == ':');
	zb->zb_level = (int)zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == ':');
	zb->zb_blkid = zfs_strtonum(buf + 1, &buf);
	ASSERT(*buf == '\0');
	}

	#ifdef _KERNEL
	static void
	zep_to_zb(uint64_t dataset, zbookmark_err_phys_t zep, zbookmark_phys_t zb)
	{
	zb->zb_objset = dataset;
	zb->zb_object = zep->zb_object;
	zb->zb_level = zep->zb_level;
	zb->zb_blkid = zep->zb_blkid;
	}
	#endif

	static void
	name_to_object(char buf, uint64_t obj)
	{
	*obj = zfs_strtonum(buf, &buf);
	ASSERT(*buf == '\0');
	}

	static int
	get_head_and_birth_txg(spa_t spa, zbookmark_err_phys_t zep, uint64_t ds_obj,
	uint64_t *head_dataset_id)
	{
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dsl_dataset_t *ds;
	objset_t *os;

	dsl_pool_config_enter(dp, FTAG);
	int error = dsl_dataset_hold_obj(dp, ds_obj, FTAG, &ds);
	if (error != 0) {
	dsl_pool_config_exit(dp, FTAG);
	return (error);
	}
	ASSERT(head_dataset_id);
	*head_dataset_id = dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj;

	error = dmu_objset_from_ds(ds, &os);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	return (error);
	}

	/*
	* If the key is not loaded dbuf_dnode_findbp() will error out with
	* EACCES. However in that case dnode_hold() will eventually call
	* dbuf_read()->zio_wait() which may call spa_log_error(). This will
	* lead to a deadlock due to us holding the mutex spa_errlist_lock.
	* Avoid this by checking here if the keys are loaded, if not return.
	* If the keys are not loaded the head_errlog feature is meaningless
	* as we cannot figure out the birth txg of the block pointer.
	*/
	if (dsl_dataset_get_keystatus(ds->ds_dir) ==
	ZFS_KEYSTATUS_UNAVAILABLE) {
	zep->zb_birth = 0;
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	return (0);
	}

	dnode_t *dn;
	blkptr_t bp;

	error = dnode_hold(os, zep->zb_object, FTAG, &dn);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	return (error);
	}

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	error = dbuf_dnode_findbp(dn, zep->zb_level, zep->zb_blkid, &bp, NULL,
	NULL);
	if (error == 0 && BP_IS_HOLE(&bp))
	error = SET_ERROR(ENOENT);

	/*
	* If the key is loaded but the encrypted filesystem is unmounted when
	* a scrub is run, then dbuf_dnode_findbp() will still error out with
	* EACCES (possibly due to the key mapping being removed upon
	* unmounting). In that case the head_errlog feature is also
	* meaningless as we cannot figure out the birth txg of the block
	* pointer.
	*/
	if (error == EACCES)
	error = 0;
	else if (!error)
	zep->zb_birth = bp.blk_birth;

	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	return (error);
	}

	/*
	* Log an uncorrectable error to the persistent error log. We add it to the
	* spa's list of pending errors. The changes are actually synced out to disk
	* during spa_errlog_sync().
	*/
	void
	spa_log_error(spa_t spa, const zbookmark_phys_t zb)
	{
	spa_error_entry_t search;
	spa_error_entry_t *new;
	avl_tree_t *tree;
	avl_index_t where;

	/*
	* If we are trying to import a pool, ignore any errors, as we won't be
	* writing to the pool any time soon.
	*/
	if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT)
	return;

	mutex_enter(&spa->spa_errlist_lock);

	/*
	* If we have had a request to rotate the log, log it to the next list
	* instead of the current one.
	*/
	if (spa->spa_scrub_active \|\| spa->spa_scrub_finished)
	tree = &spa->spa_errlist_scrub;
	else
	tree = &spa->spa_errlist_last;

	search.se_bookmark = *zb;
	if (avl_find(tree, &search, &where) != NULL) {
	mutex_exit(&spa->spa_errlist_lock);
	return;
	}

	new = kmem_zalloc(sizeof (spa_error_entry_t), KM_SLEEP);
	new->se_bookmark = *zb;
	avl_insert(tree, new, where);

	mutex_exit(&spa->spa_errlist_lock);
	}

	#ifdef _KERNEL
	static int
	find_birth_txg(dsl_dataset_t ds, zbookmark_err_phys_t zep,
	uint64_t *birth_txg)
	{
	objset_t *os;
	int error = dmu_objset_from_ds(ds, &os);
	if (error != 0)
	return (error);

	dnode_t *dn;
	blkptr_t bp;

	error = dnode_hold(os, zep->zb_object, FTAG, &dn);
	if (error != 0)
	return (error);

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	error = dbuf_dnode_findbp(dn, zep->zb_level, zep->zb_blkid, &bp, NULL,
	NULL);
	if (error == 0 && BP_IS_HOLE(&bp))
	error = SET_ERROR(ENOENT);

	*birth_txg = bp.blk_birth;
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	return (error);
	}

	/*
	* This function serves a double role. If only_count is true, it returns
	* (in *count) how many times an error block belonging to this filesystem is
	* referenced by snapshots or clones. If only_count is false, each time the
	* error block is referenced by a snapshot or clone, it fills the userspace
	* array at uaddr with the bookmarks of the error blocks. The array is filled
	* from the back and *count is modified to be the number of unused entries at
	* the beginning of the array.
	*/
	static int
	check_filesystem(spa_t spa, uint64_t head_ds, zbookmark_err_phys_t zep,
	uint64_t count, void uaddr, boolean_t only_count)
	{
	dsl_dataset_t *ds;
	dsl_pool_t *dp = spa->spa_dsl_pool;

	int error = dsl_dataset_hold_obj(dp, head_ds, FTAG, &ds);
	if (error != 0)
	return (error);

	uint64_t latest_txg;
	uint64_t txg_to_consider = spa->spa_syncing_txg;
	boolean_t check_snapshot = B_TRUE;
	error = find_birth_txg(ds, zep, &latest_txg);

	/*
	* If we cannot figure out the current birth txg of the block pointer
	* error out. If the filesystem is encrypted and the key is not loaded
	* or the encrypted filesystem is not mounted the error will be EACCES.
	* In that case do not return an error.
	*/
	if (error == EACCES) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}
	if (error) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	if (zep->zb_birth == latest_txg) {
	/* Block neither free nor rewritten. */
	if (!only_count) {
	zbookmark_phys_t zb;
	zep_to_zb(head_ds, zep, &zb);
	if (copyout(&zb, (char )uaddr + (count - 1)
	* sizeof (zbookmark_phys_t),
	sizeof (zbookmark_phys_t)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EFAULT));
	}
	(*count)--;
	} else {
	(*count)++;
	}
	check_snapshot = B_FALSE;
	} else {
	ASSERT3U(zep->zb_birth, <, latest_txg);
	txg_to_consider = latest_txg;
	}

	/* How many snapshots reference this block. */
	uint64_t snap_count;
	error = zap_count(spa->spa_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj, &snap_count);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (snap_count == 0) {
	/* File system has no snapshot. */
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	uint64_t snap_obj_array = kmem_alloc(snap_count sizeof (uint64_t),
	KM_SLEEP);

	int aff_snap_count = 0;
	uint64_t snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	uint64_t snap_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;

	/* Check only snapshots created from this file system. */
	while (snap_obj != 0 && zep->zb_birth < snap_obj_txg &&
	snap_obj_txg <= txg_to_consider) {

	dsl_dataset_rele(ds, FTAG);
	error = dsl_dataset_hold_obj(dp, snap_obj, FTAG, &ds);
	if (error != 0)
	goto out;

	if (dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj != head_ds)
	break;

	boolean_t affected = B_TRUE;
	if (check_snapshot) {
	uint64_t blk_txg;
	error = find_birth_txg(ds, zep, &blk_txg);
	affected = (error == 0 && zep->zb_birth == blk_txg);
	}

	if (affected) {
	snap_obj_array[aff_snap_count] = snap_obj;
	aff_snap_count++;

	if (!only_count) {
	zbookmark_phys_t zb;
	zep_to_zb(snap_obj, zep, &zb);
	if (copyout(&zb, (char )uaddr + (count - 1) *
	sizeof (zbookmark_phys_t),
	sizeof (zbookmark_phys_t)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	error = SET_ERROR(EFAULT);
	goto out;
	}
	(*count)--;
	} else {
	(*count)++;
	}

	/*
	* Only clones whose origins were affected could also
	* have affected snapshots.
	*/
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	dsl_dataset_phys(ds)->ds_next_clones_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	error = check_filesystem(spa,
	za.za_first_integer, zep,
	count, uaddr, only_count);

	if (error != 0) {
	zap_cursor_fini(&zc);
	goto out;
	}
	}
	zap_cursor_fini(&zc);
	}
	snap_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
	snap_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	}
	dsl_dataset_rele(ds, FTAG);

	out:
	kmem_free(snap_obj_array, sizeof (*snap_obj_array));
	return (error);
	}

	static int
	find_top_affected_fs(spa_t spa, uint64_t head_ds, zbookmark_err_phys_t zep,
	uint64_t *top_affected_fs)
	{
	uint64_t oldest_dsobj;
	int error = dsl_dataset_oldest_snapshot(spa, head_ds, zep->zb_birth,
	&oldest_dsobj);
	if (error != 0)
	return (error);

	dsl_dataset_t *ds;
	error = dsl_dataset_hold_obj(spa->spa_dsl_pool, oldest_dsobj,
	FTAG, &ds);
	if (error != 0)
	return (error);

	*top_affected_fs =
	dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj;
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	static int
	process_error_block(spa_t spa, uint64_t head_ds, zbookmark_err_phys_t zep,
	uint64_t count, void uaddr, boolean_t only_count)
	{
	dsl_pool_t *dp = spa->spa_dsl_pool;
	uint64_t top_affected_fs;

	/*
	* If the zb_birth is 0 it means we failed to retrieve the birth txg
	* of the block pointer. This happens when an encrypted filesystem is
	* not mounted or when the key is not loaded. Do not proceed to
	* check_filesystem(), instead do the accounting here.
	*/
	if (zep->zb_birth == 0) {
	if (!only_count) {
	zbookmark_phys_t zb;
	zep_to_zb(head_ds, zep, &zb);
	if (copyout(&zb, (char )uaddr + (count - 1)
	* sizeof (zbookmark_phys_t),
	sizeof (zbookmark_phys_t)) != 0) {
	return (SET_ERROR(EFAULT));
	}
	(*count)--;
	} else {
	(*count)++;
	}
	return (0);
	}

	dsl_pool_config_enter(dp, FTAG);
	int error = find_top_affected_fs(spa, head_ds, zep, &top_affected_fs);
	if (error == 0)
	error = check_filesystem(spa, top_affected_fs, zep, count,
	uaddr, only_count);

	dsl_pool_config_exit(dp, FTAG);
	return (error);
	}

	static uint64_t
	get_errlog_size(spa_t *spa, uint64_t spa_err_obj)
	{
	if (spa_err_obj == 0)
	return (0);
	uint64_t total = 0;

	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj);
	zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) {

	zap_cursor_t head_ds_cursor;
	zap_attribute_t head_ds_attr;
	zbookmark_err_phys_t head_ds_block;

	uint64_t head_ds;
	name_to_object(za.za_name, &head_ds);

	for (zap_cursor_init(&head_ds_cursor, spa->spa_meta_objset,
	za.za_first_integer); zap_cursor_retrieve(&head_ds_cursor,
	&head_ds_attr) == 0; zap_cursor_advance(&head_ds_cursor)) {

	name_to_errphys(head_ds_attr.za_name, &head_ds_block);
	(void) process_error_block(spa, head_ds, &head_ds_block,
	&total, NULL, B_TRUE);
	}
	zap_cursor_fini(&head_ds_cursor);
	}
	zap_cursor_fini(&zc);
	return (total);
	}

	static uint64_t
	get_errlist_size(spa_t spa, avl_tree_t tree)
	{
	if (avl_numnodes(tree) == 0)
	return (0);
	uint64_t total = 0;

	spa_error_entry_t *se;
	for (se = avl_first(tree); se != NULL; se = AVL_NEXT(tree, se)) {
	zbookmark_err_phys_t zep;
	zep.zb_object = se->se_bookmark.zb_object;
	zep.zb_level = se->se_bookmark.zb_level;
	zep.zb_blkid = se->se_bookmark.zb_blkid;
	zep.zb_birth = 0;

	/*
	* If we cannot find out the head dataset and birth txg of
	* the present error block, we opt not to error out. In the
	* next pool sync this information will be retrieved by
	* sync_error_list() and written to the on-disk error log.
	*/
	uint64_t head_ds_obj;
	int error = get_head_and_birth_txg(spa, &zep,
	se->se_bookmark.zb_objset, &head_ds_obj);

	if (!error)
	(void) process_error_block(spa, head_ds_obj, &zep,
	&total, NULL, B_TRUE);
	}
	return (total);
	}
	#endif

	/*
	* If a healed bookmark matches an entry in the error log we stash it in a tree
	* so that we can later remove the related log entries in sync context.
	*/
	static void
	spa_add_healed_error(spa_t spa, uint64_t obj, zbookmark_phys_t healed_zb)
	{
	char name[NAME_MAX_LEN];

	if (obj == 0)
	return;

	bookmark_to_name(healed_zb, name, sizeof (name));
	mutex_enter(&spa->spa_errlog_lock);
	if (zap_contains(spa->spa_meta_objset, obj, name) == 0) {
	/*
	* Found an error matching healed zb, add zb to our
	* tree of healed errors
	*/
	avl_tree_t *tree = &spa->spa_errlist_healed;
	spa_error_entry_t search;
	spa_error_entry_t *new;
	avl_index_t where;
	search.se_bookmark = *healed_zb;
	mutex_enter(&spa->spa_errlist_lock);
	if (avl_find(tree, &search, &where) != NULL) {
	mutex_exit(&spa->spa_errlist_lock);
	mutex_exit(&spa->spa_errlog_lock);
	return;
	}
	new = kmem_zalloc(sizeof (spa_error_entry_t), KM_SLEEP);
	new->se_bookmark = *healed_zb;
	avl_insert(tree, new, where);
	mutex_exit(&spa->spa_errlist_lock);
	}
	mutex_exit(&spa->spa_errlog_lock);
	}

	/*
	* If this error exists in the given tree remove it.
	*/
	static void
	remove_error_from_list(spa_t spa, avl_tree_t t, const zbookmark_phys_t *zb)
	{
	spa_error_entry_t search, *found;
	avl_index_t where;

	mutex_enter(&spa->spa_errlist_lock);
	search.se_bookmark = *zb;
	if ((found = avl_find(t, &search, &where)) != NULL) {
	avl_remove(t, found);
	kmem_free(found, sizeof (spa_error_entry_t));
	}
	mutex_exit(&spa->spa_errlist_lock);
	}


	/*
	* Removes all of the recv healed errors from both on-disk error logs
	*/
	static void
	spa_remove_healed_errors(spa_t spa, avl_tree_t s, avl_tree_t l, dmu_tx_t tx)
	{
	char name[NAME_MAX_LEN];
	spa_error_entry_t *se;
	void *cookie = NULL;

	ASSERT(MUTEX_HELD(&spa->spa_errlog_lock));

	while ((se = avl_destroy_nodes(&spa->spa_errlist_healed,
	&cookie)) != NULL) {
	remove_error_from_list(spa, s, &se->se_bookmark);
	remove_error_from_list(spa, l, &se->se_bookmark);
	bookmark_to_name(&se->se_bookmark, name, sizeof (name));
	kmem_free(se, sizeof (spa_error_entry_t));
	(void) zap_remove(spa->spa_meta_objset,
	spa->spa_errlog_last, name, tx);
	(void) zap_remove(spa->spa_meta_objset,
	spa->spa_errlog_scrub, name, tx);
	}
	}

	/*
	* Stash away healed bookmarks to remove them from the on-disk error logs
	* later in spa_remove_healed_errors().
	*/
	void
	spa_remove_error(spa_t spa, zbookmark_phys_t zb)
	{
	char name[NAME_MAX_LEN];

	bookmark_to_name(zb, name, sizeof (name));

	spa_add_healed_error(spa, spa->spa_errlog_last, zb);
	spa_add_healed_error(spa, spa->spa_errlog_scrub, zb);
	}

	/*
	* Return the number of errors currently in the error log. This is actually the
	* sum of both the last log and the current log, since we don't know the union
	* of these logs until we reach userland.
	*/
	uint64_t
	spa_get_errlog_size(spa_t *spa)
	{
	uint64_t total = 0;

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	mutex_enter(&spa->spa_errlog_lock);
	uint64_t count;
	if (spa->spa_errlog_scrub != 0 &&
	zap_count(spa->spa_meta_objset, spa->spa_errlog_scrub,
	&count) == 0)
	total += count;

	if (spa->spa_errlog_last != 0 && !spa->spa_scrub_finished &&
	zap_count(spa->spa_meta_objset, spa->spa_errlog_last,
	&count) == 0)
	total += count;
	mutex_exit(&spa->spa_errlog_lock);

	mutex_enter(&spa->spa_errlist_lock);
	total += avl_numnodes(&spa->spa_errlist_last);
	total += avl_numnodes(&spa->spa_errlist_scrub);
	mutex_exit(&spa->spa_errlist_lock);
	} else {
	#ifdef _KERNEL
	mutex_enter(&spa->spa_errlog_lock);
	total += get_errlog_size(spa, spa->spa_errlog_last);
	total += get_errlog_size(spa, spa->spa_errlog_scrub);
	mutex_exit(&spa->spa_errlog_lock);

	mutex_enter(&spa->spa_errlist_lock);
	total += get_errlist_size(spa, &spa->spa_errlist_last);
	total += get_errlist_size(spa, &spa->spa_errlist_scrub);
	mutex_exit(&spa->spa_errlist_lock);
	#endif
	}
	return (total);
	}

	/*
	* This function sweeps through an on-disk error log and stores all bookmarks
	* as error bookmarks in a new ZAP object. At the end we discard the old one,
	* and spa_update_errlog() will set the spa's on-disk error log to new ZAP
	* object.
	*/
	static void
	sync_upgrade_errlog(spa_t spa, uint64_t spa_err_obj, uint64_t newobj,
	dmu_tx_t *tx)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	zbookmark_phys_t zb;
	uint64_t count;

	*newobj = zap_create(spa->spa_meta_objset, DMU_OT_ERROR_LOG,
	DMU_OT_NONE, 0, tx);

	/*
	* If we cannnot perform the upgrade we should clear the old on-disk
	* error logs.
	*/
	if (zap_count(spa->spa_meta_objset, spa_err_obj, &count) != 0) {
	VERIFY0(dmu_object_free(spa->spa_meta_objset, spa_err_obj, tx));
	return;
	}

	for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	if (spa_upgrade_errlog_limit != 0 &&
	zc.zc_cd == spa_upgrade_errlog_limit)
	break;

	name_to_bookmark(za.za_name, &zb);

	zbookmark_err_phys_t zep;
	zep.zb_object = zb.zb_object;
	zep.zb_level = zb.zb_level;
	zep.zb_blkid = zb.zb_blkid;
	zep.zb_birth = 0;

	/*
	* We cannot use get_head_and_birth_txg() because it will
	* acquire the pool config lock, which we already have. In case
	* of an error we simply continue.
	*/
	uint64_t head_dataset_obj;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dsl_dataset_t *ds;
	objset_t *os;

	int error = dsl_dataset_hold_obj(dp, zb.zb_objset, FTAG, &ds);
	if (error != 0)
	continue;

	head_dataset_obj =
	dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj;

	/*
	* The objset and the dnode are required for getting the block
	* pointer, which is used to determine if BP_IS_HOLE(). If
	* getting the objset or the dnode fails, do not create a
	* zap entry (presuming we know the dataset) as this may create
	* spurious errors that we cannot ever resolve. If an error is
	* truly persistent, it should re-appear after a scan.
	*/
	if (dmu_objset_from_ds(ds, &os) != 0) {
	dsl_dataset_rele(ds, FTAG);
	continue;
	}

	dnode_t *dn;
	blkptr_t bp;

	if (dnode_hold(os, zep.zb_object, FTAG, &dn) != 0) {
	dsl_dataset_rele(ds, FTAG);
	continue;
	}

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	error = dbuf_dnode_findbp(dn, zep.zb_level, zep.zb_blkid, &bp,
	NULL, NULL);
	if (error == EACCES)
	error = 0;
	else if (!error)
	zep.zb_birth = bp.blk_birth;

	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	dsl_dataset_rele(ds, FTAG);

	if (error != 0 \|\| BP_IS_HOLE(&bp))
	continue;

	uint64_t err_obj;
	error = zap_lookup_int_key(spa->spa_meta_objset, *newobj,
	head_dataset_obj, &err_obj);

	if (error == ENOENT) {
	err_obj = zap_create(spa->spa_meta_objset,
	DMU_OT_ERROR_LOG, DMU_OT_NONE, 0, tx);

	(void) zap_update_int_key(spa->spa_meta_objset,
	*newobj, head_dataset_obj, err_obj, tx);
	}

	char buf[64];
	errphys_to_name(&zep, buf, sizeof (buf));

	const char *name = "";
	(void) zap_update(spa->spa_meta_objset, err_obj,
	buf, 1, strlen(name) + 1, name, tx);
	}
	zap_cursor_fini(&zc);

	VERIFY0(dmu_object_free(spa->spa_meta_objset, spa_err_obj, tx));
	}

	void
	spa_upgrade_errlog(spa_t spa, dmu_tx_t tx)
	{
	uint64_t newobj = 0;

	mutex_enter(&spa->spa_errlog_lock);
	if (spa->spa_errlog_last != 0) {
	sync_upgrade_errlog(spa, spa->spa_errlog_last, &newobj, tx);
	spa->spa_errlog_last = newobj;
	}

	if (spa->spa_errlog_scrub != 0) {
	sync_upgrade_errlog(spa, spa->spa_errlog_scrub, &newobj, tx);
	spa->spa_errlog_scrub = newobj;
	}
	mutex_exit(&spa->spa_errlog_lock);
	}

	#ifdef _KERNEL
	/*
	* If an error block is shared by two datasets it will be counted twice. For
	* detailed message see spa_get_errlog_size() above.
	*/
	static int
	process_error_log(spa_t spa, uint64_t obj, void uaddr, uint64_t *count)
	{
	zap_cursor_t zc;
	zap_attribute_t za;

	if (obj == 0)
	return (0);

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	for (zap_cursor_init(&zc, spa->spa_meta_objset, obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	if (*count == 0) {
	zap_cursor_fini(&zc);
	return (SET_ERROR(ENOMEM));
	}

	zbookmark_phys_t zb;
	name_to_bookmark(za.za_name, &zb);

	if (copyout(&zb, (char *)uaddr +
	(count - 1) sizeof (zbookmark_phys_t),
	sizeof (zbookmark_phys_t)) != 0) {
	zap_cursor_fini(&zc);
	return (SET_ERROR(EFAULT));
	}
	*count -= 1;

	}
	zap_cursor_fini(&zc);
	return (0);
	}

	for (zap_cursor_init(&zc, spa->spa_meta_objset, obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {

	zap_cursor_t head_ds_cursor;
	zap_attribute_t head_ds_attr;

	uint64_t head_ds_err_obj = za.za_first_integer;
	uint64_t head_ds;
	name_to_object(za.za_name, &head_ds);
	for (zap_cursor_init(&head_ds_cursor, spa->spa_meta_objset,
	head_ds_err_obj); zap_cursor_retrieve(&head_ds_cursor,
	&head_ds_attr) == 0; zap_cursor_advance(&head_ds_cursor)) {

	zbookmark_err_phys_t head_ds_block;
	name_to_errphys(head_ds_attr.za_name, &head_ds_block);
	int error = process_error_block(spa, head_ds,
	&head_ds_block, count, uaddr, B_FALSE);

	if (error != 0) {
	zap_cursor_fini(&head_ds_cursor);
	zap_cursor_fini(&zc);
	return (error);
	}
	}
	zap_cursor_fini(&head_ds_cursor);
	}
	zap_cursor_fini(&zc);
	return (0);
	}

	static int
	process_error_list(spa_t spa, avl_tree_t list, void uaddr, uint64_t count)
	{
	spa_error_entry_t *se;

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	for (se = avl_first(list); se != NULL;
	se = AVL_NEXT(list, se)) {

	if (*count == 0)
	return (SET_ERROR(ENOMEM));

	if (copyout(&se->se_bookmark, (char *)uaddr +
	(count - 1) sizeof (zbookmark_phys_t),
	sizeof (zbookmark_phys_t)) != 0)
	return (SET_ERROR(EFAULT));

	*count -= 1;
	}
	return (0);
	}

	for (se = avl_first(list); se != NULL; se = AVL_NEXT(list, se)) {
	zbookmark_err_phys_t zep;
	zep.zb_object = se->se_bookmark.zb_object;
	zep.zb_level = se->se_bookmark.zb_level;
	zep.zb_blkid = se->se_bookmark.zb_blkid;
	zep.zb_birth = 0;

	uint64_t head_ds_obj;
	int error = get_head_and_birth_txg(spa, &zep,
	se->se_bookmark.zb_objset, &head_ds_obj);

	if (!error)
	error = process_error_block(spa, head_ds_obj, &zep,
	count, uaddr, B_FALSE);
	if (error)
	return (error);
	}
	return (0);
	}
	#endif

	/*
	* Copy all known errors to userland as an array of bookmarks. This is
	* actually a union of the on-disk last log and current log, as well as any
	* pending error requests.
	*
	* Because the act of reading the on-disk log could cause errors to be
	* generated, we have two separate locks: one for the error log and one for the
	* in-core error lists. We only need the error list lock to log and error, so
	* we grab the error log lock while we read the on-disk logs, and only pick up
	* the error list lock when we are finished.
	*/
	int
	spa_get_errlog(spa_t spa, void uaddr, uint64_t *count)
	{
	int ret = 0;

	#ifdef _KERNEL
	mutex_enter(&spa->spa_errlog_lock);

	ret = process_error_log(spa, spa->spa_errlog_scrub, uaddr, count);

	if (!ret && !spa->spa_scrub_finished)
	ret = process_error_log(spa, spa->spa_errlog_last, uaddr,
	count);

	mutex_enter(&spa->spa_errlist_lock);
	if (!ret)
	ret = process_error_list(spa, &spa->spa_errlist_scrub, uaddr,
	count);
	if (!ret)
	ret = process_error_list(spa, &spa->spa_errlist_last, uaddr,
	count);
	mutex_exit(&spa->spa_errlist_lock);

	mutex_exit(&spa->spa_errlog_lock);
	#else
	(void) spa, (void) uaddr, (void) count;
	#endif

	return (ret);
	}

	/*
	* Called when a scrub completes. This simply set a bit which tells which AVL
	* tree to add new errors. spa_errlog_sync() is responsible for actually
	* syncing the changes to the underlying objects.
	*/
	void
	spa_errlog_rotate(spa_t *spa)
	{
	mutex_enter(&spa->spa_errlist_lock);
	spa->spa_scrub_finished = B_TRUE;
	mutex_exit(&spa->spa_errlist_lock);
	}

	/*
	* Discard any pending errors from the spa_t. Called when unloading a faulted
	* pool, as the errors encountered during the open cannot be synced to disk.
	*/
	void
	spa_errlog_drain(spa_t *spa)
	{
	spa_error_entry_t *se;
	void *cookie;

	mutex_enter(&spa->spa_errlist_lock);

	cookie = NULL;
	while ((se = avl_destroy_nodes(&spa->spa_errlist_last,
	&cookie)) != NULL)
	kmem_free(se, sizeof (spa_error_entry_t));
	cookie = NULL;
	while ((se = avl_destroy_nodes(&spa->spa_errlist_scrub,
	&cookie)) != NULL)
	kmem_free(se, sizeof (spa_error_entry_t));

	mutex_exit(&spa->spa_errlist_lock);
	}

	/*
	* Process a list of errors into the current on-disk log.
	*/
	void
	sync_error_list(spa_t spa, avl_tree_t t, uint64_t obj, dmu_tx_t tx)
	{
	spa_error_entry_t *se;
	char buf[NAME_MAX_LEN];
	void *cookie;

	if (avl_numnodes(t) == 0)
	return;

	/* create log if necessary */
	if (*obj == 0)
	*obj = zap_create(spa->spa_meta_objset, DMU_OT_ERROR_LOG,
	DMU_OT_NONE, 0, tx);

	/* add errors to the current log */
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	for (se = avl_first(t); se != NULL; se = AVL_NEXT(t, se)) {
	bookmark_to_name(&se->se_bookmark, buf, sizeof (buf));

	const char *name = se->se_name ? se->se_name : "";
	(void) zap_update(spa->spa_meta_objset, *obj, buf, 1,
	strlen(name) + 1, name, tx);
	}
	} else {
	for (se = avl_first(t); se != NULL; se = AVL_NEXT(t, se)) {
	zbookmark_err_phys_t zep;
	zep.zb_object = se->se_bookmark.zb_object;
	zep.zb_level = se->se_bookmark.zb_level;
	zep.zb_blkid = se->se_bookmark.zb_blkid;
	zep.zb_birth = 0;

	/*
	* If we cannot find out the head dataset and birth txg
	* of the present error block, we simply continue.
	* Reinserting that error block to the error lists,
	* even if we are not syncing the final txg, results
	* in duplicate posting of errors.
	*/
	uint64_t head_dataset_obj;
	int error = get_head_and_birth_txg(spa, &zep,
	se->se_bookmark.zb_objset, &head_dataset_obj);
	if (error)
	continue;

	uint64_t err_obj;
	error = zap_lookup_int_key(spa->spa_meta_objset,
	*obj, head_dataset_obj, &err_obj);

	if (error == ENOENT) {
	err_obj = zap_create(spa->spa_meta_objset,
	DMU_OT_ERROR_LOG, DMU_OT_NONE, 0, tx);

	(void) zap_update_int_key(spa->spa_meta_objset,
	*obj, head_dataset_obj, err_obj, tx);
	}
	errphys_to_name(&zep, buf, sizeof (buf));

	const char *name = se->se_name ? se->se_name : "";
	(void) zap_update(spa->spa_meta_objset,
	err_obj, buf, 1, strlen(name) + 1, name, tx);
	}
	}
	/* purge the error list */
	cookie = NULL;
	while ((se = avl_destroy_nodes(t, &cookie)) != NULL)
	kmem_free(se, sizeof (spa_error_entry_t));
	}

	static void
	delete_errlog(spa_t spa, uint64_t spa_err_obj, dmu_tx_t tx)
	{
	if (spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	VERIFY0(dmu_object_free(spa->spa_meta_objset,
	za.za_first_integer, tx));
	}
	zap_cursor_fini(&zc);
	}
	VERIFY0(dmu_object_free(spa->spa_meta_objset, spa_err_obj, tx));
	}

	/*
	* Sync the error log out to disk. This is a little tricky because the act of
	* writing the error log requires the spa_errlist_lock. So, we need to lock the
	* error lists, take a copy of the lists, and then reinitialize them. Then, we
	* drop the error list lock and take the error log lock, at which point we
	* do the errlog processing. Then, if we encounter an I/O error during this
	* process, we can successfully add the error to the list. Note that this will
	* result in the perpetual recycling of errors, but it is an unlikely situation
	* and not a performance critical operation.
	*/
	void
	spa_errlog_sync(spa_t *spa, uint64_t txg)
	{
	dmu_tx_t *tx;
	avl_tree_t scrub, last;
	int scrub_finished;

	mutex_enter(&spa->spa_errlist_lock);

	/*
	* Bail out early under normal circumstances.
	*/
	if (avl_numnodes(&spa->spa_errlist_scrub) == 0 &&
	avl_numnodes(&spa->spa_errlist_last) == 0 &&
	avl_numnodes(&spa->spa_errlist_healed) == 0 &&
	!spa->spa_scrub_finished) {
	mutex_exit(&spa->spa_errlist_lock);
	return;
	}

	spa_get_errlists(spa, &last, &scrub);
	scrub_finished = spa->spa_scrub_finished;
	spa->spa_scrub_finished = B_FALSE;

	mutex_exit(&spa->spa_errlist_lock);
	mutex_enter(&spa->spa_errlog_lock);

	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);

	/*
	* Remove healed errors from errors.
	*/
	spa_remove_healed_errors(spa, &last, &scrub, tx);

	/*
	* Sync out the current list of errors.
	*/
	sync_error_list(spa, &last, &spa->spa_errlog_last, tx);

	/*
	* Rotate the log if necessary.
	*/
	if (scrub_finished) {
	if (spa->spa_errlog_last != 0)
	delete_errlog(spa, spa->spa_errlog_last, tx);
	spa->spa_errlog_last = spa->spa_errlog_scrub;
	spa->spa_errlog_scrub = 0;

	sync_error_list(spa, &scrub, &spa->spa_errlog_last, tx);
	}

	/*
	* Sync out any pending scrub errors.
	*/
	sync_error_list(spa, &scrub, &spa->spa_errlog_scrub, tx);

	/*
	* Update the MOS to reflect the new values.
	*/
	(void) zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ERRLOG_LAST, sizeof (uint64_t), 1,
	&spa->spa_errlog_last, tx);
	(void) zap_update(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ERRLOG_SCRUB, sizeof (uint64_t), 1,
	&spa->spa_errlog_scrub, tx);

	dmu_tx_commit(tx);

	mutex_exit(&spa->spa_errlog_lock);
	}

	static void
	delete_dataset_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t ds,
	dmu_tx_t *tx)
	{
	if (spa_err_obj == 0)
	return;

	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj);
	zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) {
	uint64_t head_ds;
	name_to_object(za.za_name, &head_ds);
	if (head_ds == ds) {
	(void) zap_remove(spa->spa_meta_objset, spa_err_obj,
	za.za_name, tx);
	VERIFY0(dmu_object_free(spa->spa_meta_objset,
	za.za_first_integer, tx));
	break;
	}
	}
	zap_cursor_fini(&zc);
	}

	void
	spa_delete_dataset_errlog(spa_t spa, uint64_t ds, dmu_tx_t tx)
	{
	mutex_enter(&spa->spa_errlog_lock);
	delete_dataset_errlog(spa, spa->spa_errlog_scrub, ds, tx);
	delete_dataset_errlog(spa, spa->spa_errlog_last, ds, tx);
	mutex_exit(&spa->spa_errlog_lock);
	}

	static int
	find_txg_ancestor_snapshot(spa_t *spa, uint64_t new_head, uint64_t old_head,
	uint64_t *txg)
	{
	dsl_dataset_t *ds;
	dsl_pool_t *dp = spa->spa_dsl_pool;

	int error = dsl_dataset_hold_obj(dp, old_head, FTAG, &ds);
	if (error != 0)
	return (error);

	uint64_t prev_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	uint64_t prev_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;

	while (prev_obj != 0) {
	dsl_dataset_rele(ds, FTAG);
	if ((error = dsl_dataset_hold_obj(dp, prev_obj,
	FTAG, &ds)) == 0 &&
	dsl_dir_phys(ds->ds_dir)->dd_head_dataset_obj == new_head)
	break;

	if (error != 0)
	return (error);

	prev_obj_txg = dsl_dataset_phys(ds)->ds_prev_snap_txg;
	prev_obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
	}
	dsl_dataset_rele(ds, FTAG);
	ASSERT(prev_obj != 0);
	*txg = prev_obj_txg;
	return (0);
	}

	static void
	swap_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t new_head, uint64_t
	old_head, dmu_tx_t *tx)
	{
	if (spa_err_obj == 0)
	return;

	uint64_t old_head_errlog;
	int error = zap_lookup_int_key(spa->spa_meta_objset, spa_err_obj,
	old_head, &old_head_errlog);

	/* If no error log, then there is nothing to do. */
	if (error != 0)
	return;

	uint64_t txg;
	error = find_txg_ancestor_snapshot(spa, new_head, old_head, &txg);
	if (error != 0)
	return;

	/*
	* Create an error log if the file system being promoted does not
	* already have one.
	*/
	uint64_t new_head_errlog;
	error = zap_lookup_int_key(spa->spa_meta_objset, spa_err_obj, new_head,
	&new_head_errlog);

	if (error != 0) {
	new_head_errlog = zap_create(spa->spa_meta_objset,
	DMU_OT_ERROR_LOG, DMU_OT_NONE, 0, tx);

	(void) zap_update_int_key(spa->spa_meta_objset, spa_err_obj,
	new_head, new_head_errlog, tx);
	}

	zap_cursor_t zc;
	zap_attribute_t za;
	zbookmark_err_phys_t err_block;
	for (zap_cursor_init(&zc, spa->spa_meta_objset, old_head_errlog);
	zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) {

	const char *name = "";
	name_to_errphys(za.za_name, &err_block);
	if (err_block.zb_birth < txg) {
	(void) zap_update(spa->spa_meta_objset, new_head_errlog,
	za.za_name, 1, strlen(name) + 1, name, tx);

	(void) zap_remove(spa->spa_meta_objset, old_head_errlog,
	za.za_name, tx);
	}
	}
	zap_cursor_fini(&zc);
	}

	void
	spa_swap_errlog(spa_t *spa, uint64_t new_head_ds, uint64_t old_head_ds,
	dmu_tx_t *tx)
	{
	mutex_enter(&spa->spa_errlog_lock);
	swap_errlog(spa, spa->spa_errlog_scrub, new_head_ds, old_head_ds, tx);
	swap_errlog(spa, spa->spa_errlog_last, new_head_ds, old_head_ds, tx);
	mutex_exit(&spa->spa_errlog_lock);
	}

	#if defined(_KERNEL)
	/* error handling */
	EXPORT_SYMBOL(spa_log_error);
	EXPORT_SYMBOL(spa_get_errlog_size);
	EXPORT_SYMBOL(spa_get_errlog);
	EXPORT_SYMBOL(spa_errlog_rotate);
	EXPORT_SYMBOL(spa_errlog_drain);
	EXPORT_SYMBOL(spa_errlog_sync);
	EXPORT_SYMBOL(spa_get_errlists);
	EXPORT_SYMBOL(spa_delete_dataset_errlog);
	EXPORT_SYMBOL(spa_swap_errlog);
	EXPORT_SYMBOL(sync_error_list);
	EXPORT_SYMBOL(spa_upgrade_errlog);
	#endif

	/* BEGIN CSTYLED */
	-ZFS_MODULE_PARAM(zfs_spa, spa_, upgrade_errlog_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_spa, spa_, upgrade_errlog_limit, UINT, ZMOD_RW,
	"Limit the number of errors which will be upgraded to the new "
	"on-disk error log when enabling head_errlog");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c
	index decf4ddae6af..102070013404 100644
	--- a/sys/contrib/openzfs/module/zfs/spa_misc.c
	+++ b/sys/contrib/openzfs/module/zfs/spa_misc.c
	@@ -1,2959 +1,2959 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2017 Datto Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/zfs_chksum.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/zil.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_file.h>
	#include <sys/vdev_raidz.h>
	#include <sys/metaslab.h>
	#include <sys/uberblock_impl.h>
	#include <sys/txg.h>
	#include <sys/avl.h>
	#include <sys/unique.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/fm/util.h>
	#include <sys/dsl_scan.h>
	#include <sys/fs/zfs.h>
	#include <sys/metaslab_impl.h>
	#include <sys/arc.h>
	#include <sys/ddt.h>
	#include <sys/kstat.h>
	#include "zfs_prop.h"
	#include <sys/btree.h>
	#include <sys/zfeature.h>
	#include <sys/qat.h>
	#include <sys/zstd/zstd.h>

	/*
	* SPA locking
	*
	* There are three basic locks for managing spa_t structures:
	*
	* spa_namespace_lock (global mutex)
	*
	* This lock must be acquired to do any of the following:
	*
	* - Lookup a spa_t by name
	* - Add or remove a spa_t from the namespace
	* - Increase spa_refcount from non-zero
	* - Check if spa_refcount is zero
	* - Rename a spa_t
	* - add/remove/attach/detach devices
	* - Held for the duration of create/destroy/import/export
	*
	* It does not need to handle recursion. A create or destroy may
	* reference objects (files or zvols) in other pools, but by
	* definition they must have an existing reference, and will never need
	* to lookup a spa_t by name.
	*
	* spa_refcount (per-spa zfs_refcount_t protected by mutex)
	*
	* This reference count keep track of any active users of the spa_t. The
	* spa_t cannot be destroyed or freed while this is non-zero. Internally,
	* the refcount is never really 'zero' - opening a pool implicitly keeps
	* some references in the DMU. Internally we check against spa_minref, but
	* present the image of a zero/non-zero value to consumers.
	*
	* spa_config_lock[] (per-spa array of rwlocks)
	*
	* This protects the spa_t from config changes, and must be held in
	* the following circumstances:
	*
	* - RW_READER to perform I/O to the spa
	* - RW_WRITER to change the vdev config
	*
	* The locking order is fairly straightforward:
	*
	* spa_namespace_lock -> spa_refcount
	*
	* The namespace lock must be acquired to increase the refcount from 0
	* or to check if it is zero.
	*
	* spa_refcount -> spa_config_lock[]
	*
	* There must be at least one valid reference on the spa_t to acquire
	* the config lock.
	*
	* spa_namespace_lock -> spa_config_lock[]
	*
	* The namespace lock must always be taken before the config lock.
	*
	*
	* The spa_namespace_lock can be acquired directly and is globally visible.
	*
	* The namespace is manipulated using the following functions, all of which
	* require the spa_namespace_lock to be held.
	*
	* spa_lookup() Lookup a spa_t by name.
	*
	* spa_add() Create a new spa_t in the namespace.
	*
	* spa_remove() Remove a spa_t from the namespace. This also
	* frees up any memory associated with the spa_t.
	*
	* spa_next() Returns the next spa_t in the system, or the
	* first if NULL is passed.
	*
	* spa_evict_all() Shutdown and remove all spa_t structures in
	* the system.
	*
	* spa_guid_exists() Determine whether a pool/device guid exists.
	*
	* The spa_refcount is manipulated using the following functions:
	*
	* spa_open_ref() Adds a reference to the given spa_t. Must be
	* called with spa_namespace_lock held if the
	* refcount is currently zero.
	*
	* spa_close() Remove a reference from the spa_t. This will
	* not free the spa_t or remove it from the
	* namespace. No locking is required.
	*
	* spa_refcount_zero() Returns true if the refcount is currently
	* zero. Must be called with spa_namespace_lock
	* held.
	*
	* The spa_config_lock[] is an array of rwlocks, ordered as follows:
	* SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
	* spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
	*
	* To read the configuration, it suffices to hold one of these locks as reader.
	* To modify the configuration, you must hold all locks as writer. To modify
	* vdev state without altering the vdev tree's topology (e.g. online/offline),
	* you must hold SCL_STATE and SCL_ZIO as writer.
	*
	* We use these distinct config locks to avoid recursive lock entry.
	* For example, spa_sync() (which holds SCL_CONFIG as reader) induces
	* block allocations (SCL_ALLOC), which may require reading space maps
	* from disk (dmu_read() -> zio_read() -> SCL_ZIO).
	*
	* The spa config locks cannot be normal rwlocks because we need the
	* ability to hand off ownership. For example, SCL_ZIO is acquired
	* by the issuing thread and later released by an interrupt thread.
	* They do, however, obey the usual write-wanted semantics to prevent
	* writer (i.e. system administrator) starvation.
	*
	* The lock acquisition rules are as follows:
	*
	* SCL_CONFIG
	* Protects changes to the vdev tree topology, such as vdev
	* add/remove/attach/detach. Protects the dirty config list
	* (spa_config_dirty_list) and the set of spares and l2arc devices.
	*
	* SCL_STATE
	* Protects changes to pool state and vdev state, such as vdev
	* online/offline/fault/degrade/clear. Protects the dirty state list
	* (spa_state_dirty_list) and global pool state (spa_state).
	*
	* SCL_ALLOC
	* Protects changes to metaslab groups and classes.
	* Held as reader by metaslab_alloc() and metaslab_claim().
	*
	* SCL_ZIO
	* Held by bp-level zios (those which have no io_vd upon entry)
	* to prevent changes to the vdev tree. The bp-level zio implicitly
	* protects all of its vdev child zios, which do not hold SCL_ZIO.
	*
	* SCL_FREE
	* Protects changes to metaslab groups and classes.
	* Held as reader by metaslab_free(). SCL_FREE is distinct from
	* SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
	* blocks in zio_done() while another i/o that holds either
	* SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
	*
	* SCL_VDEV
	* Held as reader to prevent changes to the vdev tree during trivial
	* inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the
	* other locks, and lower than all of them, to ensure that it's safe
	* to acquire regardless of caller context.
	*
	* In addition, the following rules apply:
	*
	* (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
	* The lock ordering is SCL_CONFIG > spa_props_lock.
	*
	* (b) I/O operations on leaf vdevs. For any zio operation that takes
	* an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
	* or zio_write_phys() -- the caller must ensure that the config cannot
	* cannot change in the interim, and that the vdev cannot be reopened.
	* SCL_STATE as reader suffices for both.
	*
	* The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
	*
	* spa_vdev_enter() Acquire the namespace lock and the config lock
	* for writing.
	*
	* spa_vdev_exit() Release the config lock, wait for all I/O
	* to complete, sync the updated configs to the
	* cache, and release the namespace lock.
	*
	* vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
	* Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
	* locking is, always, based on spa_namespace_lock and spa_config_lock[].
	*/

	static avl_tree_t spa_namespace_avl;
	kmutex_t spa_namespace_lock;
	static kcondvar_t spa_namespace_cv;
	static const int spa_max_replication_override = SPA_DVAS_PER_BP;

	static kmutex_t spa_spare_lock;
	static avl_tree_t spa_spare_avl;
	static kmutex_t spa_l2cache_lock;
	static avl_tree_t spa_l2cache_avl;

	spa_mode_t spa_mode_global = SPA_MODE_UNINIT;

	#ifdef ZFS_DEBUG
	/*
	* Everything except dprintf, set_error, spa, and indirect_remap is on
	* by default in debug builds.
	*/
	int zfs_flags = ~(ZFS_DEBUG_DPRINTF \| ZFS_DEBUG_SET_ERROR \|
	ZFS_DEBUG_INDIRECT_REMAP);
	#else
	int zfs_flags = 0;
	#endif

	/*
	* zfs_recover can be set to nonzero to attempt to recover from
	* otherwise-fatal errors, typically caused by on-disk corruption. When
	* set, calls to zfs_panic_recover() will turn into warning messages.
	* This should only be used as a last resort, as it typically results
	* in leaked space, or worse.
	*/
	int zfs_recover = B_FALSE;

	/*
	* If destroy encounters an EIO while reading metadata (e.g. indirect
	* blocks), space referenced by the missing metadata can not be freed.
	* Normally this causes the background destroy to become "stalled", as
	* it is unable to make forward progress. While in this stalled state,
	* all remaining space to free from the error-encountering filesystem is
	* "temporarily leaked". Set this flag to cause it to ignore the EIO,
	* permanently leak the space from indirect blocks that can not be read,
	* and continue to free everything else that it can.
	*
	* The default, "stalling" behavior is useful if the storage partially
	* fails (i.e. some but not all i/os fail), and then later recovers. In
	* this case, we will be able to continue pool operations while it is
	* partially failed, and when it recovers, we can continue to free the
	* space, with no leaks. However, note that this case is actually
	* fairly rare.
	*
	* Typically pools either (a) fail completely (but perhaps temporarily,
	* e.g. a top-level vdev going offline), or (b) have localized,
	* permanent errors (e.g. disk returns the wrong data due to bit flip or
	* firmware bug). In case (a), this setting does not matter because the
	* pool will be suspended and the sync thread will not be able to make
	* forward progress regardless. In case (b), because the error is
	* permanent, the best we can do is leak the minimum amount of space,
	* which is what setting this flag will do. Therefore, it is reasonable
	* for this flag to normally be set, but we chose the more conservative
	* approach of not setting it, so that there is no possibility of
	* leaking space in the "partial temporary" failure case.
	*/
	int zfs_free_leak_on_eio = B_FALSE;

	/*
	* Expiration time in milliseconds. This value has two meanings. First it is
	* used to determine when the spa_deadman() logic should fire. By default the
	* spa_deadman() will fire if spa_sync() has not completed in 600 seconds.
	* Secondly, the value determines if an I/O is considered "hung". Any I/O that
	* has not completed in zfs_deadman_synctime_ms is considered "hung" resulting
	* in one of three behaviors controlled by zfs_deadman_failmode.
	*/
	unsigned long zfs_deadman_synctime_ms = 600000UL; /* 10 min. */

	/*
	* This value controls the maximum amount of time zio_wait() will block for an
	* outstanding IO. By default this is 300 seconds at which point the "hung"
	* behavior will be applied as described for zfs_deadman_synctime_ms.
	*/
	unsigned long zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */

	/*
	* Check time in milliseconds. This defines the frequency at which we check
	* for hung I/O.
	*/
	unsigned long zfs_deadman_checktime_ms = 60000UL; /* 1 min. */

	/*
	* By default the deadman is enabled.
	*/
	int zfs_deadman_enabled = B_TRUE;

	/*
	* Controls the behavior of the deadman when it detects a "hung" I/O.
	* Valid values are zfs_deadman_failmode=<wait\|continue\|panic>.
	*
	* wait - Wait for the "hung" I/O (default)
	* continue - Attempt to recover from a "hung" I/O
	* panic - Panic the system
	*/
	const char *zfs_deadman_failmode = "wait";

	/*
	* The worst case is single-sector max-parity RAID-Z blocks, in which
	* case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
	* times the size; so just assume that. Add to this the fact that
	* we can have up to 3 DVAs per bp, and one more factor of 2 because
	* the block may be dittoed with up to 3 DVAs by ddt_sync(). All together,
	* the worst case is:
	* (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24
	*/
	-int spa_asize_inflation = 24;
	+uint_t spa_asize_inflation = 24;

	/*
	* Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
	* the pool to be consumed (bounded by spa_max_slop). This ensures that we
	* don't run the pool completely out of space, due to unaccounted changes (e.g.
	* to the MOS). It also limits the worst-case time to allocate space. If we
	* have less than this amount of free space, most ZPL operations (e.g. write,
	* create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are
	* also part of this 3.2% of space which can't be consumed by normal writes;
	* the slop space "proper" (spa_get_slop_space()) is decreased by the embedded
	* log space.
	*
	* Certain operations (e.g. file removal, most administrative actions) can
	* use half the slop space. They will only return ENOSPC if less than half
	* the slop space is free. Typically, once the pool has less than the slop
	* space free, the user will use these operations to free up space in the pool.
	* These are the operations that call dsl_pool_adjustedsize() with the netfree
	* argument set to TRUE.
	*
	* Operations that are almost guaranteed to free up space in the absence of
	* a pool checkpoint can use up to three quarters of the slop space
	* (e.g zfs destroy).
	*
	* A very restricted set of operations are always permitted, regardless of
	* the amount of free space. These are the operations that call
	* dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net
	* increase in the amount of space used, it is possible to run the pool
	* completely out of space, causing it to be permanently read-only.
	*
	* Note that on very small pools, the slop space will be larger than
	* 3.2%, in an effort to have it be at least spa_min_slop (128MB),
	* but we never allow it to be more than half the pool size.
	*
	* Further, on very large pools, the slop space will be smaller than
	* 3.2%, to avoid reserving much more space than we actually need; bounded
	* by spa_max_slop (128GB).
	*
	* See also the comments in zfs_space_check_t.
	*/
	-int spa_slop_shift = 5;
	+uint_t spa_slop_shift = 5;
	static const uint64_t spa_min_slop = 128ULL * 1024 * 1024;
	static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024;
	static const int spa_allocators = 4;


	void
	spa_load_failed(spa_t spa, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name,
	spa->spa_trust_config ? "trusted" : "untrusted", buf);
	}

	void
	spa_load_note(spa_t spa, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name,
	spa->spa_trust_config ? "trusted" : "untrusted", buf);
	}

	/*
	* By default dedup and user data indirects land in the special class
	*/
	static int zfs_ddt_data_is_special = B_TRUE;
	static int zfs_user_indirect_is_special = B_TRUE;

	/*
	* The percentage of special class final space reserved for metadata only.
	* Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only
	* let metadata into the class.
	*/
	-static int zfs_special_class_metadata_reserve_pct = 25;
	+static uint_t zfs_special_class_metadata_reserve_pct = 25;

	/*
	* ==========================================================================
	* SPA config locking
	* ==========================================================================
	*/
	static void
	spa_config_lock_init(spa_t *spa)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
	scl->scl_writer = NULL;
	scl->scl_write_wanted = 0;
	scl->scl_count = 0;
	}
	}

	static void
	spa_config_lock_destroy(spa_t *spa)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	mutex_destroy(&scl->scl_lock);
	cv_destroy(&scl->scl_cv);
	ASSERT(scl->scl_writer == NULL);
	ASSERT(scl->scl_write_wanted == 0);
	ASSERT(scl->scl_count == 0);
	}
	}

	int
	spa_config_tryenter(spa_t spa, int locks, const void tag, krw_t rw)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	if (rw == RW_READER) {
	if (scl->scl_writer \|\| scl->scl_write_wanted) {
	mutex_exit(&scl->scl_lock);
	spa_config_exit(spa, locks & ((1 << i) - 1),
	tag);
	return (0);
	}
	} else {
	ASSERT(scl->scl_writer != curthread);
	if (scl->scl_count != 0) {
	mutex_exit(&scl->scl_lock);
	spa_config_exit(spa, locks & ((1 << i) - 1),
	tag);
	return (0);
	}
	scl->scl_writer = curthread;
	}
	scl->scl_count++;
	mutex_exit(&scl->scl_lock);
	}
	return (1);
	}

	void
	spa_config_enter(spa_t spa, int locks, const void tag, krw_t rw)
	{
	(void) tag;
	int wlocks_held = 0;

	ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY);

	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (scl->scl_writer == curthread)
	wlocks_held \|= (1 << i);
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	if (rw == RW_READER) {
	while (scl->scl_writer \|\| scl->scl_write_wanted) {
	cv_wait(&scl->scl_cv, &scl->scl_lock);
	}
	} else {
	ASSERT(scl->scl_writer != curthread);
	while (scl->scl_count != 0) {
	scl->scl_write_wanted++;
	cv_wait(&scl->scl_cv, &scl->scl_lock);
	scl->scl_write_wanted--;
	}
	scl->scl_writer = curthread;
	}
	scl->scl_count++;
	mutex_exit(&scl->scl_lock);
	}
	ASSERT3U(wlocks_held, <=, locks);
	}

	void
	spa_config_exit(spa_t spa, int locks, const void tag)
	{
	(void) tag;
	for (int i = SCL_LOCKS - 1; i >= 0; i--) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	ASSERT(scl->scl_count > 0);
	if (--scl->scl_count == 0) {
	ASSERT(scl->scl_writer == NULL \|\|
	scl->scl_writer == curthread);
	scl->scl_writer = NULL; /* OK in either case */
	cv_broadcast(&scl->scl_cv);
	}
	mutex_exit(&scl->scl_lock);
	}
	}

	int
	spa_config_held(spa_t *spa, int locks, krw_t rw)
	{
	int locks_held = 0;

	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	if ((rw == RW_READER && scl->scl_count != 0) \|\|
	(rw == RW_WRITER && scl->scl_writer == curthread))
	locks_held \|= 1 << i;
	}

	return (locks_held);
	}

	/*
	* ==========================================================================
	* SPA namespace functions
	* ==========================================================================
	*/

	/*
	* Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
	* Returns NULL if no matching spa_t is found.
	*/
	spa_t *
	spa_lookup(const char *name)
	{
	static spa_t search; /* spa_t is large; don't allocate on stack */
	spa_t *spa;
	avl_index_t where;
	char *cp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	(void) strlcpy(search.spa_name, name, sizeof (search.spa_name));

	/*
	* If it's a full dataset name, figure out the pool name and
	* just use that.
	*/
	cp = strpbrk(search.spa_name, "/@#");
	if (cp != NULL)
	*cp = '\0';

	spa = avl_find(&spa_namespace_avl, &search, &where);

	return (spa);
	}

	/*
	* Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
	* If the zfs_deadman_enabled flag is set then it inspects all vdev queues
	* looking for potentially hung I/Os.
	*/
	void
	spa_deadman(void *arg)
	{
	spa_t *spa = arg;

	/* Disable the deadman if the pool is suspended. */
	if (spa_suspended(spa))
	return;

	zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
	(gethrtime() - spa->spa_sync_starttime) / NANOSEC,
	(u_longlong_t)++spa->spa_deadman_calls);
	if (zfs_deadman_enabled)
	vdev_deadman(spa->spa_root_vdev, FTAG);

	spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
	spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
	MSEC_TO_TICK(zfs_deadman_checktime_ms));
	}

	static int
	spa_log_sm_sort_by_txg(const void va, const void vb)
	{
	const spa_log_sm_t *a = va;
	const spa_log_sm_t *b = vb;

	return (TREE_CMP(a->sls_txg, b->sls_txg));
	}

	/*
	* Create an uninitialized spa_t with the given name. Requires
	* spa_namespace_lock. The caller must ensure that the spa_t doesn't already
	* exist by calling spa_lookup() first.
	*/
	spa_t *
	spa_add(const char name, nvlist_t config, const char *altroot)
	{
	spa_t *spa;
	spa_config_dirent_t *dp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);

	mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL);

	cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL);

	for (int t = 0; t < TXG_SIZE; t++)
	bplist_create(&spa->spa_free_bplist[t]);

	(void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
	spa->spa_state = POOL_STATE_UNINITIALIZED;
	spa->spa_freeze_txg = UINT64_MAX;
	spa->spa_final_txg = UINT64_MAX;
	spa->spa_load_max_txg = UINT64_MAX;
	spa->spa_proc = &p0;
	spa->spa_proc_state = SPA_PROC_NONE;
	spa->spa_trust_config = B_TRUE;
	spa->spa_hostid = zone_get_hostid(NULL);

	spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms);
	spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms);
	spa_set_deadman_failmode(spa, zfs_deadman_failmode);

	zfs_refcount_create(&spa->spa_refcount);
	spa_config_lock_init(spa);
	spa_stats_init(spa);

	avl_add(&spa_namespace_avl, spa);

	/*
	* Set the alternate root, if there is one.
	*/
	if (altroot)
	spa->spa_root = spa_strdup(altroot);

	spa->spa_alloc_count = spa_allocators;
	spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count *
	sizeof (spa_alloc_t), KM_SLEEP);
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT,
	NULL);
	avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare,
	sizeof (zio_t), offsetof(zio_t, io_alloc_node));
	}
	avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed,
	sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node));
	avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg,
	sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node));
	list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t),
	offsetof(log_summary_entry_t, lse_node));

	/*
	* Every pool starts with the default cachefile
	*/
	list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
	offsetof(spa_config_dirent_t, scd_link));

	dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
	dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path);
	list_insert_head(&spa->spa_config_list, dp);

	VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME,
	KM_SLEEP) == 0);

	if (config != NULL) {
	nvlist_t *features;

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
	&features) == 0) {
	VERIFY(nvlist_dup(features, &spa->spa_label_features,
	0) == 0);
	}

	VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
	}

	if (spa->spa_label_features == NULL) {
	VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME,
	KM_SLEEP) == 0);
	}

	spa->spa_min_ashift = INT_MAX;
	spa->spa_max_ashift = 0;
	spa->spa_min_alloc = INT_MAX;

	/* Reset cached value */
	spa->spa_dedup_dspace = ~0ULL;

	/*
	* As a pool is being created, treat all features as disabled by
	* setting SPA_FEATURE_DISABLED for all entries in the feature
	* refcount cache.
	*/
	for (int i = 0; i < SPA_FEATURES; i++) {
	spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED;
	}

	list_create(&spa->spa_leaf_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_leaf_node));

	return (spa);
	}

	/*
	* Removes a spa_t from the namespace, freeing up any memory used. Requires
	* spa_namespace_lock. This is called only after the spa_t has been closed and
	* deactivated.
	*/
	void
	spa_remove(spa_t *spa)
	{
	spa_config_dirent_t *dp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED);
	ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0);
	ASSERT0(spa->spa_waiters);

	nvlist_free(spa->spa_config_splitting);

	avl_remove(&spa_namespace_avl, spa);
	cv_broadcast(&spa_namespace_cv);

	if (spa->spa_root)
	spa_strfree(spa->spa_root);

	while ((dp = list_head(&spa->spa_config_list)) != NULL) {
	list_remove(&spa->spa_config_list, dp);
	if (dp->scd_path != NULL)
	spa_strfree(dp->scd_path);
	kmem_free(dp, sizeof (spa_config_dirent_t));
	}

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	avl_destroy(&spa->spa_allocs[i].spaa_tree);
	mutex_destroy(&spa->spa_allocs[i].spaa_lock);
	}
	kmem_free(spa->spa_allocs, spa->spa_alloc_count *
	sizeof (spa_alloc_t));

	avl_destroy(&spa->spa_metaslabs_by_flushed);
	avl_destroy(&spa->spa_sm_logs_by_txg);
	list_destroy(&spa->spa_log_summary);
	list_destroy(&spa->spa_config_list);
	list_destroy(&spa->spa_leaf_list);

	nvlist_free(spa->spa_label_features);
	nvlist_free(spa->spa_load_info);
	nvlist_free(spa->spa_feat_stats);
	spa_config_set(spa, NULL);

	zfs_refcount_destroy(&spa->spa_refcount);

	spa_stats_destroy(spa);
	spa_config_lock_destroy(spa);

	for (int t = 0; t < TXG_SIZE; t++)
	bplist_destroy(&spa->spa_free_bplist[t]);

	zio_checksum_templates_free(spa);

	cv_destroy(&spa->spa_async_cv);
	cv_destroy(&spa->spa_evicting_os_cv);
	cv_destroy(&spa->spa_proc_cv);
	cv_destroy(&spa->spa_scrub_io_cv);
	cv_destroy(&spa->spa_suspend_cv);
	cv_destroy(&spa->spa_activities_cv);
	cv_destroy(&spa->spa_waiters_cv);

	mutex_destroy(&spa->spa_flushed_ms_lock);
	mutex_destroy(&spa->spa_async_lock);
	mutex_destroy(&spa->spa_errlist_lock);
	mutex_destroy(&spa->spa_errlog_lock);
	mutex_destroy(&spa->spa_evicting_os_lock);
	mutex_destroy(&spa->spa_history_lock);
	mutex_destroy(&spa->spa_proc_lock);
	mutex_destroy(&spa->spa_props_lock);
	mutex_destroy(&spa->spa_cksum_tmpls_lock);
	mutex_destroy(&spa->spa_scrub_lock);
	mutex_destroy(&spa->spa_suspend_lock);
	mutex_destroy(&spa->spa_vdev_top_lock);
	mutex_destroy(&spa->spa_feat_stats_lock);
	mutex_destroy(&spa->spa_activities_lock);

	kmem_free(spa, sizeof (spa_t));
	}

	/*
	* Given a pool, return the next pool in the namespace, or NULL if there is
	* none. If 'prev' is NULL, return the first pool.
	*/
	spa_t *
	spa_next(spa_t *prev)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	if (prev)
	return (AVL_NEXT(&spa_namespace_avl, prev));
	else
	return (avl_first(&spa_namespace_avl));
	}

	/*
	* ==========================================================================
	* SPA refcount functions
	* ==========================================================================
	*/

	/*
	* Add a reference to the given spa_t. Must have at least one reference, or
	* have the namespace lock held.
	*/
	void
	spa_open_ref(spa_t spa, const void tag)
	{
	ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref \|\|
	MUTEX_HELD(&spa_namespace_lock));
	(void) zfs_refcount_add(&spa->spa_refcount, tag);
	}

	/*
	* Remove a reference to the given spa_t. Must have at least one reference, or
	* have the namespace lock held.
	*/
	void
	spa_close(spa_t spa, const void tag)
	{
	ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref \|\|
	MUTEX_HELD(&spa_namespace_lock));
	(void) zfs_refcount_remove(&spa->spa_refcount, tag);
	}

	/*
	* Remove a reference to the given spa_t held by a dsl dir that is
	* being asynchronously released. Async releases occur from a taskq
	* performing eviction of dsl datasets and dirs. The namespace lock
	* isn't held and the hold by the object being evicted may contribute to
	* spa_minref (e.g. dataset or directory released during pool export),
	* so the asserts in spa_close() do not apply.
	*/
	void
	spa_async_close(spa_t spa, const void tag)
	{
	(void) zfs_refcount_remove(&spa->spa_refcount, tag);
	}

	/*
	* Check to see if the spa refcount is zero. Must be called with
	* spa_namespace_lock held. We really compare against spa_minref, which is the
	* number of references acquired when opening a pool
	*/
	boolean_t
	spa_refcount_zero(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref);
	}

	/*
	* ==========================================================================
	* SPA spare and l2cache tracking
	* ==========================================================================
	*/

	/*
	* Hot spares and cache devices are tracked using the same code below,
	* for 'auxiliary' devices.
	*/

	typedef struct spa_aux {
	uint64_t aux_guid;
	uint64_t aux_pool;
	avl_node_t aux_avl;
	int aux_count;
	} spa_aux_t;

	static inline int
	spa_aux_compare(const void a, const void b)
	{
	const spa_aux_t sa = (const spa_aux_t )a;
	const spa_aux_t sb = (const spa_aux_t )b;

	return (TREE_CMP(sa->aux_guid, sb->aux_guid));
	}

	static void
	spa_aux_add(vdev_t vd, avl_tree_t avl)
	{
	avl_index_t where;
	spa_aux_t search;
	spa_aux_t *aux;

	search.aux_guid = vd->vdev_guid;
	if ((aux = avl_find(avl, &search, &where)) != NULL) {
	aux->aux_count++;
	} else {
	aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
	aux->aux_guid = vd->vdev_guid;
	aux->aux_count = 1;
	avl_insert(avl, aux, where);
	}
	}

	static void
	spa_aux_remove(vdev_t vd, avl_tree_t avl)
	{
	spa_aux_t search;
	spa_aux_t *aux;
	avl_index_t where;

	search.aux_guid = vd->vdev_guid;
	aux = avl_find(avl, &search, &where);

	ASSERT(aux != NULL);

	if (--aux->aux_count == 0) {
	avl_remove(avl, aux);
	kmem_free(aux, sizeof (spa_aux_t));
	} else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
	aux->aux_pool = 0ULL;
	}
	}

	static boolean_t
	spa_aux_exists(uint64_t guid, uint64_t pool, int refcnt, avl_tree_t *avl)
	{
	spa_aux_t search, *found;

	search.aux_guid = guid;
	found = avl_find(avl, &search, NULL);

	if (pool) {
	if (found)
	*pool = found->aux_pool;
	else
	*pool = 0ULL;
	}

	if (refcnt) {
	if (found)
	*refcnt = found->aux_count;
	else
	*refcnt = 0;
	}

	return (found != NULL);
	}

	static void
	spa_aux_activate(vdev_t vd, avl_tree_t avl)
	{
	spa_aux_t search, *found;
	avl_index_t where;

	search.aux_guid = vd->vdev_guid;
	found = avl_find(avl, &search, &where);
	ASSERT(found != NULL);
	ASSERT(found->aux_pool == 0ULL);

	found->aux_pool = spa_guid(vd->vdev_spa);
	}

	/*
	* Spares are tracked globally due to the following constraints:
	*
	* - A spare may be part of multiple pools.
	* - A spare may be added to a pool even if it's actively in use within
	* another pool.
	* - A spare in use in any pool can only be the source of a replacement if
	* the target is a spare in the same pool.
	*
	* We keep track of all spares on the system through the use of a reference
	* counted AVL tree. When a vdev is added as a spare, or used as a replacement
	* spare, then we bump the reference count in the AVL tree. In addition, we set
	* the 'vdev_isspare' member to indicate that the device is a spare (active or
	* inactive). When a spare is made active (used to replace a device in the
	* pool), we also keep track of which pool its been made a part of.
	*
	* The 'spa_spare_lock' protects the AVL tree. These functions are normally
	* called under the spa_namespace lock as part of vdev reconfiguration. The
	* separate spare lock exists for the status query path, which does not need to
	* be completely consistent with respect to other vdev configuration changes.
	*/

	static int
	spa_spare_compare(const void a, const void b)
	{
	return (spa_aux_compare(a, b));
	}

	void
	spa_spare_add(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(!vd->vdev_isspare);
	spa_aux_add(vd, &spa_spare_avl);
	vd->vdev_isspare = B_TRUE;
	mutex_exit(&spa_spare_lock);
	}

	void
	spa_spare_remove(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(vd->vdev_isspare);
	spa_aux_remove(vd, &spa_spare_avl);
	vd->vdev_isspare = B_FALSE;
	mutex_exit(&spa_spare_lock);
	}

	boolean_t
	spa_spare_exists(uint64_t guid, uint64_t pool, int refcnt)
	{
	boolean_t found;

	mutex_enter(&spa_spare_lock);
	found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
	mutex_exit(&spa_spare_lock);

	return (found);
	}

	void
	spa_spare_activate(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(vd->vdev_isspare);
	spa_aux_activate(vd, &spa_spare_avl);
	mutex_exit(&spa_spare_lock);
	}

	/*
	* Level 2 ARC devices are tracked globally for the same reasons as spares.
	* Cache devices currently only support one pool per cache device, and so
	* for these devices the aux reference count is currently unused beyond 1.
	*/

	static int
	spa_l2cache_compare(const void a, const void b)
	{
	return (spa_aux_compare(a, b));
	}

	void
	spa_l2cache_add(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(!vd->vdev_isl2cache);
	spa_aux_add(vd, &spa_l2cache_avl);
	vd->vdev_isl2cache = B_TRUE;
	mutex_exit(&spa_l2cache_lock);
	}

	void
	spa_l2cache_remove(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(vd->vdev_isl2cache);
	spa_aux_remove(vd, &spa_l2cache_avl);
	vd->vdev_isl2cache = B_FALSE;
	mutex_exit(&spa_l2cache_lock);
	}

	boolean_t
	spa_l2cache_exists(uint64_t guid, uint64_t *pool)
	{
	boolean_t found;

	mutex_enter(&spa_l2cache_lock);
	found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
	mutex_exit(&spa_l2cache_lock);

	return (found);
	}

	void
	spa_l2cache_activate(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(vd->vdev_isl2cache);
	spa_aux_activate(vd, &spa_l2cache_avl);
	mutex_exit(&spa_l2cache_lock);
	}

	/*
	* ==========================================================================
	* SPA vdev locking
	* ==========================================================================
	*/

	/*
	* Lock the given spa_t for the purpose of adding or removing a vdev.
	* Grabs the global spa_namespace_lock plus the spa config lock for writing.
	* It returns the next transaction group for the spa_t.
	*/
	uint64_t
	spa_vdev_enter(spa_t *spa)
	{
	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);

	vdev_autotrim_stop_all(spa);

	return (spa_vdev_config_enter(spa));
	}

	/*
	* The same as spa_vdev_enter() above but additionally takes the guid of
	* the vdev being detached. When there is a rebuild in process it will be
	* suspended while the vdev tree is modified then resumed by spa_vdev_exit().
	* The rebuild is canceled if only a single child remains after the detach.
	*/
	uint64_t
	spa_vdev_detach_enter(spa_t *spa, uint64_t guid)
	{
	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);

	vdev_autotrim_stop_all(spa);

	if (guid != 0) {
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd) {
	vdev_rebuild_stop_wait(vd->vdev_top);
	}
	}

	return (spa_vdev_config_enter(spa));
	}

	/*
	* Internal implementation for spa_vdev_enter(). Used when a vdev
	* operation requires multiple syncs (i.e. removing a device) while
	* keeping the spa_namespace_lock held.
	*/
	uint64_t
	spa_vdev_config_enter(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);

	return (spa_last_synced_txg(spa) + 1);
	}

	/*
	* Used in combination with spa_vdev_config_enter() to allow the syncing
	* of multiple transactions without releasing the spa_namespace_lock.
	*/
	void
	spa_vdev_config_exit(spa_t spa, vdev_t vd, uint64_t txg, int error,
	const char *tag)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	int config_changed = B_FALSE;

	ASSERT(txg > spa_last_synced_txg(spa));

	spa->spa_pending_vdev = NULL;

	/*
	* Reassess the DTLs.
	*/
	vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE);

	if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
	config_changed = B_TRUE;
	spa->spa_config_generation++;
	}

	/*
	* Verify the metaslab classes.
	*/
	ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0);

	spa_config_exit(spa, SCL_ALL, spa);

	/*
	* Panic the system if the specified tag requires it. This
	* is useful for ensuring that configurations are updated
	* transactionally.
	*/
	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, tag, 0);

	/*
	* Note: this txg_wait_synced() is important because it ensures
	* that there won't be more than one config change per txg.
	* This allows us to use the txg as the generation number.
	*/
	if (error == 0)
	txg_wait_synced(spa->spa_dsl_pool, txg);

	if (vd != NULL) {
	ASSERT(!vd->vdev_detached \|\| vd->vdev_dtl_sm == NULL);
	if (vd->vdev_ops->vdev_op_leaf) {
	mutex_enter(&vd->vdev_initialize_lock);
	vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED,
	NULL);
	mutex_exit(&vd->vdev_initialize_lock);

	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL);
	mutex_exit(&vd->vdev_trim_lock);
	}

	/*
	* The vdev may be both a leaf and top-level device.
	*/
	vdev_autotrim_stop_wait(vd);

	spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER);
	vdev_free(vd);
	spa_config_exit(spa, SCL_STATE_ALL, spa);
	}

	/*
	* If the config changed, update the config cache.
	*/
	if (config_changed)
	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	}

	/*
	* Unlock the spa_t after adding or removing a vdev. Besides undoing the
	* locking of spa_vdev_enter(), we also want make sure the transactions have
	* synced to disk, and then update the global configuration cache with the new
	* information.
	*/
	int
	spa_vdev_exit(spa_t spa, vdev_t vd, uint64_t txg, int error)
	{
	vdev_autotrim_restart(spa);
	vdev_rebuild_restart(spa);

	spa_vdev_config_exit(spa, vd, txg, error, FTAG);
	mutex_exit(&spa_namespace_lock);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* Lock the given spa_t for the purpose of changing vdev state.
	*/
	void
	spa_vdev_state_enter(spa_t *spa, int oplocks)
	{
	int locks = SCL_STATE_ALL \| oplocks;

	/*
	* Root pools may need to read of the underlying devfs filesystem
	* when opening up a vdev. Unfortunately if we're holding the
	* SCL_ZIO lock it will result in a deadlock when we try to issue
	* the read from the root filesystem. Instead we "prefetch"
	* the associated vnodes that we need prior to opening the
	* underlying devices and cache them so that we can prevent
	* any I/O when we are doing the actual open.
	*/
	if (spa_is_root(spa)) {
	int low = locks & ~(SCL_ZIO - 1);
	int high = locks & ~low;

	spa_config_enter(spa, high, spa, RW_WRITER);
	vdev_hold(spa->spa_root_vdev);
	spa_config_enter(spa, low, spa, RW_WRITER);
	} else {
	spa_config_enter(spa, locks, spa, RW_WRITER);
	}
	spa->spa_vdev_locks = locks;
	}

	int
	spa_vdev_state_exit(spa_t spa, vdev_t vd, int error)
	{
	boolean_t config_changed = B_FALSE;
	vdev_t *vdev_top;

	if (vd == NULL \|\| vd == spa->spa_root_vdev) {
	vdev_top = spa->spa_root_vdev;
	} else {
	vdev_top = vd->vdev_top;
	}

	if (vd != NULL \|\| error == 0)
	vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE);

	if (vd != NULL) {
	if (vd != spa->spa_root_vdev)
	vdev_state_dirty(vdev_top);

	config_changed = B_TRUE;
	spa->spa_config_generation++;
	}

	if (spa_is_root(spa))
	vdev_rele(spa->spa_root_vdev);

	ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL);
	spa_config_exit(spa, spa->spa_vdev_locks, spa);

	/*
	* If anything changed, wait for it to sync. This ensures that,
	* from the system administrator's perspective, zpool(8) commands
	* are synchronous. This is important for things like zpool offline:
	* when the command completes, you expect no further I/O from ZFS.
	*/
	if (vd != NULL)
	txg_wait_synced(spa->spa_dsl_pool, 0);

	/*
	* If the config changed, update the config cache.
	*/
	if (config_changed) {
	mutex_enter(&spa_namespace_lock);
	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	mutex_exit(&spa_namespace_lock);
	}

	return (error);
	}

	/*
	* ==========================================================================
	* Miscellaneous functions
	* ==========================================================================
	*/

	void
	spa_activate_mos_feature(spa_t spa, const char feature, dmu_tx_t *tx)
	{
	if (!nvlist_exists(spa->spa_label_features, feature)) {
	fnvlist_add_boolean(spa->spa_label_features, feature);
	/*
	* When we are creating the pool (tx_txg==TXG_INITIAL), we can't
	* dirty the vdev config because lock SCL_CONFIG is not held.
	* Thankfully, in this case we don't need to dirty the config
	* because it will be written out anyway when we finish
	* creating the pool.
	*/
	if (tx->tx_txg != TXG_INITIAL)
	vdev_config_dirty(spa->spa_root_vdev);
	}
	}

	void
	spa_deactivate_mos_feature(spa_t spa, const char feature)
	{
	if (nvlist_remove_all(spa->spa_label_features, feature) == 0)
	vdev_config_dirty(spa->spa_root_vdev);
	}

	/*
	* Return the spa_t associated with given pool_guid, if it exists. If
	* device_guid is non-zero, determine whether the pool exists and contains
	* a device with the specified device_guid.
	*/
	spa_t *
	spa_by_guid(uint64_t pool_guid, uint64_t device_guid)
	{
	spa_t *spa;
	avl_tree_t *t = &spa_namespace_avl;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
	if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	continue;
	if (spa->spa_root_vdev == NULL)
	continue;
	if (spa_guid(spa) == pool_guid) {
	if (device_guid == 0)
	break;

	if (vdev_lookup_by_guid(spa->spa_root_vdev,
	device_guid) != NULL)
	break;

	/*
	* Check any devices we may be in the process of adding.
	*/
	if (spa->spa_pending_vdev) {
	if (vdev_lookup_by_guid(spa->spa_pending_vdev,
	device_guid) != NULL)
	break;
	}
	}
	}

	return (spa);
	}

	/*
	* Determine whether a pool with the given pool_guid exists.
	*/
	boolean_t
	spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
	{
	return (spa_by_guid(pool_guid, device_guid) != NULL);
	}

	char *
	spa_strdup(const char *s)
	{
	size_t len;
	char *new;

	len = strlen(s);
	new = kmem_alloc(len + 1, KM_SLEEP);
	memcpy(new, s, len + 1);

	return (new);
	}

	void
	spa_strfree(char *s)
	{
	kmem_free(s, strlen(s) + 1);
	}

	uint64_t
	spa_generate_guid(spa_t *spa)
	{
	uint64_t guid;

	if (spa != NULL) {
	do {
	(void) random_get_pseudo_bytes((void *)&guid,
	sizeof (guid));
	} while (guid == 0 \|\| spa_guid_exists(spa_guid(spa), guid));
	} else {
	do {
	(void) random_get_pseudo_bytes((void *)&guid,
	sizeof (guid));
	} while (guid == 0 \|\| spa_guid_exists(guid, 0));
	}

	return (guid);
	}

	void
	snprintf_blkptr(char buf, size_t buflen, const blkptr_t bp)
	{
	char type[256];
	const char *checksum = NULL;
	const char *compress = NULL;

	if (bp != NULL) {
	if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) {
	dmu_object_byteswap_t bswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
	(void) snprintf(type, sizeof (type), "bswap %s %s",
	DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ?
	"metadata" : "data",
	dmu_ot_byteswap[bswap].ob_name);
	} else {
	(void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name,
	sizeof (type));
	}
	if (!BP_IS_EMBEDDED(bp)) {
	checksum =
	zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name;
	}
	compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name;
	}

	SNPRINTF_BLKPTR(snprintf, ' ', buf, buflen, bp, type, checksum,
	compress);
	}

	void
	spa_freeze(spa_t *spa)
	{
	uint64_t freeze_txg = 0;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	if (spa->spa_freeze_txg == UINT64_MAX) {
	freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
	spa->spa_freeze_txg = freeze_txg;
	}
	spa_config_exit(spa, SCL_ALL, FTAG);
	if (freeze_txg != 0)
	txg_wait_synced(spa_get_dsl(spa), freeze_txg);
	}

	void
	zfs_panic_recover(const char *fmt, ...)
	{
	va_list adx;

	va_start(adx, fmt);
	vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
	va_end(adx);
	}

	/*
	* This is a stripped-down version of strtoull, suitable only for converting
	* lowercase hexadecimal numbers that don't overflow.
	*/
	uint64_t
	zfs_strtonum(const char str, char *nptr)
	{
	uint64_t val = 0;
	char c;
	int digit;

	while ((c = *str) != '\0') {
	if (c >= '0' && c <= '9')
	digit = c - '0';
	else if (c >= 'a' && c <= 'f')
	digit = 10 + c - 'a';
	else
	break;

	val *= 16;
	val += digit;

	str++;
	}

	if (nptr)
	nptr = (char )str;

	return (val);
	}

	void
	spa_activate_allocation_classes(spa_t spa, dmu_tx_t tx)
	{
	/*
	* We bump the feature refcount for each special vdev added to the pool
	*/
	ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES));
	spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx);
	}

	/*
	* ==========================================================================
	* Accessor functions
	* ==========================================================================
	*/

	boolean_t
	spa_shutting_down(spa_t *spa)
	{
	return (spa->spa_async_suspended);
	}

	dsl_pool_t *
	spa_get_dsl(spa_t *spa)
	{
	return (spa->spa_dsl_pool);
	}

	boolean_t
	spa_is_initializing(spa_t *spa)
	{
	return (spa->spa_is_initializing);
	}

	boolean_t
	spa_indirect_vdevs_loaded(spa_t *spa)
	{
	return (spa->spa_indirect_vdevs_loaded);
	}

	blkptr_t *
	spa_get_rootblkptr(spa_t *spa)
	{
	return (&spa->spa_ubsync.ub_rootbp);
	}

	void
	spa_set_rootblkptr(spa_t spa, const blkptr_t bp)
	{
	spa->spa_uberblock.ub_rootbp = *bp;
	}

	void
	spa_altroot(spa_t spa, char buf, size_t buflen)
	{
	if (spa->spa_root == NULL)
	buf[0] = '\0';
	else
	- (void) strncpy(buf, spa->spa_root, buflen);
	+ (void) strlcpy(buf, spa->spa_root, buflen);
	}

	-int
	+uint32_t
	spa_sync_pass(spa_t *spa)
	{
	return (spa->spa_sync_pass);
	}

	char *
	spa_name(spa_t *spa)
	{
	return (spa->spa_name);
	}

	uint64_t
	spa_guid(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);
	uint64_t guid;

	/*
	* If we fail to parse the config during spa_load(), we can go through
	* the error path (which posts an ereport) and end up here with no root
	* vdev. We stash the original pool guid in 'spa_config_guid' to handle
	* this case.
	*/
	if (spa->spa_root_vdev == NULL)
	return (spa->spa_config_guid);

	guid = spa->spa_last_synced_guid != 0 ?
	spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid;

	/*
	* Return the most recently synced out guid unless we're
	* in syncing context.
	*/
	if (dp && dsl_pool_sync_context(dp))
	return (spa->spa_root_vdev->vdev_guid);
	else
	return (guid);
	}

	uint64_t
	spa_load_guid(spa_t *spa)
	{
	/*
	* This is a GUID that exists solely as a reference for the
	* purposes of the arc. It is generated at load time, and
	* is never written to persistent storage.
	*/
	return (spa->spa_load_guid);
	}

	uint64_t
	spa_last_synced_txg(spa_t *spa)
	{
	return (spa->spa_ubsync.ub_txg);
	}

	uint64_t
	spa_first_txg(spa_t *spa)
	{
	return (spa->spa_first_txg);
	}

	uint64_t
	spa_syncing_txg(spa_t *spa)
	{
	return (spa->spa_syncing_txg);
	}

	/*
	* Return the last txg where data can be dirtied. The final txgs
	* will be used to just clear out any deferred frees that remain.
	*/
	uint64_t
	spa_final_dirty_txg(spa_t *spa)
	{
	return (spa->spa_final_txg - TXG_DEFER_SIZE);
	}

	pool_state_t
	spa_state(spa_t *spa)
	{
	return (spa->spa_state);
	}

	spa_load_state_t
	spa_load_state(spa_t *spa)
	{
	return (spa->spa_load_state);
	}

	uint64_t
	spa_freeze_txg(spa_t *spa)
	{
	return (spa->spa_freeze_txg);
	}

	/*
	* Return the inflated asize for a logical write in bytes. This is used by the
	* DMU to calculate the space a logical write will require on disk.
	* If lsize is smaller than the largest physical block size allocatable on this
	* pool we use its value instead, since the write will end up using the whole
	* block anyway.
	*/
	uint64_t
	spa_get_worst_case_asize(spa_t *spa, uint64_t lsize)
	{
	if (lsize == 0)
	return (0); /* No inflation needed */
	return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation);
	}

	/*
	* Return the amount of slop space in bytes. It is typically 1/32 of the pool
	* (3.2%), minus the embedded log space. On very small pools, it may be
	* slightly larger than this. On very large pools, it will be capped to
	* the value of spa_max_slop. The embedded log space is not included in
	* spa_dspace. By subtracting it, the usable space (per "zfs list") is a
	* constant 97% of the total space, regardless of metaslab size (assuming the
	* default spa_slop_shift=5 and a non-tiny pool).
	*
	* See the comment above spa_slop_shift for more details.
	*/
	uint64_t
	spa_get_slop_space(spa_t *spa)
	{
	uint64_t space = 0;
	uint64_t slop = 0;

	/*
	* Make sure spa_dedup_dspace has been set.
	*/
	if (spa->spa_dedup_dspace == ~0ULL)
	spa_update_dspace(spa);

	/*
	* spa_get_dspace() includes the space only logically "used" by
	* deduplicated data, so since it's not useful to reserve more
	* space with more deduplicated data, we subtract that out here.
	*/
	space = spa_get_dspace(spa) - spa->spa_dedup_dspace;
	slop = MIN(space >> spa_slop_shift, spa_max_slop);

	/*
	* Subtract the embedded log space, but no more than half the (3.2%)
	* unusable space. Note, the "no more than half" is only relevant if
	* zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by
	* default.
	*/
	uint64_t embedded_log =
	metaslab_class_get_dspace(spa_embedded_log_class(spa));
	slop -= MIN(embedded_log, slop >> 1);

	/*
	* Slop space should be at least spa_min_slop, but no more than half
	* the entire pool.
	*/
	slop = MAX(slop, MIN(space >> 1, spa_min_slop));
	return (slop);
	}

	uint64_t
	spa_get_dspace(spa_t *spa)
	{
	return (spa->spa_dspace);
	}

	uint64_t
	spa_get_checkpoint_space(spa_t *spa)
	{
	return (spa->spa_checkpoint_info.sci_dspace);
	}

	void
	spa_update_dspace(spa_t *spa)
	{
	spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) +
	ddt_get_dedup_dspace(spa);
	if (spa->spa_nonallocating_dspace > 0) {
	/*
	* Subtract the space provided by all non-allocating vdevs that
	* contribute to dspace. If a file is overwritten, its old
	* blocks are freed and new blocks are allocated. If there are
	* no snapshots of the file, the available space should remain
	* the same. The old blocks could be freed from the
	* non-allocating vdev, but the new blocks must be allocated on
	* other (allocating) vdevs. By reserving the entire size of
	* the non-allocating vdevs (including allocated space), we
	* ensure that there will be enough space on the allocating
	* vdevs for this file overwrite to succeed.
	*
	* Note that the DMU/DSL doesn't actually know or care
	* how much space is allocated (it does its own tracking
	* of how much space has been logically used). So it
	* doesn't matter that the data we are moving may be
	* allocated twice (on the old device and the new device).
	*/
	ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace);
	spa->spa_dspace -= spa->spa_nonallocating_dspace;
	}
	}

	/*
	* Return the failure mode that has been set to this pool. The default
	* behavior will be to block all I/Os when a complete failure occurs.
	*/
	uint64_t
	spa_get_failmode(spa_t *spa)
	{
	return (spa->spa_failmode);
	}

	boolean_t
	spa_suspended(spa_t *spa)
	{
	return (spa->spa_suspended != ZIO_SUSPEND_NONE);
	}

	uint64_t
	spa_version(spa_t *spa)
	{
	return (spa->spa_ubsync.ub_version);
	}

	boolean_t
	spa_deflate(spa_t *spa)
	{
	return (spa->spa_deflate);
	}

	metaslab_class_t *
	spa_normal_class(spa_t *spa)
	{
	return (spa->spa_normal_class);
	}

	metaslab_class_t *
	spa_log_class(spa_t *spa)
	{
	return (spa->spa_log_class);
	}

	metaslab_class_t *
	spa_embedded_log_class(spa_t *spa)
	{
	return (spa->spa_embedded_log_class);
	}

	metaslab_class_t *
	spa_special_class(spa_t *spa)
	{
	return (spa->spa_special_class);
	}

	metaslab_class_t *
	spa_dedup_class(spa_t *spa)
	{
	return (spa->spa_dedup_class);
	}

	/*
	* Locate an appropriate allocation class
	*/
	metaslab_class_t *
	spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype,
	uint_t level, uint_t special_smallblk)
	{
	/*
	* ZIL allocations determine their class in zio_alloc_zil().
	*/
	ASSERT(objtype != DMU_OT_INTENT_LOG);

	boolean_t has_special_class = spa->spa_special_class->mc_groups != 0;

	if (DMU_OT_IS_DDT(objtype)) {
	if (spa->spa_dedup_class->mc_groups != 0)
	return (spa_dedup_class(spa));
	else if (has_special_class && zfs_ddt_data_is_special)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	/* Indirect blocks for user data can land in special if allowed */
	if (level > 0 && (DMU_OT_IS_FILE(objtype) \|\| objtype == DMU_OT_ZVOL)) {
	if (has_special_class && zfs_user_indirect_is_special)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	if (DMU_OT_IS_METADATA(objtype) \|\| level > 0) {
	if (has_special_class)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	/*
	* Allow small file blocks in special class in some cases (like
	* for the dRAID vdev feature). But always leave a reserve of
	* zfs_special_class_metadata_reserve_pct exclusively for metadata.
	*/
	if (DMU_OT_IS_FILE(objtype) &&
	has_special_class && size <= special_smallblk) {
	metaslab_class_t *special = spa_special_class(spa);
	uint64_t alloc = metaslab_class_get_alloc(special);
	uint64_t space = metaslab_class_get_space(special);
	uint64_t limit =
	(space * (100 - zfs_special_class_metadata_reserve_pct))
	/ 100;

	if (alloc < limit)
	return (special);
	}

	return (spa_normal_class(spa));
	}

	void
	spa_evicting_os_register(spa_t spa, objset_t os)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	list_insert_head(&spa->spa_evicting_os_list, os);
	mutex_exit(&spa->spa_evicting_os_lock);
	}

	void
	spa_evicting_os_deregister(spa_t spa, objset_t os)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	list_remove(&spa->spa_evicting_os_list, os);
	cv_broadcast(&spa->spa_evicting_os_cv);
	mutex_exit(&spa->spa_evicting_os_lock);
	}

	void
	spa_evicting_os_wait(spa_t *spa)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	while (!list_is_empty(&spa->spa_evicting_os_list))
	cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock);
	mutex_exit(&spa->spa_evicting_os_lock);

	dmu_buf_user_evict_wait();
	}

	int
	spa_max_replication(spa_t *spa)
	{
	/*
	* As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
	* handle BPs with more than one DVA allocated. Set our max
	* replication level accordingly.
	*/
	if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
	return (1);
	return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
	}

	int
	spa_prev_software_version(spa_t *spa)
	{
	return (spa->spa_prev_software_version);
	}

	uint64_t
	spa_deadman_synctime(spa_t *spa)
	{
	return (spa->spa_deadman_synctime);
	}

	spa_autotrim_t
	spa_get_autotrim(spa_t *spa)
	{
	return (spa->spa_autotrim);
	}

	uint64_t
	spa_deadman_ziotime(spa_t *spa)
	{
	return (spa->spa_deadman_ziotime);
	}

	uint64_t
	spa_get_deadman_failmode(spa_t *spa)
	{
	return (spa->spa_deadman_failmode);
	}

	void
	spa_set_deadman_failmode(spa_t spa, const char failmode)
	{
	if (strcmp(failmode, "wait") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
	else if (strcmp(failmode, "continue") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE;
	else if (strcmp(failmode, "panic") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
	else
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
	}

	void
	spa_set_deadman_ziotime(hrtime_t ns)
	{
	spa_t *spa = NULL;

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa->spa_deadman_ziotime = ns;
	mutex_exit(&spa_namespace_lock);
	}
	}

	void
	spa_set_deadman_synctime(hrtime_t ns)
	{
	spa_t *spa = NULL;

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa->spa_deadman_synctime = ns;
	mutex_exit(&spa_namespace_lock);
	}
	}

	uint64_t
	dva_get_dsize_sync(spa_t spa, const dva_t dva)
	{
	uint64_t asize = DVA_GET_ASIZE(dva);
	uint64_t dsize = asize;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	if (asize != 0 && spa->spa_deflate) {
	vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
	if (vd != NULL)
	dsize = (asize >> SPA_MINBLOCKSHIFT) *
	vd->vdev_deflate_ratio;
	}

	return (dsize);
	}

	uint64_t
	bp_get_dsize_sync(spa_t spa, const blkptr_t bp)
	{
	uint64_t dsize = 0;

	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
	dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);

	return (dsize);
	}

	uint64_t
	bp_get_dsize(spa_t spa, const blkptr_t bp)
	{
	uint64_t dsize = 0;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
	dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);

	spa_config_exit(spa, SCL_VDEV, FTAG);

	return (dsize);
	}

	uint64_t
	spa_dirty_data(spa_t *spa)
	{
	return (spa->spa_dsl_pool->dp_dirty_total);
	}

	/*
	* ==========================================================================
	* SPA Import Progress Routines
	* ==========================================================================
	*/

	typedef struct spa_import_progress {
	uint64_t pool_guid; /* unique id for updates */
	char *pool_name;
	spa_load_state_t spa_load_state;
	uint64_t mmp_sec_remaining; /* MMP activity check */
	uint64_t spa_load_max_txg; /* rewind txg */
	procfs_list_node_t smh_node;
	} spa_import_progress_t;

	spa_history_list_t *spa_import_progress_list = NULL;

	static int
	spa_import_progress_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-20s %-14s %-14s %-12s %s\n", "pool_guid",
	"load_state", "multihost_secs", "max_txg",
	"pool_name");
	return (0);
	}

	static int
	spa_import_progress_show(struct seq_file f, void data)
	{
	spa_import_progress_t sip = (spa_import_progress_t )data;

	seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %s\n",
	(u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state,
	(u_longlong_t)sip->mmp_sec_remaining,
	(u_longlong_t)sip->spa_load_max_txg,
	(sip->pool_name ? sip->pool_name : "-"));

	return (0);
	}

	/* Remove oldest elements from list until there are no more than 'size' left */
	static void
	spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size)
	{
	spa_import_progress_t *sip;
	while (shl->size > size) {
	sip = list_remove_head(&shl->procfs_list.pl_list);
	if (sip->pool_name)
	spa_strfree(sip->pool_name);
	kmem_free(sip, sizeof (spa_import_progress_t));
	shl->size--;
	}

	IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list));
	}

	static void
	spa_import_progress_init(void)
	{
	spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t),
	KM_SLEEP);

	spa_import_progress_list->size = 0;

	spa_import_progress_list->procfs_list.pl_private =
	spa_import_progress_list;

	procfs_list_install("zfs",
	NULL,
	"import_progress",
	0644,
	&spa_import_progress_list->procfs_list,
	spa_import_progress_show,
	spa_import_progress_show_header,
	NULL,
	offsetof(spa_import_progress_t, smh_node));
	}

	static void
	spa_import_progress_destroy(void)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	procfs_list_uninstall(&shl->procfs_list);
	spa_import_progress_truncate(shl, 0);
	procfs_list_destroy(&shl->procfs_list);
	kmem_free(shl, sizeof (spa_history_list_t));
	}

	int
	spa_import_progress_set_state(uint64_t pool_guid,
	spa_load_state_t load_state)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->spa_load_state = load_state;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	int
	spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->spa_load_max_txg = load_max_txg;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	int
	spa_import_progress_set_mmp_check(uint64_t pool_guid,
	uint64_t mmp_sec_remaining)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->mmp_sec_remaining = mmp_sec_remaining;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	/*
	* A new import is in progress, add an entry.
	*/
	void
	spa_import_progress_add(spa_t *spa)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	char *poolname = NULL;

	sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP);
	sip->pool_guid = spa_guid(spa);

	(void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME,
	&poolname);
	if (poolname == NULL)
	poolname = spa_name(spa);
	sip->pool_name = spa_strdup(poolname);
	sip->spa_load_state = spa_load_state(spa);

	mutex_enter(&shl->procfs_list.pl_lock);
	procfs_list_add(&shl->procfs_list, sip);
	shl->size++;
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	void
	spa_import_progress_remove(uint64_t pool_guid)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	if (sip->pool_name)
	spa_strfree(sip->pool_name);
	list_remove(&shl->procfs_list.pl_list, sip);
	shl->size--;
	kmem_free(sip, sizeof (spa_import_progress_t));
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	/*
	* ==========================================================================
	* Initialization and Termination
	* ==========================================================================
	*/

	static int
	spa_name_compare(const void a1, const void a2)
	{
	const spa_t *s1 = a1;
	const spa_t *s2 = a2;
	int s;

	s = strcmp(s1->spa_name, s2->spa_name);

	return (TREE_ISIGN(s));
	}

	void
	spa_boot_init(void)
	{
	spa_config_load();
	}

	void
	spa_init(spa_mode_t mode)
	{
	mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);

	avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
	offsetof(spa_t, spa_avl));

	avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
	offsetof(spa_aux_t, aux_avl));

	avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
	offsetof(spa_aux_t, aux_avl));

	spa_mode_global = mode;

	#ifndef _KERNEL
	if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) {
	struct sigaction sa;

	sa.sa_flags = SA_SIGINFO;
	sigemptyset(&sa.sa_mask);
	sa.sa_sigaction = arc_buf_sigsegv;

	if (sigaction(SIGSEGV, &sa, NULL) == -1) {
	perror("could not enable watchpoints: "
	"sigaction(SIGSEGV, ...) = ");
	} else {
	arc_watch = B_TRUE;
	}
	}
	#endif

	fm_init();
	zfs_refcount_init();
	unique_init();
	zfs_btree_init();
	metaslab_stat_init();
	ddt_init();
	zio_init();
	dmu_init();
	zil_init();
	vdev_cache_stat_init();
	vdev_mirror_stat_init();
	vdev_raidz_math_init();
	vdev_file_init();
	zfs_prop_init();
	chksum_init();
	zpool_prop_init();
	zpool_feature_init();
	spa_config_load();
	vdev_prop_init();
	l2arc_start();
	scan_init();
	qat_init();
	spa_import_progress_init();
	}

	void
	spa_fini(void)
	{
	l2arc_stop();

	spa_evict_all();

	vdev_file_fini();
	vdev_cache_stat_fini();
	vdev_mirror_stat_fini();
	vdev_raidz_math_fini();
	chksum_fini();
	zil_fini();
	dmu_fini();
	zio_fini();
	ddt_fini();
	metaslab_stat_fini();
	zfs_btree_fini();
	unique_fini();
	zfs_refcount_fini();
	fm_fini();
	scan_fini();
	qat_fini();
	spa_import_progress_destroy();

	avl_destroy(&spa_namespace_avl);
	avl_destroy(&spa_spare_avl);
	avl_destroy(&spa_l2cache_avl);

	cv_destroy(&spa_namespace_cv);
	mutex_destroy(&spa_namespace_lock);
	mutex_destroy(&spa_spare_lock);
	mutex_destroy(&spa_l2cache_lock);
	}

	/*
	* Return whether this pool has a dedicated slog device. No locking needed.
	* It's not a problem if the wrong answer is returned as it's only for
	* performance and not correctness.
	*/
	boolean_t
	spa_has_slogs(spa_t *spa)
	{
	return (spa->spa_log_class->mc_groups != 0);
	}

	spa_log_state_t
	spa_get_log_state(spa_t *spa)
	{
	return (spa->spa_log_state);
	}

	void
	spa_set_log_state(spa_t *spa, spa_log_state_t state)
	{
	spa->spa_log_state = state;
	}

	boolean_t
	spa_is_root(spa_t *spa)
	{
	return (spa->spa_is_root);
	}

	boolean_t
	spa_writeable(spa_t *spa)
	{
	return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config);
	}

	/*
	* Returns true if there is a pending sync task in any of the current
	* syncing txg, the current quiescing txg, or the current open txg.
	*/
	boolean_t
	spa_has_pending_synctask(spa_t *spa)
	{
	return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) \|\|
	!txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks));
	}

	spa_mode_t
	spa_mode(spa_t *spa)
	{
	return (spa->spa_mode);
	}

	uint64_t
	spa_bootfs(spa_t *spa)
	{
	return (spa->spa_bootfs);
	}

	uint64_t
	spa_delegation(spa_t *spa)
	{
	return (spa->spa_delegation);
	}

	objset_t *
	spa_meta_objset(spa_t *spa)
	{
	return (spa->spa_meta_objset);
	}

	enum zio_checksum
	spa_dedup_checksum(spa_t *spa)
	{
	return (spa->spa_dedup_checksum);
	}

	/*
	* Reset pool scan stat per scan pass (or reboot).
	*/
	void
	spa_scan_stat_init(spa_t *spa)
	{
	/* data not stored on disk */
	spa->spa_scan_pass_start = gethrestime_sec();
	if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan))
	spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start;
	else
	spa->spa_scan_pass_scrub_pause = 0;
	spa->spa_scan_pass_scrub_spent_paused = 0;
	spa->spa_scan_pass_exam = 0;
	spa->spa_scan_pass_issued = 0;
	vdev_scan_stat_init(spa->spa_root_vdev);
	}

	/*
	* Get scan stats for zpool status reports
	*/
	int
	spa_scan_get_stats(spa_t spa, pool_scan_stat_t ps)
	{
	dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL;

	if (scn == NULL \|\| scn->scn_phys.scn_func == POOL_SCAN_NONE)
	return (SET_ERROR(ENOENT));
	memset(ps, 0, sizeof (pool_scan_stat_t));

	/* data stored on disk */
	ps->pss_func = scn->scn_phys.scn_func;
	ps->pss_state = scn->scn_phys.scn_state;
	ps->pss_start_time = scn->scn_phys.scn_start_time;
	ps->pss_end_time = scn->scn_phys.scn_end_time;
	ps->pss_to_examine = scn->scn_phys.scn_to_examine;
	ps->pss_examined = scn->scn_phys.scn_examined;
	ps->pss_to_process = scn->scn_phys.scn_to_process;
	ps->pss_processed = scn->scn_phys.scn_processed;
	ps->pss_errors = scn->scn_phys.scn_errors;

	/* data not stored on disk */
	ps->pss_pass_exam = spa->spa_scan_pass_exam;
	ps->pss_pass_start = spa->spa_scan_pass_start;
	ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause;
	ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused;
	ps->pss_pass_issued = spa->spa_scan_pass_issued;
	ps->pss_issued =
	scn->scn_issued_before_pass + spa->spa_scan_pass_issued;

	return (0);
	}

	int
	spa_maxblocksize(spa_t *spa)
	{
	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
	return (SPA_MAXBLOCKSIZE);
	else
	return (SPA_OLD_MAXBLOCKSIZE);
	}


	/*
	* Returns the txg that the last device removal completed. No indirect mappings
	* have been added since this txg.
	*/
	uint64_t
	spa_get_last_removal_txg(spa_t *spa)
	{
	uint64_t vdevid;
	uint64_t ret = -1ULL;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	/*
	* sr_prev_indirect_vdev is only modified while holding all the
	* config locks, so it is sufficient to hold SCL_VDEV as reader when
	* examining it.
	*/
	vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev;

	while (vdevid != -1ULL) {
	vdev_t *vd = vdev_lookup_top(spa, vdevid);
	vdev_indirect_births_t *vib = vd->vdev_indirect_births;

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	/*
	* If the removal did not remap any data, we don't care.
	*/
	if (vdev_indirect_births_count(vib) != 0) {
	ret = vdev_indirect_births_last_entry_txg(vib);
	break;
	}

	vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev;
	}
	spa_config_exit(spa, SCL_VDEV, FTAG);

	IMPLY(ret != -1ULL,
	spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));

	return (ret);
	}

	int
	spa_maxdnodesize(spa_t *spa)
	{
	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
	return (DNODE_MAX_SIZE);
	else
	return (DNODE_MIN_SIZE);
	}

	boolean_t
	spa_multihost(spa_t *spa)
	{
	return (spa->spa_multihost ? B_TRUE : B_FALSE);
	}

	uint32_t
	spa_get_hostid(spa_t *spa)
	{
	return (spa->spa_hostid);
	}

	boolean_t
	spa_trust_config(spa_t *spa)
	{
	return (spa->spa_trust_config);
	}

	uint64_t
	spa_missing_tvds_allowed(spa_t *spa)
	{
	return (spa->spa_missing_tvds_allowed);
	}

	space_map_t *
	spa_syncing_log_sm(spa_t *spa)
	{
	return (spa->spa_syncing_log_sm);
	}

	void
	spa_set_missing_tvds(spa_t *spa, uint64_t missing)
	{
	spa->spa_missing_tvds = missing;
	}

	/*
	* Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc).
	*/
	const char *
	spa_state_to_name(spa_t *spa)
	{
	ASSERT3P(spa, !=, NULL);

	/*
	* it is possible for the spa to exist, without root vdev
	* as the spa transitions during import/export
	*/
	vdev_t *rvd = spa->spa_root_vdev;
	if (rvd == NULL) {
	return ("TRANSITIONING");
	}
	vdev_state_t state = rvd->vdev_state;
	vdev_aux_t aux = rvd->vdev_stat.vs_aux;

	if (spa_suspended(spa) &&
	(spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE))
	return ("SUSPENDED");

	switch (state) {
	case VDEV_STATE_CLOSED:
	case VDEV_STATE_OFFLINE:
	return ("OFFLINE");
	case VDEV_STATE_REMOVED:
	return ("REMOVED");
	case VDEV_STATE_CANT_OPEN:
	if (aux == VDEV_AUX_CORRUPT_DATA \|\| aux == VDEV_AUX_BAD_LOG)
	return ("FAULTED");
	else if (aux == VDEV_AUX_SPLIT_POOL)
	return ("SPLIT");
	else
	return ("UNAVAIL");
	case VDEV_STATE_FAULTED:
	return ("FAULTED");
	case VDEV_STATE_DEGRADED:
	return ("DEGRADED");
	case VDEV_STATE_HEALTHY:
	return ("ONLINE");
	default:
	break;
	}

	return ("UNKNOWN");
	}

	boolean_t
	spa_top_vdevs_spacemap_addressable(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	if (!vdev_is_spacemap_addressable(rvd->vdev_child[c]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	boolean_t
	spa_has_checkpoint(spa_t *spa)
	{
	return (spa->spa_checkpoint_txg != 0);
	}

	boolean_t
	spa_importing_readonly_checkpoint(spa_t *spa)
	{
	return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) &&
	spa->spa_mode == SPA_MODE_READ);
	}

	uint64_t
	spa_min_claim_txg(spa_t *spa)
	{
	uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg;

	if (checkpoint_txg != 0)
	return (checkpoint_txg + 1);

	return (spa->spa_first_txg);
	}

	/*
	* If there is a checkpoint, async destroys may consume more space from
	* the pool instead of freeing it. In an attempt to save the pool from
	* getting suspended when it is about to run out of space, we stop
	* processing async destroys.
	*/
	boolean_t
	spa_suspend_async_destroy(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);

	uint64_t unreserved = dsl_pool_unreserved_space(dp,
	ZFS_SPACE_CHECK_EXTRA_RESERVED);
	uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes;
	uint64_t avail = (unreserved > used) ? (unreserved - used) : 0;

	if (spa_has_checkpoint(spa) && avail == 0)
	return (B_TRUE);

	return (B_FALSE);
	}

	#if defined(_KERNEL)

	int
	param_set_deadman_failmode_common(const char *val)
	{
	spa_t *spa = NULL;
	char *p;

	if (val == NULL)
	return (SET_ERROR(EINVAL));

	if ((p = strchr(val, '\n')) != NULL)
	*p = '\0';

	if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 &&
	strcmp(val, "panic"))
	return (SET_ERROR(EINVAL));

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa_set_deadman_failmode(spa, val);
	mutex_exit(&spa_namespace_lock);
	}

	return (0);
	}
	#endif

	/* Namespace manipulation */
	EXPORT_SYMBOL(spa_lookup);
	EXPORT_SYMBOL(spa_add);
	EXPORT_SYMBOL(spa_remove);
	EXPORT_SYMBOL(spa_next);

	/* Refcount functions */
	EXPORT_SYMBOL(spa_open_ref);
	EXPORT_SYMBOL(spa_close);
	EXPORT_SYMBOL(spa_refcount_zero);

	/* Pool configuration lock */
	EXPORT_SYMBOL(spa_config_tryenter);
	EXPORT_SYMBOL(spa_config_enter);
	EXPORT_SYMBOL(spa_config_exit);
	EXPORT_SYMBOL(spa_config_held);

	/* Pool vdev add/remove lock */
	EXPORT_SYMBOL(spa_vdev_enter);
	EXPORT_SYMBOL(spa_vdev_exit);

	/* Pool vdev state change lock */
	EXPORT_SYMBOL(spa_vdev_state_enter);
	EXPORT_SYMBOL(spa_vdev_state_exit);

	/* Accessor functions */
	EXPORT_SYMBOL(spa_shutting_down);
	EXPORT_SYMBOL(spa_get_dsl);
	EXPORT_SYMBOL(spa_get_rootblkptr);
	EXPORT_SYMBOL(spa_set_rootblkptr);
	EXPORT_SYMBOL(spa_altroot);
	EXPORT_SYMBOL(spa_sync_pass);
	EXPORT_SYMBOL(spa_name);
	EXPORT_SYMBOL(spa_guid);
	EXPORT_SYMBOL(spa_last_synced_txg);
	EXPORT_SYMBOL(spa_first_txg);
	EXPORT_SYMBOL(spa_syncing_txg);
	EXPORT_SYMBOL(spa_version);
	EXPORT_SYMBOL(spa_state);
	EXPORT_SYMBOL(spa_load_state);
	EXPORT_SYMBOL(spa_freeze_txg);
	EXPORT_SYMBOL(spa_get_dspace);
	EXPORT_SYMBOL(spa_update_dspace);
	EXPORT_SYMBOL(spa_deflate);
	EXPORT_SYMBOL(spa_normal_class);
	EXPORT_SYMBOL(spa_log_class);
	EXPORT_SYMBOL(spa_special_class);
	EXPORT_SYMBOL(spa_preferred_class);
	EXPORT_SYMBOL(spa_max_replication);
	EXPORT_SYMBOL(spa_prev_software_version);
	EXPORT_SYMBOL(spa_get_failmode);
	EXPORT_SYMBOL(spa_suspended);
	EXPORT_SYMBOL(spa_bootfs);
	EXPORT_SYMBOL(spa_delegation);
	EXPORT_SYMBOL(spa_meta_objset);
	EXPORT_SYMBOL(spa_maxblocksize);
	EXPORT_SYMBOL(spa_maxdnodesize);

	/* Miscellaneous support routines */
	EXPORT_SYMBOL(spa_guid_exists);
	EXPORT_SYMBOL(spa_strdup);
	EXPORT_SYMBOL(spa_strfree);
	EXPORT_SYMBOL(spa_generate_guid);
	EXPORT_SYMBOL(snprintf_blkptr);
	EXPORT_SYMBOL(spa_freeze);
	EXPORT_SYMBOL(spa_upgrade);
	EXPORT_SYMBOL(spa_evict_all);
	EXPORT_SYMBOL(spa_lookup_by_guid);
	EXPORT_SYMBOL(spa_has_spare);
	EXPORT_SYMBOL(dva_get_dsize_sync);
	EXPORT_SYMBOL(bp_get_dsize_sync);
	EXPORT_SYMBOL(bp_get_dsize);
	EXPORT_SYMBOL(spa_has_slogs);
	EXPORT_SYMBOL(spa_is_root);
	EXPORT_SYMBOL(spa_writeable);
	EXPORT_SYMBOL(spa_mode);
	EXPORT_SYMBOL(spa_namespace_lock);
	EXPORT_SYMBOL(spa_trust_config);
	EXPORT_SYMBOL(spa_missing_tvds_allowed);
	EXPORT_SYMBOL(spa_set_missing_tvds);
	EXPORT_SYMBOL(spa_state_to_name);
	EXPORT_SYMBOL(spa_importing_readonly_checkpoint);
	EXPORT_SYMBOL(spa_min_claim_txg);
	EXPORT_SYMBOL(spa_suspend_async_destroy);
	EXPORT_SYMBOL(spa_has_checkpoint);
	EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable);

	ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW,
	"Set additional debugging flags");

	ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW,
	"Set to attempt to recover from fatal errors");

	ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW,
	"Set to ignore IO errors during free and permanently leak the space");

	ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, ULONG, ZMOD_RW,
	"Dead I/O check interval in milliseconds");

	ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW,
	"Enable deadman timer");

	-ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, UINT, ZMOD_RW,
	"SPA size estimate multiplication factor");

	ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW,
	"Place DDT data into the special class");

	ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW,
	"Place user data indirect blocks into the special class");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode,
	param_set_deadman_failmode, param_get_charp, ZMOD_RW,
	"Failmode for deadman timer");

	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms,
	param_set_deadman_synctime, param_get_ulong, ZMOD_RW,
	"Pool sync expiration time in milliseconds");

	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms,
	param_set_deadman_ziotime, param_get_ulong, ZMOD_RW,
	"IO expiration time in milliseconds");

	-ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, UINT, ZMOD_RW,
	"Small file blocks in special vdevs depends on this much "
	"free space available");
	/* END CSTYLED */

	ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift,
	- param_get_int, ZMOD_RW, "Reserved free space in pool");
	+ param_get_uint, ZMOD_RW, "Reserved free space in pool");
	diff --git a/sys/contrib/openzfs/module/zfs/spa_stats.c b/sys/contrib/openzfs/module/zfs/spa_stats.c
	index 59844d5d6491..17ed2a620b1e 100644
	--- a/sys/contrib/openzfs/module/zfs/spa_stats.c
	+++ b/sys/contrib/openzfs/module/zfs/spa_stats.c
	@@ -1,1025 +1,1025 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/spa.h>
	#include <zfs_comutil.h>

	/*
	* Keeps stats on last N reads per spa_t, disabled by default.
	*/
	-static int zfs_read_history = B_FALSE;
	+static uint_t zfs_read_history = B_FALSE;

	/*
	* Include cache hits in history, disabled by default.
	*/
	static int zfs_read_history_hits = B_FALSE;

	/*
	* Keeps stats on the last 100 txgs by default.
	*/
	-static int zfs_txg_history = 100;
	+static uint_t zfs_txg_history = 100;

	/*
	* Keeps stats on the last N MMP updates, disabled by default.
	*/
	-int zfs_multihost_history = B_FALSE;
	+static uint_t zfs_multihost_history = B_FALSE;

	/*
	* ==========================================================================
	* SPA Read History Routines
	* ==========================================================================
	*/

	/*
	* Read statistics - Information exported regarding each arc_read call
	*/
	typedef struct spa_read_history {
	hrtime_t start; /* time read completed */
	uint64_t objset; /* read from this objset */
	uint64_t object; /* read of this object number */
	uint64_t level; /* block's indirection level */
	uint64_t blkid; /* read of this block id */
	char origin[24]; /* read originated from here */
	uint32_t aflags; /* ARC flags (cached, prefetch, etc.) */
	pid_t pid; /* PID of task doing read */
	char comm[16]; /* process name of task doing read */
	procfs_list_node_t srh_node;
	} spa_read_history_t;

	static int
	spa_read_history_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-8s %-16s %-8s %-8s %-8s %-8s %-8s "
	"%-24s %-8s %-16s\n", "UID", "start", "objset", "object",
	"level", "blkid", "aflags", "origin", "pid", "process");

	return (0);
	}

	static int
	spa_read_history_show(struct seq_file f, void data)
	{
	spa_read_history_t srh = (spa_read_history_t )data;

	seq_printf(f, "%-8llu %-16llu 0x%-6llx "
	"%-8lli %-8lli %-8lli 0x%-6x %-24s %-8i %-16s\n",
	(u_longlong_t)srh->srh_node.pln_id, srh->start,
	(longlong_t)srh->objset, (longlong_t)srh->object,
	(longlong_t)srh->level, (longlong_t)srh->blkid,
	srh->aflags, srh->origin, srh->pid, srh->comm);

	return (0);
	}

	/* Remove oldest elements from list until there are no more than 'size' left */
	static void
	spa_read_history_truncate(spa_history_list_t *shl, unsigned int size)
	{
	spa_read_history_t *srh;
	while (shl->size > size) {
	srh = list_remove_head(&shl->procfs_list.pl_list);
	ASSERT3P(srh, !=, NULL);
	kmem_free(srh, sizeof (spa_read_history_t));
	shl->size--;
	}

	if (size == 0)
	ASSERT(list_is_empty(&shl->procfs_list.pl_list));
	}

	static int
	spa_read_history_clear(procfs_list_t *procfs_list)
	{
	spa_history_list_t *shl = procfs_list->pl_private;
	mutex_enter(&procfs_list->pl_lock);
	spa_read_history_truncate(shl, 0);
	mutex_exit(&procfs_list->pl_lock);
	return (0);
	}

	static void
	spa_read_history_init(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.read_history;

	shl->size = 0;
	shl->procfs_list.pl_private = shl;
	procfs_list_install("zfs",
	spa_name(spa),
	"reads",
	0600,
	&shl->procfs_list,
	spa_read_history_show,
	spa_read_history_show_header,
	spa_read_history_clear,
	offsetof(spa_read_history_t, srh_node));
	}

	static void
	spa_read_history_destroy(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.read_history;
	procfs_list_uninstall(&shl->procfs_list);
	spa_read_history_truncate(shl, 0);
	procfs_list_destroy(&shl->procfs_list);
	}

	void
	spa_read_history_add(spa_t spa, const zbookmark_phys_t zb, uint32_t aflags)
	{
	spa_history_list_t *shl = &spa->spa_stats.read_history;
	spa_read_history_t *srh;

	ASSERT3P(spa, !=, NULL);
	ASSERT3P(zb, !=, NULL);

	if (zfs_read_history == 0 && shl->size == 0)
	return;

	if (zfs_read_history_hits == 0 && (aflags & ARC_FLAG_CACHED))
	return;

	srh = kmem_zalloc(sizeof (spa_read_history_t), KM_SLEEP);
	strlcpy(srh->comm, getcomm(), sizeof (srh->comm));
	srh->start = gethrtime();
	srh->objset = zb->zb_objset;
	srh->object = zb->zb_object;
	srh->level = zb->zb_level;
	srh->blkid = zb->zb_blkid;
	srh->aflags = aflags;
	srh->pid = getpid();

	mutex_enter(&shl->procfs_list.pl_lock);

	procfs_list_add(&shl->procfs_list, srh);
	shl->size++;

	spa_read_history_truncate(shl, zfs_read_history);

	mutex_exit(&shl->procfs_list.pl_lock);
	}

	/*
	* ==========================================================================
	* SPA TXG History Routines
	* ==========================================================================
	*/

	/*
	* Txg statistics - Information exported regarding each txg sync
	*/

	typedef struct spa_txg_history {
	uint64_t txg; /* txg id */
	txg_state_t state; /* active txg state */
	uint64_t nread; /* number of bytes read */
	uint64_t nwritten; /* number of bytes written */
	uint64_t reads; /* number of read operations */
	uint64_t writes; /* number of write operations */
	uint64_t ndirty; /* number of dirty bytes */
	hrtime_t times[TXG_STATE_COMMITTED]; /* completion times */
	procfs_list_node_t sth_node;
	} spa_txg_history_t;

	static int
	spa_txg_history_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-8s %-16s %-5s %-12s %-12s %-12s "
	"%-8s %-8s %-12s %-12s %-12s %-12s\n", "txg", "birth", "state",
	"ndirty", "nread", "nwritten", "reads", "writes",
	"otime", "qtime", "wtime", "stime");
	return (0);
	}

	static int
	spa_txg_history_show(struct seq_file f, void data)
	{
	spa_txg_history_t sth = (spa_txg_history_t )data;
	uint64_t open = 0, quiesce = 0, wait = 0, sync = 0;
	char state;

	switch (sth->state) {
	case TXG_STATE_BIRTH: state = 'B'; break;
	case TXG_STATE_OPEN: state = 'O'; break;
	case TXG_STATE_QUIESCED: state = 'Q'; break;
	case TXG_STATE_WAIT_FOR_SYNC: state = 'W'; break;
	case TXG_STATE_SYNCED: state = 'S'; break;
	case TXG_STATE_COMMITTED: state = 'C'; break;
	default: state = '?'; break;
	}

	if (sth->times[TXG_STATE_OPEN])
	open = sth->times[TXG_STATE_OPEN] -
	sth->times[TXG_STATE_BIRTH];

	if (sth->times[TXG_STATE_QUIESCED])
	quiesce = sth->times[TXG_STATE_QUIESCED] -
	sth->times[TXG_STATE_OPEN];

	if (sth->times[TXG_STATE_WAIT_FOR_SYNC])
	wait = sth->times[TXG_STATE_WAIT_FOR_SYNC] -
	sth->times[TXG_STATE_QUIESCED];

	if (sth->times[TXG_STATE_SYNCED])
	sync = sth->times[TXG_STATE_SYNCED] -
	sth->times[TXG_STATE_WAIT_FOR_SYNC];

	seq_printf(f, "%-8llu %-16llu %-5c %-12llu "
	"%-12llu %-12llu %-8llu %-8llu %-12llu %-12llu %-12llu %-12llu\n",
	(longlong_t)sth->txg, sth->times[TXG_STATE_BIRTH], state,
	(u_longlong_t)sth->ndirty,
	(u_longlong_t)sth->nread, (u_longlong_t)sth->nwritten,
	(u_longlong_t)sth->reads, (u_longlong_t)sth->writes,
	(u_longlong_t)open, (u_longlong_t)quiesce, (u_longlong_t)wait,
	(u_longlong_t)sync);

	return (0);
	}

	/* Remove oldest elements from list until there are no more than 'size' left */
	static void
	spa_txg_history_truncate(spa_history_list_t *shl, unsigned int size)
	{
	spa_txg_history_t *sth;
	while (shl->size > size) {
	sth = list_remove_head(&shl->procfs_list.pl_list);
	ASSERT3P(sth, !=, NULL);
	kmem_free(sth, sizeof (spa_txg_history_t));
	shl->size--;
	}

	if (size == 0)
	ASSERT(list_is_empty(&shl->procfs_list.pl_list));

	}

	static int
	spa_txg_history_clear(procfs_list_t *procfs_list)
	{
	spa_history_list_t *shl = procfs_list->pl_private;
	mutex_enter(&procfs_list->pl_lock);
	spa_txg_history_truncate(shl, 0);
	mutex_exit(&procfs_list->pl_lock);
	return (0);
	}

	static void
	spa_txg_history_init(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.txg_history;

	shl->size = 0;
	shl->procfs_list.pl_private = shl;
	procfs_list_install("zfs",
	spa_name(spa),
	"txgs",
	0644,
	&shl->procfs_list,
	spa_txg_history_show,
	spa_txg_history_show_header,
	spa_txg_history_clear,
	offsetof(spa_txg_history_t, sth_node));
	}

	static void
	spa_txg_history_destroy(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.txg_history;
	procfs_list_uninstall(&shl->procfs_list);
	spa_txg_history_truncate(shl, 0);
	procfs_list_destroy(&shl->procfs_list);
	}

	/*
	* Add a new txg to historical record.
	*/
	void
	spa_txg_history_add(spa_t *spa, uint64_t txg, hrtime_t birth_time)
	{
	spa_history_list_t *shl = &spa->spa_stats.txg_history;
	spa_txg_history_t *sth;

	if (zfs_txg_history == 0 && shl->size == 0)
	return;

	sth = kmem_zalloc(sizeof (spa_txg_history_t), KM_SLEEP);
	sth->txg = txg;
	sth->state = TXG_STATE_OPEN;
	sth->times[TXG_STATE_BIRTH] = birth_time;

	mutex_enter(&shl->procfs_list.pl_lock);
	procfs_list_add(&shl->procfs_list, sth);
	shl->size++;
	spa_txg_history_truncate(shl, zfs_txg_history);
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	/*
	* Set txg state completion time and increment current state.
	*/
	int
	spa_txg_history_set(spa_t *spa, uint64_t txg, txg_state_t completed_state,
	hrtime_t completed_time)
	{
	spa_history_list_t *shl = &spa->spa_stats.txg_history;
	spa_txg_history_t *sth;
	int error = ENOENT;

	if (zfs_txg_history == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sth = list_tail(&shl->procfs_list.pl_list); sth != NULL;
	sth = list_prev(&shl->procfs_list.pl_list, sth)) {
	if (sth->txg == txg) {
	sth->times[completed_state] = completed_time;
	sth->state++;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	/*
	* Set txg IO stats.
	*/
	static int
	spa_txg_history_set_io(spa_t *spa, uint64_t txg, uint64_t nread,
	uint64_t nwritten, uint64_t reads, uint64_t writes, uint64_t ndirty)
	{
	spa_history_list_t *shl = &spa->spa_stats.txg_history;
	spa_txg_history_t *sth;
	int error = ENOENT;

	if (zfs_txg_history == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sth = list_tail(&shl->procfs_list.pl_list); sth != NULL;
	sth = list_prev(&shl->procfs_list.pl_list, sth)) {
	if (sth->txg == txg) {
	sth->nread = nread;
	sth->nwritten = nwritten;
	sth->reads = reads;
	sth->writes = writes;
	sth->ndirty = ndirty;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	txg_stat_t *
	spa_txg_history_init_io(spa_t spa, uint64_t txg, dsl_pool_t dp)
	{
	txg_stat_t *ts;

	if (zfs_txg_history == 0)
	return (NULL);

	ts = kmem_alloc(sizeof (txg_stat_t), KM_SLEEP);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_get_stats(spa->spa_root_vdev, &ts->vs1);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	ts->txg = txg;
	ts->ndirty = dp->dp_dirty_pertxg[txg & TXG_MASK];

	spa_txg_history_set(spa, txg, TXG_STATE_WAIT_FOR_SYNC, gethrtime());

	return (ts);
	}

	void
	spa_txg_history_fini_io(spa_t spa, txg_stat_t ts)
	{
	if (ts == NULL)
	return;

	if (zfs_txg_history == 0) {
	kmem_free(ts, sizeof (txg_stat_t));
	return;
	}

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_get_stats(spa->spa_root_vdev, &ts->vs2);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	spa_txg_history_set(spa, ts->txg, TXG_STATE_SYNCED, gethrtime());
	spa_txg_history_set_io(spa, ts->txg,
	ts->vs2.vs_bytes[ZIO_TYPE_READ] - ts->vs1.vs_bytes[ZIO_TYPE_READ],
	ts->vs2.vs_bytes[ZIO_TYPE_WRITE] - ts->vs1.vs_bytes[ZIO_TYPE_WRITE],
	ts->vs2.vs_ops[ZIO_TYPE_READ] - ts->vs1.vs_ops[ZIO_TYPE_READ],
	ts->vs2.vs_ops[ZIO_TYPE_WRITE] - ts->vs1.vs_ops[ZIO_TYPE_WRITE],
	ts->ndirty);

	kmem_free(ts, sizeof (txg_stat_t));
	}

	/*
	* ==========================================================================
	* SPA TX Assign Histogram Routines
	* ==========================================================================
	*/

	/*
	* Tx statistics - Information exported regarding dmu_tx_assign time.
	*/

	/*
	* When the kstat is written zero all buckets. When the kstat is read
	* count the number of trailing buckets set to zero and update ks_ndata
	* such that they are not output.
	*/
	static int
	spa_tx_assign_update(kstat_t *ksp, int rw)
	{
	spa_t *spa = ksp->ks_private;
	spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
	int i;

	if (rw == KSTAT_WRITE) {
	for (i = 0; i < shk->count; i++)
	((kstat_named_t *)shk->priv)[i].value.ui64 = 0;
	}

	for (i = shk->count; i > 0; i--)
	if (((kstat_named_t *)shk->priv)[i-1].value.ui64 != 0)
	break;

	ksp->ks_ndata = i;
	ksp->ks_data_size = i * sizeof (kstat_named_t);

	return (0);
	}

	static void
	spa_tx_assign_init(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
	char *name;
	kstat_named_t *ks;
	kstat_t *ksp;
	int i;

	mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);

	shk->count = 42; /* power of two buckets for 1ns to 2,199s */
	shk->size = shk->count * sizeof (kstat_named_t);
	shk->priv = kmem_alloc(shk->size, KM_SLEEP);

	name = kmem_asprintf("zfs/%s", spa_name(spa));

	for (i = 0; i < shk->count; i++) {
	ks = &((kstat_named_t *)shk->priv)[i];
	ks->data_type = KSTAT_DATA_UINT64;
	ks->value.ui64 = 0;
	(void) snprintf(ks->name, KSTAT_STRLEN, "%llu ns",
	(u_longlong_t)1 << i);
	}

	ksp = kstat_create(name, 0, "dmu_tx_assign", "misc",
	KSTAT_TYPE_NAMED, 0, KSTAT_FLAG_VIRTUAL);
	shk->kstat = ksp;

	if (ksp) {
	ksp->ks_lock = &shk->lock;
	ksp->ks_data = shk->priv;
	ksp->ks_ndata = shk->count;
	ksp->ks_data_size = shk->size;
	ksp->ks_private = spa;
	ksp->ks_update = spa_tx_assign_update;
	kstat_install(ksp);
	}
	kmem_strfree(name);
	}

	static void
	spa_tx_assign_destroy(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
	kstat_t *ksp;

	ksp = shk->kstat;
	if (ksp)
	kstat_delete(ksp);

	kmem_free(shk->priv, shk->size);
	mutex_destroy(&shk->lock);
	}

	void
	spa_tx_assign_add_nsecs(spa_t *spa, uint64_t nsecs)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.tx_assign_histogram;
	uint64_t idx = 0;

	while (((1ULL << idx) < nsecs) && (idx < shk->size - 1))
	idx++;

	atomic_inc_64(&((kstat_named_t *)shk->priv)[idx].value.ui64);
	}

	/*
	* ==========================================================================
	* SPA MMP History Routines
	* ==========================================================================
	*/

	/*
	* MMP statistics - Information exported regarding attempted MMP writes
	* For MMP writes issued, fields used as per comments below.
	* For MMP writes skipped, an entry represents a span of time when
	* writes were skipped for same reason (error from mmp_random_leaf).
	* Differences are:
	* timestamp time first write skipped, if >1 skipped in a row
	* mmp_delay delay value at timestamp
	* vdev_guid number of writes skipped
	* io_error one of enum mmp_error
	* duration time span (ns) of skipped writes
	*/

	typedef struct spa_mmp_history {
	uint64_t mmp_node_id; /* unique # for updates */
	uint64_t txg; /* txg of last sync */
	uint64_t timestamp; /* UTC time MMP write issued */
	uint64_t mmp_delay; /* mmp_thread.mmp_delay at timestamp */
	uint64_t vdev_guid; /* unique ID of leaf vdev */
	char *vdev_path;
	int vdev_label; /* vdev label */
	int io_error; /* error status of MMP write */
	hrtime_t error_start; /* hrtime of start of error period */
	hrtime_t duration; /* time from submission to completion */
	procfs_list_node_t smh_node;
	} spa_mmp_history_t;

	static int
	spa_mmp_history_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-10s %-10s %-10s %-6s %-10s %-12s %-24s "
	"%-10s %s\n", "id", "txg", "timestamp", "error", "duration",
	"mmp_delay", "vdev_guid", "vdev_label", "vdev_path");
	return (0);
	}

	static int
	spa_mmp_history_show(struct seq_file f, void data)
	{
	spa_mmp_history_t smh = (spa_mmp_history_t )data;
	char skip_fmt[] = "%-10llu %-10llu %10llu %#6llx %10lld %12llu %-24llu "
	"%-10lld %s\n";
	char write_fmt[] = "%-10llu %-10llu %10llu %6lld %10lld %12llu %-24llu "
	"%-10lld %s\n";

	seq_printf(f, (smh->error_start ? skip_fmt : write_fmt),
	(u_longlong_t)smh->mmp_node_id, (u_longlong_t)smh->txg,
	(u_longlong_t)smh->timestamp, (longlong_t)smh->io_error,
	(longlong_t)smh->duration, (u_longlong_t)smh->mmp_delay,
	(u_longlong_t)smh->vdev_guid, (u_longlong_t)smh->vdev_label,
	(smh->vdev_path ? smh->vdev_path : "-"));

	return (0);
	}

	/* Remove oldest elements from list until there are no more than 'size' left */
	static void
	spa_mmp_history_truncate(spa_history_list_t *shl, unsigned int size)
	{
	spa_mmp_history_t *smh;
	while (shl->size > size) {
	smh = list_remove_head(&shl->procfs_list.pl_list);
	if (smh->vdev_path)
	kmem_strfree(smh->vdev_path);
	kmem_free(smh, sizeof (spa_mmp_history_t));
	shl->size--;
	}

	if (size == 0)
	ASSERT(list_is_empty(&shl->procfs_list.pl_list));

	}

	static int
	spa_mmp_history_clear(procfs_list_t *procfs_list)
	{
	spa_history_list_t *shl = procfs_list->pl_private;
	mutex_enter(&procfs_list->pl_lock);
	spa_mmp_history_truncate(shl, 0);
	mutex_exit(&procfs_list->pl_lock);
	return (0);
	}

	static void
	spa_mmp_history_init(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.mmp_history;

	shl->size = 0;

	shl->procfs_list.pl_private = shl;
	procfs_list_install("zfs",
	spa_name(spa),
	"multihost",
	0644,
	&shl->procfs_list,
	spa_mmp_history_show,
	spa_mmp_history_show_header,
	spa_mmp_history_clear,
	offsetof(spa_mmp_history_t, smh_node));
	}

	static void
	spa_mmp_history_destroy(spa_t *spa)
	{
	spa_history_list_t *shl = &spa->spa_stats.mmp_history;
	procfs_list_uninstall(&shl->procfs_list);
	spa_mmp_history_truncate(shl, 0);
	procfs_list_destroy(&shl->procfs_list);
	}

	/*
	* Set duration in existing "skip" record to how long we have waited for a leaf
	* vdev to become available.
	*
	* Important that we start search at the tail of the list where new
	* records are inserted, so this is normally an O(1) operation.
	*/
	int
	spa_mmp_history_set_skip(spa_t *spa, uint64_t mmp_node_id)
	{
	spa_history_list_t *shl = &spa->spa_stats.mmp_history;
	spa_mmp_history_t *smh;
	int error = ENOENT;

	if (zfs_multihost_history == 0 && shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (smh = list_tail(&shl->procfs_list.pl_list); smh != NULL;
	smh = list_prev(&shl->procfs_list.pl_list, smh)) {
	if (smh->mmp_node_id == mmp_node_id) {
	ASSERT3U(smh->io_error, !=, 0);
	smh->duration = gethrtime() - smh->error_start;
	smh->vdev_guid++;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	/*
	* Set MMP write duration and error status in existing record.
	* See comment re: search order above spa_mmp_history_set_skip().
	*/
	int
	spa_mmp_history_set(spa_t *spa, uint64_t mmp_node_id, int io_error,
	hrtime_t duration)
	{
	spa_history_list_t *shl = &spa->spa_stats.mmp_history;
	spa_mmp_history_t *smh;
	int error = ENOENT;

	if (zfs_multihost_history == 0 && shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (smh = list_tail(&shl->procfs_list.pl_list); smh != NULL;
	smh = list_prev(&shl->procfs_list.pl_list, smh)) {
	if (smh->mmp_node_id == mmp_node_id) {
	ASSERT(smh->io_error == 0);
	smh->io_error = io_error;
	smh->duration = duration;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	/*
	* Add a new MMP historical record.
	* error == 0 : a write was issued.
	* error != 0 : a write was not issued because no leaves were found.
	*/
	void
	spa_mmp_history_add(spa_t *spa, uint64_t txg, uint64_t timestamp,
	uint64_t mmp_delay, vdev_t *vd, int label, uint64_t mmp_node_id,
	int error)
	{
	spa_history_list_t *shl = &spa->spa_stats.mmp_history;
	spa_mmp_history_t *smh;

	if (zfs_multihost_history == 0 && shl->size == 0)
	return;

	smh = kmem_zalloc(sizeof (spa_mmp_history_t), KM_SLEEP);
	smh->txg = txg;
	smh->timestamp = timestamp;
	smh->mmp_delay = mmp_delay;
	if (vd) {
	smh->vdev_guid = vd->vdev_guid;
	if (vd->vdev_path)
	smh->vdev_path = kmem_strdup(vd->vdev_path);
	}
	smh->vdev_label = label;
	smh->mmp_node_id = mmp_node_id;

	if (error) {
	smh->io_error = error;
	smh->error_start = gethrtime();
	smh->vdev_guid = 1;
	}

	mutex_enter(&shl->procfs_list.pl_lock);
	procfs_list_add(&shl->procfs_list, smh);
	shl->size++;
	spa_mmp_history_truncate(shl, zfs_multihost_history);
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	static void *
	spa_state_addr(kstat_t *ksp, loff_t n)
	{
	if (n == 0)
	return (ksp->ks_private); /* return the spa_t */
	return (NULL);
	}

	static int
	spa_state_data(char buf, size_t size, void data)
	{
	spa_t spa = (spa_t )data;
	(void) snprintf(buf, size, "%s\n", spa_state_to_name(spa));
	return (0);
	}

	/*
	* Return the state of the pool in /proc/spl/kstat/zfs/<pool>/state.
	*
	* This is a lock-less read of the pool's state (unlike using 'zpool', which
	* can potentially block for seconds). Because it doesn't block, it can useful
	* as a pool heartbeat value.
	*/
	static void
	spa_state_init(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.state;
	char *name;
	kstat_t *ksp;

	mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);

	name = kmem_asprintf("zfs/%s", spa_name(spa));
	ksp = kstat_create(name, 0, "state", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);

	shk->kstat = ksp;
	if (ksp) {
	ksp->ks_lock = &shk->lock;
	ksp->ks_data = NULL;
	ksp->ks_private = spa;
	ksp->ks_flags \|= KSTAT_FLAG_NO_HEADERS;
	kstat_set_raw_ops(ksp, NULL, spa_state_data, spa_state_addr);
	kstat_install(ksp);
	}

	kmem_strfree(name);
	}

	static int
	spa_guid_data(char buf, size_t size, void data)
	{
	spa_t spa = (spa_t )data;
	(void) snprintf(buf, size, "%llu\n", (u_longlong_t)spa_guid(spa));
	return (0);
	}

	static void
	spa_guid_init(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.guid;
	char *name;
	kstat_t *ksp;

	mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);

	name = kmem_asprintf("zfs/%s", spa_name(spa));

	ksp = kstat_create(name, 0, "guid", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);

	shk->kstat = ksp;
	if (ksp) {
	ksp->ks_lock = &shk->lock;
	ksp->ks_data = NULL;
	ksp->ks_private = spa;
	ksp->ks_flags \|= KSTAT_FLAG_NO_HEADERS;
	kstat_set_raw_ops(ksp, NULL, spa_guid_data, spa_state_addr);
	kstat_install(ksp);
	}

	kmem_strfree(name);
	}

	static void
	spa_health_destroy(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.state;
	kstat_t *ksp = shk->kstat;
	if (ksp)
	kstat_delete(ksp);

	mutex_destroy(&shk->lock);
	}

	static void
	spa_guid_destroy(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.guid;
	kstat_t *ksp = shk->kstat;
	if (ksp)
	kstat_delete(ksp);

	mutex_destroy(&shk->lock);
	}

	static const spa_iostats_t spa_iostats_template = {
	{ "trim_extents_written", KSTAT_DATA_UINT64 },
	{ "trim_bytes_written", KSTAT_DATA_UINT64 },
	{ "trim_extents_skipped", KSTAT_DATA_UINT64 },
	{ "trim_bytes_skipped", KSTAT_DATA_UINT64 },
	{ "trim_extents_failed", KSTAT_DATA_UINT64 },
	{ "trim_bytes_failed", KSTAT_DATA_UINT64 },
	{ "autotrim_extents_written", KSTAT_DATA_UINT64 },
	{ "autotrim_bytes_written", KSTAT_DATA_UINT64 },
	{ "autotrim_extents_skipped", KSTAT_DATA_UINT64 },
	{ "autotrim_bytes_skipped", KSTAT_DATA_UINT64 },
	{ "autotrim_extents_failed", KSTAT_DATA_UINT64 },
	{ "autotrim_bytes_failed", KSTAT_DATA_UINT64 },
	{ "simple_trim_extents_written", KSTAT_DATA_UINT64 },
	{ "simple_trim_bytes_written", KSTAT_DATA_UINT64 },
	{ "simple_trim_extents_skipped", KSTAT_DATA_UINT64 },
	{ "simple_trim_bytes_skipped", KSTAT_DATA_UINT64 },
	{ "simple_trim_extents_failed", KSTAT_DATA_UINT64 },
	{ "simple_trim_bytes_failed", KSTAT_DATA_UINT64 },
	};

	#define SPA_IOSTATS_ADD(stat, val) \
	atomic_add_64(&iostats->stat.value.ui64, (val));

	void
	spa_iostats_trim_add(spa_t *spa, trim_type_t type,
	uint64_t extents_written, uint64_t bytes_written,
	uint64_t extents_skipped, uint64_t bytes_skipped,
	uint64_t extents_failed, uint64_t bytes_failed)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.iostats;
	kstat_t *ksp = shk->kstat;
	spa_iostats_t *iostats;

	if (ksp == NULL)
	return;

	iostats = ksp->ks_data;
	if (type == TRIM_TYPE_MANUAL) {
	SPA_IOSTATS_ADD(trim_extents_written, extents_written);
	SPA_IOSTATS_ADD(trim_bytes_written, bytes_written);
	SPA_IOSTATS_ADD(trim_extents_skipped, extents_skipped);
	SPA_IOSTATS_ADD(trim_bytes_skipped, bytes_skipped);
	SPA_IOSTATS_ADD(trim_extents_failed, extents_failed);
	SPA_IOSTATS_ADD(trim_bytes_failed, bytes_failed);
	} else if (type == TRIM_TYPE_AUTO) {
	SPA_IOSTATS_ADD(autotrim_extents_written, extents_written);
	SPA_IOSTATS_ADD(autotrim_bytes_written, bytes_written);
	SPA_IOSTATS_ADD(autotrim_extents_skipped, extents_skipped);
	SPA_IOSTATS_ADD(autotrim_bytes_skipped, bytes_skipped);
	SPA_IOSTATS_ADD(autotrim_extents_failed, extents_failed);
	SPA_IOSTATS_ADD(autotrim_bytes_failed, bytes_failed);
	} else {
	SPA_IOSTATS_ADD(simple_trim_extents_written, extents_written);
	SPA_IOSTATS_ADD(simple_trim_bytes_written, bytes_written);
	SPA_IOSTATS_ADD(simple_trim_extents_skipped, extents_skipped);
	SPA_IOSTATS_ADD(simple_trim_bytes_skipped, bytes_skipped);
	SPA_IOSTATS_ADD(simple_trim_extents_failed, extents_failed);
	SPA_IOSTATS_ADD(simple_trim_bytes_failed, bytes_failed);
	}
	}

	static int
	spa_iostats_update(kstat_t *ksp, int rw)
	{
	if (rw == KSTAT_WRITE) {
	memcpy(ksp->ks_data, &spa_iostats_template,
	sizeof (spa_iostats_t));
	}

	return (0);
	}

	static void
	spa_iostats_init(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.iostats;

	mutex_init(&shk->lock, NULL, MUTEX_DEFAULT, NULL);

	char *name = kmem_asprintf("zfs/%s", spa_name(spa));
	kstat_t *ksp = kstat_create(name, 0, "iostats", "misc",
	KSTAT_TYPE_NAMED, sizeof (spa_iostats_t) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);

	shk->kstat = ksp;
	if (ksp) {
	int size = sizeof (spa_iostats_t);
	ksp->ks_lock = &shk->lock;
	ksp->ks_private = spa;
	ksp->ks_update = spa_iostats_update;
	ksp->ks_data = kmem_alloc(size, KM_SLEEP);
	memcpy(ksp->ks_data, &spa_iostats_template, size);
	kstat_install(ksp);
	}

	kmem_strfree(name);
	}

	static void
	spa_iostats_destroy(spa_t *spa)
	{
	spa_history_kstat_t *shk = &spa->spa_stats.iostats;
	kstat_t *ksp = shk->kstat;
	if (ksp) {
	kmem_free(ksp->ks_data, sizeof (spa_iostats_t));
	kstat_delete(ksp);
	}

	mutex_destroy(&shk->lock);
	}

	void
	spa_stats_init(spa_t *spa)
	{
	spa_read_history_init(spa);
	spa_txg_history_init(spa);
	spa_tx_assign_init(spa);
	spa_mmp_history_init(spa);
	spa_state_init(spa);
	spa_guid_init(spa);
	spa_iostats_init(spa);
	}

	void
	spa_stats_destroy(spa_t *spa)
	{
	spa_iostats_destroy(spa);
	spa_health_destroy(spa);
	spa_tx_assign_destroy(spa);
	spa_txg_history_destroy(spa);
	spa_read_history_destroy(spa);
	spa_mmp_history_destroy(spa);
	spa_guid_destroy(spa);
	}

	-ZFS_MODULE_PARAM(zfs, zfs_, read_history, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, read_history, UINT, ZMOD_RW,
	"Historical statistics for the last N reads");

	ZFS_MODULE_PARAM(zfs, zfs_, read_history_hits, INT, ZMOD_RW,
	"Include cache hits in read history");

	-ZFS_MODULE_PARAM(zfs_txg, zfs_txg_, history, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_txg, zfs_txg_, history, UINT, ZMOD_RW,
	"Historical statistics for the last N txgs");

	-ZFS_MODULE_PARAM(zfs_multihost, zfs_multihost_, history, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_multihost, zfs_multihost_, history, UINT, ZMOD_RW,
	"Historical statistics for last N multihost writes");
	diff --git a/sys/contrib/openzfs/module/zfs/txg.c b/sys/contrib/openzfs/module/zfs/txg.c
	index 6e2b8b01053f..29eb9e8e8016 100644
	--- a/sys/contrib/openzfs/module/zfs/txg.c
	+++ b/sys/contrib/openzfs/module/zfs/txg.c
	@@ -1,1073 +1,1073 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Portions Copyright 2011 Martin Matuska
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/txg_impl.h>
	#include <sys/dmu_impl.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_scan.h>
	#include <sys/zil.h>
	#include <sys/callb.h>
	#include <sys/trace_zfs.h>

	/*
	* ZFS Transaction Groups
	* ----------------------
	*
	* ZFS transaction groups are, as the name implies, groups of transactions
	* that act on persistent state. ZFS asserts consistency at the granularity of
	* these transaction groups. Each successive transaction group (txg) is
	* assigned a 64-bit consecutive identifier. There are three active
	* transaction group states: open, quiescing, or syncing. At any given time,
	* there may be an active txg associated with each state; each active txg may
	* either be processing, or blocked waiting to enter the next state. There may
	* be up to three active txgs, and there is always a txg in the open state
	* (though it may be blocked waiting to enter the quiescing state). In broad
	* strokes, transactions -- operations that change in-memory structures -- are
	* accepted into the txg in the open state, and are completed while the txg is
	* in the open or quiescing states. The accumulated changes are written to
	* disk in the syncing state.
	*
	* Open
	*
	* When a new txg becomes active, it first enters the open state. New
	* transactions -- updates to in-memory structures -- are assigned to the
	* currently open txg. There is always a txg in the open state so that ZFS can
	* accept new changes (though the txg may refuse new changes if it has hit
	* some limit). ZFS advances the open txg to the next state for a variety of
	* reasons such as it hitting a time or size threshold, or the execution of an
	* administrative action that must be completed in the syncing state.
	*
	* Quiescing
	*
	* After a txg exits the open state, it enters the quiescing state. The
	* quiescing state is intended to provide a buffer between accepting new
	* transactions in the open state and writing them out to stable storage in
	* the syncing state. While quiescing, transactions can continue their
	* operation without delaying either of the other states. Typically, a txg is
	* in the quiescing state very briefly since the operations are bounded by
	* software latencies rather than, say, slower I/O latencies. After all
	* transactions complete, the txg is ready to enter the next state.
	*
	* Syncing
	*
	* In the syncing state, the in-memory state built up during the open and (to
	* a lesser degree) the quiescing states is written to stable storage. The
	* process of writing out modified data can, in turn modify more data. For
	* example when we write new blocks, we need to allocate space for them; those
	* allocations modify metadata (space maps)... which themselves must be
	* written to stable storage. During the sync state, ZFS iterates, writing out
	* data until it converges and all in-memory changes have been written out.
	* The first such pass is the largest as it encompasses all the modified user
	* data (as opposed to filesystem metadata). Subsequent passes typically have
	* far less data to write as they consist exclusively of filesystem metadata.
	*
	* To ensure convergence, after a certain number of passes ZFS begins
	* overwriting locations on stable storage that had been allocated earlier in
	* the syncing state (and subsequently freed). ZFS usually allocates new
	* blocks to optimize for large, continuous, writes. For the syncing state to
	* converge however it must complete a pass where no new blocks are allocated
	* since each allocation requires a modification of persistent metadata.
	* Further, to hasten convergence, after a prescribed number of passes, ZFS
	* also defers frees, and stops compressing.
	*
	* In addition to writing out user data, we must also execute synctasks during
	* the syncing context. A synctask is the mechanism by which some
	* administrative activities work such as creating and destroying snapshots or
	* datasets. Note that when a synctask is initiated it enters the open txg,
	* and ZFS then pushes that txg as quickly as possible to completion of the
	* syncing state in order to reduce the latency of the administrative
	* activity. To complete the syncing state, ZFS writes out a new uberblock,
	* the root of the tree of blocks that comprise all state stored on the ZFS
	* pool. Finally, if there is a quiesced txg waiting, we signal that it can
	* now transition to the syncing state.
	*/

	static __attribute__((noreturn)) void txg_sync_thread(void *arg);
	static __attribute__((noreturn)) void txg_quiesce_thread(void *arg);

	-int zfs_txg_timeout = 5; /* max seconds worth of delta per txg */
	+uint_t zfs_txg_timeout = 5; /* max seconds worth of delta per txg */

	/*
	* Prepare the txg subsystem.
	*/
	void
	txg_init(dsl_pool_t *dp, uint64_t txg)
	{
	tx_state_t *tx = &dp->dp_tx;
	int c;
	memset(tx, 0, sizeof (tx_state_t));

	tx->tx_cpu = vmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);

	for (c = 0; c < max_ncpus; c++) {
	int i;

	mutex_init(&tx->tx_cpu[c].tc_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&tx->tx_cpu[c].tc_open_lock, NULL, MUTEX_NOLOCKDEP,
	NULL);
	for (i = 0; i < TXG_SIZE; i++) {
	cv_init(&tx->tx_cpu[c].tc_cv[i], NULL, CV_DEFAULT,
	NULL);
	list_create(&tx->tx_cpu[c].tc_callbacks[i],
	sizeof (dmu_tx_callback_t),
	offsetof(dmu_tx_callback_t, dcb_node));
	}
	}

	mutex_init(&tx->tx_sync_lock, NULL, MUTEX_DEFAULT, NULL);

	cv_init(&tx->tx_sync_more_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tx->tx_sync_done_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tx->tx_quiesce_more_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tx->tx_quiesce_done_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&tx->tx_exit_cv, NULL, CV_DEFAULT, NULL);

	tx->tx_open_txg = txg;
	}

	/*
	* Close down the txg subsystem.
	*/
	void
	txg_fini(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;
	int c;

	ASSERT0(tx->tx_threads);

	mutex_destroy(&tx->tx_sync_lock);

	cv_destroy(&tx->tx_sync_more_cv);
	cv_destroy(&tx->tx_sync_done_cv);
	cv_destroy(&tx->tx_quiesce_more_cv);
	cv_destroy(&tx->tx_quiesce_done_cv);
	cv_destroy(&tx->tx_exit_cv);

	for (c = 0; c < max_ncpus; c++) {
	int i;

	mutex_destroy(&tx->tx_cpu[c].tc_open_lock);
	mutex_destroy(&tx->tx_cpu[c].tc_lock);
	for (i = 0; i < TXG_SIZE; i++) {
	cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
	list_destroy(&tx->tx_cpu[c].tc_callbacks[i]);
	}
	}

	if (tx->tx_commit_cb_taskq != NULL)
	taskq_destroy(tx->tx_commit_cb_taskq);

	vmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));

	memset(tx, 0, sizeof (tx_state_t));
	}

	/*
	* Start syncing transaction groups.
	*/
	void
	txg_sync_start(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;

	mutex_enter(&tx->tx_sync_lock);

	dprintf("pool %p\n", dp);

	ASSERT0(tx->tx_threads);

	tx->tx_threads = 2;

	tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
	dp, 0, &p0, TS_RUN, defclsyspri);

	/*
	* The sync thread can need a larger-than-default stack size on
	* 32-bit x86. This is due in part to nested pools and
	* scrub_visitbp() recursion.
	*/
	tx->tx_sync_thread = thread_create(NULL, 0, txg_sync_thread,
	dp, 0, &p0, TS_RUN, defclsyspri);

	mutex_exit(&tx->tx_sync_lock);
	}

	static void
	txg_thread_enter(tx_state_t tx, callb_cpr_t cpr)
	{
	CALLB_CPR_INIT(cpr, &tx->tx_sync_lock, callb_generic_cpr, FTAG);
	mutex_enter(&tx->tx_sync_lock);
	}

	static void
	txg_thread_exit(tx_state_t tx, callb_cpr_t cpr, kthread_t **tpp)
	{
	ASSERT(*tpp != NULL);
	*tpp = NULL;
	tx->tx_threads--;
	cv_broadcast(&tx->tx_exit_cv);
	CALLB_CPR_EXIT(cpr); /* drops &tx->tx_sync_lock */
	thread_exit();
	}

	static void
	txg_thread_wait(tx_state_t tx, callb_cpr_t cpr, kcondvar_t *cv, clock_t time)
	{
	CALLB_CPR_SAFE_BEGIN(cpr);

	if (time) {
	(void) cv_timedwait_idle(cv, &tx->tx_sync_lock,
	ddi_get_lbolt() + time);
	} else {
	cv_wait_idle(cv, &tx->tx_sync_lock);
	}

	CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
	}

	/*
	* Stop syncing transaction groups.
	*/
	void
	txg_sync_stop(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;

	dprintf("pool %p\n", dp);
	/*
	* Finish off any work in progress.
	*/
	ASSERT3U(tx->tx_threads, ==, 2);

	/*
	* We need to ensure that we've vacated the deferred metaslab trees.
	*/
	txg_wait_synced(dp, tx->tx_open_txg + TXG_DEFER_SIZE);

	/*
	* Wake all sync threads and wait for them to die.
	*/
	mutex_enter(&tx->tx_sync_lock);

	ASSERT3U(tx->tx_threads, ==, 2);

	tx->tx_exiting = 1;

	cv_broadcast(&tx->tx_quiesce_more_cv);
	cv_broadcast(&tx->tx_quiesce_done_cv);
	cv_broadcast(&tx->tx_sync_more_cv);

	while (tx->tx_threads != 0)
	cv_wait(&tx->tx_exit_cv, &tx->tx_sync_lock);

	tx->tx_exiting = 0;

	mutex_exit(&tx->tx_sync_lock);
	}

	/*
	* Get a handle on the currently open txg and keep it open.
	*
	* The txg is guaranteed to stay open until txg_rele_to_quiesce() is called for
	* the handle. Once txg_rele_to_quiesce() has been called, the txg stays
	* in quiescing state until txg_rele_to_sync() is called for the handle.
	*
	* It is guaranteed that subsequent calls return monotonically increasing
	* txgs for the same dsl_pool_t. Of course this is not strong monotonicity,
	* because the same txg can be returned multiple times in a row. This
	* guarantee holds both for subsequent calls from one thread and for multiple
	* threads. For example, it is impossible to observe the following sequence
	* of events:
	*
	* Thread 1 Thread 2
	*
	* 1 <- txg_hold_open(P, ...)
	* 2 <- txg_hold_open(P, ...)
	* 1 <- txg_hold_open(P, ...)
	*
	*/
	uint64_t
	txg_hold_open(dsl_pool_t dp, txg_handle_t th)
	{
	tx_state_t *tx = &dp->dp_tx;
	tx_cpu_t *tc;
	uint64_t txg;

	/*
	* It appears the processor id is simply used as a "random"
	* number to index into the array, and there isn't any other
	* significance to the chosen tx_cpu. Because.. Why not use
	* the current cpu to index into the array?
	*/
	tc = &tx->tx_cpu[CPU_SEQID_UNSTABLE];

	mutex_enter(&tc->tc_open_lock);
	txg = tx->tx_open_txg;

	mutex_enter(&tc->tc_lock);
	tc->tc_count[txg & TXG_MASK]++;
	mutex_exit(&tc->tc_lock);

	th->th_cpu = tc;
	th->th_txg = txg;

	return (txg);
	}

	void
	txg_rele_to_quiesce(txg_handle_t *th)
	{
	tx_cpu_t *tc = th->th_cpu;

	ASSERT(!MUTEX_HELD(&tc->tc_lock));
	mutex_exit(&tc->tc_open_lock);
	}

	void
	txg_register_callbacks(txg_handle_t th, list_t tx_callbacks)
	{
	tx_cpu_t *tc = th->th_cpu;
	int g = th->th_txg & TXG_MASK;

	mutex_enter(&tc->tc_lock);
	list_move_tail(&tc->tc_callbacks[g], tx_callbacks);
	mutex_exit(&tc->tc_lock);
	}

	void
	txg_rele_to_sync(txg_handle_t *th)
	{
	tx_cpu_t *tc = th->th_cpu;
	int g = th->th_txg & TXG_MASK;

	mutex_enter(&tc->tc_lock);
	ASSERT(tc->tc_count[g] != 0);
	if (--tc->tc_count[g] == 0)
	cv_broadcast(&tc->tc_cv[g]);
	mutex_exit(&tc->tc_lock);

	th->th_cpu = NULL; /* defensive */
	}

	/*
	* Blocks until all transactions in the group are committed.
	*
	* On return, the transaction group has reached a stable state in which it can
	* then be passed off to the syncing context.
	*/
	static void
	txg_quiesce(dsl_pool_t *dp, uint64_t txg)
	{
	tx_state_t *tx = &dp->dp_tx;
	uint64_t tx_open_time;
	int g = txg & TXG_MASK;
	int c;

	/*
	* Grab all tc_open_locks so nobody else can get into this txg.
	*/
	for (c = 0; c < max_ncpus; c++)
	mutex_enter(&tx->tx_cpu[c].tc_open_lock);

	ASSERT(txg == tx->tx_open_txg);
	tx->tx_open_txg++;
	tx->tx_open_time = tx_open_time = gethrtime();

	DTRACE_PROBE2(txg__quiescing, dsl_pool_t *, dp, uint64_t, txg);
	DTRACE_PROBE2(txg__opened, dsl_pool_t *, dp, uint64_t, tx->tx_open_txg);

	/*
	* Now that we've incremented tx_open_txg, we can let threads
	* enter the next transaction group.
	*/
	for (c = 0; c < max_ncpus; c++)
	mutex_exit(&tx->tx_cpu[c].tc_open_lock);

	spa_txg_history_set(dp->dp_spa, txg, TXG_STATE_OPEN, tx_open_time);
	spa_txg_history_add(dp->dp_spa, txg + 1, tx_open_time);

	/*
	* Quiesce the transaction group by waiting for everyone to
	* call txg_rele_to_sync() for their open transaction handles.
	*/
	for (c = 0; c < max_ncpus; c++) {
	tx_cpu_t *tc = &tx->tx_cpu[c];
	mutex_enter(&tc->tc_lock);
	while (tc->tc_count[g] != 0)
	cv_wait(&tc->tc_cv[g], &tc->tc_lock);
	mutex_exit(&tc->tc_lock);
	}

	spa_txg_history_set(dp->dp_spa, txg, TXG_STATE_QUIESCED, gethrtime());
	}

	static void
	txg_do_callbacks(list_t *cb_list)
	{
	dmu_tx_do_callbacks(cb_list, 0);

	list_destroy(cb_list);

	kmem_free(cb_list, sizeof (list_t));
	}

	/*
	* Dispatch the commit callbacks registered on this txg to worker threads.
	*
	* If no callbacks are registered for a given TXG, nothing happens.
	* This function creates a taskq for the associated pool, if needed.
	*/
	static void
	txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
	{
	int c;
	tx_state_t *tx = &dp->dp_tx;
	list_t *cb_list;

	for (c = 0; c < max_ncpus; c++) {
	tx_cpu_t *tc = &tx->tx_cpu[c];
	/*
	* No need to lock tx_cpu_t at this point, since this can
	* only be called once a txg has been synced.
	*/

	int g = txg & TXG_MASK;

	if (list_is_empty(&tc->tc_callbacks[g]))
	continue;

	if (tx->tx_commit_cb_taskq == NULL) {
	/*
	* Commit callback taskq hasn't been created yet.
	*/
	tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
	100, defclsyspri, boot_ncpus, boot_ncpus * 2,
	TASKQ_PREPOPULATE \| TASKQ_DYNAMIC \|
	TASKQ_THREADS_CPU_PCT);
	}

	cb_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
	list_create(cb_list, sizeof (dmu_tx_callback_t),
	offsetof(dmu_tx_callback_t, dcb_node));

	list_move_tail(cb_list, &tc->tc_callbacks[g]);

	(void) taskq_dispatch(tx->tx_commit_cb_taskq, (task_func_t *)
	txg_do_callbacks, cb_list, TQ_SLEEP);
	}
	}

	/*
	* Wait for pending commit callbacks of already-synced transactions to finish
	* processing.
	* Calling this function from within a commit callback will deadlock.
	*/
	void
	txg_wait_callbacks(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;

	if (tx->tx_commit_cb_taskq != NULL)
	taskq_wait_outstanding(tx->tx_commit_cb_taskq, 0);
	}

	static boolean_t
	txg_is_quiescing(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;
	ASSERT(MUTEX_HELD(&tx->tx_sync_lock));
	return (tx->tx_quiescing_txg != 0);
	}

	static boolean_t
	txg_has_quiesced_to_sync(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;
	ASSERT(MUTEX_HELD(&tx->tx_sync_lock));
	return (tx->tx_quiesced_txg != 0);
	}

	static __attribute__((noreturn)) void
	txg_sync_thread(void *arg)
	{
	dsl_pool_t *dp = arg;
	spa_t *spa = dp->dp_spa;
	tx_state_t *tx = &dp->dp_tx;
	callb_cpr_t cpr;
	clock_t start, delta;

	(void) spl_fstrans_mark();
	txg_thread_enter(tx, &cpr);

	start = delta = 0;
	for (;;) {
	clock_t timeout = zfs_txg_timeout * hz;
	clock_t timer;
	uint64_t txg;

	/*
	* We sync when we're scanning, there's someone waiting
	* on us, or the quiesce thread has handed off a txg to
	* us, or we have reached our timeout.
	*/
	timer = (delta >= timeout ? 0 : timeout - delta);
	while (!dsl_scan_active(dp->dp_scan) &&
	!tx->tx_exiting && timer > 0 &&
	tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
	!txg_has_quiesced_to_sync(dp)) {
	dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
	(u_longlong_t)tx->tx_synced_txg,
	(u_longlong_t)tx->tx_sync_txg_waiting, dp);
	txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
	delta = ddi_get_lbolt() - start;
	timer = (delta > timeout ? 0 : timeout - delta);
	}

	/*
	* Wait until the quiesce thread hands off a txg to us,
	* prompting it to do so if necessary.
	*/
	while (!tx->tx_exiting && !txg_has_quiesced_to_sync(dp)) {
	if (txg_is_quiescing(dp)) {
	txg_thread_wait(tx, &cpr,
	&tx->tx_quiesce_done_cv, 0);
	continue;
	}
	if (tx->tx_quiesce_txg_waiting < tx->tx_open_txg+1)
	tx->tx_quiesce_txg_waiting = tx->tx_open_txg+1;
	cv_broadcast(&tx->tx_quiesce_more_cv);
	txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
	}

	if (tx->tx_exiting)
	txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);

	/*
	* Consume the quiesced txg which has been handed off to
	* us. This may cause the quiescing thread to now be
	* able to quiesce another txg, so we must signal it.
	*/
	ASSERT(tx->tx_quiesced_txg != 0);
	txg = tx->tx_quiesced_txg;
	tx->tx_quiesced_txg = 0;
	tx->tx_syncing_txg = txg;
	DTRACE_PROBE2(txg__syncing, dsl_pool_t *, dp, uint64_t, txg);
	cv_broadcast(&tx->tx_quiesce_more_cv);

	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
	(u_longlong_t)txg, (u_longlong_t)tx->tx_quiesce_txg_waiting,
	(u_longlong_t)tx->tx_sync_txg_waiting);
	mutex_exit(&tx->tx_sync_lock);

	txg_stat_t *ts = spa_txg_history_init_io(spa, txg, dp);
	start = ddi_get_lbolt();
	spa_sync(spa, txg);
	delta = ddi_get_lbolt() - start;
	spa_txg_history_fini_io(spa, ts);

	mutex_enter(&tx->tx_sync_lock);
	tx->tx_synced_txg = txg;
	tx->tx_syncing_txg = 0;
	DTRACE_PROBE2(txg__synced, dsl_pool_t *, dp, uint64_t, txg);
	cv_broadcast(&tx->tx_sync_done_cv);

	/*
	* Dispatch commit callbacks to worker threads.
	*/
	txg_dispatch_callbacks(dp, txg);
	}
	}

	static __attribute__((noreturn)) void
	txg_quiesce_thread(void *arg)
	{
	dsl_pool_t *dp = arg;
	tx_state_t *tx = &dp->dp_tx;
	callb_cpr_t cpr;

	txg_thread_enter(tx, &cpr);

	for (;;) {
	uint64_t txg;

	/*
	* We quiesce when there's someone waiting on us.
	* However, we can only have one txg in "quiescing" or
	* "quiesced, waiting to sync" state. So we wait until
	* the "quiesced, waiting to sync" txg has been consumed
	* by the sync thread.
	*/
	while (!tx->tx_exiting &&
	(tx->tx_open_txg >= tx->tx_quiesce_txg_waiting \|\|
	txg_has_quiesced_to_sync(dp)))
	txg_thread_wait(tx, &cpr, &tx->tx_quiesce_more_cv, 0);

	if (tx->tx_exiting)
	txg_thread_exit(tx, &cpr, &tx->tx_quiesce_thread);

	txg = tx->tx_open_txg;
	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
	(u_longlong_t)txg,
	(u_longlong_t)tx->tx_quiesce_txg_waiting,
	(u_longlong_t)tx->tx_sync_txg_waiting);
	tx->tx_quiescing_txg = txg;

	mutex_exit(&tx->tx_sync_lock);
	txg_quiesce(dp, txg);
	mutex_enter(&tx->tx_sync_lock);

	/*
	* Hand this txg off to the sync thread.
	*/
	dprintf("quiesce done, handing off txg %llu\n",
	(u_longlong_t)txg);
	tx->tx_quiescing_txg = 0;
	tx->tx_quiesced_txg = txg;
	DTRACE_PROBE2(txg__quiesced, dsl_pool_t *, dp, uint64_t, txg);
	cv_broadcast(&tx->tx_sync_more_cv);
	cv_broadcast(&tx->tx_quiesce_done_cv);
	}
	}

	/*
	* Delay this thread by delay nanoseconds if we are still in the open
	* transaction group and there is already a waiting txg quiescing or quiesced.
	* Abort the delay if this txg stalls or enters the quiescing state.
	*/
	void
	txg_delay(dsl_pool_t *dp, uint64_t txg, hrtime_t delay, hrtime_t resolution)
	{
	tx_state_t *tx = &dp->dp_tx;
	hrtime_t start = gethrtime();

	/* don't delay if this txg could transition to quiescing immediately */
	if (tx->tx_open_txg > txg \|\|
	tx->tx_syncing_txg == txg-1 \|\| tx->tx_synced_txg == txg-1)
	return;

	mutex_enter(&tx->tx_sync_lock);
	if (tx->tx_open_txg > txg \|\| tx->tx_synced_txg == txg-1) {
	mutex_exit(&tx->tx_sync_lock);
	return;
	}

	while (gethrtime() - start < delay &&
	tx->tx_syncing_txg < txg-1 && !txg_stalled(dp)) {
	(void) cv_timedwait_hires(&tx->tx_quiesce_more_cv,
	&tx->tx_sync_lock, delay, resolution, 0);
	}

	DMU_TX_STAT_BUMP(dmu_tx_delay);

	mutex_exit(&tx->tx_sync_lock);
	}

	static boolean_t
	txg_wait_synced_impl(dsl_pool_t *dp, uint64_t txg, boolean_t wait_sig)
	{
	tx_state_t *tx = &dp->dp_tx;

	ASSERT(!dsl_pool_config_held(dp));

	mutex_enter(&tx->tx_sync_lock);
	ASSERT3U(tx->tx_threads, ==, 2);
	if (txg == 0)
	txg = tx->tx_open_txg + TXG_DEFER_SIZE;
	if (tx->tx_sync_txg_waiting < txg)
	tx->tx_sync_txg_waiting = txg;
	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
	(u_longlong_t)txg, (u_longlong_t)tx->tx_quiesce_txg_waiting,
	(u_longlong_t)tx->tx_sync_txg_waiting);
	while (tx->tx_synced_txg < txg) {
	dprintf("broadcasting sync more "
	"tx_synced=%llu waiting=%llu dp=%px\n",
	(u_longlong_t)tx->tx_synced_txg,
	(u_longlong_t)tx->tx_sync_txg_waiting, dp);
	cv_broadcast(&tx->tx_sync_more_cv);
	if (wait_sig) {
	/*
	* Condition wait here but stop if the thread receives a
	* signal. The caller may call txg_wait_synced*() again
	* to resume waiting for this txg.
	*/
	if (cv_wait_io_sig(&tx->tx_sync_done_cv,
	&tx->tx_sync_lock) == 0) {
	mutex_exit(&tx->tx_sync_lock);
	return (B_TRUE);
	}
	} else {
	cv_wait_io(&tx->tx_sync_done_cv, &tx->tx_sync_lock);
	}
	}
	mutex_exit(&tx->tx_sync_lock);
	return (B_FALSE);
	}

	void
	txg_wait_synced(dsl_pool_t *dp, uint64_t txg)
	{
	VERIFY0(txg_wait_synced_impl(dp, txg, B_FALSE));
	}

	/*
	* Similar to a txg_wait_synced but it can be interrupted from a signal.
	* Returns B_TRUE if the thread was signaled while waiting.
	*/
	boolean_t
	txg_wait_synced_sig(dsl_pool_t *dp, uint64_t txg)
	{
	return (txg_wait_synced_impl(dp, txg, B_TRUE));
	}

	/*
	* Wait for the specified open transaction group. Set should_quiesce
	* when the current open txg should be quiesced immediately.
	*/
	void
	txg_wait_open(dsl_pool_t *dp, uint64_t txg, boolean_t should_quiesce)
	{
	tx_state_t *tx = &dp->dp_tx;

	ASSERT(!dsl_pool_config_held(dp));

	mutex_enter(&tx->tx_sync_lock);
	ASSERT3U(tx->tx_threads, ==, 2);
	if (txg == 0)
	txg = tx->tx_open_txg + 1;
	if (tx->tx_quiesce_txg_waiting < txg && should_quiesce)
	tx->tx_quiesce_txg_waiting = txg;
	dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
	(u_longlong_t)txg, (u_longlong_t)tx->tx_quiesce_txg_waiting,
	(u_longlong_t)tx->tx_sync_txg_waiting);
	while (tx->tx_open_txg < txg) {
	cv_broadcast(&tx->tx_quiesce_more_cv);
	/*
	* Callers setting should_quiesce will use cv_wait_io() and
	* be accounted for as iowait time. Otherwise, the caller is
	* understood to be idle and cv_wait_sig() is used to prevent
	* incorrectly inflating the system load average.
	*/
	if (should_quiesce == B_TRUE) {
	cv_wait_io(&tx->tx_quiesce_done_cv, &tx->tx_sync_lock);
	} else {
	cv_wait_idle(&tx->tx_quiesce_done_cv,
	&tx->tx_sync_lock);
	}
	}
	mutex_exit(&tx->tx_sync_lock);
	}

	/*
	* Pass in the txg number that should be synced.
	*/
	void
	txg_kick(dsl_pool_t *dp, uint64_t txg)
	{
	tx_state_t *tx = &dp->dp_tx;

	ASSERT(!dsl_pool_config_held(dp));

	if (tx->tx_sync_txg_waiting >= txg)
	return;

	mutex_enter(&tx->tx_sync_lock);
	if (tx->tx_sync_txg_waiting < txg) {
	tx->tx_sync_txg_waiting = txg;
	cv_broadcast(&tx->tx_sync_more_cv);
	}
	mutex_exit(&tx->tx_sync_lock);
	}

	boolean_t
	txg_stalled(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;
	return (tx->tx_quiesce_txg_waiting > tx->tx_open_txg);
	}

	boolean_t
	txg_sync_waiting(dsl_pool_t *dp)
	{
	tx_state_t *tx = &dp->dp_tx;

	return (tx->tx_syncing_txg <= tx->tx_sync_txg_waiting \|\|
	tx->tx_quiesced_txg != 0);
	}

	/*
	* Verify that this txg is active (open, quiescing, syncing). Non-active
	* txg's should not be manipulated.
	*/
	#ifdef ZFS_DEBUG
	void
	txg_verify(spa_t *spa, uint64_t txg)
	{
	dsl_pool_t *dp __maybe_unused = spa_get_dsl(spa);
	if (txg <= TXG_INITIAL \|\| txg == ZILTEST_TXG)
	return;
	ASSERT3U(txg, <=, dp->dp_tx.tx_open_txg);
	ASSERT3U(txg, >=, dp->dp_tx.tx_synced_txg);
	ASSERT3U(txg, >=, dp->dp_tx.tx_open_txg - TXG_CONCURRENT_STATES);
	}
	#endif

	/*
	* Per-txg object lists.
	*/
	void
	txg_list_create(txg_list_t tl, spa_t spa, size_t offset)
	{
	int t;

	mutex_init(&tl->tl_lock, NULL, MUTEX_DEFAULT, NULL);

	tl->tl_offset = offset;
	tl->tl_spa = spa;

	for (t = 0; t < TXG_SIZE; t++)
	tl->tl_head[t] = NULL;
	}

	static boolean_t
	txg_list_empty_impl(txg_list_t *tl, uint64_t txg)
	{
	ASSERT(MUTEX_HELD(&tl->tl_lock));
	TXG_VERIFY(tl->tl_spa, txg);
	return (tl->tl_head[txg & TXG_MASK] == NULL);
	}

	boolean_t
	txg_list_empty(txg_list_t *tl, uint64_t txg)
	{
	mutex_enter(&tl->tl_lock);
	boolean_t ret = txg_list_empty_impl(tl, txg);
	mutex_exit(&tl->tl_lock);

	return (ret);
	}

	void
	txg_list_destroy(txg_list_t *tl)
	{
	int t;

	mutex_enter(&tl->tl_lock);
	for (t = 0; t < TXG_SIZE; t++)
	ASSERT(txg_list_empty_impl(tl, t));
	mutex_exit(&tl->tl_lock);

	mutex_destroy(&tl->tl_lock);
	}

	/*
	* Returns true if all txg lists are empty.
	*
	* Warning: this is inherently racy (an item could be added immediately
	* after this function returns).
	*/
	boolean_t
	txg_all_lists_empty(txg_list_t *tl)
	{
	mutex_enter(&tl->tl_lock);
	for (int i = 0; i < TXG_SIZE; i++) {
	if (!txg_list_empty_impl(tl, i)) {
	mutex_exit(&tl->tl_lock);
	return (B_FALSE);
	}
	}
	mutex_exit(&tl->tl_lock);
	return (B_TRUE);
	}

	/*
	* Add an entry to the list (unless it's already on the list).
	* Returns B_TRUE if it was actually added.
	*/
	boolean_t
	txg_list_add(txg_list_t tl, void p, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t tn = (txg_node_t )((char *)p + tl->tl_offset);
	boolean_t add;

	TXG_VERIFY(tl->tl_spa, txg);
	mutex_enter(&tl->tl_lock);
	add = (tn->tn_member[t] == 0);
	if (add) {
	tn->tn_member[t] = 1;
	tn->tn_next[t] = tl->tl_head[t];
	tl->tl_head[t] = tn;
	}
	mutex_exit(&tl->tl_lock);

	return (add);
	}

	/*
	* Add an entry to the end of the list, unless it's already on the list.
	* (walks list to find end)
	* Returns B_TRUE if it was actually added.
	*/
	boolean_t
	txg_list_add_tail(txg_list_t tl, void p, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t tn = (txg_node_t )((char *)p + tl->tl_offset);
	boolean_t add;

	TXG_VERIFY(tl->tl_spa, txg);
	mutex_enter(&tl->tl_lock);
	add = (tn->tn_member[t] == 0);
	if (add) {
	txg_node_t **tp;

	for (tp = &tl->tl_head[t]; tp != NULL; tp = &(tp)->tn_next[t])
	continue;

	tn->tn_member[t] = 1;
	tn->tn_next[t] = NULL;
	*tp = tn;
	}
	mutex_exit(&tl->tl_lock);

	return (add);
	}

	/*
	* Remove the head of the list and return it.
	*/
	void *
	txg_list_remove(txg_list_t *tl, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t *tn;
	void *p = NULL;

	TXG_VERIFY(tl->tl_spa, txg);
	mutex_enter(&tl->tl_lock);
	if ((tn = tl->tl_head[t]) != NULL) {
	ASSERT(tn->tn_member[t]);
	ASSERT(tn->tn_next[t] == NULL \|\| tn->tn_next[t]->tn_member[t]);
	p = (char *)tn - tl->tl_offset;
	tl->tl_head[t] = tn->tn_next[t];
	tn->tn_next[t] = NULL;
	tn->tn_member[t] = 0;
	}
	mutex_exit(&tl->tl_lock);

	return (p);
	}

	/*
	* Remove a specific item from the list and return it.
	*/
	void *
	txg_list_remove_this(txg_list_t tl, void p, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t tn, *tp;

	TXG_VERIFY(tl->tl_spa, txg);
	mutex_enter(&tl->tl_lock);

	for (tp = &tl->tl_head[t]; (tn = *tp) != NULL; tp = &tn->tn_next[t]) {
	if ((char *)tn - tl->tl_offset == p) {
	*tp = tn->tn_next[t];
	tn->tn_next[t] = NULL;
	tn->tn_member[t] = 0;
	mutex_exit(&tl->tl_lock);
	return (p);
	}
	}

	mutex_exit(&tl->tl_lock);

	return (NULL);
	}

	boolean_t
	txg_list_member(txg_list_t tl, void p, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t tn = (txg_node_t )((char *)p + tl->tl_offset);

	TXG_VERIFY(tl->tl_spa, txg);
	return (tn->tn_member[t] != 0);
	}

	/*
	* Walk a txg list
	*/
	void *
	txg_list_head(txg_list_t *tl, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t *tn;

	mutex_enter(&tl->tl_lock);
	tn = tl->tl_head[t];
	mutex_exit(&tl->tl_lock);

	TXG_VERIFY(tl->tl_spa, txg);
	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
	}

	void *
	txg_list_next(txg_list_t tl, void p, uint64_t txg)
	{
	int t = txg & TXG_MASK;
	txg_node_t tn = (txg_node_t )((char *)p + tl->tl_offset);

	TXG_VERIFY(tl->tl_spa, txg);

	mutex_enter(&tl->tl_lock);
	tn = tn->tn_next[t];
	mutex_exit(&tl->tl_lock);

	return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
	}

	EXPORT_SYMBOL(txg_init);
	EXPORT_SYMBOL(txg_fini);
	EXPORT_SYMBOL(txg_sync_start);
	EXPORT_SYMBOL(txg_sync_stop);
	EXPORT_SYMBOL(txg_hold_open);
	EXPORT_SYMBOL(txg_rele_to_quiesce);
	EXPORT_SYMBOL(txg_rele_to_sync);
	EXPORT_SYMBOL(txg_register_callbacks);
	EXPORT_SYMBOL(txg_delay);
	EXPORT_SYMBOL(txg_wait_synced);
	EXPORT_SYMBOL(txg_wait_open);
	EXPORT_SYMBOL(txg_wait_callbacks);
	EXPORT_SYMBOL(txg_stalled);
	EXPORT_SYMBOL(txg_sync_waiting);

	-ZFS_MODULE_PARAM(zfs_txg, zfs_txg_, timeout, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_txg, zfs_txg_, timeout, UINT, ZMOD_RW,
	"Max seconds worth of delta per txg");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev.c b/sys/contrib/openzfs/module/zfs/vdev.c
	index 048616c253c4..66cec052b669 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev.c
	@@ -1,6107 +1,6166 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, Datto Inc. All rights reserved.
	* Copyright (c) 2021, Klara Inc.
	* Copyright [2021] Hewlett Packard Enterprise Development LP
	*/

	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/bpobj.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_dir.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_draid.h>
	#include <sys/uberblock_impl.h>
	#include <sys/metaslab.h>
	#include <sys/metaslab_impl.h>
	#include <sys/space_map.h>
	#include <sys/space_reftree.h>
	#include <sys/zio.h>
	#include <sys/zap.h>
	#include <sys/fs/zfs.h>
	#include <sys/arc.h>
	#include <sys/zil.h>
	#include <sys/dsl_scan.h>
	#include <sys/vdev_raidz.h>
	#include <sys/abd.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>
	#include <sys/zvol.h>
	#include <sys/zfs_ratelimit.h>
	#include "zfs_prop.h"

	/*
	* One metaslab from each (normal-class) vdev is used by the ZIL. These are
	* called "embedded slog metaslabs", are referenced by vdev_log_mg, and are
	* part of the spa_embedded_log_class. The metaslab with the most free space
	* in each vdev is selected for this purpose when the pool is opened (or a
	* vdev is added). See vdev_metaslab_init().
	*
	* Log blocks can be allocated from the following locations. Each one is tried
	* in order until the allocation succeeds:
	* 1. dedicated log vdevs, aka "slog" (spa_log_class)
	* 2. embedded slog metaslabs (spa_embedded_log_class)
	* 3. other metaslabs in normal vdevs (spa_normal_class)
	*
	* zfs_embedded_slog_min_ms disables the embedded slog if there are fewer
	* than this number of metaslabs in the vdev. This ensures that we don't set
	* aside an unreasonable amount of space for the ZIL. If set to less than
	* 1 << (spa_slop_shift + 1), on small pools the usable space may be reduced
	* (by more than 1<<spa_slop_shift) due to the embedded slog metaslab.
	*/
	-static int zfs_embedded_slog_min_ms = 64;
	+static uint_t zfs_embedded_slog_min_ms = 64;

	/* default target for number of metaslabs per top-level vdev */
	-static int zfs_vdev_default_ms_count = 200;
	+static uint_t zfs_vdev_default_ms_count = 200;

	/* minimum number of metaslabs per top-level vdev */
	-static int zfs_vdev_min_ms_count = 16;
	+static uint_t zfs_vdev_min_ms_count = 16;

	/* practical upper limit of total metaslabs per top-level vdev */
	-static int zfs_vdev_ms_count_limit = 1ULL << 17;
	+static uint_t zfs_vdev_ms_count_limit = 1ULL << 17;

	/* lower limit for metaslab size (512M) */
	-static int zfs_vdev_default_ms_shift = 29;
	+static uint_t zfs_vdev_default_ms_shift = 29;

	/* upper limit for metaslab size (16G) */
	-static const int zfs_vdev_max_ms_shift = 34;
	+static const uint_t zfs_vdev_max_ms_shift = 34;

	int vdev_validate_skip = B_FALSE;

	/*
	* Since the DTL space map of a vdev is not expected to have a lot of
	* entries, we default its block size to 4K.
	*/
	int zfs_vdev_dtl_sm_blksz = (1 << 12);

	/*
	* Rate limit slow IO (delay) events to this many per second.
	*/
	static unsigned int zfs_slow_io_events_per_second = 20;

	/*
	* Rate limit checksum events after this many checksum errors per second.
	*/
	static unsigned int zfs_checksum_events_per_second = 20;

	/*
	* Ignore errors during scrub/resilver. Allows to work around resilver
	* upon import when there are pool errors.
	*/
	static int zfs_scan_ignore_errors = 0;

	/*
	* vdev-wide space maps that have lots of entries written to them at
	* the end of each transaction can benefit from a higher I/O bandwidth
	* (e.g. vdev_obsolete_sm), thus we default their block size to 128K.
	*/
	int zfs_vdev_standard_sm_blksz = (1 << 17);

	/*
	* Tunable parameter for debugging or performance analysis. Setting this
	* will cause pool corruption on power loss if a volatile out-of-order
	* write cache is enabled.
	*/
	int zfs_nocacheflush = 0;

	/*
	* Maximum and minimum ashift values that can be automatically set based on
	* vdev's physical ashift (disk's physical sector size). While ASHIFT_MAX
	* is higher than the maximum value, it is intentionally limited here to not
	* excessively impact pool space efficiency. Higher ashift values may still
	* be forced by vdev logical ashift or by user via ashift property, but won't
	* be set automatically as a performance optimization.
	*/
	uint64_t zfs_vdev_max_auto_ashift = 14;
	uint64_t zfs_vdev_min_auto_ashift = ASHIFT_MIN;

	void
	vdev_dbgmsg(vdev_t vd, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	if (vd->vdev_path != NULL) {
	zfs_dbgmsg("%s vdev '%s': %s", vd->vdev_ops->vdev_op_type,
	vd->vdev_path, buf);
	} else {
	zfs_dbgmsg("%s-%llu vdev (guid %llu): %s",
	vd->vdev_ops->vdev_op_type,
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)vd->vdev_guid, buf);
	}
	}

	void
	vdev_dbgmsg_print_tree(vdev_t *vd, int indent)
	{
	char state[20];

	if (vd->vdev_ishole \|\| vd->vdev_ops == &vdev_missing_ops) {
	zfs_dbgmsg("%*svdev %llu: %s", indent, "",
	(u_longlong_t)vd->vdev_id,
	vd->vdev_ops->vdev_op_type);
	return;
	}

	switch (vd->vdev_state) {
	case VDEV_STATE_UNKNOWN:
	(void) snprintf(state, sizeof (state), "unknown");
	break;
	case VDEV_STATE_CLOSED:
	(void) snprintf(state, sizeof (state), "closed");
	break;
	case VDEV_STATE_OFFLINE:
	(void) snprintf(state, sizeof (state), "offline");
	break;
	case VDEV_STATE_REMOVED:
	(void) snprintf(state, sizeof (state), "removed");
	break;
	case VDEV_STATE_CANT_OPEN:
	(void) snprintf(state, sizeof (state), "can't open");
	break;
	case VDEV_STATE_FAULTED:
	(void) snprintf(state, sizeof (state), "faulted");
	break;
	case VDEV_STATE_DEGRADED:
	(void) snprintf(state, sizeof (state), "degraded");
	break;
	case VDEV_STATE_HEALTHY:
	(void) snprintf(state, sizeof (state), "healthy");
	break;
	default:
	(void) snprintf(state, sizeof (state), "<state %u>",
	(uint_t)vd->vdev_state);
	}

	zfs_dbgmsg("%*svdev %u: %s%s, guid: %llu, path: %s, %s", indent,
	"", (int)vd->vdev_id, vd->vdev_ops->vdev_op_type,
	vd->vdev_islog ? " (log)" : "",
	(u_longlong_t)vd->vdev_guid,
	vd->vdev_path ? vd->vdev_path : "N/A", state);

	for (uint64_t i = 0; i < vd->vdev_children; i++)
	vdev_dbgmsg_print_tree(vd->vdev_child[i], indent + 2);
	}

	/*
	* Virtual device management.
	*/

	-static const vdev_ops_t *const vdev_ops_table[] = {
	+static vdev_ops_t *const vdev_ops_table[] = {
	&vdev_root_ops,
	&vdev_raidz_ops,
	&vdev_draid_ops,
	&vdev_draid_spare_ops,
	&vdev_mirror_ops,
	&vdev_replacing_ops,
	&vdev_spare_ops,
	&vdev_disk_ops,
	&vdev_file_ops,
	&vdev_missing_ops,
	&vdev_hole_ops,
	&vdev_indirect_ops,
	NULL
	};

	/*
	* Given a vdev type, return the appropriate ops vector.
	*/
	static vdev_ops_t *
	vdev_getops(const char *type)
	{
	- const vdev_ops_t ops, const *opspp;
	+ vdev_ops_t ops, const *opspp;

	for (opspp = vdev_ops_table; (ops = *opspp) != NULL; opspp++)
	if (strcmp(ops->vdev_op_type, type) == 0)
	break;

	return (ops);
	}

	/*
	* Given a vdev and a metaslab class, find which metaslab group we're
	* interested in. All vdevs may belong to two different metaslab classes.
	* Dedicated slog devices use only the primary metaslab group, rather than a
	* separate log group. For embedded slogs, the vdev_log_mg will be non-NULL.
	*/
	metaslab_group_t *
	vdev_get_mg(vdev_t vd, metaslab_class_t mc)
	{
	if (mc == spa_embedded_log_class(vd->vdev_spa) &&
	vd->vdev_log_mg != NULL)
	return (vd->vdev_log_mg);
	else
	return (vd->vdev_mg);
	}

	void
	vdev_default_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs)
	{
	(void) vd, (void) remain_rs;

	physical_rs->rs_start = logical_rs->rs_start;
	physical_rs->rs_end = logical_rs->rs_end;
	}

	/*
	* Derive the enumerated allocation bias from string input.
	* String origin is either the per-vdev zap or zpool(8).
	*/
	static vdev_alloc_bias_t
	vdev_derive_alloc_bias(const char *bias)
	{
	vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;

	if (strcmp(bias, VDEV_ALLOC_BIAS_LOG) == 0)
	alloc_bias = VDEV_BIAS_LOG;
	else if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
	alloc_bias = VDEV_BIAS_SPECIAL;
	else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
	alloc_bias = VDEV_BIAS_DEDUP;

	return (alloc_bias);
	}

	/*
	* Default asize function: return the MAX of psize with the asize of
	* all children. This is what's used by anything other than RAID-Z.
	*/
	uint64_t
	vdev_default_asize(vdev_t *vd, uint64_t psize)
	{
	uint64_t asize = P2ROUNDUP(psize, 1ULL << vd->vdev_top->vdev_ashift);
	uint64_t csize;

	for (int c = 0; c < vd->vdev_children; c++) {
	csize = vdev_psize_to_asize(vd->vdev_child[c], psize);
	asize = MAX(asize, csize);
	}

	return (asize);
	}

	uint64_t
	vdev_default_min_asize(vdev_t *vd)
	{
	return (vd->vdev_min_asize);
	}

	/*
	* Get the minimum allocatable size. We define the allocatable size as
	* the vdev's asize rounded to the nearest metaslab. This allows us to
	* replace or attach devices which don't have the same physical size but
	* can still satisfy the same number of allocations.
	*/
	uint64_t
	vdev_get_min_asize(vdev_t *vd)
	{
	vdev_t *pvd = vd->vdev_parent;

	/*
	* If our parent is NULL (inactive spare or cache) or is the root,
	* just return our own asize.
	*/
	if (pvd == NULL)
	return (vd->vdev_asize);

	/*
	* The top-level vdev just returns the allocatable size rounded
	* to the nearest metaslab.
	*/
	if (vd == vd->vdev_top)
	return (P2ALIGN(vd->vdev_asize, 1ULL << vd->vdev_ms_shift));

	return (pvd->vdev_ops->vdev_op_min_asize(pvd));
	}

	void
	vdev_set_min_asize(vdev_t *vd)
	{
	vd->vdev_min_asize = vdev_get_min_asize(vd);

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_set_min_asize(vd->vdev_child[c]);
	}

	/*
	* Get the minimal allocation size for the top-level vdev.
	*/
	uint64_t
	vdev_get_min_alloc(vdev_t *vd)
	{
	uint64_t min_alloc = 1ULL << vd->vdev_ashift;

	if (vd->vdev_ops->vdev_op_min_alloc != NULL)
	min_alloc = vd->vdev_ops->vdev_op_min_alloc(vd);

	return (min_alloc);
	}

	/*
	* Get the parity level for a top-level vdev.
	*/
	uint64_t
	vdev_get_nparity(vdev_t *vd)
	{
	uint64_t nparity = 0;

	if (vd->vdev_ops->vdev_op_nparity != NULL)
	nparity = vd->vdev_ops->vdev_op_nparity(vd);

	return (nparity);
	}

	/*
	* Get the number of data disks for a top-level vdev.
	*/
	uint64_t
	vdev_get_ndisks(vdev_t *vd)
	{
	uint64_t ndisks = 1;

	if (vd->vdev_ops->vdev_op_ndisks != NULL)
	ndisks = vd->vdev_ops->vdev_op_ndisks(vd);

	return (ndisks);
	}

	vdev_t *
	vdev_lookup_top(spa_t *spa, uint64_t vdev)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	if (vdev < rvd->vdev_children) {
	ASSERT(rvd->vdev_child[vdev] != NULL);
	return (rvd->vdev_child[vdev]);
	}

	return (NULL);
	}

	vdev_t *
	vdev_lookup_by_guid(vdev_t *vd, uint64_t guid)
	{
	vdev_t *mvd;

	if (vd->vdev_guid == guid)
	return (vd);

	for (int c = 0; c < vd->vdev_children; c++)
	if ((mvd = vdev_lookup_by_guid(vd->vdev_child[c], guid)) !=
	NULL)
	return (mvd);

	return (NULL);
	}

	static int
	vdev_count_leaves_impl(vdev_t *vd)
	{
	int n = 0;

	if (vd->vdev_ops->vdev_op_leaf)
	return (1);

	for (int c = 0; c < vd->vdev_children; c++)
	n += vdev_count_leaves_impl(vd->vdev_child[c]);

	return (n);
	}

	int
	vdev_count_leaves(spa_t *spa)
	{
	int rc;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	rc = vdev_count_leaves_impl(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	return (rc);
	}

	void
	vdev_add_child(vdev_t pvd, vdev_t cvd)
	{
	size_t oldsize, newsize;
	uint64_t id = cvd->vdev_id;
	vdev_t **newchild;

	ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
	ASSERT(cvd->vdev_parent == NULL);

	cvd->vdev_parent = pvd;

	if (pvd == NULL)
	return;

	ASSERT(id >= pvd->vdev_children \|\| pvd->vdev_child[id] == NULL);

	oldsize = pvd->vdev_children * sizeof (vdev_t *);
	pvd->vdev_children = MAX(pvd->vdev_children, id + 1);
	newsize = pvd->vdev_children * sizeof (vdev_t *);

	newchild = kmem_alloc(newsize, KM_SLEEP);
	if (pvd->vdev_child != NULL) {
	memcpy(newchild, pvd->vdev_child, oldsize);
	kmem_free(pvd->vdev_child, oldsize);
	}

	pvd->vdev_child = newchild;
	pvd->vdev_child[id] = cvd;

	cvd->vdev_top = (pvd->vdev_top ? pvd->vdev_top: cvd);
	ASSERT(cvd->vdev_top->vdev_parent->vdev_parent == NULL);

	/*
	* Walk up all ancestors to update guid sum.
	*/
	for (; pvd != NULL; pvd = pvd->vdev_parent)
	pvd->vdev_guid_sum += cvd->vdev_guid_sum;

	if (cvd->vdev_ops->vdev_op_leaf) {
	list_insert_head(&cvd->vdev_spa->spa_leaf_list, cvd);
	cvd->vdev_spa->spa_leaf_list_gen++;
	}
	}

	void
	vdev_remove_child(vdev_t pvd, vdev_t cvd)
	{
	int c;
	uint_t id = cvd->vdev_id;

	ASSERT(cvd->vdev_parent == pvd);

	if (pvd == NULL)
	return;

	ASSERT(id < pvd->vdev_children);
	ASSERT(pvd->vdev_child[id] == cvd);

	pvd->vdev_child[id] = NULL;
	cvd->vdev_parent = NULL;

	for (c = 0; c < pvd->vdev_children; c++)
	if (pvd->vdev_child[c])
	break;

	if (c == pvd->vdev_children) {
	kmem_free(pvd->vdev_child, c * sizeof (vdev_t *));
	pvd->vdev_child = NULL;
	pvd->vdev_children = 0;
	}

	if (cvd->vdev_ops->vdev_op_leaf) {
	spa_t *spa = cvd->vdev_spa;
	list_remove(&spa->spa_leaf_list, cvd);
	spa->spa_leaf_list_gen++;
	}

	/*
	* Walk up all ancestors to update guid sum.
	*/
	for (; pvd != NULL; pvd = pvd->vdev_parent)
	pvd->vdev_guid_sum -= cvd->vdev_guid_sum;
	}

	/*
	* Remove any holes in the child array.
	*/
	void
	vdev_compact_children(vdev_t *pvd)
	{
	vdev_t *newchild, cvd;
	int oldc = pvd->vdev_children;
	int newc;

	ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if (oldc == 0)
	return;

	for (int c = newc = 0; c < oldc; c++)
	if (pvd->vdev_child[c])
	newc++;

	if (newc > 0) {
	newchild = kmem_zalloc(newc * sizeof (vdev_t *), KM_SLEEP);

	for (int c = newc = 0; c < oldc; c++) {
	if ((cvd = pvd->vdev_child[c]) != NULL) {
	newchild[newc] = cvd;
	cvd->vdev_id = newc++;
	}
	}
	} else {
	newchild = NULL;
	}

	kmem_free(pvd->vdev_child, oldc * sizeof (vdev_t *));
	pvd->vdev_child = newchild;
	pvd->vdev_children = newc;
	}

	/*
	* Allocate and minimally initialize a vdev_t.
	*/
	vdev_t *
	vdev_alloc_common(spa_t spa, uint_t id, uint64_t guid, vdev_ops_t ops)
	{
	vdev_t *vd;
	vdev_indirect_config_t *vic;

	vd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
	vic = &vd->vdev_indirect_config;

	if (spa->spa_root_vdev == NULL) {
	ASSERT(ops == &vdev_root_ops);
	spa->spa_root_vdev = vd;
	spa->spa_load_guid = spa_generate_guid(NULL);
	}

	if (guid == 0 && ops != &vdev_hole_ops) {
	if (spa->spa_root_vdev == vd) {
	/*
	* The root vdev's guid will also be the pool guid,
	* which must be unique among all pools.
	*/
	guid = spa_generate_guid(NULL);
	} else {
	/*
	* Any other vdev's guid must be unique within the pool.
	*/
	guid = spa_generate_guid(spa);
	}
	ASSERT(!spa_guid_exists(spa_guid(spa), guid));
	}

	vd->vdev_spa = spa;
	vd->vdev_id = id;
	vd->vdev_guid = guid;
	vd->vdev_guid_sum = guid;
	vd->vdev_ops = ops;
	vd->vdev_state = VDEV_STATE_CLOSED;
	vd->vdev_ishole = (ops == &vdev_hole_ops);
	vic->vic_prev_indirect_vdev = UINT64_MAX;

	rw_init(&vd->vdev_indirect_rwlock, NULL, RW_DEFAULT, NULL);
	mutex_init(&vd->vdev_obsolete_lock, NULL, MUTEX_DEFAULT, NULL);
	vd->vdev_obsolete_segments = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);

	/*
	* Initialize rate limit structs for events. We rate limit ZIO delay
	* and checksum events so that we don't overwhelm ZED with thousands
	* of events when a disk is acting up.
	*/
	zfs_ratelimit_init(&vd->vdev_delay_rl, &zfs_slow_io_events_per_second,
	1);
	zfs_ratelimit_init(&vd->vdev_deadman_rl, &zfs_slow_io_events_per_second,
	1);
	zfs_ratelimit_init(&vd->vdev_checksum_rl,
	&zfs_checksum_events_per_second, 1);

	list_link_init(&vd->vdev_config_dirty_node);
	list_link_init(&vd->vdev_state_dirty_node);
	list_link_init(&vd->vdev_initialize_node);
	list_link_init(&vd->vdev_leaf_node);
	list_link_init(&vd->vdev_trim_node);

	mutex_init(&vd->vdev_dtl_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_scan_io_queue_lock, NULL, MUTEX_DEFAULT, NULL);

	mutex_init(&vd->vdev_initialize_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_initialize_io_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_initialize_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_initialize_io_cv, NULL, CV_DEFAULT, NULL);

	mutex_init(&vd->vdev_trim_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_autotrim_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_trim_io_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_trim_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_autotrim_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_trim_io_cv, NULL, CV_DEFAULT, NULL);

	mutex_init(&vd->vdev_rebuild_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_rebuild_cv, NULL, CV_DEFAULT, NULL);

	for (int t = 0; t < DTL_TYPES; t++) {
	vd->vdev_dtl[t] = range_tree_create(NULL, RANGE_SEG64, NULL, 0,
	0);
	}

	txg_list_create(&vd->vdev_ms_list, spa,
	offsetof(struct metaslab, ms_txg_node));
	txg_list_create(&vd->vdev_dtl_list, spa,
	offsetof(struct vdev, vdev_dtl_node));
	vd->vdev_stat.vs_timestamp = gethrtime();
	vdev_queue_init(vd);
	vdev_cache_init(vd);

	return (vd);
	}

	/*
	* Allocate a new vdev. The 'alloctype' is used to control whether we are
	* creating a new vdev or loading an existing one - the behavior is slightly
	* different for each case.
	*/
	int
	vdev_alloc(spa_t spa, vdev_t vdp, nvlist_t nv, vdev_t *parent, uint_t id,
	int alloctype)
	{
	vdev_ops_t *ops;
	char *type;
	uint64_t guid = 0, islog;
	vdev_t *vd;
	vdev_indirect_config_t *vic;
	char *tmp = NULL;
	int rc;
	vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;
	boolean_t top_level = (parent && !parent->vdev_parent);

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
	return (SET_ERROR(EINVAL));

	if ((ops = vdev_getops(type)) == NULL)
	return (SET_ERROR(EINVAL));

	/*
	* If this is a load, get the vdev guid from the nvlist.
	* Otherwise, vdev_alloc_common() will generate one for us.
	*/
	if (alloctype == VDEV_ALLOC_LOAD) {
	uint64_t label_id;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, &label_id) \|\|
	label_id != id)
	return (SET_ERROR(EINVAL));

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_SPARE) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_L2CACHE) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_ROOTPOOL) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* The first allocated vdev must be of type 'root'.
	*/
	if (ops != &vdev_root_ops && spa->spa_root_vdev == NULL)
	return (SET_ERROR(EINVAL));

	/*
	* Determine whether we're a log vdev.
	*/
	islog = 0;
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &islog);
	if (islog && spa_version(spa) < SPA_VERSION_SLOGS)
	return (SET_ERROR(ENOTSUP));

	if (ops == &vdev_hole_ops && spa_version(spa) < SPA_VERSION_HOLES)
	return (SET_ERROR(ENOTSUP));

	if (top_level && alloctype == VDEV_ALLOC_ADD) {
	char *bias;

	/*
	* If creating a top-level vdev, check for allocation
	* classes input.
	*/
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS,
	&bias) == 0) {
	alloc_bias = vdev_derive_alloc_bias(bias);

	/* spa_vdev_add() expects feature to be enabled */
	if (spa->spa_load_state != SPA_LOAD_CREATE &&
	!spa_feature_is_enabled(spa,
	SPA_FEATURE_ALLOCATION_CLASSES)) {
	return (SET_ERROR(ENOTSUP));
	}
	}

	/* spa_vdev_add() expects feature to be enabled */
	if (ops == &vdev_draid_ops &&
	spa->spa_load_state != SPA_LOAD_CREATE &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_DRAID)) {
	return (SET_ERROR(ENOTSUP));
	}
	}

	/*
	* Initialize the vdev specific data. This is done before calling
	* vdev_alloc_common() since it may fail and this simplifies the
	* error reporting and cleanup code paths.
	*/
	void *tsd = NULL;
	if (ops->vdev_op_init != NULL) {
	rc = ops->vdev_op_init(spa, nv, &tsd);
	if (rc != 0) {
	return (rc);
	}
	}

	vd = vdev_alloc_common(spa, id, guid, ops);
	vd->vdev_tsd = tsd;
	vd->vdev_islog = islog;

	if (top_level && alloc_bias != VDEV_BIAS_NONE)
	vd->vdev_alloc_bias = alloc_bias;

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &vd->vdev_path) == 0)
	vd->vdev_path = spa_strdup(vd->vdev_path);

	/*
	* ZPOOL_CONFIG_AUX_STATE = "external" means we previously forced a
	* fault on a vdev and want it to persist across imports (like with
	* zpool offline -f).
	*/
	rc = nvlist_lookup_string(nv, ZPOOL_CONFIG_AUX_STATE, &tmp);
	if (rc == 0 && tmp != NULL && strcmp(tmp, "external") == 0) {
	vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
	vd->vdev_faulted = 1;
	vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
	}

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &vd->vdev_devid) == 0)
	vd->vdev_devid = spa_strdup(vd->vdev_devid);
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
	&vd->vdev_physpath) == 0)
	vd->vdev_physpath = spa_strdup(vd->vdev_physpath);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
	&vd->vdev_enc_sysfs_path) == 0)
	vd->vdev_enc_sysfs_path = spa_strdup(vd->vdev_enc_sysfs_path);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_FRU, &vd->vdev_fru) == 0)
	vd->vdev_fru = spa_strdup(vd->vdev_fru);

	/*
	* Set the whole_disk property. If it's not specified, leave the value
	* as -1.
	*/
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
	&vd->vdev_wholedisk) != 0)
	vd->vdev_wholedisk = -1ULL;

	vic = &vd->vdev_indirect_config;

	ASSERT0(vic->vic_mapping_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT,
	&vic->vic_mapping_object);
	ASSERT0(vic->vic_births_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS,
	&vic->vic_births_object);
	ASSERT3U(vic->vic_prev_indirect_vdev, ==, UINT64_MAX);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV,
	&vic->vic_prev_indirect_vdev);

	/*
	* Look for the 'not present' flag. This will only be set if the device
	* was not present at the time of import.
	*/
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
	&vd->vdev_not_present);

	/*
	* Get the alignment requirement.
	*/
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &vd->vdev_ashift);

	/*
	* Retrieve the vdev creation time.
	*/
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_CREATE_TXG,
	&vd->vdev_crtxg);

	/*
	* If we're a top-level vdev, try to load the allocation parameters.
	*/
	if (top_level &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_SPLIT)) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
	&vd->vdev_ms_array);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
	&vd->vdev_ms_shift);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASIZE,
	&vd->vdev_asize);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NONALLOCATING,
	&vd->vdev_noalloc);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVING,
	&vd->vdev_removing);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
	&vd->vdev_top_zap);
	} else {
	ASSERT0(vd->vdev_top_zap);
	}

	if (top_level && alloctype != VDEV_ALLOC_ATTACH) {
	ASSERT(alloctype == VDEV_ALLOC_LOAD \|\|
	alloctype == VDEV_ALLOC_ADD \|\|
	alloctype == VDEV_ALLOC_SPLIT \|\|
	alloctype == VDEV_ALLOC_ROOTPOOL);
	/* Note: metaslab_group_create() is now deferred */
	}

	if (vd->vdev_ops->vdev_op_leaf &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_SPLIT)) {
	(void) nvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_VDEV_LEAF_ZAP, &vd->vdev_leaf_zap);
	} else {
	ASSERT0(vd->vdev_leaf_zap);
	}

	/*
	* If we're a leaf vdev, try to load the DTL object and other state.
	*/

	if (vd->vdev_ops->vdev_op_leaf &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_L2CACHE \|\|
	alloctype == VDEV_ALLOC_ROOTPOOL)) {
	if (alloctype == VDEV_ALLOC_LOAD) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
	&vd->vdev_dtl_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
	&vd->vdev_unspare);
	}

	if (alloctype == VDEV_ALLOC_ROOTPOOL) {
	uint64_t spare = 0;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
	&spare) == 0 && spare)
	spa_spare_add(vd);
	}

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE,
	&vd->vdev_offline);

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG,
	&vd->vdev_resilver_txg);

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REBUILD_TXG,
	&vd->vdev_rebuild_txg);

	if (nvlist_exists(nv, ZPOOL_CONFIG_RESILVER_DEFER))
	vdev_defer_resilver(vd);

	/*
	* In general, when importing a pool we want to ignore the
	* persistent fault state, as the diagnosis made on another
	* system may not be valid in the current context. The only
	* exception is if we forced a vdev to a persistently faulted
	* state with 'zpool offline -f'. The persistent fault will
	* remain across imports until cleared.
	*
	* Local vdevs will remain in the faulted state.
	*/
	if (spa_load_state(spa) == SPA_LOAD_OPEN \|\|
	spa_load_state(spa) == SPA_LOAD_IMPORT) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED,
	&vd->vdev_faulted);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DEGRADED,
	&vd->vdev_degraded);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED,
	&vd->vdev_removed);

	if (vd->vdev_faulted \|\| vd->vdev_degraded) {
	char *aux;

	vd->vdev_label_aux =
	VDEV_AUX_ERR_EXCEEDED;
	if (nvlist_lookup_string(nv,
	ZPOOL_CONFIG_AUX_STATE, &aux) == 0 &&
	strcmp(aux, "external") == 0)
	vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
	else
	vd->vdev_faulted = 0ULL;
	}
	}
	}

	/*
	* Add ourselves to the parent's list of children.
	*/
	vdev_add_child(parent, vd);

	*vdp = vd;

	return (0);
	}

	void
	vdev_free(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	ASSERT3P(vd->vdev_autotrim_thread, ==, NULL);
	ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);

	/*
	* Scan queues are normally destroyed at the end of a scan. If the
	* queue exists here, that implies the vdev is being removed while
	* the scan is still running.
	*/
	if (vd->vdev_scan_io_queue != NULL) {
	mutex_enter(&vd->vdev_scan_io_queue_lock);
	dsl_scan_io_queue_destroy(vd->vdev_scan_io_queue);
	vd->vdev_scan_io_queue = NULL;
	mutex_exit(&vd->vdev_scan_io_queue_lock);
	}

	/*
	* vdev_free() implies closing the vdev first. This is simpler than
	* trying to ensure complicated semantics for all callers.
	*/
	vdev_close(vd);

	ASSERT(!list_link_active(&vd->vdev_config_dirty_node));
	ASSERT(!list_link_active(&vd->vdev_state_dirty_node));

	/*
	* Free all children.
	*/
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_free(vd->vdev_child[c]);

	ASSERT(vd->vdev_child == NULL);
	ASSERT(vd->vdev_guid_sum == vd->vdev_guid);

	if (vd->vdev_ops->vdev_op_fini != NULL)
	vd->vdev_ops->vdev_op_fini(vd);

	/*
	* Discard allocation state.
	*/
	if (vd->vdev_mg != NULL) {
	vdev_metaslab_fini(vd);
	metaslab_group_destroy(vd->vdev_mg);
	vd->vdev_mg = NULL;
	}
	if (vd->vdev_log_mg != NULL) {
	ASSERT0(vd->vdev_ms_count);
	metaslab_group_destroy(vd->vdev_log_mg);
	vd->vdev_log_mg = NULL;
	}

	ASSERT0(vd->vdev_stat.vs_space);
	ASSERT0(vd->vdev_stat.vs_dspace);
	ASSERT0(vd->vdev_stat.vs_alloc);

	/*
	* Remove this vdev from its parent's child list.
	*/
	vdev_remove_child(vd->vdev_parent, vd);

	ASSERT(vd->vdev_parent == NULL);
	ASSERT(!list_link_active(&vd->vdev_leaf_node));

	/*
	* Clean up vdev structure.
	*/
	vdev_queue_fini(vd);
	vdev_cache_fini(vd);

	if (vd->vdev_path)
	spa_strfree(vd->vdev_path);
	if (vd->vdev_devid)
	spa_strfree(vd->vdev_devid);
	if (vd->vdev_physpath)
	spa_strfree(vd->vdev_physpath);

	if (vd->vdev_enc_sysfs_path)
	spa_strfree(vd->vdev_enc_sysfs_path);

	if (vd->vdev_fru)
	spa_strfree(vd->vdev_fru);

	if (vd->vdev_isspare)
	spa_spare_remove(vd);
	if (vd->vdev_isl2cache)
	spa_l2cache_remove(vd);

	txg_list_destroy(&vd->vdev_ms_list);
	txg_list_destroy(&vd->vdev_dtl_list);

	mutex_enter(&vd->vdev_dtl_lock);
	space_map_close(vd->vdev_dtl_sm);
	for (int t = 0; t < DTL_TYPES; t++) {
	range_tree_vacate(vd->vdev_dtl[t], NULL, NULL);
	range_tree_destroy(vd->vdev_dtl[t]);
	}
	mutex_exit(&vd->vdev_dtl_lock);

	EQUIV(vd->vdev_indirect_births != NULL,
	vd->vdev_indirect_mapping != NULL);
	if (vd->vdev_indirect_births != NULL) {
	vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
	vdev_indirect_births_close(vd->vdev_indirect_births);
	}

	if (vd->vdev_obsolete_sm != NULL) {
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);
	space_map_close(vd->vdev_obsolete_sm);
	vd->vdev_obsolete_sm = NULL;
	}
	range_tree_destroy(vd->vdev_obsolete_segments);
	rw_destroy(&vd->vdev_indirect_rwlock);
	mutex_destroy(&vd->vdev_obsolete_lock);

	mutex_destroy(&vd->vdev_dtl_lock);
	mutex_destroy(&vd->vdev_stat_lock);
	mutex_destroy(&vd->vdev_probe_lock);
	mutex_destroy(&vd->vdev_scan_io_queue_lock);

	mutex_destroy(&vd->vdev_initialize_lock);
	mutex_destroy(&vd->vdev_initialize_io_lock);
	cv_destroy(&vd->vdev_initialize_io_cv);
	cv_destroy(&vd->vdev_initialize_cv);

	mutex_destroy(&vd->vdev_trim_lock);
	mutex_destroy(&vd->vdev_autotrim_lock);
	mutex_destroy(&vd->vdev_trim_io_lock);
	cv_destroy(&vd->vdev_trim_cv);
	cv_destroy(&vd->vdev_autotrim_cv);
	cv_destroy(&vd->vdev_trim_io_cv);

	mutex_destroy(&vd->vdev_rebuild_lock);
	cv_destroy(&vd->vdev_rebuild_cv);

	zfs_ratelimit_fini(&vd->vdev_delay_rl);
	zfs_ratelimit_fini(&vd->vdev_deadman_rl);
	zfs_ratelimit_fini(&vd->vdev_checksum_rl);

	if (vd == spa->spa_root_vdev)
	spa->spa_root_vdev = NULL;

	kmem_free(vd, sizeof (vdev_t));
	}

	/*
	* Transfer top-level vdev state from svd to tvd.
	*/
	static void
	vdev_top_transfer(vdev_t svd, vdev_t tvd)
	{
	spa_t *spa = svd->vdev_spa;
	metaslab_t *msp;
	vdev_t *vd;
	int t;

	ASSERT(tvd == tvd->vdev_top);

	tvd->vdev_pending_fastwrite = svd->vdev_pending_fastwrite;
	tvd->vdev_ms_array = svd->vdev_ms_array;
	tvd->vdev_ms_shift = svd->vdev_ms_shift;
	tvd->vdev_ms_count = svd->vdev_ms_count;
	tvd->vdev_top_zap = svd->vdev_top_zap;

	svd->vdev_ms_array = 0;
	svd->vdev_ms_shift = 0;
	svd->vdev_ms_count = 0;
	svd->vdev_top_zap = 0;

	if (tvd->vdev_mg)
	ASSERT3P(tvd->vdev_mg, ==, svd->vdev_mg);
	if (tvd->vdev_log_mg)
	ASSERT3P(tvd->vdev_log_mg, ==, svd->vdev_log_mg);
	tvd->vdev_mg = svd->vdev_mg;
	tvd->vdev_log_mg = svd->vdev_log_mg;
	tvd->vdev_ms = svd->vdev_ms;

	svd->vdev_mg = NULL;
	svd->vdev_log_mg = NULL;
	svd->vdev_ms = NULL;

	if (tvd->vdev_mg != NULL)
	tvd->vdev_mg->mg_vd = tvd;
	if (tvd->vdev_log_mg != NULL)
	tvd->vdev_log_mg->mg_vd = tvd;

	tvd->vdev_checkpoint_sm = svd->vdev_checkpoint_sm;
	svd->vdev_checkpoint_sm = NULL;

	tvd->vdev_alloc_bias = svd->vdev_alloc_bias;
	svd->vdev_alloc_bias = VDEV_BIAS_NONE;

	tvd->vdev_stat.vs_alloc = svd->vdev_stat.vs_alloc;
	tvd->vdev_stat.vs_space = svd->vdev_stat.vs_space;
	tvd->vdev_stat.vs_dspace = svd->vdev_stat.vs_dspace;

	svd->vdev_stat.vs_alloc = 0;
	svd->vdev_stat.vs_space = 0;
	svd->vdev_stat.vs_dspace = 0;

	/*
	* State which may be set on a top-level vdev that's in the
	* process of being removed.
	*/
	ASSERT0(tvd->vdev_indirect_config.vic_births_object);
	ASSERT0(tvd->vdev_indirect_config.vic_mapping_object);
	ASSERT3U(tvd->vdev_indirect_config.vic_prev_indirect_vdev, ==, -1ULL);
	ASSERT3P(tvd->vdev_indirect_mapping, ==, NULL);
	ASSERT3P(tvd->vdev_indirect_births, ==, NULL);
	ASSERT3P(tvd->vdev_obsolete_sm, ==, NULL);
	ASSERT0(tvd->vdev_noalloc);
	ASSERT0(tvd->vdev_removing);
	ASSERT0(tvd->vdev_rebuilding);
	tvd->vdev_noalloc = svd->vdev_noalloc;
	tvd->vdev_removing = svd->vdev_removing;
	tvd->vdev_rebuilding = svd->vdev_rebuilding;
	tvd->vdev_rebuild_config = svd->vdev_rebuild_config;
	tvd->vdev_indirect_config = svd->vdev_indirect_config;
	tvd->vdev_indirect_mapping = svd->vdev_indirect_mapping;
	tvd->vdev_indirect_births = svd->vdev_indirect_births;
	range_tree_swap(&svd->vdev_obsolete_segments,
	&tvd->vdev_obsolete_segments);
	tvd->vdev_obsolete_sm = svd->vdev_obsolete_sm;
	svd->vdev_indirect_config.vic_mapping_object = 0;
	svd->vdev_indirect_config.vic_births_object = 0;
	svd->vdev_indirect_config.vic_prev_indirect_vdev = -1ULL;
	svd->vdev_indirect_mapping = NULL;
	svd->vdev_indirect_births = NULL;
	svd->vdev_obsolete_sm = NULL;
	svd->vdev_noalloc = 0;
	svd->vdev_removing = 0;
	svd->vdev_rebuilding = 0;

	for (t = 0; t < TXG_SIZE; t++) {
	while ((msp = txg_list_remove(&svd->vdev_ms_list, t)) != NULL)
	(void) txg_list_add(&tvd->vdev_ms_list, msp, t);
	while ((vd = txg_list_remove(&svd->vdev_dtl_list, t)) != NULL)
	(void) txg_list_add(&tvd->vdev_dtl_list, vd, t);
	if (txg_list_remove_this(&spa->spa_vdev_txg_list, svd, t))
	(void) txg_list_add(&spa->spa_vdev_txg_list, tvd, t);
	}

	if (list_link_active(&svd->vdev_config_dirty_node)) {
	vdev_config_clean(svd);
	vdev_config_dirty(tvd);
	}

	if (list_link_active(&svd->vdev_state_dirty_node)) {
	vdev_state_clean(svd);
	vdev_state_dirty(tvd);
	}

	tvd->vdev_deflate_ratio = svd->vdev_deflate_ratio;
	svd->vdev_deflate_ratio = 0;

	tvd->vdev_islog = svd->vdev_islog;
	svd->vdev_islog = 0;

	dsl_scan_io_queue_vdev_xfer(svd, tvd);
	}

	static void
	vdev_top_update(vdev_t tvd, vdev_t vd)
	{
	if (vd == NULL)
	return;

	vd->vdev_top = tvd;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_top_update(tvd, vd->vdev_child[c]);
	}

	/*
	* Add a mirror/replacing vdev above an existing vdev. There is no need to
	* call .vdev_op_init() since mirror/replacing vdevs do not have private state.
	*/
	vdev_t *
	vdev_add_parent(vdev_t cvd, vdev_ops_t ops)
	{
	spa_t *spa = cvd->vdev_spa;
	vdev_t *pvd = cvd->vdev_parent;
	vdev_t *mvd;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	mvd = vdev_alloc_common(spa, cvd->vdev_id, 0, ops);

	mvd->vdev_asize = cvd->vdev_asize;
	mvd->vdev_min_asize = cvd->vdev_min_asize;
	mvd->vdev_max_asize = cvd->vdev_max_asize;
	mvd->vdev_psize = cvd->vdev_psize;
	mvd->vdev_ashift = cvd->vdev_ashift;
	mvd->vdev_logical_ashift = cvd->vdev_logical_ashift;
	mvd->vdev_physical_ashift = cvd->vdev_physical_ashift;
	mvd->vdev_state = cvd->vdev_state;
	mvd->vdev_crtxg = cvd->vdev_crtxg;

	vdev_remove_child(pvd, cvd);
	vdev_add_child(pvd, mvd);
	cvd->vdev_id = mvd->vdev_children;
	vdev_add_child(mvd, cvd);
	vdev_top_update(cvd->vdev_top, cvd->vdev_top);

	if (mvd == mvd->vdev_top)
	vdev_top_transfer(cvd, mvd);

	return (mvd);
	}

	/*
	* Remove a 1-way mirror/replacing vdev from the tree.
	*/
	void
	vdev_remove_parent(vdev_t *cvd)
	{
	vdev_t *mvd = cvd->vdev_parent;
	vdev_t *pvd = mvd->vdev_parent;

	ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	ASSERT(mvd->vdev_children == 1);
	ASSERT(mvd->vdev_ops == &vdev_mirror_ops \|\|
	mvd->vdev_ops == &vdev_replacing_ops \|\|
	mvd->vdev_ops == &vdev_spare_ops);
	cvd->vdev_ashift = mvd->vdev_ashift;
	cvd->vdev_logical_ashift = mvd->vdev_logical_ashift;
	cvd->vdev_physical_ashift = mvd->vdev_physical_ashift;
	vdev_remove_child(mvd, cvd);
	vdev_remove_child(pvd, mvd);

	/*
	* If cvd will replace mvd as a top-level vdev, preserve mvd's guid.
	* Otherwise, we could have detached an offline device, and when we
	* go to import the pool we'll think we have two top-level vdevs,
	* instead of a different version of the same top-level vdev.
	*/
	if (mvd->vdev_top == mvd) {
	uint64_t guid_delta = mvd->vdev_guid - cvd->vdev_guid;
	cvd->vdev_orig_guid = cvd->vdev_guid;
	cvd->vdev_guid += guid_delta;
	cvd->vdev_guid_sum += guid_delta;

	/*
	* If pool not set for autoexpand, we need to also preserve
	* mvd's asize to prevent automatic expansion of cvd.
	* Otherwise if we are adjusting the mirror by attaching and
	* detaching children of non-uniform sizes, the mirror could
	* autoexpand, unexpectedly requiring larger devices to
	* re-establish the mirror.
	*/
	if (!cvd->vdev_spa->spa_autoexpand)
	cvd->vdev_asize = mvd->vdev_asize;
	}
	cvd->vdev_id = mvd->vdev_id;
	vdev_add_child(pvd, cvd);
	vdev_top_update(cvd->vdev_top, cvd->vdev_top);

	if (cvd == cvd->vdev_top)
	vdev_top_transfer(mvd, cvd);

	ASSERT(mvd->vdev_children == 0);
	vdev_free(mvd);
	}

	void
	vdev_metaslab_group_create(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	/*
	* metaslab_group_create was delayed until allocation bias was available
	*/
	if (vd->vdev_mg == NULL) {
	metaslab_class_t *mc;

	if (vd->vdev_islog && vd->vdev_alloc_bias == VDEV_BIAS_NONE)
	vd->vdev_alloc_bias = VDEV_BIAS_LOG;

	ASSERT3U(vd->vdev_islog, ==,
	(vd->vdev_alloc_bias == VDEV_BIAS_LOG));

	switch (vd->vdev_alloc_bias) {
	case VDEV_BIAS_LOG:
	mc = spa_log_class(spa);
	break;
	case VDEV_BIAS_SPECIAL:
	mc = spa_special_class(spa);
	break;
	case VDEV_BIAS_DEDUP:
	mc = spa_dedup_class(spa);
	break;
	default:
	mc = spa_normal_class(spa);
	}

	vd->vdev_mg = metaslab_group_create(mc, vd,
	spa->spa_alloc_count);

	if (!vd->vdev_islog) {
	vd->vdev_log_mg = metaslab_group_create(
	spa_embedded_log_class(spa), vd, 1);
	}

	/*
	* The spa ashift min/max only apply for the normal metaslab
	* class. Class destination is late binding so ashift boundary
	* setting had to wait until now.
	*/
	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
	mc == spa_normal_class(spa) && vd->vdev_aux == NULL) {
	if (vd->vdev_ashift > spa->spa_max_ashift)
	spa->spa_max_ashift = vd->vdev_ashift;
	if (vd->vdev_ashift < spa->spa_min_ashift)
	spa->spa_min_ashift = vd->vdev_ashift;

	uint64_t min_alloc = vdev_get_min_alloc(vd);
	if (min_alloc < spa->spa_min_alloc)
	spa->spa_min_alloc = min_alloc;
	}
	}
	}

	int
	vdev_metaslab_init(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t oldc = vd->vdev_ms_count;
	uint64_t newc = vd->vdev_asize >> vd->vdev_ms_shift;
	metaslab_t **mspp;
	int error;
	boolean_t expanding = (oldc != 0);

	ASSERT(txg == 0 \|\| spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	/*
	* This vdev is not being allocated from yet or is a hole.
	*/
	if (vd->vdev_ms_shift == 0)
	return (0);

	ASSERT(!vd->vdev_ishole);

	ASSERT(oldc <= newc);

	mspp = vmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);

	if (expanding) {
	memcpy(mspp, vd->vdev_ms, oldc * sizeof (*mspp));
	vmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
	}

	vd->vdev_ms = mspp;
	vd->vdev_ms_count = newc;

	for (uint64_t m = oldc; m < newc; m++) {
	uint64_t object = 0;
	/*
	* vdev_ms_array may be 0 if we are creating the "fake"
	* metaslabs for an indirect vdev for zdb's leak detection.
	* See zdb_leak_init().
	*/
	if (txg == 0 && vd->vdev_ms_array != 0) {
	error = dmu_read(spa->spa_meta_objset,
	vd->vdev_ms_array,
	m * sizeof (uint64_t), sizeof (uint64_t), &object,
	DMU_READ_PREFETCH);
	if (error != 0) {
	vdev_dbgmsg(vd, "unable to read the metaslab "
	"array [error=%d]", error);
	return (error);
	}
	}

	error = metaslab_init(vd->vdev_mg, m, object, txg,
	&(vd->vdev_ms[m]));
	if (error != 0) {
	vdev_dbgmsg(vd, "metaslab_init failed [error=%d]",
	error);
	return (error);
	}
	}

	/*
	* Find the emptiest metaslab on the vdev and mark it for use for
	* embedded slog by moving it from the regular to the log metaslab
	* group.
	*/
	if (vd->vdev_mg->mg_class == spa_normal_class(spa) &&
	vd->vdev_ms_count > zfs_embedded_slog_min_ms &&
	avl_is_empty(&vd->vdev_log_mg->mg_metaslab_tree)) {
	uint64_t slog_msid = 0;
	uint64_t smallest = UINT64_MAX;

	/*
	* Note, we only search the new metaslabs, because the old
	* (pre-existing) ones may be active (e.g. have non-empty
	* range_tree's), and we don't move them to the new
	* metaslab_t.
	*/
	for (uint64_t m = oldc; m < newc; m++) {
	uint64_t alloc =
	space_map_allocated(vd->vdev_ms[m]->ms_sm);
	if (alloc < smallest) {
	slog_msid = m;
	smallest = alloc;
	}
	}
	metaslab_t *slog_ms = vd->vdev_ms[slog_msid];
	/*
	* The metaslab was marked as dirty at the end of
	* metaslab_init(). Remove it from the dirty list so that we
	* can uninitialize and reinitialize it to the new class.
	*/
	if (txg != 0) {
	(void) txg_list_remove_this(&vd->vdev_ms_list,
	slog_ms, txg);
	}
	uint64_t sm_obj = space_map_object(slog_ms->ms_sm);
	metaslab_fini(slog_ms);
	VERIFY0(metaslab_init(vd->vdev_log_mg, slog_msid, sm_obj, txg,
	&vd->vdev_ms[slog_msid]));
	}

	if (txg == 0)
	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_WRITER);

	/*
	* If the vdev is marked as non-allocating then don't
	* activate the metaslabs since we want to ensure that
	* no allocations are performed on this device.
	*/
	if (vd->vdev_noalloc) {
	/* track non-allocating vdev space */
	spa->spa_nonallocating_dspace += spa_deflate(spa) ?
	vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
	} else if (!expanding) {
	metaslab_group_activate(vd->vdev_mg);
	if (vd->vdev_log_mg != NULL)
	metaslab_group_activate(vd->vdev_log_mg);
	}

	if (txg == 0)
	spa_config_exit(spa, SCL_ALLOC, FTAG);

	return (0);
	}

	void
	vdev_metaslab_fini(vdev_t *vd)
	{
	if (vd->vdev_checkpoint_sm != NULL) {
	ASSERT(spa_feature_is_active(vd->vdev_spa,
	SPA_FEATURE_POOL_CHECKPOINT));
	space_map_close(vd->vdev_checkpoint_sm);
	/*
	* Even though we close the space map, we need to set its
	* pointer to NULL. The reason is that vdev_metaslab_fini()
	* may be called multiple times for certain operations
	* (i.e. when destroying a pool) so we need to ensure that
	* this clause never executes twice. This logic is similar
	* to the one used for the vdev_ms clause below.
	*/
	vd->vdev_checkpoint_sm = NULL;
	}

	if (vd->vdev_ms != NULL) {
	metaslab_group_t *mg = vd->vdev_mg;

	metaslab_group_passivate(mg);
	if (vd->vdev_log_mg != NULL) {
	ASSERT(!vd->vdev_islog);
	metaslab_group_passivate(vd->vdev_log_mg);
	}

	uint64_t count = vd->vdev_ms_count;
	for (uint64_t m = 0; m < count; m++) {
	metaslab_t *msp = vd->vdev_ms[m];
	if (msp != NULL)
	metaslab_fini(msp);
	}
	vmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
	vd->vdev_ms = NULL;
	vd->vdev_ms_count = 0;

	for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	ASSERT0(mg->mg_histogram[i]);
	if (vd->vdev_log_mg != NULL)
	ASSERT0(vd->vdev_log_mg->mg_histogram[i]);
	}
	}
	ASSERT0(vd->vdev_ms_count);
	ASSERT3U(vd->vdev_pending_fastwrite, ==, 0);
	}

	typedef struct vdev_probe_stats {
	boolean_t vps_readable;
	boolean_t vps_writeable;
	int vps_flags;
	} vdev_probe_stats_t;

	static void
	vdev_probe_done(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	vdev_t *vd = zio->io_vd;
	vdev_probe_stats_t *vps = zio->io_private;

	ASSERT(vd->vdev_probe_zio != NULL);

	if (zio->io_type == ZIO_TYPE_READ) {
	if (zio->io_error == 0)
	vps->vps_readable = 1;
	if (zio->io_error == 0 && spa_writeable(spa)) {
	zio_nowait(zio_write_phys(vd->vdev_probe_zio, vd,
	zio->io_offset, zio->io_size, zio->io_abd,
	ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
	ZIO_PRIORITY_SYNC_WRITE, vps->vps_flags, B_TRUE));
	} else {
	abd_free(zio->io_abd);
	}
	} else if (zio->io_type == ZIO_TYPE_WRITE) {
	if (zio->io_error == 0)
	vps->vps_writeable = 1;
	abd_free(zio->io_abd);
	} else if (zio->io_type == ZIO_TYPE_NULL) {
	zio_t *pio;
	zio_link_t *zl;

	vd->vdev_cant_read \|= !vps->vps_readable;
	vd->vdev_cant_write \|= !vps->vps_writeable;

	if (vdev_readable(vd) &&
	(vdev_writeable(vd) \|\| !spa_writeable(spa))) {
	zio->io_error = 0;
	} else {
	ASSERT(zio->io_error != 0);
	vdev_dbgmsg(vd, "failed probe");
	(void) zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
	spa, vd, NULL, NULL, 0);
	zio->io_error = SET_ERROR(ENXIO);
	}

	mutex_enter(&vd->vdev_probe_lock);
	ASSERT(vd->vdev_probe_zio == zio);
	vd->vdev_probe_zio = NULL;
	mutex_exit(&vd->vdev_probe_lock);

	zl = NULL;
	while ((pio = zio_walk_parents(zio, &zl)) != NULL)
	if (!vdev_accessible(vd, pio))
	pio->io_error = SET_ERROR(ENXIO);

	kmem_free(vps, sizeof (*vps));
	}
	}

	/*
	* Determine whether this device is accessible.
	*
	* Read and write to several known locations: the pad regions of each
	* vdev label but the first, which we leave alone in case it contains
	* a VTOC.
	*/
	zio_t *
	vdev_probe(vdev_t vd, zio_t zio)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_probe_stats_t *vps = NULL;
	zio_t *pio;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	/*
	* Don't probe the probe.
	*/
	if (zio && (zio->io_flags & ZIO_FLAG_PROBE))
	return (NULL);

	/*
	* To prevent 'probe storms' when a device fails, we create
	* just one probe i/o at a time. All zios that want to probe
	* this vdev will become parents of the probe io.
	*/
	mutex_enter(&vd->vdev_probe_lock);

	if ((pio = vd->vdev_probe_zio) == NULL) {
	vps = kmem_zalloc(sizeof (*vps), KM_SLEEP);

	vps->vps_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_PROBE \|
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_AGGREGATE \|
	ZIO_FLAG_TRYHARD;

	if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
	/*
	* vdev_cant_read and vdev_cant_write can only
	* transition from TRUE to FALSE when we have the
	* SCL_ZIO lock as writer; otherwise they can only
	* transition from FALSE to TRUE. This ensures that
	* any zio looking at these values can assume that
	* failures persist for the life of the I/O. That's
	* important because when a device has intermittent
	* connectivity problems, we want to ensure that
	* they're ascribed to the device (ENXIO) and not
	* the zio (EIO).
	*
	* Since we hold SCL_ZIO as writer here, clear both
	* values so the probe can reevaluate from first
	* principles.
	*/
	vps->vps_flags \|= ZIO_FLAG_CONFIG_WRITER;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	}

	vd->vdev_probe_zio = pio = zio_null(NULL, spa, vd,
	vdev_probe_done, vps,
	vps->vps_flags \| ZIO_FLAG_DONT_PROPAGATE);

	/*
	* We can't change the vdev state in this context, so we
	* kick off an async task to do it on our behalf.
	*/
	if (zio != NULL) {
	vd->vdev_probe_wanted = B_TRUE;
	spa_async_request(spa, SPA_ASYNC_PROBE);
	}
	}

	if (zio != NULL)
	zio_add_child(zio, pio);

	mutex_exit(&vd->vdev_probe_lock);

	if (vps == NULL) {
	ASSERT(zio != NULL);
	return (NULL);
	}

	for (int l = 1; l < VDEV_LABELS; l++) {
	zio_nowait(zio_read_phys(pio, vd,
	vdev_label_offset(vd->vdev_psize, l,
	offsetof(vdev_label_t, vl_be)), VDEV_PAD_SIZE,
	abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE),
	ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
	ZIO_PRIORITY_SYNC_READ, vps->vps_flags, B_TRUE));
	}

	if (zio == NULL)
	return (pio);

	zio_nowait(pio);
	return (NULL);
	}

	static void
	vdev_load_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_load_error = vdev_load(vd);
	}

	static void
	vdev_open_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_open_thread = curthread;
	vd->vdev_open_error = vdev_open(vd);
	vd->vdev_open_thread = NULL;
	}

	static boolean_t
	vdev_uses_zvols(vdev_t *vd)
	{
	#ifdef _KERNEL
	if (zvol_is_zvol(vd->vdev_path))
	return (B_TRUE);
	#endif

	for (int c = 0; c < vd->vdev_children; c++)
	if (vdev_uses_zvols(vd->vdev_child[c]))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Returns B_TRUE if the passed child should be opened.
	*/
	static boolean_t
	vdev_default_open_children_func(vdev_t *vd)
	{
	(void) vd;
	return (B_TRUE);
	}

	/*
	* Open the requested child vdevs. If any of the leaf vdevs are using
	* a ZFS volume then do the opens in a single thread. This avoids a
	* deadlock when the current thread is holding the spa_namespace_lock.
	*/
	static void
	vdev_open_children_impl(vdev_t vd, vdev_open_children_func_t open_func)
	{
	int children = vd->vdev_children;

	taskq_t *tq = taskq_create("vdev_open", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	vd->vdev_nonrot = B_TRUE;

	for (int c = 0; c < children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (open_func(cvd) == B_FALSE)
	continue;

	if (tq == NULL \|\| vdev_uses_zvols(vd)) {
	cvd->vdev_open_error = vdev_open(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_open_child,
	cvd, TQ_SLEEP) != TASKQID_INVALID);
	}

	vd->vdev_nonrot &= cvd->vdev_nonrot;
	}

	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}
	}

	/*
	* Open all child vdevs.
	*/
	void
	vdev_open_children(vdev_t *vd)
	{
	vdev_open_children_impl(vd, vdev_default_open_children_func);
	}

	/*
	* Conditionally open a subset of child vdevs.
	*/
	void
	vdev_open_children_subset(vdev_t vd, vdev_open_children_func_t open_func)
	{
	vdev_open_children_impl(vd, open_func);
	}

	/*
	* Compute the raidz-deflation ratio. Note, we hard-code
	* in 128k (1 << 17) because it is the "typical" blocksize.
	* Even though SPA_MAXBLOCKSIZE changed, this algorithm can not change,
	* otherwise it would inconsistently account for existing bp's.
	*/
	static void
	vdev_set_deflate_ratio(vdev_t *vd)
	{
	if (vd == vd->vdev_top && !vd->vdev_ishole && vd->vdev_ashift != 0) {
	vd->vdev_deflate_ratio = (1 << 17) /
	(vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
	}
	}

	/*
	* Choose the best of two ashifts, preferring one between logical ashift
	* (absolute minimum) and administrator defined maximum, otherwise take
	* the biggest of the two.
	*/
	uint64_t
	vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b)
	{
	if (a > logical && a <= zfs_vdev_max_auto_ashift) {
	if (b <= logical \|\| b > zfs_vdev_max_auto_ashift)
	return (a);
	else
	return (MAX(a, b));
	} else if (b <= logical \|\| b > zfs_vdev_max_auto_ashift)
	return (MAX(a, b));
	return (b);
	}

	/*
	* Maximize performance by inflating the configured ashift for top level
	* vdevs to be as close to the physical ashift as possible while maintaining
	* administrator defined limits and ensuring it doesn't go below the
	* logical ashift.
	*/
	static void
	vdev_ashift_optimize(vdev_t *vd)
	{
	ASSERT(vd == vd->vdev_top);

	if (vd->vdev_ashift < vd->vdev_physical_ashift &&
	vd->vdev_physical_ashift <= zfs_vdev_max_auto_ashift) {
	vd->vdev_ashift = MIN(
	MAX(zfs_vdev_max_auto_ashift, vd->vdev_ashift),
	MAX(zfs_vdev_min_auto_ashift,
	vd->vdev_physical_ashift));
	} else {
	/*
	* If the logical and physical ashifts are the same, then
	* we ensure that the top-level vdev's ashift is not smaller
	* than our minimum ashift value. For the unusual case
	* where logical ashift > physical ashift, we can't cap
	* the calculated ashift based on max ashift as that
	* would cause failures.
	* We still check if we need to increase it to match
	* the min ashift.
	*/
	vd->vdev_ashift = MAX(zfs_vdev_min_auto_ashift,
	vd->vdev_ashift);
	}
	}

	/*
	* Prepare a virtual device for access.
	*/
	int
	vdev_open(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	int error;
	uint64_t osize = 0;
	uint64_t max_osize = 0;
	uint64_t asize, max_asize, psize;
	uint64_t logical_ashift = 0;
	uint64_t physical_ashift = 0;

	ASSERT(vd->vdev_open_thread == curthread \|\|
	spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
	ASSERT(vd->vdev_state == VDEV_STATE_CLOSED \|\|
	vd->vdev_state == VDEV_STATE_CANT_OPEN \|\|
	vd->vdev_state == VDEV_STATE_OFFLINE);

	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	vd->vdev_min_asize = vdev_get_min_asize(vd);

	/*
	* If this vdev is not removed, check its fault status. If it's
	* faulted, bail out of the open.
	*/
	if (!vd->vdev_removed && vd->vdev_faulted) {
	ASSERT(vd->vdev_children == 0);
	ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED \|\|
	vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	vd->vdev_label_aux);
	return (SET_ERROR(ENXIO));
	} else if (vd->vdev_offline) {
	ASSERT(vd->vdev_children == 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE, VDEV_AUX_NONE);
	return (SET_ERROR(ENXIO));
	}

	error = vd->vdev_ops->vdev_op_open(vd, &osize, &max_osize,
	&logical_ashift, &physical_ashift);
	+
	+ /* Keep the device in removed state if unplugged */
	+ if (error == ENOENT && vd->vdev_removed) {
	+ vdev_set_state(vd, B_TRUE, VDEV_STATE_REMOVED,
	+ VDEV_AUX_NONE);
	+ return (error);
	+ }
	+
	/*
	* Physical volume size should never be larger than its max size, unless
	* the disk has shrunk while we were reading it or the device is buggy
	* or damaged: either way it's not safe for use, bail out of the open.
	*/
	if (osize > max_osize) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_OPEN_FAILED);
	return (SET_ERROR(ENXIO));
	}

	/*
	* Reset the vdev_reopening flag so that we actually close
	* the vdev on error.
	*/
	vd->vdev_reopening = B_FALSE;
	if (zio_injection_enabled && error == 0)
	error = zio_handle_device_injection(vd, NULL, SET_ERROR(ENXIO));

	if (error) {
	if (vd->vdev_removed &&
	vd->vdev_stat.vs_aux != VDEV_AUX_OPEN_FAILED)
	vd->vdev_removed = B_FALSE;

	if (vd->vdev_stat.vs_aux == VDEV_AUX_CHILDREN_OFFLINE) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE,
	vd->vdev_stat.vs_aux);
	} else {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	vd->vdev_stat.vs_aux);
	}
	return (error);
	}

	vd->vdev_removed = B_FALSE;

	/*
	* Recheck the faulted flag now that we have confirmed that
	* the vdev is accessible. If we're faulted, bail.
	*/
	if (vd->vdev_faulted) {
	ASSERT(vd->vdev_children == 0);
	ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED \|\|
	vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	vd->vdev_label_aux);
	return (SET_ERROR(ENXIO));
	}

	if (vd->vdev_degraded) {
	ASSERT(vd->vdev_children == 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
	VDEV_AUX_ERR_EXCEEDED);
	} else {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_HEALTHY, 0);
	}

	/*
	* For hole or missing vdevs we just return success.
	*/
	if (vd->vdev_ishole \|\| vd->vdev_ops == &vdev_missing_ops)
	return (0);

	for (int c = 0; c < vd->vdev_children; c++) {
	if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
	VDEV_AUX_NONE);
	break;
	}
	}

	osize = P2ALIGN(osize, (uint64_t)sizeof (vdev_label_t));
	max_osize = P2ALIGN(max_osize, (uint64_t)sizeof (vdev_label_t));

	if (vd->vdev_children == 0) {
	if (osize < SPA_MINDEVSIZE) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_TOO_SMALL);
	return (SET_ERROR(EOVERFLOW));
	}
	psize = osize;
	asize = osize - (VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE);
	max_asize = max_osize - (VDEV_LABEL_START_SIZE +
	VDEV_LABEL_END_SIZE);
	} else {
	if (vd->vdev_parent != NULL && osize < SPA_MINDEVSIZE -
	(VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE)) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_TOO_SMALL);
	return (SET_ERROR(EOVERFLOW));
	}
	psize = 0;
	asize = osize;
	max_asize = max_osize;
	}

	/*
	* If the vdev was expanded, record this so that we can re-create the
	* uberblock rings in labels {2,3}, during the next sync.
	*/
	if ((psize > vd->vdev_psize) && (vd->vdev_psize != 0))
	vd->vdev_copy_uberblocks = B_TRUE;

	vd->vdev_psize = psize;

	/*
	* Make sure the allocatable size hasn't shrunk too much.
	*/
	if (asize < vd->vdev_min_asize) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	return (SET_ERROR(EINVAL));
	}

	/*
	* We can always set the logical/physical ashift members since
	* their values are only used to calculate the vdev_ashift when
	* the device is first added to the config. These values should
	* not be used for anything else since they may change whenever
	* the device is reopened and we don't store them in the label.
	*/
	vd->vdev_physical_ashift =
	MAX(physical_ashift, vd->vdev_physical_ashift);
	vd->vdev_logical_ashift = MAX(logical_ashift,
	vd->vdev_logical_ashift);

	if (vd->vdev_asize == 0) {
	/*
	* This is the first-ever open, so use the computed values.
	* For compatibility, a different ashift can be requested.
	*/
	vd->vdev_asize = asize;
	vd->vdev_max_asize = max_asize;

	/*
	* If the vdev_ashift was not overridden at creation time,
	* then set it the logical ashift and optimize the ashift.
	*/
	if (vd->vdev_ashift == 0) {
	vd->vdev_ashift = vd->vdev_logical_ashift;

	if (vd->vdev_logical_ashift > ASHIFT_MAX) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_ASHIFT_TOO_BIG);
	return (SET_ERROR(EDOM));
	}

	if (vd->vdev_top == vd) {
	vdev_ashift_optimize(vd);
	}
	}
	if (vd->vdev_ashift != 0 && (vd->vdev_ashift < ASHIFT_MIN \|\|
	vd->vdev_ashift > ASHIFT_MAX)) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_ASHIFT);
	return (SET_ERROR(EDOM));
	}
	} else {
	/*
	* Make sure the alignment required hasn't increased.
	*/
	if (vd->vdev_ashift > vd->vdev_top->vdev_ashift &&
	vd->vdev_ops->vdev_op_leaf) {
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT,
	spa, vd, NULL, NULL, 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	return (SET_ERROR(EDOM));
	}
	vd->vdev_max_asize = max_asize;
	}

	/*
	* If all children are healthy we update asize if either:
	* The asize has increased, due to a device expansion caused by dynamic
	* LUN growth or vdev replacement, and automatic expansion is enabled;
	* making the additional space available.
	*
	* The asize has decreased, due to a device shrink usually caused by a
	* vdev replace with a smaller device. This ensures that calculations
	* based of max_asize and asize e.g. esize are always valid. It's safe
	* to do this as we've already validated that asize is greater than
	* vdev_min_asize.
	*/
	if (vd->vdev_state == VDEV_STATE_HEALTHY &&
	((asize > vd->vdev_asize &&
	(vd->vdev_expanding \|\| spa->spa_autoexpand)) \|\|
	(asize < vd->vdev_asize)))
	vd->vdev_asize = asize;

	vdev_set_min_asize(vd);

	/*
	* Ensure we can issue some IO before declaring the
	* vdev open for business.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(error = zio_wait(vdev_probe(vd, NULL))) != 0) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	VDEV_AUX_ERR_EXCEEDED);
	return (error);
	}

	/*
	* Track the minimum allocation size.
	*/
	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
	vd->vdev_islog == 0 && vd->vdev_aux == NULL) {
	uint64_t min_alloc = vdev_get_min_alloc(vd);
	if (min_alloc < spa->spa_min_alloc)
	spa->spa_min_alloc = min_alloc;
	}

	/*
	* If this is a leaf vdev, assess whether a resilver is needed.
	* But don't do this if we are doing a reopen for a scrub, since
	* this would just restart the scrub we are already doing.
	*/
	if (vd->vdev_ops->vdev_op_leaf && !spa->spa_scrub_reopen)
	dsl_scan_assess_vdev(spa->spa_dsl_pool, vd);

	return (0);
	}

	static void
	vdev_validate_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_validate_thread = curthread;
	vd->vdev_validate_error = vdev_validate(vd);
	vd->vdev_validate_thread = NULL;
	}

	/*
	* Called once the vdevs are all opened, this routine validates the label
	* contents. This needs to be done before vdev_load() so that we don't
	* inadvertently do repair I/Os to the wrong device.
	*
	* This function will only return failure if one of the vdevs indicates that it
	* has since been destroyed or exported. This is only possible if
	* /etc/zfs/zpool.cache was readonly at the time. Otherwise, the vdev state
	* will be updated but the function will return 0.
	*/
	int
	vdev_validate(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	taskq_t *tq = NULL;
	nvlist_t *label;
	uint64_t guid = 0, aux_guid = 0, top_guid;
	uint64_t state;
	nvlist_t *nvl;
	uint64_t txg;
	int children = vd->vdev_children;

	if (vdev_validate_skip)
	return (0);

	if (children > 0) {
	tq = taskq_create("vdev_validate", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	}

	for (uint64_t c = 0; c < children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (tq == NULL \|\| vdev_uses_zvols(cvd)) {
	vdev_validate_child(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_validate_child, cvd,
	TQ_SLEEP) != TASKQID_INVALID);
	}
	}
	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}
	for (int c = 0; c < children; c++) {
	int error = vd->vdev_child[c]->vdev_validate_error;

	if (error != 0)
	return (SET_ERROR(EBADF));
	}


	/*
	* If the device has already failed, or was marked offline, don't do
	* any further validation. Otherwise, label I/O will fail and we will
	* overwrite the previous state.
	*/
	if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_readable(vd))
	return (0);

	/*
	* If we are performing an extreme rewind, we allow for a label that
	* was modified at a point after the current txg.
	* If config lock is not held do not check for the txg. spa_sync could
	* be updating the vdev's label before updating spa_last_synced_txg.
	*/
	if (spa->spa_extreme_rewind \|\| spa_last_synced_txg(spa) == 0 \|\|
	spa_config_held(spa, SCL_CONFIG, RW_WRITER) != SCL_CONFIG)
	txg = UINT64_MAX;
	else
	txg = spa_last_synced_txg(spa);

	if ((label = vdev_label_read_config(vd, txg)) == NULL) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	vdev_dbgmsg(vd, "vdev_validate: failed reading config for "
	"txg %llu", (u_longlong_t)txg);
	return (0);
	}

	/*
	* Determine if this vdev has been split off into another
	* pool. If so, then refuse to open it.
	*/
	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_SPLIT_GUID,
	&aux_guid) == 0 && aux_guid == spa_guid(spa)) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_SPLIT_POOL);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: vdev split into other pool");
	return (0);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &guid) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_POOL_GUID);
	return (0);
	}

	/*
	* If config is not trusted then ignore the spa guid check. This is
	* necessary because if the machine crashed during a re-guid the new
	* guid might have been written to all of the vdev labels, but not the
	* cached config. The check will be performed again once we have the
	* trusted config from the MOS.
	*/
	if (spa->spa_trust_config && guid != spa_guid(spa)) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: vdev label pool_guid doesn't "
	"match config (%llu != %llu)", (u_longlong_t)guid,
	(u_longlong_t)spa_guid(spa));
	return (0);
	}

	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvl)
	!= 0 \|\| nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_ORIG_GUID,
	&aux_guid) != 0)
	aux_guid = 0;

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_GUID);
	return (0);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_TOP_GUID, &top_guid)
	!= 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_TOP_GUID);
	return (0);
	}

	/*
	* If this vdev just became a top-level vdev because its sibling was
	* detached, it will have adopted the parent's vdev guid -- but the
	* label may or may not be on disk yet. Fortunately, either version
	* of the label will have the same top guid, so if we're a top-level
	* vdev, we can safely compare to that instead.
	* However, if the config comes from a cachefile that failed to update
	* after the detach, a top-level vdev will appear as a non top-level
	* vdev in the config. Also relax the constraints if we perform an
	* extreme rewind.
	*
	* If we split this vdev off instead, then we also check the
	* original pool's guid. We don't want to consider the vdev
	* corrupt if it is partway through a split operation.
	*/
	if (vd->vdev_guid != guid && vd->vdev_guid != aux_guid) {
	boolean_t mismatch = B_FALSE;
	if (spa->spa_trust_config && !spa->spa_extreme_rewind) {
	if (vd != vd->vdev_top \|\| vd->vdev_guid != top_guid)
	mismatch = B_TRUE;
	} else {
	if (vd->vdev_guid != top_guid &&
	vd->vdev_top->vdev_guid != guid)
	mismatch = B_TRUE;
	}

	if (mismatch) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: config guid "
	"doesn't match label guid");
	vdev_dbgmsg(vd, "CONFIG: guid %llu, top_guid %llu",
	(u_longlong_t)vd->vdev_guid,
	(u_longlong_t)vd->vdev_top->vdev_guid);
	vdev_dbgmsg(vd, "LABEL: guid %llu, top_guid %llu, "
	"aux_guid %llu", (u_longlong_t)guid,
	(u_longlong_t)top_guid, (u_longlong_t)aux_guid);
	return (0);
	}
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
	&state) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_POOL_STATE);
	return (0);
	}

	nvlist_free(label);

	/*
	* If this is a verbatim import, no need to check the
	* state of the pool.
	*/
	if (!(spa->spa_import_flags & ZFS_IMPORT_VERBATIM) &&
	spa_load_state(spa) == SPA_LOAD_OPEN &&
	state != POOL_STATE_ACTIVE) {
	vdev_dbgmsg(vd, "vdev_validate: invalid pool state (%llu) "
	"for spa %s", (u_longlong_t)state, spa->spa_name);
	return (SET_ERROR(EBADF));
	}

	/*
	* If we were able to open and validate a vdev that was
	* previously marked permanently unavailable, clear that state
	* now.
	*/
	if (vd->vdev_not_present)
	vd->vdev_not_present = 0;

	return (0);
	}

	static void
	vdev_copy_path_impl(vdev_t svd, vdev_t dvd)
	{
	char old, new;
	if (svd->vdev_path != NULL && dvd->vdev_path != NULL) {
	if (strcmp(svd->vdev_path, dvd->vdev_path) != 0) {
	zfs_dbgmsg("vdev_copy_path: vdev %llu: path changed "
	"from '%s' to '%s'", (u_longlong_t)dvd->vdev_guid,
	dvd->vdev_path, svd->vdev_path);
	spa_strfree(dvd->vdev_path);
	dvd->vdev_path = spa_strdup(svd->vdev_path);
	}
	} else if (svd->vdev_path != NULL) {
	dvd->vdev_path = spa_strdup(svd->vdev_path);
	zfs_dbgmsg("vdev_copy_path: vdev %llu: path set to '%s'",
	(u_longlong_t)dvd->vdev_guid, dvd->vdev_path);
	}

	/*
	* Our enclosure sysfs path may have changed between imports
	*/
	old = dvd->vdev_enc_sysfs_path;
	new = svd->vdev_enc_sysfs_path;
	if ((old != NULL && new == NULL) \|\|
	(old == NULL && new != NULL) \|\|
	((old != NULL && new != NULL) && strcmp(new, old) != 0)) {
	zfs_dbgmsg("vdev_copy_path: vdev %llu: vdev_enc_sysfs_path "
	"changed from '%s' to '%s'", (u_longlong_t)dvd->vdev_guid,
	old, new);

	if (dvd->vdev_enc_sysfs_path)
	spa_strfree(dvd->vdev_enc_sysfs_path);

	if (svd->vdev_enc_sysfs_path) {
	dvd->vdev_enc_sysfs_path = spa_strdup(
	svd->vdev_enc_sysfs_path);
	} else {
	dvd->vdev_enc_sysfs_path = NULL;
	}
	}
	}

	/*
	* Recursively copy vdev paths from one vdev to another. Source and destination
	* vdev trees must have same geometry otherwise return error. Intended to copy
	* paths from userland config into MOS config.
	*/
	int
	vdev_copy_path_strict(vdev_t svd, vdev_t dvd)
	{
	if ((svd->vdev_ops == &vdev_missing_ops) \|\|
	(svd->vdev_ishole && dvd->vdev_ishole) \|\|
	(dvd->vdev_ops == &vdev_indirect_ops))
	return (0);

	if (svd->vdev_ops != dvd->vdev_ops) {
	vdev_dbgmsg(svd, "vdev_copy_path: vdev type mismatch: %s != %s",
	svd->vdev_ops->vdev_op_type, dvd->vdev_ops->vdev_op_type);
	return (SET_ERROR(EINVAL));
	}

	if (svd->vdev_guid != dvd->vdev_guid) {
	vdev_dbgmsg(svd, "vdev_copy_path: guids mismatch (%llu != "
	"%llu)", (u_longlong_t)svd->vdev_guid,
	(u_longlong_t)dvd->vdev_guid);
	return (SET_ERROR(EINVAL));
	}

	if (svd->vdev_children != dvd->vdev_children) {
	vdev_dbgmsg(svd, "vdev_copy_path: children count mismatch: "
	"%llu != %llu", (u_longlong_t)svd->vdev_children,
	(u_longlong_t)dvd->vdev_children);
	return (SET_ERROR(EINVAL));
	}

	for (uint64_t i = 0; i < svd->vdev_children; i++) {
	int error = vdev_copy_path_strict(svd->vdev_child[i],
	dvd->vdev_child[i]);
	if (error != 0)
	return (error);
	}

	if (svd->vdev_ops->vdev_op_leaf)
	vdev_copy_path_impl(svd, dvd);

	return (0);
	}

	static void
	vdev_copy_path_search(vdev_t stvd, vdev_t dvd)
	{
	ASSERT(stvd->vdev_top == stvd);
	ASSERT3U(stvd->vdev_id, ==, dvd->vdev_top->vdev_id);

	for (uint64_t i = 0; i < dvd->vdev_children; i++) {
	vdev_copy_path_search(stvd, dvd->vdev_child[i]);
	}

	if (!dvd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(dvd))
	return;

	/*
	* The idea here is that while a vdev can shift positions within
	* a top vdev (when replacing, attaching mirror, etc.) it cannot
	* step outside of it.
	*/
	vdev_t *vd = vdev_lookup_by_guid(stvd, dvd->vdev_guid);

	if (vd == NULL \|\| vd->vdev_ops != dvd->vdev_ops)
	return;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	vdev_copy_path_impl(vd, dvd);
	}

	/*
	* Recursively copy vdev paths from one root vdev to another. Source and
	* destination vdev trees may differ in geometry. For each destination leaf
	* vdev, search a vdev with the same guid and top vdev id in the source.
	* Intended to copy paths from userland config into MOS config.
	*/
	void
	vdev_copy_path_relaxed(vdev_t srvd, vdev_t drvd)
	{
	uint64_t children = MIN(srvd->vdev_children, drvd->vdev_children);
	ASSERT(srvd->vdev_ops == &vdev_root_ops);
	ASSERT(drvd->vdev_ops == &vdev_root_ops);

	for (uint64_t i = 0; i < children; i++) {
	vdev_copy_path_search(srvd->vdev_child[i],
	drvd->vdev_child[i]);
	}
	}

	/*
	* Close a virtual device.
	*/
	void
	vdev_close(vdev_t *vd)
	{
	vdev_t *pvd = vd->vdev_parent;
	spa_t *spa __maybe_unused = vd->vdev_spa;

	ASSERT(vd != NULL);
	ASSERT(vd->vdev_open_thread == curthread \|\|
	spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	/*
	* If our parent is reopening, then we are as well, unless we are
	* going offline.
	*/
	if (pvd != NULL && pvd->vdev_reopening)
	vd->vdev_reopening = (pvd->vdev_reopening && !vd->vdev_offline);

	vd->vdev_ops->vdev_op_close(vd);

	vdev_cache_purge(vd);

	/*
	* We record the previous state before we close it, so that if we are
	* doing a reopen(), we don't generate FMA ereports if we notice that
	* it's still faulted.
	*/
	vd->vdev_prevstate = vd->vdev_state;

	if (vd->vdev_offline)
	vd->vdev_state = VDEV_STATE_OFFLINE;
	else
	vd->vdev_state = VDEV_STATE_CLOSED;
	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	}

	void
	vdev_hold(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_is_root(spa));
	if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	return;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_hold(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_hold != NULL)
	vd->vdev_ops->vdev_op_hold(vd);
	}

	void
	vdev_rele(vdev_t *vd)
	{
	ASSERT(spa_is_root(vd->vdev_spa));
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_rele(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_rele != NULL)
	vd->vdev_ops->vdev_op_rele(vd);
	}

	/*
	* Reopen all interior vdevs and any unopened leaves. We don't actually
	* reopen leaf vdevs which had previously been opened as they might deadlock
	* on the spa_config_lock. Instead we only obtain the leaf's physical size.
	* If the leaf has never been opened then open it, as usual.
	*/
	void
	vdev_reopen(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	/* set the reopening flag unless we're taking the vdev offline */
	vd->vdev_reopening = !vd->vdev_offline;
	vdev_close(vd);
	(void) vdev_open(vd);

	/*
	* Call vdev_validate() here to make sure we have the same device.
	* Otherwise, a device with an invalid label could be successfully
	* opened in response to vdev_reopen().
	*/
	if (vd->vdev_aux) {
	(void) vdev_validate_aux(vd);
	if (vdev_readable(vd) && vdev_writeable(vd) &&
	vd->vdev_aux == &spa->spa_l2cache) {
	/*
	* In case the vdev is present we should evict all ARC
	* buffers and pointers to log blocks and reclaim their
	* space before restoring its contents to L2ARC.
	*/
	if (l2arc_vdev_present(vd)) {
	l2arc_rebuild_vdev(vd, B_TRUE);
	} else {
	l2arc_add_vdev(spa, vd);
	}
	spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);
	spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
	}
	} else {
	(void) vdev_validate(vd);
	}

	/*
	* Reassess parent vdev's health.
	*/
	vdev_propagate_state(vd);
	}

	int
	vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
	{
	int error;

	/*
	* Normally, partial opens (e.g. of a mirror) are allowed.
	* For a create, however, we want to fail the request if
	* there are any components we can't open.
	*/
	error = vdev_open(vd);

	if (error \|\| vd->vdev_state != VDEV_STATE_HEALTHY) {
	vdev_close(vd);
	return (error ? error : SET_ERROR(ENXIO));
	}

	/*
	* Recursively load DTLs and initialize all labels.
	*/
	if ((error = vdev_dtl_load(vd)) != 0 \|\|
	(error = vdev_label_init(vd, txg, isreplacing ?
	VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
	vdev_close(vd);
	return (error);
	}

	return (0);
	}

	void
	vdev_metaslab_set_size(vdev_t *vd)
	{
	uint64_t asize = vd->vdev_asize;
	uint64_t ms_count = asize >> zfs_vdev_default_ms_shift;
	uint64_t ms_shift;

	/*
	* There are two dimensions to the metaslab sizing calculation:
	* the size of the metaslab and the count of metaslabs per vdev.
	*
	* The default values used below are a good balance between memory
	* usage (larger metaslab size means more memory needed for loaded
	* metaslabs; more metaslabs means more memory needed for the
	* metaslab_t structs), metaslab load time (larger metaslabs take
	* longer to load), and metaslab sync time (more metaslabs means
	* more time spent syncing all of them).
	*
	* In general, we aim for zfs_vdev_default_ms_count (200) metaslabs.
	* The range of the dimensions are as follows:
	*
	* 2^29 <= ms_size <= 2^34
	* 16 <= ms_count <= 131,072
	*
	* On the lower end of vdev sizes, we aim for metaslabs sizes of
	* at least 512MB (2^29) to minimize fragmentation effects when
	* testing with smaller devices. However, the count constraint
	* of at least 16 metaslabs will override this minimum size goal.
	*
	* On the upper end of vdev sizes, we aim for a maximum metaslab
	* size of 16GB. However, we will cap the total count to 2^17
	* metaslabs to keep our memory footprint in check and let the
	* metaslab size grow from there if that limit is hit.
	*
	* The net effect of applying above constrains is summarized below.
	*
	* vdev size metaslab count
	* --------------\|-----------------
	* < 8GB ~16
	* 8GB - 100GB one per 512MB
	* 100GB - 3TB ~200
	* 3TB - 2PB one per 16GB
	* > 2PB ~131,072
	* --------------------------------
	*
	* Finally, note that all of the above calculate the initial
	* number of metaslabs. Expanding a top-level vdev will result
	* in additional metaslabs being allocated making it possible
	* to exceed the zfs_vdev_ms_count_limit.
	*/

	if (ms_count < zfs_vdev_min_ms_count)
	ms_shift = highbit64(asize / zfs_vdev_min_ms_count);
	else if (ms_count > zfs_vdev_default_ms_count)
	ms_shift = highbit64(asize / zfs_vdev_default_ms_count);
	else
	ms_shift = zfs_vdev_default_ms_shift;

	if (ms_shift < SPA_MAXBLOCKSHIFT) {
	ms_shift = SPA_MAXBLOCKSHIFT;
	} else if (ms_shift > zfs_vdev_max_ms_shift) {
	ms_shift = zfs_vdev_max_ms_shift;
	/* cap the total count to constrain memory footprint */
	if ((asize >> ms_shift) > zfs_vdev_ms_count_limit)
	ms_shift = highbit64(asize / zfs_vdev_ms_count_limit);
	}

	vd->vdev_ms_shift = ms_shift;
	ASSERT3U(vd->vdev_ms_shift, >=, SPA_MAXBLOCKSHIFT);
	}

	void
	vdev_dirty(vdev_t vd, int flags, void arg, uint64_t txg)
	{
	ASSERT(vd == vd->vdev_top);
	/* indirect vdevs don't have metaslabs or dtls */
	ASSERT(vdev_is_concrete(vd) \|\| flags == 0);
	ASSERT(ISP2(flags));
	ASSERT(spa_writeable(vd->vdev_spa));

	if (flags & VDD_METASLAB)
	(void) txg_list_add(&vd->vdev_ms_list, arg, txg);

	if (flags & VDD_DTL)
	(void) txg_list_add(&vd->vdev_dtl_list, arg, txg);

	(void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
	}

	void
	vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg)
	{
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_dirty_leaves(vd->vdev_child[c], flags, txg);

	if (vd->vdev_ops->vdev_op_leaf)
	vdev_dirty(vd->vdev_top, flags, vd, txg);
	}

	/*
	* DTLs.
	*
	* A vdev's DTL (dirty time log) is the set of transaction groups for which
	* the vdev has less than perfect replication. There are four kinds of DTL:
	*
	* DTL_MISSING: txgs for which the vdev has no valid copies of the data
	*
	* DTL_PARTIAL: txgs for which data is available, but not fully replicated
	*
	* DTL_SCRUB: the txgs that could not be repaired by the last scrub; upon
	* scrub completion, DTL_SCRUB replaces DTL_MISSING in the range of
	* txgs that was scrubbed.
	*
	* DTL_OUTAGE: txgs which cannot currently be read, whether due to
	* persistent errors or just some device being offline.
	* Unlike the other three, the DTL_OUTAGE map is not generally
	* maintained; it's only computed when needed, typically to
	* determine whether a device can be detached.
	*
	* For leaf vdevs, DTL_MISSING and DTL_PARTIAL are identical: the device
	* either has the data or it doesn't.
	*
	* For interior vdevs such as mirror and RAID-Z the picture is more complex.
	* A vdev's DTL_PARTIAL is the union of its children's DTL_PARTIALs, because
	* if any child is less than fully replicated, then so is its parent.
	* A vdev's DTL_MISSING is a modified union of its children's DTL_MISSINGs,
	* comprising only those txgs which appear in 'maxfaults' or more children;
	* those are the txgs we don't have enough replication to read. For example,
	* double-parity RAID-Z can tolerate up to two missing devices (maxfaults == 2);
	* thus, its DTL_MISSING consists of the set of txgs that appear in more than
	* two child DTL_MISSING maps.
	*
	* It should be clear from the above that to compute the DTLs and outage maps
	* for all vdevs, it suffices to know just the leaf vdevs' DTL_MISSING maps.
	* Therefore, that is all we keep on disk. When loading the pool, or after
	* a configuration change, we generate all other DTLs from first principles.
	*/
	void
	vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
	{
	range_tree_t *rt = vd->vdev_dtl[t];

	ASSERT(t < DTL_TYPES);
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);
	ASSERT(spa_writeable(vd->vdev_spa));

	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_contains(rt, txg, size))
	range_tree_add(rt, txg, size);
	mutex_exit(&vd->vdev_dtl_lock);
	}

	boolean_t
	vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
	{
	range_tree_t *rt = vd->vdev_dtl[t];
	boolean_t dirty = B_FALSE;

	ASSERT(t < DTL_TYPES);
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);

	/*
	* While we are loading the pool, the DTLs have not been loaded yet.
	* This isn't a problem but it can result in devices being tried
	* which are known to not have the data. In which case, the import
	* is relying on the checksum to ensure that we get the right data.
	* Note that while importing we are only reading the MOS, which is
	* always checksummed.
	*/
	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_is_empty(rt))
	dirty = range_tree_contains(rt, txg, size);
	mutex_exit(&vd->vdev_dtl_lock);

	return (dirty);
	}

	boolean_t
	vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
	{
	range_tree_t *rt = vd->vdev_dtl[t];
	boolean_t empty;

	mutex_enter(&vd->vdev_dtl_lock);
	empty = range_tree_is_empty(rt);
	mutex_exit(&vd->vdev_dtl_lock);

	return (empty);
	}

	/*
	* Check if the txg falls within the range which must be
	* resilvered. DVAs outside this range can always be skipped.
	*/
	boolean_t
	vdev_default_need_resilver(vdev_t vd, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	(void) dva, (void) psize;

	/* Set by sequential resilver. */
	if (phys_birth == TXG_UNKNOWN)
	return (B_TRUE);

	return (vdev_dtl_contains(vd, DTL_PARTIAL, phys_birth, 1));
	}

	/*
	* Returns B_TRUE if the vdev determines the DVA needs to be resilvered.
	*/
	boolean_t
	vdev_dtl_need_resilver(vdev_t vd, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);

	if (vd->vdev_ops->vdev_op_need_resilver == NULL \|\|
	vd->vdev_ops->vdev_op_leaf)
	return (B_TRUE);

	return (vd->vdev_ops->vdev_op_need_resilver(vd, dva, psize,
	phys_birth));
	}

	/*
	* Returns the lowest txg in the DTL range.
	*/
	static uint64_t
	vdev_dtl_min(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
	ASSERT0(vd->vdev_children);

	return (range_tree_min(vd->vdev_dtl[DTL_MISSING]) - 1);
	}

	/*
	* Returns the highest txg in the DTL.
	*/
	static uint64_t
	vdev_dtl_max(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
	ASSERT0(vd->vdev_children);

	return (range_tree_max(vd->vdev_dtl[DTL_MISSING]));
	}

	/*
	* Determine if a resilvering vdev should remove any DTL entries from
	* its range. If the vdev was resilvering for the entire duration of the
	* scan then it should excise that range from its DTLs. Otherwise, this
	* vdev is considered partially resilvered and should leave its DTL
	* entries intact. The comment in vdev_dtl_reassess() describes how we
	* excise the DTLs.
	*/
	static boolean_t
	vdev_dtl_should_excise(vdev_t *vd, boolean_t rebuild_done)
	{
	ASSERT0(vd->vdev_children);

	if (vd->vdev_state < VDEV_STATE_DEGRADED)
	return (B_FALSE);

	if (vd->vdev_resilver_deferred)
	return (B_FALSE);

	if (range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]))
	return (B_TRUE);

	if (rebuild_done) {
	vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	/* Rebuild not initiated by attach */
	if (vd->vdev_rebuild_txg == 0)
	return (B_TRUE);

	/*
	* When a rebuild completes without error then all missing data
	* up to the rebuild max txg has been reconstructed and the DTL
	* is eligible for excision.
	*/
	if (vrp->vrp_rebuild_state == VDEV_REBUILD_COMPLETE &&
	vdev_dtl_max(vd) <= vrp->vrp_max_txg) {
	ASSERT3U(vrp->vrp_min_txg, <=, vdev_dtl_min(vd));
	ASSERT3U(vrp->vrp_min_txg, <, vd->vdev_rebuild_txg);
	ASSERT3U(vd->vdev_rebuild_txg, <=, vrp->vrp_max_txg);
	return (B_TRUE);
	}
	} else {
	dsl_scan_t *scn = vd->vdev_spa->spa_dsl_pool->dp_scan;
	dsl_scan_phys_t *scnp __maybe_unused = &scn->scn_phys;

	/* Resilver not initiated by attach */
	if (vd->vdev_resilver_txg == 0)
	return (B_TRUE);

	/*
	* When a resilver is initiated the scan will assign the
	* scn_max_txg value to the highest txg value that exists
	* in all DTLs. If this device's max DTL is not part of this
	* scan (i.e. it is not in the range (scn_min_txg, scn_max_txg]
	* then it is not eligible for excision.
	*/
	if (vdev_dtl_max(vd) <= scn->scn_phys.scn_max_txg) {
	ASSERT3U(scnp->scn_min_txg, <=, vdev_dtl_min(vd));
	ASSERT3U(scnp->scn_min_txg, <, vd->vdev_resilver_txg);
	ASSERT3U(vd->vdev_resilver_txg, <=, scnp->scn_max_txg);
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}

	/*
	* Reassess DTLs after a config change or scrub completion. If txg == 0 no
	* write operations will be issued to the pool.
	*/
	void
	vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
	boolean_t scrub_done, boolean_t rebuild_done)
	{
	spa_t *spa = vd->vdev_spa;
	avl_tree_t reftree;
	int minref;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_dtl_reassess(vd->vdev_child[c], txg,
	scrub_txg, scrub_done, rebuild_done);

	if (vd == spa->spa_root_vdev \|\| !vdev_is_concrete(vd) \|\| vd->vdev_aux)
	return;

	if (vd->vdev_ops->vdev_op_leaf) {
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
	vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
	boolean_t check_excise = B_FALSE;
	boolean_t wasempty = B_TRUE;

	mutex_enter(&vd->vdev_dtl_lock);

	/*
	* If requested, pretend the scan or rebuild completed cleanly.
	*/
	if (zfs_scan_ignore_errors) {
	if (scn != NULL)
	scn->scn_phys.scn_errors = 0;
	if (vr != NULL)
	vr->vr_rebuild_phys.vrp_errors = 0;
	}

	if (scrub_txg != 0 &&
	!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
	wasempty = B_FALSE;
	zfs_dbgmsg("guid:%llu txg:%llu scrub:%llu started:%d "
	"dtl:%llu/%llu errors:%llu",
	(u_longlong_t)vd->vdev_guid, (u_longlong_t)txg,
	(u_longlong_t)scrub_txg, spa->spa_scrub_started,
	(u_longlong_t)vdev_dtl_min(vd),
	(u_longlong_t)vdev_dtl_max(vd),
	(u_longlong_t)(scn ? scn->scn_phys.scn_errors : 0));
	}

	/*
	* If we've completed a scrub/resilver or a rebuild cleanly
	* then determine if this vdev should remove any DTLs. We
	* only want to excise regions on vdevs that were available
	* during the entire duration of this scan.
	*/
	if (rebuild_done &&
	vr != NULL && vr->vr_rebuild_phys.vrp_errors == 0) {
	check_excise = B_TRUE;
	} else {
	if (spa->spa_scrub_started \|\|
	(scn != NULL && scn->scn_phys.scn_errors == 0)) {
	check_excise = B_TRUE;
	}
	}

	if (scrub_txg && check_excise &&
	vdev_dtl_should_excise(vd, rebuild_done)) {
	/*
	* We completed a scrub, resilver or rebuild up to
	* scrub_txg. If we did it without rebooting, then
	* the scrub dtl will be valid, so excise the old
	* region and fold in the scrub dtl. Otherwise,
	* leave the dtl as-is if there was an error.
	*
	* There's little trick here: to excise the beginning
	* of the DTL_MISSING map, we put it into a reference
	* tree and then add a segment with refcnt -1 that
	* covers the range [0, scrub_txg). This means
	* that each txg in that range has refcnt -1 or 0.
	* We then add DTL_SCRUB with a refcnt of 2, so that
	* entries in the range [0, scrub_txg) will have a
	* positive refcnt -- either 1 or 2. We then convert
	* the reference tree into the new DTL_MISSING map.
	*/
	space_reftree_create(&reftree);
	space_reftree_add_map(&reftree,
	vd->vdev_dtl[DTL_MISSING], 1);
	space_reftree_add_seg(&reftree, 0, scrub_txg, -1);
	space_reftree_add_map(&reftree,
	vd->vdev_dtl[DTL_SCRUB], 2);
	space_reftree_generate_map(&reftree,
	vd->vdev_dtl[DTL_MISSING], 1);
	space_reftree_destroy(&reftree);

	if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
	zfs_dbgmsg("update DTL_MISSING:%llu/%llu",
	(u_longlong_t)vdev_dtl_min(vd),
	(u_longlong_t)vdev_dtl_max(vd));
	} else if (!wasempty) {
	zfs_dbgmsg("DTL_MISSING is now empty");
	}
	}
	range_tree_vacate(vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
	range_tree_walk(vd->vdev_dtl[DTL_MISSING],
	range_tree_add, vd->vdev_dtl[DTL_PARTIAL]);
	if (scrub_done)
	range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
	range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
	if (!vdev_readable(vd))
	range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
	else
	range_tree_walk(vd->vdev_dtl[DTL_MISSING],
	range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);

	/*
	* If the vdev was resilvering or rebuilding and no longer
	* has any DTLs then reset the appropriate flag and dirty
	* the top level so that we persist the change.
	*/
	if (txg != 0 &&
	range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
	range_tree_is_empty(vd->vdev_dtl[DTL_OUTAGE])) {
	if (vd->vdev_rebuild_txg != 0) {
	vd->vdev_rebuild_txg = 0;
	vdev_config_dirty(vd->vdev_top);
	} else if (vd->vdev_resilver_txg != 0) {
	vd->vdev_resilver_txg = 0;
	vdev_config_dirty(vd->vdev_top);
	}
	}

	mutex_exit(&vd->vdev_dtl_lock);

	if (txg != 0)
	vdev_dirty(vd->vdev_top, VDD_DTL, vd, txg);
	return;
	}

	mutex_enter(&vd->vdev_dtl_lock);
	for (int t = 0; t < DTL_TYPES; t++) {
	/* account for child's outage in parent's missing map */
	int s = (t == DTL_MISSING) ? DTL_OUTAGE: t;
	if (t == DTL_SCRUB)
	continue; /* leaf vdevs only */
	if (t == DTL_PARTIAL)
	minref = 1; /* i.e. non-zero */
	else if (vdev_get_nparity(vd) != 0)
	minref = vdev_get_nparity(vd) + 1; /* RAID-Z, dRAID */
	else
	minref = vd->vdev_children; /* any kind of mirror */
	space_reftree_create(&reftree);
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	mutex_enter(&cvd->vdev_dtl_lock);
	space_reftree_add_map(&reftree, cvd->vdev_dtl[s], 1);
	mutex_exit(&cvd->vdev_dtl_lock);
	}
	space_reftree_generate_map(&reftree, vd->vdev_dtl[t], minref);
	space_reftree_destroy(&reftree);
	}
	mutex_exit(&vd->vdev_dtl_lock);
	}

	+/*
	+ * Iterate over all the vdevs except spare, and post kobj events
	+ */
	+void
	+vdev_post_kobj_evt(vdev_t *vd)
	+{
	+ if (vd->vdev_ops->vdev_op_kobj_evt_post &&
	+ vd->vdev_kobj_flag == B_FALSE) {
	+ vd->vdev_kobj_flag = B_TRUE;
	+ vd->vdev_ops->vdev_op_kobj_evt_post(vd);
	+ }
	+
	+ for (int c = 0; c < vd->vdev_children; c++)
	+ vdev_post_kobj_evt(vd->vdev_child[c]);
	+}
	+
	+/*
	+ * Iterate over all the vdevs except spare, and clear kobj events
	+ */
	+void
	+vdev_clear_kobj_evt(vdev_t *vd)
	+{
	+ vd->vdev_kobj_flag = B_FALSE;
	+
	+ for (int c = 0; c < vd->vdev_children; c++)
	+ vdev_clear_kobj_evt(vd->vdev_child[c]);
	+}
	+
	int
	vdev_dtl_load(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	range_tree_t *rt;
	int error = 0;

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
	ASSERT(vdev_is_concrete(vd));

	/*
	* If the dtl cannot be sync'd there is no need to open it.
	*/
	if (spa->spa_mode == SPA_MODE_READ && !spa->spa_read_spacemaps)
	return (0);

	error = space_map_open(&vd->vdev_dtl_sm, mos,
	vd->vdev_dtl_object, 0, -1ULL, 0);
	if (error)
	return (error);
	ASSERT(vd->vdev_dtl_sm != NULL);

	rt = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	error = space_map_load(vd->vdev_dtl_sm, rt, SM_ALLOC);
	if (error == 0) {
	mutex_enter(&vd->vdev_dtl_lock);
	range_tree_walk(rt, range_tree_add,
	vd->vdev_dtl[DTL_MISSING]);
	mutex_exit(&vd->vdev_dtl_lock);
	}

	range_tree_vacate(rt, NULL, NULL);
	range_tree_destroy(rt);

	return (error);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	error = vdev_dtl_load(vd->vdev_child[c]);
	if (error != 0)
	break;
	}

	return (error);
	}

	static void
	vdev_zap_allocation_data(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	vdev_alloc_bias_t alloc_bias = vd->vdev_alloc_bias;
	const char *string;

	ASSERT(alloc_bias != VDEV_BIAS_NONE);

	string =
	(alloc_bias == VDEV_BIAS_LOG) ? VDEV_ALLOC_BIAS_LOG :
	(alloc_bias == VDEV_BIAS_SPECIAL) ? VDEV_ALLOC_BIAS_SPECIAL :
	(alloc_bias == VDEV_BIAS_DEDUP) ? VDEV_ALLOC_BIAS_DEDUP : NULL;

	ASSERT(string != NULL);
	VERIFY0(zap_add(mos, vd->vdev_top_zap, VDEV_TOP_ZAP_ALLOCATION_BIAS,
	1, strlen(string) + 1, string, tx));

	if (alloc_bias == VDEV_BIAS_SPECIAL \|\| alloc_bias == VDEV_BIAS_DEDUP) {
	spa_activate_allocation_classes(spa, tx);
	}
	}

	void
	vdev_destroy_unlink_zap(vdev_t vd, uint64_t zapobj, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;

	VERIFY0(zap_destroy(spa->spa_meta_objset, zapobj, tx));
	VERIFY0(zap_remove_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
	zapobj, tx));
	}

	uint64_t
	vdev_create_link_zap(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t zap = zap_create(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA,
	DMU_OT_NONE, 0, tx);

	ASSERT(zap != 0);
	VERIFY0(zap_add_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
	zap, tx));

	return (zap);
	}

	void
	vdev_construct_zaps(vdev_t vd, dmu_tx_t tx)
	{
	if (vd->vdev_ops != &vdev_hole_ops &&
	vd->vdev_ops != &vdev_missing_ops &&
	vd->vdev_ops != &vdev_root_ops &&
	!vd->vdev_top->vdev_removing) {
	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_leaf_zap == 0) {
	vd->vdev_leaf_zap = vdev_create_link_zap(vd, tx);
	}
	if (vd == vd->vdev_top && vd->vdev_top_zap == 0) {
	vd->vdev_top_zap = vdev_create_link_zap(vd, tx);
	if (vd->vdev_alloc_bias != VDEV_BIAS_NONE)
	vdev_zap_allocation_data(vd, tx);
	}
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_construct_zaps(vd->vdev_child[i], tx);
	}
	}

	static void
	vdev_dtl_sync(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
	objset_t *mos = spa->spa_meta_objset;
	range_tree_t *rtsync;
	dmu_tx_t *tx;
	uint64_t object = space_map_object(vd->vdev_dtl_sm);

	ASSERT(vdev_is_concrete(vd));
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);

	if (vd->vdev_detached \|\| vd->vdev_top->vdev_removing) {
	mutex_enter(&vd->vdev_dtl_lock);
	space_map_free(vd->vdev_dtl_sm, tx);
	space_map_close(vd->vdev_dtl_sm);
	vd->vdev_dtl_sm = NULL;
	mutex_exit(&vd->vdev_dtl_lock);

	/*
	* We only destroy the leaf ZAP for detached leaves or for
	* removed log devices. Removed data devices handle leaf ZAP
	* cleanup later, once cancellation is no longer possible.
	*/
	if (vd->vdev_leaf_zap != 0 && (vd->vdev_detached \|\|
	vd->vdev_top->vdev_islog)) {
	vdev_destroy_unlink_zap(vd, vd->vdev_leaf_zap, tx);
	vd->vdev_leaf_zap = 0;
	}

	dmu_tx_commit(tx);
	return;
	}

	if (vd->vdev_dtl_sm == NULL) {
	uint64_t new_object;

	new_object = space_map_alloc(mos, zfs_vdev_dtl_sm_blksz, tx);
	VERIFY3U(new_object, !=, 0);

	VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
	0, -1ULL, 0));
	ASSERT(vd->vdev_dtl_sm != NULL);
	}

	rtsync = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);

	mutex_enter(&vd->vdev_dtl_lock);
	range_tree_walk(rt, range_tree_add, rtsync);
	mutex_exit(&vd->vdev_dtl_lock);

	space_map_truncate(vd->vdev_dtl_sm, zfs_vdev_dtl_sm_blksz, tx);
	space_map_write(vd->vdev_dtl_sm, rtsync, SM_ALLOC, SM_NO_VDEVID, tx);
	range_tree_vacate(rtsync, NULL, NULL);

	range_tree_destroy(rtsync);

	/*
	* If the object for the space map has changed then dirty
	* the top level so that we update the config.
	*/
	if (object != space_map_object(vd->vdev_dtl_sm)) {
	vdev_dbgmsg(vd, "txg %llu, spa %s, DTL old object %llu, "
	"new object %llu", (u_longlong_t)txg, spa_name(spa),
	(u_longlong_t)object,
	(u_longlong_t)space_map_object(vd->vdev_dtl_sm));
	vdev_config_dirty(vd->vdev_top);
	}

	dmu_tx_commit(tx);
	}

	/*
	* Determine whether the specified vdev can be offlined/detached/removed
	* without losing data.
	*/
	boolean_t
	vdev_dtl_required(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *tvd = vd->vdev_top;
	uint8_t cant_read = vd->vdev_cant_read;
	boolean_t required;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	if (vd == spa->spa_root_vdev \|\| vd == tvd)
	return (B_TRUE);

	/*
	* Temporarily mark the device as unreadable, and then determine
	* whether this results in any DTL outages in the top-level vdev.
	* If not, we can safely offline/detach/remove the device.
	*/
	vd->vdev_cant_read = B_TRUE;
	vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
	required = !vdev_dtl_empty(tvd, DTL_OUTAGE);
	vd->vdev_cant_read = cant_read;
	vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);

	if (!required && zio_injection_enabled) {
	required = !!zio_handle_device_injection(vd, NULL,
	SET_ERROR(ECHILD));
	}

	return (required);
	}

	/*
	* Determine if resilver is needed, and if so the txg range.
	*/
	boolean_t
	vdev_resilver_needed(vdev_t vd, uint64_t minp, uint64_t *maxp)
	{
	boolean_t needed = B_FALSE;
	uint64_t thismin = UINT64_MAX;
	uint64_t thismax = 0;

	if (vd->vdev_children == 0) {
	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
	vdev_writeable(vd)) {

	thismin = vdev_dtl_min(vd);
	thismax = vdev_dtl_max(vd);
	needed = B_TRUE;
	}
	mutex_exit(&vd->vdev_dtl_lock);
	} else {
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	uint64_t cmin, cmax;

	if (vdev_resilver_needed(cvd, &cmin, &cmax)) {
	thismin = MIN(thismin, cmin);
	thismax = MAX(thismax, cmax);
	needed = B_TRUE;
	}
	}
	}

	if (needed && minp) {
	*minp = thismin;
	*maxp = thismax;
	}
	return (needed);
	}

	/*
	* Gets the checkpoint space map object from the vdev's ZAP. On success sm_obj
	* will contain either the checkpoint spacemap object or zero if none exists.
	* All other errors are returned to the caller.
	*/
	int
	vdev_checkpoint_sm_object(vdev_t vd, uint64_t sm_obj)
	{
	ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_top_zap == 0) {
	*sm_obj = 0;
	return (0);
	}

	int error = zap_lookup(spa_meta_objset(vd->vdev_spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM, sizeof (uint64_t), 1, sm_obj);
	if (error == ENOENT) {
	*sm_obj = 0;
	error = 0;
	}

	return (error);
	}

	int
	vdev_load(vdev_t *vd)
	{
	int children = vd->vdev_children;
	int error = 0;
	taskq_t *tq = NULL;

	/*
	* It's only worthwhile to use the taskq for the root vdev, because the
	* slow part is metaslab_init, and that only happens for top-level
	* vdevs.
	*/
	if (vd->vdev_ops == &vdev_root_ops && vd->vdev_children > 0) {
	tq = taskq_create("vdev_load", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	}

	/*
	* Recursively load all children.
	*/
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (tq == NULL \|\| vdev_uses_zvols(cvd)) {
	cvd->vdev_load_error = vdev_load(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_load_child,
	cvd, TQ_SLEEP) != TASKQID_INVALID);
	}
	}

	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	int error = vd->vdev_child[c]->vdev_load_error;

	if (error != 0)
	return (error);
	}

	vdev_set_deflate_ratio(vd);

	/*
	* On spa_load path, grab the allocation bias from our zap
	*/
	if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
	spa_t *spa = vd->vdev_spa;
	char bias_str[64];

	error = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_ALLOCATION_BIAS, 1, sizeof (bias_str),
	bias_str);
	if (error == 0) {
	ASSERT(vd->vdev_alloc_bias == VDEV_BIAS_NONE);
	vd->vdev_alloc_bias = vdev_derive_alloc_bias(bias_str);
	} else if (error != ENOENT) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(top_zap=%llu) "
	"failed [error=%d]",
	(u_longlong_t)vd->vdev_top_zap, error);
	return (error);
	}
	}

	/*
	* Load any rebuild state from the top-level vdev zap.
	*/
	if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
	error = vdev_rebuild_load(vd);
	if (error && error != ENOTSUP) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: vdev_rebuild_load "
	"failed [error=%d]", error);
	return (error);
	}
	}

	/*
	* If this is a top-level vdev, initialize its metaslabs.
	*/
	if (vd == vd->vdev_top && vdev_is_concrete(vd)) {
	vdev_metaslab_group_create(vd);

	if (vd->vdev_ashift == 0 \|\| vd->vdev_asize == 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: invalid size. ashift=%llu, "
	"asize=%llu", (u_longlong_t)vd->vdev_ashift,
	(u_longlong_t)vd->vdev_asize);
	return (SET_ERROR(ENXIO));
	}

	error = vdev_metaslab_init(vd, 0);
	if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: metaslab_init failed "
	"[error=%d]", error);
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	return (error);
	}

	uint64_t checkpoint_sm_obj;
	error = vdev_checkpoint_sm_object(vd, &checkpoint_sm_obj);
	if (error == 0 && checkpoint_sm_obj != 0) {
	objset_t *mos = spa_meta_objset(vd->vdev_spa);
	ASSERT(vd->vdev_asize != 0);
	ASSERT3P(vd->vdev_checkpoint_sm, ==, NULL);

	error = space_map_open(&vd->vdev_checkpoint_sm,
	mos, checkpoint_sm_obj, 0, vd->vdev_asize,
	vd->vdev_ashift);
	if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: space_map_open "
	"failed for checkpoint spacemap (obj %llu) "
	"[error=%d]",
	(u_longlong_t)checkpoint_sm_obj, error);
	return (error);
	}
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* Since the checkpoint_sm contains free entries
	* exclusively we can use space_map_allocated() to
	* indicate the cumulative checkpointed space that
	* has been freed.
	*/
	vd->vdev_stat.vs_checkpoint_space =
	-space_map_allocated(vd->vdev_checkpoint_sm);
	vd->vdev_spa->spa_checkpoint_info.sci_dspace +=
	vd->vdev_stat.vs_checkpoint_space;
	} else if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: failed to retrieve "
	"checkpoint space map object from vdev ZAP "
	"[error=%d]", error);
	return (error);
	}
	}

	/*
	* If this is a leaf vdev, load its DTL.
	*/
	if (vd->vdev_ops->vdev_op_leaf && (error = vdev_dtl_load(vd)) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: vdev_dtl_load failed "
	"[error=%d]", error);
	return (error);
	}

	uint64_t obsolete_sm_object;
	error = vdev_obsolete_sm_object(vd, &obsolete_sm_object);
	if (error == 0 && obsolete_sm_object != 0) {
	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	ASSERT(vd->vdev_asize != 0);
	ASSERT3P(vd->vdev_obsolete_sm, ==, NULL);

	if ((error = space_map_open(&vd->vdev_obsolete_sm, mos,
	obsolete_sm_object, 0, vd->vdev_asize, 0))) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: space_map_open failed for "
	"obsolete spacemap (obj %llu) [error=%d]",
	(u_longlong_t)obsolete_sm_object, error);
	return (error);
	}
	} else if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: failed to retrieve obsolete "
	"space map object from vdev ZAP [error=%d]", error);
	return (error);
	}

	return (0);
	}

	/*
	* The special vdev case is used for hot spares and l2cache devices. Its
	* sole purpose it to set the vdev state for the associated vdev. To do this,
	* we make sure that we can open the underlying device, then try to read the
	* label, and make sure that the label is sane and that it hasn't been
	* repurposed to another pool.
	*/
	int
	vdev_validate_aux(vdev_t *vd)
	{
	nvlist_t *label;
	uint64_t guid, version;
	uint64_t state;

	if (!vdev_readable(vd))
	return (0);

	if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	return (-1);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_VERSION, &version) != 0 \|\|
	!SPA_VERSION_IS_SUPPORTED(version) \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0 \|\|
	guid != vd->vdev_guid \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, &state) != 0) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	return (-1);
	}

	/*
	* We don't actually check the pool state here. If it's in fact in
	* use by another pool, we update this fact on the fly when requested.
	*/
	nvlist_free(label);
	return (0);
	}

	static void
	vdev_destroy_ms_flush_data(vdev_t vd, dmu_tx_t tx)
	{
	objset_t *mos = spa_meta_objset(vd->vdev_spa);

	if (vd->vdev_top_zap == 0)
	return;

	uint64_t object = 0;
	int err = zap_lookup(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1, &object);
	if (err == ENOENT)
	return;
	VERIFY0(err);

	VERIFY0(dmu_object_free(mos, object, tx));
	VERIFY0(zap_remove(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, tx));
	}

	/*
	* Free the objects used to store this vdev's spacemaps, and the array
	* that points to them.
	*/
	void
	vdev_destroy_spacemaps(vdev_t vd, dmu_tx_t tx)
	{
	if (vd->vdev_ms_array == 0)
	return;

	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	uint64_t array_count = vd->vdev_asize >> vd->vdev_ms_shift;
	size_t array_bytes = array_count * sizeof (uint64_t);
	uint64_t *smobj_array = kmem_alloc(array_bytes, KM_SLEEP);
	VERIFY0(dmu_read(mos, vd->vdev_ms_array, 0,
	array_bytes, smobj_array, 0));

	for (uint64_t i = 0; i < array_count; i++) {
	uint64_t smobj = smobj_array[i];
	if (smobj == 0)
	continue;

	space_map_free_obj(mos, smobj, tx);
	}

	kmem_free(smobj_array, array_bytes);
	VERIFY0(dmu_object_free(mos, vd->vdev_ms_array, tx));
	vdev_destroy_ms_flush_data(vd, tx);
	vd->vdev_ms_array = 0;
	}

	static void
	vdev_remove_empty_log(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(vd->vdev_islog);
	ASSERT(vd == vd->vdev_top);
	ASSERT3U(txg, ==, spa_syncing_txg(spa));

	dmu_tx_t *tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	vdev_destroy_spacemaps(vd, tx);
	if (vd->vdev_top_zap != 0) {
	vdev_destroy_unlink_zap(vd, vd->vdev_top_zap, tx);
	vd->vdev_top_zap = 0;
	}

	dmu_tx_commit(tx);
	}

	void
	vdev_sync_done(vdev_t *vd, uint64_t txg)
	{
	metaslab_t *msp;
	boolean_t reassess = !txg_list_empty(&vd->vdev_ms_list, TXG_CLEAN(txg));

	ASSERT(vdev_is_concrete(vd));

	while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg)))
	!= NULL)
	metaslab_sync_done(msp, txg);

	if (reassess) {
	metaslab_sync_reassess(vd->vdev_mg);
	if (vd->vdev_log_mg != NULL)
	metaslab_sync_reassess(vd->vdev_log_mg);
	}
	}

	void
	vdev_sync(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *lvd;
	metaslab_t *msp;

	ASSERT3U(txg, ==, spa->spa_syncing_txg);
	dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	if (range_tree_space(vd->vdev_obsolete_segments) > 0) {
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);

	vdev_indirect_sync_obsolete(vd, tx);

	/*
	* If the vdev is indirect, it can't have dirty
	* metaslabs or DTLs.
	*/
	if (vd->vdev_ops == &vdev_indirect_ops) {
	ASSERT(txg_list_empty(&vd->vdev_ms_list, txg));
	ASSERT(txg_list_empty(&vd->vdev_dtl_list, txg));
	dmu_tx_commit(tx);
	return;
	}
	}

	ASSERT(vdev_is_concrete(vd));

	if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0 &&
	!vd->vdev_removing) {
	ASSERT(vd == vd->vdev_top);
	ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
	vd->vdev_ms_array = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_OBJECT_ARRAY, 0, DMU_OT_NONE, 0, tx);
	ASSERT(vd->vdev_ms_array != 0);
	vdev_config_dirty(vd);
	}

	while ((msp = txg_list_remove(&vd->vdev_ms_list, txg)) != NULL) {
	metaslab_sync(msp, txg);
	(void) txg_list_add(&vd->vdev_ms_list, msp, TXG_CLEAN(txg));
	}

	while ((lvd = txg_list_remove(&vd->vdev_dtl_list, txg)) != NULL)
	vdev_dtl_sync(lvd, txg);

	/*
	* If this is an empty log device being removed, destroy the
	* metadata associated with it.
	*/
	if (vd->vdev_islog && vd->vdev_stat.vs_alloc == 0 && vd->vdev_removing)
	vdev_remove_empty_log(vd, txg);

	(void) txg_list_add(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg));
	dmu_tx_commit(tx);
	}

	uint64_t
	vdev_psize_to_asize(vdev_t *vd, uint64_t psize)
	{
	return (vd->vdev_ops->vdev_op_asize(vd, psize));
	}

	/*
	* Mark the given vdev faulted. A faulted vdev behaves as if the device could
	* not be opened, and no I/O is attempted.
	*/
	int
	vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux)
	{
	vdev_t vd, tvd;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	tvd = vd->vdev_top;

	/*
	* If user did a 'zpool offline -f' then make the fault persist across
	* reboots.
	*/
	if (aux == VDEV_AUX_EXTERNAL_PERSIST) {
	/*
	* There are two kinds of forced faults: temporary and
	* persistent. Temporary faults go away at pool import, while
	* persistent faults stay set. Both types of faults can be
	* cleared with a zpool clear.
	*
	* We tell if a vdev is persistently faulted by looking at the
	* ZPOOL_CONFIG_AUX_STATE nvpair. If it's set to "external" at
	* import then it's a persistent fault. Otherwise, it's
	* temporary. We get ZPOOL_CONFIG_AUX_STATE set to "external"
	* by setting vd.vdev_stat.vs_aux to VDEV_AUX_EXTERNAL. This
	* tells vdev_config_generate() (which gets run later) to set
	* ZPOOL_CONFIG_AUX_STATE to "external" in the nvlist.
	*/
	vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
	vd->vdev_tmpoffline = B_FALSE;
	aux = VDEV_AUX_EXTERNAL;
	} else {
	vd->vdev_tmpoffline = B_TRUE;
	}

	/*
	* We don't directly use the aux state here, but if we do a
	* vdev_reopen(), we need this value to be present to remember why we
	* were faulted.
	*/
	vd->vdev_label_aux = aux;

	/*
	* Faulted state takes precedence over degraded.
	*/
	vd->vdev_delayed_close = B_FALSE;
	vd->vdev_faulted = 1ULL;
	vd->vdev_degraded = 0ULL;
	vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED, aux);

	/*
	* If this device has the only valid copy of the data, then
	* back off and simply mark the vdev as degraded instead.
	*/
	if (!tvd->vdev_islog && vd->vdev_aux == NULL && vdev_dtl_required(vd)) {
	vd->vdev_degraded = 1ULL;
	vd->vdev_faulted = 0ULL;

	/*
	* If we reopen the device and it's not dead, only then do we
	* mark it degraded.
	*/
	vdev_reopen(tvd);

	if (vdev_readable(vd))
	vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, aux);
	}

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	/*
	* Mark the given vdev degraded. A degraded vdev is purely an indication to the
	* user that something is wrong. The vdev continues to operate as normal as far
	* as I/O is concerned.
	*/
	int
	vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux)
	{
	vdev_t *vd;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	/*
	* If the vdev is already faulted, then don't do anything.
	*/
	if (vd->vdev_faulted \|\| vd->vdev_degraded)
	return (spa_vdev_state_exit(spa, NULL, 0));

	vd->vdev_degraded = 1ULL;
	if (!vdev_is_dead(vd))
	vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED,
	aux);

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	+int
	+vdev_remove_wanted(spa_t *spa, uint64_t guid)
	+{
	+ vdev_t *vd;
	+
	+ spa_vdev_state_enter(spa, SCL_NONE);
	+
	+ if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	+ return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));
	+
	+ /*
	+ * If the vdev is already removed, then don't do anything.
	+ */
	+ if (vd->vdev_removed)
	+ return (spa_vdev_state_exit(spa, NULL, 0));
	+
	+ vd->vdev_remove_wanted = B_TRUE;
	+ spa_async_request(spa, SPA_ASYNC_REMOVE);
	+
	+ return (spa_vdev_state_exit(spa, vd, 0));
	+}
	+
	+
	/*
	* Online the given vdev.
	*
	* If 'ZFS_ONLINE_UNSPARE' is set, it implies two things. First, any attached
	* spare device should be detached when the device finishes resilvering.
	* Second, the online should be treated like a 'test' online case, so no FMA
	* events are generated if the device fails to open.
	*/
	int
	vdev_online(spa_t spa, uint64_t guid, uint64_t flags, vdev_state_t newstate)
	{
	vdev_t vd, tvd, pvd, rvd = spa->spa_root_vdev;
	boolean_t wasoffline;
	vdev_state_t oldstate;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	wasoffline = (vd->vdev_offline \|\| vd->vdev_tmpoffline);
	oldstate = vd->vdev_state;

	tvd = vd->vdev_top;
	vd->vdev_offline = B_FALSE;
	vd->vdev_tmpoffline = B_FALSE;
	vd->vdev_checkremove = !!(flags & ZFS_ONLINE_CHECKREMOVE);
	vd->vdev_forcefault = !!(flags & ZFS_ONLINE_FORCEFAULT);

	/* XXX - L2ARC 1.0 does not support expansion */
	if (!vd->vdev_aux) {
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	pvd->vdev_expanding = !!((flags & ZFS_ONLINE_EXPAND) \|\|
	spa->spa_autoexpand);
	vd->vdev_expansion_time = gethrestime_sec();
	}

	vdev_reopen(tvd);
	vd->vdev_checkremove = vd->vdev_forcefault = B_FALSE;

	if (!vd->vdev_aux) {
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	pvd->vdev_expanding = B_FALSE;
	}

	if (newstate)
	*newstate = vd->vdev_state;
	if ((flags & ZFS_ONLINE_UNSPARE) &&
	!vdev_is_dead(vd) && vd->vdev_parent &&
	vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
	vd->vdev_parent->vdev_child[0] == vd)
	vd->vdev_unspare = B_TRUE;

	if ((flags & ZFS_ONLINE_EXPAND) \|\| spa->spa_autoexpand) {

	/* XXX - L2ARC 1.0 does not support expansion */
	if (vd->vdev_aux)
	return (spa_vdev_state_exit(spa, vd, ENOTSUP));
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	/* Restart initializing if necessary */
	mutex_enter(&vd->vdev_initialize_lock);
	if (vdev_writeable(vd) &&
	vd->vdev_initialize_thread == NULL &&
	vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) {
	(void) vdev_initialize(vd);
	}
	mutex_exit(&vd->vdev_initialize_lock);

	/*
	* Restart trimming if necessary. We do not restart trimming for cache
	* devices here. This is triggered by l2arc_rebuild_vdev()
	* asynchronously for the whole device or in l2arc_evict() as it evicts
	* space for upcoming writes.
	*/
	mutex_enter(&vd->vdev_trim_lock);
	if (vdev_writeable(vd) && !vd->vdev_isl2cache &&
	vd->vdev_trim_thread == NULL &&
	vd->vdev_trim_state == VDEV_TRIM_ACTIVE) {
	(void) vdev_trim(vd, vd->vdev_trim_rate, vd->vdev_trim_partial,
	vd->vdev_trim_secure);
	}
	mutex_exit(&vd->vdev_trim_lock);

	if (wasoffline \|\|
	(oldstate < VDEV_STATE_DEGRADED &&
	vd->vdev_state >= VDEV_STATE_DEGRADED))
	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_ONLINE);

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	static int
	vdev_offline_locked(spa_t *spa, uint64_t guid, uint64_t flags)
	{
	vdev_t vd, tvd;
	int error = 0;
	uint64_t generation;
	metaslab_group_t *mg;

	top:
	spa_vdev_state_enter(spa, SCL_ALLOC);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	if (vd->vdev_ops == &vdev_draid_spare_ops)
	return (spa_vdev_state_exit(spa, NULL, ENOTSUP));

	tvd = vd->vdev_top;
	mg = tvd->vdev_mg;
	generation = spa->spa_config_generation + 1;

	/*
	* If the device isn't already offline, try to offline it.
	*/
	if (!vd->vdev_offline) {
	/*
	* If this device has the only valid copy of some data,
	* don't allow it to be offlined. Log devices are always
	* expendable.
	*/
	if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
	vdev_dtl_required(vd))
	return (spa_vdev_state_exit(spa, NULL,
	SET_ERROR(EBUSY)));

	/*
	* If the top-level is a slog and it has had allocations
	* then proceed. We check that the vdev's metaslab group
	* is not NULL since it's possible that we may have just
	* added this vdev but not yet initialized its metaslabs.
	*/
	if (tvd->vdev_islog && mg != NULL) {
	/*
	* Prevent any future allocations.
	*/
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_passivate(mg);
	(void) spa_vdev_state_exit(spa, vd, 0);

	error = spa_reset_logs(spa);

	/*
	* If the log device was successfully reset but has
	* checkpointed data, do not offline it.
	*/
	if (error == 0 &&
	tvd->vdev_checkpoint_sm != NULL) {
	ASSERT3U(space_map_allocated(
	tvd->vdev_checkpoint_sm), !=, 0);
	error = ZFS_ERR_CHECKPOINT_EXISTS;
	}

	spa_vdev_state_enter(spa, SCL_ALLOC);

	/*
	* Check to see if the config has changed.
	*/
	if (error \|\| generation != spa->spa_config_generation) {
	metaslab_group_activate(mg);
	if (error)
	return (spa_vdev_state_exit(spa,
	vd, error));
	(void) spa_vdev_state_exit(spa, vd, 0);
	goto top;
	}
	ASSERT0(tvd->vdev_stat.vs_alloc);
	}

	/*
	* Offline this device and reopen its top-level vdev.
	* If the top-level vdev is a log device then just offline
	* it. Otherwise, if this action results in the top-level
	* vdev becoming unusable, undo it and fail the request.
	*/
	vd->vdev_offline = B_TRUE;
	vdev_reopen(tvd);

	if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
	vdev_is_dead(tvd)) {
	vd->vdev_offline = B_FALSE;
	vdev_reopen(tvd);
	return (spa_vdev_state_exit(spa, NULL,
	SET_ERROR(EBUSY)));
	}

	/*
	* Add the device back into the metaslab rotor so that
	* once we online the device it's open for business.
	*/
	if (tvd->vdev_islog && mg != NULL)
	metaslab_group_activate(mg);
	}

	vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	int
	vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
	{
	int error;

	mutex_enter(&spa->spa_vdev_top_lock);
	error = vdev_offline_locked(spa, guid, flags);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* Clear the error counts associated with this vdev. Unlike vdev_online() and
	* vdev_offline(), we assume the spa config is locked. We also clear all
	* children. If 'vd' is NULL, then the user wants to clear all vdevs.
	*/
	void
	vdev_clear(spa_t spa, vdev_t vd)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	if (vd == NULL)
	vd = rvd;

	vd->vdev_stat.vs_read_errors = 0;
	vd->vdev_stat.vs_write_errors = 0;
	vd->vdev_stat.vs_checksum_errors = 0;
	vd->vdev_stat.vs_slow_ios = 0;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_clear(spa, vd->vdev_child[c]);

	/*
	* It makes no sense to "clear" an indirect vdev.
	*/
	if (!vdev_is_concrete(vd))
	return;

	/*
	* If we're in the FAULTED state or have experienced failed I/O, then
	* clear the persistent state and attempt to reopen the device. We
	* also mark the vdev config dirty, so that the new faulted state is
	* written out to disk.
	*/
	if (vd->vdev_faulted \|\| vd->vdev_degraded \|\|
	!vdev_readable(vd) \|\| !vdev_writeable(vd)) {
	/*
	* When reopening in response to a clear event, it may be due to
	* a fmadm repair request. In this case, if the device is
	* still broken, we want to still post the ereport again.
	*/
	vd->vdev_forcefault = B_TRUE;

	vd->vdev_faulted = vd->vdev_degraded = 0ULL;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	vd->vdev_stat.vs_aux = 0;

	vdev_reopen(vd == rvd ? rvd : vd->vdev_top);

	vd->vdev_forcefault = B_FALSE;

	if (vd != rvd && vdev_writeable(vd->vdev_top))
	vdev_state_dirty(vd->vdev_top);

	/* If a resilver isn't required, check if vdevs can be culled */
	if (vd->vdev_aux == NULL && !vdev_is_dead(vd) &&
	!dsl_scan_resilvering(spa->spa_dsl_pool) &&
	!dsl_scan_resilver_scheduled(spa->spa_dsl_pool))
	spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);

	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_CLEAR);
	}

	/*
	* When clearing a FMA-diagnosed fault, we always want to
	* unspare the device, as we assume that the original spare was
	* done in response to the FMA fault.
	*/
	if (!vdev_is_dead(vd) && vd->vdev_parent != NULL &&
	vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
	vd->vdev_parent->vdev_child[0] == vd)
	vd->vdev_unspare = B_TRUE;

	/* Clear recent error events cache (i.e. duplicate events tracking) */
	zfs_ereport_clear(spa, vd);
	}

	boolean_t
	vdev_is_dead(vdev_t *vd)
	{
	/*
	* Holes and missing devices are always considered "dead".
	* This simplifies the code since we don't have to check for
	* these types of devices in the various code paths.
	* Instead we rely on the fact that we skip over dead devices
	* before issuing I/O to them.
	*/
	return (vd->vdev_state < VDEV_STATE_DEGRADED \|\|
	vd->vdev_ops == &vdev_hole_ops \|\|
	vd->vdev_ops == &vdev_missing_ops);
	}

	boolean_t
	vdev_readable(vdev_t *vd)
	{
	return (!vdev_is_dead(vd) && !vd->vdev_cant_read);
	}

	boolean_t
	vdev_writeable(vdev_t *vd)
	{
	return (!vdev_is_dead(vd) && !vd->vdev_cant_write &&
	vdev_is_concrete(vd));
	}

	boolean_t
	vdev_allocatable(vdev_t *vd)
	{
	uint64_t state = vd->vdev_state;

	/*
	* We currently allow allocations from vdevs which may be in the
	* process of reopening (i.e. VDEV_STATE_CLOSED). If the device
	* fails to reopen then we'll catch it later when we're holding
	* the proper locks. Note that we have to get the vdev state
	* in a local variable because although it changes atomically,
	* we're asking two separate questions about it.
	*/
	return (!(state < VDEV_STATE_DEGRADED && state != VDEV_STATE_CLOSED) &&
	!vd->vdev_cant_write && vdev_is_concrete(vd) &&
	vd->vdev_mg->mg_initialized);
	}

	boolean_t
	vdev_accessible(vdev_t vd, zio_t zio)
	{
	ASSERT(zio->io_vd == vd);

	if (vdev_is_dead(vd) \|\| vd->vdev_remove_wanted)
	return (B_FALSE);

	if (zio->io_type == ZIO_TYPE_READ)
	return (!vd->vdev_cant_read);

	if (zio->io_type == ZIO_TYPE_WRITE)
	return (!vd->vdev_cant_write);

	return (B_TRUE);
	}

	static void
	vdev_get_child_stat(vdev_t cvd, vdev_stat_t vs, vdev_stat_t *cvs)
	{
	/*
	* Exclude the dRAID spare when aggregating to avoid double counting
	* the ops and bytes. These IOs are counted by the physical leaves.
	*/
	if (cvd->vdev_ops == &vdev_draid_spare_ops)
	return;

	for (int t = 0; t < VS_ZIO_TYPES; t++) {
	vs->vs_ops[t] += cvs->vs_ops[t];
	vs->vs_bytes[t] += cvs->vs_bytes[t];
	}

	cvs->vs_scan_removing = cvd->vdev_removing;
	}

	/*
	* Get extended stats
	*/
	static void
	vdev_get_child_stat_ex(vdev_t cvd, vdev_stat_ex_t vsx, vdev_stat_ex_t *cvsx)
	{
	(void) cvd;

	int t, b;
	for (t = 0; t < ZIO_TYPES; t++) {
	for (b = 0; b < ARRAY_SIZE(vsx->vsx_disk_histo[0]); b++)
	vsx->vsx_disk_histo[t][b] += cvsx->vsx_disk_histo[t][b];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_total_histo[0]); b++) {
	vsx->vsx_total_histo[t][b] +=
	cvsx->vsx_total_histo[t][b];
	}
	}

	for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) {
	for (b = 0; b < ARRAY_SIZE(vsx->vsx_queue_histo[0]); b++) {
	vsx->vsx_queue_histo[t][b] +=
	cvsx->vsx_queue_histo[t][b];
	}
	vsx->vsx_active_queue[t] += cvsx->vsx_active_queue[t];
	vsx->vsx_pend_queue[t] += cvsx->vsx_pend_queue[t];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_ind_histo[0]); b++)
	vsx->vsx_ind_histo[t][b] += cvsx->vsx_ind_histo[t][b];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_agg_histo[0]); b++)
	vsx->vsx_agg_histo[t][b] += cvsx->vsx_agg_histo[t][b];
	}

	}

	boolean_t
	vdev_is_spacemap_addressable(vdev_t *vd)
	{
	if (spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_SPACEMAP_V2))
	return (B_TRUE);

	/*
	* If double-word space map entries are not enabled we assume
	* 47 bits of the space map entry are dedicated to the entry's
	* offset (see SM_OFFSET_BITS in space_map.h). We then use that
	* to calculate the maximum address that can be described by a
	* space map entry for the given device.
	*/
	uint64_t shift = vd->vdev_ashift + SM_OFFSET_BITS;

	if (shift >= 63) /* detect potential overflow */
	return (B_TRUE);

	return (vd->vdev_asize < (1ULL << shift));
	}

	/*
	* Get statistics for the given vdev.
	*/
	static void
	vdev_get_stats_ex_impl(vdev_t vd, vdev_stat_t vs, vdev_stat_ex_t *vsx)
	{
	int t;
	/*
	* If we're getting stats on the root vdev, aggregate the I/O counts
	* over all top-level vdevs (i.e. the direct children of the root).
	*/
	if (!vd->vdev_ops->vdev_op_leaf) {
	if (vs) {
	memset(vs->vs_ops, 0, sizeof (vs->vs_ops));
	memset(vs->vs_bytes, 0, sizeof (vs->vs_bytes));
	}
	if (vsx)
	memset(vsx, 0, sizeof (*vsx));

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	vdev_stat_t *cvs = &cvd->vdev_stat;
	vdev_stat_ex_t *cvsx = &cvd->vdev_stat_ex;

	vdev_get_stats_ex_impl(cvd, cvs, cvsx);
	if (vs)
	vdev_get_child_stat(cvd, vs, cvs);
	if (vsx)
	vdev_get_child_stat_ex(cvd, vsx, cvsx);
	}
	} else {
	/*
	* We're a leaf. Just copy our ZIO active queue stats in. The
	* other leaf stats are updated in vdev_stat_update().
	*/
	if (!vsx)
	return;

	memcpy(vsx, &vd->vdev_stat_ex, sizeof (vd->vdev_stat_ex));

	for (t = 0; t < ARRAY_SIZE(vd->vdev_queue.vq_class); t++) {
	vsx->vsx_active_queue[t] =
	vd->vdev_queue.vq_class[t].vqc_active;
	vsx->vsx_pend_queue[t] = avl_numnodes(
	&vd->vdev_queue.vq_class[t].vqc_queued_tree);
	}
	}
	}

	void
	vdev_get_stats_ex(vdev_t vd, vdev_stat_t vs, vdev_stat_ex_t *vsx)
	{
	vdev_t *tvd = vd->vdev_top;
	mutex_enter(&vd->vdev_stat_lock);
	if (vs) {
	memcpy(vs, &vd->vdev_stat, sizeof (*vs));
	vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
	vs->vs_state = vd->vdev_state;
	vs->vs_rsize = vdev_get_min_asize(vd);

	if (vd->vdev_ops->vdev_op_leaf) {
	vs->vs_pspace = vd->vdev_psize;
	vs->vs_rsize += VDEV_LABEL_START_SIZE +
	VDEV_LABEL_END_SIZE;
	/*
	* Report initializing progress. Since we don't
	* have the initializing locks held, this is only
	* an estimate (although a fairly accurate one).
	*/
	vs->vs_initialize_bytes_done =
	vd->vdev_initialize_bytes_done;
	vs->vs_initialize_bytes_est =
	vd->vdev_initialize_bytes_est;
	vs->vs_initialize_state = vd->vdev_initialize_state;
	vs->vs_initialize_action_time =
	vd->vdev_initialize_action_time;

	/*
	* Report manual TRIM progress. Since we don't have
	* the manual TRIM locks held, this is only an
	* estimate (although fairly accurate one).
	*/
	vs->vs_trim_notsup = !vd->vdev_has_trim;
	vs->vs_trim_bytes_done = vd->vdev_trim_bytes_done;
	vs->vs_trim_bytes_est = vd->vdev_trim_bytes_est;
	vs->vs_trim_state = vd->vdev_trim_state;
	vs->vs_trim_action_time = vd->vdev_trim_action_time;

	/* Set when there is a deferred resilver. */
	vs->vs_resilver_deferred = vd->vdev_resilver_deferred;
	}

	/*
	* Report expandable space on top-level, non-auxiliary devices
	* only. The expandable space is reported in terms of metaslab
	* sized units since that determines how much space the pool
	* can expand.
	*/
	if (vd->vdev_aux == NULL && tvd != NULL) {
	vs->vs_esize = P2ALIGN(
	vd->vdev_max_asize - vd->vdev_asize,
	1ULL << tvd->vdev_ms_shift);
	}

	vs->vs_configured_ashift = vd->vdev_top != NULL
	? vd->vdev_top->vdev_ashift : vd->vdev_ashift;
	vs->vs_logical_ashift = vd->vdev_logical_ashift;
	if (vd->vdev_physical_ashift <= ASHIFT_MAX)
	vs->vs_physical_ashift = vd->vdev_physical_ashift;
	else
	vs->vs_physical_ashift = 0;

	/*
	* Report fragmentation and rebuild progress for top-level,
	* non-auxiliary, concrete devices.
	*/
	if (vd->vdev_aux == NULL && vd == vd->vdev_top &&
	vdev_is_concrete(vd)) {
	/*
	* The vdev fragmentation rating doesn't take into
	* account the embedded slog metaslab (vdev_log_mg).
	* Since it's only one metaslab, it would have a tiny
	* impact on the overall fragmentation.
	*/
	vs->vs_fragmentation = (vd->vdev_mg != NULL) ?
	vd->vdev_mg->mg_fragmentation : 0;
	}
	vs->vs_noalloc = MAX(vd->vdev_noalloc,
	tvd ? tvd->vdev_noalloc : 0);
	}

	vdev_get_stats_ex_impl(vd, vs, vsx);
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_get_stats(vdev_t vd, vdev_stat_t vs)
	{
	return (vdev_get_stats_ex(vd, vs, NULL));
	}

	void
	vdev_clear_stats(vdev_t *vd)
	{
	mutex_enter(&vd->vdev_stat_lock);
	vd->vdev_stat.vs_space = 0;
	vd->vdev_stat.vs_dspace = 0;
	vd->vdev_stat.vs_alloc = 0;
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_scan_stat_init(vdev_t *vd)
	{
	vdev_stat_t *vs = &vd->vdev_stat;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_scan_stat_init(vd->vdev_child[c]);

	mutex_enter(&vd->vdev_stat_lock);
	vs->vs_scan_processed = 0;
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_stat_update(zio_t *zio, uint64_t psize)
	{
	spa_t *spa = zio->io_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t *vd = zio->io_vd ? zio->io_vd : rvd;
	vdev_t *pvd;
	uint64_t txg = zio->io_txg;
	vdev_stat_t *vs = vd ? &vd->vdev_stat : NULL;
	vdev_stat_ex_t *vsx = vd ? &vd->vdev_stat_ex : NULL;
	zio_type_t type = zio->io_type;
	int flags = zio->io_flags;

	/*
	* If this i/o is a gang leader, it didn't do any actual work.
	*/
	if (zio->io_gang_tree)
	return;

	if (zio->io_error == 0) {
	/*
	* If this is a root i/o, don't count it -- we've already
	* counted the top-level vdevs, and vdev_get_stats() will
	* aggregate them when asked. This reduces contention on
	* the root vdev_stat_lock and implicitly handles blocks
	* that compress away to holes, for which there is no i/o.
	* (Holes never create vdev children, so all the counters
	* remain zero, which is what we want.)
	*
	* Note: this only applies to successful i/o (io_error == 0)
	* because unlike i/o counts, errors are not additive.
	* When reading a ditto block, for example, failure of
	* one top-level vdev does not imply a root-level error.
	*/
	if (vd == rvd)
	return;

	ASSERT(vd == zio->io_vd);

	if (flags & ZIO_FLAG_IO_BYPASS)
	return;

	mutex_enter(&vd->vdev_stat_lock);

	if (flags & ZIO_FLAG_IO_REPAIR) {
	/*
	* Repair is the result of a resilver issued by the
	* scan thread (spa_sync).
	*/
	if (flags & ZIO_FLAG_SCAN_THREAD) {
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
	dsl_scan_phys_t *scn_phys = &scn->scn_phys;
	uint64_t *processed = &scn_phys->scn_processed;

	if (vd->vdev_ops->vdev_op_leaf)
	atomic_add_64(processed, psize);
	vs->vs_scan_processed += psize;
	}

	/*
	* Repair is the result of a rebuild issued by the
	* rebuild thread (vdev_rebuild_thread). To avoid
	* double counting repaired bytes the virtual dRAID
	* spare vdev is excluded from the processed bytes.
	*/
	if (zio->io_priority == ZIO_PRIORITY_REBUILD) {
	vdev_t *tvd = vd->vdev_top;
	vdev_rebuild_t *vr = &tvd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	uint64_t *rebuilt = &vrp->vrp_bytes_rebuilt;

	if (vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops) {
	atomic_add_64(rebuilt, psize);
	}
	vs->vs_rebuild_processed += psize;
	}

	if (flags & ZIO_FLAG_SELF_HEAL)
	vs->vs_self_healed += psize;
	}

	/*
	* The bytes/ops/histograms are recorded at the leaf level and
	* aggregated into the higher level vdevs in vdev_get_stats().
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(zio->io_priority < ZIO_PRIORITY_NUM_QUEUEABLE)) {
	zio_type_t vs_type = type;
	zio_priority_t priority = zio->io_priority;

	/*
	* TRIM ops and bytes are reported to user space as
	* ZIO_TYPE_IOCTL. This is done to preserve the
	* vdev_stat_t structure layout for user space.
	*/
	if (type == ZIO_TYPE_TRIM)
	vs_type = ZIO_TYPE_IOCTL;

	/*
	* Solely for the purposes of 'zpool iostat -lqrw'
	* reporting use the priority to categorize the IO.
	* Only the following are reported to user space:
	*
	* ZIO_PRIORITY_SYNC_READ,
	* ZIO_PRIORITY_SYNC_WRITE,
	* ZIO_PRIORITY_ASYNC_READ,
	* ZIO_PRIORITY_ASYNC_WRITE,
	* ZIO_PRIORITY_SCRUB,
	* ZIO_PRIORITY_TRIM,
	* ZIO_PRIORITY_REBUILD.
	*/
	if (priority == ZIO_PRIORITY_INITIALIZING) {
	ASSERT3U(type, ==, ZIO_TYPE_WRITE);
	priority = ZIO_PRIORITY_ASYNC_WRITE;
	} else if (priority == ZIO_PRIORITY_REMOVAL) {
	priority = ((type == ZIO_TYPE_WRITE) ?
	ZIO_PRIORITY_ASYNC_WRITE :
	ZIO_PRIORITY_ASYNC_READ);
	}

	vs->vs_ops[vs_type]++;
	vs->vs_bytes[vs_type] += psize;

	if (flags & ZIO_FLAG_DELEGATED) {
	vsx->vsx_agg_histo[priority]
	[RQ_HISTO(zio->io_size)]++;
	} else {
	vsx->vsx_ind_histo[priority]
	[RQ_HISTO(zio->io_size)]++;
	}

	if (zio->io_delta && zio->io_delay) {
	vsx->vsx_queue_histo[priority]
	[L_HISTO(zio->io_delta - zio->io_delay)]++;
	vsx->vsx_disk_histo[type]
	[L_HISTO(zio->io_delay)]++;
	vsx->vsx_total_histo[type]
	[L_HISTO(zio->io_delta)]++;
	}
	}

	mutex_exit(&vd->vdev_stat_lock);
	return;
	}

	if (flags & ZIO_FLAG_SPECULATIVE)
	return;

	/*
	* If this is an I/O error that is going to be retried, then ignore the
	* error. Otherwise, the user may interpret B_FAILFAST I/O errors as
	* hard errors, when in reality they can happen for any number of
	* innocuous reasons (bus resets, MPxIO link failure, etc).
	*/
	if (zio->io_error == EIO &&
	!(zio->io_flags & ZIO_FLAG_IO_RETRY))
	return;

	/*
	* Intent logs writes won't propagate their error to the root
	* I/O so don't mark these types of failures as pool-level
	* errors.
	*/
	if (zio->io_vd == NULL && (zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
	return;

	if (type == ZIO_TYPE_WRITE && txg != 0 &&
	(!(flags & ZIO_FLAG_IO_REPAIR) \|\|
	(flags & ZIO_FLAG_SCAN_THREAD) \|\|
	spa->spa_claiming)) {
	/*
	* This is either a normal write (not a repair), or it's
	* a repair induced by the scrub thread, or it's a repair
	* made by zil_claim() during spa_load() in the first txg.
	* In the normal case, we commit the DTL change in the same
	* txg as the block was born. In the scrub-induced repair
	* case, we know that scrubs run in first-pass syncing context,
	* so we commit the DTL change in spa_syncing_txg(spa).
	* In the zil_claim() case, we commit in spa_first_txg(spa).
	*
	* We currently do not make DTL entries for failed spontaneous
	* self-healing writes triggered by normal (non-scrubbing)
	* reads, because we have no transactional context in which to
	* do so -- and it's not clear that it'd be desirable anyway.
	*/
	if (vd->vdev_ops->vdev_op_leaf) {
	uint64_t commit_txg = txg;
	if (flags & ZIO_FLAG_SCAN_THREAD) {
	ASSERT(flags & ZIO_FLAG_IO_REPAIR);
	ASSERT(spa_sync_pass(spa) == 1);
	vdev_dtl_dirty(vd, DTL_SCRUB, txg, 1);
	commit_txg = spa_syncing_txg(spa);
	} else if (spa->spa_claiming) {
	ASSERT(flags & ZIO_FLAG_IO_REPAIR);
	commit_txg = spa_first_txg(spa);
	}
	ASSERT(commit_txg >= spa_syncing_txg(spa));
	if (vdev_dtl_contains(vd, DTL_MISSING, txg, 1))
	return;
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	vdev_dtl_dirty(pvd, DTL_PARTIAL, txg, 1);
	vdev_dirty(vd->vdev_top, VDD_DTL, vd, commit_txg);
	}
	if (vd != rvd)
	vdev_dtl_dirty(vd, DTL_MISSING, txg, 1);
	}
	}

	int64_t
	vdev_deflated_space(vdev_t *vd, int64_t space)
	{
	ASSERT((space & (SPA_MINBLOCKSIZE-1)) == 0);
	ASSERT(vd->vdev_deflate_ratio != 0 \|\| vd->vdev_isl2cache);

	return ((space >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio);
	}

	/*
	* Update the in-core space usage stats for this vdev, its metaslab class,
	* and the root vdev.
	*/
	void
	vdev_space_update(vdev_t *vd, int64_t alloc_delta, int64_t defer_delta,
	int64_t space_delta)
	{
	(void) defer_delta;
	int64_t dspace_delta;
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(vd == vd->vdev_top);

	/*
	* Apply the inverse of the psize-to-asize (ie. RAID-Z) space-expansion
	* factor. We must calculate this here and not at the root vdev
	* because the root vdev's psize-to-asize is simply the max of its
	* children's, thus not accurate enough for us.
	*/
	dspace_delta = vdev_deflated_space(vd, space_delta);

	mutex_enter(&vd->vdev_stat_lock);
	/* ensure we won't underflow */
	if (alloc_delta < 0) {
	ASSERT3U(vd->vdev_stat.vs_alloc, >=, -alloc_delta);
	}

	vd->vdev_stat.vs_alloc += alloc_delta;
	vd->vdev_stat.vs_space += space_delta;
	vd->vdev_stat.vs_dspace += dspace_delta;
	mutex_exit(&vd->vdev_stat_lock);

	/* every class but log contributes to root space stats */
	if (vd->vdev_mg != NULL && !vd->vdev_islog) {
	ASSERT(!vd->vdev_isl2cache);
	mutex_enter(&rvd->vdev_stat_lock);
	rvd->vdev_stat.vs_alloc += alloc_delta;
	rvd->vdev_stat.vs_space += space_delta;
	rvd->vdev_stat.vs_dspace += dspace_delta;
	mutex_exit(&rvd->vdev_stat_lock);
	}
	/* Note: metaslab_class_space_update moved to metaslab_space_update */
	}

	/*
	* Mark a top-level vdev's config as dirty, placing it on the dirty list
	* so that it will be written out next time the vdev configuration is synced.
	* If the root vdev is specified (vdev_top == NULL), dirty all top-level vdevs.
	*/
	void
	vdev_config_dirty(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	int c;

	ASSERT(spa_writeable(spa));

	/*
	* If this is an aux vdev (as with l2cache and spare devices), then we
	* update the vdev config manually and set the sync flag.
	*/
	if (vd->vdev_aux != NULL) {
	spa_aux_vdev_t *sav = vd->vdev_aux;
	nvlist_t **aux;
	uint_t naux;

	for (c = 0; c < sav->sav_count; c++) {
	if (sav->sav_vdevs[c] == vd)
	break;
	}

	if (c == sav->sav_count) {
	/*
	* We're being removed. There's nothing more to do.
	*/
	ASSERT(sav->sav_sync == B_TRUE);
	return;
	}

	sav->sav_sync = B_TRUE;

	if (nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_L2CACHE, &aux, &naux) != 0) {
	VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_SPARES, &aux, &naux) == 0);
	}

	ASSERT(c < naux);

	/*
	* Setting the nvlist in the middle if the array is a little
	* sketchy, but it will work.
	*/
	nvlist_free(aux[c]);
	aux[c] = vdev_config_generate(spa, vd, B_TRUE, 0);

	return;
	}

	/*
	* The dirty list is protected by the SCL_CONFIG lock. The caller
	* must either hold SCL_CONFIG as writer, or must be the sync thread
	* (which holds SCL_CONFIG as reader). There's only one sync thread,
	* so this is sufficient to ensure mutual exclusion.
	*/
	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_CONFIG, RW_READER)));

	if (vd == rvd) {
	for (c = 0; c < rvd->vdev_children; c++)
	vdev_config_dirty(rvd->vdev_child[c]);
	} else {
	ASSERT(vd == vd->vdev_top);

	if (!list_link_active(&vd->vdev_config_dirty_node) &&
	vdev_is_concrete(vd)) {
	list_insert_head(&spa->spa_config_dirty_list, vd);
	}
	}
	}

	void
	vdev_config_clean(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_CONFIG, RW_READER)));

	ASSERT(list_link_active(&vd->vdev_config_dirty_node));
	list_remove(&spa->spa_config_dirty_list, vd);
	}

	/*
	* Mark a top-level vdev's state as dirty, so that the next pass of
	* spa_sync() can convert this into vdev_config_dirty(). We distinguish
	* the state changes from larger config changes because they require
	* much less locking, and are often needed for administrative actions.
	*/
	void
	vdev_state_dirty(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_writeable(spa));
	ASSERT(vd == vd->vdev_top);

	/*
	* The state list is protected by the SCL_STATE lock. The caller
	* must either hold SCL_STATE as writer, or must be the sync thread
	* (which holds SCL_STATE as reader). There's only one sync thread,
	* so this is sufficient to ensure mutual exclusion.
	*/
	ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_STATE, RW_READER)));

	if (!list_link_active(&vd->vdev_state_dirty_node) &&
	vdev_is_concrete(vd))
	list_insert_head(&spa->spa_state_dirty_list, vd);
	}

	void
	vdev_state_clean(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_STATE, RW_READER)));

	ASSERT(list_link_active(&vd->vdev_state_dirty_node));
	list_remove(&spa->spa_state_dirty_list, vd);
	}

	/*
	* Propagate vdev state up from children to parent.
	*/
	void
	vdev_propagate_state(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	int degraded = 0, faulted = 0;
	int corrupted = 0;
	vdev_t *child;

	if (vd->vdev_children > 0) {
	for (int c = 0; c < vd->vdev_children; c++) {
	child = vd->vdev_child[c];

	/*
	* Don't factor holes or indirect vdevs into the
	* decision.
	*/
	if (!vdev_is_concrete(child))
	continue;

	if (!vdev_readable(child) \|\|
	(!vdev_writeable(child) && spa_writeable(spa))) {
	/*
	* Root special: if there is a top-level log
	* device, treat the root vdev as if it were
	* degraded.
	*/
	if (child->vdev_islog && vd == rvd)
	degraded++;
	else
	faulted++;
	} else if (child->vdev_state <= VDEV_STATE_DEGRADED) {
	degraded++;
	}

	if (child->vdev_stat.vs_aux == VDEV_AUX_CORRUPT_DATA)
	corrupted++;
	}

	vd->vdev_ops->vdev_op_state_change(vd, faulted, degraded);

	/*
	* Root special: if there is a top-level vdev that cannot be
	* opened due to corrupted metadata, then propagate the root
	* vdev's aux state as 'corrupt' rather than 'insufficient
	* replicas'.
	*/
	if (corrupted && vd == rvd &&
	rvd->vdev_state == VDEV_STATE_CANT_OPEN)
	vdev_set_state(rvd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	}

	if (vd->vdev_parent)
	vdev_propagate_state(vd->vdev_parent);
	}

	/*
	* Set a vdev's state. If this is during an open, we don't update the parent
	* state, because we're in the process of opening children depth-first.
	* Otherwise, we propagate the change to the parent.
	*
	* If this routine places a device in a faulted state, an appropriate ereport is
	* generated.
	*/
	void
	vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
	{
	uint64_t save_state;
	spa_t *spa = vd->vdev_spa;

	if (state == vd->vdev_state) {
	/*
	* Since vdev_offline() code path is already in an offline
	* state we can miss a statechange event to OFFLINE. Check
	* the previous state to catch this condition.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(state == VDEV_STATE_OFFLINE) &&
	(vd->vdev_prevstate >= VDEV_STATE_FAULTED)) {
	/* post an offline state change */
	zfs_post_state_change(spa, vd, vd->vdev_prevstate);
	}
	vd->vdev_stat.vs_aux = aux;
	return;
	}

	save_state = vd->vdev_state;

	vd->vdev_state = state;
	vd->vdev_stat.vs_aux = aux;

	/*
	* If we are setting the vdev state to anything but an open state, then
	* always close the underlying device unless the device has requested
	* a delayed close (i.e. we're about to remove or fault the device).
	* Otherwise, we keep accessible but invalid devices open forever.
	* We don't call vdev_close() itself, because that implies some extra
	* checks (offline, etc) that we don't want here. This is limited to
	* leaf devices, because otherwise closing the device will affect other
	* children.
	*/
	if (!vd->vdev_delayed_close && vdev_is_dead(vd) &&
	vd->vdev_ops->vdev_op_leaf)
	vd->vdev_ops->vdev_op_close(vd);

	if (vd->vdev_removed &&
	state == VDEV_STATE_CANT_OPEN &&
	(aux == VDEV_AUX_OPEN_FAILED \|\| vd->vdev_checkremove)) {
	/*
	* If the previous state is set to VDEV_STATE_REMOVED, then this
	* device was previously marked removed and someone attempted to
	* reopen it. If this failed due to a nonexistent device, then
	* keep the device in the REMOVED state. We also let this be if
	* it is one of our special test online cases, which is only
	* attempting to online the device and shouldn't generate an FMA
	* fault.
	*/
	vd->vdev_state = VDEV_STATE_REMOVED;
	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	} else if (state == VDEV_STATE_REMOVED) {
	vd->vdev_removed = B_TRUE;
	} else if (state == VDEV_STATE_CANT_OPEN) {
	/*
	* If we fail to open a vdev during an import or recovery, we
	* mark it as "not available", which signifies that it was
	* never there to begin with. Failure to open such a device
	* is not considered an error.
	*/
	if ((spa_load_state(spa) == SPA_LOAD_IMPORT \|\|
	spa_load_state(spa) == SPA_LOAD_RECOVER) &&
	vd->vdev_ops->vdev_op_leaf)
	vd->vdev_not_present = 1;

	/*
	* Post the appropriate ereport. If the 'prevstate' field is
	* set to something other than VDEV_STATE_UNKNOWN, it indicates
	* that this is part of a vdev_reopen(). In this case, we don't
	* want to post the ereport if the device was already in the
	* CANT_OPEN state beforehand.
	*
	* If the 'checkremove' flag is set, then this is an attempt to
	* online the device in response to an insertion event. If we
	* hit this case, then we have detected an insertion event for a
	* faulted or offline device that wasn't in the removed state.
	* In this scenario, we don't post an ereport because we are
	* about to replace the device, or attempt an online with
	* vdev_forcefault, which will generate the fault for us.
	*/
	if ((vd->vdev_prevstate != state \|\| vd->vdev_forcefault) &&
	!vd->vdev_not_present && !vd->vdev_checkremove &&
	vd != spa->spa_root_vdev) {
	const char *class;

	switch (aux) {
	case VDEV_AUX_OPEN_FAILED:
	class = FM_EREPORT_ZFS_DEVICE_OPEN_FAILED;
	break;
	case VDEV_AUX_CORRUPT_DATA:
	class = FM_EREPORT_ZFS_DEVICE_CORRUPT_DATA;
	break;
	case VDEV_AUX_NO_REPLICAS:
	class = FM_EREPORT_ZFS_DEVICE_NO_REPLICAS;
	break;
	case VDEV_AUX_BAD_GUID_SUM:
	class = FM_EREPORT_ZFS_DEVICE_BAD_GUID_SUM;
	break;
	case VDEV_AUX_TOO_SMALL:
	class = FM_EREPORT_ZFS_DEVICE_TOO_SMALL;
	break;
	case VDEV_AUX_BAD_LABEL:
	class = FM_EREPORT_ZFS_DEVICE_BAD_LABEL;
	break;
	case VDEV_AUX_BAD_ASHIFT:
	class = FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT;
	break;
	default:
	class = FM_EREPORT_ZFS_DEVICE_UNKNOWN;
	}

	(void) zfs_ereport_post(class, spa, vd, NULL, NULL,
	save_state);
	}

	/* Erase any notion of persistent removed state */
	vd->vdev_removed = B_FALSE;
	} else {
	vd->vdev_removed = B_FALSE;
	}

	/*
	* Notify ZED of any significant state-change on a leaf vdev.
	*
	*/
	if (vd->vdev_ops->vdev_op_leaf) {
	/* preserve original state from a vdev_reopen() */
	if ((vd->vdev_prevstate != VDEV_STATE_UNKNOWN) &&
	(vd->vdev_prevstate != vd->vdev_state) &&
	(save_state <= VDEV_STATE_CLOSED))
	save_state = vd->vdev_prevstate;

	/* filter out state change due to initial vdev_open */
	if (save_state > VDEV_STATE_CLOSED)
	zfs_post_state_change(spa, vd, save_state);
	}

	if (!isopen && vd->vdev_parent)
	vdev_propagate_state(vd->vdev_parent);
	}

	boolean_t
	vdev_children_are_offline(vdev_t *vd)
	{
	ASSERT(!vd->vdev_ops->vdev_op_leaf);

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	if (vd->vdev_child[i]->vdev_state != VDEV_STATE_OFFLINE)
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	/*
	* Check the vdev configuration to ensure that it's capable of supporting
	* a root pool. We do not support partial configuration.
	*/
	boolean_t
	vdev_is_bootable(vdev_t *vd)
	{
	if (!vd->vdev_ops->vdev_op_leaf) {
	const char *vdev_type = vd->vdev_ops->vdev_op_type;

	if (strcmp(vdev_type, VDEV_TYPE_MISSING) == 0)
	return (B_FALSE);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	if (!vdev_is_bootable(vd->vdev_child[c]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	boolean_t
	vdev_is_concrete(vdev_t *vd)
	{
	vdev_ops_t *ops = vd->vdev_ops;
	if (ops == &vdev_indirect_ops \|\| ops == &vdev_hole_ops \|\|
	ops == &vdev_missing_ops \|\| ops == &vdev_root_ops) {
	return (B_FALSE);
	} else {
	return (B_TRUE);
	}
	}

	/*
	* Determine if a log device has valid content. If the vdev was
	* removed or faulted in the MOS config then we know that
	* the content on the log device has already been written to the pool.
	*/
	boolean_t
	vdev_log_state_valid(vdev_t *vd)
	{
	if (vd->vdev_ops->vdev_op_leaf && !vd->vdev_faulted &&
	!vd->vdev_removed)
	return (B_TRUE);

	for (int c = 0; c < vd->vdev_children; c++)
	if (vdev_log_state_valid(vd->vdev_child[c]))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Expand a vdev if possible.
	*/
	void
	vdev_expand(vdev_t *vd, uint64_t txg)
	{
	ASSERT(vd->vdev_top == vd);
	ASSERT(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
	ASSERT(vdev_is_concrete(vd));

	vdev_set_deflate_ratio(vd);

	if ((vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count &&
	vdev_is_concrete(vd)) {
	vdev_metaslab_group_create(vd);
	VERIFY(vdev_metaslab_init(vd, txg) == 0);
	vdev_config_dirty(vd);
	}
	}

	/*
	* Split a vdev.
	*/
	void
	vdev_split(vdev_t *vd)
	{
	vdev_t cvd, pvd = vd->vdev_parent;

	vdev_remove_child(pvd, vd);
	vdev_compact_children(pvd);

	cvd = pvd->vdev_child[0];
	if (pvd->vdev_children == 1) {
	vdev_remove_parent(cvd);
	cvd->vdev_splitting = B_TRUE;
	}
	vdev_propagate_state(cvd);
	}

	void
	vdev_deadman(vdev_t vd, const char tag)
	{
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	vdev_deadman(cvd, tag);
	}

	if (vd->vdev_ops->vdev_op_leaf) {
	vdev_queue_t *vq = &vd->vdev_queue;

	mutex_enter(&vq->vq_lock);
	if (avl_numnodes(&vq->vq_active_tree) > 0) {
	spa_t *spa = vd->vdev_spa;
	zio_t *fio;
	uint64_t delta;

	zfs_dbgmsg("slow vdev: %s has %lu active IOs",
	vd->vdev_path, avl_numnodes(&vq->vq_active_tree));

	/*
	* Look at the head of all the pending queues,
	* if any I/O has been outstanding for longer than
	* the spa_deadman_synctime invoke the deadman logic.
	*/
	fio = avl_first(&vq->vq_active_tree);
	delta = gethrtime() - fio->io_timestamp;
	if (delta > spa_deadman_synctime(spa))
	zio_deadman(fio, tag);
	}
	mutex_exit(&vq->vq_lock);
	}
	}

	void
	vdev_defer_resilver(vdev_t *vd)
	{
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	vd->vdev_resilver_deferred = B_TRUE;
	vd->vdev_spa->spa_resilver_deferred = B_TRUE;
	}

	/*
	* Clears the resilver deferred flag on all leaf devs under vd. Returns
	* B_TRUE if we have devices that need to be resilvered and are available to
	* accept resilver I/Os.
	*/
	boolean_t
	vdev_clear_resilver_deferred(vdev_t vd, dmu_tx_t tx)
	{
	boolean_t resilver_needed = B_FALSE;
	spa_t *spa = vd->vdev_spa;

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	resilver_needed \|= vdev_clear_resilver_deferred(cvd, tx);
	}

	if (vd == spa->spa_root_vdev &&
	spa_feature_is_active(spa, SPA_FEATURE_RESILVER_DEFER)) {
	spa_feature_decr(spa, SPA_FEATURE_RESILVER_DEFER, tx);
	vdev_config_dirty(vd);
	spa->spa_resilver_deferred = B_FALSE;
	return (resilver_needed);
	}

	if (!vdev_is_concrete(vd) \|\| vd->vdev_aux \|\|
	!vd->vdev_ops->vdev_op_leaf)
	return (resilver_needed);

	vd->vdev_resilver_deferred = B_FALSE;

	return (!vdev_is_dead(vd) && !vd->vdev_offline &&
	vdev_resilver_needed(vd, NULL, NULL));
	}

	boolean_t
	vdev_xlate_is_empty(range_seg64_t *rs)
	{
	return (rs->rs_start == rs->rs_end);
	}

	/*
	* Translate a logical range to the first contiguous physical range for the
	* specified vdev_t. This function is initially called with a leaf vdev and
	* will walk each parent vdev until it reaches a top-level vdev. Once the
	* top-level is reached the physical range is initialized and the recursive
	* function begins to unwind. As it unwinds it calls the parent's vdev
	* specific translation function to do the real conversion.
	*/
	void
	vdev_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs)
	{
	/*
	* Walk up the vdev tree
	*/
	if (vd != vd->vdev_top) {
	vdev_xlate(vd->vdev_parent, logical_rs, physical_rs,
	remain_rs);
	} else {
	/*
	* We've reached the top-level vdev, initialize the physical
	* range to the logical range and set an empty remaining
	* range then start to unwind.
	*/
	physical_rs->rs_start = logical_rs->rs_start;
	physical_rs->rs_end = logical_rs->rs_end;

	remain_rs->rs_start = logical_rs->rs_start;
	remain_rs->rs_end = logical_rs->rs_start;

	return;
	}

	vdev_t *pvd = vd->vdev_parent;
	ASSERT3P(pvd, !=, NULL);
	ASSERT3P(pvd->vdev_ops->vdev_op_xlate, !=, NULL);

	/*
	* As this recursive function unwinds, translate the logical
	* range into its physical and any remaining components by calling
	* the vdev specific translate function.
	*/
	range_seg64_t intermediate = { 0 };
	pvd->vdev_ops->vdev_op_xlate(vd, physical_rs, &intermediate, remain_rs);

	physical_rs->rs_start = intermediate.rs_start;
	physical_rs->rs_end = intermediate.rs_end;
	}

	void
	vdev_xlate_walk(vdev_t vd, const range_seg64_t logical_rs,
	vdev_xlate_func_t func, void arg)
	{
	range_seg64_t iter_rs = *logical_rs;
	range_seg64_t physical_rs;
	range_seg64_t remain_rs;

	while (!vdev_xlate_is_empty(&iter_rs)) {

	vdev_xlate(vd, &iter_rs, &physical_rs, &remain_rs);

	/*
	* With raidz and dRAID, it's possible that the logical range
	* does not live on this leaf vdev. Only when there is a non-
	* zero physical size call the provided function.
	*/
	if (!vdev_xlate_is_empty(&physical_rs))
	func(arg, &physical_rs);

	iter_rs = remain_rs;
	}
	}

	static char *
	vdev_name(vdev_t vd, char buf, int buflen)
	{
	if (vd->vdev_path == NULL) {
	if (strcmp(vd->vdev_ops->vdev_op_type, "root") == 0) {
	strlcpy(buf, vd->vdev_spa->spa_name, buflen);
	} else if (!vd->vdev_ops->vdev_op_leaf) {
	snprintf(buf, buflen, "%s-%llu",
	vd->vdev_ops->vdev_op_type,
	(u_longlong_t)vd->vdev_id);
	}
	} else {
	strlcpy(buf, vd->vdev_path, buflen);
	}
	return (buf);
	}

	/*
	* Look at the vdev tree and determine whether any devices are currently being
	* replaced.
	*/
	boolean_t
	vdev_replace_in_progress(vdev_t *vdev)
	{
	ASSERT(spa_config_held(vdev->vdev_spa, SCL_ALL, RW_READER) != 0);

	if (vdev->vdev_ops == &vdev_replacing_ops)
	return (B_TRUE);

	/*
	* A 'spare' vdev indicates that we have a replace in progress, unless
	* it has exactly two children, and the second, the hot spare, has
	* finished being resilvered.
	*/
	if (vdev->vdev_ops == &vdev_spare_ops && (vdev->vdev_children > 2 \|\|
	!vdev_dtl_empty(vdev->vdev_child[1], DTL_MISSING)))
	return (B_TRUE);

	for (int i = 0; i < vdev->vdev_children; i++) {
	if (vdev_replace_in_progress(vdev->vdev_child[i]))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Add a (source=src, propname=propval) list to an nvlist.
	*/
	static void
	vdev_prop_add_list(nvlist_t nvl, const char propname, char *strval,
	uint64_t intval, zprop_source_t src)
	{
	nvlist_t *propval;

	propval = fnvlist_alloc();
	fnvlist_add_uint64(propval, ZPROP_SOURCE, src);

	if (strval != NULL)
	fnvlist_add_string(propval, ZPROP_VALUE, strval);
	else
	fnvlist_add_uint64(propval, ZPROP_VALUE, intval);

	fnvlist_add_nvlist(nvl, propname, propval);
	nvlist_free(propval);
	}

	static void
	vdev_props_set_sync(void arg, dmu_tx_t tx)
	{
	vdev_t *vd;
	nvlist_t *nvp = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = spa->spa_meta_objset;
	nvpair_t *elem = NULL;
	uint64_t vdev_guid;
	nvlist_t *nvprops;

	vdev_guid = fnvlist_lookup_uint64(nvp, ZPOOL_VDEV_PROPS_SET_VDEV);
	nvprops = fnvlist_lookup_nvlist(nvp, ZPOOL_VDEV_PROPS_SET_PROPS);
	vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE);

	/* this vdev could get removed while waiting for this sync task */
	if (vd == NULL)
	return;

	mutex_enter(&spa->spa_props_lock);

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	uint64_t intval, objid = 0;
	char *strval;
	vdev_prop_t prop;
	const char *propname = nvpair_name(elem);
	zprop_type_t proptype;

	/*
	* Set vdev property values in the vdev props mos object.
	*/
	if (vd->vdev_top_zap != 0) {
	objid = vd->vdev_top_zap;
	} else if (vd->vdev_leaf_zap != 0) {
	objid = vd->vdev_leaf_zap;
	} else {
	panic("vdev not top or leaf");
	}

	switch (prop = vdev_name_to_prop(propname)) {
	case VDEV_PROP_USERPROP:
	if (vdev_prop_user(propname)) {
	strval = fnvpair_value_string(elem);
	if (strlen(strval) == 0) {
	/* remove the property if value == "" */
	(void) zap_remove(mos, objid, propname,
	tx);
	} else {
	VERIFY0(zap_update(mos, objid, propname,
	1, strlen(strval) + 1, strval, tx));
	}
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%s",
	(u_longlong_t)vdev_guid, nvpair_name(elem),
	strval);
	}
	break;
	default:
	/* normalize the property name */
	propname = vdev_prop_to_name(prop);
	proptype = vdev_prop_get_type(prop);

	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	ASSERT(proptype == PROP_TYPE_STRING);
	strval = fnvpair_value_string(elem);
	VERIFY0(zap_update(mos, objid, propname,
	1, strlen(strval) + 1, strval, tx));
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%s",
	(u_longlong_t)vdev_guid, nvpair_name(elem),
	strval);
	} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
	intval = fnvpair_value_uint64(elem);

	if (proptype == PROP_TYPE_INDEX) {
	const char *unused;
	VERIFY0(vdev_prop_index_to_string(
	prop, intval, &unused));
	}
	VERIFY0(zap_update(mos, objid, propname,
	sizeof (uint64_t), 1, &intval, tx));
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%lld",
	(u_longlong_t)vdev_guid,
	nvpair_name(elem), (longlong_t)intval);
	} else {
	panic("invalid vdev property type %u",
	nvpair_type(elem));
	}
	}

	}

	mutex_exit(&spa->spa_props_lock);
	}

	int
	vdev_prop_set(vdev_t vd, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa = vd->vdev_spa;
	nvpair_t *elem = NULL;
	uint64_t vdev_guid;
	nvlist_t *nvprops;
	int error;

	ASSERT(vd != NULL);

	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_SET_PROPS,
	&nvprops) != 0)
	return (SET_ERROR(EINVAL));

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL)
	return (SET_ERROR(EINVAL));

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	char *propname = nvpair_name(elem);
	vdev_prop_t prop = vdev_name_to_prop(propname);
	uint64_t intval = 0;
	char *strval = NULL;

	if (prop == VDEV_PROP_USERPROP && !vdev_prop_user(propname)) {
	error = EINVAL;
	goto end;
	}

	if (vdev_prop_readonly(prop)) {
	error = EROFS;
	goto end;
	}

	/* Special Processing */
	switch (prop) {
	case VDEV_PROP_PATH:
	if (vd->vdev_path == NULL) {
	error = EROFS;
	break;
	}
	if (nvpair_value_string(elem, &strval) != 0) {
	error = EINVAL;
	break;
	}
	/* New path must start with /dev/ */
	if (strncmp(strval, "/dev/", 5)) {
	error = EINVAL;
	break;
	}
	error = spa_vdev_setpath(spa, vdev_guid, strval);
	break;
	case VDEV_PROP_ALLOCATING:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	if (intval != vd->vdev_noalloc)
	break;
	if (intval == 0)
	error = spa_vdev_noalloc(spa, vdev_guid);
	else
	error = spa_vdev_alloc(spa, vdev_guid);
	break;
	default:
	/* Most processing is done in vdev_props_set_sync */
	break;
	}
	end:
	if (error != 0) {
	intval = error;
	vdev_prop_add_list(outnvl, propname, strval, intval, 0);
	return (error);
	}
	}

	return (dsl_sync_task(spa->spa_name, NULL, vdev_props_set_sync,
	innvl, 6, ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	int
	vdev_prop_get(vdev_t vd, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	int err = 0;
	uint64_t objid;
	uint64_t vdev_guid;
	nvpair_t *elem = NULL;
	nvlist_t *nvprops = NULL;
	uint64_t intval = 0;
	char *strval = NULL;
	const char *propname = NULL;
	vdev_prop_t prop;

	ASSERT(vd != NULL);
	ASSERT(mos != NULL);

	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_GET_PROPS, &nvprops);

	if (vd->vdev_top_zap != 0) {
	objid = vd->vdev_top_zap;
	} else if (vd->vdev_leaf_zap != 0) {
	objid = vd->vdev_leaf_zap;
	} else {
	return (SET_ERROR(EINVAL));
	}
	ASSERT(objid != 0);

	mutex_enter(&spa->spa_props_lock);

	if (nvprops != NULL) {
	char namebuf[64] = { 0 };

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	intval = 0;
	strval = NULL;
	propname = nvpair_name(elem);
	prop = vdev_name_to_prop(propname);
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	uint64_t integer_size, num_integers;

	switch (prop) {
	/* Special Read-only Properties */
	case VDEV_PROP_NAME:
	strval = vdev_name(vd, namebuf,
	sizeof (namebuf));
	if (strval == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname, strval, 0,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_CAPACITY:
	/* percent used */
	intval = (vd->vdev_stat.vs_dspace == 0) ? 0 :
	(vd->vdev_stat.vs_alloc * 100 /
	vd->vdev_stat.vs_dspace);
	vdev_prop_add_list(outnvl, propname, NULL,
	intval, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_STATE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_state, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_GUID:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_guid, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ASIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_asize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PSIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_psize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ASHIFT:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_ashift, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_SIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_dspace, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_dspace -
	vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ALLOCATED:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_EXPANDSZ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_esize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FRAGMENTATION:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_fragmentation,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PARITY:
	vdev_prop_add_list(outnvl, propname, NULL,
	vdev_get_nparity(vd), ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PATH:
	if (vd->vdev_path == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_path, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_DEVID:
	if (vd->vdev_devid == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_devid, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PHYS_PATH:
	if (vd->vdev_physpath == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_physpath, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ENC_PATH:
	if (vd->vdev_enc_sysfs_path == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_enc_sysfs_path, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FRU:
	if (vd->vdev_fru == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_fru, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PARENT:
	if (vd->vdev_parent != NULL) {
	strval = vdev_name(vd->vdev_parent,
	namebuf, sizeof (namebuf));
	vdev_prop_add_list(outnvl, propname,
	strval, 0, ZPROP_SRC_NONE);
	}
	continue;
	case VDEV_PROP_CHILDREN:
	if (vd->vdev_children > 0)
	strval = kmem_zalloc(ZAP_MAXVALUELEN,
	KM_SLEEP);
	for (uint64_t i = 0; i < vd->vdev_children;
	i++) {
	const char *vname;

	vname = vdev_name(vd->vdev_child[i],
	namebuf, sizeof (namebuf));
	if (vname == NULL)
	vname = "(unknown)";
	if (strlen(strval) > 0)
	strlcat(strval, ",",
	ZAP_MAXVALUELEN);
	strlcat(strval, vname, ZAP_MAXVALUELEN);
	}
	if (strval != NULL) {
	vdev_prop_add_list(outnvl, propname,
	strval, 0, ZPROP_SRC_NONE);
	kmem_free(strval, ZAP_MAXVALUELEN);
	}
	continue;
	case VDEV_PROP_NUMCHILDREN:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_children, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_READ_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_read_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_WRITE_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_write_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_CHECKSUM_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_checksum_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_INITIALIZE_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_initialize_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_NULL:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_NULL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_READ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_READ],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_WRITE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_WRITE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_FREE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_CLAIM:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_CLAIM],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_TRIM:
	/*
	* TRIM ops and bytes are reported to user
	* space as ZIO_TYPE_IOCTL. This is done to
	* preserve the vdev_stat_t structure layout
	* for user space.
	*/
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_IOCTL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_NULL:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_NULL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_READ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_READ],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_WRITE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_WRITE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_FREE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_CLAIM:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_CLAIM],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_TRIM:
	/*
	* TRIM ops and bytes are reported to user
	* space as ZIO_TYPE_IOCTL. This is done to
	* preserve the vdev_stat_t structure layout
	* for user space.
	*/
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_IOCTL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_REMOVING:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_removing, ZPROP_SRC_NONE);
	continue;
	/* Numeric Properites */
	case VDEV_PROP_ALLOCATING:
	src = ZPROP_SRC_LOCAL;
	strval = NULL;

	err = zap_lookup(mos, objid, nvpair_name(elem),
	sizeof (uint64_t), 1, &intval);
	if (err == ENOENT) {
	intval =
	vdev_prop_default_numeric(prop);
	err = 0;
	} else if (err)
	break;
	if (intval == vdev_prop_default_numeric(prop))
	src = ZPROP_SRC_DEFAULT;

	/* Leaf vdevs cannot have this property */
	if (vd->vdev_mg == NULL &&
	vd->vdev_top != NULL) {
	src = ZPROP_SRC_NONE;
	intval = ZPROP_BOOLEAN_NA;
	}

	vdev_prop_add_list(outnvl, propname, strval,
	intval, src);
	break;
	/* Text Properties */
	case VDEV_PROP_COMMENT:
	/* Exists in the ZAP below */
	/* FALLTHRU */
	case VDEV_PROP_USERPROP:
	/* User Properites */
	src = ZPROP_SRC_LOCAL;

	err = zap_length(mos, objid, nvpair_name(elem),
	&integer_size, &num_integers);
	if (err)
	break;

	switch (integer_size) {
	case 8:
	/* User properties cannot be integers */
	err = EINVAL;
	break;
	case 1:
	/* string property */
	strval = kmem_alloc(num_integers,
	KM_SLEEP);
	err = zap_lookup(mos, objid,
	nvpair_name(elem), 1,
	num_integers, strval);
	if (err) {
	kmem_free(strval,
	num_integers);
	break;
	}
	vdev_prop_add_list(outnvl, propname,
	strval, 0, src);
	kmem_free(strval, num_integers);
	break;
	}
	break;
	default:
	err = ENOENT;
	break;
	}
	if (err)
	break;
	}
	} else {
	/*
	* Get all properties from the MOS vdev property object.
	*/
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, mos, objid);
	(err = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	intval = 0;
	strval = NULL;
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	propname = za.za_name;
	prop = vdev_name_to_prop(propname);

	switch (za.za_integer_length) {
	case 8:
	/* We do not allow integer user properties */
	/* This is likely an internal value */
	break;
	case 1:
	/* string property */
	strval = kmem_alloc(za.za_num_integers,
	KM_SLEEP);
	err = zap_lookup(mos, objid, za.za_name, 1,
	za.za_num_integers, strval);
	if (err) {
	kmem_free(strval, za.za_num_integers);
	break;
	}
	vdev_prop_add_list(outnvl, propname, strval, 0,
	src);
	kmem_free(strval, za.za_num_integers);
	break;

	default:
	break;
	}
	}
	zap_cursor_fini(&zc);
	}

	mutex_exit(&spa->spa_props_lock);
	if (err && err != ENOENT) {
	return (err);
	}

	return (0);
	}

	EXPORT_SYMBOL(vdev_fault);
	EXPORT_SYMBOL(vdev_degrade);
	EXPORT_SYMBOL(vdev_online);
	EXPORT_SYMBOL(vdev_offline);
	EXPORT_SYMBOL(vdev_clear);

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_count, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_count, UINT, ZMOD_RW,
	"Target number of metaslabs per top-level vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_shift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_shift, UINT, ZMOD_RW,
	"Default limit for metaslab size");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, min_ms_count, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, min_ms_count, UINT, ZMOD_RW,
	"Minimum number of metaslabs per top-level vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, ms_count_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, ms_count_limit, UINT, ZMOD_RW,
	"Practical upper limit of total metaslabs per top-level vdev");

	ZFS_MODULE_PARAM(zfs, zfs_, slow_io_events_per_second, UINT, ZMOD_RW,
	"Rate limit slow IO (delay) events to this many per second");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, zfs_, checksum_events_per_second, UINT, ZMOD_RW,
	"Rate limit checksum events to this many checksum errors per second "
	"(do not set below ZED threshold).");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs, zfs_, scan_ignore_errors, INT, ZMOD_RW,
	"Ignore errors during resilver/scrub");

	ZFS_MODULE_PARAM(zfs_vdev, vdev_, validate_skip, INT, ZMOD_RW,
	"Bypass vdev_validate()");

	ZFS_MODULE_PARAM(zfs, zfs_, nocacheflush, INT, ZMOD_RW,
	"Disable cache flushes");

	-ZFS_MODULE_PARAM(zfs, zfs_, embedded_slog_min_ms, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, embedded_slog_min_ms, UINT, ZMOD_RW,
	"Minimum number of metaslabs required to dedicate one for log blocks");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, min_auto_ashift,
	param_set_min_auto_ashift, param_get_ulong, ZMOD_RW,
	"Minimum ashift used when creating new top-level vdevs");

	ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, max_auto_ashift,
	param_set_max_auto_ashift, param_get_ulong, ZMOD_RW,
	"Maximum ashift used when optimizing for logical -> physical sector "
	"size on new top-level vdevs");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_cache.c b/sys/contrib/openzfs/module/zfs/vdev_cache.c
	index d1b96b5e2eb4..f0a17600d58e 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_cache.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_cache.c
	@@ -1,436 +1,436 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/
	/*
	* Copyright (c) 2013, 2016 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/vdev_impl.h>
	#include <sys/zio.h>
	#include <sys/kstat.h>
	#include <sys/abd.h>

	/*
	* Virtual device read-ahead caching.
	*
	* This file implements a simple LRU read-ahead cache. When the DMU reads
	* a given block, it will often want other, nearby blocks soon thereafter.
	* We take advantage of this by reading a larger disk region and caching
	* the result. In the best case, this can turn 128 back-to-back 512-byte
	* reads into a single 64k read followed by 127 cache hits; this reduces
	* latency dramatically. In the worst case, it can turn an isolated 512-byte
	* read into a 64k read, which doesn't affect latency all that much but is
	* terribly wasteful of bandwidth. A more intelligent version of the cache
	* could keep track of access patterns and not do read-ahead unless it sees
	* at least two temporally close I/Os to the same region. Currently, only
	* metadata I/O is inflated. A further enhancement could take advantage of
	* more semantic information about the I/O. And it could use something
	* faster than an AVL tree; that was chosen solely for convenience.
	*
	* There are five cache operations: allocate, fill, read, write, evict.
	*
	* (1) Allocate. This reserves a cache entry for the specified region.
	* We separate the allocate and fill operations so that multiple threads
	* don't generate I/O for the same cache miss.
	*
	* (2) Fill. When the I/O for a cache miss completes, the fill routine
	* places the data in the previously allocated cache entry.
	*
	* (3) Read. Read data from the cache.
	*
	* (4) Write. Update cache contents after write completion.
	*
	* (5) Evict. When allocating a new entry, we evict the oldest (LRU) entry
	* if the total cache size exceeds zfs_vdev_cache_size.
	*/

	/*
	* These tunables are for performance analysis.
	*/
	/*
	* All i/os smaller than zfs_vdev_cache_max will be turned into
	* 1<<zfs_vdev_cache_bshift byte reads by the vdev_cache (aka software
	* track buffer). At most zfs_vdev_cache_size bytes will be kept in each
	* vdev's vdev_cache.
	*
	* TODO: Note that with the current ZFS code, it turns out that the
	* vdev cache is not helpful, and in some cases actually harmful. It
	* is better if we disable this. Once some time has passed, we should
	* actually remove this to simplify the code. For now we just disable
	* it by setting the zfs_vdev_cache_size to zero. Note that Solaris 11
	* has made these same changes.
	*/
	-static int zfs_vdev_cache_max = 1 << 14; /* 16KB */
	-static int zfs_vdev_cache_size = 0;
	-static int zfs_vdev_cache_bshift = 16;
	+static uint_t zfs_vdev_cache_max = 1 << 14; /* 16KB */
	+static uint_t zfs_vdev_cache_size = 0;
	+static uint_t zfs_vdev_cache_bshift = 16;

	#define VCBS (1 << zfs_vdev_cache_bshift) /* 64KB */

	static kstat_t *vdc_ksp = NULL;

	typedef struct vdc_stats {
	kstat_named_t vdc_stat_delegations;
	kstat_named_t vdc_stat_hits;
	kstat_named_t vdc_stat_misses;
	} vdc_stats_t;

	static vdc_stats_t vdc_stats = {
	{ "delegations", KSTAT_DATA_UINT64 },
	{ "hits", KSTAT_DATA_UINT64 },
	{ "misses", KSTAT_DATA_UINT64 }
	};

	#define VDCSTAT_BUMP(stat) atomic_inc_64(&vdc_stats.stat.value.ui64);

	static inline int
	vdev_cache_offset_compare(const void a1, const void a2)
	{
	const vdev_cache_entry_t ve1 = (const vdev_cache_entry_t )a1;
	const vdev_cache_entry_t ve2 = (const vdev_cache_entry_t )a2;

	return (TREE_CMP(ve1->ve_offset, ve2->ve_offset));
	}

	static int
	vdev_cache_lastused_compare(const void a1, const void a2)
	{
	const vdev_cache_entry_t ve1 = (const vdev_cache_entry_t )a1;
	const vdev_cache_entry_t ve2 = (const vdev_cache_entry_t )a2;

	int cmp = TREE_CMP(ve1->ve_lastused, ve2->ve_lastused);
	if (likely(cmp))
	return (cmp);

	/*
	* Among equally old entries, sort by offset to ensure uniqueness.
	*/
	return (vdev_cache_offset_compare(a1, a2));
	}

	/*
	* Evict the specified entry from the cache.
	*/
	static void
	vdev_cache_evict(vdev_cache_t vc, vdev_cache_entry_t ve)
	{
	ASSERT(MUTEX_HELD(&vc->vc_lock));
	ASSERT3P(ve->ve_fill_io, ==, NULL);
	ASSERT3P(ve->ve_abd, !=, NULL);

	avl_remove(&vc->vc_lastused_tree, ve);
	avl_remove(&vc->vc_offset_tree, ve);
	abd_free(ve->ve_abd);
	kmem_free(ve, sizeof (vdev_cache_entry_t));
	}

	/*
	* Allocate an entry in the cache. At the point we don't have the data,
	* we're just creating a placeholder so that multiple threads don't all
	* go off and read the same blocks.
	*/
	static vdev_cache_entry_t *
	vdev_cache_allocate(zio_t *zio)
	{
	vdev_cache_t *vc = &zio->io_vd->vdev_cache;
	uint64_t offset = P2ALIGN(zio->io_offset, VCBS);
	vdev_cache_entry_t *ve;

	ASSERT(MUTEX_HELD(&vc->vc_lock));

	if (zfs_vdev_cache_size == 0)
	return (NULL);

	/*
	* If adding a new entry would exceed the cache size,
	* evict the oldest entry (LRU).
	*/
	if ((avl_numnodes(&vc->vc_lastused_tree) << zfs_vdev_cache_bshift) >
	zfs_vdev_cache_size) {
	ve = avl_first(&vc->vc_lastused_tree);
	if (ve->ve_fill_io != NULL)
	return (NULL);
	ASSERT3U(ve->ve_hits, !=, 0);
	vdev_cache_evict(vc, ve);
	}

	ve = kmem_zalloc(sizeof (vdev_cache_entry_t), KM_SLEEP);
	ve->ve_offset = offset;
	ve->ve_lastused = ddi_get_lbolt();
	ve->ve_abd = abd_alloc_for_io(VCBS, B_TRUE);

	avl_add(&vc->vc_offset_tree, ve);
	avl_add(&vc->vc_lastused_tree, ve);

	return (ve);
	}

	static void
	vdev_cache_hit(vdev_cache_t vc, vdev_cache_entry_t ve, zio_t *zio)
	{
	uint64_t cache_phase = P2PHASE(zio->io_offset, VCBS);

	ASSERT(MUTEX_HELD(&vc->vc_lock));
	ASSERT3P(ve->ve_fill_io, ==, NULL);

	if (ve->ve_lastused != ddi_get_lbolt()) {
	avl_remove(&vc->vc_lastused_tree, ve);
	ve->ve_lastused = ddi_get_lbolt();
	avl_add(&vc->vc_lastused_tree, ve);
	}

	ve->ve_hits++;
	abd_copy_off(zio->io_abd, ve->ve_abd, 0, cache_phase, zio->io_size);
	}

	/*
	* Fill a previously allocated cache entry with data.
	*/
	static void
	vdev_cache_fill(zio_t *fio)
	{
	vdev_t *vd = fio->io_vd;
	vdev_cache_t *vc = &vd->vdev_cache;
	vdev_cache_entry_t *ve = fio->io_private;
	zio_t *pio;

	ASSERT3U(fio->io_size, ==, VCBS);

	/*
	* Add data to the cache.
	*/
	mutex_enter(&vc->vc_lock);

	ASSERT3P(ve->ve_fill_io, ==, fio);
	ASSERT3U(ve->ve_offset, ==, fio->io_offset);
	ASSERT3P(ve->ve_abd, ==, fio->io_abd);

	ve->ve_fill_io = NULL;

	/*
	* Even if this cache line was invalidated by a missed write update,
	* any reads that were queued up before the missed update are still
	* valid, so we can satisfy them from this line before we evict it.
	*/
	zio_link_t *zl = NULL;
	while ((pio = zio_walk_parents(fio, &zl)) != NULL)
	vdev_cache_hit(vc, ve, pio);

	if (fio->io_error \|\| ve->ve_missed_update)
	vdev_cache_evict(vc, ve);

	mutex_exit(&vc->vc_lock);
	}

	/*
	* Read data from the cache. Returns B_TRUE cache hit, B_FALSE on miss.
	*/
	boolean_t
	vdev_cache_read(zio_t *zio)
	{
	vdev_cache_t *vc = &zio->io_vd->vdev_cache;
	vdev_cache_entry_t *ve, ve_search;
	uint64_t cache_offset = P2ALIGN(zio->io_offset, VCBS);
	zio_t *fio;
	uint64_t cache_phase __maybe_unused = P2PHASE(zio->io_offset, VCBS);

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);

	if (zfs_vdev_cache_size == 0)
	return (B_FALSE);

	if (zio->io_flags & ZIO_FLAG_DONT_CACHE)
	return (B_FALSE);

	if (zio->io_size > zfs_vdev_cache_max)
	return (B_FALSE);

	/*
	* If the I/O straddles two or more cache blocks, don't cache it.
	*/
	if (P2BOUNDARY(zio->io_offset, zio->io_size, VCBS))
	return (B_FALSE);

	ASSERT3U(cache_phase + zio->io_size, <=, VCBS);

	mutex_enter(&vc->vc_lock);

	ve_search.ve_offset = cache_offset;
	ve = avl_find(&vc->vc_offset_tree, &ve_search, NULL);

	if (ve != NULL) {
	if (ve->ve_missed_update) {
	mutex_exit(&vc->vc_lock);
	return (B_FALSE);
	}

	if ((fio = ve->ve_fill_io) != NULL) {
	zio_vdev_io_bypass(zio);
	zio_add_child(zio, fio);
	mutex_exit(&vc->vc_lock);
	VDCSTAT_BUMP(vdc_stat_delegations);
	return (B_TRUE);
	}

	vdev_cache_hit(vc, ve, zio);
	zio_vdev_io_bypass(zio);

	mutex_exit(&vc->vc_lock);
	VDCSTAT_BUMP(vdc_stat_hits);
	return (B_TRUE);
	}

	ve = vdev_cache_allocate(zio);

	if (ve == NULL) {
	mutex_exit(&vc->vc_lock);
	return (B_FALSE);
	}

	fio = zio_vdev_delegated_io(zio->io_vd, cache_offset,
	ve->ve_abd, VCBS, ZIO_TYPE_READ, ZIO_PRIORITY_NOW,
	ZIO_FLAG_DONT_CACHE, vdev_cache_fill, ve);

	ve->ve_fill_io = fio;
	zio_vdev_io_bypass(zio);
	zio_add_child(zio, fio);

	mutex_exit(&vc->vc_lock);
	zio_nowait(fio);
	VDCSTAT_BUMP(vdc_stat_misses);

	return (B_TRUE);
	}

	/*
	* Update cache contents upon write completion.
	*/
	void
	vdev_cache_write(zio_t *zio)
	{
	vdev_cache_t *vc = &zio->io_vd->vdev_cache;
	vdev_cache_entry_t *ve, ve_search;
	uint64_t io_start = zio->io_offset;
	uint64_t io_end = io_start + zio->io_size;
	uint64_t min_offset = P2ALIGN(io_start, VCBS);
	uint64_t max_offset = P2ROUNDUP(io_end, VCBS);
	avl_index_t where;

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);

	mutex_enter(&vc->vc_lock);

	ve_search.ve_offset = min_offset;
	ve = avl_find(&vc->vc_offset_tree, &ve_search, &where);

	if (ve == NULL)
	ve = avl_nearest(&vc->vc_offset_tree, where, AVL_AFTER);

	while (ve != NULL && ve->ve_offset < max_offset) {
	uint64_t start = MAX(ve->ve_offset, io_start);
	uint64_t end = MIN(ve->ve_offset + VCBS, io_end);

	if (ve->ve_fill_io != NULL) {
	ve->ve_missed_update = 1;
	} else {
	abd_copy_off(ve->ve_abd, zio->io_abd,
	start - ve->ve_offset, start - io_start,
	end - start);
	}
	ve = AVL_NEXT(&vc->vc_offset_tree, ve);
	}
	mutex_exit(&vc->vc_lock);
	}

	void
	vdev_cache_purge(vdev_t *vd)
	{
	vdev_cache_t *vc = &vd->vdev_cache;
	vdev_cache_entry_t *ve;

	mutex_enter(&vc->vc_lock);
	while ((ve = avl_first(&vc->vc_offset_tree)) != NULL)
	vdev_cache_evict(vc, ve);
	mutex_exit(&vc->vc_lock);
	}

	void
	vdev_cache_init(vdev_t *vd)
	{
	vdev_cache_t *vc = &vd->vdev_cache;

	mutex_init(&vc->vc_lock, NULL, MUTEX_DEFAULT, NULL);

	avl_create(&vc->vc_offset_tree, vdev_cache_offset_compare,
	sizeof (vdev_cache_entry_t),
	offsetof(struct vdev_cache_entry, ve_offset_node));

	avl_create(&vc->vc_lastused_tree, vdev_cache_lastused_compare,
	sizeof (vdev_cache_entry_t),
	offsetof(struct vdev_cache_entry, ve_lastused_node));
	}

	void
	vdev_cache_fini(vdev_t *vd)
	{
	vdev_cache_t *vc = &vd->vdev_cache;

	vdev_cache_purge(vd);

	avl_destroy(&vc->vc_offset_tree);
	avl_destroy(&vc->vc_lastused_tree);

	mutex_destroy(&vc->vc_lock);
	}

	void
	vdev_cache_stat_init(void)
	{
	vdc_ksp = kstat_create("zfs", 0, "vdev_cache_stats", "misc",
	KSTAT_TYPE_NAMED, sizeof (vdc_stats) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);
	if (vdc_ksp != NULL) {
	vdc_ksp->ks_data = &vdc_stats;
	kstat_install(vdc_ksp);
	}
	}

	void
	vdev_cache_stat_fini(void)
	{
	if (vdc_ksp != NULL) {
	kstat_delete(vdc_ksp);
	vdc_ksp = NULL;
	}
	}

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_max, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_max, UINT, ZMOD_RW,
	"Inflate reads small than max");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_size, INT, ZMOD_RD,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_size, UINT, ZMOD_RD,
	"Total size of the per-disk cache");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_bshift, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, cache_bshift, UINT, ZMOD_RW,
	"Shift size to inflate reads too");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_indirect.c b/sys/contrib/openzfs/module/zfs/vdev_indirect.c
	index 9189d3f31241..0ca0c245e952 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_indirect.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_indirect.c
	@@ -1,1910 +1,1910 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2014, 2017 by Delphix. All rights reserved.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2014, 2020 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/metaslab.h>
	#include <sys/dmu.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zap.h>
	#include <sys/abd.h>
	#include <sys/zthr.h>

	/*
	* An indirect vdev corresponds to a vdev that has been removed. Since
	* we cannot rewrite block pointers of snapshots, etc., we keep a
	* mapping from old location on the removed device to the new location
	* on another device in the pool and use this mapping whenever we need
	* to access the DVA. Unfortunately, this mapping did not respect
	* logical block boundaries when it was first created, and so a DVA on
	* this indirect vdev may be "split" into multiple sections that each
	* map to a different location. As a consequence, not all DVAs can be
	* translated to an equivalent new DVA. Instead we must provide a
	* "vdev_remap" operation that executes a callback on each contiguous
	* segment of the new location. This function is used in multiple ways:
	*
	* - I/Os to this vdev use the callback to determine where the
	* data is now located, and issue child I/Os for each segment's new
	* location.
	*
	* - frees and claims to this vdev use the callback to free or claim
	* each mapped segment. (Note that we don't actually need to claim
	* log blocks on indirect vdevs, because we don't allocate to
	* removing vdevs. However, zdb uses zio_claim() for its leak
	* detection.)
	*/

	/*
	* "Big theory statement" for how we mark blocks obsolete.
	*
	* When a block on an indirect vdev is freed or remapped, a section of
	* that vdev's mapping may no longer be referenced (aka "obsolete"). We
	* keep track of how much of each mapping entry is obsolete. When
	* an entry becomes completely obsolete, we can remove it, thus reducing
	* the memory used by the mapping. The complete picture of obsolescence
	* is given by the following data structures, described below:
	* - the entry-specific obsolete count
	* - the vdev-specific obsolete spacemap
	* - the pool-specific obsolete bpobj
	*
	* == On disk data structures used ==
	*
	* We track the obsolete space for the pool using several objects. Each
	* of these objects is created on demand and freed when no longer
	* needed, and is assumed to be empty if it does not exist.
	* SPA_FEATURE_OBSOLETE_COUNTS includes the count of these objects.
	*
	* - Each vic_mapping_object (associated with an indirect vdev) can
	* have a vimp_counts_object. This is an array of uint32_t's
	* with the same number of entries as the vic_mapping_object. When
	* the mapping is condensed, entries from the vic_obsolete_sm_object
	* (see below) are folded into the counts. Therefore, each
	* obsolete_counts entry tells us the number of bytes in the
	* corresponding mapping entry that were not referenced when the
	* mapping was last condensed.
	*
	* - Each indirect or removing vdev can have a vic_obsolete_sm_object.
	* This is a space map containing an alloc entry for every DVA that
	* has been obsoleted since the last time this indirect vdev was
	* condensed. We use this object in order to improve performance
	* when marking a DVA as obsolete. Instead of modifying an arbitrary
	* offset of the vimp_counts_object, we only need to append an entry
	* to the end of this object. When a DVA becomes obsolete, it is
	* added to the obsolete space map. This happens when the DVA is
	* freed, remapped and not referenced by a snapshot, or the last
	* snapshot referencing it is destroyed.
	*
	* - Each dataset can have a ds_remap_deadlist object. This is a
	* deadlist object containing all blocks that were remapped in this
	* dataset but referenced in a previous snapshot. Blocks can only
	* appear on this list if they were remapped (dsl_dataset_block_remapped);
	* blocks that were killed in a head dataset are put on the normal
	* ds_deadlist and marked obsolete when they are freed.
	*
	* - The pool can have a dp_obsolete_bpobj. This is a list of blocks
	* in the pool that need to be marked obsolete. When a snapshot is
	* destroyed, we move some of the ds_remap_deadlist to the obsolete
	* bpobj (see dsl_destroy_snapshot_handle_remaps()). We then
	* asynchronously process the obsolete bpobj, moving its entries to
	* the specific vdevs' obsolete space maps.
	*
	* == Summary of how we mark blocks as obsolete ==
	*
	* - When freeing a block: if any DVA is on an indirect vdev, append to
	* vic_obsolete_sm_object.
	* - When remapping a block, add dva to ds_remap_deadlist (if prev snap
	* references; otherwise append to vic_obsolete_sm_object).
	* - When freeing a snapshot: move parts of ds_remap_deadlist to
	* dp_obsolete_bpobj (same algorithm as ds_deadlist).
	* - When syncing the spa: process dp_obsolete_bpobj, moving ranges to
	* individual vdev's vic_obsolete_sm_object.
	*/

	/*
	* "Big theory statement" for how we condense indirect vdevs.
	*
	* Condensing an indirect vdev's mapping is the process of determining
	* the precise counts of obsolete space for each mapping entry (by
	* integrating the obsolete spacemap into the obsolete counts) and
	* writing out a new mapping that contains only referenced entries.
	*
	* We condense a vdev when we expect the mapping to shrink (see
	* vdev_indirect_should_condense()), but only perform one condense at a
	* time to limit the memory usage. In addition, we use a separate
	* open-context thread (spa_condense_indirect_thread) to incrementally
	* create the new mapping object in a way that minimizes the impact on
	* the rest of the system.
	*
	* == Generating a new mapping ==
	*
	* To generate a new mapping, we follow these steps:
	*
	* 1. Save the old obsolete space map and create a new mapping object
	* (see spa_condense_indirect_start_sync()). This initializes the
	* spa_condensing_indirect_phys with the "previous obsolete space map",
	* which is now read only. Newly obsolete DVAs will be added to a
	* new (initially empty) obsolete space map, and will not be
	* considered as part of this condense operation.
	*
	* 2. Construct in memory the precise counts of obsolete space for each
	* mapping entry, by incorporating the obsolete space map into the
	* counts. (See vdev_indirect_mapping_load_obsolete_{counts,spacemap}().)
	*
	* 3. Iterate through each mapping entry, writing to the new mapping any
	* entries that are not completely obsolete (i.e. which don't have
	* obsolete count == mapping length). (See
	* spa_condense_indirect_generate_new_mapping().)
	*
	* 4. Destroy the old mapping object and switch over to the new one
	* (spa_condense_indirect_complete_sync).
	*
	* == Restarting from failure ==
	*
	* To restart the condense when we import/open the pool, we must start
	* at the 2nd step above: reconstruct the precise counts in memory,
	* based on the space map + counts. Then in the 3rd step, we start
	* iterating where we left off: at vimp_max_offset of the new mapping
	* object.
	*/

	static int zfs_condense_indirect_vdevs_enable = B_TRUE;

	/*
	* Condense if at least this percent of the bytes in the mapping is
	* obsolete. With the default of 25%, the amount of space mapped
	* will be reduced to 1% of its original size after at most 16
	* condenses. Higher values will condense less often (causing less
	* i/o); lower values will reduce the mapping size more quickly.
	*/
	-static int zfs_condense_indirect_obsolete_pct = 25;
	+static uint_t zfs_condense_indirect_obsolete_pct = 25;

	/*
	* Condense if the obsolete space map takes up more than this amount of
	* space on disk (logically). This limits the amount of disk space
	* consumed by the obsolete space map; the default of 1GB is small enough
	* that we typically don't mind "wasting" it.
	*/
	static unsigned long zfs_condense_max_obsolete_bytes = 1024 * 1024 * 1024;

	/*
	* Don't bother condensing if the mapping uses less than this amount of
	* memory. The default of 128KB is considered a "trivial" amount of
	* memory and not worth reducing.
	*/
	static unsigned long zfs_condense_min_mapping_bytes = 128 * 1024;

	/*
	* This is used by the test suite so that it can ensure that certain
	* actions happen while in the middle of a condense (which might otherwise
	* complete too quickly). If used to reduce the performance impact of
	* condensing in production, a maximum value of 1 should be sufficient.
	*/
	-static int zfs_condense_indirect_commit_entry_delay_ms = 0;
	+static uint_t zfs_condense_indirect_commit_entry_delay_ms = 0;

	/*
	* If an indirect split block contains more than this many possible unique
	* combinations when being reconstructed, consider it too computationally
	* expensive to check them all. Instead, try at most 100 randomly-selected
	* combinations each time the block is accessed. This allows all segment
	* copies to participate fairly in the reconstruction when all combinations
	* cannot be checked and prevents repeated use of one bad copy.
	*/
	-int zfs_reconstruct_indirect_combinations_max = 4096;
	+uint_t zfs_reconstruct_indirect_combinations_max = 4096;

	/*
	* Enable to simulate damaged segments and validate reconstruction. This
	* is intentionally not exposed as a module parameter.
	*/
	unsigned long zfs_reconstruct_indirect_damage_fraction = 0;

	/*
	* The indirect_child_t represents the vdev that we will read from, when we
	* need to read all copies of the data (e.g. for scrub or reconstruction).
	* For plain (non-mirror) top-level vdevs (i.e. is_vdev is not a mirror),
	* ic_vdev is the same as is_vdev. However, for mirror top-level vdevs,
	* ic_vdev is a child of the mirror.
	*/
	typedef struct indirect_child {
	abd_t *ic_data;
	vdev_t *ic_vdev;

	/*
	* ic_duplicate is NULL when the ic_data contents are unique, when it
	* is determined to be a duplicate it references the primary child.
	*/
	struct indirect_child *ic_duplicate;
	list_node_t ic_node; /* node on is_unique_child */
	int ic_error; /* set when a child does not contain the data */
	} indirect_child_t;

	/*
	* The indirect_split_t represents one mapped segment of an i/o to the
	* indirect vdev. For non-split (contiguously-mapped) blocks, there will be
	* only one indirect_split_t, with is_split_offset==0 and is_size==io_size.
	* For split blocks, there will be several of these.
	*/
	typedef struct indirect_split {
	list_node_t is_node; /* link on iv_splits */

	/*
	* is_split_offset is the offset into the i/o.
	* This is the sum of the previous splits' is_size's.
	*/
	uint64_t is_split_offset;

	vdev_t is_vdev; / top-level vdev */
	uint64_t is_target_offset; /* offset on is_vdev */
	uint64_t is_size;
	int is_children; /* number of entries in is_child[] */
	int is_unique_children; /* number of entries in is_unique_child */
	list_t is_unique_child;

	/*
	* is_good_child is the child that we are currently using to
	* attempt reconstruction.
	*/
	indirect_child_t *is_good_child;

	indirect_child_t is_child[1]; /* variable-length */
	} indirect_split_t;

	/*
	* The indirect_vsd_t is associated with each i/o to the indirect vdev.
	* It is the "Vdev-Specific Data" in the zio_t's io_vsd.
	*/
	typedef struct indirect_vsd {
	boolean_t iv_split_block;
	boolean_t iv_reconstruct;
	uint64_t iv_unique_combinations;
	uint64_t iv_attempts;
	uint64_t iv_attempts_max;

	list_t iv_splits; /* list of indirect_split_t's */
	} indirect_vsd_t;

	static void
	vdev_indirect_map_free(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;

	indirect_split_t *is;
	while ((is = list_head(&iv->iv_splits)) != NULL) {
	for (int c = 0; c < is->is_children; c++) {
	indirect_child_t *ic = &is->is_child[c];
	if (ic->ic_data != NULL)
	abd_free(ic->ic_data);
	}
	list_remove(&iv->iv_splits, is);

	indirect_child_t *ic;
	while ((ic = list_head(&is->is_unique_child)) != NULL)
	list_remove(&is->is_unique_child, ic);

	list_destroy(&is->is_unique_child);

	kmem_free(is,
	offsetof(indirect_split_t, is_child[is->is_children]));
	}
	kmem_free(iv, sizeof (*iv));
	}

	static const zio_vsd_ops_t vdev_indirect_vsd_ops = {
	.vsd_free = vdev_indirect_map_free,
	};

	/*
	* Mark the given offset and size as being obsolete.
	*/
	void
	vdev_indirect_mark_obsolete(vdev_t *vd, uint64_t offset, uint64_t size)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT3U(vd->vdev_indirect_config.vic_mapping_object, !=, 0);
	ASSERT(vd->vdev_removing \|\| vd->vdev_ops == &vdev_indirect_ops);
	ASSERT(size > 0);
	VERIFY(vdev_indirect_mapping_entry_for_offset(
	vd->vdev_indirect_mapping, offset) != NULL);

	if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
	mutex_enter(&vd->vdev_obsolete_lock);
	range_tree_add(vd->vdev_obsolete_segments, offset, size);
	mutex_exit(&vd->vdev_obsolete_lock);
	vdev_dirty(vd, 0, NULL, spa_syncing_txg(spa));
	}
	}

	/*
	* Mark the DVA vdev_id:offset:size as being obsolete in the given tx. This
	* wrapper is provided because the DMU does not know about vdev_t's and
	* cannot directly call vdev_indirect_mark_obsolete.
	*/
	void
	spa_vdev_indirect_mark_obsolete(spa_t *spa, uint64_t vdev_id, uint64_t offset,
	uint64_t size, dmu_tx_t *tx)
	{
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	ASSERT(dmu_tx_is_syncing(tx));

	/* The DMU can only remap indirect vdevs. */
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	vdev_indirect_mark_obsolete(vd, offset, size);
	}

	static spa_condensing_indirect_t *
	spa_condensing_indirect_create(spa_t *spa)
	{
	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	spa_condensing_indirect_t sci = kmem_zalloc(sizeof (sci), KM_SLEEP);
	objset_t *mos = spa->spa_meta_objset;

	for (int i = 0; i < TXG_SIZE; i++) {
	list_create(&sci->sci_new_mapping_entries[i],
	sizeof (vdev_indirect_mapping_entry_t),
	offsetof(vdev_indirect_mapping_entry_t, vime_node));
	}

	sci->sci_new_mapping =
	vdev_indirect_mapping_open(mos, scip->scip_next_mapping_object);

	return (sci);
	}

	static void
	spa_condensing_indirect_destroy(spa_condensing_indirect_t *sci)
	{
	for (int i = 0; i < TXG_SIZE; i++)
	list_destroy(&sci->sci_new_mapping_entries[i]);

	if (sci->sci_new_mapping != NULL)
	vdev_indirect_mapping_close(sci->sci_new_mapping);

	kmem_free(sci, sizeof (*sci));
	}

	boolean_t
	vdev_indirect_should_condense(vdev_t *vd)
	{
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	spa_t *spa = vd->vdev_spa;

	ASSERT(dsl_pool_sync_context(spa->spa_dsl_pool));

	if (!zfs_condense_indirect_vdevs_enable)
	return (B_FALSE);

	/*
	* We can only condense one indirect vdev at a time.
	*/
	if (spa->spa_condensing_indirect != NULL)
	return (B_FALSE);

	if (spa_shutting_down(spa))
	return (B_FALSE);

	/*
	* The mapping object size must not change while we are
	* condensing, so we can only condense indirect vdevs
	* (not vdevs that are still in the middle of being removed).
	*/
	if (vd->vdev_ops != &vdev_indirect_ops)
	return (B_FALSE);

	/*
	* If nothing new has been marked obsolete, there is no
	* point in condensing.
	*/
	uint64_t obsolete_sm_obj __maybe_unused;
	ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_obj));
	if (vd->vdev_obsolete_sm == NULL) {
	ASSERT0(obsolete_sm_obj);
	return (B_FALSE);
	}

	ASSERT(vd->vdev_obsolete_sm != NULL);

	ASSERT3U(obsolete_sm_obj, ==, space_map_object(vd->vdev_obsolete_sm));

	uint64_t bytes_mapped = vdev_indirect_mapping_bytes_mapped(vim);
	uint64_t bytes_obsolete = space_map_allocated(vd->vdev_obsolete_sm);
	uint64_t mapping_size = vdev_indirect_mapping_size(vim);
	uint64_t obsolete_sm_size = space_map_length(vd->vdev_obsolete_sm);

	ASSERT3U(bytes_obsolete, <=, bytes_mapped);

	/*
	* If a high percentage of the bytes that are mapped have become
	* obsolete, condense (unless the mapping is already small enough).
	* This has a good chance of reducing the amount of memory used
	* by the mapping.
	*/
	if (bytes_obsolete * 100 / bytes_mapped >=
	zfs_condense_indirect_obsolete_pct &&
	mapping_size > zfs_condense_min_mapping_bytes) {
	zfs_dbgmsg("should condense vdev %llu because obsolete "
	"spacemap covers %d%% of %lluMB mapping",
	(u_longlong_t)vd->vdev_id,
	(int)(bytes_obsolete * 100 / bytes_mapped),
	(u_longlong_t)bytes_mapped / 1024 / 1024);
	return (B_TRUE);
	}

	/*
	* If the obsolete space map takes up too much space on disk,
	* condense in order to free up this disk space.
	*/
	if (obsolete_sm_size >= zfs_condense_max_obsolete_bytes) {
	zfs_dbgmsg("should condense vdev %llu because obsolete sm "
	"length %lluMB >= max size %lluMB",
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)obsolete_sm_size / 1024 / 1024,
	(u_longlong_t)zfs_condense_max_obsolete_bytes /
	1024 / 1024);
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* This sync task completes (finishes) a condense, deleting the old
	* mapping and replacing it with the new one.
	*/
	static void
	spa_condense_indirect_complete_sync(void arg, dmu_tx_t tx)
	{
	spa_condensing_indirect_t *sci = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	vdev_t *vd = vdev_lookup_top(spa, scip->scip_vdev);
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
	objset_t *mos = spa->spa_meta_objset;
	vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
	uint64_t old_count = vdev_indirect_mapping_num_entries(old_mapping);
	uint64_t new_count =
	vdev_indirect_mapping_num_entries(sci->sci_new_mapping);

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	ASSERT3P(sci, ==, spa->spa_condensing_indirect);
	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT(list_is_empty(&sci->sci_new_mapping_entries[i]));
	}
	ASSERT(vic->vic_mapping_object != 0);
	ASSERT3U(vd->vdev_id, ==, scip->scip_vdev);
	ASSERT(scip->scip_next_mapping_object != 0);
	ASSERT(scip->scip_prev_obsolete_sm_object != 0);

	/*
	* Reset vdev_indirect_mapping to refer to the new object.
	*/
	rw_enter(&vd->vdev_indirect_rwlock, RW_WRITER);
	vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
	vd->vdev_indirect_mapping = sci->sci_new_mapping;
	rw_exit(&vd->vdev_indirect_rwlock);

	sci->sci_new_mapping = NULL;
	vdev_indirect_mapping_free(mos, vic->vic_mapping_object, tx);
	vic->vic_mapping_object = scip->scip_next_mapping_object;
	scip->scip_next_mapping_object = 0;

	space_map_free_obj(mos, scip->scip_prev_obsolete_sm_object, tx);
	spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	scip->scip_prev_obsolete_sm_object = 0;

	scip->scip_vdev = 0;

	VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CONDENSING_INDIRECT, tx));
	spa_condensing_indirect_destroy(spa->spa_condensing_indirect);
	spa->spa_condensing_indirect = NULL;

	zfs_dbgmsg("finished condense of vdev %llu in txg %llu: "
	"new mapping object %llu has %llu entries "
	"(was %llu entries)",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx),
	(u_longlong_t)vic->vic_mapping_object,
	(u_longlong_t)new_count, (u_longlong_t)old_count);

	vdev_config_dirty(spa->spa_root_vdev);
	}

	/*
	* This sync task appends entries to the new mapping object.
	*/
	static void
	spa_condense_indirect_commit_sync(void arg, dmu_tx_t tx)
	{
	spa_condensing_indirect_t *sci = arg;
	uint64_t txg = dmu_tx_get_txg(tx);
	spa_t *spa __maybe_unused = dmu_tx_pool(tx)->dp_spa;

	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT3P(sci, ==, spa->spa_condensing_indirect);

	vdev_indirect_mapping_add_entries(sci->sci_new_mapping,
	&sci->sci_new_mapping_entries[txg & TXG_MASK], tx);
	ASSERT(list_is_empty(&sci->sci_new_mapping_entries[txg & TXG_MASK]));
	}

	/*
	* Open-context function to add one entry to the new mapping. The new
	* entry will be remembered and written from syncing context.
	*/
	static void
	spa_condense_indirect_commit_entry(spa_t *spa,
	vdev_indirect_mapping_entry_phys_t *vimep, uint32_t count)
	{
	spa_condensing_indirect_t *sci = spa->spa_condensing_indirect;

	ASSERT3U(count, <, DVA_GET_ASIZE(&vimep->vimep_dst));

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	dmu_tx_hold_space(tx, sizeof (*vimep) + sizeof (count));
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;

	/*
	* If we are the first entry committed this txg, kick off the sync
	* task to write to the MOS on our behalf.
	*/
	if (list_is_empty(&sci->sci_new_mapping_entries[txgoff])) {
	dsl_sync_task_nowait(dmu_tx_pool(tx),
	spa_condense_indirect_commit_sync, sci, tx);
	}

	vdev_indirect_mapping_entry_t *vime =
	kmem_alloc(sizeof (*vime), KM_SLEEP);
	vime->vime_mapping = *vimep;
	vime->vime_obsolete_count = count;
	list_insert_tail(&sci->sci_new_mapping_entries[txgoff], vime);

	dmu_tx_commit(tx);
	}

	static void
	spa_condense_indirect_generate_new_mapping(vdev_t *vd,
	uint32_t obsolete_counts, uint64_t start_index, zthr_t zthr)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t mapi = start_index;
	vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
	uint64_t old_num_entries =
	vdev_indirect_mapping_num_entries(old_mapping);

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	ASSERT3U(vd->vdev_id, ==, spa->spa_condensing_indirect_phys.scip_vdev);

	zfs_dbgmsg("starting condense of vdev %llu from index %llu",
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)mapi);

	while (mapi < old_num_entries) {

	if (zthr_iscancelled(zthr)) {
	zfs_dbgmsg("pausing condense of vdev %llu "
	"at index %llu", (u_longlong_t)vd->vdev_id,
	(u_longlong_t)mapi);
	break;
	}

	vdev_indirect_mapping_entry_phys_t *entry =
	&old_mapping->vim_entries[mapi];
	uint64_t entry_size = DVA_GET_ASIZE(&entry->vimep_dst);
	ASSERT3U(obsolete_counts[mapi], <=, entry_size);
	if (obsolete_counts[mapi] < entry_size) {
	spa_condense_indirect_commit_entry(spa, entry,
	obsolete_counts[mapi]);

	/*
	* This delay may be requested for testing, debugging,
	* or performance reasons.
	*/
	hrtime_t now = gethrtime();
	hrtime_t sleep_until = now + MSEC2NSEC(
	zfs_condense_indirect_commit_entry_delay_ms);
	zfs_sleep_until(sleep_until);
	}

	mapi++;
	}
	}

	static boolean_t
	spa_condense_indirect_thread_check(void arg, zthr_t zthr)
	{
	(void) zthr;
	spa_t *spa = arg;

	return (spa->spa_condensing_indirect != NULL);
	}

	static void
	spa_condense_indirect_thread(void arg, zthr_t zthr)
	{
	spa_t *spa = arg;
	vdev_t *vd;

	ASSERT3P(spa->spa_condensing_indirect, !=, NULL);
	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	vd = vdev_lookup_top(spa, spa->spa_condensing_indirect_phys.scip_vdev);
	ASSERT3P(vd, !=, NULL);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	spa_condensing_indirect_t *sci = spa->spa_condensing_indirect;
	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	uint32_t *counts;
	uint64_t start_index;
	vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
	space_map_t *prev_obsolete_sm = NULL;

	ASSERT3U(vd->vdev_id, ==, scip->scip_vdev);
	ASSERT(scip->scip_next_mapping_object != 0);
	ASSERT(scip->scip_prev_obsolete_sm_object != 0);
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	for (int i = 0; i < TXG_SIZE; i++) {
	/*
	* The list must start out empty in order for the
	* _commit_sync() sync task to be properly registered
	* on the first call to _commit_entry(); so it's wise
	* to double check and ensure we actually are starting
	* with empty lists.
	*/
	ASSERT(list_is_empty(&sci->sci_new_mapping_entries[i]));
	}

	VERIFY0(space_map_open(&prev_obsolete_sm, spa->spa_meta_objset,
	scip->scip_prev_obsolete_sm_object, 0, vd->vdev_asize, 0));
	counts = vdev_indirect_mapping_load_obsolete_counts(old_mapping);
	if (prev_obsolete_sm != NULL) {
	vdev_indirect_mapping_load_obsolete_spacemap(old_mapping,
	counts, prev_obsolete_sm);
	}
	space_map_close(prev_obsolete_sm);

	/*
	* Generate new mapping. Determine what index to continue from
	* based on the max offset that we've already written in the
	* new mapping.
	*/
	uint64_t max_offset =
	vdev_indirect_mapping_max_offset(sci->sci_new_mapping);
	if (max_offset == 0) {
	/* We haven't written anything to the new mapping yet. */
	start_index = 0;
	} else {
	/*
	* Pick up from where we left off. _entry_for_offset()
	* returns a pointer into the vim_entries array. If
	* max_offset is greater than any of the mappings
	* contained in the table NULL will be returned and
	* that indicates we've exhausted our iteration of the
	* old_mapping.
	*/

	vdev_indirect_mapping_entry_phys_t *entry =
	vdev_indirect_mapping_entry_for_offset_or_next(old_mapping,
	max_offset);

	if (entry == NULL) {
	/*
	* We've already written the whole new mapping.
	* This special value will cause us to skip the
	* generate_new_mapping step and just do the sync
	* task to complete the condense.
	*/
	start_index = UINT64_MAX;
	} else {
	start_index = entry - old_mapping->vim_entries;
	ASSERT3U(start_index, <,
	vdev_indirect_mapping_num_entries(old_mapping));
	}
	}

	spa_condense_indirect_generate_new_mapping(vd, counts,
	start_index, zthr);

	vdev_indirect_mapping_free_obsolete_counts(old_mapping, counts);

	/*
	* If the zthr has received a cancellation signal while running
	* in generate_new_mapping() or at any point after that, then bail
	* early. We don't want to complete the condense if the spa is
	* shutting down.
	*/
	if (zthr_iscancelled(zthr))
	return;

	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	spa_condense_indirect_complete_sync, sci, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	/*
	* Sync task to begin the condensing process.
	*/
	void
	spa_condense_indirect_start_sync(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;

	ASSERT0(scip->scip_next_mapping_object);
	ASSERT0(scip->scip_prev_obsolete_sm_object);
	ASSERT0(scip->scip_vdev);
	ASSERT(dmu_tx_is_syncing(tx));
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_OBSOLETE_COUNTS));
	ASSERT(vdev_indirect_mapping_num_entries(vd->vdev_indirect_mapping));

	uint64_t obsolete_sm_obj;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_obj));
	ASSERT3U(obsolete_sm_obj, !=, 0);

	scip->scip_vdev = vd->vdev_id;
	scip->scip_next_mapping_object =
	vdev_indirect_mapping_alloc(spa->spa_meta_objset, tx);

	scip->scip_prev_obsolete_sm_object = obsolete_sm_obj;

	/*
	* We don't need to allocate a new space map object, since
	* vdev_indirect_sync_obsolete will allocate one when needed.
	*/
	space_map_close(vd->vdev_obsolete_sm);
	vd->vdev_obsolete_sm = NULL;
	VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, tx));

	VERIFY0(zap_add(spa->spa_dsl_pool->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t),
	sizeof (*scip) / sizeof (uint64_t), scip, tx));

	ASSERT3P(spa->spa_condensing_indirect, ==, NULL);
	spa->spa_condensing_indirect = spa_condensing_indirect_create(spa);

	zfs_dbgmsg("starting condense of vdev %llu in txg %llu: "
	"posm=%llu nm=%llu",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx),
	(u_longlong_t)scip->scip_prev_obsolete_sm_object,
	(u_longlong_t)scip->scip_next_mapping_object);

	zthr_wakeup(spa->spa_condense_zthr);
	}

	/*
	* Sync to the given vdev's obsolete space map any segments that are no longer
	* referenced as of the given txg.
	*
	* If the obsolete space map doesn't exist yet, create and open it.
	*/
	void
	vdev_indirect_sync_obsolete(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_indirect_config_t *vic __maybe_unused = &vd->vdev_indirect_config;

	ASSERT3U(vic->vic_mapping_object, !=, 0);
	ASSERT(range_tree_space(vd->vdev_obsolete_segments) > 0);
	ASSERT(vd->vdev_removing \|\| vd->vdev_ops == &vdev_indirect_ops);
	ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS));

	uint64_t obsolete_sm_object;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object == 0) {
	obsolete_sm_object = space_map_alloc(spa->spa_meta_objset,
	zfs_vdev_standard_sm_blksz, tx);

	ASSERT(vd->vdev_top_zap != 0);
	VERIFY0(zap_add(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM,
	sizeof (obsolete_sm_object), 1, &obsolete_sm_object, tx));
	ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	ASSERT3U(obsolete_sm_object, !=, 0);

	spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	VERIFY0(space_map_open(&vd->vdev_obsolete_sm,
	spa->spa_meta_objset, obsolete_sm_object,
	0, vd->vdev_asize, 0));
	}

	ASSERT(vd->vdev_obsolete_sm != NULL);
	ASSERT3U(obsolete_sm_object, ==,
	space_map_object(vd->vdev_obsolete_sm));

	space_map_write(vd->vdev_obsolete_sm,
	vd->vdev_obsolete_segments, SM_ALLOC, SM_NO_VDEVID, tx);
	range_tree_vacate(vd->vdev_obsolete_segments, NULL, NULL);
	}

	int
	spa_condense_init(spa_t *spa)
	{
	int error = zap_lookup(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t),
	sizeof (spa->spa_condensing_indirect_phys) / sizeof (uint64_t),
	&spa->spa_condensing_indirect_phys);
	if (error == 0) {
	if (spa_writeable(spa)) {
	spa->spa_condensing_indirect =
	spa_condensing_indirect_create(spa);
	}
	return (0);
	} else if (error == ENOENT) {
	return (0);
	} else {
	return (error);
	}
	}

	void
	spa_condense_fini(spa_t *spa)
	{
	if (spa->spa_condensing_indirect != NULL) {
	spa_condensing_indirect_destroy(spa->spa_condensing_indirect);
	spa->spa_condensing_indirect = NULL;
	}
	}

	void
	spa_start_indirect_condensing_thread(spa_t *spa)
	{
	ASSERT3P(spa->spa_condense_zthr, ==, NULL);
	spa->spa_condense_zthr = zthr_create("z_indirect_condense",
	spa_condense_indirect_thread_check,
	spa_condense_indirect_thread, spa, minclsyspri);
	}

	/*
	* Gets the obsolete spacemap object from the vdev's ZAP. On success sm_obj
	* will contain either the obsolete spacemap object or zero if none exists.
	* All other errors are returned to the caller.
	*/
	int
	vdev_obsolete_sm_object(vdev_t vd, uint64_t sm_obj)
	{
	ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_top_zap == 0) {
	*sm_obj = 0;
	return (0);
	}

	int error = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, sizeof (uint64_t), 1, sm_obj);
	if (error == ENOENT) {
	*sm_obj = 0;
	error = 0;
	}

	return (error);
	}

	/*
	* Gets the obsolete count are precise spacemap object from the vdev's ZAP.
	* On success are_precise will be set to reflect if the counts are precise.
	* All other errors are returned to the caller.
	*/
	int
	vdev_obsolete_counts_are_precise(vdev_t vd, boolean_t are_precise)
	{
	ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_top_zap == 0) {
	*are_precise = B_FALSE;
	return (0);
	}

	uint64_t val = 0;
	int error = zap_lookup(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, sizeof (val), 1, &val);
	if (error == 0) {
	*are_precise = (val != 0);
	} else if (error == ENOENT) {
	*are_precise = B_FALSE;
	error = 0;
	}

	return (error);
	}

	static void
	vdev_indirect_close(vdev_t *vd)
	{
	(void) vd;
	}

	static int
	vdev_indirect_open(vdev_t vd, uint64_t psize, uint64_t *max_psize,
	uint64_t logical_ashift, uint64_t physical_ashift)
	{
	psize = max_psize = vd->vdev_asize +
	VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE;
	*logical_ashift = vd->vdev_ashift;
	*physical_ashift = vd->vdev_physical_ashift;
	return (0);
	}

	typedef struct remap_segment {
	vdev_t *rs_vd;
	uint64_t rs_offset;
	uint64_t rs_asize;
	uint64_t rs_split_offset;
	list_node_t rs_node;
	} remap_segment_t;

	static remap_segment_t *
	rs_alloc(vdev_t *vd, uint64_t offset, uint64_t asize, uint64_t split_offset)
	{
	remap_segment_t *rs = kmem_alloc(sizeof (remap_segment_t), KM_SLEEP);
	rs->rs_vd = vd;
	rs->rs_offset = offset;
	rs->rs_asize = asize;
	rs->rs_split_offset = split_offset;
	return (rs);
	}

	/*
	* Given an indirect vdev and an extent on that vdev, it duplicates the
	* physical entries of the indirect mapping that correspond to the extent
	* to a new array and returns a pointer to it. In addition, copied_entries
	* is populated with the number of mapping entries that were duplicated.
	*
	* Note that the function assumes that the caller holds vdev_indirect_rwlock.
	* This ensures that the mapping won't change due to condensing as we
	* copy over its contents.
	*
	* Finally, since we are doing an allocation, it is up to the caller to
	* free the array allocated in this function.
	*/
	static vdev_indirect_mapping_entry_phys_t *
	vdev_indirect_mapping_duplicate_adjacent_entries(vdev_t *vd, uint64_t offset,
	uint64_t asize, uint64_t *copied_entries)
	{
	vdev_indirect_mapping_entry_phys_t *duplicate_mappings = NULL;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	uint64_t entries = 0;

	ASSERT(RW_READ_HELD(&vd->vdev_indirect_rwlock));

	vdev_indirect_mapping_entry_phys_t *first_mapping =
	vdev_indirect_mapping_entry_for_offset(vim, offset);
	ASSERT3P(first_mapping, !=, NULL);

	vdev_indirect_mapping_entry_phys_t *m = first_mapping;
	while (asize > 0) {
	uint64_t size = DVA_GET_ASIZE(&m->vimep_dst);

	ASSERT3U(offset, >=, DVA_MAPPING_GET_SRC_OFFSET(m));
	ASSERT3U(offset, <, DVA_MAPPING_GET_SRC_OFFSET(m) + size);

	uint64_t inner_offset = offset - DVA_MAPPING_GET_SRC_OFFSET(m);
	uint64_t inner_size = MIN(asize, size - inner_offset);

	offset += inner_size;
	asize -= inner_size;
	entries++;
	m++;
	}

	size_t copy_length = entries * sizeof (*first_mapping);
	duplicate_mappings = kmem_alloc(copy_length, KM_SLEEP);
	memcpy(duplicate_mappings, first_mapping, copy_length);
	*copied_entries = entries;

	return (duplicate_mappings);
	}

	/*
	* Goes through the relevant indirect mappings until it hits a concrete vdev
	* and issues the callback. On the way to the concrete vdev, if any other
	* indirect vdevs are encountered, then the callback will also be called on
	* each of those indirect vdevs. For example, if the segment is mapped to
	* segment A on indirect vdev 1, and then segment A on indirect vdev 1 is
	* mapped to segment B on concrete vdev 2, then the callback will be called on
	* both vdev 1 and vdev 2.
	*
	* While the callback passed to vdev_indirect_remap() is called on every vdev
	* the function encounters, certain callbacks only care about concrete vdevs.
	* These types of callbacks should return immediately and explicitly when they
	* are called on an indirect vdev.
	*
	* Because there is a possibility that a DVA section in the indirect device
	* has been split into multiple sections in our mapping, we keep track
	* of the relevant contiguous segments of the new location (remap_segment_t)
	* in a stack. This way we can call the callback for each of the new sections
	* created by a single section of the indirect device. Note though, that in
	* this scenario the callbacks in each split block won't occur in-order in
	* terms of offset, so callers should not make any assumptions about that.
	*
	* For callbacks that don't handle split blocks and immediately return when
	* they encounter them (as is the case for remap_blkptr_cb), the caller can
	* assume that its callback will be applied from the first indirect vdev
	* encountered to the last one and then the concrete vdev, in that order.
	*/
	static void
	vdev_indirect_remap(vdev_t *vd, uint64_t offset, uint64_t asize,
	void (func)(uint64_t, vdev_t , uint64_t, uint64_t, void ), void arg)
	{
	list_t stack;
	spa_t *spa = vd->vdev_spa;

	list_create(&stack, sizeof (remap_segment_t),
	offsetof(remap_segment_t, rs_node));

	for (remap_segment_t *rs = rs_alloc(vd, offset, asize, 0);
	rs != NULL; rs = list_remove_head(&stack)) {
	vdev_t *v = rs->rs_vd;
	uint64_t num_entries = 0;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
	ASSERT(rs->rs_asize > 0);

	/*
	* Note: As this function can be called from open context
	* (e.g. zio_read()), we need the following rwlock to
	* prevent the mapping from being changed by condensing.
	*
	* So we grab the lock and we make a copy of the entries
	* that are relevant to the extent that we are working on.
	* Once that is done, we drop the lock and iterate over
	* our copy of the mapping. Once we are done with the with
	* the remap segment and we free it, we also free our copy
	* of the indirect mapping entries that are relevant to it.
	*
	* This way we don't need to wait until the function is
	* finished with a segment, to condense it. In addition, we
	* don't need a recursive rwlock for the case that a call to
	* vdev_indirect_remap() needs to call itself (through the
	* codepath of its callback) for the same vdev in the middle
	* of its execution.
	*/
	rw_enter(&v->vdev_indirect_rwlock, RW_READER);
	ASSERT3P(v->vdev_indirect_mapping, !=, NULL);

	vdev_indirect_mapping_entry_phys_t *mapping =
	vdev_indirect_mapping_duplicate_adjacent_entries(v,
	rs->rs_offset, rs->rs_asize, &num_entries);
	ASSERT3P(mapping, !=, NULL);
	ASSERT3U(num_entries, >, 0);
	rw_exit(&v->vdev_indirect_rwlock);

	for (uint64_t i = 0; i < num_entries; i++) {
	/*
	* Note: the vdev_indirect_mapping can not change
	* while we are running. It only changes while the
	* removal is in progress, and then only from syncing
	* context. While a removal is in progress, this
	* function is only called for frees, which also only
	* happen from syncing context.
	*/
	vdev_indirect_mapping_entry_phys_t *m = &mapping[i];

	ASSERT3P(m, !=, NULL);
	ASSERT3U(rs->rs_asize, >, 0);

	uint64_t size = DVA_GET_ASIZE(&m->vimep_dst);
	uint64_t dst_offset = DVA_GET_OFFSET(&m->vimep_dst);
	uint64_t dst_vdev = DVA_GET_VDEV(&m->vimep_dst);

	ASSERT3U(rs->rs_offset, >=,
	DVA_MAPPING_GET_SRC_OFFSET(m));
	ASSERT3U(rs->rs_offset, <,
	DVA_MAPPING_GET_SRC_OFFSET(m) + size);
	ASSERT3U(dst_vdev, !=, v->vdev_id);

	uint64_t inner_offset = rs->rs_offset -
	DVA_MAPPING_GET_SRC_OFFSET(m);
	uint64_t inner_size =
	MIN(rs->rs_asize, size - inner_offset);

	vdev_t *dst_v = vdev_lookup_top(spa, dst_vdev);
	ASSERT3P(dst_v, !=, NULL);

	if (dst_v->vdev_ops == &vdev_indirect_ops) {
	list_insert_head(&stack,
	rs_alloc(dst_v, dst_offset + inner_offset,
	inner_size, rs->rs_split_offset));

	}

	if ((zfs_flags & ZFS_DEBUG_INDIRECT_REMAP) &&
	IS_P2ALIGNED(inner_size, 2 * SPA_MINBLOCKSIZE)) {
	/*
	* Note: This clause exists only solely for
	* testing purposes. We use it to ensure that
	* split blocks work and that the callbacks
	* using them yield the same result if issued
	* in reverse order.
	*/
	uint64_t inner_half = inner_size / 2;

	func(rs->rs_split_offset + inner_half, dst_v,
	dst_offset + inner_offset + inner_half,
	inner_half, arg);

	func(rs->rs_split_offset, dst_v,
	dst_offset + inner_offset,
	inner_half, arg);
	} else {
	func(rs->rs_split_offset, dst_v,
	dst_offset + inner_offset,
	inner_size, arg);
	}

	rs->rs_offset += inner_size;
	rs->rs_asize -= inner_size;
	rs->rs_split_offset += inner_size;
	}
	VERIFY0(rs->rs_asize);

	kmem_free(mapping, num_entries * sizeof (*mapping));
	kmem_free(rs, sizeof (remap_segment_t));
	}
	list_destroy(&stack);
	}

	static void
	vdev_indirect_child_io_done(zio_t *zio)
	{
	zio_t *pio = zio->io_private;

	mutex_enter(&pio->io_lock);
	pio->io_error = zio_worst_error(pio->io_error, zio->io_error);
	mutex_exit(&pio->io_lock);

	abd_free(zio->io_abd);
	}

	/*
	* This is a callback for vdev_indirect_remap() which allocates an
	* indirect_split_t for each split segment and adds it to iv_splits.
	*/
	static void
	vdev_indirect_gather_splits(uint64_t split_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	zio_t *zio = arg;
	indirect_vsd_t *iv = zio->io_vsd;

	ASSERT3P(vd, !=, NULL);

	if (vd->vdev_ops == &vdev_indirect_ops)
	return;

	int n = 1;
	if (vd->vdev_ops == &vdev_mirror_ops)
	n = vd->vdev_children;

	indirect_split_t *is =
	kmem_zalloc(offsetof(indirect_split_t, is_child[n]), KM_SLEEP);

	is->is_children = n;
	is->is_size = size;
	is->is_split_offset = split_offset;
	is->is_target_offset = offset;
	is->is_vdev = vd;
	list_create(&is->is_unique_child, sizeof (indirect_child_t),
	offsetof(indirect_child_t, ic_node));

	/*
	* Note that we only consider multiple copies of the data for
	* mirror vdevs. We don't for "replacing" or "spare" vdevs, even
	* though they use the same ops as mirror, because there's only one
	* "good" copy under the replacing/spare.
	*/
	if (vd->vdev_ops == &vdev_mirror_ops) {
	for (int i = 0; i < n; i++) {
	is->is_child[i].ic_vdev = vd->vdev_child[i];
	list_link_init(&is->is_child[i].ic_node);
	}
	} else {
	is->is_child[0].ic_vdev = vd;
	}

	list_insert_tail(&iv->iv_splits, is);
	}

	static void
	vdev_indirect_read_split_done(zio_t *zio)
	{
	indirect_child_t *ic = zio->io_private;

	if (zio->io_error != 0) {
	/*
	* Clear ic_data to indicate that we do not have data for this
	* child.
	*/
	abd_free(ic->ic_data);
	ic->ic_data = NULL;
	}
	}

	/*
	* Issue reads for all copies (mirror children) of all splits.
	*/
	static void
	vdev_indirect_read_all(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	for (int i = 0; i < is->is_children; i++) {
	indirect_child_t *ic = &is->is_child[i];

	if (!vdev_readable(ic->ic_vdev))
	continue;

	/*
	* If a child is missing the data, set ic_error. Used
	* in vdev_indirect_repair(). We perform the read
	* nevertheless which provides the opportunity to
	* reconstruct the split block if at all possible.
	*/
	if (vdev_dtl_contains(ic->ic_vdev, DTL_MISSING,
	zio->io_txg, 1))
	ic->ic_error = SET_ERROR(ESTALE);

	ic->ic_data = abd_alloc_sametype(zio->io_abd,
	is->is_size);
	ic->ic_duplicate = NULL;

	zio_nowait(zio_vdev_child_io(zio, NULL,
	ic->ic_vdev, is->is_target_offset, ic->ic_data,
	is->is_size, zio->io_type, zio->io_priority, 0,
	vdev_indirect_read_split_done, ic));
	}
	}
	iv->iv_reconstruct = B_TRUE;
	}

	static void
	vdev_indirect_io_start(zio_t *zio)
	{
	spa_t *spa __maybe_unused = zio->io_spa;
	indirect_vsd_t iv = kmem_zalloc(sizeof (iv), KM_SLEEP);
	list_create(&iv->iv_splits,
	sizeof (indirect_split_t), offsetof(indirect_split_t, is_node));

	zio->io_vsd = iv;
	zio->io_vsd_ops = &vdev_indirect_vsd_ops;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
	if (zio->io_type != ZIO_TYPE_READ) {
	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
	/*
	* Note: this code can handle other kinds of writes,
	* but we don't expect them.
	*/
	ASSERT((zio->io_flags & (ZIO_FLAG_SELF_HEAL \|
	ZIO_FLAG_RESILVER \| ZIO_FLAG_INDUCE_DAMAGE)) != 0);
	}

	vdev_indirect_remap(zio->io_vd, zio->io_offset, zio->io_size,
	vdev_indirect_gather_splits, zio);

	indirect_split_t *first = list_head(&iv->iv_splits);
	if (first->is_size == zio->io_size) {
	/*
	* This is not a split block; we are pointing to the entire
	* data, which will checksum the same as the original data.
	* Pass the BP down so that the child i/o can verify the
	* checksum, and try a different location if available
	* (e.g. on a mirror).
	*
	* While this special case could be handled the same as the
	* general (split block) case, doing it this way ensures
	* that the vast majority of blocks on indirect vdevs
	* (which are not split) are handled identically to blocks
	* on non-indirect vdevs. This allows us to be less strict
	* about performance in the general (but rare) case.
	*/
	ASSERT0(first->is_split_offset);
	ASSERT3P(list_next(&iv->iv_splits, first), ==, NULL);
	zio_nowait(zio_vdev_child_io(zio, zio->io_bp,
	first->is_vdev, first->is_target_offset,
	abd_get_offset(zio->io_abd, 0),
	zio->io_size, zio->io_type, zio->io_priority, 0,
	vdev_indirect_child_io_done, zio));
	} else {
	iv->iv_split_block = B_TRUE;
	if (zio->io_type == ZIO_TYPE_READ &&
	zio->io_flags & (ZIO_FLAG_SCRUB \| ZIO_FLAG_RESILVER)) {
	/*
	* Read all copies. Note that for simplicity,
	* we don't bother consulting the DTL in the
	* resilver case.
	*/
	vdev_indirect_read_all(zio);
	} else {
	/*
	* If this is a read zio, we read one copy of each
	* split segment, from the top-level vdev. Since
	* we don't know the checksum of each split
	* individually, the child zio can't ensure that
	* we get the right data. E.g. if it's a mirror,
	* it will just read from a random (healthy) leaf
	* vdev. We have to verify the checksum in
	* vdev_indirect_io_done().
	*
	* For write zios, the vdev code will ensure we write
	* to all children.
	*/
	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	zio_nowait(zio_vdev_child_io(zio, NULL,
	is->is_vdev, is->is_target_offset,
	abd_get_offset(zio->io_abd,
	is->is_split_offset), is->is_size,
	zio->io_type, zio->io_priority, 0,
	vdev_indirect_child_io_done, zio));
	}

	}
	}

	zio_execute(zio);
	}

	/*
	* Report a checksum error for a child.
	*/
	static void
	vdev_indirect_checksum_error(zio_t *zio,
	indirect_split_t is, indirect_child_t ic)
	{
	vdev_t *vd = ic->ic_vdev;

	if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
	return;

	mutex_enter(&vd->vdev_stat_lock);
	vd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&vd->vdev_stat_lock);

	zio_bad_cksum_t zbc = {{{ 0 }}};
	abd_t *bad_abd = ic->ic_data;
	abd_t *good_abd = is->is_good_child->ic_data;
	(void) zfs_ereport_post_checksum(zio->io_spa, vd, NULL, zio,
	is->is_target_offset, is->is_size, good_abd, bad_abd, &zbc);
	}

	/*
	* Issue repair i/os for any incorrect copies. We do this by comparing
	* each split segment's correct data (is_good_child's ic_data) with each
	* other copy of the data. If they differ, then we overwrite the bad data
	* with the good copy. The DTL is checked in vdev_indirect_read_all() and
	* if a vdev is missing a copy of the data we set ic_error and the read is
	* performed. This provides the opportunity to reconstruct the split block
	* if at all possible. ic_error is checked here and if set it suppresses
	* incrementing the checksum counter. Aside from this DTLs are not checked,
	* which simplifies this code and also issues the optimal number of writes
	* (based on which copies actually read bad data, as opposed to which we
	* think might be wrong). For the same reason, we always use
	* ZIO_FLAG_SELF_HEAL, to bypass the DTL check in zio_vdev_io_start().
	*/
	static void
	vdev_indirect_repair(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;

	if (!spa_writeable(zio->io_spa))
	return;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	for (int c = 0; c < is->is_children; c++) {
	indirect_child_t *ic = &is->is_child[c];
	if (ic == is->is_good_child)
	continue;
	if (ic->ic_data == NULL)
	continue;
	if (ic->ic_duplicate == is->is_good_child)
	continue;

	zio_nowait(zio_vdev_child_io(zio, NULL,
	ic->ic_vdev, is->is_target_offset,
	is->is_good_child->ic_data, is->is_size,
	ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_IO_REPAIR \| ZIO_FLAG_SELF_HEAL,
	NULL, NULL));

	/*
	* If ic_error is set the current child does not have
	* a copy of the data, so suppress incrementing the
	* checksum counter.
	*/
	if (ic->ic_error == ESTALE)
	continue;

	vdev_indirect_checksum_error(zio, is, ic);
	}
	}
	}

	/*
	* Report checksum errors on all children that we read from.
	*/
	static void
	vdev_indirect_all_checksum_errors(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;

	if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
	return;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	for (int c = 0; c < is->is_children; c++) {
	indirect_child_t *ic = &is->is_child[c];

	if (ic->ic_data == NULL)
	continue;

	vdev_t *vd = ic->ic_vdev;

	(void) zfs_ereport_post_checksum(zio->io_spa, vd,
	NULL, zio, is->is_target_offset, is->is_size,
	NULL, NULL, NULL);
	mutex_enter(&vd->vdev_stat_lock);
	vd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&vd->vdev_stat_lock);
	}
	}
	}

	/*
	* Copy data from all the splits to a main zio then validate the checksum.
	* If then checksum is successfully validated return success.
	*/
	static int
	vdev_indirect_splits_checksum_validate(indirect_vsd_t iv, zio_t zio)
	{
	zio_bad_cksum_t zbc;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {

	ASSERT3P(is->is_good_child->ic_data, !=, NULL);
	ASSERT3P(is->is_good_child->ic_duplicate, ==, NULL);

	abd_copy_off(zio->io_abd, is->is_good_child->ic_data,
	is->is_split_offset, 0, is->is_size);
	}

	return (zio_checksum_error(zio, &zbc));
	}

	/*
	* There are relatively few possible combinations making it feasible to
	* deterministically check them all. We do this by setting the good_child
	* to the next unique split version. If we reach the end of the list then
	* "carry over" to the next unique split version (like counting in base
	* is_unique_children, but each digit can have a different base).
	*/
	static int
	vdev_indirect_splits_enumerate_all(indirect_vsd_t iv, zio_t zio)
	{
	boolean_t more = B_TRUE;

	iv->iv_attempts = 0;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is))
	is->is_good_child = list_head(&is->is_unique_child);

	while (more == B_TRUE) {
	iv->iv_attempts++;
	more = B_FALSE;

	if (vdev_indirect_splits_checksum_validate(iv, zio) == 0)
	return (0);

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	is->is_good_child = list_next(&is->is_unique_child,
	is->is_good_child);
	if (is->is_good_child != NULL) {
	more = B_TRUE;
	break;
	}

	is->is_good_child = list_head(&is->is_unique_child);
	}
	}

	ASSERT3S(iv->iv_attempts, <=, iv->iv_unique_combinations);

	return (SET_ERROR(ECKSUM));
	}

	/*
	* There are too many combinations to try all of them in a reasonable amount
	* of time. So try a fixed number of random combinations from the unique
	* split versions, after which we'll consider the block unrecoverable.
	*/
	static int
	vdev_indirect_splits_enumerate_randomly(indirect_vsd_t iv, zio_t zio)
	{
	iv->iv_attempts = 0;

	while (iv->iv_attempts < iv->iv_attempts_max) {
	iv->iv_attempts++;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	indirect_child_t *ic = list_head(&is->is_unique_child);
	int children = is->is_unique_children;

	for (int i = random_in_range(children); i > 0; i--)
	ic = list_next(&is->is_unique_child, ic);

	ASSERT3P(ic, !=, NULL);
	is->is_good_child = ic;
	}

	if (vdev_indirect_splits_checksum_validate(iv, zio) == 0)
	return (0);
	}

	return (SET_ERROR(ECKSUM));
	}

	/*
	* This is a validation function for reconstruction. It randomly selects
	* a good combination, if one can be found, and then it intentionally
	* damages all other segment copes by zeroing them. This forces the
	* reconstruction algorithm to locate the one remaining known good copy.
	*/
	static int
	vdev_indirect_splits_damage(indirect_vsd_t iv, zio_t zio)
	{
	int error;

	/* Presume all the copies are unique for initial selection. */
	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	is->is_unique_children = 0;

	for (int i = 0; i < is->is_children; i++) {
	indirect_child_t *ic = &is->is_child[i];
	if (ic->ic_data != NULL) {
	is->is_unique_children++;
	list_insert_tail(&is->is_unique_child, ic);
	}
	}

	if (list_is_empty(&is->is_unique_child)) {
	error = SET_ERROR(EIO);
	goto out;
	}
	}

	/*
	* Set each is_good_child to a randomly-selected child which
	* is known to contain validated data.
	*/
	error = vdev_indirect_splits_enumerate_randomly(iv, zio);
	if (error)
	goto out;

	/*
	* Damage all but the known good copy by zeroing it. This will
	* result in two or less unique copies per indirect_child_t.
	* Both may need to be checked in order to reconstruct the block.
	* Set iv->iv_attempts_max such that all unique combinations will
	* enumerated, but limit the damage to at most 12 indirect splits.
	*/
	iv->iv_attempts_max = 1;

	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	for (int c = 0; c < is->is_children; c++) {
	indirect_child_t *ic = &is->is_child[c];

	if (ic == is->is_good_child)
	continue;
	if (ic->ic_data == NULL)
	continue;

	abd_zero(ic->ic_data, abd_get_size(ic->ic_data));
	}

	iv->iv_attempts_max *= 2;
	if (iv->iv_attempts_max >= (1ULL << 12)) {
	iv->iv_attempts_max = UINT64_MAX;
	break;
	}
	}

	out:
	/* Empty the unique children lists so they can be reconstructed. */
	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	indirect_child_t *ic;
	while ((ic = list_head(&is->is_unique_child)) != NULL)
	list_remove(&is->is_unique_child, ic);

	is->is_unique_children = 0;
	}

	return (error);
	}

	/*
	* This function is called when we have read all copies of the data and need
	* to try to find a combination of copies that gives us the right checksum.
	*
	* If we pointed to any mirror vdevs, this effectively does the job of the
	* mirror. The mirror vdev code can't do its own job because we don't know
	* the checksum of each split segment individually.
	*
	* We have to try every unique combination of copies of split segments, until
	* we find one that checksums correctly. Duplicate segment copies are first
	* identified and latter skipped during reconstruction. This optimization
	* reduces the search space and ensures that of the remaining combinations
	* at most one is correct.
	*
	* When the total number of combinations is small they can all be checked.
	* For example, if we have 3 segments in the split, and each points to a
	* 2-way mirror with unique copies, we will have the following pieces of data:
	*
	* \| mirror child
	* split \| [0] [1]
	* ======\|=====================
	* A \| data_A_0 data_A_1
	* B \| data_B_0 data_B_1
	* C \| data_C_0 data_C_1
	*
	* We will try the following (mirror children)^(number of splits) (2^3=8)
	* combinations, which is similar to bitwise-little-endian counting in
	* binary. In general each "digit" corresponds to a split segment, and the
	* base of each digit is is_children, which can be different for each
	* digit.
	*
	* "low bit" "high bit"
	* v v
	* data_A_0 data_B_0 data_C_0
	* data_A_1 data_B_0 data_C_0
	* data_A_0 data_B_1 data_C_0
	* data_A_1 data_B_1 data_C_0
	* data_A_0 data_B_0 data_C_1
	* data_A_1 data_B_0 data_C_1
	* data_A_0 data_B_1 data_C_1
	* data_A_1 data_B_1 data_C_1
	*
	* Note that the split segments may be on the same or different top-level
	* vdevs. In either case, we may need to try lots of combinations (see
	* zfs_reconstruct_indirect_combinations_max). This ensures that if a mirror
	* has small silent errors on all of its children, we can still reconstruct
	* the correct data, as long as those errors are at sufficiently-separated
	* offsets (specifically, separated by the largest block size - default of
	* 128KB, but up to 16MB).
	*/
	static void
	vdev_indirect_reconstruct_io_done(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;
	boolean_t known_good = B_FALSE;
	int error;

	iv->iv_unique_combinations = 1;
	iv->iv_attempts_max = UINT64_MAX;

	if (zfs_reconstruct_indirect_combinations_max > 0)
	iv->iv_attempts_max = zfs_reconstruct_indirect_combinations_max;

	/*
	* If nonzero, every 1/x blocks will be damaged, in order to validate
	* reconstruction when there are split segments with damaged copies.
	* Known_good will be TRUE when reconstruction is known to be possible.
	*/
	if (zfs_reconstruct_indirect_damage_fraction != 0 &&
	random_in_range(zfs_reconstruct_indirect_damage_fraction) == 0)
	known_good = (vdev_indirect_splits_damage(iv, zio) == 0);

	/*
	* Determine the unique children for a split segment and add them
	* to the is_unique_child list. By restricting reconstruction
	* to these children, only unique combinations will be considered.
	* This can vastly reduce the search space when there are a large
	* number of indirect splits.
	*/
	for (indirect_split_t *is = list_head(&iv->iv_splits);
	is != NULL; is = list_next(&iv->iv_splits, is)) {
	is->is_unique_children = 0;

	for (int i = 0; i < is->is_children; i++) {
	indirect_child_t *ic_i = &is->is_child[i];

	if (ic_i->ic_data == NULL \|\|
	ic_i->ic_duplicate != NULL)
	continue;

	for (int j = i + 1; j < is->is_children; j++) {
	indirect_child_t *ic_j = &is->is_child[j];

	if (ic_j->ic_data == NULL \|\|
	ic_j->ic_duplicate != NULL)
	continue;

	if (abd_cmp(ic_i->ic_data, ic_j->ic_data) == 0)
	ic_j->ic_duplicate = ic_i;
	}

	is->is_unique_children++;
	list_insert_tail(&is->is_unique_child, ic_i);
	}

	/* Reconstruction is impossible, no valid children */
	EQUIV(list_is_empty(&is->is_unique_child),
	is->is_unique_children == 0);
	if (list_is_empty(&is->is_unique_child)) {
	zio->io_error = EIO;
	vdev_indirect_all_checksum_errors(zio);
	zio_checksum_verified(zio);
	return;
	}

	iv->iv_unique_combinations *= is->is_unique_children;
	}

	if (iv->iv_unique_combinations <= iv->iv_attempts_max)
	error = vdev_indirect_splits_enumerate_all(iv, zio);
	else
	error = vdev_indirect_splits_enumerate_randomly(iv, zio);

	if (error != 0) {
	/* All attempted combinations failed. */
	ASSERT3B(known_good, ==, B_FALSE);
	zio->io_error = error;
	vdev_indirect_all_checksum_errors(zio);
	} else {
	/*
	* The checksum has been successfully validated. Issue
	* repair I/Os to any copies of splits which don't match
	* the validated version.
	*/
	ASSERT0(vdev_indirect_splits_checksum_validate(iv, zio));
	vdev_indirect_repair(zio);
	zio_checksum_verified(zio);
	}
	}

	static void
	vdev_indirect_io_done(zio_t *zio)
	{
	indirect_vsd_t *iv = zio->io_vsd;

	if (iv->iv_reconstruct) {
	/*
	* We have read all copies of the data (e.g. from mirrors),
	* either because this was a scrub/resilver, or because the
	* one-copy read didn't checksum correctly.
	*/
	vdev_indirect_reconstruct_io_done(zio);
	return;
	}

	if (!iv->iv_split_block) {
	/*
	* This was not a split block, so we passed the BP down,
	* and the checksum was handled by the (one) child zio.
	*/
	return;
	}

	zio_bad_cksum_t zbc;
	int ret = zio_checksum_error(zio, &zbc);
	if (ret == 0) {
	zio_checksum_verified(zio);
	return;
	}

	/*
	* The checksum didn't match. Read all copies of all splits, and
	* then we will try to reconstruct. The next time
	* vdev_indirect_io_done() is called, iv_reconstruct will be set.
	*/
	vdev_indirect_read_all(zio);

	zio_vdev_io_redone(zio);
	}

	vdev_ops_t vdev_indirect_ops = {
	.vdev_op_init = NULL,
	.vdev_op_fini = NULL,
	.vdev_op_open = vdev_indirect_open,
	.vdev_op_close = vdev_indirect_close,
	.vdev_op_asize = vdev_default_asize,
	.vdev_op_min_asize = vdev_default_min_asize,
	.vdev_op_min_alloc = NULL,
	.vdev_op_io_start = vdev_indirect_io_start,
	.vdev_op_io_done = vdev_indirect_io_done,
	.vdev_op_state_change = NULL,
	.vdev_op_need_resilver = NULL,
	.vdev_op_hold = NULL,
	.vdev_op_rele = NULL,
	.vdev_op_remap = vdev_indirect_remap,
	.vdev_op_xlate = NULL,
	.vdev_op_rebuild_asize = NULL,
	.vdev_op_metaslab_init = NULL,
	.vdev_op_config_generate = NULL,
	.vdev_op_nparity = NULL,
	.vdev_op_ndisks = NULL,
	.vdev_op_type = VDEV_TYPE_INDIRECT, /* name of this vdev type */
	.vdev_op_leaf = B_FALSE /* leaf vdev */
	};

	EXPORT_SYMBOL(spa_condense_fini);
	EXPORT_SYMBOL(spa_start_indirect_condensing_thread);
	EXPORT_SYMBOL(spa_condense_indirect_start_sync);
	EXPORT_SYMBOL(spa_condense_init);
	EXPORT_SYMBOL(spa_vdev_indirect_mark_obsolete);
	EXPORT_SYMBOL(vdev_indirect_mark_obsolete);
	EXPORT_SYMBOL(vdev_indirect_should_condense);
	EXPORT_SYMBOL(vdev_indirect_sync_obsolete);
	EXPORT_SYMBOL(vdev_obsolete_counts_are_precise);
	EXPORT_SYMBOL(vdev_obsolete_sm_object);

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_vdevs_enable, INT,
	ZMOD_RW, "Whether to attempt condensing indirect vdev mappings");

	-ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_obsolete_pct, INT,
	+ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_obsolete_pct, UINT,
	ZMOD_RW,
	"Minimum obsolete percent of bytes in the mapping "
	"to attempt condensing");

	ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, min_mapping_bytes, ULONG, ZMOD_RW,
	"Don't bother condensing if the mapping uses less than this amount of "
	"memory");

	ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, max_obsolete_bytes, ULONG,
	ZMOD_RW,
	"Minimum size obsolete spacemap to attempt condensing");

	ZFS_MODULE_PARAM(zfs_condense, zfs_condense_, indirect_commit_entry_delay_ms,
	- INT, ZMOD_RW,
	+ UINT, ZMOD_RW,
	"Used by tests to ensure certain actions happen in the middle of a "
	"condense. A maximum value of 1 should be sufficient.");

	ZFS_MODULE_PARAM(zfs_reconstruct, zfs_reconstruct_, indirect_combinations_max,
	- INT, ZMOD_RW,
	+ UINT, ZMOD_RW,
	"Maximum number of combinations when reconstructing split segments");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_queue.c b/sys/contrib/openzfs/module/zfs/vdev_queue.c
	index 9a805f2c3181..1acb89cea393 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_queue.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_queue.c
	@@ -1,1119 +1,1119 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/vdev_impl.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/avl.h>
	#include <sys/dsl_pool.h>
	#include <sys/metaslab_impl.h>
	#include <sys/spa.h>
	#include <sys/abd.h>

	/*
	* ZFS I/O Scheduler
	* ---------------
	*
	* ZFS issues I/O operations to leaf vdevs to satisfy and complete zios. The
	* I/O scheduler determines when and in what order those operations are
	* issued. The I/O scheduler divides operations into five I/O classes
	* prioritized in the following order: sync read, sync write, async read,
	* async write, and scrub/resilver. Each queue defines the minimum and
	* maximum number of concurrent operations that may be issued to the device.
	* In addition, the device has an aggregate maximum. Note that the sum of the
	* per-queue minimums must not exceed the aggregate maximum. If the
	* sum of the per-queue maximums exceeds the aggregate maximum, then the
	* number of active i/os may reach zfs_vdev_max_active, in which case no
	* further i/os will be issued regardless of whether all per-queue
	* minimums have been met.
	*
	* For many physical devices, throughput increases with the number of
	* concurrent operations, but latency typically suffers. Further, physical
	* devices typically have a limit at which more concurrent operations have no
	* effect on throughput or can actually cause it to decrease.
	*
	* The scheduler selects the next operation to issue by first looking for an
	* I/O class whose minimum has not been satisfied. Once all are satisfied and
	* the aggregate maximum has not been hit, the scheduler looks for classes
	* whose maximum has not been satisfied. Iteration through the I/O classes is
	* done in the order specified above. No further operations are issued if the
	* aggregate maximum number of concurrent operations has been hit or if there
	* are no operations queued for an I/O class that has not hit its maximum.
	* Every time an i/o is queued or an operation completes, the I/O scheduler
	* looks for new operations to issue.
	*
	* All I/O classes have a fixed maximum number of outstanding operations
	* except for the async write class. Asynchronous writes represent the data
	* that is committed to stable storage during the syncing stage for
	* transaction groups (see txg.c). Transaction groups enter the syncing state
	* periodically so the number of queued async writes will quickly burst up and
	* then bleed down to zero. Rather than servicing them as quickly as possible,
	* the I/O scheduler changes the maximum number of active async write i/os
	* according to the amount of dirty data in the pool (see dsl_pool.c). Since
	* both throughput and latency typically increase with the number of
	* concurrent operations issued to physical devices, reducing the burstiness
	* in the number of concurrent operations also stabilizes the response time of
	* operations from other -- and in particular synchronous -- queues. In broad
	* strokes, the I/O scheduler will issue more concurrent operations from the
	* async write queue as there's more dirty data in the pool.
	*
	* Async Writes
	*
	* The number of concurrent operations issued for the async write I/O class
	* follows a piece-wise linear function defined by a few adjustable points.
	*
	* \| o---------\| <-- zfs_vdev_async_write_max_active
	* ^ \| /^ \|
	* \| \| / \| \|
	* active \| / \| \|
	* I/O \| / \| \|
	* count \| / \| \|
	* \| / \| \|
	* \|------------o \| \| <-- zfs_vdev_async_write_min_active
	* 0\|____________^______\|_________\|
	* 0% \| \| 100% of zfs_dirty_data_max
	* \| \|
	* \| `-- zfs_vdev_async_write_active_max_dirty_percent
	* `--------- zfs_vdev_async_write_active_min_dirty_percent
	*
	* Until the amount of dirty data exceeds a minimum percentage of the dirty
	* data allowed in the pool, the I/O scheduler will limit the number of
	* concurrent operations to the minimum. As that threshold is crossed, the
	* number of concurrent operations issued increases linearly to the maximum at
	* the specified maximum percentage of the dirty data allowed in the pool.
	*
	* Ideally, the amount of dirty data on a busy pool will stay in the sloped
	* part of the function between zfs_vdev_async_write_active_min_dirty_percent
	* and zfs_vdev_async_write_active_max_dirty_percent. If it exceeds the
	* maximum percentage, this indicates that the rate of incoming data is
	* greater than the rate that the backend storage can handle. In this case, we
	* must further throttle incoming writes (see dmu_tx_delay() for details).
	*/

	/*
	* The maximum number of i/os active to each device. Ideally, this will be >=
	* the sum of each queue's max_active.
	*/
	-uint32_t zfs_vdev_max_active = 1000;
	+uint_t zfs_vdev_max_active = 1000;

	/*
	* Per-queue limits on the number of i/os active to each device. If the
	* number of active i/os is < zfs_vdev_max_active, then the min_active comes
	* into play. We will send min_active from each queue round-robin, and then
	* send from queues in the order defined by zio_priority_t up to max_active.
	* Some queues have additional mechanisms to limit number of active I/Os in
	* addition to min_active and max_active, see below.
	*
	* In general, smaller max_active's will lead to lower latency of synchronous
	* operations. Larger max_active's may lead to higher overall throughput,
	* depending on underlying storage.
	*
	* The ratio of the queues' max_actives determines the balance of performance
	* between reads, writes, and scrubs. E.g., increasing
	* zfs_vdev_scrub_max_active will cause the scrub or resilver to complete
	* more quickly, but reads and writes to have higher latency and lower
	* throughput.
	*/
	-static uint32_t zfs_vdev_sync_read_min_active = 10;
	-static uint32_t zfs_vdev_sync_read_max_active = 10;
	-static uint32_t zfs_vdev_sync_write_min_active = 10;
	-static uint32_t zfs_vdev_sync_write_max_active = 10;
	-static uint32_t zfs_vdev_async_read_min_active = 1;
	-/* */ uint32_t zfs_vdev_async_read_max_active = 3;
	-static uint32_t zfs_vdev_async_write_min_active = 2;
	-/* */ uint32_t zfs_vdev_async_write_max_active = 10;
	-static uint32_t zfs_vdev_scrub_min_active = 1;
	-static uint32_t zfs_vdev_scrub_max_active = 3;
	-static uint32_t zfs_vdev_removal_min_active = 1;
	-static uint32_t zfs_vdev_removal_max_active = 2;
	-static uint32_t zfs_vdev_initializing_min_active = 1;
	-static uint32_t zfs_vdev_initializing_max_active = 1;
	-static uint32_t zfs_vdev_trim_min_active = 1;
	-static uint32_t zfs_vdev_trim_max_active = 2;
	-static uint32_t zfs_vdev_rebuild_min_active = 1;
	-static uint32_t zfs_vdev_rebuild_max_active = 3;
	+static uint_t zfs_vdev_sync_read_min_active = 10;
	+static uint_t zfs_vdev_sync_read_max_active = 10;
	+static uint_t zfs_vdev_sync_write_min_active = 10;
	+static uint_t zfs_vdev_sync_write_max_active = 10;
	+static uint_t zfs_vdev_async_read_min_active = 1;
	+/* */ uint_t zfs_vdev_async_read_max_active = 3;
	+static uint_t zfs_vdev_async_write_min_active = 2;
	+/* */ uint_t zfs_vdev_async_write_max_active = 10;
	+static uint_t zfs_vdev_scrub_min_active = 1;
	+static uint_t zfs_vdev_scrub_max_active = 3;
	+static uint_t zfs_vdev_removal_min_active = 1;
	+static uint_t zfs_vdev_removal_max_active = 2;
	+static uint_t zfs_vdev_initializing_min_active = 1;
	+static uint_t zfs_vdev_initializing_max_active = 1;
	+static uint_t zfs_vdev_trim_min_active = 1;
	+static uint_t zfs_vdev_trim_max_active = 2;
	+static uint_t zfs_vdev_rebuild_min_active = 1;
	+static uint_t zfs_vdev_rebuild_max_active = 3;

	/*
	* When the pool has less than zfs_vdev_async_write_active_min_dirty_percent
	* dirty data, use zfs_vdev_async_write_min_active. When it has more than
	* zfs_vdev_async_write_active_max_dirty_percent, use
	* zfs_vdev_async_write_max_active. The value is linearly interpolated
	* between min and max.
	*/
	-int zfs_vdev_async_write_active_min_dirty_percent = 30;
	-int zfs_vdev_async_write_active_max_dirty_percent = 60;
	+uint_t zfs_vdev_async_write_active_min_dirty_percent = 30;
	+uint_t zfs_vdev_async_write_active_max_dirty_percent = 60;

	/*
	* For non-interactive I/O (scrub, resilver, removal, initialize and rebuild),
	* the number of concurrently-active I/O's is limited to *_min_active, unless
	* the vdev is "idle". When there are no interactive I/Os active (sync or
	* async), and zfs_vdev_nia_delay I/Os have completed since the last
	* interactive I/O, then the vdev is considered to be "idle", and the number
	* of concurrently-active non-interactive I/O's is increased to *_max_active.
	*/
	static uint_t zfs_vdev_nia_delay = 5;

	/*
	* Some HDDs tend to prioritize sequential I/O so high that concurrent
	* random I/O latency reaches several seconds. On some HDDs it happens
	* even if sequential I/Os are submitted one at a time, and so setting
	* *_max_active to 1 does not help. To prevent non-interactive I/Os, like
	* scrub, from monopolizing the device no more than zfs_vdev_nia_credit
	* I/Os can be sent while there are outstanding incomplete interactive
	* I/Os. This enforced wait ensures the HDD services the interactive I/O
	* within a reasonable amount of time.
	*/
	static uint_t zfs_vdev_nia_credit = 5;

	/*
	* To reduce IOPs, we aggregate small adjacent I/Os into one large I/O.
	* For read I/Os, we also aggregate across small adjacency gaps; for writes
	* we include spans of optional I/Os to aid aggregation at the disk even when
	* they aren't able to help us aggregate at this level.
	*/
	-static int zfs_vdev_aggregation_limit = 1 << 20;
	-static int zfs_vdev_aggregation_limit_non_rotating = SPA_OLD_MAXBLOCKSIZE;
	-static int zfs_vdev_read_gap_limit = 32 << 10;
	-static int zfs_vdev_write_gap_limit = 4 << 10;
	+static uint_t zfs_vdev_aggregation_limit = 1 << 20;
	+static uint_t zfs_vdev_aggregation_limit_non_rotating = SPA_OLD_MAXBLOCKSIZE;
	+static uint_t zfs_vdev_read_gap_limit = 32 << 10;
	+static uint_t zfs_vdev_write_gap_limit = 4 << 10;

	/*
	* Define the queue depth percentage for each top-level. This percentage is
	* used in conjunction with zfs_vdev_async_max_active to determine how many
	* allocations a specific top-level vdev should handle. Once the queue depth
	* reaches zfs_vdev_queue_depth_pct * zfs_vdev_async_write_max_active / 100
	* then allocator will stop allocating blocks on that top-level device.
	* The default kernel setting is 1000% which will yield 100 allocations per
	* device. For userland testing, the default setting is 300% which equates
	* to 30 allocations per device.
	*/
	#ifdef _KERNEL
	-int zfs_vdev_queue_depth_pct = 1000;
	+uint_t zfs_vdev_queue_depth_pct = 1000;
	#else
	-int zfs_vdev_queue_depth_pct = 300;
	+uint_t zfs_vdev_queue_depth_pct = 300;
	#endif

	/*
	* When performing allocations for a given metaslab, we want to make sure that
	* there are enough IOs to aggregate together to improve throughput. We want to
	* ensure that there are at least 128k worth of IOs that can be aggregated, and
	* we assume that the average allocation size is 4k, so we need the queue depth
	* to be 32 per allocator to get good aggregation of sequential writes.
	*/
	-int zfs_vdev_def_queue_depth = 32;
	+uint_t zfs_vdev_def_queue_depth = 32;

	/*
	* Allow TRIM I/Os to be aggregated. This should normally not be needed since
	* TRIM I/O for extents up to zfs_trim_extent_bytes_max (128M) can be submitted
	* by the TRIM code in zfs_trim.c.
	*/
	-static int zfs_vdev_aggregate_trim = 0;
	+static uint_t zfs_vdev_aggregate_trim = 0;

	static int
	vdev_queue_offset_compare(const void x1, const void x2)
	{
	const zio_t z1 = (const zio_t )x1;
	const zio_t z2 = (const zio_t )x2;

	int cmp = TREE_CMP(z1->io_offset, z2->io_offset);

	if (likely(cmp))
	return (cmp);

	return (TREE_PCMP(z1, z2));
	}

	static inline avl_tree_t *
	vdev_queue_class_tree(vdev_queue_t *vq, zio_priority_t p)
	{
	return (&vq->vq_class[p].vqc_queued_tree);
	}

	static inline avl_tree_t *
	vdev_queue_type_tree(vdev_queue_t *vq, zio_type_t t)
	{
	ASSERT(t == ZIO_TYPE_READ \|\| t == ZIO_TYPE_WRITE \|\| t == ZIO_TYPE_TRIM);
	if (t == ZIO_TYPE_READ)
	return (&vq->vq_read_offset_tree);
	else if (t == ZIO_TYPE_WRITE)
	return (&vq->vq_write_offset_tree);
	else
	return (&vq->vq_trim_offset_tree);
	}

	static int
	vdev_queue_timestamp_compare(const void x1, const void x2)
	{
	const zio_t z1 = (const zio_t )x1;
	const zio_t z2 = (const zio_t )x2;

	int cmp = TREE_CMP(z1->io_timestamp, z2->io_timestamp);

	if (likely(cmp))
	return (cmp);

	return (TREE_PCMP(z1, z2));
	}

	-static int
	+static uint_t
	vdev_queue_class_min_active(vdev_queue_t *vq, zio_priority_t p)
	{
	switch (p) {
	case ZIO_PRIORITY_SYNC_READ:
	return (zfs_vdev_sync_read_min_active);
	case ZIO_PRIORITY_SYNC_WRITE:
	return (zfs_vdev_sync_write_min_active);
	case ZIO_PRIORITY_ASYNC_READ:
	return (zfs_vdev_async_read_min_active);
	case ZIO_PRIORITY_ASYNC_WRITE:
	return (zfs_vdev_async_write_min_active);
	case ZIO_PRIORITY_SCRUB:
	return (vq->vq_ia_active == 0 ? zfs_vdev_scrub_min_active :
	MIN(vq->vq_nia_credit, zfs_vdev_scrub_min_active));
	case ZIO_PRIORITY_REMOVAL:
	return (vq->vq_ia_active == 0 ? zfs_vdev_removal_min_active :
	MIN(vq->vq_nia_credit, zfs_vdev_removal_min_active));
	case ZIO_PRIORITY_INITIALIZING:
	return (vq->vq_ia_active == 0 ?zfs_vdev_initializing_min_active:
	MIN(vq->vq_nia_credit, zfs_vdev_initializing_min_active));
	case ZIO_PRIORITY_TRIM:
	return (zfs_vdev_trim_min_active);
	case ZIO_PRIORITY_REBUILD:
	return (vq->vq_ia_active == 0 ? zfs_vdev_rebuild_min_active :
	MIN(vq->vq_nia_credit, zfs_vdev_rebuild_min_active));
	default:
	panic("invalid priority %u", p);
	return (0);
	}
	}

	-static int
	+static uint_t
	vdev_queue_max_async_writes(spa_t *spa)
	{
	- int writes;
	+ uint_t writes;
	uint64_t dirty = 0;
	dsl_pool_t *dp = spa_get_dsl(spa);
	uint64_t min_bytes = zfs_dirty_data_max *
	zfs_vdev_async_write_active_min_dirty_percent / 100;
	uint64_t max_bytes = zfs_dirty_data_max *
	zfs_vdev_async_write_active_max_dirty_percent / 100;

	/*
	* Async writes may occur before the assignment of the spa's
	* dsl_pool_t if a self-healing zio is issued prior to the
	* completion of dmu_objset_open_impl().
	*/
	if (dp == NULL)
	return (zfs_vdev_async_write_max_active);

	/*
	* Sync tasks correspond to interactive user actions. To reduce the
	* execution time of those actions we push data out as fast as possible.
	*/
	dirty = dp->dp_dirty_total;
	if (dirty > max_bytes \|\| spa_has_pending_synctask(spa))
	return (zfs_vdev_async_write_max_active);

	if (dirty < min_bytes)
	return (zfs_vdev_async_write_min_active);

	/*
	* linear interpolation:
	* slope = (max_writes - min_writes) / (max_bytes - min_bytes)
	* move right by min_bytes
	* move up by min_writes
	*/
	writes = (dirty - min_bytes) *
	(zfs_vdev_async_write_max_active -
	zfs_vdev_async_write_min_active) /
	(max_bytes - min_bytes) +
	zfs_vdev_async_write_min_active;
	ASSERT3U(writes, >=, zfs_vdev_async_write_min_active);
	ASSERT3U(writes, <=, zfs_vdev_async_write_max_active);
	return (writes);
	}

	-static int
	+static uint_t
	vdev_queue_class_max_active(spa_t spa, vdev_queue_t vq, zio_priority_t p)
	{
	switch (p) {
	case ZIO_PRIORITY_SYNC_READ:
	return (zfs_vdev_sync_read_max_active);
	case ZIO_PRIORITY_SYNC_WRITE:
	return (zfs_vdev_sync_write_max_active);
	case ZIO_PRIORITY_ASYNC_READ:
	return (zfs_vdev_async_read_max_active);
	case ZIO_PRIORITY_ASYNC_WRITE:
	return (vdev_queue_max_async_writes(spa));
	case ZIO_PRIORITY_SCRUB:
	if (vq->vq_ia_active > 0) {
	return (MIN(vq->vq_nia_credit,
	zfs_vdev_scrub_min_active));
	} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
	return (MAX(1, zfs_vdev_scrub_min_active));
	return (zfs_vdev_scrub_max_active);
	case ZIO_PRIORITY_REMOVAL:
	if (vq->vq_ia_active > 0) {
	return (MIN(vq->vq_nia_credit,
	zfs_vdev_removal_min_active));
	} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
	return (MAX(1, zfs_vdev_removal_min_active));
	return (zfs_vdev_removal_max_active);
	case ZIO_PRIORITY_INITIALIZING:
	if (vq->vq_ia_active > 0) {
	return (MIN(vq->vq_nia_credit,
	zfs_vdev_initializing_min_active));
	} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
	return (MAX(1, zfs_vdev_initializing_min_active));
	return (zfs_vdev_initializing_max_active);
	case ZIO_PRIORITY_TRIM:
	return (zfs_vdev_trim_max_active);
	case ZIO_PRIORITY_REBUILD:
	if (vq->vq_ia_active > 0) {
	return (MIN(vq->vq_nia_credit,
	zfs_vdev_rebuild_min_active));
	} else if (vq->vq_nia_credit < zfs_vdev_nia_delay)
	return (MAX(1, zfs_vdev_rebuild_min_active));
	return (zfs_vdev_rebuild_max_active);
	default:
	panic("invalid priority %u", p);
	return (0);
	}
	}

	/*
	* Return the i/o class to issue from, or ZIO_PRIORITY_NUM_QUEUEABLE if
	* there is no eligible class.
	*/
	static zio_priority_t
	vdev_queue_class_to_issue(vdev_queue_t *vq)
	{
	spa_t *spa = vq->vq_vdev->vdev_spa;
	zio_priority_t p, n;

	if (avl_numnodes(&vq->vq_active_tree) >= zfs_vdev_max_active)
	return (ZIO_PRIORITY_NUM_QUEUEABLE);

	/*
	* Find a queue that has not reached its minimum # outstanding i/os.
	* Do round-robin to reduce starvation due to zfs_vdev_max_active
	* and vq_nia_credit limits.
	*/
	for (n = 0; n < ZIO_PRIORITY_NUM_QUEUEABLE; n++) {
	p = (vq->vq_last_prio + n + 1) % ZIO_PRIORITY_NUM_QUEUEABLE;
	if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
	vq->vq_class[p].vqc_active <
	vdev_queue_class_min_active(vq, p)) {
	vq->vq_last_prio = p;
	return (p);
	}
	}

	/*
	* If we haven't found a queue, look for one that hasn't reached its
	* maximum # outstanding i/os.
	*/
	for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
	if (avl_numnodes(vdev_queue_class_tree(vq, p)) > 0 &&
	vq->vq_class[p].vqc_active <
	vdev_queue_class_max_active(spa, vq, p)) {
	vq->vq_last_prio = p;
	return (p);
	}
	}

	/* No eligible queued i/os */
	return (ZIO_PRIORITY_NUM_QUEUEABLE);
	}

	void
	vdev_queue_init(vdev_t *vd)
	{
	vdev_queue_t *vq = &vd->vdev_queue;
	zio_priority_t p;

	mutex_init(&vq->vq_lock, NULL, MUTEX_DEFAULT, NULL);
	vq->vq_vdev = vd;
	taskq_init_ent(&vd->vdev_queue.vq_io_search.io_tqent);

	avl_create(&vq->vq_active_tree, vdev_queue_offset_compare,
	sizeof (zio_t), offsetof(struct zio, io_queue_node));
	avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_READ),
	vdev_queue_offset_compare, sizeof (zio_t),
	offsetof(struct zio, io_offset_node));
	avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE),
	vdev_queue_offset_compare, sizeof (zio_t),
	offsetof(struct zio, io_offset_node));
	avl_create(vdev_queue_type_tree(vq, ZIO_TYPE_TRIM),
	vdev_queue_offset_compare, sizeof (zio_t),
	offsetof(struct zio, io_offset_node));

	for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) {
	int (compfn) (const void , const void *);

	/*
	* The synchronous/trim i/o queues are dispatched in FIFO rather
	* than LBA order. This provides more consistent latency for
	* these i/os.
	*/
	if (p == ZIO_PRIORITY_SYNC_READ \|\|
	p == ZIO_PRIORITY_SYNC_WRITE \|\|
	p == ZIO_PRIORITY_TRIM) {
	compfn = vdev_queue_timestamp_compare;
	} else {
	compfn = vdev_queue_offset_compare;
	}
	avl_create(vdev_queue_class_tree(vq, p), compfn,
	sizeof (zio_t), offsetof(struct zio, io_queue_node));
	}

	vq->vq_last_offset = 0;
	}

	void
	vdev_queue_fini(vdev_t *vd)
	{
	vdev_queue_t *vq = &vd->vdev_queue;

	for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++)
	avl_destroy(vdev_queue_class_tree(vq, p));
	avl_destroy(&vq->vq_active_tree);
	avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_READ));
	avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_WRITE));
	avl_destroy(vdev_queue_type_tree(vq, ZIO_TYPE_TRIM));

	mutex_destroy(&vq->vq_lock);
	}

	static void
	vdev_queue_io_add(vdev_queue_t vq, zio_t zio)
	{
	ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
	avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
	avl_add(vdev_queue_type_tree(vq, zio->io_type), zio);
	}

	static void
	vdev_queue_io_remove(vdev_queue_t vq, zio_t zio)
	{
	ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
	avl_remove(vdev_queue_class_tree(vq, zio->io_priority), zio);
	avl_remove(vdev_queue_type_tree(vq, zio->io_type), zio);
	}

	static boolean_t
	vdev_queue_is_interactive(zio_priority_t p)
	{
	switch (p) {
	case ZIO_PRIORITY_SCRUB:
	case ZIO_PRIORITY_REMOVAL:
	case ZIO_PRIORITY_INITIALIZING:
	case ZIO_PRIORITY_REBUILD:
	return (B_FALSE);
	default:
	return (B_TRUE);
	}
	}

	static void
	vdev_queue_pending_add(vdev_queue_t vq, zio_t zio)
	{
	ASSERT(MUTEX_HELD(&vq->vq_lock));
	ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
	vq->vq_class[zio->io_priority].vqc_active++;
	if (vdev_queue_is_interactive(zio->io_priority)) {
	if (++vq->vq_ia_active == 1)
	vq->vq_nia_credit = 1;
	} else if (vq->vq_ia_active > 0) {
	vq->vq_nia_credit--;
	}
	avl_add(&vq->vq_active_tree, zio);
	}

	static void
	vdev_queue_pending_remove(vdev_queue_t vq, zio_t zio)
	{
	ASSERT(MUTEX_HELD(&vq->vq_lock));
	ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
	vq->vq_class[zio->io_priority].vqc_active--;
	if (vdev_queue_is_interactive(zio->io_priority)) {
	if (--vq->vq_ia_active == 0)
	vq->vq_nia_credit = 0;
	else
	vq->vq_nia_credit = zfs_vdev_nia_credit;
	} else if (vq->vq_ia_active == 0)
	vq->vq_nia_credit++;
	avl_remove(&vq->vq_active_tree, zio);
	}

	static void
	vdev_queue_agg_io_done(zio_t *aio)
	{
	abd_free(aio->io_abd);
	}

	/*
	* Compute the range spanned by two i/os, which is the endpoint of the last
	* (lio->io_offset + lio->io_size) minus start of the first (fio->io_offset).
	* Conveniently, the gap between fio and lio is given by -IO_SPAN(lio, fio);
	* thus fio and lio are adjacent if and only if IO_SPAN(lio, fio) == 0.
	*/
	#define IO_SPAN(fio, lio) ((lio)->io_offset + (lio)->io_size - (fio)->io_offset)
	#define IO_GAP(fio, lio) (-IO_SPAN(lio, fio))

	/*
	* Sufficiently adjacent io_offset's in ZIOs will be aggregated. We do this
	* by creating a gang ABD from the adjacent ZIOs io_abd's. By using
	* a gang ABD we avoid doing memory copies to and from the parent,
	* child ZIOs. The gang ABD also accounts for gaps between adjacent
	* io_offsets by simply getting the zero ABD for writes or allocating
	* a new ABD for reads and placing them in the gang ABD as well.
	*/
	static zio_t *
	vdev_queue_aggregate(vdev_queue_t vq, zio_t zio)
	{
	zio_t first, last, aio, dio, mandatory, nio;
	uint64_t maxgap = 0;
	uint64_t size;
	uint64_t limit;
	int maxblocksize;
	boolean_t stretch = B_FALSE;
	avl_tree_t *t = vdev_queue_type_tree(vq, zio->io_type);
	enum zio_flag flags = zio->io_flags & ZIO_FLAG_AGG_INHERIT;
	uint64_t next_offset;
	abd_t *abd;

	maxblocksize = spa_maxblocksize(vq->vq_vdev->vdev_spa);
	if (vq->vq_vdev->vdev_nonrot)
	limit = zfs_vdev_aggregation_limit_non_rotating;
	else
	limit = zfs_vdev_aggregation_limit;
	- limit = MAX(MIN(limit, maxblocksize), 0);
	+ limit = MIN(limit, maxblocksize);

	if (zio->io_flags & ZIO_FLAG_DONT_AGGREGATE \|\| limit == 0)
	return (NULL);

	/*
	* While TRIM commands could be aggregated based on offset this
	* behavior is disabled until it's determined to be beneficial.
	*/
	if (zio->io_type == ZIO_TYPE_TRIM && !zfs_vdev_aggregate_trim)
	return (NULL);

	/*
	* I/Os to distributed spares are directly dispatched to the dRAID
	* leaf vdevs for aggregation. See the comment at the end of the
	* zio_vdev_io_start() function.
	*/
	ASSERT(vq->vq_vdev->vdev_ops != &vdev_draid_spare_ops);

	first = last = zio;

	if (zio->io_type == ZIO_TYPE_READ)
	maxgap = zfs_vdev_read_gap_limit;

	/*
	* We can aggregate I/Os that are sufficiently adjacent and of
	* the same flavor, as expressed by the AGG_INHERIT flags.
	* The latter requirement is necessary so that certain
	* attributes of the I/O, such as whether it's a normal I/O
	* or a scrub/resilver, can be preserved in the aggregate.
	* We can include optional I/Os, but don't allow them
	* to begin a range as they add no benefit in that situation.
	*/

	/*
	* We keep track of the last non-optional I/O.
	*/
	mandatory = (first->io_flags & ZIO_FLAG_OPTIONAL) ? NULL : first;

	/*
	* Walk backwards through sufficiently contiguous I/Os
	* recording the last non-optional I/O.
	*/
	while ((dio = AVL_PREV(t, first)) != NULL &&
	(dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
	IO_SPAN(dio, last) <= limit &&
	IO_GAP(dio, first) <= maxgap &&
	dio->io_type == zio->io_type) {
	first = dio;
	if (mandatory == NULL && !(first->io_flags & ZIO_FLAG_OPTIONAL))
	mandatory = first;
	}

	/*
	* Skip any initial optional I/Os.
	*/
	while ((first->io_flags & ZIO_FLAG_OPTIONAL) && first != last) {
	first = AVL_NEXT(t, first);
	ASSERT(first != NULL);
	}


	/*
	* Walk forward through sufficiently contiguous I/Os.
	* The aggregation limit does not apply to optional i/os, so that
	* we can issue contiguous writes even if they are larger than the
	* aggregation limit.
	*/
	while ((dio = AVL_NEXT(t, last)) != NULL &&
	(dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
	(IO_SPAN(first, dio) <= limit \|\|
	(dio->io_flags & ZIO_FLAG_OPTIONAL)) &&
	IO_SPAN(first, dio) <= maxblocksize &&
	IO_GAP(last, dio) <= maxgap &&
	dio->io_type == zio->io_type) {
	last = dio;
	if (!(last->io_flags & ZIO_FLAG_OPTIONAL))
	mandatory = last;
	}

	/*
	* Now that we've established the range of the I/O aggregation
	* we must decide what to do with trailing optional I/Os.
	* For reads, there's nothing to do. While we are unable to
	* aggregate further, it's possible that a trailing optional
	* I/O would allow the underlying device to aggregate with
	* subsequent I/Os. We must therefore determine if the next
	* non-optional I/O is close enough to make aggregation
	* worthwhile.
	*/
	if (zio->io_type == ZIO_TYPE_WRITE && mandatory != NULL) {
	zio_t *nio = last;
	while ((dio = AVL_NEXT(t, nio)) != NULL &&
	IO_GAP(nio, dio) == 0 &&
	IO_GAP(mandatory, dio) <= zfs_vdev_write_gap_limit) {
	nio = dio;
	if (!(nio->io_flags & ZIO_FLAG_OPTIONAL)) {
	stretch = B_TRUE;
	break;
	}
	}
	}

	if (stretch) {
	/*
	* We are going to include an optional io in our aggregated
	* span, thus closing the write gap. Only mandatory i/os can
	* start aggregated spans, so make sure that the next i/o
	* after our span is mandatory.
	*/
	dio = AVL_NEXT(t, last);
	dio->io_flags &= ~ZIO_FLAG_OPTIONAL;
	} else {
	/* do not include the optional i/o */
	while (last != mandatory && last != first) {
	ASSERT(last->io_flags & ZIO_FLAG_OPTIONAL);
	last = AVL_PREV(t, last);
	ASSERT(last != NULL);
	}
	}

	if (first == last)
	return (NULL);

	size = IO_SPAN(first, last);
	ASSERT3U(size, <=, maxblocksize);

	abd = abd_alloc_gang();
	if (abd == NULL)
	return (NULL);

	aio = zio_vdev_delegated_io(first->io_vd, first->io_offset,
	abd, size, first->io_type, zio->io_priority,
	flags \| ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_QUEUE,
	vdev_queue_agg_io_done, NULL);
	aio->io_timestamp = first->io_timestamp;

	nio = first;
	next_offset = first->io_offset;
	do {
	dio = nio;
	nio = AVL_NEXT(t, dio);
	zio_add_child(dio, aio);
	vdev_queue_io_remove(vq, dio);

	if (dio->io_offset != next_offset) {
	/* allocate a buffer for a read gap */
	ASSERT3U(dio->io_type, ==, ZIO_TYPE_READ);
	ASSERT3U(dio->io_offset, >, next_offset);
	abd = abd_alloc_for_io(
	dio->io_offset - next_offset, B_TRUE);
	abd_gang_add(aio->io_abd, abd, B_TRUE);
	}
	if (dio->io_abd &&
	(dio->io_size != abd_get_size(dio->io_abd))) {
	/* abd size not the same as IO size */
	ASSERT3U(abd_get_size(dio->io_abd), >, dio->io_size);
	abd = abd_get_offset_size(dio->io_abd, 0, dio->io_size);
	abd_gang_add(aio->io_abd, abd, B_TRUE);
	} else {
	if (dio->io_flags & ZIO_FLAG_NODATA) {
	/* allocate a buffer for a write gap */
	ASSERT3U(dio->io_type, ==, ZIO_TYPE_WRITE);
	ASSERT3P(dio->io_abd, ==, NULL);
	abd_gang_add(aio->io_abd,
	abd_get_zeros(dio->io_size), B_TRUE);
	} else {
	/*
	* We pass B_FALSE to abd_gang_add()
	* because we did not allocate a new
	* ABD, so it is assumed the caller
	* will free this ABD.
	*/
	abd_gang_add(aio->io_abd, dio->io_abd,
	B_FALSE);
	}
	}
	next_offset = dio->io_offset + dio->io_size;
	} while (dio != last);
	ASSERT3U(abd_get_size(aio->io_abd), ==, aio->io_size);

	/*
	* Callers must call zio_vdev_io_bypass() and zio_execute() for
	* aggregated (parent) I/Os so that we could avoid dropping the
	* queue's lock here to avoid a deadlock that we could encounter
	* due to lock order reversal between vq_lock and io_lock in
	* zio_change_priority().
	*/
	return (aio);
	}

	static zio_t *
	vdev_queue_io_to_issue(vdev_queue_t *vq)
	{
	zio_t zio, aio;
	zio_priority_t p;
	avl_index_t idx;
	avl_tree_t *tree;

	again:
	ASSERT(MUTEX_HELD(&vq->vq_lock));

	p = vdev_queue_class_to_issue(vq);

	if (p == ZIO_PRIORITY_NUM_QUEUEABLE) {
	/* No eligible queued i/os */
	return (NULL);
	}

	/*
	* For LBA-ordered queues (async / scrub / initializing), issue the
	* i/o which follows the most recently issued i/o in LBA (offset) order.
	*
	* For FIFO queues (sync/trim), issue the i/o with the lowest timestamp.
	*/
	tree = vdev_queue_class_tree(vq, p);
	vq->vq_io_search.io_timestamp = 0;
	vq->vq_io_search.io_offset = vq->vq_last_offset - 1;
	VERIFY3P(avl_find(tree, &vq->vq_io_search, &idx), ==, NULL);
	zio = avl_nearest(tree, idx, AVL_AFTER);
	if (zio == NULL)
	zio = avl_first(tree);
	ASSERT3U(zio->io_priority, ==, p);

	aio = vdev_queue_aggregate(vq, zio);
	if (aio != NULL) {
	zio = aio;
	} else {
	vdev_queue_io_remove(vq, zio);

	/*
	* If the I/O is or was optional and therefore has no data, we
	* need to simply discard it. We need to drop the vdev queue's
	* lock to avoid a deadlock that we could encounter since this
	* I/O will complete immediately.
	*/
	if (zio->io_flags & ZIO_FLAG_NODATA) {
	mutex_exit(&vq->vq_lock);
	zio_vdev_io_bypass(zio);
	zio_execute(zio);
	mutex_enter(&vq->vq_lock);
	goto again;
	}
	}

	vdev_queue_pending_add(vq, zio);
	vq->vq_last_offset = zio->io_offset + zio->io_size;

	return (zio);
	}

	zio_t *
	vdev_queue_io(zio_t *zio)
	{
	vdev_queue_t *vq = &zio->io_vd->vdev_queue;
	zio_t dio, nio;
	zio_link_t *zl = NULL;

	if (zio->io_flags & ZIO_FLAG_DONT_QUEUE)
	return (zio);

	/*
	* Children i/os inherent their parent's priority, which might
	* not match the child's i/o type. Fix it up here.
	*/
	if (zio->io_type == ZIO_TYPE_READ) {
	ASSERT(zio->io_priority != ZIO_PRIORITY_TRIM);

	if (zio->io_priority != ZIO_PRIORITY_SYNC_READ &&
	zio->io_priority != ZIO_PRIORITY_ASYNC_READ &&
	zio->io_priority != ZIO_PRIORITY_SCRUB &&
	zio->io_priority != ZIO_PRIORITY_REMOVAL &&
	zio->io_priority != ZIO_PRIORITY_INITIALIZING &&
	zio->io_priority != ZIO_PRIORITY_REBUILD) {
	zio->io_priority = ZIO_PRIORITY_ASYNC_READ;
	}
	} else if (zio->io_type == ZIO_TYPE_WRITE) {
	ASSERT(zio->io_priority != ZIO_PRIORITY_TRIM);

	if (zio->io_priority != ZIO_PRIORITY_SYNC_WRITE &&
	zio->io_priority != ZIO_PRIORITY_ASYNC_WRITE &&
	zio->io_priority != ZIO_PRIORITY_REMOVAL &&
	zio->io_priority != ZIO_PRIORITY_INITIALIZING &&
	zio->io_priority != ZIO_PRIORITY_REBUILD) {
	zio->io_priority = ZIO_PRIORITY_ASYNC_WRITE;
	}
	} else {
	ASSERT(zio->io_type == ZIO_TYPE_TRIM);
	ASSERT(zio->io_priority == ZIO_PRIORITY_TRIM);
	}

	zio->io_flags \|= ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_QUEUE;
	zio->io_timestamp = gethrtime();

	mutex_enter(&vq->vq_lock);
	vdev_queue_io_add(vq, zio);
	nio = vdev_queue_io_to_issue(vq);
	mutex_exit(&vq->vq_lock);

	if (nio == NULL)
	return (NULL);

	if (nio->io_done == vdev_queue_agg_io_done) {
	while ((dio = zio_walk_parents(nio, &zl)) != NULL) {
	ASSERT3U(dio->io_type, ==, nio->io_type);
	zio_vdev_io_bypass(dio);
	zio_execute(dio);
	}
	zio_nowait(nio);
	return (NULL);
	}

	return (nio);
	}

	void
	vdev_queue_io_done(zio_t *zio)
	{
	vdev_queue_t *vq = &zio->io_vd->vdev_queue;
	zio_t dio, nio;
	zio_link_t *zl = NULL;

	hrtime_t now = gethrtime();
	vq->vq_io_complete_ts = now;
	vq->vq_io_delta_ts = zio->io_delta = now - zio->io_timestamp;

	mutex_enter(&vq->vq_lock);
	vdev_queue_pending_remove(vq, zio);

	while ((nio = vdev_queue_io_to_issue(vq)) != NULL) {
	mutex_exit(&vq->vq_lock);
	if (nio->io_done == vdev_queue_agg_io_done) {
	while ((dio = zio_walk_parents(nio, &zl)) != NULL) {
	ASSERT3U(dio->io_type, ==, nio->io_type);
	zio_vdev_io_bypass(dio);
	zio_execute(dio);
	}
	zio_nowait(nio);
	} else {
	zio_vdev_io_reissue(nio);
	zio_execute(nio);
	}
	mutex_enter(&vq->vq_lock);
	}

	mutex_exit(&vq->vq_lock);
	}

	void
	vdev_queue_change_io_priority(zio_t *zio, zio_priority_t priority)
	{
	vdev_queue_t *vq = &zio->io_vd->vdev_queue;
	avl_tree_t *tree;

	/*
	* ZIO_PRIORITY_NOW is used by the vdev cache code and the aggregate zio
	* code to issue IOs without adding them to the vdev queue. In this
	* case, the zio is already going to be issued as quickly as possible
	* and so it doesn't need any reprioritization to help.
	*/
	if (zio->io_priority == ZIO_PRIORITY_NOW)
	return;

	ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);
	ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);

	if (zio->io_type == ZIO_TYPE_READ) {
	if (priority != ZIO_PRIORITY_SYNC_READ &&
	priority != ZIO_PRIORITY_ASYNC_READ &&
	priority != ZIO_PRIORITY_SCRUB)
	priority = ZIO_PRIORITY_ASYNC_READ;
	} else {
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	if (priority != ZIO_PRIORITY_SYNC_WRITE &&
	priority != ZIO_PRIORITY_ASYNC_WRITE)
	priority = ZIO_PRIORITY_ASYNC_WRITE;
	}

	mutex_enter(&vq->vq_lock);

	/*
	* If the zio is in none of the queues we can simply change
	* the priority. If the zio is waiting to be submitted we must
	* remove it from the queue and re-insert it with the new priority.
	* Otherwise, the zio is currently active and we cannot change its
	* priority.
	*/
	tree = vdev_queue_class_tree(vq, zio->io_priority);
	if (avl_find(tree, zio, NULL) == zio) {
	avl_remove(vdev_queue_class_tree(vq, zio->io_priority), zio);
	zio->io_priority = priority;
	avl_add(vdev_queue_class_tree(vq, zio->io_priority), zio);
	} else if (avl_find(&vq->vq_active_tree, zio, NULL) != zio) {
	zio->io_priority = priority;
	}

	mutex_exit(&vq->vq_lock);
	}

	/*
	* As these two methods are only used for load calculations we're not
	* concerned if we get an incorrect value on 32bit platforms due to lack of
	* vq_lock mutex use here, instead we prefer to keep it lock free for
	* performance.
	*/
	int
	vdev_queue_length(vdev_t *vd)
	{
	return (avl_numnodes(&vd->vdev_queue.vq_active_tree));
	}

	uint64_t
	vdev_queue_last_offset(vdev_t *vd)
	{
	return (vd->vdev_queue.vq_last_offset);
	}

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit, UINT, ZMOD_RW,
	"Max vdev I/O aggregation size");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit_non_rotating, INT,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregation_limit_non_rotating, UINT,
	ZMOD_RW, "Max vdev I/O aggregation size for non-rotating media");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregate_trim, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, aggregate_trim, UINT, ZMOD_RW,
	"Allow TRIM I/O to be aggregated");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, read_gap_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, read_gap_limit, UINT, ZMOD_RW,
	"Aggregate read I/O over gap");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, write_gap_limit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, write_gap_limit, UINT, ZMOD_RW,
	"Aggregate write I/O over gap");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, max_active, UINT, ZMOD_RW,
	"Maximum number of active I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_max_dirty_percent, INT,
	- ZMOD_RW, "Async write concurrency max threshold");
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_max_dirty_percent,
	+ UINT, ZMOD_RW, "Async write concurrency max threshold");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_min_dirty_percent, INT,
	- ZMOD_RW, "Async write concurrency min threshold");
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_active_min_dirty_percent,
	+ UINT, ZMOD_RW, "Async write concurrency min threshold");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_max_active, UINT, ZMOD_RW,
	"Max active async read I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_read_min_active, UINT, ZMOD_RW,
	"Min active async read I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_max_active, UINT, ZMOD_RW,
	"Max active async write I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, async_write_min_active, UINT, ZMOD_RW,
	"Min active async write I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_max_active, UINT, ZMOD_RW,
	"Max active initializing I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, initializing_min_active, UINT, ZMOD_RW,
	"Min active initializing I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_max_active, UINT, ZMOD_RW,
	"Max active removal I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, removal_min_active, UINT, ZMOD_RW,
	"Min active removal I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_max_active, UINT, ZMOD_RW,
	"Max active scrub I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, scrub_min_active, UINT, ZMOD_RW,
	"Min active scrub I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_max_active, UINT, ZMOD_RW,
	"Max active sync read I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_read_min_active, UINT, ZMOD_RW,
	"Min active sync read I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_max_active, UINT, ZMOD_RW,
	"Max active sync write I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, sync_write_min_active, UINT, ZMOD_RW,
	"Min active sync write I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_max_active, UINT, ZMOD_RW,
	"Max active trim/discard I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, trim_min_active, UINT, ZMOD_RW,
	"Min active trim/discard I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_max_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_max_active, UINT, ZMOD_RW,
	"Max active rebuild I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_min_active, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, rebuild_min_active, UINT, ZMOD_RW,
	"Min active rebuild I/Os per vdev");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_credit, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_credit, UINT, ZMOD_RW,
	"Number of non-interactive I/Os to allow in sequence");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_delay, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, nia_delay, UINT, ZMOD_RW,
	"Number of non-interactive I/Os before _max_active");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, queue_depth_pct, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, queue_depth_pct, UINT, ZMOD_RW,
	"Queue depth percentage for each top-level vdev");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_removal.c b/sys/contrib/openzfs/module/zfs/vdev_removal.c
	index a48b47a50540..53592dbfdfa0 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_removal.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_removal.c
	@@ -1,2568 +1,2568 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/vdev_impl.h>
	#include <sys/metaslab.h>
	#include <sys/metaslab_impl.h>
	#include <sys/uberblock_impl.h>
	#include <sys/txg.h>
	#include <sys/avl.h>
	#include <sys/bpobj.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_dir.h>
	#include <sys/arc.h>
	#include <sys/zfeature.h>
	#include <sys/vdev_indirect_births.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/abd.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>
	#include <sys/trace_zfs.h>

	/*
	* This file contains the necessary logic to remove vdevs from a
	* storage pool. Currently, the only devices that can be removed
	* are log, cache, and spare devices; and top level vdevs from a pool
	* w/o raidz or mirrors. (Note that members of a mirror can be removed
	* by the detach operation.)
	*
	* Log vdevs are removed by evacuating them and then turning the vdev
	* into a hole vdev while holding spa config locks.
	*
	* Top level vdevs are removed and converted into an indirect vdev via
	* a multi-step process:
	*
	* - Disable allocations from this device (spa_vdev_remove_top).
	*
	* - From a new thread (spa_vdev_remove_thread), copy data from
	* the removing vdev to a different vdev. The copy happens in open
	* context (spa_vdev_copy_impl) and issues a sync task
	* (vdev_mapping_sync) so the sync thread can update the partial
	* indirect mappings in core and on disk.
	*
	* - If a free happens during a removal, it is freed from the
	* removing vdev, and if it has already been copied, from the new
	* location as well (free_from_removing_vdev).
	*
	* - After the removal is completed, the copy thread converts the vdev
	* into an indirect vdev (vdev_remove_complete) before instructing
	* the sync thread to destroy the space maps and finish the removal
	* (spa_finish_removal).
	*/

	typedef struct vdev_copy_arg {
	metaslab_t *vca_msp;
	uint64_t vca_outstanding_bytes;
	uint64_t vca_read_error_bytes;
	uint64_t vca_write_error_bytes;
	kcondvar_t vca_cv;
	kmutex_t vca_lock;
	} vdev_copy_arg_t;

	/*
	* The maximum amount of memory we can use for outstanding i/o while
	* doing a device removal. This determines how much i/o we can have
	* in flight concurrently.
	*/
	-static const int zfs_remove_max_copy_bytes = 64 * 1024 * 1024;
	+static const uint_t zfs_remove_max_copy_bytes = 64 * 1024 * 1024;

	/*
	* The largest contiguous segment that we will attempt to allocate when
	* removing a device. This can be no larger than SPA_MAXBLOCKSIZE. If
	* there is a performance problem with attempting to allocate large blocks,
	* consider decreasing this.
	*
	* See also the accessor function spa_remove_max_segment().
	*/
	-int zfs_remove_max_segment = SPA_MAXBLOCKSIZE;
	+uint_t zfs_remove_max_segment = SPA_MAXBLOCKSIZE;

	/*
	* Ignore hard IO errors during device removal. When set if a device
	* encounters hard IO error during the removal process the removal will
	* not be cancelled. This can result in a normally recoverable block
	* becoming permanently damaged and is not recommended.
	*/
	static int zfs_removal_ignore_errors = 0;

	/*
	* Allow a remap segment to span free chunks of at most this size. The main
	* impact of a larger span is that we will read and write larger, more
	* contiguous chunks, with more "unnecessary" data -- trading off bandwidth
	* for iops. The value here was chosen to align with
	* zfs_vdev_read_gap_limit, which is a similar concept when doing regular
	* reads (but there's no reason it has to be the same).
	*
	* Additionally, a higher span will have the following relatively minor
	* effects:
	* - the mapping will be smaller, since one entry can cover more allocated
	* segments
	* - more of the fragmentation in the removing device will be preserved
	* - we'll do larger allocations, which may fail and fall back on smaller
	* allocations
	*/
	-int vdev_removal_max_span = 32 * 1024;
	+uint_t vdev_removal_max_span = 32 * 1024;

	/*
	* This is used by the test suite so that it can ensure that certain
	* actions happen while in the middle of a removal.
	*/
	int zfs_removal_suspend_progress = 0;

	#define VDEV_REMOVAL_ZAP_OBJS "lzap"

	static __attribute__((noreturn)) void spa_vdev_remove_thread(void *arg);
	static int spa_vdev_remove_cancel_impl(spa_t *spa);

	static void
	spa_sync_removing_state(spa_t spa, dmu_tx_t tx)
	{
	VERIFY0(zap_update(spa->spa_dsl_pool->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_REMOVING, sizeof (uint64_t),
	sizeof (spa->spa_removing_phys) / sizeof (uint64_t),
	&spa->spa_removing_phys, tx));
	}

	static nvlist_t *
	spa_nvlist_lookup_by_guid(nvlist_t **nvpp, int count, uint64_t target_guid)
	{
	for (int i = 0; i < count; i++) {
	uint64_t guid =
	fnvlist_lookup_uint64(nvpp[i], ZPOOL_CONFIG_GUID);

	if (guid == target_guid)
	return (nvpp[i]);
	}

	return (NULL);
	}

	static void
	vdev_activate(vdev_t *vd)
	{
	metaslab_group_t *mg = vd->vdev_mg;
	spa_t *spa = vd->vdev_spa;
	uint64_t vdev_space = spa_deflate(spa) ?
	vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;

	ASSERT(!vd->vdev_islog);
	ASSERT(vd->vdev_noalloc);

	metaslab_group_activate(mg);
	metaslab_group_activate(vd->vdev_log_mg);

	ASSERT3U(spa->spa_nonallocating_dspace, >=, vdev_space);

	spa->spa_nonallocating_dspace -= vdev_space;

	vd->vdev_noalloc = B_FALSE;
	}

	static int
	vdev_passivate(vdev_t vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	int error;

	ASSERT(!vd->vdev_noalloc);

	vdev_t *rvd = spa->spa_root_vdev;
	metaslab_group_t *mg = vd->vdev_mg;
	metaslab_class_t *normal = spa_normal_class(spa);
	if (mg->mg_class == normal) {
	/*
	* We must check that this is not the only allocating device in
	* the pool before passivating, otherwise we will not be able
	* to make progress because we can't allocate from any vdevs.
	*/
	boolean_t last = B_TRUE;
	for (uint64_t id = 0; id < rvd->vdev_children; id++) {
	vdev_t *cvd = rvd->vdev_child[id];

	if (cvd == vd \|\|
	cvd->vdev_ops == &vdev_indirect_ops)
	continue;

	metaslab_class_t *mc = cvd->vdev_mg->mg_class;
	if (mc != normal)
	continue;

	if (!cvd->vdev_noalloc) {
	last = B_FALSE;
	break;
	}
	}
	if (last)
	return (SET_ERROR(EINVAL));
	}

	metaslab_group_passivate(mg);
	ASSERT(!vd->vdev_islog);
	metaslab_group_passivate(vd->vdev_log_mg);

	/*
	* Wait for the youngest allocations and frees to sync,
	* and then wait for the deferral of those frees to finish.
	*/
	spa_vdev_config_exit(spa, NULL,
	*txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);

	/*
	* We must ensure that no "stubby" log blocks are allocated
	* on the device to be removed. These blocks could be
	* written at any time, including while we are in the middle
	* of copying them.
	*/
	error = spa_reset_logs(spa);

	*txg = spa_vdev_config_enter(spa);

	if (error != 0) {
	metaslab_group_activate(mg);
	ASSERT(!vd->vdev_islog);
	if (vd->vdev_log_mg != NULL)
	metaslab_group_activate(vd->vdev_log_mg);
	return (error);
	}

	spa->spa_nonallocating_dspace += spa_deflate(spa) ?
	vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
	vd->vdev_noalloc = B_TRUE;

	return (0);
	}

	/*
	* Turn off allocations for a top-level device from the pool.
	*
	* Turning off allocations for a top-level device can take a significant
	* amount of time. As a result we use the spa_vdev_config_[enter/exit]
	* functions which allow us to grab and release the spa_config_lock while
	* still holding the namespace lock. During each step the configuration
	* is synced out.
	*/
	int
	spa_vdev_noalloc(spa_t *spa, uint64_t guid)
	{
	vdev_t *vd;
	uint64_t txg;
	int error = 0;

	ASSERT(!MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	vd = spa_lookup_by_guid(spa, guid, B_FALSE);

	if (vd == NULL)
	error = SET_ERROR(ENOENT);
	else if (vd->vdev_mg == NULL)
	error = SET_ERROR(ZFS_ERR_VDEV_NOTSUP);
	else if (!vd->vdev_noalloc)
	error = vdev_passivate(vd, &txg);

	if (error == 0) {
	vdev_dirty_leaves(vd, VDD_DTL, txg);
	vdev_config_dirty(vd);
	}

	error = spa_vdev_exit(spa, NULL, txg, error);

	return (error);
	}

	int
	spa_vdev_alloc(spa_t *spa, uint64_t guid)
	{
	vdev_t *vd;
	uint64_t txg;
	int error = 0;

	ASSERT(!MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	vd = spa_lookup_by_guid(spa, guid, B_FALSE);

	if (vd == NULL)
	error = SET_ERROR(ENOENT);
	else if (vd->vdev_mg == NULL)
	error = SET_ERROR(ZFS_ERR_VDEV_NOTSUP);
	else if (!vd->vdev_removing)
	vdev_activate(vd);

	if (error == 0) {
	vdev_dirty_leaves(vd, VDD_DTL, txg);
	vdev_config_dirty(vd);
	}

	(void) spa_vdev_exit(spa, NULL, txg, error);

	return (error);
	}

	static void
	spa_vdev_remove_aux(nvlist_t config, const char name, nvlist_t **dev,
	int count, nvlist_t *dev_to_remove)
	{
	nvlist_t **newdev = NULL;

	if (count > 1)
	newdev = kmem_alloc((count - 1) * sizeof (void *), KM_SLEEP);

	for (int i = 0, j = 0; i < count; i++) {
	if (dev[i] == dev_to_remove)
	continue;
	VERIFY(nvlist_dup(dev[i], &newdev[j++], KM_SLEEP) == 0);
	}

	VERIFY(nvlist_remove(config, name, DATA_TYPE_NVLIST_ARRAY) == 0);
	fnvlist_add_nvlist_array(config, name, (const nvlist_t * const *)newdev,
	count - 1);

	for (int i = 0; i < count - 1; i++)
	nvlist_free(newdev[i]);

	if (count > 1)
	kmem_free(newdev, (count - 1) * sizeof (void *));
	}

	static spa_vdev_removal_t *
	spa_vdev_removal_create(vdev_t *vd)
	{
	spa_vdev_removal_t svr = kmem_zalloc(sizeof (svr), KM_SLEEP);
	mutex_init(&svr->svr_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&svr->svr_cv, NULL, CV_DEFAULT, NULL);
	svr->svr_allocd_segs = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	svr->svr_vdev_id = vd->vdev_id;

	for (int i = 0; i < TXG_SIZE; i++) {
	svr->svr_frees[i] = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);
	list_create(&svr->svr_new_segments[i],
	sizeof (vdev_indirect_mapping_entry_t),
	offsetof(vdev_indirect_mapping_entry_t, vime_node));
	}

	return (svr);
	}

	void
	spa_vdev_removal_destroy(spa_vdev_removal_t *svr)
	{
	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT0(svr->svr_bytes_done[i]);
	ASSERT0(svr->svr_max_offset_to_sync[i]);
	range_tree_destroy(svr->svr_frees[i]);
	list_destroy(&svr->svr_new_segments[i]);
	}

	range_tree_destroy(svr->svr_allocd_segs);
	mutex_destroy(&svr->svr_lock);
	cv_destroy(&svr->svr_cv);
	kmem_free(svr, sizeof (*svr));
	}

	/*
	* This is called as a synctask in the txg in which we will mark this vdev
	* as removing (in the config stored in the MOS).
	*
	* It begins the evacuation of a toplevel vdev by:
	* - initializing the spa_removing_phys which tracks this removal
	* - computing the amount of space to remove for accounting purposes
	* - dirtying all dbufs in the spa_config_object
	* - creating the spa_vdev_removal
	* - starting the spa_vdev_remove_thread
	*/
	static void
	vdev_remove_initiate_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
	objset_t *mos = spa->spa_dsl_pool->dp_meta_objset;
	spa_vdev_removal_t *svr = NULL;
	uint64_t txg __maybe_unused = dmu_tx_get_txg(tx);

	ASSERT0(vdev_get_nparity(vd));
	svr = spa_vdev_removal_create(vd);

	ASSERT(vd->vdev_removing);
	ASSERT3P(vd->vdev_indirect_mapping, ==, NULL);

	spa_feature_incr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx);
	if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
	/*
	* By activating the OBSOLETE_COUNTS feature, we prevent
	* the pool from being downgraded and ensure that the
	* refcounts are precise.
	*/
	spa_feature_incr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	uint64_t one = 1;
	VERIFY0(zap_add(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, sizeof (one), 1,
	&one, tx));
	boolean_t are_precise __maybe_unused;
	ASSERT0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	ASSERT3B(are_precise, ==, B_TRUE);
	}

	vic->vic_mapping_object = vdev_indirect_mapping_alloc(mos, tx);
	vd->vdev_indirect_mapping =
	vdev_indirect_mapping_open(mos, vic->vic_mapping_object);
	vic->vic_births_object = vdev_indirect_births_alloc(mos, tx);
	vd->vdev_indirect_births =
	vdev_indirect_births_open(mos, vic->vic_births_object);
	spa->spa_removing_phys.sr_removing_vdev = vd->vdev_id;
	spa->spa_removing_phys.sr_start_time = gethrestime_sec();
	spa->spa_removing_phys.sr_end_time = 0;
	spa->spa_removing_phys.sr_state = DSS_SCANNING;
	spa->spa_removing_phys.sr_to_copy = 0;
	spa->spa_removing_phys.sr_copied = 0;

	/*
	* Note: We can't use vdev_stat's vs_alloc for sr_to_copy, because
	* there may be space in the defer tree, which is free, but still
	* counted in vs_alloc.
	*/
	for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
	metaslab_t *ms = vd->vdev_ms[i];
	if (ms->ms_sm == NULL)
	continue;

	spa->spa_removing_phys.sr_to_copy +=
	metaslab_allocated_space(ms);

	/*
	* Space which we are freeing this txg does not need to
	* be copied.
	*/
	spa->spa_removing_phys.sr_to_copy -=
	range_tree_space(ms->ms_freeing);

	ASSERT0(range_tree_space(ms->ms_freed));
	for (int t = 0; t < TXG_SIZE; t++)
	ASSERT0(range_tree_space(ms->ms_allocating[t]));
	}

	/*
	* Sync tasks are called before metaslab_sync(), so there should
	* be no already-synced metaslabs in the TXG_CLEAN list.
	*/
	ASSERT3P(txg_list_head(&vd->vdev_ms_list, TXG_CLEAN(txg)), ==, NULL);

	spa_sync_removing_state(spa, tx);

	/*
	* All blocks that we need to read the most recent mapping must be
	* stored on concrete vdevs. Therefore, we must dirty anything that
	* is read before spa_remove_init(). Specifically, the
	* spa_config_object. (Note that although we already modified the
	* spa_config_object in spa_sync_removing_state, that may not have
	* modified all blocks of the object.)
	*/
	dmu_object_info_t doi;
	VERIFY0(dmu_object_info(mos, DMU_POOL_DIRECTORY_OBJECT, &doi));
	for (uint64_t offset = 0; offset < doi.doi_max_offset; ) {
	dmu_buf_t *dbuf;
	VERIFY0(dmu_buf_hold(mos, DMU_POOL_DIRECTORY_OBJECT,
	offset, FTAG, &dbuf, 0));
	dmu_buf_will_dirty(dbuf, tx);
	offset += dbuf->db_size;
	dmu_buf_rele(dbuf, FTAG);
	}

	/*
	* Now that we've allocated the im_object, dirty the vdev to ensure
	* that the object gets written to the config on disk.
	*/
	vdev_config_dirty(vd);

	zfs_dbgmsg("starting removal thread for vdev %llu (%px) in txg %llu "
	"im_obj=%llu", (u_longlong_t)vd->vdev_id, vd,
	(u_longlong_t)dmu_tx_get_txg(tx),
	(u_longlong_t)vic->vic_mapping_object);

	spa_history_log_internal(spa, "vdev remove started", tx,
	"%s vdev %llu %s", spa_name(spa), (u_longlong_t)vd->vdev_id,
	(vd->vdev_path != NULL) ? vd->vdev_path : "-");
	/*
	* Setting spa_vdev_removal causes subsequent frees to call
	* free_from_removing_vdev(). Note that we don't need any locking
	* because we are the sync thread, and metaslab_free_impl() is only
	* called from syncing context (potentially from a zio taskq thread,
	* but in any case only when there are outstanding free i/os, which
	* there are not).
	*/
	ASSERT3P(spa->spa_vdev_removal, ==, NULL);
	spa->spa_vdev_removal = svr;
	svr->svr_thread = thread_create(NULL, 0,
	spa_vdev_remove_thread, spa, 0, &p0, TS_RUN, minclsyspri);
	}

	/*
	* When we are opening a pool, we must read the mapping for each
	* indirect vdev in order from most recently removed to least
	* recently removed. We do this because the blocks for the mapping
	* of older indirect vdevs may be stored on more recently removed vdevs.
	* In order to read each indirect mapping object, we must have
	* initialized all more recently removed vdevs.
	*/
	int
	spa_remove_init(spa_t *spa)
	{
	int error;

	error = zap_lookup(spa->spa_dsl_pool->dp_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_REMOVING, sizeof (uint64_t),
	sizeof (spa->spa_removing_phys) / sizeof (uint64_t),
	&spa->spa_removing_phys);

	if (error == ENOENT) {
	spa->spa_removing_phys.sr_state = DSS_NONE;
	spa->spa_removing_phys.sr_removing_vdev = -1;
	spa->spa_removing_phys.sr_prev_indirect_vdev = -1;
	spa->spa_indirect_vdevs_loaded = B_TRUE;
	return (0);
	} else if (error != 0) {
	return (error);
	}

	if (spa->spa_removing_phys.sr_state == DSS_SCANNING) {
	/*
	* We are currently removing a vdev. Create and
	* initialize a spa_vdev_removal_t from the bonus
	* buffer of the removing vdevs vdev_im_object, and
	* initialize its partial mapping.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	vdev_t *vd = vdev_lookup_top(spa,
	spa->spa_removing_phys.sr_removing_vdev);

	if (vd == NULL) {
	spa_config_exit(spa, SCL_STATE, FTAG);
	return (EINVAL);
	}

	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;

	ASSERT(vdev_is_concrete(vd));
	spa_vdev_removal_t *svr = spa_vdev_removal_create(vd);
	ASSERT3U(svr->svr_vdev_id, ==, vd->vdev_id);
	ASSERT(vd->vdev_removing);

	vd->vdev_indirect_mapping = vdev_indirect_mapping_open(
	spa->spa_meta_objset, vic->vic_mapping_object);
	vd->vdev_indirect_births = vdev_indirect_births_open(
	spa->spa_meta_objset, vic->vic_births_object);
	spa_config_exit(spa, SCL_STATE, FTAG);

	spa->spa_vdev_removal = svr;
	}

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	uint64_t indirect_vdev_id =
	spa->spa_removing_phys.sr_prev_indirect_vdev;
	while (indirect_vdev_id != UINT64_MAX) {
	vdev_t *vd = vdev_lookup_top(spa, indirect_vdev_id);
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	vd->vdev_indirect_mapping = vdev_indirect_mapping_open(
	spa->spa_meta_objset, vic->vic_mapping_object);
	vd->vdev_indirect_births = vdev_indirect_births_open(
	spa->spa_meta_objset, vic->vic_births_object);

	indirect_vdev_id = vic->vic_prev_indirect_vdev;
	}
	spa_config_exit(spa, SCL_STATE, FTAG);

	/*
	* Now that we've loaded all the indirect mappings, we can allow
	* reads from other blocks (e.g. via predictive prefetch).
	*/
	spa->spa_indirect_vdevs_loaded = B_TRUE;
	return (0);
	}

	void
	spa_restart_removal(spa_t *spa)
	{
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;

	if (svr == NULL)
	return;

	/*
	* In general when this function is called there is no
	* removal thread running. The only scenario where this
	* is not true is during spa_import() where this function
	* is called twice [once from spa_import_impl() and
	* spa_async_resume()]. Thus, in the scenario where we
	* import a pool that has an ongoing removal we don't
	* want to spawn a second thread.
	*/
	if (svr->svr_thread != NULL)
	return;

	if (!spa_writeable(spa))
	return;

	zfs_dbgmsg("restarting removal of %llu",
	(u_longlong_t)svr->svr_vdev_id);
	svr->svr_thread = thread_create(NULL, 0, spa_vdev_remove_thread, spa,
	0, &p0, TS_RUN, minclsyspri);
	}

	/*
	* Process freeing from a device which is in the middle of being removed.
	* We must handle this carefully so that we attempt to copy freed data,
	* and we correctly free already-copied data.
	*/
	void
	free_from_removing_vdev(vdev_t *vd, uint64_t offset, uint64_t size)
	{
	spa_t *spa = vd->vdev_spa;
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	uint64_t txg = spa_syncing_txg(spa);
	uint64_t max_offset_yet = 0;

	ASSERT(vd->vdev_indirect_config.vic_mapping_object != 0);
	ASSERT3U(vd->vdev_indirect_config.vic_mapping_object, ==,
	vdev_indirect_mapping_object(vim));
	ASSERT3U(vd->vdev_id, ==, svr->svr_vdev_id);

	mutex_enter(&svr->svr_lock);

	/*
	* Remove the segment from the removing vdev's spacemap. This
	* ensures that we will not attempt to copy this space (if the
	* removal thread has not yet visited it), and also ensures
	* that we know what is actually allocated on the new vdevs
	* (needed if we cancel the removal).
	*
	* Note: we must do the metaslab_free_concrete() with the svr_lock
	* held, so that the remove_thread can not load this metaslab and then
	* visit this offset between the time that we metaslab_free_concrete()
	* and when we check to see if it has been visited.
	*
	* Note: The checkpoint flag is set to false as having/taking
	* a checkpoint and removing a device can't happen at the same
	* time.
	*/
	ASSERT(!spa_has_checkpoint(spa));
	metaslab_free_concrete(vd, offset, size, B_FALSE);

	uint64_t synced_size = 0;
	uint64_t synced_offset = 0;
	uint64_t max_offset_synced = vdev_indirect_mapping_max_offset(vim);
	if (offset < max_offset_synced) {
	/*
	* The mapping for this offset is already on disk.
	* Free from the new location.
	*
	* Note that we use svr_max_synced_offset because it is
	* updated atomically with respect to the in-core mapping.
	* By contrast, vim_max_offset is not.
	*
	* This block may be split between a synced entry and an
	* in-flight or unvisited entry. Only process the synced
	* portion of it here.
	*/
	synced_size = MIN(size, max_offset_synced - offset);
	synced_offset = offset;

	ASSERT3U(max_offset_yet, <=, max_offset_synced);
	max_offset_yet = max_offset_synced;

	DTRACE_PROBE3(remove__free__synced,
	spa_t *, spa,
	uint64_t, offset,
	uint64_t, synced_size);

	size -= synced_size;
	offset += synced_size;
	}

	/*
	* Look at all in-flight txgs starting from the currently syncing one
	* and see if a section of this free is being copied. By starting from
	* this txg and iterating forward, we might find that this region
	* was copied in two different txgs and handle it appropriately.
	*/
	for (int i = 0; i < TXG_CONCURRENT_STATES; i++) {
	int txgoff = (txg + i) & TXG_MASK;
	if (size > 0 && offset < svr->svr_max_offset_to_sync[txgoff]) {
	/*
	* The mapping for this offset is in flight, and
	* will be synced in txg+i.
	*/
	uint64_t inflight_size = MIN(size,
	svr->svr_max_offset_to_sync[txgoff] - offset);

	DTRACE_PROBE4(remove__free__inflight,
	spa_t *, spa,
	uint64_t, offset,
	uint64_t, inflight_size,
	uint64_t, txg + i);

	/*
	* We copy data in order of increasing offset.
	* Therefore the max_offset_to_sync[] must increase
	* (or be zero, indicating that nothing is being
	* copied in that txg).
	*/
	if (svr->svr_max_offset_to_sync[txgoff] != 0) {
	ASSERT3U(svr->svr_max_offset_to_sync[txgoff],
	>=, max_offset_yet);
	max_offset_yet =
	svr->svr_max_offset_to_sync[txgoff];
	}

	/*
	* We've already committed to copying this segment:
	* we have allocated space elsewhere in the pool for
	* it and have an IO outstanding to copy the data. We
	* cannot free the space before the copy has
	* completed, or else the copy IO might overwrite any
	* new data. To free that space, we record the
	* segment in the appropriate svr_frees tree and free
	* the mapped space later, in the txg where we have
	* completed the copy and synced the mapping (see
	* vdev_mapping_sync).
	*/
	range_tree_add(svr->svr_frees[txgoff],
	offset, inflight_size);
	size -= inflight_size;
	offset += inflight_size;

	/*
	* This space is already accounted for as being
	* done, because it is being copied in txg+i.
	* However, if i!=0, then it is being copied in
	* a future txg. If we crash after this txg
	* syncs but before txg+i syncs, then the space
	* will be free. Therefore we must account
	* for the space being done in this txg
	* (when it is freed) rather than the future txg
	* (when it will be copied).
	*/
	ASSERT3U(svr->svr_bytes_done[txgoff], >=,
	inflight_size);
	svr->svr_bytes_done[txgoff] -= inflight_size;
	svr->svr_bytes_done[txg & TXG_MASK] += inflight_size;
	}
	}
	ASSERT0(svr->svr_max_offset_to_sync[TXG_CLEAN(txg) & TXG_MASK]);

	if (size > 0) {
	/*
	* The copy thread has not yet visited this offset. Ensure
	* that it doesn't.
	*/

	DTRACE_PROBE3(remove__free__unvisited,
	spa_t *, spa,
	uint64_t, offset,
	uint64_t, size);

	if (svr->svr_allocd_segs != NULL)
	range_tree_clear(svr->svr_allocd_segs, offset, size);

	/*
	* Since we now do not need to copy this data, for
	* accounting purposes we have done our job and can count
	* it as completed.
	*/
	svr->svr_bytes_done[txg & TXG_MASK] += size;
	}
	mutex_exit(&svr->svr_lock);

	/*
	* Now that we have dropped svr_lock, process the synced portion
	* of this free.
	*/
	if (synced_size > 0) {
	vdev_indirect_mark_obsolete(vd, synced_offset, synced_size);

	/*
	* Note: this can only be called from syncing context,
	* and the vdev_indirect_mapping is only changed from the
	* sync thread, so we don't need svr_lock while doing
	* metaslab_free_impl_cb.
	*/
	boolean_t checkpoint = B_FALSE;
	vdev_indirect_ops.vdev_op_remap(vd, synced_offset, synced_size,
	metaslab_free_impl_cb, &checkpoint);
	}
	}

	/*
	* Stop an active removal and update the spa_removing phys.
	*/
	static void
	spa_finish_removal(spa_t spa, dsl_scan_state_t state, dmu_tx_t tx)
	{
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	ASSERT3U(dmu_tx_get_txg(tx), ==, spa_syncing_txg(spa));

	/* Ensure the removal thread has completed before we free the svr. */
	spa_vdev_remove_suspend(spa);

	ASSERT(state == DSS_FINISHED \|\| state == DSS_CANCELED);

	if (state == DSS_FINISHED) {
	spa_removing_phys_t *srp = &spa->spa_removing_phys;
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;

	if (srp->sr_prev_indirect_vdev != -1) {
	vdev_t *pvd;
	pvd = vdev_lookup_top(spa,
	srp->sr_prev_indirect_vdev);
	ASSERT3P(pvd->vdev_ops, ==, &vdev_indirect_ops);
	}

	vic->vic_prev_indirect_vdev = srp->sr_prev_indirect_vdev;
	srp->sr_prev_indirect_vdev = vd->vdev_id;
	}
	spa->spa_removing_phys.sr_state = state;
	spa->spa_removing_phys.sr_end_time = gethrestime_sec();

	spa->spa_vdev_removal = NULL;
	spa_vdev_removal_destroy(svr);

	spa_sync_removing_state(spa, tx);
	spa_notify_waiters(spa);

	vdev_config_dirty(spa->spa_root_vdev);
	}

	static void
	free_mapped_segment_cb(void *arg, uint64_t offset, uint64_t size)
	{
	vdev_t *vd = arg;
	vdev_indirect_mark_obsolete(vd, offset, size);
	boolean_t checkpoint = B_FALSE;
	vdev_indirect_ops.vdev_op_remap(vd, offset, size,
	metaslab_free_impl_cb, &checkpoint);
	}

	/*
	* On behalf of the removal thread, syncs an incremental bit more of
	* the indirect mapping to disk and updates the in-memory mapping.
	* Called as a sync task in every txg that the removal thread makes progress.
	*/
	static void
	vdev_mapping_sync(void arg, dmu_tx_t tx)
	{
	spa_vdev_removal_t *svr = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
	vdev_indirect_config_t *vic __maybe_unused = &vd->vdev_indirect_config;
	uint64_t txg = dmu_tx_get_txg(tx);
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	ASSERT(vic->vic_mapping_object != 0);
	ASSERT3U(txg, ==, spa_syncing_txg(spa));

	vdev_indirect_mapping_add_entries(vim,
	&svr->svr_new_segments[txg & TXG_MASK], tx);
	vdev_indirect_births_add_entry(vd->vdev_indirect_births,
	vdev_indirect_mapping_max_offset(vim), dmu_tx_get_txg(tx), tx);

	/*
	* Free the copied data for anything that was freed while the
	* mapping entries were in flight.
	*/
	mutex_enter(&svr->svr_lock);
	range_tree_vacate(svr->svr_frees[txg & TXG_MASK],
	free_mapped_segment_cb, vd);
	ASSERT3U(svr->svr_max_offset_to_sync[txg & TXG_MASK], >=,
	vdev_indirect_mapping_max_offset(vim));
	svr->svr_max_offset_to_sync[txg & TXG_MASK] = 0;
	mutex_exit(&svr->svr_lock);

	spa_sync_removing_state(spa, tx);
	}

	typedef struct vdev_copy_segment_arg {
	spa_t *vcsa_spa;
	dva_t *vcsa_dest_dva;
	uint64_t vcsa_txg;
	range_tree_t *vcsa_obsolete_segs;
	} vdev_copy_segment_arg_t;

	static void
	unalloc_seg(void *arg, uint64_t start, uint64_t size)
	{
	vdev_copy_segment_arg_t *vcsa = arg;
	spa_t *spa = vcsa->vcsa_spa;
	blkptr_t bp = { { { {0} } } };

	BP_SET_BIRTH(&bp, TXG_INITIAL, TXG_INITIAL);
	BP_SET_LSIZE(&bp, size);
	BP_SET_PSIZE(&bp, size);
	BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_OFF);
	BP_SET_TYPE(&bp, DMU_OT_NONE);
	BP_SET_LEVEL(&bp, 0);
	BP_SET_DEDUP(&bp, 0);
	BP_SET_BYTEORDER(&bp, ZFS_HOST_BYTEORDER);

	DVA_SET_VDEV(&bp.blk_dva[0], DVA_GET_VDEV(vcsa->vcsa_dest_dva));
	DVA_SET_OFFSET(&bp.blk_dva[0],
	DVA_GET_OFFSET(vcsa->vcsa_dest_dva) + start);
	DVA_SET_ASIZE(&bp.blk_dva[0], size);

	zio_free(spa, vcsa->vcsa_txg, &bp);
	}

	/*
	* All reads and writes associated with a call to spa_vdev_copy_segment()
	* are done.
	*/
	static void
	spa_vdev_copy_segment_done(zio_t *zio)
	{
	vdev_copy_segment_arg_t *vcsa = zio->io_private;

	range_tree_vacate(vcsa->vcsa_obsolete_segs,
	unalloc_seg, vcsa);
	range_tree_destroy(vcsa->vcsa_obsolete_segs);
	kmem_free(vcsa, sizeof (*vcsa));

	spa_config_exit(zio->io_spa, SCL_STATE, zio->io_spa);
	}

	/*
	* The write of the new location is done.
	*/
	static void
	spa_vdev_copy_segment_write_done(zio_t *zio)
	{
	vdev_copy_arg_t *vca = zio->io_private;

	abd_free(zio->io_abd);

	mutex_enter(&vca->vca_lock);
	vca->vca_outstanding_bytes -= zio->io_size;

	if (zio->io_error != 0)
	vca->vca_write_error_bytes += zio->io_size;

	cv_signal(&vca->vca_cv);
	mutex_exit(&vca->vca_lock);
	}

	/*
	* The read of the old location is done. The parent zio is the write to
	* the new location. Allow it to start.
	*/
	static void
	spa_vdev_copy_segment_read_done(zio_t *zio)
	{
	vdev_copy_arg_t *vca = zio->io_private;

	if (zio->io_error != 0) {
	mutex_enter(&vca->vca_lock);
	vca->vca_read_error_bytes += zio->io_size;
	mutex_exit(&vca->vca_lock);
	}

	zio_nowait(zio_unique_parent(zio));
	}

	/*
	* If the old and new vdevs are mirrors, we will read both sides of the old
	* mirror, and write each copy to the corresponding side of the new mirror.
	* If the old and new vdevs have a different number of children, we will do
	* this as best as possible. Since we aren't verifying checksums, this
	* ensures that as long as there's a good copy of the data, we'll have a
	* good copy after the removal, even if there's silent damage to one side
	* of the mirror. If we're removing a mirror that has some silent damage,
	* we'll have exactly the same damage in the new location (assuming that
	* the new location is also a mirror).
	*
	* We accomplish this by creating a tree of zio_t's, with as many writes as
	* there are "children" of the new vdev (a non-redundant vdev counts as one
	* child, a 2-way mirror has 2 children, etc). Each write has an associated
	* read from a child of the old vdev. Typically there will be the same
	* number of children of the old and new vdevs. However, if there are more
	* children of the new vdev, some child(ren) of the old vdev will be issued
	* multiple reads. If there are more children of the old vdev, some copies
	* will be dropped.
	*
	* For example, the tree of zio_t's for a 2-way mirror is:
	*
	* null
	* / \
	* write(new vdev, child 0) write(new vdev, child 1)
	* \| \|
	* read(old vdev, child 0) read(old vdev, child 1)
	*
	* Child zio's complete before their parents complete. However, zio's
	* created with zio_vdev_child_io() may be issued before their children
	* complete. In this case we need to make sure that the children (reads)
	* complete before the parents (writes) are issued. We do this by not
	* calling zio_nowait() on each write until its corresponding read has
	* completed.
	*
	* The spa_config_lock must be held while zio's created by
	* zio_vdev_child_io() are in progress, to ensure that the vdev tree does
	* not change (e.g. due to a concurrent "zpool attach/detach"). The "null"
	* zio is needed to release the spa_config_lock after all the reads and
	* writes complete. (Note that we can't grab the config lock for each read,
	* because it is not reentrant - we could deadlock with a thread waiting
	* for a write lock.)
	*/
	static void
	spa_vdev_copy_one_child(vdev_copy_arg_t vca, zio_t nzio,
	vdev_t *source_vd, uint64_t source_offset,
	vdev_t *dest_child_vd, uint64_t dest_offset, int dest_id, uint64_t size)
	{
	ASSERT3U(spa_config_held(nzio->io_spa, SCL_ALL, RW_READER), !=, 0);

	/*
	* If the destination child in unwritable then there is no point
	* in issuing the source reads which cannot be written.
	*/
	if (!vdev_writeable(dest_child_vd))
	return;

	mutex_enter(&vca->vca_lock);
	vca->vca_outstanding_bytes += size;
	mutex_exit(&vca->vca_lock);

	abd_t *abd = abd_alloc_for_io(size, B_FALSE);

	vdev_t *source_child_vd = NULL;
	if (source_vd->vdev_ops == &vdev_mirror_ops && dest_id != -1) {
	/*
	* Source and dest are both mirrors. Copy from the same
	* child id as we are copying to (wrapping around if there
	* are more dest children than source children). If the
	* preferred source child is unreadable select another.
	*/
	for (int i = 0; i < source_vd->vdev_children; i++) {
	source_child_vd = source_vd->vdev_child[
	(dest_id + i) % source_vd->vdev_children];
	if (vdev_readable(source_child_vd))
	break;
	}
	} else {
	source_child_vd = source_vd;
	}

	/*
	* There should always be at least one readable source child or
	* the pool would be in a suspended state. Somehow selecting an
	* unreadable child would result in IO errors, the removal process
	* being cancelled, and the pool reverting to its pre-removal state.
	*/
	ASSERT3P(source_child_vd, !=, NULL);

	zio_t *write_zio = zio_vdev_child_io(nzio, NULL,
	dest_child_vd, dest_offset, abd, size,
	ZIO_TYPE_WRITE, ZIO_PRIORITY_REMOVAL,
	ZIO_FLAG_CANFAIL,
	spa_vdev_copy_segment_write_done, vca);

	zio_nowait(zio_vdev_child_io(write_zio, NULL,
	source_child_vd, source_offset, abd, size,
	ZIO_TYPE_READ, ZIO_PRIORITY_REMOVAL,
	ZIO_FLAG_CANFAIL,
	spa_vdev_copy_segment_read_done, vca));
	}

	/*
	* Allocate a new location for this segment, and create the zio_t's to
	* read from the old location and write to the new location.
	*/
	static int
	spa_vdev_copy_segment(vdev_t vd, range_tree_t segs,
	uint64_t maxalloc, uint64_t txg,
	vdev_copy_arg_t vca, zio_alloc_list_t zal)
	{
	metaslab_group_t *mg = vd->vdev_mg;
	spa_t *spa = vd->vdev_spa;
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	vdev_indirect_mapping_entry_t *entry;
	dva_t dst = {{ 0 }};
	uint64_t start = range_tree_min(segs);
	ASSERT0(P2PHASE(start, 1 << spa->spa_min_ashift));

	ASSERT3U(maxalloc, <=, SPA_MAXBLOCKSIZE);
	ASSERT0(P2PHASE(maxalloc, 1 << spa->spa_min_ashift));

	uint64_t size = range_tree_span(segs);
	if (range_tree_span(segs) > maxalloc) {
	/*
	* We can't allocate all the segments. Prefer to end
	* the allocation at the end of a segment, thus avoiding
	* additional split blocks.
	*/
	range_seg_max_t search;
	zfs_btree_index_t where;
	rs_set_start(&search, segs, start + maxalloc);
	rs_set_end(&search, segs, start + maxalloc);
	(void) zfs_btree_find(&segs->rt_root, &search, &where);
	range_seg_t *rs = zfs_btree_prev(&segs->rt_root, &where,
	&where);
	if (rs != NULL) {
	size = rs_get_end(rs, segs) - start;
	} else {
	/*
	* There are no segments that end before maxalloc.
	* I.e. the first segment is larger than maxalloc,
	* so we must split it.
	*/
	size = maxalloc;
	}
	}
	ASSERT3U(size, <=, maxalloc);
	ASSERT0(P2PHASE(size, 1 << spa->spa_min_ashift));

	/*
	* An allocation class might not have any remaining vdevs or space
	*/
	metaslab_class_t *mc = mg->mg_class;
	if (mc->mc_groups == 0)
	mc = spa_normal_class(spa);
	int error = metaslab_alloc_dva(spa, mc, size, &dst, 0, NULL, txg, 0,
	zal, 0);
	if (error == ENOSPC && mc != spa_normal_class(spa)) {
	error = metaslab_alloc_dva(spa, spa_normal_class(spa), size,
	&dst, 0, NULL, txg, 0, zal, 0);
	}
	if (error != 0)
	return (error);

	/*
	* Determine the ranges that are not actually needed. Offsets are
	* relative to the start of the range to be copied (i.e. relative to the
	* local variable "start").
	*/
	range_tree_t *obsolete_segs = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);

	zfs_btree_index_t where;
	range_seg_t *rs = zfs_btree_first(&segs->rt_root, &where);
	ASSERT3U(rs_get_start(rs, segs), ==, start);
	uint64_t prev_seg_end = rs_get_end(rs, segs);
	while ((rs = zfs_btree_next(&segs->rt_root, &where, &where)) != NULL) {
	if (rs_get_start(rs, segs) >= start + size) {
	break;
	} else {
	range_tree_add(obsolete_segs,
	prev_seg_end - start,
	rs_get_start(rs, segs) - prev_seg_end);
	}
	prev_seg_end = rs_get_end(rs, segs);
	}
	/* We don't end in the middle of an obsolete range */
	ASSERT3U(start + size, <=, prev_seg_end);

	range_tree_clear(segs, start, size);

	/*
	* We can't have any padding of the allocated size, otherwise we will
	* misunderstand what's allocated, and the size of the mapping. We
	* prevent padding by ensuring that all devices in the pool have the
	* same ashift, and the allocation size is a multiple of the ashift.
	*/
	VERIFY3U(DVA_GET_ASIZE(&dst), ==, size);

	entry = kmem_zalloc(sizeof (vdev_indirect_mapping_entry_t), KM_SLEEP);
	DVA_MAPPING_SET_SRC_OFFSET(&entry->vime_mapping, start);
	entry->vime_mapping.vimep_dst = dst;
	if (spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS)) {
	entry->vime_obsolete_count = range_tree_space(obsolete_segs);
	}

	vdev_copy_segment_arg_t vcsa = kmem_zalloc(sizeof (vcsa), KM_SLEEP);
	vcsa->vcsa_dest_dva = &entry->vime_mapping.vimep_dst;
	vcsa->vcsa_obsolete_segs = obsolete_segs;
	vcsa->vcsa_spa = spa;
	vcsa->vcsa_txg = txg;

	/*
	* See comment before spa_vdev_copy_one_child().
	*/
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);
	zio_t *nzio = zio_null(spa->spa_txg_zio[txg & TXG_MASK], spa, NULL,
	spa_vdev_copy_segment_done, vcsa, 0);
	vdev_t *dest_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dst));
	if (dest_vd->vdev_ops == &vdev_mirror_ops) {
	for (int i = 0; i < dest_vd->vdev_children; i++) {
	vdev_t *child = dest_vd->vdev_child[i];
	spa_vdev_copy_one_child(vca, nzio, vd, start,
	child, DVA_GET_OFFSET(&dst), i, size);
	}
	} else {
	spa_vdev_copy_one_child(vca, nzio, vd, start,
	dest_vd, DVA_GET_OFFSET(&dst), -1, size);
	}
	zio_nowait(nzio);

	list_insert_tail(&svr->svr_new_segments[txg & TXG_MASK], entry);
	ASSERT3U(start + size, <=, vd->vdev_ms_count << vd->vdev_ms_shift);
	vdev_dirty(vd, 0, NULL, txg);

	return (0);
	}

	/*
	* Complete the removal of a toplevel vdev. This is called as a
	* synctask in the same txg that we will sync out the new config (to the
	* MOS object) which indicates that this vdev is indirect.
	*/
	static void
	vdev_remove_complete_sync(void arg, dmu_tx_t tx)
	{
	spa_vdev_removal_t *svr = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT0(svr->svr_bytes_done[i]);
	}

	ASSERT3U(spa->spa_removing_phys.sr_copied, ==,
	spa->spa_removing_phys.sr_to_copy);

	vdev_destroy_spacemaps(vd, tx);

	/* destroy leaf zaps, if any */
	ASSERT3P(svr->svr_zaplist, !=, NULL);
	for (nvpair_t *pair = nvlist_next_nvpair(svr->svr_zaplist, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(svr->svr_zaplist, pair)) {
	vdev_destroy_unlink_zap(vd, fnvpair_value_uint64(pair), tx);
	}
	fnvlist_free(svr->svr_zaplist);

	spa_finish_removal(dmu_tx_pool(tx)->dp_spa, DSS_FINISHED, tx);
	/* vd->vdev_path is not available here */
	spa_history_log_internal(spa, "vdev remove completed", tx,
	"%s vdev %llu", spa_name(spa), (u_longlong_t)vd->vdev_id);
	}

	static void
	vdev_remove_enlist_zaps(vdev_t vd, nvlist_t zlist)
	{
	ASSERT3P(zlist, !=, NULL);
	ASSERT0(vdev_get_nparity(vd));

	if (vd->vdev_leaf_zap != 0) {
	char zkey[32];
	(void) snprintf(zkey, sizeof (zkey), "%s-%llu",
	VDEV_REMOVAL_ZAP_OBJS, (u_longlong_t)vd->vdev_leaf_zap);
	fnvlist_add_uint64(zlist, zkey, vd->vdev_leaf_zap);
	}

	for (uint64_t id = 0; id < vd->vdev_children; id++) {
	vdev_remove_enlist_zaps(vd->vdev_child[id], zlist);
	}
	}

	static void
	vdev_remove_replace_with_indirect(vdev_t *vd, uint64_t txg)
	{
	vdev_t *ivd;
	dmu_tx_t *tx;
	spa_t *spa = vd->vdev_spa;
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;

	/*
	* First, build a list of leaf zaps to be destroyed.
	* This is passed to the sync context thread,
	* which does the actual unlinking.
	*/
	svr->svr_zaplist = fnvlist_alloc();
	vdev_remove_enlist_zaps(vd, svr->svr_zaplist);

	ivd = vdev_add_parent(vd, &vdev_indirect_ops);
	ivd->vdev_removing = 0;

	vd->vdev_leaf_zap = 0;

	vdev_remove_child(ivd, vd);
	vdev_compact_children(ivd);

	ASSERT(!list_link_active(&vd->vdev_state_dirty_node));

	mutex_enter(&svr->svr_lock);
	svr->svr_thread = NULL;
	cv_broadcast(&svr->svr_cv);
	mutex_exit(&svr->svr_lock);

	/* After this, we can not use svr. */
	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	dsl_sync_task_nowait(spa->spa_dsl_pool,
	vdev_remove_complete_sync, svr, tx);
	dmu_tx_commit(tx);
	}

	/*
	* Complete the removal of a toplevel vdev. This is called in open
	* context by the removal thread after we have copied all vdev's data.
	*/
	static void
	vdev_remove_complete(spa_t *spa)
	{
	uint64_t txg;

	/*
	* Wait for any deferred frees to be synced before we call
	* vdev_metaslab_fini()
	*/
	txg_wait_synced(spa->spa_dsl_pool, 0);
	txg = spa_vdev_enter(spa);
	vdev_t *vd = vdev_lookup_top(spa, spa->spa_vdev_removal->svr_vdev_id);
	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	ASSERT3P(vd->vdev_autotrim_thread, ==, NULL);
	vdev_rebuild_stop_wait(vd);
	ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);
	uint64_t vdev_space = spa_deflate(spa) ?
	vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;

	sysevent_t *ev = spa_event_create(spa, vd, NULL,
	ESC_ZFS_VDEV_REMOVE_DEV);

	zfs_dbgmsg("finishing device removal for vdev %llu in txg %llu",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)txg);

	ASSERT3U(0, !=, vdev_space);
	ASSERT3U(spa->spa_nonallocating_dspace, >=, vdev_space);

	/* the vdev is no longer part of the dspace */
	spa->spa_nonallocating_dspace -= vdev_space;

	/*
	* Discard allocation state.
	*/
	if (vd->vdev_mg != NULL) {
	vdev_metaslab_fini(vd);
	metaslab_group_destroy(vd->vdev_mg);
	vd->vdev_mg = NULL;
	}
	if (vd->vdev_log_mg != NULL) {
	ASSERT0(vd->vdev_ms_count);
	metaslab_group_destroy(vd->vdev_log_mg);
	vd->vdev_log_mg = NULL;
	}
	ASSERT0(vd->vdev_stat.vs_space);
	ASSERT0(vd->vdev_stat.vs_dspace);

	vdev_remove_replace_with_indirect(vd, txg);

	/*
	* We now release the locks, allowing spa_sync to run and finish the
	* removal via vdev_remove_complete_sync in syncing context.
	*
	* Note that we hold on to the vdev_t that has been replaced. Since
	* it isn't part of the vdev tree any longer, it can't be concurrently
	* manipulated, even while we don't have the config lock.
	*/
	(void) spa_vdev_exit(spa, NULL, txg, 0);

	/*
	* Top ZAP should have been transferred to the indirect vdev in
	* vdev_remove_replace_with_indirect.
	*/
	ASSERT0(vd->vdev_top_zap);

	/*
	* Leaf ZAP should have been moved in vdev_remove_replace_with_indirect.
	*/
	ASSERT0(vd->vdev_leaf_zap);

	txg = spa_vdev_enter(spa);
	(void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);
	/*
	* Request to update the config and the config cachefile.
	*/
	vdev_config_dirty(spa->spa_root_vdev);
	(void) spa_vdev_exit(spa, vd, txg, 0);

	if (ev != NULL)
	spa_event_post(ev);
	}

	/*
	* Evacuates a segment of size at most max_alloc from the vdev
	* via repeated calls to spa_vdev_copy_segment. If an allocation
	* fails, the pool is probably too fragmented to handle such a
	* large size, so decrease max_alloc so that the caller will not try
	* this size again this txg.
	*/
	static void
	spa_vdev_copy_impl(vdev_t vd, spa_vdev_removal_t svr, vdev_copy_arg_t *vca,
	uint64_t max_alloc, dmu_tx_t tx)
	{
	uint64_t txg = dmu_tx_get_txg(tx);
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	mutex_enter(&svr->svr_lock);

	/*
	* Determine how big of a chunk to copy. We can allocate up
	* to max_alloc bytes, and we can span up to vdev_removal_max_span
	* bytes of unallocated space at a time. "segs" will track the
	* allocated segments that we are copying. We may also be copying
	* free segments (of up to vdev_removal_max_span bytes).
	*/
	range_tree_t *segs = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	for (;;) {
	range_tree_t *rt = svr->svr_allocd_segs;
	range_seg_t *rs = range_tree_first(rt);

	if (rs == NULL)
	break;

	uint64_t seg_length;

	if (range_tree_is_empty(segs)) {
	/* need to truncate the first seg based on max_alloc */
	seg_length = MIN(rs_get_end(rs, rt) - rs_get_start(rs,
	rt), *max_alloc);
	} else {
	if (rs_get_start(rs, rt) - range_tree_max(segs) >
	vdev_removal_max_span) {
	/*
	* Including this segment would cause us to
	* copy a larger unneeded chunk than is allowed.
	*/
	break;
	} else if (rs_get_end(rs, rt) - range_tree_min(segs) >
	*max_alloc) {
	/*
	* This additional segment would extend past
	* max_alloc. Rather than splitting this
	* segment, leave it for the next mapping.
	*/
	break;
	} else {
	seg_length = rs_get_end(rs, rt) -
	rs_get_start(rs, rt);
	}
	}

	range_tree_add(segs, rs_get_start(rs, rt), seg_length);
	range_tree_remove(svr->svr_allocd_segs,
	rs_get_start(rs, rt), seg_length);
	}

	if (range_tree_is_empty(segs)) {
	mutex_exit(&svr->svr_lock);
	range_tree_destroy(segs);
	return;
	}

	if (svr->svr_max_offset_to_sync[txg & TXG_MASK] == 0) {
	dsl_sync_task_nowait(dmu_tx_pool(tx), vdev_mapping_sync,
	svr, tx);
	}

	svr->svr_max_offset_to_sync[txg & TXG_MASK] = range_tree_max(segs);

	/*
	* Note: this is the amount of allocated space
	* that we are taking care of each txg.
	*/
	svr->svr_bytes_done[txg & TXG_MASK] += range_tree_space(segs);

	mutex_exit(&svr->svr_lock);

	zio_alloc_list_t zal;
	metaslab_trace_init(&zal);
	uint64_t thismax = SPA_MAXBLOCKSIZE;
	while (!range_tree_is_empty(segs)) {
	int error = spa_vdev_copy_segment(vd,
	segs, thismax, txg, vca, &zal);

	if (error == ENOSPC) {
	/*
	* Cut our segment in half, and don't try this
	* segment size again this txg. Note that the
	* allocation size must be aligned to the highest
	* ashift in the pool, so that the allocation will
	* not be padded out to a multiple of the ashift,
	* which could cause us to think that this mapping
	* is larger than we intended.
	*/
	ASSERT3U(spa->spa_max_ashift, >=, SPA_MINBLOCKSHIFT);
	ASSERT3U(spa->spa_max_ashift, ==, spa->spa_min_ashift);
	uint64_t attempted =
	MIN(range_tree_span(segs), thismax);
	thismax = P2ROUNDUP(attempted / 2,
	1 << spa->spa_max_ashift);
	/*
	* The minimum-size allocation can not fail.
	*/
	ASSERT3U(attempted, >, 1 << spa->spa_max_ashift);
	*max_alloc = attempted - (1 << spa->spa_max_ashift);
	} else {
	ASSERT0(error);

	/*
	* We've performed an allocation, so reset the
	* alloc trace list.
	*/
	metaslab_trace_fini(&zal);
	metaslab_trace_init(&zal);
	}
	}
	metaslab_trace_fini(&zal);
	range_tree_destroy(segs);
	}

	/*
	* The size of each removal mapping is limited by the tunable
	* zfs_remove_max_segment, but we must adjust this to be a multiple of the
	* pool's ashift, so that we don't try to split individual sectors regardless
	* of the tunable value. (Note that device removal requires that all devices
	* have the same ashift, so there's no difference between spa_min_ashift and
	* spa_max_ashift.) The raw tunable should not be used elsewhere.
	*/
	uint64_t
	spa_remove_max_segment(spa_t *spa)
	{
	return (P2ROUNDUP(zfs_remove_max_segment, 1 << spa->spa_max_ashift));
	}

	/*
	* The removal thread operates in open context. It iterates over all
	* allocated space in the vdev, by loading each metaslab's spacemap.
	* For each contiguous segment of allocated space (capping the segment
	* size at SPA_MAXBLOCKSIZE), we:
	* - Allocate space for it on another vdev.
	* - Create a new mapping from the old location to the new location
	* (as a record in svr_new_segments).
	* - Initiate a physical read zio to get the data off the removing disk.
	* - In the read zio's done callback, initiate a physical write zio to
	* write it to the new vdev.
	* Note that all of this will take effect when a particular TXG syncs.
	* The sync thread ensures that all the phys reads and writes for the syncing
	* TXG have completed (see spa_txg_zio) and writes the new mappings to disk
	* (see vdev_mapping_sync()).
	*/
	static __attribute__((noreturn)) void
	spa_vdev_remove_thread(void *arg)
	{
	spa_t *spa = arg;
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	vdev_copy_arg_t vca;
	uint64_t max_alloc = spa_remove_max_segment(spa);
	uint64_t last_txg = 0;

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	uint64_t start_offset = vdev_indirect_mapping_max_offset(vim);

	ASSERT3P(vd->vdev_ops, !=, &vdev_indirect_ops);
	ASSERT(vdev_is_concrete(vd));
	ASSERT(vd->vdev_removing);
	ASSERT(vd->vdev_indirect_config.vic_mapping_object != 0);
	ASSERT(vim != NULL);

	mutex_init(&vca.vca_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vca.vca_cv, NULL, CV_DEFAULT, NULL);
	vca.vca_outstanding_bytes = 0;
	vca.vca_read_error_bytes = 0;
	vca.vca_write_error_bytes = 0;

	mutex_enter(&svr->svr_lock);

	/*
	* Start from vim_max_offset so we pick up where we left off
	* if we are restarting the removal after opening the pool.
	*/
	uint64_t msi;
	for (msi = start_offset >> vd->vdev_ms_shift;
	msi < vd->vdev_ms_count && !svr->svr_thread_exit; msi++) {
	metaslab_t *msp = vd->vdev_ms[msi];
	ASSERT3U(msi, <=, vd->vdev_ms_count);

	ASSERT0(range_tree_space(svr->svr_allocd_segs));

	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	/*
	* Assert nothing in flight -- ms_*tree is empty.
	*/
	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT0(range_tree_space(msp->ms_allocating[i]));
	}

	/*
	* If the metaslab has ever been allocated from (ms_sm!=NULL),
	* read the allocated segments from the space map object
	* into svr_allocd_segs. Since we do this while holding
	* svr_lock and ms_sync_lock, concurrent frees (which
	* would have modified the space map) will wait for us
	* to finish loading the spacemap, and then take the
	* appropriate action (see free_from_removing_vdev()).
	*/
	if (msp->ms_sm != NULL) {
	VERIFY0(space_map_load(msp->ms_sm,
	svr->svr_allocd_segs, SM_ALLOC));

	range_tree_walk(msp->ms_unflushed_allocs,
	range_tree_add, svr->svr_allocd_segs);
	range_tree_walk(msp->ms_unflushed_frees,
	range_tree_remove, svr->svr_allocd_segs);
	range_tree_walk(msp->ms_freeing,
	range_tree_remove, svr->svr_allocd_segs);

	/*
	* When we are resuming from a paused removal (i.e.
	* when importing a pool with a removal in progress),
	* discard any state that we have already processed.
	*/
	range_tree_clear(svr->svr_allocd_segs, 0, start_offset);
	}
	mutex_exit(&msp->ms_lock);
	mutex_exit(&msp->ms_sync_lock);

	vca.vca_msp = msp;
	zfs_dbgmsg("copying %llu segments for metaslab %llu",
	(u_longlong_t)zfs_btree_numnodes(
	&svr->svr_allocd_segs->rt_root),
	(u_longlong_t)msp->ms_id);

	while (!svr->svr_thread_exit &&
	!range_tree_is_empty(svr->svr_allocd_segs)) {

	mutex_exit(&svr->svr_lock);

	/*
	* We need to periodically drop the config lock so that
	* writers can get in. Additionally, we can't wait
	* for a txg to sync while holding a config lock
	* (since a waiting writer could cause a 3-way deadlock
	* with the sync thread, which also gets a config
	* lock for reader). So we can't hold the config lock
	* while calling dmu_tx_assign().
	*/
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* This delay will pause the removal around the point
	* specified by zfs_removal_suspend_progress. We do this
	* solely from the test suite or during debugging.
	*/
	while (zfs_removal_suspend_progress &&
	!svr->svr_thread_exit)
	delay(hz);

	mutex_enter(&vca.vca_lock);
	while (vca.vca_outstanding_bytes >
	zfs_remove_max_copy_bytes) {
	cv_wait(&vca.vca_cv, &vca.vca_lock);
	}
	mutex_exit(&vca.vca_lock);

	dmu_tx_t *tx =
	dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);

	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);

	/*
	* Reacquire the vdev_config lock. The vdev_t
	* that we're removing may have changed, e.g. due
	* to a vdev_attach or vdev_detach.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vd = vdev_lookup_top(spa, svr->svr_vdev_id);

	if (txg != last_txg)
	max_alloc = spa_remove_max_segment(spa);
	last_txg = txg;

	spa_vdev_copy_impl(vd, svr, &vca, &max_alloc, tx);

	dmu_tx_commit(tx);
	mutex_enter(&svr->svr_lock);
	}

	mutex_enter(&vca.vca_lock);
	if (zfs_removal_ignore_errors == 0 &&
	(vca.vca_read_error_bytes > 0 \|\|
	vca.vca_write_error_bytes > 0)) {
	svr->svr_thread_exit = B_TRUE;
	}
	mutex_exit(&vca.vca_lock);
	}

	mutex_exit(&svr->svr_lock);

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* Wait for all copies to finish before cleaning up the vca.
	*/
	txg_wait_synced(spa->spa_dsl_pool, 0);
	ASSERT0(vca.vca_outstanding_bytes);

	mutex_destroy(&vca.vca_lock);
	cv_destroy(&vca.vca_cv);

	if (svr->svr_thread_exit) {
	mutex_enter(&svr->svr_lock);
	range_tree_vacate(svr->svr_allocd_segs, NULL, NULL);
	svr->svr_thread = NULL;
	cv_broadcast(&svr->svr_cv);
	mutex_exit(&svr->svr_lock);

	/*
	* During the removal process an unrecoverable read or write
	* error was encountered. The removal process must be
	* cancelled or this damage may become permanent.
	*/
	if (zfs_removal_ignore_errors == 0 &&
	(vca.vca_read_error_bytes > 0 \|\|
	vca.vca_write_error_bytes > 0)) {
	zfs_dbgmsg("canceling removal due to IO errors: "
	"[read_error_bytes=%llu] [write_error_bytes=%llu]",
	(u_longlong_t)vca.vca_read_error_bytes,
	(u_longlong_t)vca.vca_write_error_bytes);
	spa_vdev_remove_cancel_impl(spa);
	}
	} else {
	ASSERT0(range_tree_space(svr->svr_allocd_segs));
	vdev_remove_complete(spa);
	}

	thread_exit();
	}

	void
	spa_vdev_remove_suspend(spa_t *spa)
	{
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;

	if (svr == NULL)
	return;

	mutex_enter(&svr->svr_lock);
	svr->svr_thread_exit = B_TRUE;
	while (svr->svr_thread != NULL)
	cv_wait(&svr->svr_cv, &svr->svr_lock);
	svr->svr_thread_exit = B_FALSE;
	mutex_exit(&svr->svr_lock);
	}

	/*
	* Return true if the "allocating" property has been set to "off"
	*/
	static boolean_t
	vdev_prop_allocating_off(vdev_t *vd)
	{
	uint64_t objid = vd->vdev_top_zap;
	uint64_t allocating = 1;

	/* no vdev property object => no props */
	if (objid != 0) {
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;

	mutex_enter(&spa->spa_props_lock);
	(void) zap_lookup(mos, objid, "allocating", sizeof (uint64_t),
	1, &allocating);
	mutex_exit(&spa->spa_props_lock);
	}
	return (allocating == 0);
	}

	static int
	spa_vdev_remove_cancel_check(void arg, dmu_tx_t tx)
	{
	(void) arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	if (spa->spa_vdev_removal == NULL)
	return (ENOTACTIVE);
	return (0);
	}

	/*
	* Cancel a removal by freeing all entries from the partial mapping
	* and marking the vdev as no longer being removing.
	*/
	static void
	spa_vdev_remove_cancel_sync(void arg, dmu_tx_t tx)
	{
	(void) arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	vdev_t *vd = vdev_lookup_top(spa, svr->svr_vdev_id);
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
	objset_t *mos = spa->spa_meta_objset;

	ASSERT3P(svr->svr_thread, ==, NULL);

	spa_feature_decr(spa, SPA_FEATURE_DEVICE_REMOVAL, tx);

	boolean_t are_precise;
	VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	if (are_precise) {
	spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_OBSOLETE_COUNTS_ARE_PRECISE, tx));
	}

	uint64_t obsolete_sm_object;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	ASSERT(vd->vdev_obsolete_sm != NULL);
	ASSERT3U(obsolete_sm_object, ==,
	space_map_object(vd->vdev_obsolete_sm));

	space_map_free(vd->vdev_obsolete_sm, tx);
	VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, tx));
	space_map_close(vd->vdev_obsolete_sm);
	vd->vdev_obsolete_sm = NULL;
	spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
	}
	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT(list_is_empty(&svr->svr_new_segments[i]));
	ASSERT3U(svr->svr_max_offset_to_sync[i], <=,
	vdev_indirect_mapping_max_offset(vim));
	}

	for (uint64_t msi = 0; msi < vd->vdev_ms_count; msi++) {
	metaslab_t *msp = vd->vdev_ms[msi];

	if (msp->ms_start >= vdev_indirect_mapping_max_offset(vim))
	break;

	ASSERT0(range_tree_space(svr->svr_allocd_segs));

	mutex_enter(&msp->ms_lock);

	/*
	* Assert nothing in flight -- ms_*tree is empty.
	*/
	for (int i = 0; i < TXG_SIZE; i++)
	ASSERT0(range_tree_space(msp->ms_allocating[i]));
	for (int i = 0; i < TXG_DEFER_SIZE; i++)
	ASSERT0(range_tree_space(msp->ms_defer[i]));
	ASSERT0(range_tree_space(msp->ms_freed));

	if (msp->ms_sm != NULL) {
	mutex_enter(&svr->svr_lock);
	VERIFY0(space_map_load(msp->ms_sm,
	svr->svr_allocd_segs, SM_ALLOC));

	range_tree_walk(msp->ms_unflushed_allocs,
	range_tree_add, svr->svr_allocd_segs);
	range_tree_walk(msp->ms_unflushed_frees,
	range_tree_remove, svr->svr_allocd_segs);
	range_tree_walk(msp->ms_freeing,
	range_tree_remove, svr->svr_allocd_segs);

	/*
	* Clear everything past what has been synced,
	* because we have not allocated mappings for it yet.
	*/
	uint64_t syncd = vdev_indirect_mapping_max_offset(vim);
	uint64_t sm_end = msp->ms_sm->sm_start +
	msp->ms_sm->sm_size;
	if (sm_end > syncd)
	range_tree_clear(svr->svr_allocd_segs,
	syncd, sm_end - syncd);

	mutex_exit(&svr->svr_lock);
	}
	mutex_exit(&msp->ms_lock);

	mutex_enter(&svr->svr_lock);
	range_tree_vacate(svr->svr_allocd_segs,
	free_mapped_segment_cb, vd);
	mutex_exit(&svr->svr_lock);
	}

	/*
	* Note: this must happen after we invoke free_mapped_segment_cb,
	* because it adds to the obsolete_segments.
	*/
	range_tree_vacate(vd->vdev_obsolete_segments, NULL, NULL);

	ASSERT3U(vic->vic_mapping_object, ==,
	vdev_indirect_mapping_object(vd->vdev_indirect_mapping));
	vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
	vd->vdev_indirect_mapping = NULL;
	vdev_indirect_mapping_free(mos, vic->vic_mapping_object, tx);
	vic->vic_mapping_object = 0;

	ASSERT3U(vic->vic_births_object, ==,
	vdev_indirect_births_object(vd->vdev_indirect_births));
	vdev_indirect_births_close(vd->vdev_indirect_births);
	vd->vdev_indirect_births = NULL;
	vdev_indirect_births_free(mos, vic->vic_births_object, tx);
	vic->vic_births_object = 0;

	/*
	* We may have processed some frees from the removing vdev in this
	* txg, thus increasing svr_bytes_done; discard that here to
	* satisfy the assertions in spa_vdev_removal_destroy().
	* Note that future txg's can not have any bytes_done, because
	* future TXG's are only modified from open context, and we have
	* already shut down the copying thread.
	*/
	svr->svr_bytes_done[dmu_tx_get_txg(tx) & TXG_MASK] = 0;
	spa_finish_removal(spa, DSS_CANCELED, tx);

	vd->vdev_removing = B_FALSE;

	if (!vdev_prop_allocating_off(vd)) {
	spa_config_enter(spa, SCL_ALLOC \| SCL_VDEV, FTAG, RW_WRITER);
	vdev_activate(vd);
	spa_config_exit(spa, SCL_ALLOC \| SCL_VDEV, FTAG);
	}

	vdev_config_dirty(vd);

	zfs_dbgmsg("canceled device removal for vdev %llu in %llu",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)dmu_tx_get_txg(tx));
	spa_history_log_internal(spa, "vdev remove canceled", tx,
	"%s vdev %llu %s", spa_name(spa),
	(u_longlong_t)vd->vdev_id,
	(vd->vdev_path != NULL) ? vd->vdev_path : "-");
	}

	static int
	spa_vdev_remove_cancel_impl(spa_t *spa)
	{
	int error = dsl_sync_task(spa->spa_name, spa_vdev_remove_cancel_check,
	spa_vdev_remove_cancel_sync, NULL, 0,
	ZFS_SPACE_CHECK_EXTRA_RESERVED);
	return (error);
	}

	int
	spa_vdev_remove_cancel(spa_t *spa)
	{
	spa_vdev_remove_suspend(spa);

	if (spa->spa_vdev_removal == NULL)
	return (ENOTACTIVE);

	return (spa_vdev_remove_cancel_impl(spa));
	}

	void
	svr_sync(spa_t spa, dmu_tx_t tx)
	{
	spa_vdev_removal_t *svr = spa->spa_vdev_removal;
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;

	if (svr == NULL)
	return;

	/*
	* This check is necessary so that we do not dirty the
	* DIRECTORY_OBJECT via spa_sync_removing_state() when there
	* is nothing to do. Dirtying it every time would prevent us
	* from syncing-to-convergence.
	*/
	if (svr->svr_bytes_done[txgoff] == 0)
	return;

	/*
	* Update progress accounting.
	*/
	spa->spa_removing_phys.sr_copied += svr->svr_bytes_done[txgoff];
	svr->svr_bytes_done[txgoff] = 0;

	spa_sync_removing_state(spa, tx);
	}

	static void
	vdev_remove_make_hole_and_free(vdev_t *vd)
	{
	uint64_t id = vd->vdev_id;
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	vdev_free(vd);

	vd = vdev_alloc_common(spa, id, 0, &vdev_hole_ops);
	vdev_add_child(rvd, vd);
	vdev_config_dirty(rvd);

	/*
	* Reassess the health of our root vdev.
	*/
	vdev_reopen(rvd);
	}

	/*
	* Remove a log device. The config lock is held for the specified TXG.
	*/
	static int
	spa_vdev_remove_log(vdev_t vd, uint64_t txg)
	{
	metaslab_group_t *mg = vd->vdev_mg;
	spa_t *spa = vd->vdev_spa;
	int error = 0;

	ASSERT(vd->vdev_islog);
	ASSERT(vd == vd->vdev_top);
	ASSERT3P(vd->vdev_log_mg, ==, NULL);
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	/*
	* Stop allocating from this vdev.
	*/
	metaslab_group_passivate(mg);

	/*
	* Wait for the youngest allocations and frees to sync,
	* and then wait for the deferral of those frees to finish.
	*/
	spa_vdev_config_exit(spa, NULL,
	*txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);

	/*
	* Cancel any initialize or TRIM which was in progress.
	*/
	vdev_initialize_stop_all(vd, VDEV_INITIALIZE_CANCELED);
	vdev_trim_stop_all(vd, VDEV_TRIM_CANCELED);
	vdev_autotrim_stop_wait(vd);

	/*
	* Evacuate the device. We don't hold the config lock as
	* writer since we need to do I/O but we do keep the
	* spa_namespace_lock held. Once this completes the device
	* should no longer have any blocks allocated on it.
	*/
	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (vd->vdev_stat.vs_alloc != 0)
	error = spa_reset_logs(spa);

	*txg = spa_vdev_config_enter(spa);

	if (error != 0) {
	metaslab_group_activate(mg);
	ASSERT3P(vd->vdev_log_mg, ==, NULL);
	return (error);
	}
	ASSERT0(vd->vdev_stat.vs_alloc);

	/*
	* The evacuation succeeded. Remove any remaining MOS metadata
	* associated with this vdev, and wait for these changes to sync.
	*/
	vd->vdev_removing = B_TRUE;

	vdev_dirty_leaves(vd, VDD_DTL, *txg);
	vdev_config_dirty(vd);

	/*
	* When the log space map feature is enabled we look at
	* the vdev's top_zap to find the on-disk flush data of
	* the metaslab we just flushed. Thus, while removing a
	* log vdev we make sure to call vdev_metaslab_fini()
	* first, which removes all metaslabs of this vdev from
	* spa_metaslabs_by_flushed before vdev_remove_empty()
	* destroys the top_zap of this log vdev.
	*
	* This avoids the scenario where we flush a metaslab
	* from the log vdev being removed that doesn't have a
	* top_zap and end up failing to lookup its on-disk flush
	* data.
	*
	* We don't call metaslab_group_destroy() right away
	* though (it will be called in vdev_free() later) as
	* during metaslab_sync() of metaslabs from other vdevs
	* we may touch the metaslab group of this vdev through
	* metaslab_class_histogram_verify()
	*/
	vdev_metaslab_fini(vd);

	spa_vdev_config_exit(spa, NULL, *txg, 0, FTAG);
	*txg = spa_vdev_config_enter(spa);

	sysevent_t *ev = spa_event_create(spa, vd, NULL,
	ESC_ZFS_VDEV_REMOVE_DEV);
	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	/* The top ZAP should have been destroyed by vdev_remove_empty. */
	ASSERT0(vd->vdev_top_zap);
	/* The leaf ZAP should have been destroyed by vdev_dtl_sync. */
	ASSERT0(vd->vdev_leaf_zap);

	(void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);

	if (list_link_active(&vd->vdev_state_dirty_node))
	vdev_state_clean(vd);
	if (list_link_active(&vd->vdev_config_dirty_node))
	vdev_config_clean(vd);

	ASSERT0(vd->vdev_stat.vs_alloc);

	/*
	* Clean up the vdev namespace.
	*/
	vdev_remove_make_hole_and_free(vd);

	if (ev != NULL)
	spa_event_post(ev);

	return (0);
	}

	static int
	spa_vdev_remove_top_check(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	if (vd != vd->vdev_top)
	return (SET_ERROR(ENOTSUP));

	if (!vdev_is_concrete(vd))
	return (SET_ERROR(ENOTSUP));

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REMOVAL))
	return (SET_ERROR(ENOTSUP));

	/*
	* This device is already being removed
	*/
	if (vd->vdev_removing)
	return (SET_ERROR(EALREADY));

	metaslab_class_t *mc = vd->vdev_mg->mg_class;
	metaslab_class_t *normal = spa_normal_class(spa);
	if (mc != normal) {
	/*
	* Space allocated from the special (or dedup) class is
	* included in the DMU's space usage, but it's not included
	* in spa_dspace (or dsl_pool_adjustedsize()). Therefore
	* there is always at least as much free space in the normal
	* class, as is allocated from the special (and dedup) class.
	* As a backup check, we will return ENOSPC if this is
	* violated. See also spa_update_dspace().
	*/
	uint64_t available = metaslab_class_get_space(normal) -
	metaslab_class_get_alloc(normal);
	ASSERT3U(available, >=, vd->vdev_stat.vs_alloc);
	if (available < vd->vdev_stat.vs_alloc)
	return (SET_ERROR(ENOSPC));
	} else if (!vd->vdev_noalloc) {
	/* available space in the pool's normal class */
	uint64_t available = dsl_dir_space_available(
	spa->spa_dsl_pool->dp_root_dir, NULL, 0, B_TRUE);
	if (available < vd->vdev_stat.vs_dspace)
	return (SET_ERROR(ENOSPC));
	}

	/*
	* There can not be a removal in progress.
	*/
	if (spa->spa_removing_phys.sr_state == DSS_SCANNING)
	return (SET_ERROR(EBUSY));

	/*
	* The device must have all its data.
	*/
	if (!vdev_dtl_empty(vd, DTL_MISSING) \|\|
	!vdev_dtl_empty(vd, DTL_OUTAGE))
	return (SET_ERROR(EBUSY));

	/*
	* The device must be healthy.
	*/
	if (!vdev_readable(vd))
	return (SET_ERROR(EIO));

	/*
	* All vdevs in normal class must have the same ashift.
	*/
	if (spa->spa_max_ashift != spa->spa_min_ashift) {
	return (SET_ERROR(EINVAL));
	}

	/*
	* A removed special/dedup vdev must have same ashift as normal class.
	*/
	ASSERT(!vd->vdev_islog);
	if (vd->vdev_alloc_bias != VDEV_BIAS_NONE &&
	vd->vdev_ashift != spa->spa_max_ashift) {
	return (SET_ERROR(EINVAL));
	}

	/*
	* All vdevs in normal class must have the same ashift
	* and not be raidz or draid.
	*/
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t id = 0; id < rvd->vdev_children; id++) {
	vdev_t *cvd = rvd->vdev_child[id];

	/*
	* A removed special/dedup vdev must have the same ashift
	* across all vdevs in its class.
	*/
	if (vd->vdev_alloc_bias != VDEV_BIAS_NONE &&
	cvd->vdev_alloc_bias == vd->vdev_alloc_bias &&
	cvd->vdev_ashift != vd->vdev_ashift) {
	return (SET_ERROR(EINVAL));
	}
	if (cvd->vdev_ashift != 0 &&
	cvd->vdev_alloc_bias == VDEV_BIAS_NONE)
	ASSERT3U(cvd->vdev_ashift, ==, spa->spa_max_ashift);
	if (!vdev_is_concrete(cvd))
	continue;
	if (vdev_get_nparity(cvd) != 0)
	return (SET_ERROR(EINVAL));
	/*
	* Need the mirror to be mirror of leaf vdevs only
	*/
	if (cvd->vdev_ops == &vdev_mirror_ops) {
	for (uint64_t cid = 0;
	cid < cvd->vdev_children; cid++) {
	if (!cvd->vdev_child[cid]->vdev_ops->
	vdev_op_leaf)
	return (SET_ERROR(EINVAL));
	}
	}
	}

	return (0);
	}

	/*
	* Initiate removal of a top-level vdev, reducing the total space in the pool.
	* The config lock is held for the specified TXG. Once initiated,
	* evacuation of all allocated space (copying it to other vdevs) happens
	* in the background (see spa_vdev_remove_thread()), and can be canceled
	* (see spa_vdev_remove_cancel()). If successful, the vdev will
	* be transformed to an indirect vdev (see spa_vdev_remove_complete()).
	*/
	static int
	spa_vdev_remove_top(vdev_t vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	boolean_t set_noalloc = B_FALSE;
	int error;

	/*
	* Check for errors up-front, so that we don't waste time
	* passivating the metaslab group and clearing the ZIL if there
	* are errors.
	*/
	error = spa_vdev_remove_top_check(vd);

	/*
	* Stop allocating from this vdev. Note that we must check
	* that this is not the only device in the pool before
	* passivating, otherwise we will not be able to make
	* progress because we can't allocate from any vdevs.
	* The above check for sufficient free space serves this
	* purpose.
	*/
	if (error == 0 && !vd->vdev_noalloc) {
	set_noalloc = B_TRUE;
	error = vdev_passivate(vd, txg);
	}

	if (error != 0)
	return (error);

	/*
	* We stop any initializing and TRIM that is currently in progress
	* but leave the state as "active". This will allow the process to
	* resume if the removal is canceled sometime later.
	*/

	spa_vdev_config_exit(spa, NULL, *txg, 0, FTAG);

	vdev_initialize_stop_all(vd, VDEV_INITIALIZE_ACTIVE);
	vdev_trim_stop_all(vd, VDEV_TRIM_ACTIVE);
	vdev_autotrim_stop_wait(vd);

	*txg = spa_vdev_config_enter(spa);

	/*
	* Things might have changed while the config lock was dropped
	* (e.g. space usage). Check for errors again.
	*/
	error = spa_vdev_remove_top_check(vd);

	if (error != 0) {
	if (set_noalloc)
	vdev_activate(vd);
	spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
	spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
	spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);
	return (error);
	}

	vd->vdev_removing = B_TRUE;

	vdev_dirty_leaves(vd, VDD_DTL, *txg);
	vdev_config_dirty(vd);
	dmu_tx_t tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	dsl_sync_task_nowait(spa->spa_dsl_pool,
	vdev_remove_initiate_sync, (void *)(uintptr_t)vd->vdev_id, tx);
	dmu_tx_commit(tx);

	return (0);
	}

	/*
	* Remove a device from the pool.
	*
	* Removing a device from the vdev namespace requires several steps
	* and can take a significant amount of time. As a result we use
	* the spa_vdev_config_[enter/exit] functions which allow us to
	* grab and release the spa_config_lock while still holding the namespace
	* lock. During each step the configuration is synced out.
	*/
	int
	spa_vdev_remove(spa_t *spa, uint64_t guid, boolean_t unspare)
	{
	vdev_t *vd;
	nvlist_t spares, l2cache, *nv;
	uint64_t txg = 0;
	uint_t nspares, nl2cache;
	int error = 0, error_log;
	boolean_t locked = MUTEX_HELD(&spa_namespace_lock);
	sysevent_t *ev = NULL;
	const char *vd_type = NULL;
	char *vd_path = NULL;

	ASSERT(spa_writeable(spa));

	if (!locked)
	txg = spa_vdev_enter(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;

	if (!locked)
	return (spa_vdev_exit(spa, NULL, txg, error));

	return (error);
	}

	vd = spa_lookup_by_guid(spa, guid, B_FALSE);

	if (spa->spa_spares.sav_vdevs != NULL &&
	nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0 &&
	(nv = spa_nvlist_lookup_by_guid(spares, nspares, guid)) != NULL) {
	/*
	* Only remove the hot spare if it's not currently in use
	* in this pool.
	*/
	if (vd == NULL \|\| unspare) {
	char *type;
	boolean_t draid_spare = B_FALSE;

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type)
	== 0 && strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0)
	draid_spare = B_TRUE;

	if (vd == NULL && draid_spare) {
	error = SET_ERROR(ENOTSUP);
	} else {
	if (vd == NULL)
	vd = spa_lookup_by_guid(spa,
	guid, B_TRUE);
	ev = spa_event_create(spa, vd, NULL,
	ESC_ZFS_VDEV_REMOVE_AUX);

	vd_type = VDEV_TYPE_SPARE;
	vd_path = spa_strdup(fnvlist_lookup_string(
	nv, ZPOOL_CONFIG_PATH));
	spa_vdev_remove_aux(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, spares, nspares, nv);
	spa_load_spares(spa);
	spa->spa_spares.sav_sync = B_TRUE;
	}
	} else {
	error = SET_ERROR(EBUSY);
	}
	} else if (spa->spa_l2cache.sav_vdevs != NULL &&
	nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache) == 0 &&
	(nv = spa_nvlist_lookup_by_guid(l2cache, nl2cache, guid)) != NULL) {
	vd_type = VDEV_TYPE_L2CACHE;
	vd_path = spa_strdup(fnvlist_lookup_string(
	nv, ZPOOL_CONFIG_PATH));
	/*
	* Cache devices can always be removed.
	*/
	vd = spa_lookup_by_guid(spa, guid, B_TRUE);

	/*
	* Stop trimming the cache device. We need to release the
	* config lock to allow the syncing of TRIM transactions
	* without releasing the spa_namespace_lock. The same
	* strategy is employed in spa_vdev_remove_top().
	*/
	spa_vdev_config_exit(spa, NULL,
	txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);
	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL);
	mutex_exit(&vd->vdev_trim_lock);
	txg = spa_vdev_config_enter(spa);

	ev = spa_event_create(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE_AUX);
	spa_vdev_remove_aux(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, l2cache, nl2cache, nv);
	spa_load_l2cache(spa);
	spa->spa_l2cache.sav_sync = B_TRUE;
	} else if (vd != NULL && vd->vdev_islog) {
	ASSERT(!locked);
	vd_type = VDEV_TYPE_LOG;
	vd_path = spa_strdup((vd->vdev_path != NULL) ?
	vd->vdev_path : "-");
	error = spa_vdev_remove_log(vd, &txg);
	} else if (vd != NULL) {
	ASSERT(!locked);
	error = spa_vdev_remove_top(vd, &txg);
	} else {
	/*
	* There is no vdev of any kind with the specified guid.
	*/
	error = SET_ERROR(ENOENT);
	}

	error_log = error;

	if (!locked)
	error = spa_vdev_exit(spa, NULL, txg, error);

	/*
	* Logging must be done outside the spa config lock. Otherwise,
	* this code path could end up holding the spa config lock while
	* waiting for a txg_sync so it can write to the internal log.
	* Doing that would prevent the txg sync from actually happening,
	* causing a deadlock.
	*/
	if (error_log == 0 && vd_type != NULL && vd_path != NULL) {
	spa_history_log_internal(spa, "vdev remove", NULL,
	"%s vdev (%s) %s", spa_name(spa), vd_type, vd_path);
	}
	if (vd_path != NULL)
	spa_strfree(vd_path);

	if (ev != NULL)
	spa_event_post(ev);

	return (error);
	}

	int
	spa_removal_get_stats(spa_t spa, pool_removal_stat_t prs)
	{
	prs->prs_state = spa->spa_removing_phys.sr_state;

	if (prs->prs_state == DSS_NONE)
	return (SET_ERROR(ENOENT));

	prs->prs_removing_vdev = spa->spa_removing_phys.sr_removing_vdev;
	prs->prs_start_time = spa->spa_removing_phys.sr_start_time;
	prs->prs_end_time = spa->spa_removing_phys.sr_end_time;
	prs->prs_to_copy = spa->spa_removing_phys.sr_to_copy;
	prs->prs_copied = spa->spa_removing_phys.sr_copied;

	prs->prs_mapping_memory = 0;
	uint64_t indirect_vdev_id =
	spa->spa_removing_phys.sr_prev_indirect_vdev;
	while (indirect_vdev_id != -1) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[indirect_vdev_id];
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
	prs->prs_mapping_memory += vdev_indirect_mapping_size(vim);
	indirect_vdev_id = vic->vic_prev_indirect_vdev;
	}

	return (0);
	}

	ZFS_MODULE_PARAM(zfs_vdev, zfs_, removal_ignore_errors, INT, ZMOD_RW,
	"Ignore hard IO errors when removing device");

	-ZFS_MODULE_PARAM(zfs_vdev, zfs_, remove_max_segment, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_, remove_max_segment, UINT, ZMOD_RW,
	"Largest contiguous segment to allocate when removing device");

	-ZFS_MODULE_PARAM(zfs_vdev, vdev_, removal_max_span, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, vdev_, removal_max_span, UINT, ZMOD_RW,
	"Largest span of free chunks a remap segment can span");

	/* BEGIN CSTYLED */
	-ZFS_MODULE_PARAM(zfs_vdev, zfs_, removal_suspend_progress, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_vdev, zfs_, removal_suspend_progress, UINT, ZMOD_RW,
	"Pause device removal after this many bytes are copied "
	"(debug use only - causes removal to hang)");
	/* END CSTYLED */

	EXPORT_SYMBOL(free_from_removing_vdev);
	EXPORT_SYMBOL(spa_removal_get_stats);
	EXPORT_SYMBOL(spa_remove_init);
	EXPORT_SYMBOL(spa_restart_removal);
	EXPORT_SYMBOL(spa_vdev_removal_destroy);
	EXPORT_SYMBOL(spa_vdev_remove);
	EXPORT_SYMBOL(spa_vdev_remove_cancel);
	EXPORT_SYMBOL(spa_vdev_remove_suspend);
	EXPORT_SYMBOL(svr_sync);
	diff --git a/sys/contrib/openzfs/module/zfs/zcp_get.c b/sys/contrib/openzfs/module/zfs/zcp_get.c
	index 0a0466d46969..cd17374eb422 100644
	--- a/sys/contrib/openzfs/module/zfs/zcp_get.c
	+++ b/sys/contrib/openzfs/module/zfs/zcp_get.c
	@@ -1,809 +1,809 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2016 by Delphix. All rights reserved.
	*/

	#include <sys/lua/lua.h>
	#include <sys/lua/lualib.h>
	#include <sys/lua/lauxlib.h>

	#include <zfs_prop.h>

	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dmu_objset.h>
	#include <sys/mntent.h>
	#include <sys/sunddi.h>
	#include <sys/zap.h>
	#include <sys/zcp.h>
	#include <sys/zcp_iter.h>
	#include <sys/zcp_global.h>
	#include <sys/zcp_prop.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_znode.h>
	#include <sys/zvol.h>

	#ifdef _KERNEL
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#endif

	static int
	get_objset_type(dsl_dataset_t ds, zfs_type_t type)
	{
	int error;
	objset_t *os;
	error = dmu_objset_from_ds(ds, &os);
	if (error != 0)
	return (error);
	if (ds->ds_is_snapshot) {
	*type = ZFS_TYPE_SNAPSHOT;
	} else {
	switch (os->os_phys->os_type) {
	case DMU_OST_ZFS:
	*type = ZFS_TYPE_FILESYSTEM;
	break;
	case DMU_OST_ZVOL:
	*type = ZFS_TYPE_VOLUME;
	break;
	default:
	return (EINVAL);
	}
	}
	return (0);
	}

	/*
	* Returns the string name of ds's type in str (a buffer which should be
	* at least 12 bytes long).
	*/
	static int
	get_objset_type_name(dsl_dataset_t ds, char str)
	{
	zfs_type_t type = ZFS_TYPE_INVALID;
	int error = get_objset_type(ds, &type);
	if (error != 0)
	return (error);
	switch (type) {
	case ZFS_TYPE_SNAPSHOT:
	(void) strlcpy(str, "snapshot", ZAP_MAXVALUELEN);
	break;
	case ZFS_TYPE_FILESYSTEM:
	(void) strlcpy(str, "filesystem", ZAP_MAXVALUELEN);
	break;
	case ZFS_TYPE_VOLUME:
	(void) strlcpy(str, "volume", ZAP_MAXVALUELEN);
	break;
	default:
	return (EINVAL);
	}
	return (0);
	}

	/*
	* Determines the source of a property given its setpoint and
	* property type. It pushes the source to the lua stack.
	*/
	static void
	get_prop_src(lua_State state, const char setpoint, zfs_prop_t prop)
	{
	if (zfs_prop_readonly(prop) \|\| (prop == ZFS_PROP_VERSION)) {
	lua_pushnil(state);
	} else {
	const char *src;
	if (strcmp("", setpoint) == 0) {
	src = "default";
	} else {
	src = setpoint;
	}
	(void) lua_pushstring(state, src);
	}
	}

	/*
	* Given an error encountered while getting properties, either longjmp's for
	* a fatal error or pushes nothing to the stack for a non fatal one.
	*/
	static int
	zcp_handle_error(lua_State state, const char dataset_name,
	const char *property_name, int error)
	{
	ASSERT3S(error, !=, 0);
	if (error == ENOENT) {
	return (0);
	} else if (error == EINVAL) {
	return (luaL_error(state,
	"property '%s' is not a valid property on dataset '%s'",
	property_name, dataset_name));
	} else if (error == EIO) {
	return (luaL_error(state,
	"I/O error while retrieving property '%s' on dataset '%s'",
	property_name, dataset_name));
	} else {
	return (luaL_error(state, "unexpected error %d while "
	"retrieving property '%s' on dataset '%s'",
	error, property_name, dataset_name));
	}
	}

	/*
	* Look up a user defined property in the zap object. If it exists, push it
	* and the setpoint onto the stack, otherwise don't push anything.
	*/
	static int
	zcp_get_user_prop(lua_State state, dsl_pool_t dp, const char *dataset_name,
	const char *property_name)
	{
	int error;
	char *buf;
	char setpoint[ZFS_MAX_DATASET_NAME_LEN];
	/*
	* zcp_dataset_hold will either successfully return the requested
	* dataset or throw a lua error and longjmp out of the zfs.get_prop call
	* without returning.
	*/
	dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
	if (ds == NULL)
	return (1); /* not reached; zcp_dataset_hold() longjmp'd */

	buf = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
	error = dsl_prop_get_ds(ds, property_name, 1, ZAP_MAXVALUELEN,
	buf, setpoint);
	dsl_dataset_rele(ds, FTAG);

	if (error != 0) {
	kmem_free(buf, ZAP_MAXVALUELEN);
	return (zcp_handle_error(state, dataset_name, property_name,
	error));
	}
	(void) lua_pushstring(state, buf);
	(void) lua_pushstring(state, setpoint);
	kmem_free(buf, ZAP_MAXVALUELEN);
	return (2);
	}

	/*
	* Check if the property we're looking for is stored in the ds_dir. If so,
	* return it in the 'val' argument. Return 0 on success and ENOENT and if
	* the property is not present.
	*/
	static int
	get_dsl_dir_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop,
	uint64_t *val)
	{
	dsl_dir_t *dd = ds->ds_dir;
	mutex_enter(&dd->dd_lock);
	switch (zfs_prop) {
	case ZFS_PROP_USEDSNAP:
	*val = dsl_dir_get_usedsnap(dd);
	break;
	case ZFS_PROP_USEDCHILD:
	*val = dsl_dir_get_usedchild(dd);
	break;
	case ZFS_PROP_USEDDS:
	*val = dsl_dir_get_usedds(dd);
	break;
	case ZFS_PROP_USEDREFRESERV:
	*val = dsl_dir_get_usedrefreserv(dd);
	break;
	case ZFS_PROP_LOGICALUSED:
	*val = dsl_dir_get_logicalused(dd);
	break;
	default:
	mutex_exit(&dd->dd_lock);
	return (SET_ERROR(ENOENT));
	}
	mutex_exit(&dd->dd_lock);
	return (0);
	}

	/*
	* Check if the property we're looking for is stored at the dsl_dataset or
	* dsl_dir level. If so, push the property value and source onto the lua stack
	* and return 0. If it is not present or a failure occurs in lookup, return a
	* non-zero error value.
	*/
	static int
	get_special_prop(lua_State state, dsl_dataset_t ds, const char *dsname,
	zfs_prop_t zfs_prop)
	{
	int error = 0;
	objset_t *os;
	uint64_t numval = 0;
	char *strval = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
	char setpoint[ZFS_MAX_DATASET_NAME_LEN] =
	"Internal error - setpoint not determined";
	zfs_type_t ds_type = ZFS_TYPE_INVALID;
	zprop_type_t prop_type = zfs_prop_get_type(zfs_prop);
	(void) get_objset_type(ds, &ds_type);

	switch (zfs_prop) {
	case ZFS_PROP_REFRATIO:
	numval = dsl_get_refratio(ds);
	break;
	case ZFS_PROP_USED:
	numval = dsl_get_used(ds);
	break;
	case ZFS_PROP_CLONES: {
	nvlist_t *clones = fnvlist_alloc();
	error = get_clones_stat_impl(ds, clones);
	if (error == 0) {
	/* push list to lua stack */
	VERIFY0(zcp_nvlist_to_lua(state, clones, NULL, 0ULL));
	/* source */
	(void) lua_pushnil(state);
	}
	nvlist_free(clones);
	kmem_free(strval, ZAP_MAXVALUELEN);
	return (error);
	}
	case ZFS_PROP_COMPRESSRATIO:
	numval = dsl_get_compressratio(ds);
	break;
	case ZFS_PROP_CREATION:
	numval = dsl_get_creation(ds);
	break;
	case ZFS_PROP_REFERENCED:
	numval = dsl_get_referenced(ds);
	break;
	case ZFS_PROP_AVAILABLE:
	numval = dsl_get_available(ds);
	break;
	case ZFS_PROP_LOGICALREFERENCED:
	numval = dsl_get_logicalreferenced(ds);
	break;
	case ZFS_PROP_CREATETXG:
	numval = dsl_get_creationtxg(ds);
	break;
	case ZFS_PROP_GUID:
	numval = dsl_get_guid(ds);
	break;
	case ZFS_PROP_UNIQUE:
	numval = dsl_get_unique(ds);
	break;
	case ZFS_PROP_OBJSETID:
	numval = dsl_get_objsetid(ds);
	break;
	case ZFS_PROP_ORIGIN:
	dsl_dir_get_origin(ds->ds_dir, strval);
	break;
	case ZFS_PROP_USERACCOUNTING:
	error = dmu_objset_from_ds(ds, &os);
	if (error == 0)
	numval = dmu_objset_userspace_present(os);
	break;
	case ZFS_PROP_WRITTEN:
	error = dsl_get_written(ds, &numval);
	break;
	case ZFS_PROP_TYPE:
	error = get_objset_type_name(ds, strval);
	break;
	case ZFS_PROP_PREV_SNAP:
	error = dsl_get_prev_snap(ds, strval);
	break;
	case ZFS_PROP_NAME:
	dsl_dataset_name(ds, strval);
	break;
	case ZFS_PROP_MOUNTPOINT:
	error = dsl_get_mountpoint(ds, dsname, strval, setpoint);
	break;
	case ZFS_PROP_VERSION:
	/* should be a snapshot or filesystem */
	ASSERT(ds_type != ZFS_TYPE_VOLUME);
	error = dmu_objset_from_ds(ds, &os);
	/* look in the master node for the version */
	if (error == 0) {
	error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
	sizeof (numval), 1, &numval);
	}
	break;
	case ZFS_PROP_DEFER_DESTROY:
	numval = dsl_get_defer_destroy(ds);
	break;
	case ZFS_PROP_USERREFS:
	numval = dsl_get_userrefs(ds);
	break;
	case ZFS_PROP_FILESYSTEM_COUNT:
	error = dsl_dir_get_filesystem_count(ds->ds_dir, &numval);
	(void) strlcpy(setpoint, "", ZFS_MAX_DATASET_NAME_LEN);
	break;
	case ZFS_PROP_SNAPSHOT_COUNT:
	error = dsl_dir_get_snapshot_count(ds->ds_dir, &numval);
	(void) strlcpy(setpoint, "", ZFS_MAX_DATASET_NAME_LEN);
	break;
	case ZFS_PROP_NUMCLONES:
	numval = dsl_get_numclones(ds);
	break;
	case ZFS_PROP_INCONSISTENT:
	numval = dsl_get_inconsistent(ds);
	break;
	case ZFS_PROP_IVSET_GUID:
	if (dsl_dataset_is_zapified(ds)) {
	error = zap_lookup(ds->ds_dir->dd_pool->dp_meta_objset,
	ds->ds_object, DS_FIELD_IVSET_GUID,
	sizeof (numval), 1, &numval);
	} else {
	error = ENOENT;
	}
	break;
	case ZFS_PROP_RECEIVE_RESUME_TOKEN: {
	char *token = get_receive_resume_token(ds);
	if (token != NULL) {
	(void) strlcpy(strval, token, ZAP_MAXVALUELEN);
	kmem_strfree(token);
	} else {
	error = ENOENT;
	}
	break;
	}
	case ZFS_PROP_VOLSIZE:
	ASSERT(ds_type == ZFS_TYPE_VOLUME \|\|
	ds_type == ZFS_TYPE_SNAPSHOT);
	error = dmu_objset_from_ds(ds, &os);
	if (error == 0) {
	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size",
	sizeof (numval), 1, &numval);
	}
	if (error == 0)
	(void) strlcpy(setpoint, dsname,
	ZFS_MAX_DATASET_NAME_LEN);

	break;
	case ZFS_PROP_VOLBLOCKSIZE: {
	ASSERT(ds_type == ZFS_TYPE_VOLUME);
	dmu_object_info_t doi;
	error = dmu_objset_from_ds(ds, &os);
	if (error == 0) {
	error = dmu_object_info(os, ZVOL_OBJ, &doi);
	if (error == 0)
	numval = doi.doi_data_block_size;
	}
	break;
	}

	case ZFS_PROP_KEYSTATUS:
	case ZFS_PROP_KEYFORMAT: {
	/* provide defaults in case no crypto obj exists */
	setpoint[0] = '\0';
	if (zfs_prop == ZFS_PROP_KEYSTATUS)
	numval = ZFS_KEYSTATUS_NONE;
	else
	numval = ZFS_KEYFORMAT_NONE;

	nvlist_t nvl, propval;
	nvl = fnvlist_alloc();
	dsl_dataset_crypt_stats(ds, nvl);
	if (nvlist_lookup_nvlist(nvl, zfs_prop_to_name(zfs_prop),
	&propval) == 0) {
	char *source;

	(void) nvlist_lookup_uint64(propval, ZPROP_VALUE,
	&numval);
	if (nvlist_lookup_string(propval, ZPROP_SOURCE,
	&source) == 0)
	strlcpy(setpoint, source, sizeof (setpoint));
	}
	nvlist_free(nvl);
	break;
	}

	case ZFS_PROP_SNAPSHOTS_CHANGED:
	numval = dsl_dir_snap_cmtime(ds->ds_dir).tv_sec;
	break;

	default:
	/* Did not match these props, check in the dsl_dir */
	error = get_dsl_dir_prop(ds, zfs_prop, &numval);
	}
	if (error != 0) {
	kmem_free(strval, ZAP_MAXVALUELEN);
	return (error);
	}

	switch (prop_type) {
	case PROP_TYPE_NUMBER: {
	(void) lua_pushnumber(state, numval);
	break;
	}
	case PROP_TYPE_STRING: {
	(void) lua_pushstring(state, strval);
	break;
	}
	case PROP_TYPE_INDEX: {
	const char *propval;
	error = zfs_prop_index_to_string(zfs_prop, numval, &propval);
	if (error != 0) {
	kmem_free(strval, ZAP_MAXVALUELEN);
	return (error);
	}
	(void) lua_pushstring(state, propval);
	break;
	}
	}
	kmem_free(strval, ZAP_MAXVALUELEN);

	/* Push the source to the stack */
	get_prop_src(state, setpoint, zfs_prop);
	return (0);
	}

	/*
	* Look up a property and its source in the zap object. If the value is
	* present and successfully retrieved, push the value and source on the
	* lua stack and return 0. On failure, return a non-zero error value.
	*/
	static int
	get_zap_prop(lua_State state, dsl_dataset_t ds, zfs_prop_t zfs_prop)
	{
	int error = 0;
	char setpoint[ZFS_MAX_DATASET_NAME_LEN];
	char *strval = kmem_alloc(ZAP_MAXVALUELEN, KM_SLEEP);
	uint64_t numval;
	const char *prop_name = zfs_prop_to_name(zfs_prop);
	zprop_type_t prop_type = zfs_prop_get_type(zfs_prop);

	if (prop_type == PROP_TYPE_STRING) {
	/* Push value to lua stack */
	error = dsl_prop_get_ds(ds, prop_name, 1,
	ZAP_MAXVALUELEN, strval, setpoint);
	if (error == 0)
	(void) lua_pushstring(state, strval);
	} else {
	error = dsl_prop_get_ds(ds, prop_name, sizeof (numval),
	1, &numval, setpoint);

	#ifdef _KERNEL
	/* Fill in temporary value for prop, if applicable */
	(void) zfs_get_temporary_prop(ds, zfs_prop, &numval, setpoint);
	#else
	+ kmem_free(strval, ZAP_MAXVALUELEN);
	return (luaL_error(state,
	"temporary properties only supported in kernel mode",
	prop_name));
	#endif
	/* Push value to lua stack */
	if (prop_type == PROP_TYPE_INDEX) {
	const char *propval;
	error = zfs_prop_index_to_string(zfs_prop, numval,
	&propval);
	if (error == 0)
	(void) lua_pushstring(state, propval);
	} else {
	if (error == 0)
	(void) lua_pushnumber(state, numval);
	}
	}
	kmem_free(strval, ZAP_MAXVALUELEN);
	if (error == 0)
	get_prop_src(state, setpoint, zfs_prop);
	return (error);
	}

	/*
	* Determine whether property is valid for a given dataset
	*/
	boolean_t
	prop_valid_for_ds(dsl_dataset_t *ds, zfs_prop_t zfs_prop)
	{
	zfs_type_t zfs_type = ZFS_TYPE_INVALID;

	/* properties not supported */
	if ((zfs_prop == ZFS_PROP_ISCSIOPTIONS) \|\|
	(zfs_prop == ZFS_PROP_MOUNTED))
	return (B_FALSE);

	/* if we want the origin prop, ds must be a clone */
	if ((zfs_prop == ZFS_PROP_ORIGIN) && (!dsl_dir_is_clone(ds->ds_dir)))
	return (B_FALSE);

	int error = get_objset_type(ds, &zfs_type);
	if (error != 0)
	return (B_FALSE);
	return (zfs_prop_valid_for_type(zfs_prop, zfs_type, B_FALSE));
	}

	/*
	* Look up a given dataset property. On success return 2, the number of
	* values pushed to the lua stack (property value and source). On a fatal
	* error, longjmp. On a non fatal error push nothing.
	*/
	static int
	zcp_get_system_prop(lua_State state, dsl_pool_t dp, const char *dataset_name,
	zfs_prop_t zfs_prop)
	{
	int error;
	/*
	* zcp_dataset_hold will either successfully return the requested
	* dataset or throw a lua error and longjmp out of the zfs.get_prop call
	* without returning.
	*/
	dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
	if (ds == NULL)
	return (1); /* not reached; zcp_dataset_hold() longjmp'd */

	/* Check that the property is valid for the given dataset */
	const char *prop_name = zfs_prop_to_name(zfs_prop);
	if (!prop_valid_for_ds(ds, zfs_prop)) {
	dsl_dataset_rele(ds, FTAG);
	return (0);
	}

	/* Check if the property can be accessed directly */
	error = get_special_prop(state, ds, dataset_name, zfs_prop);
	if (error == 0) {
	dsl_dataset_rele(ds, FTAG);
	/* The value and source have been pushed by get_special_prop */
	return (2);
	}
	if (error != ENOENT) {
	dsl_dataset_rele(ds, FTAG);
	return (zcp_handle_error(state, dataset_name,
	prop_name, error));
	}

	/* If we were unable to find it, look in the zap object */
	error = get_zap_prop(state, ds, zfs_prop);
	dsl_dataset_rele(ds, FTAG);
	if (error != 0) {
	return (zcp_handle_error(state, dataset_name,
	prop_name, error));
	}
	/* The value and source have been pushed by get_zap_prop */
	return (2);
	}

	#ifdef _KERNEL
	static zfs_userquota_prop_t
	get_userquota_prop(const char *prop_name)
	{
	zfs_userquota_prop_t type;
	/* Figure out the property type ({user\|group}{quota\|used}) */
	for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) {
	if (strncmp(prop_name, zfs_userquota_prop_prefixes[type],
	strlen(zfs_userquota_prop_prefixes[type])) == 0)
	break;
	}
	return (type);
	}

	/*
	* Given the name of a zfs_userquota_prop, this function determines the
	* prop type as well as the numeric group/user ids based on the string
	* following the '@' in the property name. On success, returns 0. On failure,
	* returns a non-zero error.
	* 'domain' must be free'd by caller using kmem_strfree()
	*/
	static int
	parse_userquota_prop(const char prop_name, zfs_userquota_prop_t type,
	char *domain, uint64_t rid)
	{
	char cp, end, *domain_val;

	*type = get_userquota_prop(prop_name);
	if (*type >= ZFS_NUM_USERQUOTA_PROPS)
	return (EINVAL);

	*rid = 0;
	cp = strchr(prop_name, '@') + 1;
	if (strncmp(cp, "S-1-", 4) == 0) {
	/*
	* It's a numeric SID (eg "S-1-234-567-89") and we want to
	* separate the domain id and the rid
	*/
	int domain_len = strrchr(cp, '-') - cp;
	domain_val = kmem_alloc(domain_len + 1, KM_SLEEP);
	- (void) strncpy(domain_val, cp, domain_len);
	- domain_val[domain_len] = '\0';
	+ (void) strlcpy(domain_val, cp, domain_len + 1);
	cp += domain_len + 1;

	(void) ddi_strtoll(cp, &end, 10, (longlong_t *)rid);
	if (*end != '\0') {
	kmem_strfree(domain_val);
	return (EINVAL);
	}
	} else {
	/* It's only a user/group ID (eg "12345"), just get the rid */
	domain_val = NULL;
	(void) ddi_strtoll(cp, &end, 10, (longlong_t *)rid);
	if (*end != '\0')
	return (EINVAL);
	}
	*domain = domain_val;
	return (0);
	}

	/*
	* Look up {user\|group}{quota\|used} property for given dataset. On success
	* push the value (quota or used amount) and the setpoint. On failure, push
	* a lua error.
	*/
	static int
	zcp_get_userquota_prop(lua_State state, dsl_pool_t dp,
	const char dataset_name, const char prop_name)
	{
	zfsvfs_t *zfvp;
	zfsvfs_t *zfsvfs;
	int error;
	zfs_userquota_prop_t type;
	char *domain;
	uint64_t rid, value = 0;
	objset_t *os;

	dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
	if (ds == NULL)
	return (1); /* not reached; zcp_dataset_hold() longjmp'd */

	error = parse_userquota_prop(prop_name, &type, &domain, &rid);
	if (error == 0) {
	error = dmu_objset_from_ds(ds, &os);
	if (error == 0) {
	zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
	error = zfsvfs_create_impl(&zfvp, zfsvfs, os);
	if (error == 0) {
	error = zfs_userspace_one(zfvp, type, domain,
	rid, &value);
	zfsvfs_free(zfvp);
	}
	}
	if (domain != NULL)
	kmem_strfree(domain);
	}
	dsl_dataset_rele(ds, FTAG);

	if ((value == 0) && ((type == ZFS_PROP_USERQUOTA) \|\|
	(type == ZFS_PROP_GROUPQUOTA)))
	error = SET_ERROR(ENOENT);
	if (error != 0) {
	return (zcp_handle_error(state, dataset_name,
	prop_name, error));
	}

	(void) lua_pushnumber(state, value);
	(void) lua_pushstring(state, dataset_name);
	return (2);
	}
	#endif

	/*
	* Determines the name of the snapshot referenced in the written property
	* name. Returns snapshot name in snap_name, a buffer that must be at least
	* as large as ZFS_MAX_DATASET_NAME_LEN
	*/
	static void
	parse_written_prop(const char dataset_name, const char prop_name,
	char *snap_name)
	{
	ASSERT(zfs_prop_written(prop_name));
	const char *name = prop_name + ZFS_WRITTEN_PROP_PREFIX_LEN;
	if (strchr(name, '@') == NULL) {
	(void) snprintf(snap_name, ZFS_MAX_DATASET_NAME_LEN, "%s@%s",
	dataset_name, name);
	} else {
	(void) strlcpy(snap_name, name, ZFS_MAX_DATASET_NAME_LEN);
	}
	}

	/*
	* Look up written@ property for given dataset. On success
	* push the value and the setpoint. If error is fatal, we will
	* longjmp, otherwise push nothing.
	*/
	static int
	zcp_get_written_prop(lua_State state, dsl_pool_t dp,
	const char dataset_name, const char prop_name)
	{
	char snap_name[ZFS_MAX_DATASET_NAME_LEN];
	uint64_t used, comp, uncomp;
	dsl_dataset_t *old;
	int error = 0;

	parse_written_prop(dataset_name, prop_name, snap_name);
	dsl_dataset_t *new = zcp_dataset_hold(state, dp, dataset_name, FTAG);
	if (new == NULL)
	return (1); /* not reached; zcp_dataset_hold() longjmp'd */

	error = dsl_dataset_hold(dp, snap_name, FTAG, &old);
	if (error != 0) {
	dsl_dataset_rele(new, FTAG);
	return (zcp_dataset_hold_error(state, dp, snap_name,
	error));
	}
	error = dsl_dataset_space_written(old, new,
	&used, &comp, &uncomp);

	dsl_dataset_rele(old, FTAG);
	dsl_dataset_rele(new, FTAG);

	if (error != 0) {
	return (zcp_handle_error(state, dataset_name,
	snap_name, error));
	}
	(void) lua_pushnumber(state, used);
	(void) lua_pushstring(state, dataset_name);
	return (2);
	}

	static int zcp_get_prop(lua_State *state);
	static const zcp_lib_info_t zcp_get_prop_info = {
	.name = "get_prop",
	.func = zcp_get_prop,
	.pargs = {
	{ .za_name = "dataset", .za_lua_type = LUA_TSTRING },
	{ .za_name = "property", .za_lua_type = LUA_TSTRING },
	{NULL, 0}
	},
	.kwargs = {
	{NULL, 0}
	}
	};

	static int
	zcp_get_prop(lua_State *state)
	{
	const char *dataset_name;
	const char *property_name;
	dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
	const zcp_lib_info_t *libinfo = &zcp_get_prop_info;

	zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);

	dataset_name = lua_tostring(state, 1);
	property_name = lua_tostring(state, 2);

	/* User defined property */
	if (zfs_prop_user(property_name)) {
	return (zcp_get_user_prop(state, dp,
	dataset_name, property_name));
	}
	/* userspace property */
	if (zfs_prop_userquota(property_name)) {
	#ifdef _KERNEL
	return (zcp_get_userquota_prop(state, dp,
	dataset_name, property_name));
	#else
	return (luaL_error(state,
	"user quota properties only supported in kernel mode",
	property_name));
	#endif
	}
	/* written@ property */
	if (zfs_prop_written(property_name)) {
	return (zcp_get_written_prop(state, dp,
	dataset_name, property_name));
	}

	zfs_prop_t zfs_prop = zfs_name_to_prop(property_name);
	/* Valid system property */
	if (zfs_prop != ZPROP_INVAL) {
	return (zcp_get_system_prop(state, dp, dataset_name,
	zfs_prop));
	}

	/* Invalid property name */
	return (luaL_error(state,
	"'%s' is not a valid property", property_name));
	}

	int
	zcp_load_get_lib(lua_State *state)
	{
	lua_pushcclosure(state, zcp_get_prop_info.func, 0);
	lua_setfield(state, -2, zcp_get_prop_info.name);

	return (1);
	}
	diff --git a/sys/contrib/openzfs/module/zfs/zfs_fm.c b/sys/contrib/openzfs/module/zfs/zfs_fm.c
	index 47bb79a5d21a..06aa1214ace8 100644
	--- a/sys/contrib/openzfs/module/zfs/zfs_fm.c
	+++ b/sys/contrib/openzfs/module/zfs/zfs_fm.c
	@@ -1,1521 +1,1521 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Copyright (c) 2012,2021 by Delphix. All rights reserved.
	*/

	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>

	#include <sys/fm/fs/zfs.h>
	#include <sys/fm/protocol.h>
	#include <sys/fm/util.h>
	#include <sys/sysevent.h>

	/*
	* This general routine is responsible for generating all the different ZFS
	* ereports. The payload is dependent on the class, and which arguments are
	* supplied to the function:
	*
	* EREPORT POOL VDEV IO
	* block X X X
	* data X X
	* device X X
	* pool X
	*
	* If we are in a loading state, all errors are chained together by the same
	* SPA-wide ENA (Error Numeric Association).
	*
	* For isolated I/O requests, we get the ENA from the zio_t. The propagation
	* gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want
	* to chain together all ereports associated with a logical piece of data. For
	* read I/Os, there are basically three 'types' of I/O, which form a roughly
	* layered diagram:
	*
	* +---------------+
	* \| Aggregate I/O \| No associated logical data or device
	* +---------------+
	* \|
	* V
	* +---------------+ Reads associated with a piece of logical data.
	* \| Read I/O \| This includes reads on behalf of RAID-Z,
	* +---------------+ mirrors, gang blocks, retries, etc.
	* \|
	* V
	* +---------------+ Reads associated with a particular device, but
	* \| Physical I/O \| no logical data. Issued as part of vdev caching
	* +---------------+ and I/O aggregation.
	*
	* Note that 'physical I/O' here is not the same terminology as used in the rest
	* of ZIO. Typically, 'physical I/O' simply means that there is no attached
	* blockpointer. But I/O with no associated block pointer can still be related
	* to a logical piece of data (i.e. RAID-Z requests).
	*
	* Purely physical I/O always have unique ENAs. They are not related to a
	* particular piece of logical data, and therefore cannot be chained together.
	* We still generate an ereport, but the DE doesn't correlate it with any
	* logical piece of data. When such an I/O fails, the delegated I/O requests
	* will issue a retry, which will trigger the 'real' ereport with the correct
	* ENA.
	*
	* We keep track of the ENA for a ZIO chain through the 'io_logical' member.
	* When a new logical I/O is issued, we set this to point to itself. Child I/Os
	* then inherit this pointer, so that when it is first set subsequent failures
	* will use the same ENA. For vdev cache fill and queue aggregation I/O,
	* this pointer is set to NULL, and no ereport will be generated (since it
	* doesn't actually correspond to any particular device or piece of data,
	* and the caller will always retry without caching or queueing anyway).
	*
	* For checksum errors, we want to include more information about the actual
	* error which occurs. Accordingly, we build an ereport when the error is
	* noticed, but instead of sending it in immediately, we hang it off of the
	* io_cksum_report field of the logical IO. When the logical IO completes
	* (successfully or not), zfs_ereport_finish_checksum() is called with the
	* good and bad versions of the buffer (if available), and we annotate the
	* ereport with information about the differences.
	*/

	#ifdef _KERNEL
	/*
	* Duplicate ereport Detection
	*
	* Some ereports are retained momentarily for detecting duplicates. These
	* are kept in a recent_events_node_t in both a time-ordered list and an AVL
	* tree of recent unique ereports.
	*
	* The lifespan of these recent ereports is bounded (15 mins) and a cleaner
	* task is used to purge stale entries.
	*/
	static list_t recent_events_list;
	static avl_tree_t recent_events_tree;
	static kmutex_t recent_events_lock;
	static taskqid_t recent_events_cleaner_tqid;

	/*
	* Each node is about 128 bytes so 2,000 would consume 1/4 MiB.
	*
	* This setting can be changed dynamically and setting it to zero
	* disables duplicate detection.
	*/
	static unsigned int zfs_zevent_retain_max = 2000;

	/*
	* The lifespan for a recent ereport entry. The default of 15 minutes is
	* intended to outlive the zfs diagnosis engine's threshold of 10 errors
	* over a period of 10 minutes.
	*/
	static unsigned int zfs_zevent_retain_expire_secs = 900;

	typedef enum zfs_subclass {
	ZSC_IO,
	ZSC_DATA,
	ZSC_CHECKSUM
	} zfs_subclass_t;

	typedef struct {
	/* common criteria */
	uint64_t re_pool_guid;
	uint64_t re_vdev_guid;
	int re_io_error;
	uint64_t re_io_size;
	uint64_t re_io_offset;
	zfs_subclass_t re_subclass;
	zio_priority_t re_io_priority;

	/* logical zio criteria (optional) */
	zbookmark_phys_t re_io_bookmark;

	/* internal state */
	avl_node_t re_tree_link;
	list_node_t re_list_link;
	uint64_t re_timestamp;
	} recent_events_node_t;

	static int
	recent_events_compare(const void a, const void b)
	{
	const recent_events_node_t *node1 = a;
	const recent_events_node_t *node2 = b;
	int cmp;

	/*
	* The comparison order here is somewhat arbitrary.
	* What's important is that if every criteria matches, then it
	* is a duplicate (i.e. compare returns 0)
	*/
	if ((cmp = TREE_CMP(node1->re_subclass, node2->re_subclass)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_pool_guid, node2->re_pool_guid)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_vdev_guid, node2->re_vdev_guid)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_io_error, node2->re_io_error)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_io_priority, node2->re_io_priority)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_io_size, node2->re_io_size)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(node1->re_io_offset, node2->re_io_offset)) != 0)
	return (cmp);

	const zbookmark_phys_t *zb1 = &node1->re_io_bookmark;
	const zbookmark_phys_t *zb2 = &node2->re_io_bookmark;

	if ((cmp = TREE_CMP(zb1->zb_objset, zb2->zb_objset)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(zb1->zb_object, zb2->zb_object)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(zb1->zb_level, zb2->zb_level)) != 0)
	return (cmp);
	if ((cmp = TREE_CMP(zb1->zb_blkid, zb2->zb_blkid)) != 0)
	return (cmp);

	return (0);
	}

	static void zfs_ereport_schedule_cleaner(void);

	/*
	* background task to clean stale recent event nodes.
	*/
	static void
	zfs_ereport_cleaner(void *arg)
	{
	recent_events_node_t *entry;
	uint64_t now = gethrtime();

	/*
	* purge expired entries
	*/
	mutex_enter(&recent_events_lock);
	while ((entry = list_tail(&recent_events_list)) != NULL) {
	uint64_t age = NSEC2SEC(now - entry->re_timestamp);
	if (age <= zfs_zevent_retain_expire_secs)
	break;

	/* remove expired node */
	avl_remove(&recent_events_tree, entry);
	list_remove(&recent_events_list, entry);
	kmem_free(entry, sizeof (*entry));
	}

	/* Restart the cleaner if more entries remain */
	recent_events_cleaner_tqid = 0;
	if (!list_is_empty(&recent_events_list))
	zfs_ereport_schedule_cleaner();

	mutex_exit(&recent_events_lock);
	}

	static void
	zfs_ereport_schedule_cleaner(void)
	{
	ASSERT(MUTEX_HELD(&recent_events_lock));

	uint64_t timeout = SEC2NSEC(zfs_zevent_retain_expire_secs + 1);

	recent_events_cleaner_tqid = taskq_dispatch_delay(
	system_delay_taskq, zfs_ereport_cleaner, NULL, TQ_SLEEP,
	ddi_get_lbolt() + NSEC_TO_TICK(timeout));
	}

	/*
	* Clear entries for a given vdev or all vdevs in a pool when vdev == NULL
	*/
	void
	zfs_ereport_clear(spa_t spa, vdev_t vd)
	{
	uint64_t vdev_guid, pool_guid;
	int cnt = 0;

	ASSERT(vd != NULL \|\| spa != NULL);
	if (vd == NULL) {
	vdev_guid = 0;
	pool_guid = spa_guid(spa);
	} else {
	vdev_guid = vd->vdev_guid;
	pool_guid = 0;
	}

	mutex_enter(&recent_events_lock);

	recent_events_node_t *next = list_head(&recent_events_list);
	while (next != NULL) {
	recent_events_node_t *entry = next;

	next = list_next(&recent_events_list, next);

	if (entry->re_vdev_guid == vdev_guid \|\|
	entry->re_pool_guid == pool_guid) {
	avl_remove(&recent_events_tree, entry);
	list_remove(&recent_events_list, entry);
	kmem_free(entry, sizeof (*entry));
	cnt++;
	}
	}

	mutex_exit(&recent_events_lock);
	}

	/*
	* Check if an ereport would be a duplicate of one recently posted.
	*
	* An ereport is considered a duplicate if the set of criteria in
	* recent_events_node_t all match.
	*
	* Only FM_EREPORT_ZFS_IO, FM_EREPORT_ZFS_DATA, and FM_EREPORT_ZFS_CHECKSUM
	* are candidates for duplicate checking.
	*/
	static boolean_t
	zfs_ereport_is_duplicate(const char subclass, spa_t spa, vdev_t *vd,
	const zbookmark_phys_t zb, zio_t zio, uint64_t offset, uint64_t size)
	{
	recent_events_node_t search = {0}, *entry;

	if (vd == NULL \|\| zio == NULL)
	return (B_FALSE);

	if (zfs_zevent_retain_max == 0)
	return (B_FALSE);

	if (strcmp(subclass, FM_EREPORT_ZFS_IO) == 0)
	search.re_subclass = ZSC_IO;
	else if (strcmp(subclass, FM_EREPORT_ZFS_DATA) == 0)
	search.re_subclass = ZSC_DATA;
	else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0)
	search.re_subclass = ZSC_CHECKSUM;
	else
	return (B_FALSE);

	search.re_pool_guid = spa_guid(spa);
	search.re_vdev_guid = vd->vdev_guid;
	search.re_io_error = zio->io_error;
	search.re_io_priority = zio->io_priority;
	/* if size is supplied use it over what's in zio */
	if (size) {
	search.re_io_size = size;
	search.re_io_offset = offset;
	} else {
	search.re_io_size = zio->io_size;
	search.re_io_offset = zio->io_offset;
	}

	/* grab optional logical zio criteria */
	if (zb != NULL) {
	search.re_io_bookmark.zb_objset = zb->zb_objset;
	search.re_io_bookmark.zb_object = zb->zb_object;
	search.re_io_bookmark.zb_level = zb->zb_level;
	search.re_io_bookmark.zb_blkid = zb->zb_blkid;
	}

	uint64_t now = gethrtime();

	mutex_enter(&recent_events_lock);

	/* check if we have seen this one recently */
	entry = avl_find(&recent_events_tree, &search, NULL);
	if (entry != NULL) {
	uint64_t age = NSEC2SEC(now - entry->re_timestamp);

	/*
	* There is still an active cleaner (since we're here).
	* Reset the last seen time for this duplicate entry
	* so that its lifespand gets extended.
	*/
	list_remove(&recent_events_list, entry);
	list_insert_head(&recent_events_list, entry);
	entry->re_timestamp = now;

	zfs_zevent_track_duplicate();
	mutex_exit(&recent_events_lock);

	return (age <= zfs_zevent_retain_expire_secs);
	}

	if (avl_numnodes(&recent_events_tree) >= zfs_zevent_retain_max) {
	/* recycle oldest node */
	entry = list_tail(&recent_events_list);
	ASSERT(entry != NULL);
	list_remove(&recent_events_list, entry);
	avl_remove(&recent_events_tree, entry);
	} else {
	entry = kmem_alloc(sizeof (recent_events_node_t), KM_SLEEP);
	}

	/* record this as a recent ereport */
	*entry = search;
	avl_add(&recent_events_tree, entry);
	list_insert_head(&recent_events_list, entry);
	entry->re_timestamp = now;

	/* Start a cleaner if not already scheduled */
	if (recent_events_cleaner_tqid == 0)
	zfs_ereport_schedule_cleaner();

	mutex_exit(&recent_events_lock);
	return (B_FALSE);
	}

	void
	zfs_zevent_post_cb(nvlist_t nvl, nvlist_t detector)
	{
	if (nvl)
	fm_nvlist_destroy(nvl, FM_NVA_FREE);

	if (detector)
	fm_nvlist_destroy(detector, FM_NVA_FREE);
	}

	/*
	* We want to rate limit ZIO delay, deadman, and checksum events so as to not
	* flood zevent consumers when a disk is acting up.
	*
	* Returns 1 if we're ratelimiting, 0 if not.
	*/
	static int
	zfs_is_ratelimiting_event(const char subclass, vdev_t vd)
	{
	int rc = 0;
	/*
	* zfs_ratelimit() returns 1 if we're not ratelimiting and 0 if we
	* are. Invert it to get our return value.
	*/
	if (strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) {
	rc = !zfs_ratelimit(&vd->vdev_delay_rl);
	} else if (strcmp(subclass, FM_EREPORT_ZFS_DEADMAN) == 0) {
	rc = !zfs_ratelimit(&vd->vdev_deadman_rl);
	} else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0) {
	rc = !zfs_ratelimit(&vd->vdev_checksum_rl);
	}

	if (rc) {
	/* We're rate limiting */
	fm_erpt_dropped_increment();
	}

	return (rc);
	}

	/*
	* Return B_TRUE if the event actually posted, B_FALSE if not.
	*/
	static boolean_t
	zfs_ereport_start(nvlist_t ereport_out, nvlist_t detector_out,
	const char subclass, spa_t spa, vdev_t vd, const zbookmark_phys_t zb,
	zio_t *zio, uint64_t stateoroffset, uint64_t size)
	{
	nvlist_t ereport, detector;

	uint64_t ena;
	char class[64];

	if ((ereport = fm_nvlist_create(NULL)) == NULL)
	return (B_FALSE);

	if ((detector = fm_nvlist_create(NULL)) == NULL) {
	fm_nvlist_destroy(ereport, FM_NVA_FREE);
	return (B_FALSE);
	}

	/*
	* Serialize ereport generation
	*/
	mutex_enter(&spa->spa_errlist_lock);

	/*
	* Determine the ENA to use for this event. If we are in a loading
	* state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use
	* a root zio-wide ENA. Otherwise, simply use a unique ENA.
	*/
	if (spa_load_state(spa) != SPA_LOAD_NONE) {
	if (spa->spa_ena == 0)
	spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
	ena = spa->spa_ena;
	} else if (zio != NULL && zio->io_logical != NULL) {
	if (zio->io_logical->io_ena == 0)
	zio->io_logical->io_ena =
	fm_ena_generate(0, FM_ENA_FMT1);
	ena = zio->io_logical->io_ena;
	} else {
	ena = fm_ena_generate(0, FM_ENA_FMT1);
	}

	/*
	* Construct the full class, detector, and other standard FMA fields.
	*/
	(void) snprintf(class, sizeof (class), "%s.%s",
	ZFS_ERROR_CLASS, subclass);

	fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa),
	vd != NULL ? vd->vdev_guid : 0);

	fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL);

	/*
	* Construct the per-ereport payload, depending on which parameters are
	* passed in.
	*/

	/*
	* Generic payload members common to all ereports.
	*/
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_POOL, DATA_TYPE_STRING, spa_name(spa),
	FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, DATA_TYPE_UINT64, spa_guid(spa),
	FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, DATA_TYPE_UINT64,
	(uint64_t)spa_state(spa),
	FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
	(int32_t)spa_load_state(spa), NULL);

	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
	DATA_TYPE_STRING,
	spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
	FM_EREPORT_FAILMODE_WAIT :
	spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
	FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC,
	NULL);

	if (vd != NULL) {
	vdev_t *pvd = vd->vdev_parent;
	vdev_queue_t *vq = &vd->vdev_queue;
	vdev_stat_t *vs = &vd->vdev_stat;
	vdev_t *spare_vd;
	uint64_t *spare_guids;
	char **spare_paths;
	int i, spare_count;

	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
	DATA_TYPE_UINT64, vd->vdev_guid,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
	DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
	if (vd->vdev_path != NULL)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
	DATA_TYPE_STRING, vd->vdev_path, NULL);
	if (vd->vdev_devid != NULL)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
	DATA_TYPE_STRING, vd->vdev_devid, NULL);
	if (vd->vdev_fru != NULL)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU,
	DATA_TYPE_STRING, vd->vdev_fru, NULL);
	if (vd->vdev_enc_sysfs_path != NULL)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
	DATA_TYPE_STRING, vd->vdev_enc_sysfs_path, NULL);
	if (vd->vdev_ashift)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_ASHIFT,
	DATA_TYPE_UINT64, vd->vdev_ashift, NULL);

	if (vq != NULL) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_COMP_TS,
	DATA_TYPE_UINT64, vq->vq_io_complete_ts, NULL);
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_DELTA_TS,
	DATA_TYPE_UINT64, vq->vq_io_delta_ts, NULL);
	}

	if (vs != NULL) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_READ_ERRORS,
	DATA_TYPE_UINT64, vs->vs_read_errors,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_WRITE_ERRORS,
	DATA_TYPE_UINT64, vs->vs_write_errors,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_CKSUM_ERRORS,
	DATA_TYPE_UINT64, vs->vs_checksum_errors,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_DELAYS,
	DATA_TYPE_UINT64, vs->vs_slow_ios,
	NULL);
	}

	if (pvd != NULL) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID,
	DATA_TYPE_UINT64, pvd->vdev_guid,
	FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
	DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type,
	NULL);
	if (pvd->vdev_path)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
	DATA_TYPE_STRING, pvd->vdev_path, NULL);
	if (pvd->vdev_devid)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
	DATA_TYPE_STRING, pvd->vdev_devid, NULL);
	}

	spare_count = spa->spa_spares.sav_count;
	spare_paths = kmem_zalloc(sizeof (char ) spare_count,
	KM_SLEEP);
	spare_guids = kmem_zalloc(sizeof (uint64_t) * spare_count,
	KM_SLEEP);

	for (i = 0; i < spare_count; i++) {
	spare_vd = spa->spa_spares.sav_vdevs[i];
	if (spare_vd) {
	spare_paths[i] = spare_vd->vdev_path;
	spare_guids[i] = spare_vd->vdev_guid;
	}
	}

	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_PATHS,
	DATA_TYPE_STRING_ARRAY, spare_count, spare_paths,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_GUIDS,
	DATA_TYPE_UINT64_ARRAY, spare_count, spare_guids, NULL);

	kmem_free(spare_guids, sizeof (uint64_t) * spare_count);
	kmem_free(spare_paths, sizeof (char ) spare_count);
	}

	if (zio != NULL) {
	/*
	* Payload common to all I/Os.
	*/
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
	DATA_TYPE_INT32, zio->io_error, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS,
	DATA_TYPE_INT32, zio->io_flags, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_STAGE,
	DATA_TYPE_UINT32, zio->io_stage, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PIPELINE,
	DATA_TYPE_UINT32, zio->io_pipeline, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELAY,
	DATA_TYPE_UINT64, zio->io_delay, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP,
	DATA_TYPE_UINT64, zio->io_timestamp, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA,
	DATA_TYPE_UINT64, zio->io_delta, NULL);
	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PRIORITY,
	DATA_TYPE_UINT32, zio->io_priority, NULL);

	/*
	* If the 'size' parameter is non-zero, it indicates this is a
	* RAID-Z or other I/O where the physical offset and length are
	* provided for us, instead of within the zio_t.
	*/
	if (vd != NULL) {
	if (size)
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
	DATA_TYPE_UINT64, stateoroffset,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
	DATA_TYPE_UINT64, size, NULL);
	else
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
	DATA_TYPE_UINT64, zio->io_offset,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
	DATA_TYPE_UINT64, zio->io_size, NULL);
	}
	} else if (vd != NULL) {
	/*
	* If we have a vdev but no zio, this is a device fault, and the
	* 'stateoroffset' parameter indicates the previous state of the
	* vdev.
	*/
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
	DATA_TYPE_UINT64, stateoroffset, NULL);
	}

	/*
	* Payload for I/Os with corresponding logical information.
	*/
	if (zb != NULL && (zio == NULL \|\| zio->io_logical != NULL)) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET,
	DATA_TYPE_UINT64, zb->zb_objset,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
	DATA_TYPE_UINT64, zb->zb_object,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
	DATA_TYPE_INT64, zb->zb_level,
	FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
	DATA_TYPE_UINT64, zb->zb_blkid, NULL);
	}

	mutex_exit(&spa->spa_errlist_lock);

	*ereport_out = ereport;
	*detector_out = detector;
	return (B_TRUE);
	}

	/* if it's <= 128 bytes, save the corruption directly */
	#define ZFM_MAX_INLINE (128 / sizeof (uint64_t))

	#define MAX_RANGES 16

	typedef struct zfs_ecksum_info {
	/* histograms of set and cleared bits by bit number in a 64-bit word */
	uint32_t zei_histogram_set[sizeof (uint64_t) * NBBY];
	uint32_t zei_histogram_cleared[sizeof (uint64_t) * NBBY];

	/* inline arrays of bits set and cleared. */
	uint64_t zei_bits_set[ZFM_MAX_INLINE];
	uint64_t zei_bits_cleared[ZFM_MAX_INLINE];

	/*
	* for each range, the number of bits set and cleared. The Hamming
	* distance between the good and bad buffers is the sum of them all.
	*/
	uint32_t zei_range_sets[MAX_RANGES];
	uint32_t zei_range_clears[MAX_RANGES];

	struct zei_ranges {
	uint32_t zr_start;
	uint32_t zr_end;
	} zei_ranges[MAX_RANGES];

	size_t zei_range_count;
	uint32_t zei_mingap;
	uint32_t zei_allowed_mingap;

	} zfs_ecksum_info_t;

	static void
	update_histogram(uint64_t value_arg, uint32_t hist, uint32_t count)
	{
	size_t i;
	size_t bits = 0;
	uint64_t value = BE_64(value_arg);

	/* We store the bits in big-endian (largest-first) order */
	for (i = 0; i < 64; i++) {
	if (value & (1ull << i)) {
	hist[63 - i]++;
	++bits;
	}
	}
	/* update the count of bits changed */
	*count += bits;
	}

	/*
	* We've now filled up the range array, and need to increase "mingap" and
	* shrink the range list accordingly. zei_mingap is always the smallest
	* distance between array entries, so we set the new_allowed_gap to be
	* one greater than that. We then go through the list, joining together
	* any ranges which are closer than the new_allowed_gap.
	*
	* By construction, there will be at least one. We also update zei_mingap
	* to the new smallest gap, to prepare for our next invocation.
	*/
	static void
	zei_shrink_ranges(zfs_ecksum_info_t *eip)
	{
	uint32_t mingap = UINT32_MAX;
	uint32_t new_allowed_gap = eip->zei_mingap + 1;

	size_t idx, output;
	size_t max = eip->zei_range_count;

	struct zei_ranges *r = eip->zei_ranges;

	ASSERT3U(eip->zei_range_count, >, 0);
	ASSERT3U(eip->zei_range_count, <=, MAX_RANGES);

	output = idx = 0;
	while (idx < max - 1) {
	uint32_t start = r[idx].zr_start;
	uint32_t end = r[idx].zr_end;

	while (idx < max - 1) {
	idx++;

	uint32_t nstart = r[idx].zr_start;
	uint32_t nend = r[idx].zr_end;

	uint32_t gap = nstart - end;
	if (gap < new_allowed_gap) {
	end = nend;
	continue;
	}
	if (gap < mingap)
	mingap = gap;
	break;
	}
	r[output].zr_start = start;
	r[output].zr_end = end;
	output++;
	}
	ASSERT3U(output, <, eip->zei_range_count);
	eip->zei_range_count = output;
	eip->zei_mingap = mingap;
	eip->zei_allowed_mingap = new_allowed_gap;
	}

	static void
	zei_add_range(zfs_ecksum_info_t *eip, int start, int end)
	{
	struct zei_ranges *r = eip->zei_ranges;
	size_t count = eip->zei_range_count;

	if (count >= MAX_RANGES) {
	zei_shrink_ranges(eip);
	count = eip->zei_range_count;
	}
	if (count == 0) {
	eip->zei_mingap = UINT32_MAX;
	eip->zei_allowed_mingap = 1;
	} else {
	int gap = start - r[count - 1].zr_end;

	if (gap < eip->zei_allowed_mingap) {
	r[count - 1].zr_end = end;
	return;
	}
	if (gap < eip->zei_mingap)
	eip->zei_mingap = gap;
	}
	r[count].zr_start = start;
	r[count].zr_end = end;
	eip->zei_range_count++;
	}

	static size_t
	zei_range_total_size(zfs_ecksum_info_t *eip)
	{
	struct zei_ranges *r = eip->zei_ranges;
	size_t count = eip->zei_range_count;
	size_t result = 0;
	size_t idx;

	for (idx = 0; idx < count; idx++)
	result += (r[idx].zr_end - r[idx].zr_start);

	return (result);
	}

	static zfs_ecksum_info_t *
	annotate_ecksum(nvlist_t ereport, zio_bad_cksum_t info,
	const abd_t goodabd, const abd_t badabd, size_t size,
	boolean_t drop_if_identical)
	{
	const uint64_t *good;
	const uint64_t *bad;

	size_t nui64s = size / sizeof (uint64_t);

	size_t inline_size;
	int no_inline = 0;
	size_t idx;
	size_t range;

	size_t offset = 0;
	ssize_t start = -1;

	zfs_ecksum_info_t eip = kmem_zalloc(sizeof (eip), KM_SLEEP);

	/* don't do any annotation for injected checksum errors */
	if (info != NULL && info->zbc_injected)
	return (eip);

	if (info != NULL && info->zbc_has_cksum) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED,
	DATA_TYPE_UINT64_ARRAY,
	sizeof (info->zbc_expected) / sizeof (uint64_t),
	(uint64_t *)&info->zbc_expected,
	FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL,
	DATA_TYPE_UINT64_ARRAY,
	sizeof (info->zbc_actual) / sizeof (uint64_t),
	(uint64_t *)&info->zbc_actual,
	FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO,
	DATA_TYPE_STRING,
	info->zbc_checksum_name,
	NULL);

	if (info->zbc_byteswapped) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP,
	DATA_TYPE_BOOLEAN, 1,
	NULL);
	}
	}

	if (badabd == NULL \|\| goodabd == NULL)
	return (eip);

	ASSERT3U(nui64s, <=, UINT32_MAX);
	ASSERT3U(size, ==, nui64s * sizeof (uint64_t));
	ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
	ASSERT3U(size, <=, UINT32_MAX);

	good = (const uint64_t ) abd_borrow_buf_copy((abd_t )goodabd, size);
	bad = (const uint64_t ) abd_borrow_buf_copy((abd_t )badabd, size);

	/* build up the range list by comparing the two buffers. */
	for (idx = 0; idx < nui64s; idx++) {
	if (good[idx] == bad[idx]) {
	if (start == -1)
	continue;

	zei_add_range(eip, start, idx);
	start = -1;
	} else {
	if (start != -1)
	continue;

	start = idx;
	}
	}
	if (start != -1)
	zei_add_range(eip, start, idx);

	/* See if it will fit in our inline buffers */
	inline_size = zei_range_total_size(eip);
	if (inline_size > ZFM_MAX_INLINE)
	no_inline = 1;

	/*
	* If there is no change and we want to drop if the buffers are
	* identical, do so.
	*/
	if (inline_size == 0 && drop_if_identical) {
	kmem_free(eip, sizeof (*eip));
	abd_return_buf((abd_t )goodabd, (void )good, size);
	abd_return_buf((abd_t )badabd, (void )bad, size);
	return (NULL);
	}

	/*
	* Now walk through the ranges, filling in the details of the
	* differences. Also convert our uint64_t-array offsets to byte
	* offsets.
	*/
	for (range = 0; range < eip->zei_range_count; range++) {
	size_t start = eip->zei_ranges[range].zr_start;
	size_t end = eip->zei_ranges[range].zr_end;

	for (idx = start; idx < end; idx++) {
	uint64_t set, cleared;

	// bits set in bad, but not in good
	set = ((~good[idx]) & bad[idx]);
	// bits set in good, but not in bad
	cleared = (good[idx] & (~bad[idx]));

	if (!no_inline) {
	ASSERT3U(offset, <, inline_size);
	eip->zei_bits_set[offset] = set;
	eip->zei_bits_cleared[offset] = cleared;
	offset++;
	}

	update_histogram(set, eip->zei_histogram_set,
	&eip->zei_range_sets[range]);
	update_histogram(cleared, eip->zei_histogram_cleared,
	&eip->zei_range_clears[range]);
	}

	/* convert to byte offsets */
	eip->zei_ranges[range].zr_start *= sizeof (uint64_t);
	eip->zei_ranges[range].zr_end *= sizeof (uint64_t);
	}

	abd_return_buf((abd_t )goodabd, (void )good, size);
	abd_return_buf((abd_t )badabd, (void )bad, size);

	eip->zei_allowed_mingap *= sizeof (uint64_t);
	inline_size *= sizeof (uint64_t);

	/* fill in ereport */
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES,
	DATA_TYPE_UINT32_ARRAY, 2 * eip->zei_range_count,
	(uint32_t *)eip->zei_ranges,
	FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_MIN_GAP,
	DATA_TYPE_UINT32, eip->zei_allowed_mingap,
	FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_SETS,
	DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_sets,
	FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS,
	DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_clears,
	NULL);

	if (!no_inline) {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS,
	DATA_TYPE_UINT8_ARRAY,
	inline_size, (uint8_t *)eip->zei_bits_set,
	FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS,
	DATA_TYPE_UINT8_ARRAY,
	inline_size, (uint8_t *)eip->zei_bits_cleared,
	NULL);
	} else {
	fm_payload_set(ereport,
	FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM,
	DATA_TYPE_UINT32_ARRAY,
	NBBY * sizeof (uint64_t), eip->zei_histogram_set,
	FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM,
	DATA_TYPE_UINT32_ARRAY,
	NBBY * sizeof (uint64_t), eip->zei_histogram_cleared,
	NULL);
	}
	return (eip);
	}
	#else
	void
	zfs_ereport_clear(spa_t spa, vdev_t vd)
	{
	(void) spa, (void) vd;
	}
	#endif

	/*
	* Make sure our event is still valid for the given zio/vdev/pool. For example,
	* we don't want to keep logging events for a faulted or missing vdev.
	*/
	boolean_t
	zfs_ereport_is_valid(const char subclass, spa_t spa, vdev_t vd, zio_t zio)
	{
	#ifdef _KERNEL
	/*
	* If we are doing a spa_tryimport() or in recovery mode,
	* ignore errors.
	*/
	if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT \|\|
	spa_load_state(spa) == SPA_LOAD_RECOVER)
	return (B_FALSE);

	/*
	* If we are in the middle of opening a pool, and the previous attempt
	* failed, don't bother logging any new ereports - we're just going to
	* get the same diagnosis anyway.
	*/
	if (spa_load_state(spa) != SPA_LOAD_NONE &&
	spa->spa_last_open_failed)
	return (B_FALSE);

	if (zio != NULL) {
	/*
	* If this is not a read or write zio, ignore the error. This
	* can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
	*/
	if (zio->io_type != ZIO_TYPE_READ &&
	zio->io_type != ZIO_TYPE_WRITE)
	return (B_FALSE);

	if (vd != NULL) {
	/*
	* If the vdev has already been marked as failing due
	* to a failed probe, then ignore any subsequent I/O
	* errors, as the DE will automatically fault the vdev
	* on the first such failure. This also catches cases
	* where vdev_remove_wanted is set and the device has
	* not yet been asynchronously placed into the REMOVED
	* state.
	*/
	if (zio->io_vd == vd && !vdev_accessible(vd, zio))
	return (B_FALSE);

	/*
	* Ignore checksum errors for reads from DTL regions of
	* leaf vdevs.
	*/
	if (zio->io_type == ZIO_TYPE_READ &&
	zio->io_error == ECKSUM &&
	vd->vdev_ops->vdev_op_leaf &&
	vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1))
	return (B_FALSE);
	}
	}

	/*
	* For probe failure, we want to avoid posting ereports if we've
	* already removed the device in the meantime.
	*/
	if (vd != NULL &&
	strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) == 0 &&
	(vd->vdev_remove_wanted \|\| vd->vdev_state == VDEV_STATE_REMOVED))
	return (B_FALSE);

	/* Ignore bogus delay events (like from ioctls or unqueued IOs) */
	if ((strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) &&
	(zio != NULL) && (!zio->io_timestamp)) {
	return (B_FALSE);
	}
	#else
	(void) subclass, (void) spa, (void) vd, (void) zio;
	#endif
	return (B_TRUE);
	}

	/*
	* Post an ereport for the given subclass
	*
	* Returns
	* - 0 if an event was posted
	* - EINVAL if there was a problem posting event
	* - EBUSY if the event was rate limited
	* - EALREADY if the event was already posted (duplicate)
	*/
	int
	zfs_ereport_post(const char subclass, spa_t spa, vdev_t *vd,
	const zbookmark_phys_t zb, zio_t zio, uint64_t state)
	{
	int rc = 0;
	#ifdef _KERNEL
	nvlist_t *ereport = NULL;
	nvlist_t *detector = NULL;

	if (!zfs_ereport_is_valid(subclass, spa, vd, zio))
	return (EINVAL);

	if (zfs_ereport_is_duplicate(subclass, spa, vd, zb, zio, 0, 0))
	return (SET_ERROR(EALREADY));

	if (zfs_is_ratelimiting_event(subclass, vd))
	return (SET_ERROR(EBUSY));

	if (!zfs_ereport_start(&ereport, &detector, subclass, spa, vd,
	zb, zio, state, 0))
	return (SET_ERROR(EINVAL)); /* couldn't post event */

	if (ereport == NULL)
	return (SET_ERROR(EINVAL));

	/* Cleanup is handled by the callback function */
	rc = zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
	#else
	(void) subclass, (void) spa, (void) vd, (void) zb, (void) zio,
	(void) state;
	#endif
	return (rc);
	}

	/*
	* Prepare a checksum ereport
	*
	* Returns
	* - 0 if an event was posted
	* - EINVAL if there was a problem posting event
	* - EBUSY if the event was rate limited
	* - EALREADY if the event was already posted (duplicate)
	*/
	int
	zfs_ereport_start_checksum(spa_t spa, vdev_t vd, const zbookmark_phys_t *zb,
	struct zio zio, uint64_t offset, uint64_t length, zio_bad_cksum_t info)
	{
	zio_cksum_report_t *report;

	#ifdef _KERNEL
	if (!zfs_ereport_is_valid(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio))
	return (SET_ERROR(EINVAL));

	if (zfs_ereport_is_duplicate(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio,
	offset, length))
	return (SET_ERROR(EALREADY));

	if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd))
	return (SET_ERROR(EBUSY));
	#else
	(void) zb, (void) offset;
	#endif

	report = kmem_zalloc(sizeof (*report), KM_SLEEP);

	zio_vsd_default_cksum_report(zio, report);

	/* copy the checksum failure information if it was provided */
	if (info != NULL) {
	report->zcr_ckinfo = kmem_zalloc(sizeof (*info), KM_SLEEP);
	memcpy(report->zcr_ckinfo, info, sizeof (*info));
	}

	report->zcr_sector = 1ULL << vd->vdev_top->vdev_ashift;
	report->zcr_align =
	vdev_psize_to_asize(vd->vdev_top, report->zcr_sector);
	report->zcr_length = length;

	#ifdef _KERNEL
	(void) zfs_ereport_start(&report->zcr_ereport, &report->zcr_detector,
	FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio, offset, length);

	if (report->zcr_ereport == NULL) {
	zfs_ereport_free_checksum(report);
	return (0);
	}
	#endif

	mutex_enter(&spa->spa_errlist_lock);
	report->zcr_next = zio->io_logical->io_cksum_report;
	zio->io_logical->io_cksum_report = report;
	mutex_exit(&spa->spa_errlist_lock);
	return (0);
	}

	void
	zfs_ereport_finish_checksum(zio_cksum_report_t report, const abd_t good_data,
	const abd_t *bad_data, boolean_t drop_if_identical)
	{
	#ifdef _KERNEL
	zfs_ecksum_info_t *info;

	info = annotate_ecksum(report->zcr_ereport, report->zcr_ckinfo,
	good_data, bad_data, report->zcr_length, drop_if_identical);
	if (info != NULL)
	zfs_zevent_post(report->zcr_ereport,
	report->zcr_detector, zfs_zevent_post_cb);
	else
	zfs_zevent_post_cb(report->zcr_ereport, report->zcr_detector);

	report->zcr_ereport = report->zcr_detector = NULL;
	if (info != NULL)
	kmem_free(info, sizeof (*info));
	#else
	(void) report, (void) good_data, (void) bad_data,
	(void) drop_if_identical;
	#endif
	}

	void
	zfs_ereport_free_checksum(zio_cksum_report_t *rpt)
	{
	#ifdef _KERNEL
	if (rpt->zcr_ereport != NULL) {
	fm_nvlist_destroy(rpt->zcr_ereport,
	FM_NVA_FREE);
	fm_nvlist_destroy(rpt->zcr_detector,
	FM_NVA_FREE);
	}
	#endif
	rpt->zcr_free(rpt->zcr_cbdata, rpt->zcr_cbinfo);

	if (rpt->zcr_ckinfo != NULL)
	kmem_free(rpt->zcr_ckinfo, sizeof (*rpt->zcr_ckinfo));

	kmem_free(rpt, sizeof (*rpt));
	}

	/*
	* Post a checksum ereport
	*
	* Returns
	* - 0 if an event was posted
	* - EINVAL if there was a problem posting event
	* - EBUSY if the event was rate limited
	* - EALREADY if the event was already posted (duplicate)
	*/
	int
	zfs_ereport_post_checksum(spa_t spa, vdev_t vd, const zbookmark_phys_t *zb,
	struct zio *zio, uint64_t offset, uint64_t length,
	const abd_t good_data, const abd_t bad_data, zio_bad_cksum_t *zbc)
	{
	int rc = 0;
	#ifdef _KERNEL
	nvlist_t *ereport = NULL;
	nvlist_t *detector = NULL;
	zfs_ecksum_info_t *info;

	if (!zfs_ereport_is_valid(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio))
	return (SET_ERROR(EINVAL));

	if (zfs_ereport_is_duplicate(FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio,
	offset, length))
	return (SET_ERROR(EALREADY));

	if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd))
	return (SET_ERROR(EBUSY));

	if (!zfs_ereport_start(&ereport, &detector, FM_EREPORT_ZFS_CHECKSUM,
	spa, vd, zb, zio, offset, length) \|\| (ereport == NULL)) {
	return (SET_ERROR(EINVAL));
	}

	info = annotate_ecksum(ereport, zbc, good_data, bad_data, length,
	B_FALSE);

	if (info != NULL) {
	rc = zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
	kmem_free(info, sizeof (*info));
	}
	#else
	(void) spa, (void) vd, (void) zb, (void) zio, (void) offset,
	(void) length, (void) good_data, (void) bad_data, (void) zbc;
	#endif
	return (rc);
	}

	/*
	* The 'sysevent.fs.zfs.*' events are signals posted to notify user space of
	* change in the pool. All sysevents are listed in sys/sysevent/eventdefs.h
	* and are designed to be consumed by the ZFS Event Daemon (ZED). For
	* additional details refer to the zed(8) man page.
	*/
	nvlist_t *
	zfs_event_create(spa_t spa, vdev_t vd, const char type, const char name,
	nvlist_t *aux)
	{
	nvlist_t *resource = NULL;
	#ifdef _KERNEL
	char class[64];

	if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT)
	return (NULL);

	if ((resource = fm_nvlist_create(NULL)) == NULL)
	return (NULL);

	(void) snprintf(class, sizeof (class), "%s.%s.%s", type,
	ZFS_ERROR_CLASS, name);
	VERIFY0(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION));
	VERIFY0(nvlist_add_string(resource, FM_CLASS, class));
	VERIFY0(nvlist_add_string(resource,
	FM_EREPORT_PAYLOAD_ZFS_POOL, spa_name(spa)));
	VERIFY0(nvlist_add_uint64(resource,
	FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa)));
	VERIFY0(nvlist_add_uint64(resource,
	FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, spa_state(spa)));
	VERIFY0(nvlist_add_int32(resource,
	FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, spa_load_state(spa)));

	if (vd) {
	VERIFY0(nvlist_add_uint64(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid));
	VERIFY0(nvlist_add_uint64(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, vd->vdev_state));
	if (vd->vdev_path != NULL)
	VERIFY0(nvlist_add_string(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, vd->vdev_path));
	if (vd->vdev_devid != NULL)
	VERIFY0(nvlist_add_string(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, vd->vdev_devid));
	if (vd->vdev_fru != NULL)
	VERIFY0(nvlist_add_string(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU, vd->vdev_fru));
	if (vd->vdev_enc_sysfs_path != NULL)
	VERIFY0(nvlist_add_string(resource,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
	vd->vdev_enc_sysfs_path));
	}

	/* also copy any optional payload data */
	if (aux) {
	nvpair_t *elem = NULL;

	while ((elem = nvlist_next_nvpair(aux, elem)) != NULL)
	(void) nvlist_add_nvpair(resource, elem);
	}
	#else
	(void) spa, (void) vd, (void) type, (void) name, (void) aux;
	#endif
	return (resource);
	}

	static void
	zfs_post_common(spa_t spa, vdev_t vd, const char type, const char name,
	nvlist_t *aux)
	{
	#ifdef _KERNEL
	nvlist_t *resource;

	resource = zfs_event_create(spa, vd, type, name, aux);
	if (resource)
	zfs_zevent_post(resource, NULL, zfs_zevent_post_cb);
	#else
	(void) spa, (void) vd, (void) type, (void) name, (void) aux;
	#endif
	}

	/*
	* The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
	* has been removed from the system. This will cause the DE to ignore any
	* recent I/O errors, inferring that they are due to the asynchronous device
	* removal.
	*/
	void
	zfs_post_remove(spa_t spa, vdev_t vd)
	{
	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_REMOVED, NULL);
	}

	/*
	* The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
	* has the 'autoreplace' property set, and therefore any broken vdevs will be
	* handled by higher level logic, and no vdev fault should be generated.
	*/
	void
	zfs_post_autoreplace(spa_t spa, vdev_t vd)
	{
	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_AUTOREPLACE, NULL);
	}

	/*
	* The 'resource.fs.zfs.statechange' event is an internal signal that the
	* given vdev has transitioned its state to DEGRADED or HEALTHY. This will
	* cause the retire agent to repair any outstanding fault management cases
	* open because the device was not found (fault.fs.zfs.device).
	*/
	void
	zfs_post_state_change(spa_t spa, vdev_t vd, uint64_t laststate)
	{
	#ifdef _KERNEL
	nvlist_t *aux;

	/*
	* Add optional supplemental keys to payload
	*/
	aux = fm_nvlist_create(NULL);
	if (vd && aux) {
	if (vd->vdev_physpath) {
	- (void) nvlist_add_string(aux,
	+ fnvlist_add_string(aux,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_PHYSPATH,
	vd->vdev_physpath);
	}
	if (vd->vdev_enc_sysfs_path) {
	- (void) nvlist_add_string(aux,
	+ fnvlist_add_string(aux,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
	vd->vdev_enc_sysfs_path);
	}

	- (void) nvlist_add_uint64(aux,
	+ fnvlist_add_uint64(aux,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE, laststate);
	}

	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_STATECHANGE,
	aux);

	if (aux)
	fm_nvlist_destroy(aux, FM_NVA_FREE);
	#else
	(void) spa, (void) vd, (void) laststate;
	#endif
	}

	#ifdef _KERNEL
	void
	zfs_ereport_init(void)
	{
	mutex_init(&recent_events_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&recent_events_list, sizeof (recent_events_node_t),
	offsetof(recent_events_node_t, re_list_link));
	avl_create(&recent_events_tree, recent_events_compare,
	sizeof (recent_events_node_t), offsetof(recent_events_node_t,
	re_tree_link));
	}

	/*
	* This 'early' fini needs to run before zfs_fini() which on Linux waits
	* for the system_delay_taskq to drain.
	*/
	void
	zfs_ereport_taskq_fini(void)
	{
	mutex_enter(&recent_events_lock);
	if (recent_events_cleaner_tqid != 0) {
	taskq_cancel_id(system_delay_taskq, recent_events_cleaner_tqid);
	recent_events_cleaner_tqid = 0;
	}
	mutex_exit(&recent_events_lock);
	}

	void
	zfs_ereport_fini(void)
	{
	recent_events_node_t *entry;

	while ((entry = list_head(&recent_events_list)) != NULL) {
	avl_remove(&recent_events_tree, entry);
	list_remove(&recent_events_list, entry);
	kmem_free(entry, sizeof (*entry));
	}
	avl_destroy(&recent_events_tree);
	list_destroy(&recent_events_list);
	mutex_destroy(&recent_events_lock);
	}

	void
	zfs_ereport_snapshot_post(const char subclass, spa_t spa, const char *name)
	{
	nvlist_t *aux;

	aux = fm_nvlist_create(NULL);
	- nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_SNAPSHOT_NAME, name);
	+ fnvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_SNAPSHOT_NAME, name);

	zfs_post_common(spa, NULL, FM_RSRC_CLASS, subclass, aux);
	fm_nvlist_destroy(aux, FM_NVA_FREE);
	}

	/*
	* Post when a event when a zvol is created or removed
	*
	* This is currently only used by macOS, since it uses the event to create
	* symlinks between the volume name (mypool/myvol) and the actual /dev
	* device (/dev/disk3). For example:
	*
	* /var/run/zfs/dsk/mypool/myvol -> /dev/disk3
	*
	* name: The full name of the zvol ("mypool/myvol")
	* dev_name: The full /dev name for the zvol ("/dev/disk3")
	* raw_name: The raw /dev name for the zvol ("/dev/rdisk3")
	*/
	void
	zfs_ereport_zvol_post(const char subclass, const char name,
	const char dev_name, const char raw_name)
	{
	nvlist_t *aux;
	char *r;

	boolean_t locked = mutex_owned(&spa_namespace_lock);
	if (!locked) mutex_enter(&spa_namespace_lock);
	spa_t *spa = spa_lookup(name);
	if (!locked) mutex_exit(&spa_namespace_lock);

	if (spa == NULL)
	return;

	aux = fm_nvlist_create(NULL);
	- nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_DEVICE_NAME, dev_name);
	- nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_RAW_DEVICE_NAME,
	+ fnvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_DEVICE_NAME, dev_name);
	+ fnvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_RAW_DEVICE_NAME,
	raw_name);
	r = strchr(name, '/');
	if (r && r[1])
	- nvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_VOLUME, &r[1]);
	+ fnvlist_add_string(aux, FM_EREPORT_PAYLOAD_ZFS_VOLUME, &r[1]);

	zfs_post_common(spa, NULL, FM_RSRC_CLASS, subclass, aux);
	fm_nvlist_destroy(aux, FM_NVA_FREE);
	}

	EXPORT_SYMBOL(zfs_ereport_post);
	EXPORT_SYMBOL(zfs_ereport_is_valid);
	EXPORT_SYMBOL(zfs_ereport_post_checksum);
	EXPORT_SYMBOL(zfs_post_remove);
	EXPORT_SYMBOL(zfs_post_autoreplace);
	EXPORT_SYMBOL(zfs_post_state_change);

	ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, retain_max, UINT, ZMOD_RW,
	"Maximum recent zevents records to retain for duplicate checking");
	ZFS_MODULE_PARAM(zfs_zevent, zfs_zevent_, retain_expire_secs, UINT, ZMOD_RW,
	"Expiration time for recent zevents records");
	#endif /* _KERNEL */
	diff --git a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	index 259d68c477de..c3266c09306b 100644
	--- a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	+++ b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	@@ -1,7887 +1,7891 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Portions Copyright 2011 Martin Matuska
	* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
	* Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
	* Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
	* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright 2017 RackTop Systems.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
	* Copyright (c) 2019, 2021, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	/*
	* ZFS ioctls.
	*
	* This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
	* pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
	*
	* There are two ways that we handle ioctls: the legacy way where almost
	* all of the logic is in the ioctl callback, and the new way where most
	* of the marshalling is handled in the common entry point, zfsdev_ioctl().
	*
	* Non-legacy ioctls should be registered by calling
	* zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
	* from userland by lzc_ioctl().
	*
	* The registration arguments are as follows:
	*
	* const char *name
	* The name of the ioctl. This is used for history logging. If the
	* ioctl returns successfully (the callback returns 0), and allow_log
	* is true, then a history log entry will be recorded with the input &
	* output nvlists. The log entry can be printed with "zpool history -i".
	*
	* zfs_ioc_t ioc
	* The ioctl request number, which userland will pass to ioctl(2).
	* We want newer versions of libzfs and libzfs_core to run against
	* existing zfs kernel modules (i.e. a deferred reboot after an update).
	* Therefore the ioctl numbers cannot change from release to release.
	*
	* zfs_secpolicy_func_t *secpolicy
	* This function will be called before the zfs_ioc_func_t, to
	* determine if this operation is permitted. It should return EPERM
	* on failure, and 0 on success. Checks include determining if the
	* dataset is visible in this zone, and if the user has either all
	* zfs privileges in the zone (SYS_MOUNT), or has been granted permission
	* to do this operation on this dataset with "zfs allow".
	*
	* zfs_ioc_namecheck_t namecheck
	* This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
	* name, a dataset name, or nothing. If the name is not well-formed,
	* the ioctl will fail and the callback will not be called.
	* Therefore, the callback can assume that the name is well-formed
	* (e.g. is null-terminated, doesn't have more than one '@' character,
	* doesn't have invalid characters).
	*
	* zfs_ioc_poolcheck_t pool_check
	* This specifies requirements on the pool state. If the pool does
	* not meet them (is suspended or is readonly), the ioctl will fail
	* and the callback will not be called. If any checks are specified
	* (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
	* Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED \|
	* POOL_CHECK_READONLY).
	*
	* zfs_ioc_key_t *nvl_keys
	* The list of expected/allowable innvl input keys. This list is used
	* to validate the nvlist input to the ioctl.
	*
	* boolean_t smush_outnvlist
	* If smush_outnvlist is true, then the output is presumed to be a
	* list of errors, and it will be "smushed" down to fit into the
	* caller's buffer, by removing some entries and replacing them with a
	* single "N_MORE_ERRORS" entry indicating how many were removed. See
	* nvlist_smush() for details. If smush_outnvlist is false, and the
	* outnvlist does not fit into the userland-provided buffer, then the
	* ioctl will fail with ENOMEM.
	*
	* zfs_ioc_func_t *func
	* The callback function that will perform the operation.
	*
	* The callback should return 0 on success, or an error number on
	* failure. If the function fails, the userland ioctl will return -1,
	* and errno will be set to the callback's return value. The callback
	* will be called with the following arguments:
	*
	* const char *name
	* The name of the pool or dataset to operate on, from
	* zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
	* expected type (pool, dataset, or none).
	*
	* nvlist_t *innvl
	* The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
	* NULL if no input nvlist was provided. Changes to this nvlist are
	* ignored. If the input nvlist could not be deserialized, the
	* ioctl will fail and the callback will not be called.
	*
	* nvlist_t *outnvl
	* The output nvlist, initially empty. The callback can fill it in,
	* and it will be returned to userland by serializing it into
	* zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
	* fails (e.g. because the caller didn't supply a large enough
	* buffer), then the overall ioctl will fail. See the
	* 'smush_nvlist' argument above for additional behaviors.
	*
	* There are two typical uses of the output nvlist:
	* - To return state, e.g. property values. In this case,
	* smush_outnvlist should be false. If the buffer was not large
	* enough, the caller will reallocate a larger buffer and try
	* the ioctl again.
	*
	* - To return multiple errors from an ioctl which makes on-disk
	* changes. In this case, smush_outnvlist should be true.
	* Ioctls which make on-disk modifications should generally not
	* use the outnvl if they succeed, because the caller can not
	* distinguish between the operation failing, and
	* deserialization failing.
	*
	* IOCTL Interface Errors
	*
	* The following ioctl input errors can be returned:
	* ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel
	* ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
	* ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
	* ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/errno.h>
	#include <sys/uio_impl.h>
	#include <sys/file.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/stat.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_znode.h>
	#include <sys/zap.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/dmu.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_deleg.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_redact.h>
	#include <sys/dmu_tx.h>
	#include <sys/sunddi.h>
	#include <sys/policy.h>
	#include <sys/zone.h>
	#include <sys/nvpair.h>
	#include <sys/pathname.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_onexit.h>
	#include <sys/zvol.h>
	#include <sys/dsl_scan.h>
	#include <sys/fm/util.h>
	#include <sys/dsl_crypt.h>
	#include <sys/rrwlock.h>
	#include <sys/zfs_file.h>

	#include <sys/dmu_recv.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/dsl_userhold.h>
	#include <sys/zfeature.h>
	#include <sys/zcp.h>
	#include <sys/zio_checksum.h>
	#include <sys/vdev_removal.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "zfs_deleg.h"
	#include "zfs_comutil.h"

	#include <sys/lua/lua.h>
	#include <sys/lua/lauxlib.h>
	#include <sys/zfs_ioctl_impl.h>

	kmutex_t zfsdev_state_lock;
	static zfsdev_state_t *zfsdev_state_list;

	/*
	* Limit maximum nvlist size. We don't want users passing in insane values
	* for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
	* Defaults to 0=auto which is handled by platform code.
	*/
	unsigned long zfs_max_nvlist_src_size = 0;

	/*
	* When logging the output nvlist of an ioctl in the on-disk history, limit
	* the logged size to this many bytes. This must be less than DMU_MAX_ACCESS.
	* This applies primarily to zfs_ioc_channel_program().
	*/
	static unsigned long zfs_history_output_max = 1024 * 1024;

	uint_t zfs_fsyncer_key;
	uint_t zfs_allow_log_key;

	/* DATA_TYPE_ANY is used when zkey_type can vary. */
	#define DATA_TYPE_ANY DATA_TYPE_UNKNOWN

	typedef struct zfs_ioc_vec {
	zfs_ioc_legacy_func_t *zvec_legacy_func;
	zfs_ioc_func_t *zvec_func;
	zfs_secpolicy_func_t *zvec_secpolicy;
	zfs_ioc_namecheck_t zvec_namecheck;
	boolean_t zvec_allow_log;
	zfs_ioc_poolcheck_t zvec_pool_check;
	boolean_t zvec_smush_outnvlist;
	const char *zvec_name;
	const zfs_ioc_key_t *zvec_nvl_keys;
	size_t zvec_nvl_key_count;
	} zfs_ioc_vec_t;

	/* This array is indexed by zfs_userquota_prop_t */
	static const char *userquota_perms[] = {
	ZFS_DELEG_PERM_USERUSED,
	ZFS_DELEG_PERM_USERQUOTA,
	ZFS_DELEG_PERM_GROUPUSED,
	ZFS_DELEG_PERM_GROUPQUOTA,
	ZFS_DELEG_PERM_USEROBJUSED,
	ZFS_DELEG_PERM_USEROBJQUOTA,
	ZFS_DELEG_PERM_GROUPOBJUSED,
	ZFS_DELEG_PERM_GROUPOBJQUOTA,
	ZFS_DELEG_PERM_PROJECTUSED,
	ZFS_DELEG_PERM_PROJECTQUOTA,
	ZFS_DELEG_PERM_PROJECTOBJUSED,
	ZFS_DELEG_PERM_PROJECTOBJQUOTA,
	};

	static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
	static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
	static int zfs_check_settable(const char name, nvpair_t property,
	cred_t *cr);
	static int zfs_check_clearable(const char dataset, nvlist_t props,
	nvlist_t **errors);
	static int zfs_fill_zplprops_root(uint64_t, nvlist_t , nvlist_t ,
	boolean_t *);
	int zfs_set_prop_nvlist(const char , zprop_source_t, nvlist_t , nvlist_t *);
	static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);

	static void
	history_str_free(char *buf)
	{
	kmem_free(buf, HIS_MAX_RECORD_LEN);
	}

	static char *
	history_str_get(zfs_cmd_t *zc)
	{
	char *buf;

	if (zc->zc_history == 0)
	return (NULL);

	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
	if (copyinstr((void *)(uintptr_t)zc->zc_history,
	buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
	history_str_free(buf);
	return (NULL);
	}

	buf[HIS_MAX_RECORD_LEN -1] = '\0';

	return (buf);
	}

	/*
	* Return non-zero if the spa version is less than requested version.
	*/
	static int
	zfs_earlier_version(const char *name, int version)
	{
	spa_t *spa;

	if (spa_open(name, &spa, FTAG) == 0) {
	if (spa_version(spa) < version) {
	spa_close(spa, FTAG);
	return (1);
	}
	spa_close(spa, FTAG);
	}
	return (0);
	}

	/*
	* Return TRUE if the ZPL version is less than requested version.
	*/
	static boolean_t
	zpl_earlier_version(const char *name, int version)
	{
	objset_t *os;
	boolean_t rc = B_TRUE;

	if (dmu_objset_hold(name, FTAG, &os) == 0) {
	uint64_t zplversion;

	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele(os, FTAG);
	return (B_TRUE);
	}
	/* XXX reading from non-owned objset */
	if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
	rc = zplversion < version;
	dmu_objset_rele(os, FTAG);
	}
	return (rc);
	}

	static void
	zfs_log_history(zfs_cmd_t *zc)
	{
	spa_t *spa;
	char *buf;

	if ((buf = history_str_get(zc)) == NULL)
	return;

	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
	if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
	(void) spa_history_log(spa, buf);
	spa_close(spa, FTAG);
	}
	history_str_free(buf);
	}

	/*
	* Policy for top-level read operations (list pools). Requires no privileges,
	* and can be used in the local zone, as there is no associated dataset.
	*/
	static int
	zfs_secpolicy_none(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (0);
	}

	/*
	* Policy for dataset read operations (list children, get statistics). Requires
	* no privileges, but must be visible in the local zone.
	*/
	static int
	zfs_secpolicy_read(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl, (void) cr;
	if (INGLOBALZONE(curproc) \|\|
	zone_dataset_visible(zc->zc_name, NULL))
	return (0);

	return (SET_ERROR(ENOENT));
	}

	static int
	zfs_dozonecheck_impl(const char dataset, uint64_t zoned, cred_t cr)
	{
	int writable = 1;

	/*
	* The dataset must be visible by this zone -- check this first
	* so they don't see EPERM on something they shouldn't know about.
	*/
	if (!INGLOBALZONE(curproc) &&
	!zone_dataset_visible(dataset, &writable))
	return (SET_ERROR(ENOENT));

	if (INGLOBALZONE(curproc)) {
	/*
	* If the fs is zoned, only root can access it from the
	* global zone.
	*/
	if (secpolicy_zfs(cr) && zoned)
	return (SET_ERROR(EPERM));
	} else {
	/*
	* If we are in a local zone, the 'zoned' property must be set.
	*/
	if (!zoned)
	return (SET_ERROR(EPERM));

	/* must be writable by this zone */
	if (!writable)
	return (SET_ERROR(EPERM));
	}
	return (0);
	}

	static int
	zfs_dozonecheck(const char dataset, cred_t cr)
	{
	uint64_t zoned;

	if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
	&zoned, NULL))
	return (SET_ERROR(ENOENT));

	return (zfs_dozonecheck_impl(dataset, zoned, cr));
	}

	static int
	zfs_dozonecheck_ds(const char dataset, dsl_dataset_t ds, cred_t *cr)
	{
	uint64_t zoned;

	if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
	return (SET_ERROR(ENOENT));

	return (zfs_dozonecheck_impl(dataset, zoned, cr));
	}

	static int
	zfs_secpolicy_write_perms_ds(const char name, dsl_dataset_t ds,
	const char perm, cred_t cr)
	{
	int error;

	error = zfs_dozonecheck_ds(name, ds, cr);
	if (error == 0) {
	error = secpolicy_zfs(cr);
	if (error != 0)
	error = dsl_deleg_access_impl(ds, perm, cr);
	}
	return (error);
	}

	static int
	zfs_secpolicy_write_perms(const char name, const char perm, cred_t *cr)
	{
	int error;
	dsl_dataset_t *ds;
	dsl_pool_t *dp;

	/*
	* First do a quick check for root in the global zone, which
	* is allowed to do all write_perms. This ensures that zfs_ioc_*
	* will get to handle nonexistent datasets.
	*/
	if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
	return (0);

	error = dsl_pool_hold(name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);

	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	/*
	* Policy for setting the security label property.
	*
	* Returns 0 for success, non-zero for access and other errors.
	*/
	static int
	zfs_set_slabel_policy(const char name, const char strval, cred_t *cr)
	{
	#ifdef HAVE_MLSLABEL
	char ds_hexsl[MAXNAMELEN];
	bslabel_t ds_sl, new_sl;
	boolean_t new_default = FALSE;
	uint64_t zoned;
	int needed_priv = -1;
	int error;

	/* First get the existing dataset label. */
	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
	1, sizeof (ds_hexsl), &ds_hexsl, NULL);
	if (error != 0)
	return (SET_ERROR(EPERM));

	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
	new_default = TRUE;

	/* The label must be translatable */
	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
	return (SET_ERROR(EINVAL));

	/*
	* In a non-global zone, disallow attempts to set a label that
	* doesn't match that of the zone; otherwise no other checks
	* are needed.
	*/
	if (!INGLOBALZONE(curproc)) {
	if (new_default \|\| !blequal(&new_sl, CR_SL(CRED())))
	return (SET_ERROR(EPERM));
	return (0);
	}

	/*
	* For global-zone datasets (i.e., those whose zoned property is
	* "off", verify that the specified new label is valid for the
	* global zone.
	*/
	if (dsl_prop_get_integer(name,
	zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
	return (SET_ERROR(EPERM));
	if (!zoned) {
	if (zfs_check_global_label(name, strval) != 0)
	return (SET_ERROR(EPERM));
	}

	/*
	* If the existing dataset label is nondefault, check if the
	* dataset is mounted (label cannot be changed while mounted).
	* Get the zfsvfs_t; if there isn't one, then the dataset isn't
	* mounted (or isn't a dataset, doesn't exist, ...).
	*/
	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
	objset_t *os;
	static const char *setsl_tag = "setsl_tag";

	/*
	* Try to own the dataset; abort if there is any error,
	* (e.g., already mounted, in use, or other error).
	*/
	error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
	setsl_tag, &os);
	if (error != 0)
	return (SET_ERROR(EPERM));

	dmu_objset_disown(os, B_TRUE, setsl_tag);

	if (new_default) {
	needed_priv = PRIV_FILE_DOWNGRADE_SL;
	goto out_check;
	}

	if (hexstr_to_label(strval, &new_sl) != 0)
	return (SET_ERROR(EPERM));

	if (blstrictdom(&ds_sl, &new_sl))
	needed_priv = PRIV_FILE_DOWNGRADE_SL;
	else if (blstrictdom(&new_sl, &ds_sl))
	needed_priv = PRIV_FILE_UPGRADE_SL;
	} else {
	/* dataset currently has a default label */
	if (!new_default)
	needed_priv = PRIV_FILE_UPGRADE_SL;
	}

	out_check:
	if (needed_priv != -1)
	return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
	return (0);
	#else
	return (SET_ERROR(ENOTSUP));
	#endif /* HAVE_MLSLABEL */
	}

	static int
	zfs_secpolicy_setprop(const char dsname, zfs_prop_t prop, nvpair_t propval,
	cred_t *cr)
	{
	char *strval;

	/*
	* Check permissions for special properties.
	*/
	switch (prop) {
	default:
	break;
	case ZFS_PROP_ZONED:
	/*
	* Disallow setting of 'zoned' from within a local zone.
	*/
	if (!INGLOBALZONE(curproc))
	return (SET_ERROR(EPERM));
	break;

	case ZFS_PROP_QUOTA:
	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	if (!INGLOBALZONE(curproc)) {
	uint64_t zoned;
	char setpoint[ZFS_MAX_DATASET_NAME_LEN];
	/*
	* Unprivileged users are allowed to modify the
	* limit on things under (ie. contained by)
	* the thing they own.
	*/
	if (dsl_prop_get_integer(dsname,
	zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
	return (SET_ERROR(EPERM));
	if (!zoned \|\| strlen(dsname) <= strlen(setpoint))
	return (SET_ERROR(EPERM));
	}
	break;

	case ZFS_PROP_MLSLABEL:
	if (!is_system_labeled())
	return (SET_ERROR(EPERM));

	if (nvpair_value_string(propval, &strval) == 0) {
	int err;

	err = zfs_set_slabel_policy(dsname, strval, CRED());
	if (err != 0)
	return (err);
	}
	break;
	}

	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
	}

	static int
	zfs_secpolicy_set_fsacl(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	/*
	* permission to set permissions will be evaluated later in
	* dsl_deleg_can_allow()
	*/
	(void) innvl;
	return (zfs_dozonecheck(zc->zc_name, cr));
	}

	static int
	zfs_secpolicy_rollback(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_ROLLBACK, cr));
	}

	static int
	zfs_secpolicy_send(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	const char *cp;
	int error;

	/*
	* Generate the current snapshot name from the given objsetid, then
	* use that name for the secpolicy/zone checks.
	*/
	cp = strchr(zc->zc_name, '@');
	if (cp == NULL)
	return (SET_ERROR(EINVAL));
	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dsl_dataset_name(ds, zc->zc_name);

	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
	ZFS_DELEG_PERM_SEND, cr);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	return (error);
	}

	static int
	zfs_secpolicy_send_new(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_SEND, cr));
	}

	static int
	zfs_secpolicy_share(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (SET_ERROR(ENOTSUP));
	}

	static int
	zfs_secpolicy_smb_acl(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (SET_ERROR(ENOTSUP));
	}

	static int
	zfs_get_parent(const char datasetname, char parent, int parentsize)
	{
	char *cp;

	/*
	* Remove the @bla or /bla from the end of the name to get the parent.
	*/
	- (void) strncpy(parent, datasetname, parentsize);
	+ (void) strlcpy(parent, datasetname, parentsize);
	cp = strrchr(parent, '@');
	if (cp != NULL) {
	cp[0] = '\0';
	} else {
	cp = strrchr(parent, '/');
	if (cp == NULL)
	return (SET_ERROR(ENOENT));
	cp[0] = '\0';
	}

	return (0);
	}

	int
	zfs_secpolicy_destroy_perms(const char name, cred_t cr)
	{
	int error;

	if ((error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
	}

	static int
	zfs_secpolicy_destroy(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
	}

	/*
	* Destroying snapshots with delegated permissions requires
	* descendant mount and destroy permissions.
	*/
	static int
	zfs_secpolicy_destroy_snaps(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvlist_t *snaps;
	nvpair_t pair, nextpair;
	int error = 0;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");

	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nextpair) {
	nextpair = nvlist_next_nvpair(snaps, pair);
	error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
	if (error == ENOENT) {
	/*
	* Ignore any snapshots that don't exist (we consider
	* them "already destroyed"). Remove the name from the
	* nvl here in case the snapshot is created between
	* now and when we try to destroy it (in which case
	* we don't want to destroy it since we haven't
	* checked for permission).
	*/
	fnvlist_remove_nvpair(snaps, pair);
	error = 0;
	}
	if (error != 0)
	break;
	}

	return (error);
	}

	int
	zfs_secpolicy_rename_perms(const char from, const char to, cred_t *cr)
	{
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	if ((error = zfs_secpolicy_write_perms(from,
	ZFS_DELEG_PERM_RENAME, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(from,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	if ((error = zfs_get_parent(to, parentname,
	sizeof (parentname))) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_CREATE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (error);
	}

	static int
	zfs_secpolicy_rename(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
	}

	static int
	zfs_secpolicy_promote(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	dsl_pool_t *dp;
	dsl_dataset_t *clone;
	int error;

	error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_PROMOTE, cr);
	if (error != 0)
	return (error);

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);

	if (error == 0) {
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_t *origin = NULL;
	dsl_dir_t *dd;
	dd = clone->ds_dir;

	error = dsl_dataset_hold_obj(dd->dd_pool,
	dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
	if (error != 0) {
	dsl_dataset_rele(clone, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
	ZFS_DELEG_PERM_MOUNT, cr);

	dsl_dataset_name(origin, parentname);
	if (error == 0) {
	error = zfs_secpolicy_write_perms_ds(parentname, origin,
	ZFS_DELEG_PERM_PROMOTE, cr);
	}
	dsl_dataset_rele(clone, FTAG);
	dsl_dataset_rele(origin, FTAG);
	}
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	static int
	zfs_secpolicy_recv(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	int error;

	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_CREATE, cr));
	}

	int
	zfs_secpolicy_snapshot_perms(const char name, cred_t cr)
	{
	return (zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_SNAPSHOT, cr));
	}

	/*
	* Check for permission to create each snapshot in the nvlist.
	*/
	static int
	zfs_secpolicy_snapshot(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvlist_t *snaps;
	int error = 0;
	nvpair_t *pair;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");

	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	char *name = nvpair_name(pair);
	char *atp = strchr(name, '@');

	if (atp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}
	*atp = '\0';
	error = zfs_secpolicy_snapshot_perms(name, cr);
	*atp = '@';
	if (error != 0)
	break;
	}
	return (error);
	}

	/*
	* Check for permission to create each bookmark in the nvlist.
	*/
	static int
	zfs_secpolicy_bookmark(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	int error = 0;

	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	char *name = nvpair_name(pair);
	char *hashp = strchr(name, '#');

	if (hashp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}
	*hashp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_BOOKMARK, cr);
	*hashp = '#';
	if (error != 0)
	break;
	}
	return (error);
	}

	static int
	zfs_secpolicy_destroy_bookmarks(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t pair, nextpair;
	int error = 0;

	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
	pair = nextpair) {
	char *name = nvpair_name(pair);
	char *hashp = strchr(name, '#');
	nextpair = nvlist_next_nvpair(innvl, pair);

	if (hashp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}

	*hashp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_DESTROY, cr);
	*hashp = '#';
	if (error == ENOENT) {
	/*
	* Ignore any filesystems that don't exist (we consider
	* their bookmarks "already destroyed"). Remove
	* the name from the nvl here in case the filesystem
	* is created between now and when we try to destroy
	* the bookmark (in which case we don't want to
	* destroy it since we haven't checked for permission).
	*/
	fnvlist_remove_nvpair(innvl, pair);
	error = 0;
	}
	if (error != 0)
	break;
	}

	return (error);
	}

	static int
	zfs_secpolicy_log_history(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	/*
	* Even root must have a proper TSD so that we know what pool
	* to log to.
	*/
	if (tsd_get(zfs_allow_log_key) == NULL)
	return (SET_ERROR(EPERM));
	return (0);
	}

	static int
	zfs_secpolicy_create_clone(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	int error;
	char *origin;

	if ((error = zfs_get_parent(zc->zc_name, parentname,
	sizeof (parentname))) != 0)
	return (error);

	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
	(error = zfs_secpolicy_write_perms(origin,
	ZFS_DELEG_PERM_CLONE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_CREATE, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_MOUNT, cr));
	}

	/*
	* Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
	* SYS_CONFIG privilege, which is not available in a local zone.
	*/
	int
	zfs_secpolicy_config(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl;

	if (secpolicy_sys_config(cr, B_FALSE) != 0)
	return (SET_ERROR(EPERM));

	return (0);
	}

	/*
	* Policy for object to name lookups.
	*/
	static int
	zfs_secpolicy_diff(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	int error;

	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
	return (0);

	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
	return (error);
	}

	/*
	* Policy for fault injection. Requires all privileges.
	*/
	static int
	zfs_secpolicy_inject(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl;
	return (secpolicy_zinject(cr));
	}

	static int
	zfs_secpolicy_inherit_prop(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(zc->zc_value))
	return (SET_ERROR(EINVAL));
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_USERPROP, cr));
	} else {
	return (zfs_secpolicy_setprop(zc->zc_name, prop,
	NULL, cr));
	}
	}

	static int
	zfs_secpolicy_userspace_one(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	int err = zfs_secpolicy_read(zc, innvl, cr);
	if (err)
	return (err);

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	if (zc->zc_value[0] == 0) {
	/*
	* They are asking about a posix uid/gid. If it's
	* themself, allow it.
	*/
	if (zc->zc_objset_type == ZFS_PROP_USERUSED \|\|
	zc->zc_objset_type == ZFS_PROP_USERQUOTA \|\|
	zc->zc_objset_type == ZFS_PROP_USEROBJUSED \|\|
	zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
	if (zc->zc_guid == crgetuid(cr))
	return (0);
	} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPQUOTA \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
	if (groupmember(zc->zc_guid, cr))
	return (0);
	}
	/* else is for project quota/used */
	}

	return (zfs_secpolicy_write_perms(zc->zc_name,
	userquota_perms[zc->zc_objset_type], cr));
	}

	static int
	zfs_secpolicy_userspace_many(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	int err = zfs_secpolicy_read(zc, innvl, cr);
	if (err)
	return (err);

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	return (zfs_secpolicy_write_perms(zc->zc_name,
	userquota_perms[zc->zc_objset_type], cr));
	}

	static int
	zfs_secpolicy_userspace_upgrade(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
	NULL, cr));
	}

	static int
	zfs_secpolicy_hold(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t *pair;
	nvlist_t *holds;
	int error;

	holds = fnvlist_lookup_nvlist(innvl, "holds");

	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
	pair = nvlist_next_nvpair(holds, pair)) {
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	error = dmu_fsname(nvpair_name(pair), fsname);
	if (error != 0)
	return (error);
	error = zfs_secpolicy_write_perms(fsname,
	ZFS_DELEG_PERM_HOLD, cr);
	if (error != 0)
	return (error);
	}
	return (0);
	}

	static int
	zfs_secpolicy_release(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t *pair;
	int error;

	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
	pair = nvlist_next_nvpair(innvl, pair)) {
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	error = dmu_fsname(nvpair_name(pair), fsname);
	if (error != 0)
	return (error);
	error = zfs_secpolicy_write_perms(fsname,
	ZFS_DELEG_PERM_RELEASE, cr);
	if (error != 0)
	return (error);
	}
	return (0);
	}

	/*
	* Policy for allowing temporary snapshots to be taken or released
	*/
	static int
	zfs_secpolicy_tmp_snapshot(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	/*
	* A temporary snapshot is the same as a snapshot,
	* hold, destroy and release all rolled into one.
	* Delegated diff alone is sufficient that we allow this.
	*/
	int error;

	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_DIFF, cr)) == 0)
	return (0);

	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);

	if (innvl != NULL) {
	if (error == 0)
	error = zfs_secpolicy_hold(zc, innvl, cr);
	if (error == 0)
	error = zfs_secpolicy_release(zc, innvl, cr);
	if (error == 0)
	error = zfs_secpolicy_destroy(zc, innvl, cr);
	}
	return (error);
	}

	static int
	zfs_secpolicy_load_key(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_LOAD_KEY, cr));
	}

	static int
	zfs_secpolicy_change_key(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_CHANGE_KEY, cr));
	}

	/*
	* Returns the nvlist as specified by the user in the zfs_cmd_t.
	*/
	static int
	get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
	{
	char *packed;
	int error;
	nvlist_t *list = NULL;

	/*
	* Read in and unpack the user-supplied nvlist.
	*/
	if (size == 0)
	return (SET_ERROR(EINVAL));

	packed = vmem_alloc(size, KM_SLEEP);

	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
	iflag)) != 0) {
	vmem_free(packed, size);
	return (SET_ERROR(EFAULT));
	}

	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
	vmem_free(packed, size);
	return (error);
	}

	vmem_free(packed, size);

	*nvp = list;
	return (0);
	}

	/*
	* Reduce the size of this nvlist until it can be serialized in 'max' bytes.
	* Entries will be removed from the end of the nvlist, and one int32 entry
	* named "N_MORE_ERRORS" will be added indicating how many entries were
	* removed.
	*/
	static int
	nvlist_smush(nvlist_t *errors, size_t max)
	{
	size_t size;

	size = fnvlist_size(errors);

	if (size > max) {
	nvpair_t *more_errors;
	int n = 0;

	if (max < 1024)
	return (SET_ERROR(ENOMEM));

	fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
	more_errors = nvlist_prev_nvpair(errors, NULL);

	do {
	nvpair_t *pair = nvlist_prev_nvpair(errors,
	more_errors);
	fnvlist_remove_nvpair(errors, pair);
	n++;
	size = fnvlist_size(errors);
	} while (size > max);

	fnvlist_remove_nvpair(errors, more_errors);
	fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
	ASSERT3U(fnvlist_size(errors), <=, max);
	}

	return (0);
	}

	static int
	put_nvlist(zfs_cmd_t zc, nvlist_t nvl)
	{
	char *packed = NULL;
	int error = 0;
	size_t size;

	size = fnvlist_size(nvl);

	if (size > zc->zc_nvlist_dst_size) {
	error = SET_ERROR(ENOMEM);
	} else {
	packed = fnvlist_pack(nvl, &size);
	if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
	size, zc->zc_iflags) != 0)
	error = SET_ERROR(EFAULT);
	fnvlist_pack_free(packed, size);
	}

	zc->zc_nvlist_dst_size = size;
	zc->zc_nvlist_dst_filled = B_TRUE;
	return (error);
	}

	int
	getzfsvfs_impl(objset_t os, zfsvfs_t *zfvp)
	{
	int error = 0;
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	return (SET_ERROR(EINVAL));
	}

	mutex_enter(&os->os_user_ptr_lock);
	*zfvp = dmu_objset_get_user(os);
	/* bump s_active only when non-zero to prevent umount race */
	error = zfs_vfs_ref(zfvp);
	mutex_exit(&os->os_user_ptr_lock);
	return (error);
	}

	int
	getzfsvfs(const char dsname, zfsvfs_t *zfvp)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold(dsname, FTAG, &os);
	if (error != 0)
	return (error);

	error = getzfsvfs_impl(os, zfvp);
	dmu_objset_rele(os, FTAG);
	return (error);
	}

	/*
	* Find a zfsvfs_t for a mounted filesystem, or create our own, in which
	* case its z_sb will be NULL, and it will be opened as the owner.
	* If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
	* which prevents all inode ops from running.
	*/
	static int
	zfsvfs_hold(const char name, const void tag, zfsvfs_t **zfvp,
	boolean_t writer)
	{
	int error = 0;

	if (getzfsvfs(name, zfvp) != 0)
	error = zfsvfs_create(name, B_FALSE, zfvp);
	if (error == 0) {
	if (writer)
	ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
	else
	ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
	if ((*zfvp)->z_unmounted) {
	/*
	* XXX we could probably try again, since the unmounting
	* thread should be just about to disassociate the
	* objset from the zfsvfs.
	*/
	ZFS_TEARDOWN_EXIT(*zfvp, tag);
	return (SET_ERROR(EBUSY));
	}
	}
	return (error);
	}

	static void
	zfsvfs_rele(zfsvfs_t zfsvfs, const void tag)
	{
	ZFS_TEARDOWN_EXIT(zfsvfs, tag);

	if (zfs_vfs_held(zfsvfs)) {
	zfs_vfs_rele(zfsvfs);
	} else {
	dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
	zfsvfs_free(zfsvfs);
	}
	}

	static int
	zfs_ioc_pool_create(zfs_cmd_t *zc)
	{
	int error;
	nvlist_t config, props = NULL;
	nvlist_t *rootprops = NULL;
	nvlist_t *zplprops = NULL;
	dsl_crypto_params_t *dcp = NULL;
	const char *spa_name = zc->zc_name;
	boolean_t unload_wkey = B_TRUE;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)))
	return (error);

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	nvlist_free(config);
	return (error);
	}

	if (props) {
	nvlist_t *nvl = NULL;
	nvlist_t *hidden_args = NULL;
	uint64_t version = SPA_VERSION;
	char *tname;

	(void) nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
	if (!SPA_VERSION_IS_SUPPORTED(version)) {
	error = SET_ERROR(EINVAL);
	goto pool_props_bad;
	}
	(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
	if (nvl) {
	error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
	if (error != 0)
	goto pool_props_bad;
	(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
	}

	(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
	&hidden_args);
	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
	rootprops, hidden_args, &dcp);
	if (error != 0)
	goto pool_props_bad;
	(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);

	VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	error = zfs_fill_zplprops_root(version, rootprops,
	zplprops, NULL);
	if (error != 0)
	goto pool_props_bad;

	if (nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
	spa_name = tname;
	}

	error = spa_create(zc->zc_name, config, props, zplprops, dcp);

	/*
	* Set the remaining root properties
	*/
	if (!error && (error = zfs_set_prop_nvlist(spa_name,
	ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
	(void) spa_destroy(spa_name);
	unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
	}

	pool_props_bad:
	nvlist_free(rootprops);
	nvlist_free(zplprops);
	nvlist_free(config);
	nvlist_free(props);
	dsl_crypto_params_free(dcp, unload_wkey && !!error);

	return (error);
	}

	static int
	zfs_ioc_pool_destroy(zfs_cmd_t *zc)
	{
	int error;
	zfs_log_history(zc);
	error = spa_destroy(zc->zc_name);

	return (error);
	}

	static int
	zfs_ioc_pool_import(zfs_cmd_t *zc)
	{
	nvlist_t config, props = NULL;
	uint64_t guid;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)) != 0)
	return (error);

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	nvlist_free(config);
	return (error);
	}

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 \|\|
	guid != zc->zc_guid)
	error = SET_ERROR(EINVAL);
	else
	error = spa_import(zc->zc_name, config, props, zc->zc_cookie);

	if (zc->zc_nvlist_dst != 0) {
	int err;

	if ((err = put_nvlist(zc, config)) != 0)
	error = err;
	}

	nvlist_free(config);
	nvlist_free(props);

	return (error);
	}

	static int
	zfs_ioc_pool_export(zfs_cmd_t *zc)
	{
	int error;
	boolean_t force = (boolean_t)zc->zc_cookie;
	boolean_t hardforce = (boolean_t)zc->zc_guid;

	zfs_log_history(zc);
	error = spa_export(zc->zc_name, NULL, force, hardforce);

	return (error);
	}

	static int
	zfs_ioc_pool_configs(zfs_cmd_t *zc)
	{
	nvlist_t *configs;
	int error;

	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
	return (SET_ERROR(EEXIST));

	error = put_nvlist(zc, configs);

	nvlist_free(configs);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	*
	* outputs:
	* zc_cookie real errno
	* zc_nvlist_dst config nvlist
	* zc_nvlist_dst_size size of config nvlist
	*/
	static int
	zfs_ioc_pool_stats(zfs_cmd_t *zc)
	{
	nvlist_t *config;
	int error;
	int ret = 0;

	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
	sizeof (zc->zc_value));

	if (config != NULL) {
	ret = put_nvlist(zc, config);
	nvlist_free(config);

	/*
	* The config may be present even if 'error' is non-zero.
	* In this case we return success, and preserve the real errno
	* in 'zc_cookie'.
	*/
	zc->zc_cookie = error;
	} else {
	ret = error;
	}

	return (ret);
	}

	/*
	* Try to import the given pool, returning pool stats as appropriate so that
	* user land knows which devices are available and overall pool health.
	*/
	static int
	zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
	{
	nvlist_t tryconfig, config = NULL;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &tryconfig)) != 0)
	return (error);

	config = spa_tryimport(tryconfig);

	nvlist_free(tryconfig);

	if (config == NULL)
	return (SET_ERROR(EINVAL));

	error = put_nvlist(zc, config);
	nvlist_free(config);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	* zc_cookie scan func (pool_scan_func_t)
	* zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
	*/
	static int
	zfs_ioc_pool_scan(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (zc->zc_flags == POOL_SCRUB_PAUSE)
	error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
	else if (zc->zc_cookie == POOL_SCAN_NONE)
	error = spa_scan_stop(spa);
	else
	error = spa_scan(spa, zc->zc_cookie);

	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_pool_freeze(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error == 0) {
	spa_freeze(spa);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	static int
	zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (zc->zc_cookie < spa_version(spa) \|\|
	!SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(EINVAL));
	}

	spa_upgrade(spa, zc->zc_cookie);
	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_pool_get_history(zfs_cmd_t *zc)
	{
	spa_t *spa;
	char *hist_buf;
	uint64_t size;
	int error;

	if ((size = zc->zc_history_len) == 0)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	hist_buf = vmem_alloc(size, KM_SLEEP);
	if ((error = spa_history_get(spa, &zc->zc_history_offset,
	&zc->zc_history_len, hist_buf)) == 0) {
	error = ddi_copyout(hist_buf,
	(void *)(uintptr_t)zc->zc_history,
	zc->zc_history_len, zc->zc_iflags);
	}

	spa_close(spa, FTAG);
	vmem_free(hist_buf, size);
	return (error);
	}

	static int
	zfs_ioc_pool_reguid(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error == 0) {
	error = spa_change_guid(spa);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	static int
	zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
	{
	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_obj object to find
	*
	* outputs:
	* zc_value name of object
	*/
	static int
	zfs_ioc_obj_to_path(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	/* XXX reading from objset not owned */
	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
	FTAG, &os)) != 0)
	return (error);
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele_flags(os, B_TRUE, FTAG);
	return (SET_ERROR(EINVAL));
	}
	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
	sizeof (zc->zc_value));
	dmu_objset_rele_flags(os, B_TRUE, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_obj object to find
	*
	* outputs:
	* zc_stat stats on object
	* zc_value path to object
	*/
	static int
	zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	/* XXX reading from objset not owned */
	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
	FTAG, &os)) != 0)
	return (error);
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele_flags(os, B_TRUE, FTAG);
	return (SET_ERROR(EINVAL));
	}
	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
	sizeof (zc->zc_value));
	dmu_objset_rele_flags(os, B_TRUE, FTAG);

	return (error);
	}

	static int
	zfs_ioc_vdev_add(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	nvlist_t *config;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config);
	if (error == 0) {
	error = spa_vdev_add(spa, config);
	nvlist_free(config);
	}
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	* zc_guid guid of vdev to remove
	* zc_cookie cancel removal
	*/
	static int
	zfs_ioc_vdev_remove(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);
	if (zc->zc_cookie != 0) {
	error = spa_vdev_remove_cancel(spa);
	} else {
	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
	}
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	vdev_state_t newstate = VDEV_STATE_UNKNOWN;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);
	switch (zc->zc_cookie) {
	case VDEV_STATE_ONLINE:
	error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
	break;

	case VDEV_STATE_OFFLINE:
	error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
	break;

	case VDEV_STATE_FAULTED:
	if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
	zc->zc_obj != VDEV_AUX_EXTERNAL &&
	zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
	zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;

	error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
	break;

	case VDEV_STATE_DEGRADED:
	if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
	zc->zc_obj != VDEV_AUX_EXTERNAL)
	zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;

	error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
	break;

	+ case VDEV_STATE_REMOVED:
	+ error = vdev_remove_wanted(spa, zc->zc_guid);
	+ break;
	+
	default:
	error = SET_ERROR(EINVAL);
	}
	zc->zc_cookie = newstate;
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_attach(zfs_cmd_t *zc)
	{
	spa_t *spa;
	nvlist_t *config;
	int replacing = zc->zc_cookie;
	int rebuild = zc->zc_simple;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)) == 0) {
	error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
	rebuild);
	nvlist_free(config);
	}

	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_detach(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);

	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_split(zfs_cmd_t *zc)
	{
	spa_t *spa;
	nvlist_t config, props = NULL;
	int error;
	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config))) {
	spa_close(spa, FTAG);
	return (error);
	}

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	spa_close(spa, FTAG);
	nvlist_free(config);
	return (error);
	}

	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);

	spa_close(spa, FTAG);

	nvlist_free(config);
	nvlist_free(props);

	return (error);
	}

	static int
	zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
	{
	spa_t *spa;
	const char *path = zc->zc_value;
	uint64_t guid = zc->zc_guid;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = spa_vdev_setpath(spa, guid, path);
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
	{
	spa_t *spa;
	const char *fru = zc->zc_value;
	uint64_t guid = zc->zc_guid;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = spa_vdev_setfru(spa, guid, fru);
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_objset_stats_impl(zfs_cmd_t zc, objset_t os)
	{
	int error = 0;
	nvlist_t *nv;

	dmu_objset_fast_stat(os, &zc->zc_objset_stats);

	if (zc->zc_nvlist_dst != 0 &&
	(error = dsl_prop_get_all(os, &nv)) == 0) {
	dmu_objset_stats(os, nv);
	/*
	* NB: zvol_get_stats() will read the objset contents,
	* which we aren't supposed to do with a
	* DS_MODE_USER hold, because it could be
	* inconsistent. So this is a bit of a workaround...
	* XXX reading without owning
	*/
	if (!zc->zc_objset_stats.dds_inconsistent &&
	dmu_objset_type(os) == DMU_OST_ZVOL) {
	error = zvol_get_stats(os, nv);
	if (error == EIO) {
	nvlist_free(nv);
	return (error);
	}
	VERIFY0(error);
	}
	if (error == 0)
	error = put_nvlist(zc, nv);
	nvlist_free(nv);
	}

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_objset_stats(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error == 0) {
	error = zfs_ioc_objset_stats_impl(zc, os);
	dmu_objset_rele(os, FTAG);
	}

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_nvlist_dst received property nvlist
	* zc_nvlist_dst_size size of received property nvlist
	*
	* Gets received properties (distinct from local properties on or after
	* SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
	* local property values.
	*/
	static int
	zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
	{
	int error = 0;
	nvlist_t *nv;

	/*
	* Without this check, we would return local property values if the
	* caller has not already received properties on or after
	* SPA_VERSION_RECVD_PROPS.
	*/
	if (!dsl_prop_get_hasrecvd(zc->zc_name))
	return (SET_ERROR(ENOTSUP));

	if (zc->zc_nvlist_dst != 0 &&
	(error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
	error = put_nvlist(zc, nv);
	nvlist_free(nv);
	}

	return (error);
	}

	static int
	nvl_add_zplprop(objset_t os, nvlist_t props, zfs_prop_t prop)
	{
	uint64_t value;
	int error;

	/*
	* zfs_get_zplprop() will either find a value or give us
	* the default value (if there is one).
	*/
	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
	return (0);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for zpl property nvlist
	*
	* outputs:
	* zc_nvlist_dst zpl property nvlist
	* zc_nvlist_dst_size size of zpl property nvlist
	*/
	static int
	zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
	{
	objset_t *os;
	int err;

	/* XXX reading without owning */
	if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
	return (err);

	dmu_objset_fast_stat(os, &zc->zc_objset_stats);

	/*
	* NB: nvl_add_zplprop() will read the objset contents,
	* which we aren't supposed to do with a DS_MODE_USER
	* hold, because it could be inconsistent.
	*/
	if (zc->zc_nvlist_dst != 0 &&
	!zc->zc_objset_stats.dds_inconsistent &&
	dmu_objset_type(os) == DMU_OST_ZFS) {
	nvlist_t *nv;

	VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
	err = put_nvlist(zc, nv);
	nvlist_free(nv);
	} else {
	err = SET_ERROR(ENOENT);
	}
	dmu_objset_rele(os, FTAG);
	return (err);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_name name of next filesystem
	* zc_cookie zap cursor
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;
	char *p;
	size_t orig_len = strlen(zc->zc_name);

	top:
	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
	if (error == ENOENT)
	error = SET_ERROR(ESRCH);
	return (error);
	}

	p = strrchr(zc->zc_name, '/');
	if (p == NULL \|\| p[1] != '\0')
	(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
	p = zc->zc_name + strlen(zc->zc_name);

	do {
	error = dmu_dir_list_next(os,
	sizeof (zc->zc_name) - (p - zc->zc_name), p,
	NULL, &zc->zc_cookie);
	if (error == ENOENT)
	error = SET_ERROR(ESRCH);
	} while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
	dmu_objset_rele(os, FTAG);

	/*
	* If it's an internal dataset (ie. with a '$' in its name),
	* don't try to get stats for it, otherwise we'll return ENOENT.
	*/
	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
	error = zfs_ioc_objset_stats(zc); /* fill in the stats */
	if (error == ENOENT) {
	/* We lost a race with destroy, get the next one. */
	zc->zc_name[orig_len] = '\0';
	goto top;
	}
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_nvlist_src iteration range nvlist
	* zc_nvlist_src_size size of iteration range nvlist
	*
	* outputs:
	* zc_name name of next snapshot
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
	{
	int error;
	objset_t os, ossnap;
	dsl_dataset_t *ds;
	uint64_t min_txg = 0, max_txg = 0;

	if (zc->zc_nvlist_src_size != 0) {
	nvlist_t *props = NULL;
	error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props);
	if (error != 0)
	return (error);
	(void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
	&min_txg);
	(void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
	&max_txg);
	nvlist_free(props);
	}

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error != 0) {
	return (error == ENOENT ? SET_ERROR(ESRCH) : error);
	}

	/*
	* A dataset name of maximum length cannot have any snapshots,
	* so exit immediately.
	*/
	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
	ZFS_MAX_DATASET_NAME_LEN) {
	dmu_objset_rele(os, FTAG);
	return (SET_ERROR(ESRCH));
	}

	while (error == 0) {
	if (issig(JUSTLOOKING) && issig(FORREAL)) {
	error = SET_ERROR(EINTR);
	break;
	}

	error = dmu_snapshot_list_next(os,
	sizeof (zc->zc_name) - strlen(zc->zc_name),
	zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
	&zc->zc_cookie, NULL);
	if (error == ENOENT) {
	error = SET_ERROR(ESRCH);
	break;
	} else if (error != 0) {
	break;
	}

	error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
	FTAG, &ds);
	if (error != 0)
	break;

	if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) \|\|
	(max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
	dsl_dataset_rele(ds, FTAG);
	/* undo snapshot name append */
	*(strchr(zc->zc_name, '@') + 1) = '\0';
	/* skip snapshot */
	continue;
	}

	if (zc->zc_simple) {
	zc->zc_objset_stats.dds_creation_txg =
	dsl_get_creationtxg(ds);
	dsl_dataset_rele(ds, FTAG);
	break;
	}

	if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	break;
	}
	if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	break;
	}
	dsl_dataset_rele(ds, FTAG);
	break;
	}

	dmu_objset_rele(os, FTAG);
	/* if we failed, undo the @ that we tacked on to zc_name */
	if (error != 0)
	*strchr(zc->zc_name, '@') = '\0';
	return (error);
	}

	static int
	zfs_prop_set_userquota(const char dsname, nvpair_t pair)
	{
	const char *propname = nvpair_name(pair);
	uint64_t *valary;
	unsigned int vallen;
	const char dash, domain;
	zfs_userquota_prop_t type;
	uint64_t rid;
	uint64_t quota;
	zfsvfs_t *zfsvfs;
	int err;

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* A correctly constructed propname is encoded as
	* userquota@<rid>-<domain>.
	*/
	if ((dash = strchr(propname, '-')) == NULL \|\|
	nvpair_value_uint64_array(pair, &valary, &vallen) != 0 \|\|
	vallen != 3)
	return (SET_ERROR(EINVAL));

	domain = dash + 1;
	type = valary[0];
	rid = valary[1];
	quota = valary[2];

	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
	if (err == 0) {
	err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
	zfsvfs_rele(zfsvfs, FTAG);
	}

	return (err);
	}

	/*
	* If the named property is one that has a special function to set its value,
	* return 0 on success and a positive error code on failure; otherwise if it is
	* not one of the special properties handled by this function, return -1.
	*
	* XXX: It would be better for callers of the property interface if we handled
	* these special cases in dsl_prop.c (in the dsl layer).
	*/
	static int
	zfs_prop_set_special(const char *dsname, zprop_source_t source,
	nvpair_t *pair)
	{
	const char *propname = nvpair_name(pair);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	uint64_t intval = 0;
	const char *strval = NULL;
	int err = -1;

	if (prop == ZPROP_USERPROP) {
	if (zfs_prop_userquota(propname))
	return (zfs_prop_set_userquota(dsname, pair));
	return (-1);
	}

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) == 0);
	}

	/* all special properties are numeric except for keylocation */
	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
	strval = fnvpair_value_string(pair);
	} else {
	intval = fnvpair_value_uint64(pair);
	}

	switch (prop) {
	case ZFS_PROP_QUOTA:
	err = dsl_dir_set_quota(dsname, source, intval);
	break;
	case ZFS_PROP_REFQUOTA:
	err = dsl_dataset_set_refquota(dsname, source, intval);
	break;
	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	if (intval == UINT64_MAX) {
	/* clearing the limit, just do it */
	err = 0;
	} else {
	err = dsl_dir_activate_fs_ss_limit(dsname);
	}
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_KEYLOCATION:
	err = dsl_crypto_can_set_keylocation(dsname, strval);

	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_RESERVATION:
	err = dsl_dir_set_reservation(dsname, source, intval);
	break;
	case ZFS_PROP_REFRESERVATION:
	err = dsl_dataset_set_refreservation(dsname, source, intval);
	break;
	case ZFS_PROP_COMPRESSION:
	err = dsl_dataset_set_compression(dsname, source, intval);
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_VOLSIZE:
	err = zvol_set_volsize(dsname, intval);
	break;
	case ZFS_PROP_SNAPDEV:
	err = zvol_set_snapdev(dsname, source, intval);
	break;
	case ZFS_PROP_VOLMODE:
	err = zvol_set_volmode(dsname, source, intval);
	break;
	case ZFS_PROP_VERSION:
	{
	zfsvfs_t *zfsvfs;

	if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
	break;

	err = zfs_set_version(zfsvfs, intval);
	zfsvfs_rele(zfsvfs, FTAG);

	if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
	zfs_cmd_t *zc;

	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
	(void) strlcpy(zc->zc_name, dsname,
	sizeof (zc->zc_name));
	(void) zfs_ioc_userspace_upgrade(zc);
	(void) zfs_ioc_id_quota_upgrade(zc);
	kmem_free(zc, sizeof (zfs_cmd_t));
	}
	break;
	}
	default:
	err = -1;
	}

	return (err);
	}

	static boolean_t
	zfs_is_namespace_prop(zfs_prop_t prop)
	{
	switch (prop) {

	case ZFS_PROP_ATIME:
	case ZFS_PROP_RELATIME:
	case ZFS_PROP_DEVICES:
	case ZFS_PROP_EXEC:
	case ZFS_PROP_SETUID:
	case ZFS_PROP_READONLY:
	case ZFS_PROP_XATTR:
	case ZFS_PROP_NBMAND:
	return (B_TRUE);

	default:
	return (B_FALSE);
	}
	}

	/*
	* This function is best effort. If it fails to set any of the given properties,
	* it continues to set as many as it can and returns the last error
	* encountered. If the caller provides a non-NULL errlist, it will be filled in
	* with the list of names of all the properties that failed along with the
	* corresponding error numbers.
	*
	* If every property is set successfully, zero is returned and errlist is not
	* modified.
	*/
	int
	zfs_set_prop_nvlist(const char dsname, zprop_source_t source, nvlist_t nvl,
	nvlist_t *errlist)
	{
	nvpair_t *pair;
	nvpair_t *propval;
	int rv = 0;
	int err;
	uint64_t intval;
	const char *strval;
	boolean_t should_update_mount_cache = B_FALSE;

	nvlist_t *genericnvl = fnvlist_alloc();
	nvlist_t *retrynvl = fnvlist_alloc();
	retry:
	pair = NULL;
	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	err = 0;

	/* decode the property value */
	propval = pair;
	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	attrs = fnvpair_value_nvlist(pair);
	if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&propval) != 0)
	err = SET_ERROR(EINVAL);
	}

	/* Validate value type */
	if (err == 0 && source == ZPROP_SRC_INHERITED) {
	/* inherited properties are expected to be booleans */
	if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
	err = SET_ERROR(EINVAL);
	} else if (err == 0 && prop == ZPROP_USERPROP) {
	if (zfs_prop_user(propname)) {
	if (nvpair_type(propval) != DATA_TYPE_STRING)
	err = SET_ERROR(EINVAL);
	} else if (zfs_prop_userquota(propname)) {
	if (nvpair_type(propval) !=
	DATA_TYPE_UINT64_ARRAY)
	err = SET_ERROR(EINVAL);
	} else {
	err = SET_ERROR(EINVAL);
	}
	} else if (err == 0) {
	if (nvpair_type(propval) == DATA_TYPE_STRING) {
	if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
	err = SET_ERROR(EINVAL);
	} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
	const char *unused;

	intval = fnvpair_value_uint64(propval);

	switch (zfs_prop_get_type(prop)) {
	case PROP_TYPE_NUMBER:
	break;
	case PROP_TYPE_STRING:
	err = SET_ERROR(EINVAL);
	break;
	case PROP_TYPE_INDEX:
	if (zfs_prop_index_to_string(prop,
	intval, &unused) != 0)
	err =
	SET_ERROR(ZFS_ERR_BADPROP);
	break;
	default:
	cmn_err(CE_PANIC,
	"unknown property type");
	}
	} else {
	err = SET_ERROR(EINVAL);
	}
	}

	/* Validate permissions */
	if (err == 0)
	err = zfs_check_settable(dsname, pair, CRED());

	if (err == 0) {
	if (source == ZPROP_SRC_INHERITED)
	err = -1; /* does not need special handling */
	else
	err = zfs_prop_set_special(dsname, source,
	pair);
	if (err == -1) {
	/*
	* For better performance we build up a list of
	* properties to set in a single transaction.
	*/
	err = nvlist_add_nvpair(genericnvl, pair);
	} else if (err != 0 && nvl != retrynvl) {
	/*
	* This may be a spurious error caused by
	* receiving quota and reservation out of order.
	* Try again in a second pass.
	*/
	err = nvlist_add_nvpair(retrynvl, pair);
	}
	}

	if (err != 0) {
	if (errlist != NULL)
	fnvlist_add_int32(errlist, propname, err);
	rv = err;
	}

	if (zfs_is_namespace_prop(prop))
	should_update_mount_cache = B_TRUE;
	}

	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
	nvl = retrynvl;
	goto retry;
	}

	if (nvlist_empty(genericnvl))
	goto out;

	/*
	* Try to set them all in one batch.
	*/
	err = dsl_props_set(dsname, source, genericnvl);
	if (err == 0)
	goto out;

	/*
	* If batching fails, we still want to set as many properties as we
	* can, so try setting them individually.
	*/
	pair = NULL;
	while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);
	err = 0;

	propval = pair;
	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	attrs = fnvpair_value_nvlist(pair);
	propval = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE);
	}

	if (nvpair_type(propval) == DATA_TYPE_STRING) {
	strval = fnvpair_value_string(propval);
	err = dsl_prop_set_string(dsname, propname,
	source, strval);
	} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
	err = dsl_prop_inherit(dsname, propname, source);
	} else {
	intval = fnvpair_value_uint64(propval);
	err = dsl_prop_set_int(dsname, propname, source,
	intval);
	}

	if (err != 0) {
	if (errlist != NULL) {
	fnvlist_add_int32(errlist, propname, err);
	}
	rv = err;
	}
	}

	out:
	if (should_update_mount_cache)
	zfs_ioctl_update_mount_cache(dsname);

	nvlist_free(genericnvl);
	nvlist_free(retrynvl);

	return (rv);
	}

	/*
	* Check that all the properties are valid user properties.
	*/
	static int
	zfs_check_userprops(nvlist_t *nvl)
	{
	nvpair_t *pair = NULL;

	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);

	if (!zfs_prop_user(propname) \|\|
	nvpair_type(pair) != DATA_TYPE_STRING)
	return (SET_ERROR(EINVAL));

	if (strlen(propname) >= ZAP_MAXNAMELEN)
	return (SET_ERROR(ENAMETOOLONG));

	if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
	return (SET_ERROR(E2BIG));
	}
	return (0);
	}

	static void
	props_skip(nvlist_t props, nvlist_t skipped, nvlist_t **newprops)
	{
	nvpair_t *pair;

	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	pair = NULL;
	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
	if (nvlist_exists(skipped, nvpair_name(pair)))
	continue;

	VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
	}
	}

	static int
	clear_received_props(const char dsname, nvlist_t props,
	nvlist_t *skipped)
	{
	int err = 0;
	nvlist_t *cleared_props = NULL;
	props_skip(props, skipped, &cleared_props);
	if (!nvlist_empty(cleared_props)) {
	/*
	* Acts on local properties until the dataset has received
	* properties at least once on or after SPA_VERSION_RECVD_PROPS.
	*/
	zprop_source_t flags = (ZPROP_SRC_NONE \|
	(dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
	err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
	}
	nvlist_free(cleared_props);
	return (err);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value name of property to set
	* zc_nvlist_src{_size} nvlist of properties to apply
	* zc_cookie received properties flag
	*
	* outputs:
	* zc_nvlist_dst{_size} error for each unapplied received property
	*/
	static int
	zfs_ioc_set_prop(zfs_cmd_t *zc)
	{
	nvlist_t *nvl;
	boolean_t received = zc->zc_cookie;
	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
	ZPROP_SRC_LOCAL);
	nvlist_t *errors;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &nvl)) != 0)
	return (error);

	if (received) {
	nvlist_t *origprops;

	if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
	(void) clear_received_props(zc->zc_name,
	origprops, nvl);
	nvlist_free(origprops);
	}

	error = dsl_prop_set_hasrecvd(zc->zc_name);
	}

	errors = fnvlist_alloc();
	if (error == 0)
	error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);

	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
	(void) put_nvlist(zc, errors);
	}

	nvlist_free(errors);
	nvlist_free(nvl);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value name of property to inherit
	* zc_cookie revert to received value if TRUE
	*
	* outputs: none
	*/
	static int
	zfs_ioc_inherit_prop(zfs_cmd_t *zc)
	{
	const char *propname = zc->zc_value;
	zfs_prop_t prop = zfs_name_to_prop(propname);
	boolean_t received = zc->zc_cookie;
	zprop_source_t source = (received
	? ZPROP_SRC_NONE /* revert to received value, if any */
	: ZPROP_SRC_INHERITED); /* explicitly inherit */
	nvlist_t *dummy;
	nvpair_t *pair;
	zprop_type_t type;
	int err;

	if (!received) {
	/*
	* Only check this in the non-received case. We want to allow
	* 'inherit -S' to revert non-inheritable properties like quota
	* and reservation to the received or default values even though
	* they are not considered inheritable.
	*/
	if (prop != ZPROP_USERPROP && !zfs_prop_inheritable(prop))
	return (SET_ERROR(EINVAL));
	}

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(propname))
	return (SET_ERROR(EINVAL));

	type = PROP_TYPE_STRING;
	} else if (prop == ZFS_PROP_VOLSIZE \|\| prop == ZFS_PROP_VERSION) {
	return (SET_ERROR(EINVAL));
	} else {
	type = zfs_prop_get_type(prop);
	}

	/*
	* zfs_prop_set_special() expects properties in the form of an
	* nvpair with type info.
	*/
	dummy = fnvlist_alloc();

	switch (type) {
	case PROP_TYPE_STRING:
	VERIFY(0 == nvlist_add_string(dummy, propname, ""));
	break;
	case PROP_TYPE_NUMBER:
	case PROP_TYPE_INDEX:
	VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
	break;
	default:
	err = SET_ERROR(EINVAL);
	goto errout;
	}

	pair = nvlist_next_nvpair(dummy, NULL);
	if (pair == NULL) {
	err = SET_ERROR(EINVAL);
	} else {
	err = zfs_prop_set_special(zc->zc_name, source, pair);
	if (err == -1) /* property is not "special", needs handling */
	err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
	source);
	}

	errout:
	nvlist_free(dummy);
	return (err);
	}

	static int
	zfs_ioc_pool_set_props(zfs_cmd_t *zc)
	{
	nvlist_t *props;
	spa_t *spa;
	int error;
	nvpair_t *pair;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props)))
	return (error);

	/*
	* If the only property is the configfile, then just do a spa_lookup()
	* to handle the faulted case.
	*/
	pair = nvlist_next_nvpair(props, NULL);
	if (pair != NULL && strcmp(nvpair_name(pair),
	zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
	nvlist_next_nvpair(props, pair) == NULL) {
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(zc->zc_name)) != NULL) {
	spa_configfile_set(spa, props, B_FALSE);
	- spa_write_cachefile(spa, B_FALSE, B_TRUE);
	+ spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	}
	mutex_exit(&spa_namespace_lock);
	if (spa != NULL) {
	nvlist_free(props);
	return (0);
	}
	}

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
	nvlist_free(props);
	return (error);
	}

	error = spa_prop_set(spa, props);

	nvlist_free(props);
	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_pool_get_props(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	nvlist_t *nvp = NULL;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
	/*
	* If the pool is faulted, there may be properties we can still
	* get (such as altroot and cachefile), so attempt to get them
	* anyway.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(zc->zc_name)) != NULL)
	error = spa_prop_get(spa, &nvp);
	mutex_exit(&spa_namespace_lock);
	} else {
	error = spa_prop_get(spa, &nvp);
	spa_close(spa, FTAG);
	}

	if (error == 0 && zc->zc_nvlist_dst != 0)
	error = put_nvlist(zc, nvp);
	else
	error = SET_ERROR(EFAULT);

	nvlist_free(nvp);
	return (error);
	}

	/*
	* innvl: {
	* "vdevprops_set_vdev" -> guid
	* "vdevprops_set_props" -> { prop -> value }
	* }
	*
	* outnvl: propname -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
	{ZPOOL_VDEV_PROPS_SET_VDEV, DATA_TYPE_UINT64, 0},
	{ZPOOL_VDEV_PROPS_SET_PROPS, DATA_TYPE_NVLIST, 0}
	};

	static int
	zfs_ioc_vdev_set_props(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;
	vdev_t *vd;
	uint64_t vdev_guid;

	/* Early validation */
	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (outnvl == NULL)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
	return (error);

	ASSERT(spa_writeable(spa));

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOENT));
	}

	error = vdev_prop_set(vd, innvl, outnvl);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* innvl: {
	* "vdevprops_get_vdev" -> guid
	* (optional) "vdevprops_get_props" -> { propname -> propid }
	* }
	*
	* outnvl: propname -> value
	*/
	static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
	{ZPOOL_VDEV_PROPS_GET_VDEV, DATA_TYPE_UINT64, 0},
	{ZPOOL_VDEV_PROPS_GET_PROPS, DATA_TYPE_NVLIST, ZK_OPTIONAL}
	};

	static int
	zfs_ioc_vdev_get_props(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;
	vdev_t *vd;
	uint64_t vdev_guid;

	/* Early validation */
	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (outnvl == NULL)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
	return (error);

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOENT));
	}

	error = vdev_prop_get(vd, innvl, outnvl);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_src{_size} nvlist of delegated permissions
	* zc_perm_action allow/unallow flag
	*
	* outputs: none
	*/
	static int
	zfs_ioc_set_fsacl(zfs_cmd_t *zc)
	{
	int error;
	nvlist_t *fsaclnv = NULL;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &fsaclnv)) != 0)
	return (error);

	/*
	* Verify nvlist is constructed correctly
	*/
	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
	nvlist_free(fsaclnv);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If we don't have PRIV_SYS_MOUNT, then validate
	* that user is allowed to hand out each permission in
	* the nvlist(s)
	*/

	error = secpolicy_zfs(CRED());
	if (error != 0) {
	if (zc->zc_perm_action == B_FALSE) {
	error = dsl_deleg_can_allow(zc->zc_name,
	fsaclnv, CRED());
	} else {
	error = dsl_deleg_can_unallow(zc->zc_name,
	fsaclnv, CRED());
	}
	}

	if (error == 0)
	error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);

	nvlist_free(fsaclnv);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* zc_nvlist_src{_size} nvlist of delegated permissions
	*/
	static int
	zfs_ioc_get_fsacl(zfs_cmd_t *zc)
	{
	nvlist_t *nvp;
	int error;

	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
	error = put_nvlist(zc, nvp);
	nvlist_free(nvp);
	}

	return (error);
	}

	static void
	zfs_create_cb(objset_t os, void arg, cred_t cr, dmu_tx_t tx)
	{
	zfs_creat_t *zct = arg;

	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
	}

	#define ZFS_PROP_UNDEFINED ((uint64_t)-1)

	/*
	* inputs:
	* os parent objset pointer (NULL if root fs)
	* fuids_ok fuids allowed in this version of the spa?
	* sa_ok SAs allowed in this version of the spa?
	* createprops list of properties requested by creator
	*
	* outputs:
	* zplprops values for the zplprops we attach to the master node object
	* is_ci true if requested file system will be purely case-insensitive
	*
	* Determine the settings for utf8only, normalization and
	* casesensitivity. Specific values may have been requested by the
	* creator and/or we can inherit values from the parent dataset. If
	* the file system is of too early a vintage, a creator can not
	* request settings for these properties, even if the requested
	* setting is the default value. We don't actually want to create dsl
	* properties for these, so remove them from the source nvlist after
	* processing.
	*/
	static int
	zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
	boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	uint64_t sense = ZFS_PROP_UNDEFINED;
	uint64_t norm = ZFS_PROP_UNDEFINED;
	uint64_t u8 = ZFS_PROP_UNDEFINED;
	int error;

	ASSERT(zplprops != NULL);

	/* parent dataset must be a filesystem */
	if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));

	/*
	* Pull out creator prop choices, if any.
	*/
	if (createprops) {
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE));
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_CASE), &sense);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_CASE));
	}

	/*
	* If the zpl version requested is whacky or the file system
	* or pool is version is too "young" to support normalization
	* and the creator tried to set a value for one of the props,
	* error out.
	*/
	if ((zplver < ZPL_VERSION_INITIAL \|\| zplver > ZPL_VERSION) \|\|
	(zplver >= ZPL_VERSION_FUID && !fuids_ok) \|\|
	(zplver >= ZPL_VERSION_SA && !sa_ok) \|\|
	(zplver < ZPL_VERSION_NORMALIZATION &&
	(norm != ZFS_PROP_UNDEFINED \|\| u8 != ZFS_PROP_UNDEFINED \|\|
	sense != ZFS_PROP_UNDEFINED)))
	return (SET_ERROR(ENOTSUP));

	/*
	* Put the version in the zplprops
	*/
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);

	if (norm == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);

	/*
	* If we're normalizing, names must always be valid UTF-8 strings.
	*/
	if (norm)
	u8 = 1;
	if (u8 == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);

	if (sense == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);

	if (is_ci)
	*is_ci = (sense == ZFS_CASE_INSENSITIVE);

	return (0);
	}

	static int
	zfs_fill_zplprops(const char dataset, nvlist_t createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	boolean_t fuids_ok, sa_ok;
	uint64_t zplver = ZPL_VERSION;
	objset_t *os = NULL;
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	spa_t *spa;
	uint64_t spa_vers;
	int error;

	zfs_get_parent(dataset, parentname, sizeof (parentname));

	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
	return (error);

	spa_vers = spa_version(spa);
	spa_close(spa, FTAG);

	zplver = zfs_zpl_version_map(spa_vers);
	fuids_ok = (zplver >= ZPL_VERSION_FUID);
	sa_ok = (zplver >= ZPL_VERSION_SA);

	/*
	* Open parent object set so we can inherit zplprop values.
	*/
	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
	return (error);

	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
	zplprops, is_ci);
	dmu_objset_rele(os, FTAG);
	return (error);
	}

	static int
	zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	boolean_t fuids_ok;
	boolean_t sa_ok;
	uint64_t zplver = ZPL_VERSION;
	int error;

	zplver = zfs_zpl_version_map(spa_vers);
	fuids_ok = (zplver >= ZPL_VERSION_FUID);
	sa_ok = (zplver >= ZPL_VERSION_SA);

	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
	createprops, zplprops, is_ci);
	return (error);
	}

	/*
	* innvl: {
	* "type" -> dmu_objset_type_t (int32)
	* (optional) "props" -> { prop -> value }
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* }
	*
	* outnvl: propname -> error code (int32)
	*/

	static const zfs_ioc_key_t zfs_keys_create[] = {
	{"type", DATA_TYPE_INT32, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_create(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error = 0;
	zfs_creat_t zct = { 0 };
	nvlist_t *nvprops = NULL;
	nvlist_t *hidden_args = NULL;
	void (cbfunc)(objset_t os, void arg, cred_t cr, dmu_tx_t *tx);
	dmu_objset_type_t type;
	boolean_t is_insensitive = B_FALSE;
	dsl_crypto_params_t *dcp = NULL;

	type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);

	switch (type) {
	case DMU_OST_ZFS:
	cbfunc = zfs_create_cb;
	break;

	case DMU_OST_ZVOL:
	cbfunc = zvol_create_cb;
	break;

	default:
	cbfunc = NULL;
	break;
	}
	if (strchr(fsname, '@') \|\|
	strchr(fsname, '%'))
	return (SET_ERROR(EINVAL));

	zct.zct_props = nvprops;

	if (cbfunc == NULL)
	return (SET_ERROR(EINVAL));

	if (type == DMU_OST_ZVOL) {
	uint64_t volsize, volblocksize;

	if (nvprops == NULL)
	return (SET_ERROR(EINVAL));
	if (nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
	return (SET_ERROR(EINVAL));

	if ((error = nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	&volblocksize)) != 0 && error != ENOENT)
	return (SET_ERROR(EINVAL));

	if (error != 0)
	volblocksize = zfs_prop_default_numeric(
	ZFS_PROP_VOLBLOCKSIZE);

	if ((error = zvol_check_volblocksize(fsname,
	volblocksize)) != 0 \|\|
	(error = zvol_check_volsize(volsize,
	volblocksize)) != 0)
	return (error);
	} else if (type == DMU_OST_ZFS) {
	int error;

	/*
	* We have to have normalization and
	* case-folding flags correct when we do the
	* file system creation, so go figure them out
	* now.
	*/
	VERIFY(nvlist_alloc(&zct.zct_zplprops,
	NV_UNIQUE_NAME, KM_SLEEP) == 0);
	error = zfs_fill_zplprops(fsname, nvprops,
	zct.zct_zplprops, &is_insensitive);
	if (error != 0) {
	nvlist_free(zct.zct_zplprops);
	return (error);
	}
	}

	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
	hidden_args, &dcp);
	if (error != 0) {
	nvlist_free(zct.zct_zplprops);
	return (error);
	}

	error = dmu_objset_create(fsname, type,
	is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);

	nvlist_free(zct.zct_zplprops);
	dsl_crypto_params_free(dcp, !!error);

	/*
	* It would be nice to do this atomically.
	*/
	if (error == 0) {
	error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
	nvprops, outnvl);
	if (error != 0) {
	spa_t *spa;
	int error2;

	/*
	* Volumes will return EBUSY and cannot be destroyed
	* until all asynchronous minor handling (e.g. from
	* setting the volmode property) has completed. Wait for
	* the spa_zvol_taskq to drain then retry.
	*/
	error2 = dsl_destroy_head(fsname);
	while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
	error2 = spa_open(fsname, &spa, FTAG);
	if (error2 == 0) {
	taskq_wait(spa->spa_zvol_taskq);
	spa_close(spa, FTAG);
	}
	error2 = dsl_destroy_head(fsname);
	}
	}
	}
	return (error);
	}

	/*
	* innvl: {
	* "origin" -> name of origin snapshot
	* (optional) "props" -> { prop -> value }
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* }
	*
	* outputs:
	* outnvl: propname -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_clone[] = {
	{"origin", DATA_TYPE_STRING, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_clone(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error = 0;
	nvlist_t *nvprops = NULL;
	const char *origin_name;

	origin_name = fnvlist_lookup_string(innvl, "origin");
	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);

	if (strchr(fsname, '@') \|\|
	strchr(fsname, '%'))
	return (SET_ERROR(EINVAL));

	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	error = dmu_objset_clone(fsname, origin_name);

	/*
	* It would be nice to do this atomically.
	*/
	if (error == 0) {
	error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
	nvprops, outnvl);
	if (error != 0)
	(void) dsl_destroy_head(fsname);
	}
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_remap[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_remap(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	/* This IOCTL is no longer supported. */
	(void) fsname, (void) innvl, (void) outnvl;
	return (0);
	}

	/*
	* innvl: {
	* "snaps" -> { snapshot1, snapshot2 }
	* (optional) "props" -> { prop -> value (string) }
	* }
	*
	* outnvl: snapshot -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_snapshot[] = {
	{"snaps", DATA_TYPE_NVLIST, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_snapshot(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	nvlist_t *snaps;
	nvlist_t *props = NULL;
	int error, poollen;
	nvpair_t *pair;

	(void) nvlist_lookup_nvlist(innvl, "props", &props);
	if (!nvlist_empty(props) &&
	zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
	return (SET_ERROR(ENOTSUP));
	if ((error = zfs_check_userprops(props)) != 0)
	return (error);

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
	poollen = strlen(poolname);
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	const char *name = nvpair_name(pair);
	char *cp = strchr(name, '@');

	/*
	* The snap name must contain an @, and the part after it must
	* contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	/*
	* The snap must be in the specified pool.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '@'))
	return (SET_ERROR(EXDEV));

	/*
	* Check for permission to set the properties on the fs.
	*/
	if (!nvlist_empty(props)) {
	*cp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_USERPROP, CRED());
	*cp = '@';
	if (error != 0)
	return (error);
	}

	/* This must be the only snap of this fs. */
	for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
	pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
	if (strncmp(name, nvpair_name(pair2), cp - name + 1)
	== 0) {
	return (SET_ERROR(EXDEV));
	}
	}
	}

	error = dsl_dataset_snapshot(snaps, props, outnvl);

	return (error);
	}

	/*
	* innvl: "message" -> string
	*/
	static const zfs_ioc_key_t zfs_keys_log_history[] = {
	{"message", DATA_TYPE_STRING, 0},
	};

	static int
	zfs_ioc_log_history(const char unused, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) unused, (void) outnvl;
	const char *message;
	char *poolname;
	spa_t *spa;
	int error;

	/*
	* The poolname in the ioctl is not set, we get it from the TSD,
	* which was set at the end of the last successful ioctl that allows
	* logging. The secpolicy func already checked that it is set.
	* Only one log ioctl is allowed after each successful ioctl, so
	* we clear the TSD here.
	*/
	poolname = tsd_get(zfs_allow_log_key);
	if (poolname == NULL)
	return (SET_ERROR(EINVAL));
	(void) tsd_set(zfs_allow_log_key, NULL);
	error = spa_open(poolname, &spa, FTAG);
	kmem_strfree(poolname);
	if (error != 0)
	return (error);

	message = fnvlist_lookup_string(innvl, "message");

	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	error = spa_history_log(spa, message);
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* This ioctl is used to set the bootenv configuration on the current
	* pool. This configuration is stored in the second padding area of the label,
	* and it is used by the bootloader(s) to store the bootloader and/or system
	* specific data.
	* The data is stored as nvlist data stream, and is protected by
	* an embedded checksum.
	* The version can have two possible values:
	* VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
	* VB_NVLIST: nvlist with arbitrary <key, value> pairs.
	*/
	static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
	{"version", DATA_TYPE_UINT64, 0},
	{"<keys>", DATA_TYPE_ANY, ZK_OPTIONAL \| ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_set_bootenv(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error;
	spa_t *spa;

	if ((error = spa_open(name, &spa, FTAG)) != 0)
	return (error);
	spa_vdev_state_enter(spa, SCL_ALL);
	error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_bootenv(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(name, &spa, FTAG)) != 0)
	return (error);
	spa_vdev_state_enter(spa, SCL_ALL);
	error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* The dp_config_rwlock must not be held when calling this, because the
	* unmount may need to write out data.
	*
	* This function is best-effort. Callers must deal gracefully if it
	* remains mounted (or is remounted after this call).
	*
	* Returns 0 if the argument is not a snapshot, or it is not currently a
	* filesystem, or we were able to unmount it. Returns error code otherwise.
	*/
	void
	zfs_unmount_snap(const char *snapname)
	{
	if (strchr(snapname, '@') == NULL)
	return;

	(void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
	}

	static int
	zfs_unmount_snap_cb(const char snapname, void arg)
	{
	(void) arg;
	zfs_unmount_snap(snapname);
	return (0);
	}

	/*
	* When a clone is destroyed, its origin may also need to be destroyed,
	* in which case it must be unmounted. This routine will do that unmount
	* if necessary.
	*/
	void
	zfs_destroy_unmount_origin(const char *fsname)
	{
	int error;
	objset_t *os;
	dsl_dataset_t *ds;

	error = dmu_objset_hold(fsname, FTAG, &os);
	if (error != 0)
	return;
	ds = dmu_objset_ds(os);
	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
	char originname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(ds->ds_prev, originname);
	dmu_objset_rele(os, FTAG);
	zfs_unmount_snap(originname);
	} else {
	dmu_objset_rele(os, FTAG);
	}
	}

	/*
	* innvl: {
	* "snaps" -> { snapshot1, snapshot2 }
	* (optional boolean) "defer"
	* }
	*
	* outnvl: snapshot -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
	{"snaps", DATA_TYPE_NVLIST, 0},
	{"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_destroy_snaps(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int poollen;
	nvlist_t *snaps;
	nvpair_t *pair;
	boolean_t defer;
	spa_t *spa;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
	defer = nvlist_exists(innvl, "defer");

	poollen = strlen(poolname);
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	const char *name = nvpair_name(pair);

	/*
	* The snap must be in the specified pool to prevent the
	* invalid removal of zvol minors below.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '@'))
	return (SET_ERROR(EXDEV));

	zfs_unmount_snap(nvpair_name(pair));
	if (spa_open(name, &spa, FTAG) == 0) {
	zvol_remove_minors(spa, name, B_TRUE);
	spa_close(spa, FTAG);
	}
	}

	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
	}

	/*
	* Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
	* All bookmarks and snapshots must be in the same pool.
	* dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
	*
	* innvl: {
	* new_bookmark1 -> existing_snapshot,
	* new_bookmark2 -> existing_bookmark,
	* }
	*
	* outnvl: bookmark -> error code (int32)
	*
	*/
	static const zfs_ioc_key_t zfs_keys_bookmark[] = {
	{"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_bookmark(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) poolname;
	return (dsl_bookmark_create(innvl, outnvl));
	}

	/*
	* innvl: {
	* property 1, property 2, ...
	* }
	*
	* outnvl: {
	* bookmark name 1 -> { property 1, property 2, ... },
	* bookmark name 2 -> { property 1, property 2, ... }
	* }
	*
	*/
	static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
	{"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST \| ZK_OPTIONAL},
	};

	static int
	zfs_ioc_get_bookmarks(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	return (dsl_get_bookmarks(fsname, innvl, outnvl));
	}

	/*
	* innvl is not used.
	*
	* outnvl: {
	* property 1, property 2, ...
	* }
	*
	*/
	static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_bookmark_props(const char bookmark, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	(void) innvl;
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	char *bmname;

	bmname = strchr(bookmark, '#');
	if (bmname == NULL)
	return (SET_ERROR(EINVAL));
	bmname++;

	(void) strlcpy(fsname, bookmark, sizeof (fsname));
	*(strchr(fsname, '#')) = '\0';

	return (dsl_get_bookmark_props(fsname, bmname, outnvl));
	}

	/*
	* innvl: {
	* bookmark name 1, bookmark name 2
	* }
	*
	* outnvl: bookmark -> error code (int32)
	*
	*/
	static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
	{"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_destroy_bookmarks(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	int error, poollen;

	poollen = strlen(poolname);
	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	const char *name = nvpair_name(pair);
	const char *cp = strchr(name, '#');

	/*
	* The bookmark name must contain an #, and the part after it
	* must contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	/*
	* The bookmark must be in the specified pool.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '#'))
	return (SET_ERROR(EXDEV));
	}

	error = dsl_bookmark_destroy(innvl, outnvl);
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_channel_program[] = {
	{"program", DATA_TYPE_STRING, 0},
	{"arg", DATA_TYPE_ANY, 0},
	{"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	{"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_channel_program(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	char *program;
	uint64_t instrlimit, memlimit;
	boolean_t sync_flag;
	nvpair_t *nvarg = NULL;

	program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
	sync_flag = B_TRUE;
	}
	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
	instrlimit = ZCP_DEFAULT_INSTRLIMIT;
	}
	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
	memlimit = ZCP_DEFAULT_MEMLIMIT;
	}
	nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);

	if (instrlimit == 0 \|\| instrlimit > zfs_lua_max_instrlimit)
	return (SET_ERROR(EINVAL));
	if (memlimit == 0 \|\| memlimit > zfs_lua_max_memlimit)
	return (SET_ERROR(EINVAL));

	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
	nvarg, outnvl));
	}

	/*
	* innvl: unused
	* outnvl: empty
	*/
	static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_pool_checkpoint(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	return (spa_checkpoint(poolname));
	}

	/*
	* innvl: unused
	* outnvl: empty
	*/
	static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_pool_discard_checkpoint(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	return (spa_checkpoint_discard(poolname));
	}

	/*
	* inputs:
	* zc_name name of dataset to destroy
	* zc_defer_destroy mark for deferred destroy
	*
	* outputs: none
	*/
	static int
	zfs_ioc_destroy(zfs_cmd_t *zc)
	{
	objset_t *os;
	dmu_objset_type_t ost;
	int err;

	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (err != 0)
	return (err);
	ost = dmu_objset_type(os);
	dmu_objset_rele(os, FTAG);

	if (ost == DMU_OST_ZFS)
	zfs_unmount_snap(zc->zc_name);

	if (strchr(zc->zc_name, '@')) {
	err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
	} else {
	err = dsl_destroy_head(zc->zc_name);
	if (err == EEXIST) {
	/*
	* It is possible that the given DS may have
	* hidden child (%recv) datasets - "leftovers"
	* resulting from the previously interrupted
	* 'zfs receive'.
	*
	* 6 extra bytes for /%recv
	*/
	char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];

	if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
	zc->zc_name, recv_clone_name) >=
	sizeof (namebuf))
	return (SET_ERROR(EINVAL));

	/*
	* Try to remove the hidden child (%recv) and after
	* that try to remove the target dataset.
	* If the hidden child (%recv) does not exist
	* the original error (EEXIST) will be returned
	*/
	err = dsl_destroy_head(namebuf);
	if (err == 0)
	err = dsl_destroy_head(zc->zc_name);
	else if (err == ENOENT)
	err = SET_ERROR(EEXIST);
	}
	}

	return (err);
	}

	/*
	* innvl: {
	* "initialize_command" -> POOL_INITIALIZE_{CANCEL\|START\|SUSPEND} (uint64)
	* "initialize_vdevs": { -> guids to initialize (nvlist)
	* "vdev_path_1": vdev_guid_1, (uint64),
	* "vdev_path_2": vdev_guid_2, (uint64),
	* ...
	* },
	* }
	*
	* outnvl: {
	* "initialize_vdevs": { -> initialization errors (nvlist)
	* "vdev_path_1": errno, see function body for possible errnos (uint64)
	* "vdev_path_2": errno, ... (uint64)
	* ...
	* }
	* }
	*
	* EINVAL is returned for an unknown commands or if any of the provided vdev
	* guids have be specified with a type other than uint64.
	*/
	static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
	{ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0},
	{ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0}
	};

	static int
	zfs_ioc_pool_initialize(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	uint64_t cmd_type;
	if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
	&cmd_type) != 0) {
	return (SET_ERROR(EINVAL));
	}

	if (!(cmd_type == POOL_INITIALIZE_CANCEL \|\|
	cmd_type == POOL_INITIALIZE_START \|\|
	cmd_type == POOL_INITIALIZE_SUSPEND)) {
	return (SET_ERROR(EINVAL));
	}

	nvlist_t *vdev_guids;
	if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
	&vdev_guids) != 0) {
	return (SET_ERROR(EINVAL));
	}

	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
	pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
	uint64_t vdev_guid;
	if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	spa_t *spa;
	int error = spa_open(poolname, &spa, FTAG);
	if (error != 0)
	return (error);

	nvlist_t *vdev_errlist = fnvlist_alloc();
	int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
	vdev_errlist);

	if (fnvlist_size(vdev_errlist) > 0) {
	fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
	vdev_errlist);
	}
	fnvlist_free(vdev_errlist);

	spa_close(spa, FTAG);
	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
	}

	/*
	* innvl: {
	* "trim_command" -> POOL_TRIM_{CANCEL\|START\|SUSPEND} (uint64)
	* "trim_vdevs": { -> guids to TRIM (nvlist)
	* "vdev_path_1": vdev_guid_1, (uint64),
	* "vdev_path_2": vdev_guid_2, (uint64),
	* ...
	* },
	* "trim_rate" -> Target TRIM rate in bytes/sec.
	* "trim_secure" -> Set to request a secure TRIM.
	* }
	*
	* outnvl: {
	* "trim_vdevs": { -> TRIM errors (nvlist)
	* "vdev_path_1": errno, see function body for possible errnos (uint64)
	* "vdev_path_2": errno, ... (uint64)
	* ...
	* }
	* }
	*
	* EINVAL is returned for an unknown commands or if any of the provided vdev
	* guids have be specified with a type other than uint64.
	*/
	static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
	{ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0},
	{ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0},
	{ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL},
	{ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_pool_trim(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	uint64_t cmd_type;
	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
	return (SET_ERROR(EINVAL));

	if (!(cmd_type == POOL_TRIM_CANCEL \|\|
	cmd_type == POOL_TRIM_START \|\|
	cmd_type == POOL_TRIM_SUSPEND)) {
	return (SET_ERROR(EINVAL));
	}

	nvlist_t *vdev_guids;
	if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
	return (SET_ERROR(EINVAL));

	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
	pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
	uint64_t vdev_guid;
	if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	/* Optional, defaults to maximum rate when not provided */
	uint64_t rate;
	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
	rate = 0;

	/* Optional, defaults to standard TRIM when not provided */
	boolean_t secure;
	if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
	&secure) != 0) {
	secure = B_FALSE;
	}

	spa_t *spa;
	int error = spa_open(poolname, &spa, FTAG);
	if (error != 0)
	return (error);

	nvlist_t *vdev_errlist = fnvlist_alloc();
	int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
	rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);

	if (fnvlist_size(vdev_errlist) > 0)
	fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);

	fnvlist_free(vdev_errlist);

	spa_close(spa, FTAG);
	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
	}

	/*
	* This ioctl waits for activity of a particular type to complete. If there is
	* no activity of that type in progress, it returns immediately, and the
	* returned value "waited" is false. If there is activity in progress, and no
	* tag is passed in, the ioctl blocks until all activity of that type is
	* complete, and then returns with "waited" set to true.
	*
	* If a tag is provided, it identifies a particular instance of an activity to
	* wait for. Currently, this is only valid for use with 'initialize', because
	* that is the only activity for which there can be multiple instances running
	* concurrently. In the case of 'initialize', the tag corresponds to the guid of
	* the vdev on which to wait.
	*
	* If a thread waiting in the ioctl receives a signal, the call will return
	* immediately, and the return value will be EINTR.
	*
	* innvl: {
	* "wait_activity" -> int32_t
	* (optional) "wait_tag" -> uint64_t
	* }
	*
	* outnvl: "waited" -> boolean_t
	*/
	static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
	{ZPOOL_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
	{ZPOOL_WAIT_TAG, DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_wait(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int32_t activity;
	uint64_t tag;
	boolean_t waited;
	int error;

	if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
	return (EINVAL);

	if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
	error = spa_wait_tag(name, activity, tag, &waited);
	else
	error = spa_wait(name, activity, &waited);

	if (error == 0)
	fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);

	return (error);
	}

	/*
	* This ioctl waits for activity of a particular type to complete. If there is
	* no activity of that type in progress, it returns immediately, and the
	* returned value "waited" is false. If there is activity in progress, and no
	* tag is passed in, the ioctl blocks until all activity of that type is
	* complete, and then returns with "waited" set to true.
	*
	* If a thread waiting in the ioctl receives a signal, the call will return
	* immediately, and the return value will be EINTR.
	*
	* innvl: {
	* "wait_activity" -> int32_t
	* }
	*
	* outnvl: "waited" -> boolean_t
	*/
	static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
	{ZFS_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
	};

	static int
	zfs_ioc_wait_fs(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int32_t activity;
	boolean_t waited = B_FALSE;
	int error;
	dsl_pool_t *dp;
	dsl_dir_t *dd;
	dsl_dataset_t *ds;

	if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
	return (SET_ERROR(EINVAL));

	if (activity >= ZFS_WAIT_NUM_ACTIVITIES \|\| activity < 0)
	return (SET_ERROR(EINVAL));

	if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
	return (error);

	if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dd = ds->ds_dir;
	mutex_enter(&dd->dd_activity_lock);
	dd->dd_activity_waiters++;

	/*
	* We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
	* aren't evicted while we're waiting. Normally this is prevented by
	* holding the pool, but we can't do that while we're waiting since
	* that would prevent TXGs from syncing out. Some of the functionality
	* of long-holds (e.g. preventing deletion) is unnecessary for this
	* case, since we would cancel the waiters before proceeding with a
	* deletion. An alternative mechanism for keeping the dataset around
	* could be developed but this is simpler.
	*/
	dsl_dataset_long_hold(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	error = dsl_dir_wait(dd, ds, activity, &waited);

	dsl_dataset_long_rele(ds, FTAG);
	dd->dd_activity_waiters--;
	if (dd->dd_activity_waiters == 0)
	cv_signal(&dd->dd_activity_cv);
	mutex_exit(&dd->dd_activity_lock);

	dsl_dataset_rele(ds, FTAG);

	if (error == 0)
	fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);

	return (error);
	}

	/*
	* fsname is name of dataset to rollback (to most recent snapshot)
	*
	* innvl may contain name of expected target snapshot
	*
	* outnvl: "target" -> name of most recent snapshot
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_rollback[] = {
	{"target", DATA_TYPE_STRING, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_rollback(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	zfsvfs_t *zfsvfs;
	zvol_state_handle_t *zv;
	char *target = NULL;
	int error;

	(void) nvlist_lookup_string(innvl, "target", &target);
	if (target != NULL) {
	const char *cp = strchr(target, '@');

	/*
	* The snap name must contain an @, and the part after it must
	* contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));
	}

	if (getzfsvfs(fsname, &zfsvfs) == 0) {
	dsl_dataset_t *ds;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	if (error == 0) {
	int resume_err;

	error = dsl_dataset_rollback(fsname, target, zfsvfs,
	outnvl);
	resume_err = zfs_resume_fs(zfsvfs, ds);
	error = error ? error : resume_err;
	}
	zfs_vfs_rele(zfsvfs);
	} else if ((zv = zvol_suspend(fsname)) != NULL) {
	error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
	outnvl);
	zvol_resume(zv);
	} else {
	error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
	}
	return (error);
	}

	static int
	recursive_unmount(const char fsname, void arg)
	{
	const char *snapname = arg;
	char *fullname;

	fullname = kmem_asprintf("%s@%s", fsname, snapname);
	zfs_unmount_snap(fullname);
	kmem_strfree(fullname);

	return (0);
	}

	/*
	*
	* snapname is the snapshot to redact.
	* innvl: {
	* "bookname" -> (string)
	* shortname of the redaction bookmark to generate
	* "snapnv" -> (nvlist, values ignored)
	* snapshots to redact snapname with respect to
	* }
	*
	* outnvl is unused
	*/

	static const zfs_ioc_key_t zfs_keys_redact[] = {
	{"bookname", DATA_TYPE_STRING, 0},
	{"snapnv", DATA_TYPE_NVLIST, 0},
	};

	static int
	zfs_ioc_redact(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	nvlist_t *redactnvl = NULL;
	char *redactbook = NULL;

	if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
	return (SET_ERROR(EINVAL));
	if (fnvlist_num_pairs(redactnvl) == 0)
	return (SET_ERROR(ENXIO));
	if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
	return (SET_ERROR(EINVAL));

	return (dmu_redact_snap(snapname, redactnvl, redactbook));
	}

	/*
	* inputs:
	* zc_name old name of dataset
	* zc_value new name of dataset
	* zc_cookie recursive flag (only valid for snapshots)
	*
	* outputs: none
	*/
	static int
	zfs_ioc_rename(zfs_cmd_t *zc)
	{
	objset_t *os;
	dmu_objset_type_t ost;
	boolean_t recursive = zc->zc_cookie & 1;
	boolean_t nounmount = !!(zc->zc_cookie & 2);
	char *at;
	int err;

	/* "zfs rename" from and to ...%recv datasets should both fail */
	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 \|\|
	dataset_namecheck(zc->zc_value, NULL, NULL) != 0 \|\|
	strchr(zc->zc_name, '%') \|\| strchr(zc->zc_value, '%'))
	return (SET_ERROR(EINVAL));

	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (err != 0)
	return (err);
	ost = dmu_objset_type(os);
	dmu_objset_rele(os, FTAG);

	at = strchr(zc->zc_name, '@');
	if (at != NULL) {
	/* snaps must be in same fs */
	int error;

	if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
	return (SET_ERROR(EXDEV));
	*at = '\0';
	if (ost == DMU_OST_ZFS && !nounmount) {
	error = dmu_objset_find(zc->zc_name,
	recursive_unmount, at + 1,
	recursive ? DS_FIND_CHILDREN : 0);
	if (error != 0) {
	*at = '@';
	return (error);
	}
	}
	error = dsl_dataset_rename_snapshot(zc->zc_name,
	at + 1, strchr(zc->zc_value, '@') + 1, recursive);
	*at = '@';

	return (error);
	} else {
	return (dsl_dir_rename(zc->zc_name, zc->zc_value));
	}
	}

	static int
	zfs_check_settable(const char dsname, nvpair_t pair, cred_t *cr)
	{
	const char *propname = nvpair_name(pair);
	boolean_t issnap = (strchr(dsname, '@') != NULL);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	uint64_t intval, compval;
	int err;

	if (prop == ZPROP_USERPROP) {
	if (zfs_prop_user(propname)) {
	if ((err = zfs_secpolicy_write_perms(dsname,
	ZFS_DELEG_PERM_USERPROP, cr)))
	return (err);
	return (0);
	}

	if (!issnap && zfs_prop_userquota(propname)) {
	const char *perm = NULL;
	const char *uq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
	const char *gq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
	const char *uiq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
	const char *giq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
	const char *pq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
	const char *piq_prefix = zfs_userquota_prop_prefixes[\
	ZFS_PROP_PROJECTOBJQUOTA];

	if (strncmp(propname, uq_prefix,
	strlen(uq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_USERQUOTA;
	} else if (strncmp(propname, uiq_prefix,
	strlen(uiq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_USEROBJQUOTA;
	} else if (strncmp(propname, gq_prefix,
	strlen(gq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_GROUPQUOTA;
	} else if (strncmp(propname, giq_prefix,
	strlen(giq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
	} else if (strncmp(propname, pq_prefix,
	strlen(pq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_PROJECTQUOTA;
	} else if (strncmp(propname, piq_prefix,
	strlen(piq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
	} else {
	/* {USER\|GROUP\|PROJECT}USED are read-only */
	return (SET_ERROR(EINVAL));
	}

	if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
	return (err);
	return (0);
	}

	return (SET_ERROR(EINVAL));
	}

	if (issnap)
	return (SET_ERROR(EINVAL));

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	/*
	* dsl_prop_get_all_impl() returns properties in this
	* format.
	*/
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) == 0);
	}

	/*
	* Check that this value is valid for this pool version
	*/
	switch (prop) {
	case ZFS_PROP_COMPRESSION:
	/*
	* If the user specified gzip compression, make sure
	* the SPA supports it. We ignore any errors here since
	* we'll catch them later.
	*/
	if (nvpair_value_uint64(pair, &intval) == 0) {
	compval = ZIO_COMPRESS_ALGO(intval);
	if (compval >= ZIO_COMPRESS_GZIP_1 &&
	compval <= ZIO_COMPRESS_GZIP_9 &&
	zfs_earlier_version(dsname,
	SPA_VERSION_GZIP_COMPRESSION)) {
	return (SET_ERROR(ENOTSUP));
	}

	if (compval == ZIO_COMPRESS_ZLE &&
	zfs_earlier_version(dsname,
	SPA_VERSION_ZLE_COMPRESSION))
	return (SET_ERROR(ENOTSUP));

	if (compval == ZIO_COMPRESS_LZ4) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LZ4_COMPRESS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}

	if (compval == ZIO_COMPRESS_ZSTD) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_ZSTD_COMPRESS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	}
	break;

	case ZFS_PROP_COPIES:
	if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
	return (SET_ERROR(ENOTSUP));
	break;

	case ZFS_PROP_VOLBLOCKSIZE:
	case ZFS_PROP_RECORDSIZE:
	/* Record sizes above 128k need the feature to be enabled */
	if (nvpair_value_uint64(pair, &intval) == 0 &&
	intval > SPA_OLD_MAXBLOCKSIZE) {
	spa_t *spa;

	/*
	* We don't allow setting the property above 1MB,
	* unless the tunable has been changed.
	*/
	if (intval > zfs_max_recordsize \|\|
	intval > SPA_MAXBLOCKSIZE)
	return (SET_ERROR(ERANGE));

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LARGE_BLOCKS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;

	case ZFS_PROP_DNODESIZE:
	/* Dnode sizes above 512 need the feature to be enabled */
	if (nvpair_value_uint64(pair, &intval) == 0 &&
	intval != ZFS_DNSIZE_LEGACY) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LARGE_DNODE)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;

	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
	/*
	* This property could require the allocation classes
	* feature to be active for setting, however we allow
	* it so that tests of settable properties succeed.
	* The CLI will issue a warning in this case.
	*/
	break;

	case ZFS_PROP_SHARESMB:
	if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
	return (SET_ERROR(ENOTSUP));
	break;

	case ZFS_PROP_ACLINHERIT:
	if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
	nvpair_value_uint64(pair, &intval) == 0) {
	if (intval == ZFS_ACL_PASSTHROUGH_X &&
	zfs_earlier_version(dsname,
	SPA_VERSION_PASSTHROUGH_X))
	return (SET_ERROR(ENOTSUP));
	}
	break;
	case ZFS_PROP_CHECKSUM:
	case ZFS_PROP_DEDUP:
	{
	spa_feature_t feature;
	spa_t *spa;
	int err;

	/* dedup feature version checks */
	if (prop == ZFS_PROP_DEDUP &&
	zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
	return (SET_ERROR(ENOTSUP));

	if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
	nvpair_value_uint64(pair, &intval) == 0) {
	/* check prop value is enabled in features */
	feature = zio_checksum_to_feature(
	intval & ZIO_CHECKSUM_MASK);
	if (feature == SPA_FEATURE_NONE)
	break;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa, feature)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;
	}

	default:
	break;
	}

	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
	}

	/*
	* Removes properties from the given props list that fail permission checks
	* needed to clear them and to restore them in case of a receive error. For each
	* property, make sure we have both set and inherit permissions.
	*
	* Returns the first error encountered if any permission checks fail. If the
	* caller provides a non-NULL errlist, it also gives the complete list of names
	* of all the properties that failed a permission check along with the
	* corresponding error numbers. The caller is responsible for freeing the
	* returned errlist.
	*
	* If every property checks out successfully, zero is returned and the list
	* pointed at by errlist is NULL.
	*/
	static int
	zfs_check_clearable(const char dataset, nvlist_t props, nvlist_t **errlist)
	{
	zfs_cmd_t *zc;
	nvpair_t pair, next_pair;
	nvlist_t *errors;
	int err, rv = 0;

	if (props == NULL)
	return (0);

	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
	(void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
	pair = nvlist_next_nvpair(props, NULL);
	while (pair != NULL) {
	next_pair = nvlist_next_nvpair(props, pair);

	(void) strlcpy(zc->zc_value, nvpair_name(pair),
	sizeof (zc->zc_value));
	if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 \|\|
	(err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
	VERIFY(nvlist_remove_nvpair(props, pair) == 0);
	VERIFY(nvlist_add_int32(errors,
	zc->zc_value, err) == 0);
	}
	pair = next_pair;
	}
	kmem_free(zc, sizeof (zfs_cmd_t));

	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
	nvlist_free(errors);
	errors = NULL;
	} else {
	VERIFY(nvpair_value_int32(pair, &rv) == 0);
	}

	if (errlist == NULL)
	nvlist_free(errors);
	else
	*errlist = errors;

	return (rv);
	}

	static boolean_t
	propval_equals(nvpair_t p1, nvpair_t p2)
	{
	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
	/* dsl_prop_get_all_impl() format */
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&p1) == 0);
	}

	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&p2) == 0);
	}

	if (nvpair_type(p1) != nvpair_type(p2))
	return (B_FALSE);

	if (nvpair_type(p1) == DATA_TYPE_STRING) {
	char valstr1, valstr2;

	VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
	VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
	return (strcmp(valstr1, valstr2) == 0);
	} else {
	uint64_t intval1, intval2;

	VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
	VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
	return (intval1 == intval2);
	}
	}

	/*
	* Remove properties from props if they are not going to change (as determined
	* by comparison with origprops). Remove them from origprops as well, since we
	* do not need to clear or restore properties that won't change.
	*/
	static void
	props_reduce(nvlist_t props, nvlist_t origprops)
	{
	nvpair_t pair, next_pair;

	if (origprops == NULL)
	return; /* all props need to be received */

	pair = nvlist_next_nvpair(props, NULL);
	while (pair != NULL) {
	const char *propname = nvpair_name(pair);
	nvpair_t *match;

	next_pair = nvlist_next_nvpair(props, pair);

	if ((nvlist_lookup_nvpair(origprops, propname,
	&match) != 0) \|\| !propval_equals(pair, match))
	goto next; /* need to set received value */

	/* don't clear the existing received value */
	(void) nvlist_remove_nvpair(origprops, match);
	/* don't bother receiving the property */
	(void) nvlist_remove_nvpair(props, pair);
	next:
	pair = next_pair;
	}
	}

	/*
	* Extract properties that cannot be set PRIOR to the receipt of a dataset.
	* For example, refquota cannot be set until after the receipt of a dataset,
	* because in replication streams, an older/earlier snapshot may exceed the
	* refquota. We want to receive the older/earlier snapshot, but setting
	* refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
	* the older/earlier snapshot from being received (with EDQUOT).
	*
	* The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
	*
	* libzfs will need to be judicious handling errors encountered by props
	* extracted by this function.
	*/
	static nvlist_t *
	extract_delay_props(nvlist_t *props)
	{
	nvlist_t *delayprops;
	nvpair_t nvp, tmp;
	static const zfs_prop_t delayable[] = {
	ZFS_PROP_REFQUOTA,
	ZFS_PROP_KEYLOCATION,
	/*
	* Setting ZFS_PROP_SHARESMB requires the objset type to be
	* known, which is not possible prior to receipt of raw sends.
	*/
	ZFS_PROP_SHARESMB,
	0
	};
	int i;

	VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
	nvp = nvlist_next_nvpair(props, nvp)) {
	/*
	* strcmp() is safe because zfs_prop_to_name() always returns
	* a bounded string.
	*/
	for (i = 0; delayable[i] != 0; i++) {
	if (strcmp(zfs_prop_to_name(delayable[i]),
	nvpair_name(nvp)) == 0) {
	break;
	}
	}
	if (delayable[i] != 0) {
	tmp = nvlist_prev_nvpair(props, nvp);
	VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
	VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
	nvp = tmp;
	}
	}

	if (nvlist_empty(delayprops)) {
	nvlist_free(delayprops);
	delayprops = NULL;
	}
	return (delayprops);
	}

	static void
	zfs_allow_log_destroy(void *arg)
	{
	char *poolname = arg;

	if (poolname != NULL)
	kmem_strfree(poolname);
	}

	#ifdef ZFS_DEBUG
	static boolean_t zfs_ioc_recv_inject_err;
	#endif

	/*
	* nvlist 'errors' is always allocated. It will contain descriptions of
	* encountered errors, if any. It's the callers responsibility to free.
	*/
	static int
	zfs_ioc_recv_impl(char tofs, char tosnap, char origin, nvlist_t recvprops,
	nvlist_t localprops, nvlist_t hidden_args, boolean_t force,
	boolean_t heal, boolean_t resumable, int input_fd,
	dmu_replay_record_t begin_record, uint64_t read_bytes,
	uint64_t errflags, nvlist_t *errors)
	{
	dmu_recv_cookie_t drc;
	int error = 0;
	int props_error = 0;
	offset_t off, noff;
	nvlist_t *local_delayprops = NULL;
	nvlist_t *recv_delayprops = NULL;
	nvlist_t *inherited_delayprops = NULL;
	nvlist_t origprops = NULL; / existing properties */
	nvlist_t origrecvd = NULL; / existing received properties */
	boolean_t first_recvd_props = B_FALSE;
	boolean_t tofs_was_redacted;
	zfs_file_t *input_fp;

	*read_bytes = 0;
	*errflags = 0;
	*errors = fnvlist_alloc();
	off = 0;

	if ((input_fp = zfs_file_get(input_fd)) == NULL)
	return (SET_ERROR(EBADF));

	noff = off = zfs_file_off(input_fp);
	error = dmu_recv_begin(tofs, tosnap, begin_record, force, heal,
	resumable, localprops, hidden_args, origin, &drc, input_fp,
	&off);
	if (error != 0)
	goto out;
	tofs_was_redacted = dsl_get_redacted(drc.drc_ds);

	/*
	* Set properties before we receive the stream so that they are applied
	* to the new data. Note that we must call dmu_recv_stream() if
	* dmu_recv_begin() succeeds.
	*/
	if (recvprops != NULL && !drc.drc_newfs) {
	if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
	SPA_VERSION_RECVD_PROPS &&
	!dsl_prop_get_hasrecvd(tofs))
	first_recvd_props = B_TRUE;

	/*
	* If new received properties are supplied, they are to
	* completely replace the existing received properties,
	* so stash away the existing ones.
	*/
	if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
	nvlist_t *errlist = NULL;
	/*
	* Don't bother writing a property if its value won't
	* change (and avoid the unnecessary security checks).
	*
	* The first receive after SPA_VERSION_RECVD_PROPS is a
	* special case where we blow away all local properties
	* regardless.
	*/
	if (!first_recvd_props)
	props_reduce(recvprops, origrecvd);
	if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
	(void) nvlist_merge(*errors, errlist, 0);
	nvlist_free(errlist);

	if (clear_received_props(tofs, origrecvd,
	first_recvd_props ? NULL : recvprops) != 0)
	*errflags \|= ZPROP_ERR_NOCLEAR;
	} else {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	}

	/*
	* Stash away existing properties so we can restore them on error unless
	* we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
	* case "origrecvd" will take care of that.
	*/
	if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
	objset_t *os;
	if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
	if (dsl_prop_get_all(os, &origprops) != 0) {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	dmu_objset_rele(os, FTAG);
	} else {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	}

	if (recvprops != NULL) {
	props_error = dsl_prop_set_hasrecvd(tofs);

	if (props_error == 0) {
	recv_delayprops = extract_delay_props(recvprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
	recvprops, *errors);
	}
	}

	if (localprops != NULL) {
	nvlist_t *oprops = fnvlist_alloc();
	nvlist_t *xprops = fnvlist_alloc();
	nvpair_t *nvp = NULL;

	while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
	if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
	/* -x property */
	const char *name = nvpair_name(nvp);
	zfs_prop_t prop = zfs_name_to_prop(name);
	if (prop != ZPROP_USERPROP) {
	if (!zfs_prop_inheritable(prop))
	continue;
	} else if (!zfs_prop_user(name))
	continue;
	fnvlist_add_boolean(xprops, name);
	} else {
	/* -o property=value */
	fnvlist_add_nvpair(oprops, nvp);
	}
	}

	local_delayprops = extract_delay_props(oprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
	oprops, *errors);
	inherited_delayprops = extract_delay_props(xprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
	xprops, *errors);

	nvlist_free(oprops);
	nvlist_free(xprops);
	}

	error = dmu_recv_stream(&drc, &off);

	if (error == 0) {
	zfsvfs_t *zfsvfs = NULL;
	zvol_state_handle_t *zv = NULL;

	if (getzfsvfs(tofs, &zfsvfs) == 0) {
	/* online recv */
	dsl_dataset_t *ds;
	int end_err;
	boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
	begin_record->drr_u.drr_begin.
	drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	/*
	* If the suspend fails, then the recv_end will
	* likely also fail, and clean up after itself.
	*/
	end_err = dmu_recv_end(&drc, zfsvfs);
	/*
	* If the dataset was not redacted, but we received a
	* redacted stream onto it, we need to unmount the
	* dataset. Otherwise, resume the filesystem.
	*/
	if (error == 0 && !drc.drc_newfs &&
	stream_is_redacted && !tofs_was_redacted) {
	error = zfs_end_fs(zfsvfs, ds);
	} else if (error == 0) {
	error = zfs_resume_fs(zfsvfs, ds);
	}
	error = error ? error : end_err;
	zfs_vfs_rele(zfsvfs);
	} else if ((zv = zvol_suspend(tofs)) != NULL) {
	error = dmu_recv_end(&drc, zvol_tag(zv));
	zvol_resume(zv);
	} else {
	error = dmu_recv_end(&drc, NULL);
	}

	/* Set delayed properties now, after we're done receiving. */
	if (recv_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
	recv_delayprops, *errors);
	}
	if (local_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
	local_delayprops, *errors);
	}
	if (inherited_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
	inherited_delayprops, *errors);
	}
	}

	/*
	* Merge delayed props back in with initial props, in case
	* we're DEBUG and zfs_ioc_recv_inject_err is set (which means
	* we have to make sure clear_received_props() includes
	* the delayed properties).
	*
	* Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
	* using ASSERT() will be just like a VERIFY.
	*/
	if (recv_delayprops != NULL) {
	ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
	nvlist_free(recv_delayprops);
	}
	if (local_delayprops != NULL) {
	ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
	nvlist_free(local_delayprops);
	}
	if (inherited_delayprops != NULL) {
	ASSERT(nvlist_merge(localprops, inherited_delayprops, 0) == 0);
	nvlist_free(inherited_delayprops);
	}
	*read_bytes = off - noff;

	#ifdef ZFS_DEBUG
	if (zfs_ioc_recv_inject_err) {
	zfs_ioc_recv_inject_err = B_FALSE;
	error = 1;
	}
	#endif

	/*
	* On error, restore the original props.
	*/
	if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
	if (clear_received_props(tofs, recvprops, NULL) != 0) {
	/*
	* We failed to clear the received properties.
	* Since we may have left a $recvd value on the
	* system, we can't clear the $hasrecvd flag.
	*/
	*errflags \|= ZPROP_ERR_NORESTORE;
	} else if (first_recvd_props) {
	dsl_prop_unset_hasrecvd(tofs);
	}

	if (origrecvd == NULL && !drc.drc_newfs) {
	/* We failed to stash the original properties. */
	*errflags \|= ZPROP_ERR_NORESTORE;
	}

	/*
	* dsl_props_set() will not convert RECEIVED to LOCAL on or
	* after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
	* explicitly if we're restoring local properties cleared in the
	* first new-style receive.
	*/
	if (origrecvd != NULL &&
	zfs_set_prop_nvlist(tofs, (first_recvd_props ?
	ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
	origrecvd, NULL) != 0) {
	/*
	* We stashed the original properties but failed to
	* restore them.
	*/
	*errflags \|= ZPROP_ERR_NORESTORE;
	}
	}
	if (error != 0 && localprops != NULL && !drc.drc_newfs &&
	!first_recvd_props) {
	nvlist_t *setprops;
	nvlist_t *inheritprops;
	nvpair_t *nvp;

	if (origprops == NULL) {
	/* We failed to stash the original properties. */
	*errflags \|= ZPROP_ERR_NORESTORE;
	goto out;
	}

	/* Restore original props */
	setprops = fnvlist_alloc();
	inheritprops = fnvlist_alloc();
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	const char *source;
	nvlist_t *attrs;

	if (!nvlist_exists(origprops, name)) {
	/*
	* Property was not present or was explicitly
	* inherited before the receive, restore this.
	*/
	fnvlist_add_boolean(inheritprops, name);
	continue;
	}
	attrs = fnvlist_lookup_nvlist(origprops, name);
	source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);

	/* Skip received properties */
	if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
	continue;

	if (strcmp(source, tofs) == 0) {
	/* Property was locally set */
	fnvlist_add_nvlist(setprops, name, attrs);
	} else {
	/* Property was implicitly inherited */
	fnvlist_add_boolean(inheritprops, name);
	}
	}

	if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
	NULL) != 0)
	*errflags \|= ZPROP_ERR_NORESTORE;
	if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
	NULL) != 0)
	*errflags \|= ZPROP_ERR_NORESTORE;

	nvlist_free(setprops);
	nvlist_free(inheritprops);
	}
	out:
	zfs_file_put(input_fp);
	nvlist_free(origrecvd);
	nvlist_free(origprops);

	if (error == 0)
	error = props_error;

	return (error);
	}

	/*
	* inputs:
	* zc_name name of containing filesystem (unused)
	* zc_nvlist_src{_size} nvlist of properties to apply
	* zc_nvlist_conf{_size} nvlist of properties to exclude
	* (DATA_TYPE_BOOLEAN) and override (everything else)
	* zc_value name of snapshot to create
	* zc_string name of clone origin (if DRR_FLAG_CLONE)
	* zc_cookie file descriptor to recv from
	* zc_begin_record the BEGIN record of the stream (not byteswapped)
	* zc_guid force flag
	*
	* outputs:
	* zc_cookie number of bytes read
	* zc_obj zprop_errflags_t
	* zc_nvlist_dst{_size} error for each unapplied received property
	*/
	static int
	zfs_ioc_recv(zfs_cmd_t *zc)
	{
	dmu_replay_record_t begin_record;
	nvlist_t *errors = NULL;
	nvlist_t *recvdprops = NULL;
	nvlist_t *localprops = NULL;
	char *origin = NULL;
	char *tosnap;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	int error = 0;

	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 \|\|
	strchr(zc->zc_value, '@') == NULL \|\|
	strchr(zc->zc_value, '%'))
	return (SET_ERROR(EINVAL));

	(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
	tosnap = strchr(tofs, '@');
	*tosnap++ = '\0';

	if (zc->zc_nvlist_src != 0 &&
	(error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &recvdprops)) != 0)
	return (error);

	if (zc->zc_nvlist_conf != 0 &&
	(error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &localprops)) != 0)
	return (error);

	if (zc->zc_string[0])
	origin = zc->zc_string;

	begin_record.drr_type = DRR_BEGIN;
	begin_record.drr_payloadlen = 0;
	begin_record.drr_u.drr_begin = zc->zc_begin_record;

	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
	NULL, zc->zc_guid, B_FALSE, B_FALSE, zc->zc_cookie, &begin_record,
	&zc->zc_cookie, &zc->zc_obj, &errors);
	nvlist_free(recvdprops);
	nvlist_free(localprops);

	/*
	* Now that all props, initial and delayed, are set, report the prop
	* errors to the caller.
	*/
	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
	(nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 \|\|
	put_nvlist(zc, errors) != 0)) {
	/*
	* Caller made zc->zc_nvlist_dst less than the minimum expected
	* size or supplied an invalid address.
	*/
	error = SET_ERROR(EINVAL);
	}

	nvlist_free(errors);

	return (error);
	}

	/*
	* innvl: {
	* "snapname" -> full name of the snapshot to create
	* (optional) "props" -> received properties to set (nvlist)
	* (optional) "localprops" -> override and exclude properties (nvlist)
	* (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
	* "begin_record" -> non-byteswapped dmu_replay_record_t
	* "input_fd" -> file descriptor to read stream from (int32)
	* (optional) "force" -> force flag (value ignored)
	* (optional) "heal" -> use send stream to heal data corruption
	* (optional) "resumable" -> resumable flag (value ignored)
	* (optional) "cleanup_fd" -> unused
	* (optional) "action_handle" -> unused
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* }
	*
	* outnvl: {
	* "read_bytes" -> number of bytes read
	* "error_flags" -> zprop_errflags_t
	* "errors" -> error for each unapplied received property (nvlist)
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_recv_new[] = {
	{"snapname", DATA_TYPE_STRING, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"localprops", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"origin", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"begin_record", DATA_TYPE_BYTE_ARRAY, 0},
	{"input_fd", DATA_TYPE_INT32, 0},
	{"force", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"heal", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"resumable", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	{"action_handle", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_recv_new(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	dmu_replay_record_t *begin_record;
	uint_t begin_record_size;
	nvlist_t *errors = NULL;
	nvlist_t *recvprops = NULL;
	nvlist_t *localprops = NULL;
	nvlist_t *hidden_args = NULL;
	char *snapname;
	char *origin = NULL;
	char *tosnap;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	boolean_t force;
	boolean_t heal;
	boolean_t resumable;
	uint64_t read_bytes = 0;
	uint64_t errflags = 0;
	int input_fd = -1;
	int error;

	snapname = fnvlist_lookup_string(innvl, "snapname");

	if (dataset_namecheck(snapname, NULL, NULL) != 0 \|\|
	strchr(snapname, '@') == NULL \|\|
	strchr(snapname, '%'))
	return (SET_ERROR(EINVAL));

	(void) strlcpy(tofs, snapname, sizeof (tofs));
	tosnap = strchr(tofs, '@');
	*tosnap++ = '\0';

	error = nvlist_lookup_string(innvl, "origin", &origin);
	if (error && error != ENOENT)
	return (error);

	error = nvlist_lookup_byte_array(innvl, "begin_record",
	(uchar_t **)&begin_record, &begin_record_size);
	if (error != 0 \|\| begin_record_size != sizeof (*begin_record))
	return (SET_ERROR(EINVAL));

	input_fd = fnvlist_lookup_int32(innvl, "input_fd");

	force = nvlist_exists(innvl, "force");
	heal = nvlist_exists(innvl, "heal");
	resumable = nvlist_exists(innvl, "resumable");

	/* we still use "props" here for backwards compatibility */
	error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
	if (error && error != ENOENT)
	return (error);

	error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
	if (error && error != ENOENT)
	return (error);

	error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
	if (error && error != ENOENT)
	return (error);

	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
	hidden_args, force, heal, resumable, input_fd, begin_record,
	&read_bytes, &errflags, &errors);

	fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
	fnvlist_add_uint64(outnvl, "error_flags", errflags);
	fnvlist_add_nvlist(outnvl, "errors", errors);

	nvlist_free(errors);
	nvlist_free(recvprops);
	nvlist_free(localprops);

	return (error);
	}

	typedef struct dump_bytes_io {
	zfs_file_t *dbi_fp;
	caddr_t dbi_buf;
	int dbi_len;
	int dbi_err;
	} dump_bytes_io_t;

	static void
	dump_bytes_cb(void *arg)
	{
	dump_bytes_io_t dbi = (dump_bytes_io_t )arg;
	zfs_file_t *fp;
	caddr_t buf;

	fp = dbi->dbi_fp;
	buf = dbi->dbi_buf;

	dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
	}

	static int
	dump_bytes(objset_t os, void buf, int len, void *arg)
	{
	dump_bytes_io_t dbi;

	dbi.dbi_fp = arg;
	dbi.dbi_buf = buf;
	dbi.dbi_len = len;

	#if defined(HAVE_LARGE_STACKS)
	dump_bytes_cb(&dbi);
	#else
	/*
	* The vn_rdwr() call is performed in a taskq to ensure that there is
	* always enough stack space to write safely to the target filesystem.
	* The ZIO_TYPE_FREE threads are used because there can be a lot of
	* them and they are used in vdev_file.c for a similar purpose.
	*/
	spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
	ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
	#endif /* HAVE_LARGE_STACKS */

	return (dbi.dbi_err);
	}

	/*
	* inputs:
	* zc_name name of snapshot to send
	* zc_cookie file descriptor to send stream to
	* zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
	* zc_sendobj objsetid of snapshot to send
	* zc_fromobj objsetid of incremental fromsnap (may be zero)
	* zc_guid if set, estimate size of stream only. zc_cookie is ignored.
	* output size in zc_objset_type.
	* zc_flags lzc_send_flags
	*
	* outputs:
	* zc_objset_type estimated size, if zc_guid is set
	*
	* NOTE: This is no longer the preferred interface, any new functionality
	* should be added to zfs_ioc_send_new() instead.
	*/
	static int
	zfs_ioc_send(zfs_cmd_t *zc)
	{
	int error;
	offset_t off;
	boolean_t estimate = (zc->zc_guid != 0);
	boolean_t embedok = (zc->zc_flags & 0x1);
	boolean_t large_block_ok = (zc->zc_flags & 0x2);
	boolean_t compressok = (zc->zc_flags & 0x4);
	boolean_t rawok = (zc->zc_flags & 0x8);
	boolean_t savedok = (zc->zc_flags & 0x10);

	if (zc->zc_obj != 0) {
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (dsl_dir_is_clone(tosnap->ds_dir))
	zc->zc_fromobj =
	dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	}

	if (estimate) {
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;
	dsl_dataset_t *fromsnap = NULL;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
	FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (zc->zc_fromobj != 0) {
	error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
	FTAG, &fromsnap);
	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	}

	error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
	compressok \|\| rawok, savedok, &zc->zc_objset_type);

	if (fromsnap != NULL)
	dsl_dataset_rele(fromsnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	} else {
	zfs_file_t *fp;
	dmu_send_outparams_t out = {0};

	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
	return (SET_ERROR(EBADF));

	off = zfs_file_off(fp);
	out.dso_outfunc = dump_bytes;
	out.dso_arg = fp;
	out.dso_dryrun = B_FALSE;
	error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
	zc->zc_fromobj, embedok, large_block_ok, compressok,
	rawok, savedok, zc->zc_cookie, &off, &out);

	zfs_file_put(fp);
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of snapshot on which to report progress
	* zc_cookie file descriptor of send stream
	*
	* outputs:
	* zc_cookie number of bytes written in send stream thus far
	* zc_objset_type logical size of data traversed by send thus far
	*/
	static int
	zfs_ioc_send_progress(zfs_cmd_t *zc)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	dmu_sendstatus_t *dsp = NULL;
	int error;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	mutex_enter(&ds->ds_sendstream_lock);

	/*
	* Iterate over all the send streams currently active on this dataset.
	* If there's one which matches the specified file descriptor _and_ the
	* stream was started by the current process, return the progress of
	* that stream.
	*/

	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
	dsp = list_next(&ds->ds_sendstreams, dsp)) {
	if (dsp->dss_outfd == zc->zc_cookie &&
	zfs_proc_is_caller(dsp->dss_proc))
	break;
	}

	if (dsp != NULL) {
	zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
	0, 0);
	/* This is the closest thing we have to atomic_read_64. */
	zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
	} else {
	error = SET_ERROR(ENOENT);
	}

	mutex_exit(&ds->ds_sendstream_lock);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	static int
	zfs_ioc_inject_fault(zfs_cmd_t *zc)
	{
	int id, error;

	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
	&zc->zc_inject_record);

	if (error == 0)
	zc->zc_guid = (uint64_t)id;

	return (error);
	}

	static int
	zfs_ioc_clear_fault(zfs_cmd_t *zc)
	{
	return (zio_clear_fault((int)zc->zc_guid));
	}

	static int
	zfs_ioc_inject_list_next(zfs_cmd_t *zc)
	{
	int id = (int)zc->zc_guid;
	int error;

	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
	&zc->zc_inject_record);

	zc->zc_guid = id;

	return (error);
	}

	static int
	zfs_ioc_error_log(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	uint64_t count = zc->zc_nvlist_dst_size;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
	&count);
	if (error == 0)
	zc->zc_nvlist_dst_size = count;
	else
	zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);

	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_clear(zfs_cmd_t *zc)
	{
	spa_t *spa;
	vdev_t *vd;
	int error;

	/*
	* On zpool clear we also fix up missing slogs
	*/
	mutex_enter(&spa_namespace_lock);
	spa = spa_lookup(zc->zc_name);
	if (spa == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EIO));
	}
	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
	/* we need to let spa_open/spa_load clear the chains */
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	}
	spa->spa_last_open_failed = 0;
	mutex_exit(&spa_namespace_lock);

	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
	error = spa_open(zc->zc_name, &spa, FTAG);
	} else {
	nvlist_t *policy;
	nvlist_t *config = NULL;

	if (zc->zc_nvlist_src == 0)
	return (SET_ERROR(EINVAL));

	if ((error = get_nvlist(zc->zc_nvlist_src,
	zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
	error = spa_open_rewind(zc->zc_name, &spa, FTAG,
	policy, &config);
	if (config != NULL) {
	int err;

	if ((err = put_nvlist(zc, config)) != 0)
	error = err;
	nvlist_free(config);
	}
	nvlist_free(policy);
	}
	}

	if (error != 0)
	return (error);

	/*
	* If multihost is enabled, resuming I/O is unsafe as another
	* host may have imported the pool.
	*/
	if (spa_multihost(spa) && spa_suspended(spa))
	return (SET_ERROR(EINVAL));

	spa_vdev_state_enter(spa, SCL_NONE);

	if (zc->zc_guid == 0) {
	vd = NULL;
	} else {
	vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
	if (vd == NULL) {
	error = SET_ERROR(ENODEV);
	(void) spa_vdev_state_exit(spa, NULL, error);
	spa_close(spa, FTAG);
	return (error);
	}
	}

	vdev_clear(spa, vd);

	(void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
	NULL : spa->spa_root_vdev, 0);

	/*
	* Resume any suspended I/Os.
	*/
	if (zio_resume(spa) != 0)
	error = SET_ERROR(EIO);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* Reopen all the vdevs associated with the pool.
	*
	* innvl: {
	* "scrub_restart" -> when true and scrub is running, allow to restart
	* scrub as the side effect of the reopen (boolean).
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
	{"scrub_restart", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_pool_reopen(const char pool, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	spa_t *spa;
	int error;
	boolean_t rc, scrub_restart = B_TRUE;

	if (innvl) {
	error = nvlist_lookup_boolean_value(innvl,
	"scrub_restart", &rc);
	if (error == 0)
	scrub_restart = rc;
	}

	error = spa_open(pool, &spa, FTAG);
	if (error != 0)
	return (error);

	spa_vdev_state_enter(spa, SCL_NONE);

	/*
	* If the scrub_restart flag is B_FALSE and a scrub is already
	* in progress then set spa_scrub_reopen flag to B_TRUE so that
	* we don't restart the scrub as a side effect of the reopen.
	* Otherwise, let vdev_open() decided if a resilver is required.
	*/

	spa->spa_scrub_reopen = (!scrub_restart &&
	dsl_scan_scrubbing(spa->spa_dsl_pool));
	vdev_reopen(spa->spa_root_vdev);
	spa->spa_scrub_reopen = B_FALSE;

	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (0);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* zc_string name of conflicting snapshot, if there is one
	*/
	static int
	zfs_ioc_promote(zfs_cmd_t *zc)
	{
	dsl_pool_t *dp;
	dsl_dataset_t ds, ods;
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	char *cp;
	int error;

	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 \|\|
	strchr(zc->zc_name, '%'))
	return (SET_ERROR(EINVAL));

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (!dsl_dir_is_clone(ds->ds_dir)) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = dsl_dataset_hold_obj(dp,
	dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dsl_dataset_name(ods, origin);
	dsl_dataset_rele(ods, FTAG);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	/*
	* We don't need to unmount all the origin fs's snapshots, but
	* it's easier.
	*/
	cp = strchr(origin, '@');
	if (cp)
	*cp = '\0';
	(void) dmu_objset_find(origin,
	zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
	}

	/*
	* Retrieve a single {user\|group\|project}{used\|quota}@... property.
	*
	* inputs:
	* zc_name name of filesystem
	* zc_objset_type zfs_userquota_prop_t
	* zc_value domain name (eg. "S-1-234-567-89")
	* zc_guid RID/UID/GID
	*
	* outputs:
	* zc_cookie property value
	*/
	static int
	zfs_ioc_userspace_one(zfs_cmd_t *zc)
	{
	zfsvfs_t *zfsvfs;
	int error;

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
	if (error != 0)
	return (error);

	error = zfs_userspace_one(zfsvfs,
	zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
	zfsvfs_rele(zfsvfs, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_objset_type zfs_userquota_prop_t
	* zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
	*
	* outputs:
	* zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
	* zc_cookie zap cursor
	*/
	static int
	zfs_ioc_userspace_many(zfs_cmd_t *zc)
	{
	zfsvfs_t *zfsvfs;
	int bufsize = zc->zc_nvlist_dst_size;

	if (bufsize <= 0)
	return (SET_ERROR(ENOMEM));

	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
	if (error != 0)
	return (error);

	void *buf = vmem_alloc(bufsize, KM_SLEEP);

	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
	buf, &zc->zc_nvlist_dst_size);

	if (error == 0) {
	error = xcopyout(buf,
	(void *)(uintptr_t)zc->zc_nvlist_dst,
	zc->zc_nvlist_dst_size);
	}
	vmem_free(buf, bufsize);
	zfsvfs_rele(zfsvfs, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* none
	*/
	static int
	zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
	{
	int error = 0;
	zfsvfs_t *zfsvfs;

	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
	if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
	/*
	* If userused is not enabled, it may be because the
	* objset needs to be closed & reopened (to grow the
	* objset_phys_t). Suspend/resume the fs will do that.
	*/
	dsl_dataset_t ds, newds;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	if (error == 0) {
	dmu_objset_refresh_ownership(ds, &newds,
	B_TRUE, zfsvfs);
	error = zfs_resume_fs(zfsvfs, newds);
	}
	}
	if (error == 0) {
	mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
	if (zfsvfs->z_os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	zfsvfs->z_os->os_upgrade_status = 0;
	mutex_exit(&zfsvfs->z_os->os_upgrade_lock);

	dsl_pool_config_enter(
	dmu_objset_pool(zfsvfs->z_os), FTAG);
	dmu_objset_userspace_upgrade(zfsvfs->z_os);
	dsl_pool_config_exit(
	dmu_objset_pool(zfsvfs->z_os), FTAG);
	} else {
	mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
	}

	taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
	zfsvfs->z_os->os_upgrade_id);
	error = zfsvfs->z_os->os_upgrade_status;
	}
	zfs_vfs_rele(zfsvfs);
	} else {
	objset_t *os;

	/* XXX kind of reading contents without owning */
	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
	if (error != 0)
	return (error);

	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	os->os_upgrade_status = 0;
	mutex_exit(&os->os_upgrade_lock);

	dmu_objset_userspace_upgrade(os);
	} else {
	mutex_exit(&os->os_upgrade_lock);
	}

	dsl_pool_rele(dmu_objset_pool(os), FTAG);

	taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
	error = os->os_upgrade_status;

	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
	FTAG);
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* none
	*/
	static int
	zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
	if (error != 0)
	return (error);

	if (dmu_objset_userobjspace_upgradable(os) \|\|
	dmu_objset_projectquota_upgradable(os)) {
	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	os->os_upgrade_status = 0;
	mutex_exit(&os->os_upgrade_lock);

	dmu_objset_id_quota_upgrade(os);
	} else {
	mutex_exit(&os->os_upgrade_lock);
	}

	dsl_pool_rele(dmu_objset_pool(os), FTAG);

	taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
	error = os->os_upgrade_status;
	} else {
	dsl_pool_rele(dmu_objset_pool(os), FTAG);
	}

	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);

	return (error);
	}

	static int
	zfs_ioc_share(zfs_cmd_t *zc)
	{
	return (SET_ERROR(ENOSYS));
	}

	/*
	* inputs:
	* zc_name name of containing filesystem
	* zc_obj object # beyond which we want next in-use object #
	*
	* outputs:
	* zc_obj next in-use object #
	*/
	static int
	zfs_ioc_next_obj(zfs_cmd_t *zc)
	{
	objset_t *os = NULL;
	int error;

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error != 0)
	return (error);

	error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);

	dmu_objset_rele(os, FTAG);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value prefix name for snapshot
	* zc_cleanup_fd cleanup-on-exit file descriptor for calling process
	*
	* outputs:
	* zc_value short name of new snapshot
	*/
	static int
	zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
	{
	char *snap_name;
	char *hold_name;
	minor_t minor;

	zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
	(u_longlong_t)ddi_get_lbolt64());
	hold_name = kmem_asprintf("%%%s", zc->zc_value);

	int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
	hold_name);
	if (error == 0)
	(void) strlcpy(zc->zc_value, snap_name,
	sizeof (zc->zc_value));
	kmem_strfree(snap_name);
	kmem_strfree(hold_name);
	zfs_onexit_fd_rele(fp);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of "to" snapshot
	* zc_value name of "from" snapshot
	* zc_cookie file descriptor to write diff data on
	*
	* outputs:
	* dmu_diff_record_t's to the file descriptor
	*/
	static int
	zfs_ioc_diff(zfs_cmd_t *zc)
	{
	zfs_file_t *fp;
	offset_t off;
	int error;

	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
	return (SET_ERROR(EBADF));

	off = zfs_file_off(fp);
	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);

	zfs_file_put(fp);

	return (error);
	}

	static int
	zfs_ioc_smb_acl(zfs_cmd_t *zc)
	{
	return (SET_ERROR(ENOTSUP));
	}

	/*
	* innvl: {
	* "holds" -> { snapname -> holdname (string), ... }
	* (optional) "cleanup_fd" -> fd (int32)
	* }
	*
	* outnvl: {
	* snapname -> error value (int32)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_hold[] = {
	{"holds", DATA_TYPE_NVLIST, 0},
	{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_hold(const char pool, nvlist_t args, nvlist_t *errlist)
	{
	(void) pool;
	nvpair_t *pair;
	nvlist_t *holds;
	int cleanup_fd = -1;
	int error;
	minor_t minor = 0;
	zfs_file_t *fp = NULL;

	holds = fnvlist_lookup_nvlist(args, "holds");

	/* make sure the user didn't pass us any invalid (empty) tags */
	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
	pair = nvlist_next_nvpair(holds, pair)) {
	char *htag;

	error = nvpair_value_string(pair, &htag);
	if (error != 0)
	return (SET_ERROR(error));

	if (strlen(htag) == 0)
	return (SET_ERROR(EINVAL));
	}

	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
	fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
	if (fp == NULL)
	return (SET_ERROR(EBADF));
	}

	error = dsl_dataset_user_hold(holds, minor, errlist);
	if (fp != NULL) {
	ASSERT3U(minor, !=, 0);
	zfs_onexit_fd_rele(fp);
	}
	return (SET_ERROR(error));
	}

	/*
	* innvl is not used.
	*
	* outnvl: {
	* holdname -> time added (uint64 seconds since epoch)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_get_holds[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_holds(const char snapname, nvlist_t args, nvlist_t *outnvl)
	{
	(void) args;
	return (dsl_dataset_get_holds(snapname, outnvl));
	}

	/*
	* innvl: {
	* snapname -> { holdname, ... }
	* ...
	* }
	*
	* outnvl: {
	* snapname -> error value (int32)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_release[] = {
	{"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_release(const char pool, nvlist_t holds, nvlist_t *errlist)
	{
	(void) pool;
	return (dsl_dataset_user_release(holds, errlist));
	}

	/*
	* inputs:
	* zc_guid flags (ZEVENT_NONBLOCK)
	* zc_cleanup_fd zevent file descriptor
	*
	* outputs:
	* zc_nvlist_dst next nvlist event
	* zc_cookie dropped events since last get
	*/
	static int
	zfs_ioc_events_next(zfs_cmd_t *zc)
	{
	zfs_zevent_t *ze;
	nvlist_t *event = NULL;
	minor_t minor;
	uint64_t dropped = 0;
	int error;

	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	do {
	error = zfs_zevent_next(ze, &event,
	&zc->zc_nvlist_dst_size, &dropped);
	if (event != NULL) {
	zc->zc_cookie = dropped;
	error = put_nvlist(zc, event);
	nvlist_free(event);
	}

	if (zc->zc_guid & ZEVENT_NONBLOCK)
	break;

	if ((error == 0) \|\| (error != ENOENT))
	break;

	error = zfs_zevent_wait(ze);
	if (error != 0)
	break;
	} while (1);

	zfs_zevent_fd_rele(fp);

	return (error);
	}

	/*
	* outputs:
	* zc_cookie cleared events count
	*/
	static int
	zfs_ioc_events_clear(zfs_cmd_t *zc)
	{
	- int count;
	+ uint_t count;

	zfs_zevent_drain_all(&count);
	zc->zc_cookie = count;

	return (0);
	}

	/*
	* inputs:
	* zc_guid eid \| ZEVENT_SEEK_START \| ZEVENT_SEEK_END
	* zc_cleanup zevent file descriptor
	*/
	static int
	zfs_ioc_events_seek(zfs_cmd_t *zc)
	{
	zfs_zevent_t *ze;
	minor_t minor;
	int error;

	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	error = zfs_zevent_seek(ze, zc->zc_guid);
	zfs_zevent_fd_rele(fp);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of later filesystem or snapshot
	* zc_value full name of old snapshot or bookmark
	*
	* outputs:
	* zc_cookie space in bytes
	* zc_objset_type compressed space in bytes
	* zc_perm_action uncompressed space in bytes
	*/
	static int
	zfs_ioc_space_written(zfs_cmd_t *zc)
	{
	int error;
	dsl_pool_t *dp;
	dsl_dataset_t *new;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);
	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	if (strchr(zc->zc_value, '#') != NULL) {
	zfs_bookmark_phys_t bmp;
	error = dsl_bookmark_lookup(dp, zc->zc_value,
	new, &bmp);
	if (error == 0) {
	error = dsl_dataset_space_written_bookmark(&bmp, new,
	&zc->zc_cookie,
	&zc->zc_objset_type, &zc->zc_perm_action);
	}
	} else {
	dsl_dataset_t *old;
	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);

	if (error == 0) {
	error = dsl_dataset_space_written(old, new,
	&zc->zc_cookie,
	&zc->zc_objset_type, &zc->zc_perm_action);
	dsl_dataset_rele(old, FTAG);
	}
	}
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	/*
	* innvl: {
	* "firstsnap" -> snapshot name
	* }
	*
	* outnvl: {
	* "used" -> space in bytes
	* "compressed" -> compressed space in bytes
	* "uncompressed" -> uncompressed space in bytes
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
	{"firstsnap", DATA_TYPE_STRING, 0},
	};

	static int
	zfs_ioc_space_snaps(const char lastsnap, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error;
	dsl_pool_t *dp;
	dsl_dataset_t new, old;
	char *firstsnap;
	uint64_t used, comp, uncomp;

	firstsnap = fnvlist_lookup_string(innvl, "firstsnap");

	error = dsl_pool_hold(lastsnap, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
	if (error == 0 && !new->ds_is_snapshot) {
	dsl_dataset_rele(new, FTAG);
	error = SET_ERROR(EINVAL);
	}
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
	if (error == 0 && !old->ds_is_snapshot) {
	dsl_dataset_rele(old, FTAG);
	error = SET_ERROR(EINVAL);
	}
	if (error != 0) {
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
	dsl_dataset_rele(old, FTAG);
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	fnvlist_add_uint64(outnvl, "used", used);
	fnvlist_add_uint64(outnvl, "compressed", comp);
	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
	return (error);
	}

	/*
	* innvl: {
	* "fd" -> file descriptor to write stream to (int32)
	* (optional) "fromsnap" -> full snap name to send an incremental from
	* (optional) "largeblockok" -> (value ignored)
	* indicates that blocks > 128KB are permitted
	* (optional) "embedok" -> (value ignored)
	* presence indicates DRR_WRITE_EMBEDDED records are permitted
	* (optional) "compressok" -> (value ignored)
	* presence indicates compressed DRR_WRITE records are permitted
	* (optional) "rawok" -> (value ignored)
	* presence indicates raw encrypted records should be used.
	* (optional) "savedok" -> (value ignored)
	* presence indicates we should send a partially received snapshot
	* (optional) "resume_object" and "resume_offset" -> (uint64)
	* if present, resume send stream from specified object and offset.
	* (optional) "redactbook" -> (string)
	* if present, use this bookmark's redaction list to generate a redacted
	* send stream
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_send_new[] = {
	{"fd", DATA_TYPE_INT32, 0},
	{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"savedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_send_new(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int error;
	offset_t off;
	char *fromname = NULL;
	int fd;
	zfs_file_t *fp;
	boolean_t largeblockok;
	boolean_t embedok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t resumeobj = 0;
	uint64_t resumeoff = 0;
	char *redactbook = NULL;

	fd = fnvlist_lookup_int32(innvl, "fd");

	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);

	largeblockok = nvlist_exists(innvl, "largeblockok");
	embedok = nvlist_exists(innvl, "embedok");
	compressok = nvlist_exists(innvl, "compressok");
	rawok = nvlist_exists(innvl, "rawok");
	savedok = nvlist_exists(innvl, "savedok");

	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);

	(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);

	if ((fp = zfs_file_get(fd)) == NULL)
	return (SET_ERROR(EBADF));

	off = zfs_file_off(fp);

	dmu_send_outparams_t out = {0};
	out.dso_outfunc = dump_bytes;
	out.dso_arg = fp;
	out.dso_dryrun = B_FALSE;
	error = dmu_send(snapname, fromname, embedok, largeblockok,
	compressok, rawok, savedok, resumeobj, resumeoff,
	redactbook, fd, &off, &out);

	zfs_file_put(fp);
	return (error);
	}

	static int
	send_space_sum(objset_t os, void buf, int len, void *arg)
	{
	(void) os, (void) buf;
	uint64_t *size = arg;

	*size += len;
	return (0);
	}

	/*
	* Determine approximately how large a zfs send stream will be -- the number
	* of bytes that will be written to the fd supplied to zfs_ioc_send_new().
	*
	* innvl: {
	* (optional) "from" -> full snap or bookmark name to send an incremental
	* from
	* (optional) "largeblockok" -> (value ignored)
	* indicates that blocks > 128KB are permitted
	* (optional) "embedok" -> (value ignored)
	* presence indicates DRR_WRITE_EMBEDDED records are permitted
	* (optional) "compressok" -> (value ignored)
	* presence indicates compressed DRR_WRITE records are permitted
	* (optional) "rawok" -> (value ignored)
	* presence indicates raw encrypted records should be used.
	* (optional) "resume_object" and "resume_offset" -> (uint64)
	* if present, resume send stream from specified object and offset.
	* (optional) "fd" -> file descriptor to use as a cookie for progress
	* tracking (int32)
	* }
	*
	* outnvl: {
	* "space" -> bytes of space (uint64)
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_send_space[] = {
	{"from", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"bytes", DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_send_space(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;
	dsl_dataset_t *fromsnap = NULL;
	int error;
	char *fromname = NULL;
	char *redactlist_book = NULL;
	boolean_t largeblockok;
	boolean_t embedok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t space = 0;
	boolean_t full_estimate = B_FALSE;
	uint64_t resumeobj = 0;
	uint64_t resumeoff = 0;
	uint64_t resume_bytes = 0;
	int32_t fd = -1;
	zfs_bookmark_phys_t zbm = {0};

	error = dsl_pool_hold(snapname, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	(void) nvlist_lookup_int32(innvl, "fd", &fd);

	largeblockok = nvlist_exists(innvl, "largeblockok");
	embedok = nvlist_exists(innvl, "embedok");
	compressok = nvlist_exists(innvl, "compressok");
	rawok = nvlist_exists(innvl, "rawok");
	savedok = nvlist_exists(innvl, "savedok");
	boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
	boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
	&redactlist_book) == 0);

	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
	(void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);

	if (altbook) {
	full_estimate = B_TRUE;
	} else if (from) {
	if (strchr(fromname, '#')) {
	error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);

	/*
	* dsl_bookmark_lookup() will fail with EXDEV if
	* the from-bookmark and tosnap are at the same txg.
	* However, it's valid to do a send (and therefore,
	* a send estimate) from and to the same time point,
	* if the bookmark is redacted (the incremental send
	* can change what's redacted on the target). In
	* this case, dsl_bookmark_lookup() fills in zbm
	* but returns EXDEV. Ignore this error.
	*/
	if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
	zbm.zbm_guid ==
	dsl_dataset_phys(tosnap)->ds_guid)
	error = 0;

	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	if (zbm.zbm_redaction_obj != 0 \|\| !(zbm.zbm_flags &
	ZBM_FLAG_HAS_FBN)) {
	full_estimate = B_TRUE;
	}
	} else if (strchr(fromname, '@')) {
	error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
	full_estimate = B_TRUE;
	dsl_dataset_rele(fromsnap, FTAG);
	}
	} else {
	/*
	* from is not properly formatted as a snapshot or
	* bookmark
	*/
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	if (full_estimate) {
	dmu_send_outparams_t out = {0};
	offset_t off = 0;
	out.dso_outfunc = send_space_sum;
	out.dso_arg = &space;
	out.dso_dryrun = B_TRUE;
	/*
	* We have to release these holds so dmu_send can take them. It
	* will do all the error checking we need.
	*/
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	error = dmu_send(snapname, fromname, embedok, largeblockok,
	compressok, rawok, savedok, resumeobj, resumeoff,
	redactlist_book, fd, &off, &out);
	} else {
	error = dmu_send_estimate_fast(tosnap, fromsnap,
	(from && strchr(fromname, '#') != NULL ? &zbm : NULL),
	compressok \|\| rawok, savedok, &space);
	space -= resume_bytes;
	if (fromsnap != NULL)
	dsl_dataset_rele(fromsnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	}

	fnvlist_add_uint64(outnvl, "space", space);

	return (error);
	}

	/*
	* Sync the currently open TXG to disk for the specified pool.
	* This is somewhat similar to 'zfs_sync()'.
	* For cases that do not result in error this ioctl will wait for
	* the currently open TXG to commit before returning back to the caller.
	*
	* innvl: {
	* "force" -> when true, force uberblock update even if there is no dirty data.
	* In addition this will cause the vdev configuration to be written
	* out including updating the zpool cache file. (boolean_t)
	* }
	*
	* onvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
	{"force", DATA_TYPE_BOOLEAN_VALUE, 0},
	};

	static int
	zfs_ioc_pool_sync(const char pool, nvlist_t innvl, nvlist_t *onvl)
	{
	(void) onvl;
	int err;
	boolean_t rc, force = B_FALSE;
	spa_t *spa;

	if ((err = spa_open(pool, &spa, FTAG)) != 0)
	return (err);

	if (innvl) {
	err = nvlist_lookup_boolean_value(innvl, "force", &rc);
	if (err == 0)
	force = rc;
	}

	if (force) {
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
	vdev_config_dirty(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}
	txg_wait_synced(spa_get_dsl(spa), 0);

	spa_close(spa, FTAG);

	return (0);
	}

	/*
	* Load a user's wrapping key into the kernel.
	* innvl: {
	* "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* (optional) "noop" -> (value ignored)
	* presence indicated key should only be verified, not loaded
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_load_key[] = {
	{"hidden_args", DATA_TYPE_NVLIST, 0},
	{"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_load_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int ret;
	dsl_crypto_params_t *dcp = NULL;
	nvlist_t *hidden_args;
	boolean_t noop = nvlist_exists(innvl, "noop");

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = SET_ERROR(EINVAL);
	goto error;
	}

	hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);

	ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
	hidden_args, &dcp);
	if (ret != 0)
	goto error;

	ret = spa_keystore_load_wkey(dsname, dcp, noop);
	if (ret != 0)
	goto error;

	dsl_crypto_params_free(dcp, noop);

	return (0);

	error:
	dsl_crypto_params_free(dcp, B_TRUE);
	return (ret);
	}

	/*
	* Unload a user's wrapping key from the kernel.
	* Both innvl and outnvl are unused.
	*/
	static const zfs_ioc_key_t zfs_keys_unload_key[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_unload_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	int ret = 0;

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = (SET_ERROR(EINVAL));
	goto out;
	}

	ret = spa_keystore_unload_wkey(dsname);
	if (ret != 0)
	goto out;

	out:
	return (ret);
	}

	/*
	* Changes a user's wrapping key used to decrypt a dataset. The keyformat,
	* keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
	* here to change how the key is derived in userspace.
	*
	* innvl: {
	* "hidden_args" (optional) -> { "wkeydata" -> value }
	* raw uint8_t array of new encryption wrapping key data (32 bytes)
	* "props" (optional) -> { prop -> value }
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_change_key[] = {
	{"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_change_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int ret;
	uint64_t cmd = DCP_CMD_NONE;
	dsl_crypto_params_t *dcp = NULL;
	nvlist_t args = NULL, hidden_args = NULL;

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = (SET_ERROR(EINVAL));
	goto error;
	}

	(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
	(void) nvlist_lookup_nvlist(innvl, "props", &args);
	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);

	ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
	if (ret != 0)
	goto error;

	ret = spa_keystore_change_key(dsname, dcp);
	if (ret != 0)
	goto error;

	dsl_crypto_params_free(dcp, B_FALSE);

	return (0);

	error:
	dsl_crypto_params_free(dcp, B_TRUE);
	return (ret);
	}

	static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];

	static void
	zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
	boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];

	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
	ASSERT3U(ioc, <, ZFS_IOC_LAST);
	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
	ASSERT3P(vec->zvec_func, ==, NULL);

	vec->zvec_legacy_func = func;
	vec->zvec_secpolicy = secpolicy;
	vec->zvec_namecheck = namecheck;
	vec->zvec_allow_log = log_history;
	vec->zvec_pool_check = pool_check;
	}

	/*
	* See the block comment at the beginning of this file for details on
	* each argument to this function.
	*/
	void
	zfs_ioctl_register(const char name, zfs_ioc_t ioc, zfs_ioc_func_t func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
	zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
	boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
	{
	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];

	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
	ASSERT3U(ioc, <, ZFS_IOC_LAST);
	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
	ASSERT3P(vec->zvec_func, ==, NULL);

	/* if we are logging, the name must be valid */
	ASSERT(!allow_log \|\| namecheck != NO_NAME);

	vec->zvec_name = name;
	vec->zvec_func = func;
	vec->zvec_secpolicy = secpolicy;
	vec->zvec_namecheck = namecheck;
	vec->zvec_pool_check = pool_check;
	vec->zvec_smush_outnvlist = smush_outnvlist;
	vec->zvec_allow_log = allow_log;
	vec->zvec_nvl_keys = nvl_keys;
	vec->zvec_nvl_key_count = num_keys;
	}

	static void
	zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
	zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	POOL_NAME, log_history, pool_check);
	}

	void
	zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_FALSE, pool_check);
	}

	static void
	zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
	{
	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
	POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	}

	static void
	zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	NO_NAME, B_FALSE, POOL_CHECK_NONE);
	}

	static void
	zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
	zfs_ioc_legacy_func_t func, zfs_secpolicy_func_t secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
	}

	static void
	zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
	{
	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
	zfs_secpolicy_read);
	}

	static void
	zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	}

	static void
	zfs_ioctl_init(void)
	{
	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
	zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));

	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
	zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));

	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
	zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));

	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
	zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));

	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
	zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));

	zfs_ioctl_register("create", ZFS_IOC_CREATE,
	zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_create, ARRAY_SIZE(zfs_keys_create));

	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
	zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));

	zfs_ioctl_register("remap", ZFS_IOC_REMAP,
	zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));

	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
	zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));

	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
	zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
	zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_release, ARRAY_SIZE(zfs_keys_release));

	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
	zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));

	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
	zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));

	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
	zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));

	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
	zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));

	zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
	zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
	ARRAY_SIZE(zfs_keys_get_bookmark_props));

	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
	zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
	POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_destroy_bookmarks,
	ARRAY_SIZE(zfs_keys_destroy_bookmarks));

	zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
	zfs_ioc_recv_new, zfs_secpolicy_recv, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
	zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
	zfs_ioc_load_key, zfs_secpolicy_load_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
	zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
	zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
	zfs_ioc_unload_key, zfs_secpolicy_load_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
	zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
	zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
	zfs_ioc_change_key, zfs_secpolicy_change_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY,
	B_TRUE, B_TRUE, zfs_keys_change_key,
	ARRAY_SIZE(zfs_keys_change_key));

	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
	zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
	zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
	zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
	B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));

	zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
	zfs_ioc_channel_program, zfs_secpolicy_config,
	POOL_NAME, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE,
	B_TRUE, zfs_keys_channel_program,
	ARRAY_SIZE(zfs_keys_channel_program));

	zfs_ioctl_register("redact", ZFS_IOC_REDACT,
	zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));

	zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
	zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));

	zfs_ioctl_register("zpool_discard_checkpoint",
	ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
	zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_discard_checkpoint,
	ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));

	zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
	zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));

	zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
	zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));

	zfs_ioctl_register("wait", ZFS_IOC_WAIT,
	zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));

	zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
	zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));

	zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
	zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));

	zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
	zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
	zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));

	zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
	zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
	POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
	ARRAY_SIZE(zfs_keys_vdev_get_props));

	zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
	zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));

	/* IOCTLS that use the legacy function signature */

	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
	zfs_ioc_pool_scan);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
	zfs_ioc_pool_upgrade);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
	zfs_ioc_vdev_add);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
	zfs_ioc_vdev_remove);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
	zfs_ioc_vdev_set_state);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
	zfs_ioc_vdev_attach);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
	zfs_ioc_vdev_detach);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
	zfs_ioc_vdev_setpath);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
	zfs_ioc_vdev_setfru);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
	zfs_ioc_pool_set_props);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
	zfs_ioc_vdev_split);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
	zfs_ioc_pool_reguid);

	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
	zfs_ioc_pool_configs, zfs_secpolicy_none);
	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
	zfs_ioc_pool_tryimport, zfs_secpolicy_config);
	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
	zfs_ioc_inject_fault, zfs_secpolicy_inject);
	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
	zfs_ioc_clear_fault, zfs_secpolicy_inject);
	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
	zfs_ioc_inject_list_next, zfs_secpolicy_inject);

	/*
	* pool destroy, and export don't log the history as part of
	* zfsdev_ioctl, but rather zfs_ioc_pool_export
	* does the logging of those commands.
	*/
	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
	zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
	zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);

	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
	zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
	zfs_ioc_dsobj_to_dsname,
	zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
	zfs_ioc_pool_get_history,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);

	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);

	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
	zfs_ioc_space_written);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
	zfs_ioc_objset_recvd_props);
	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
	zfs_ioc_next_obj);
	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
	zfs_ioc_get_fsacl);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
	zfs_ioc_objset_stats);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
	zfs_ioc_objset_zplprops);
	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
	zfs_ioc_dataset_list_next);
	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
	zfs_ioc_snapshot_list_next);
	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
	zfs_ioc_send_progress);

	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
	zfs_ioc_diff, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
	zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
	zfs_ioc_obj_to_path, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
	zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
	zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
	zfs_ioc_send, zfs_secpolicy_send);

	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
	zfs_secpolicy_none);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
	zfs_secpolicy_destroy);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
	zfs_secpolicy_rename);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
	zfs_secpolicy_recv);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
	zfs_secpolicy_promote);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
	zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
	zfs_secpolicy_set_fsacl);

	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
	zfs_secpolicy_share, POOL_CHECK_NONE);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
	zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
	zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
	zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);

	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);

	zfs_ioctl_init_os();
	}

	/*
	* Verify that for non-legacy ioctls the input nvlist
	* pairs match against the expected input.
	*
	* Possible errors are:
	* ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered
	* ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing
	* ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair
	*/
	static int
	zfs_check_input_nvpairs(nvlist_t innvl, const zfs_ioc_vec_t vec)
	{
	const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
	boolean_t required_keys_found = B_FALSE;

	/*
	* examine each input pair
	*/
	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	char *name = nvpair_name(pair);
	data_type_t type = nvpair_type(pair);
	boolean_t identified = B_FALSE;

	/*
	* check pair against the documented names and type
	*/
	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
	/* if not a wild card name, check for an exact match */
	if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
	strcmp(nvl_keys[k].zkey_name, name) != 0)
	continue;

	identified = B_TRUE;

	if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
	nvl_keys[k].zkey_type != type) {
	return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
	}

	if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
	continue;

	required_keys_found = B_TRUE;
	break;
	}

	/* allow an 'optional' key, everything else is invalid */
	if (!identified &&
	(strcmp(name, "optional") != 0 \|\|
	type != DATA_TYPE_NVLIST)) {
	return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
	}
	}

	/* verify that all required keys were found */
	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
	if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
	continue;

	if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
	/* at least one non-optional key is expected here */
	if (!required_keys_found)
	return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
	continue;
	}

	if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
	return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
	}

	return (0);
	}

	static int
	pool_status_check(const char *name, zfs_ioc_namecheck_t type,
	zfs_ioc_poolcheck_t check)
	{
	spa_t *spa;
	int error;

	ASSERT(type == POOL_NAME \|\| type == DATASET_NAME \|\|
	type == ENTITY_NAME);

	if (check & POOL_CHECK_NONE)
	return (0);

	error = spa_open(name, &spa, FTAG);
	if (error == 0) {
	if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
	error = SET_ERROR(EAGAIN);
	else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
	error = SET_ERROR(EROFS);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	int
	zfsdev_getminor(zfs_file_t fp, minor_t minorp)
	{
	zfsdev_state_t zs, fpd;

	ASSERT(!MUTEX_HELD(&zfsdev_state_lock));

	fpd = zfs_file_private(fp);
	if (fpd == NULL)
	return (SET_ERROR(EBADF));

	mutex_enter(&zfsdev_state_lock);

	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {

	if (zs->zs_minor == -1)
	continue;

	if (fpd == zs) {
	*minorp = fpd->zs_minor;
	mutex_exit(&zfsdev_state_lock);
	return (0);
	}
	}

	mutex_exit(&zfsdev_state_lock);

	return (SET_ERROR(EBADF));
	}

	void *
	zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
	{
	zfsdev_state_t *zs;

	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
	if (zs->zs_minor == minor) {
	membar_consumer();
	switch (which) {
	case ZST_ONEXIT:
	return (zs->zs_onexit);
	case ZST_ZEVENT:
	return (zs->zs_zevent);
	case ZST_ALL:
	return (zs);
	}
	}
	}

	return (NULL);
	}

	/*
	* Find a free minor number. The zfsdev_state_list is expected to
	* be short since it is only a list of currently open file handles.
	*/
	static minor_t
	zfsdev_minor_alloc(void)
	{
	static minor_t last_minor = 0;
	minor_t m;

	ASSERT(MUTEX_HELD(&zfsdev_state_lock));

	for (m = last_minor + 1; m != last_minor; m++) {
	if (m > ZFSDEV_MAX_MINOR)
	m = 1;
	if (zfsdev_get_state(m, ZST_ALL) == NULL) {
	last_minor = m;
	return (m);
	}
	}

	return (0);
	}

	int
	zfsdev_state_init(void *priv)
	{
	zfsdev_state_t zs, zsprev = NULL;
	minor_t minor;
	boolean_t newzs = B_FALSE;

	ASSERT(MUTEX_HELD(&zfsdev_state_lock));

	minor = zfsdev_minor_alloc();
	if (minor == 0)
	return (SET_ERROR(ENXIO));

	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
	if (zs->zs_minor == -1)
	break;
	zsprev = zs;
	}

	if (!zs) {
	zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
	newzs = B_TRUE;
	}

	zfsdev_private_set_state(priv, zs);

	zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
	zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);

	/*
	* In order to provide for lock-free concurrent read access
	* to the minor list in zfsdev_get_state(), new entries
	* must be completely written before linking them into the
	* list whereas existing entries are already linked; the last
	* operation must be updating zs_minor (from -1 to the new
	* value).
	*/
	if (newzs) {
	zs->zs_minor = minor;
	membar_producer();
	zsprev->zs_next = zs;
	} else {
	membar_producer();
	zs->zs_minor = minor;
	}

	return (0);
	}

	void
	zfsdev_state_destroy(void *priv)
	{
	zfsdev_state_t *zs = zfsdev_private_get_state(priv);

	ASSERT(zs != NULL);
	ASSERT3S(zs->zs_minor, >, 0);

	/*
	* The last reference to this zfsdev file descriptor is being dropped.
	* We don't have to worry about lookup grabbing this state object, and
	* zfsdev_state_init() will not try to reuse this object until it is
	* invalidated by setting zs_minor to -1. Invalidation must be done
	* last, with a memory barrier to ensure ordering. This lets us avoid
	* taking the global zfsdev state lock around destruction.
	*/
	zfs_onexit_destroy(zs->zs_onexit);
	zfs_zevent_destroy(zs->zs_zevent);
	zs->zs_onexit = NULL;
	zs->zs_zevent = NULL;
	membar_producer();
	zs->zs_minor = -1;
	}

	long
	zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
	{
	int error, cmd;
	const zfs_ioc_vec_t *vec;
	char *saved_poolname = NULL;
	uint64_t max_nvlist_src_size;
	size_t saved_poolname_len = 0;
	nvlist_t *innvl = NULL;
	fstrans_cookie_t cookie;
	hrtime_t start_time = gethrtime();

	cmd = vecnum;
	error = 0;
	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
	return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));

	vec = &zfs_ioc_vec[vecnum];

	/*
	* The registered ioctl list may be sparse, verify that either
	* a normal or legacy handler are registered.
	*/
	if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
	return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));

	zc->zc_iflags = flag & FKIOCTL;
	max_nvlist_src_size = zfs_max_nvlist_src_size_os();
	if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
	/*
	* Make sure the user doesn't pass in an insane value for
	* zc_nvlist_src_size. We have to check, since we will end
	* up allocating that much memory inside of get_nvlist(). This
	* prevents a nefarious user from allocating tons of kernel
	* memory.
	*
	* Also, we return EINVAL instead of ENOMEM here. The reason
	* being that returning ENOMEM from an ioctl() has a special
	* connotation; that the user's size value is too small and
	* needs to be expanded to hold the nvlist. See
	* zcmd_expand_dst_nvlist() for details.
	*/
	error = SET_ERROR(EINVAL); /* User's size too big */

	} else if (zc->zc_nvlist_src_size != 0) {
	error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &innvl);
	if (error != 0)
	goto out;
	}

	/*
	* Ensure that all pool/dataset names are valid before we pass down to
	* the lower layers.
	*/
	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	switch (vec->zvec_namecheck) {
	case POOL_NAME:
	if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
	error = SET_ERROR(EINVAL);
	else
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	break;

	case DATASET_NAME:
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
	error = SET_ERROR(EINVAL);
	else
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	break;

	case ENTITY_NAME:
	if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
	error = SET_ERROR(EINVAL);
	} else {
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	}
	break;

	case NO_NAME:
	break;
	}
	/*
	* Ensure that all input pairs are valid before we pass them down
	* to the lower layers.
	*
	* The vectored functions can use fnvlist_lookup_{type} for any
	* required pairs since zfs_check_input_nvpairs() confirmed that
	* they exist and are of the correct type.
	*/
	if (error == 0 && vec->zvec_func != NULL) {
	error = zfs_check_input_nvpairs(innvl, vec);
	if (error != 0)
	goto out;
	}

	if (error == 0) {
	cookie = spl_fstrans_mark();
	error = vec->zvec_secpolicy(zc, innvl, CRED());
	spl_fstrans_unmark(cookie);
	}

	if (error != 0)
	goto out;

	/* legacy ioctls can modify zc_name */
	/*
	* Can't use kmem_strdup() as we might truncate the string and
	* kmem_strfree() would then free with incorrect size.
	*/
	saved_poolname_len = strlen(zc->zc_name) + 1;
	saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);

	strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
	saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';

	if (vec->zvec_func != NULL) {
	nvlist_t *outnvl;
	int puterror = 0;
	spa_t *spa;
	nvlist_t *lognv = NULL;

	ASSERT(vec->zvec_legacy_func == NULL);

	/*
	* Add the innvl to the lognv before calling the func,
	* in case the func changes the innvl.
	*/
	if (vec->zvec_allow_log) {
	lognv = fnvlist_alloc();
	fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
	vec->zvec_name);
	if (!nvlist_empty(innvl)) {
	fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
	innvl);
	}
	}

	outnvl = fnvlist_alloc();
	cookie = spl_fstrans_mark();
	error = vec->zvec_func(zc->zc_name, innvl, outnvl);
	spl_fstrans_unmark(cookie);

	/*
	* Some commands can partially execute, modify state, and still
	* return an error. In these cases, attempt to record what
	* was modified.
	*/
	if ((error == 0 \|\|
	(cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
	vec->zvec_allow_log &&
	spa_open(zc->zc_name, &spa, FTAG) == 0) {
	if (!nvlist_empty(outnvl)) {
	size_t out_size = fnvlist_size(outnvl);
	if (out_size > zfs_history_output_max) {
	fnvlist_add_int64(lognv,
	ZPOOL_HIST_OUTPUT_SIZE, out_size);
	} else {
	fnvlist_add_nvlist(lognv,
	ZPOOL_HIST_OUTPUT_NVL, outnvl);
	}
	}
	if (error != 0) {
	fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
	error);
	}
	fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
	gethrtime() - start_time);
	(void) spa_history_log_nvl(spa, lognv);
	spa_close(spa, FTAG);
	}
	fnvlist_free(lognv);

	if (!nvlist_empty(outnvl) \|\| zc->zc_nvlist_dst_size != 0) {
	int smusherror = 0;
	if (vec->zvec_smush_outnvlist) {
	smusherror = nvlist_smush(outnvl,
	zc->zc_nvlist_dst_size);
	}
	if (smusherror == 0)
	puterror = put_nvlist(zc, outnvl);
	}

	if (puterror != 0)
	error = puterror;

	nvlist_free(outnvl);
	} else {
	cookie = spl_fstrans_mark();
	error = vec->zvec_legacy_func(zc);
	spl_fstrans_unmark(cookie);
	}

	out:
	nvlist_free(innvl);
	if (error == 0 && vec->zvec_allow_log) {
	char *s = tsd_get(zfs_allow_log_key);
	if (s != NULL)
	kmem_strfree(s);
	(void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
	}
	if (saved_poolname != NULL)
	kmem_free(saved_poolname, saved_poolname_len);

	return (error);
	}

	int
	zfs_kmod_init(void)
	{
	int error;

	if ((error = zvol_init()) != 0)
	return (error);

	spa_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	zfs_init();

	zfs_ioctl_init();

	mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
	zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
	zfsdev_state_list->zs_minor = -1;

	if ((error = zfsdev_attach()) != 0)
	goto out;

	tsd_create(&zfs_fsyncer_key, NULL);
	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);

	return (0);
	out:
	zfs_fini();
	spa_fini();
	zvol_fini();

	return (error);
	}

	void
	zfs_kmod_fini(void)
	{
	zfsdev_state_t zs, zsnext = NULL;

	zfsdev_detach();

	mutex_destroy(&zfsdev_state_lock);

	for (zs = zfsdev_state_list; zs != NULL; zs = zsnext) {
	zsnext = zs->zs_next;
	if (zs->zs_onexit)
	zfs_onexit_destroy(zs->zs_onexit);
	if (zs->zs_zevent)
	zfs_zevent_destroy(zs->zs_zevent);
	kmem_free(zs, sizeof (zfsdev_state_t));
	}

	zfs_ereport_taskq_fini(); /* run before zfs_fini() on Linux */
	zfs_fini();
	spa_fini();
	zvol_fini();

	tsd_destroy(&zfs_fsyncer_key);
	tsd_destroy(&rrw_tsd_key);
	tsd_destroy(&zfs_allow_log_key);
	}

	ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, ULONG, ZMOD_RW,
	"Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");

	ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, ULONG, ZMOD_RW,
	"Maximum size in bytes of ZFS ioctl output that will be logged");
	diff --git a/sys/contrib/openzfs/module/zfs/zil.c b/sys/contrib/openzfs/module/zfs/zil.c
	index 4864e0ccad53..dc5b8018e16e 100644
	--- a/sys/contrib/openzfs/module/zfs/zil.c
	+++ b/sys/contrib/openzfs/module/zfs/zil.c
	@@ -1,3952 +1,3952 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright (c) 2018 Datto Inc.
	*/

	/* Portions Copyright 2010 Robert Milkowski */

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/zap.h>
	#include <sys/arc.h>
	#include <sys/stat.h>
	#include <sys/zil.h>
	#include <sys/zil_impl.h>
	#include <sys/dsl_dataset.h>
	#include <sys/vdev_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_pool.h>
	#include <sys/metaslab.h>
	#include <sys/trace_zfs.h>
	#include <sys/abd.h>
	#include <sys/wmsum.h>

	/*
	* The ZFS Intent Log (ZIL) saves "transaction records" (itxs) of system
	* calls that change the file system. Each itx has enough information to
	* be able to replay them after a system crash, power loss, or
	* equivalent failure mode. These are stored in memory until either:
	*
	* 1. they are committed to the pool by the DMU transaction group
	* (txg), at which point they can be discarded; or
	* 2. they are committed to the on-disk ZIL for the dataset being
	* modified (e.g. due to an fsync, O_DSYNC, or other synchronous
	* requirement).
	*
	* In the event of a crash or power loss, the itxs contained by each
	* dataset's on-disk ZIL will be replayed when that dataset is first
	* instantiated (e.g. if the dataset is a normal filesystem, when it is
	* first mounted).
	*
	* As hinted at above, there is one ZIL per dataset (both the in-memory
	* representation, and the on-disk representation). The on-disk format
	* consists of 3 parts:
	*
	* - a single, per-dataset, ZIL header; which points to a chain of
	* - zero or more ZIL blocks; each of which contains
	* - zero or more ZIL records
	*
	* A ZIL record holds the information necessary to replay a single
	* system call transaction. A ZIL block can hold many ZIL records, and
	* the blocks are chained together, similarly to a singly linked list.
	*
	* Each ZIL block contains a block pointer (blkptr_t) to the next ZIL
	* block in the chain, and the ZIL header points to the first block in
	* the chain.
	*
	* Note, there is not a fixed place in the pool to hold these ZIL
	* blocks; they are dynamically allocated and freed as needed from the
	* blocks available on the pool, though they can be preferentially
	* allocated from a dedicated "log" vdev.
	*/

	/*
	* This controls the amount of time that a ZIL block (lwb) will remain
	* "open" when it isn't "full", and it has a thread waiting for it to be
	* committed to stable storage. Please refer to the zil_commit_waiter()
	* function (and the comments within it) for more details.
	*/
	-static int zfs_commit_timeout_pct = 5;
	+static uint_t zfs_commit_timeout_pct = 5;

	/*
	* See zil.h for more information about these fields.
	*/
	static zil_kstat_values_t zil_stats = {
	{ "zil_commit_count", KSTAT_DATA_UINT64 },
	{ "zil_commit_writer_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_indirect_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_indirect_bytes", KSTAT_DATA_UINT64 },
	{ "zil_itx_copied_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_copied_bytes", KSTAT_DATA_UINT64 },
	{ "zil_itx_needcopy_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_needcopy_bytes", KSTAT_DATA_UINT64 },
	{ "zil_itx_metaslab_normal_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_metaslab_normal_bytes", KSTAT_DATA_UINT64 },
	{ "zil_itx_metaslab_slog_count", KSTAT_DATA_UINT64 },
	{ "zil_itx_metaslab_slog_bytes", KSTAT_DATA_UINT64 },
	};

	static zil_sums_t zil_sums_global;
	static kstat_t *zil_kstats_global;

	/*
	* Disable intent logging replay. This global ZIL switch affects all pools.
	*/
	int zil_replay_disable = 0;

	/*
	* Disable the DKIOCFLUSHWRITECACHE commands that are normally sent to
	* the disk(s) by the ZIL after an LWB write has completed. Setting this
	* will cause ZIL corruption on power loss if a volatile out-of-order
	* write cache is enabled.
	*/
	static int zil_nocacheflush = 0;

	/*
	* Limit SLOG write size per commit executed with synchronous priority.
	* Any writes above that will be executed with lower (asynchronous) priority
	* to limit potential SLOG device abuse by single active ZIL writer.
	*/
	static unsigned long zil_slog_bulk = 768 * 1024;

	static kmem_cache_t *zil_lwb_cache;
	static kmem_cache_t *zil_zcw_cache;

	#define LWB_EMPTY(lwb) ((BP_GET_LSIZE(&lwb->lwb_blk) - \
	sizeof (zil_chain_t)) == (lwb->lwb_sz - lwb->lwb_nused))

	static int
	zil_bp_compare(const void x1, const void x2)
	{
	const dva_t dva1 = &((zil_bp_node_t )x1)->zn_dva;
	const dva_t dva2 = &((zil_bp_node_t )x2)->zn_dva;

	int cmp = TREE_CMP(DVA_GET_VDEV(dva1), DVA_GET_VDEV(dva2));
	if (likely(cmp))
	return (cmp);

	return (TREE_CMP(DVA_GET_OFFSET(dva1), DVA_GET_OFFSET(dva2)));
	}

	static void
	zil_bp_tree_init(zilog_t *zilog)
	{
	avl_create(&zilog->zl_bp_tree, zil_bp_compare,
	sizeof (zil_bp_node_t), offsetof(zil_bp_node_t, zn_node));
	}

	static void
	zil_bp_tree_fini(zilog_t *zilog)
	{
	avl_tree_t *t = &zilog->zl_bp_tree;
	zil_bp_node_t *zn;
	void *cookie = NULL;

	while ((zn = avl_destroy_nodes(t, &cookie)) != NULL)
	kmem_free(zn, sizeof (zil_bp_node_t));

	avl_destroy(t);
	}

	int
	zil_bp_tree_add(zilog_t zilog, const blkptr_t bp)
	{
	avl_tree_t *t = &zilog->zl_bp_tree;
	const dva_t *dva;
	zil_bp_node_t *zn;
	avl_index_t where;

	if (BP_IS_EMBEDDED(bp))
	return (0);

	dva = BP_IDENTITY(bp);

	if (avl_find(t, dva, &where) != NULL)
	return (SET_ERROR(EEXIST));

	zn = kmem_alloc(sizeof (zil_bp_node_t), KM_SLEEP);
	zn->zn_dva = *dva;
	avl_insert(t, zn, where);

	return (0);
	}

	static zil_header_t *
	zil_header_in_syncing_context(zilog_t *zilog)
	{
	return ((zil_header_t *)zilog->zl_header);
	}

	static void
	zil_init_log_chain(zilog_t zilog, blkptr_t bp)
	{
	zio_cksum_t *zc = &bp->blk_cksum;

	(void) random_get_pseudo_bytes((void *)&zc->zc_word[ZIL_ZC_GUID_0],
	sizeof (zc->zc_word[ZIL_ZC_GUID_0]));
	(void) random_get_pseudo_bytes((void *)&zc->zc_word[ZIL_ZC_GUID_1],
	sizeof (zc->zc_word[ZIL_ZC_GUID_1]));
	zc->zc_word[ZIL_ZC_OBJSET] = dmu_objset_id(zilog->zl_os);
	zc->zc_word[ZIL_ZC_SEQ] = 1ULL;
	}

	static int
	zil_kstats_global_update(kstat_t *ksp, int rw)
	{
	zil_kstat_values_t *zs = ksp->ks_data;
	ASSERT3P(&zil_stats, ==, zs);

	if (rw == KSTAT_WRITE) {
	return (SET_ERROR(EACCES));
	}

	zil_kstat_values_update(zs, &zil_sums_global);

	return (0);
	}

	/*
	* Read a log block and make sure it's valid.
	*/
	static int
	zil_read_log_block(zilog_t zilog, boolean_t decrypt, const blkptr_t bp,
	blkptr_t nbp, void dst, char **end)
	{
	enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
	arc_flags_t aflags = ARC_FLAG_WAIT;
	arc_buf_t *abuf = NULL;
	zbookmark_phys_t zb;
	int error;

	if (zilog->zl_header->zh_claim_txg == 0)
	zio_flags \|= ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_SCRUB;

	if (!(zilog->zl_header->zh_flags & ZIL_CLAIM_LR_SEQ_VALID))
	zio_flags \|= ZIO_FLAG_SPECULATIVE;

	if (!decrypt)
	zio_flags \|= ZIO_FLAG_RAW;

	SET_BOOKMARK(&zb, bp->blk_cksum.zc_word[ZIL_ZC_OBJSET],
	ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);

	error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func,
	&abuf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);

	if (error == 0) {
	zio_cksum_t cksum = bp->blk_cksum;

	/*
	* Validate the checksummed log block.
	*
	* Sequence numbers should be... sequential. The checksum
	* verifier for the next block should be bp's checksum plus 1.
	*
	* Also check the log chain linkage and size used.
	*/
	cksum.zc_word[ZIL_ZC_SEQ]++;

	if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) {
	zil_chain_t *zilc = abuf->b_data;
	char lr = (char )(zilc + 1);
	uint64_t len = zilc->zc_nused - sizeof (zil_chain_t);

	if (memcmp(&cksum, &zilc->zc_next_blk.blk_cksum,
	sizeof (cksum)) \|\| BP_IS_HOLE(&zilc->zc_next_blk)) {
	error = SET_ERROR(ECKSUM);
	} else {
	ASSERT3U(len, <=, SPA_OLD_MAXBLOCKSIZE);
	memcpy(dst, lr, len);
	end = (char )dst + len;
	*nbp = zilc->zc_next_blk;
	}
	} else {
	char *lr = abuf->b_data;
	uint64_t size = BP_GET_LSIZE(bp);
	zil_chain_t zilc = (zil_chain_t )(lr + size) - 1;

	if (memcmp(&cksum, &zilc->zc_next_blk.blk_cksum,
	sizeof (cksum)) \|\| BP_IS_HOLE(&zilc->zc_next_blk) \|\|
	(zilc->zc_nused > (size - sizeof (*zilc)))) {
	error = SET_ERROR(ECKSUM);
	} else {
	ASSERT3U(zilc->zc_nused, <=,
	SPA_OLD_MAXBLOCKSIZE);
	memcpy(dst, lr, zilc->zc_nused);
	end = (char )dst + zilc->zc_nused;
	*nbp = zilc->zc_next_blk;
	}
	}

	arc_buf_destroy(abuf, &abuf);
	}

	return (error);
	}

	/*
	* Read a TX_WRITE log data block.
	*/
	static int
	zil_read_log_data(zilog_t zilog, const lr_write_t lr, void *wbuf)
	{
	enum zio_flag zio_flags = ZIO_FLAG_CANFAIL;
	const blkptr_t *bp = &lr->lr_blkptr;
	arc_flags_t aflags = ARC_FLAG_WAIT;
	arc_buf_t *abuf = NULL;
	zbookmark_phys_t zb;
	int error;

	if (BP_IS_HOLE(bp)) {
	if (wbuf != NULL)
	memset(wbuf, 0, MAX(BP_GET_LSIZE(bp), lr->lr_length));
	return (0);
	}

	if (zilog->zl_header->zh_claim_txg == 0)
	zio_flags \|= ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_SCRUB;

	/*
	* If we are not using the resulting data, we are just checking that
	* it hasn't been corrupted so we don't need to waste CPU time
	* decompressing and decrypting it.
	*/
	if (wbuf == NULL)
	zio_flags \|= ZIO_FLAG_RAW;

	SET_BOOKMARK(&zb, dmu_objset_id(zilog->zl_os), lr->lr_foid,
	ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp));

	error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func, &abuf,
	ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);

	if (error == 0) {
	if (wbuf != NULL)
	memcpy(wbuf, abuf->b_data, arc_buf_size(abuf));
	arc_buf_destroy(abuf, &abuf);
	}

	return (error);
	}

	void
	zil_sums_init(zil_sums_t *zs)
	{
	wmsum_init(&zs->zil_commit_count, 0);
	wmsum_init(&zs->zil_commit_writer_count, 0);
	wmsum_init(&zs->zil_itx_count, 0);
	wmsum_init(&zs->zil_itx_indirect_count, 0);
	wmsum_init(&zs->zil_itx_indirect_bytes, 0);
	wmsum_init(&zs->zil_itx_copied_count, 0);
	wmsum_init(&zs->zil_itx_copied_bytes, 0);
	wmsum_init(&zs->zil_itx_needcopy_count, 0);
	wmsum_init(&zs->zil_itx_needcopy_bytes, 0);
	wmsum_init(&zs->zil_itx_metaslab_normal_count, 0);
	wmsum_init(&zs->zil_itx_metaslab_normal_bytes, 0);
	wmsum_init(&zs->zil_itx_metaslab_slog_count, 0);
	wmsum_init(&zs->zil_itx_metaslab_slog_bytes, 0);
	}

	void
	zil_sums_fini(zil_sums_t *zs)
	{
	wmsum_fini(&zs->zil_commit_count);
	wmsum_fini(&zs->zil_commit_writer_count);
	wmsum_fini(&zs->zil_itx_count);
	wmsum_fini(&zs->zil_itx_indirect_count);
	wmsum_fini(&zs->zil_itx_indirect_bytes);
	wmsum_fini(&zs->zil_itx_copied_count);
	wmsum_fini(&zs->zil_itx_copied_bytes);
	wmsum_fini(&zs->zil_itx_needcopy_count);
	wmsum_fini(&zs->zil_itx_needcopy_bytes);
	wmsum_fini(&zs->zil_itx_metaslab_normal_count);
	wmsum_fini(&zs->zil_itx_metaslab_normal_bytes);
	wmsum_fini(&zs->zil_itx_metaslab_slog_count);
	wmsum_fini(&zs->zil_itx_metaslab_slog_bytes);
	}

	void
	zil_kstat_values_update(zil_kstat_values_t zs, zil_sums_t zil_sums)
	{
	zs->zil_commit_count.value.ui64 =
	wmsum_value(&zil_sums->zil_commit_count);
	zs->zil_commit_writer_count.value.ui64 =
	wmsum_value(&zil_sums->zil_commit_writer_count);
	zs->zil_itx_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_count);
	zs->zil_itx_indirect_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_indirect_count);
	zs->zil_itx_indirect_bytes.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_indirect_bytes);
	zs->zil_itx_copied_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_copied_count);
	zs->zil_itx_copied_bytes.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_copied_bytes);
	zs->zil_itx_needcopy_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_needcopy_count);
	zs->zil_itx_needcopy_bytes.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_needcopy_bytes);
	zs->zil_itx_metaslab_normal_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_metaslab_normal_count);
	zs->zil_itx_metaslab_normal_bytes.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_metaslab_normal_bytes);
	zs->zil_itx_metaslab_slog_count.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_metaslab_slog_count);
	zs->zil_itx_metaslab_slog_bytes.value.ui64 =
	wmsum_value(&zil_sums->zil_itx_metaslab_slog_bytes);
	}

	/*
	* Parse the intent log, and call parse_func for each valid record within.
	*/
	int
	zil_parse(zilog_t zilog, zil_parse_blk_func_t parse_blk_func,
	zil_parse_lr_func_t parse_lr_func, void arg, uint64_t txg,
	boolean_t decrypt)
	{
	const zil_header_t *zh = zilog->zl_header;
	boolean_t claimed = !!zh->zh_claim_txg;
	uint64_t claim_blk_seq = claimed ? zh->zh_claim_blk_seq : UINT64_MAX;
	uint64_t claim_lr_seq = claimed ? zh->zh_claim_lr_seq : UINT64_MAX;
	uint64_t max_blk_seq = 0;
	uint64_t max_lr_seq = 0;
	uint64_t blk_count = 0;
	uint64_t lr_count = 0;
	blkptr_t blk, next_blk = {{{{0}}}};
	char lrbuf, lrp;
	int error = 0;

	/*
	* Old logs didn't record the maximum zh_claim_lr_seq.
	*/
	if (!(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID))
	claim_lr_seq = UINT64_MAX;

	/*
	* Starting at the block pointed to by zh_log we read the log chain.
	* For each block in the chain we strongly check that block to
	* ensure its validity. We stop when an invalid block is found.
	* For each block pointer in the chain we call parse_blk_func().
	* For each record in each valid block we call parse_lr_func().
	* If the log has been claimed, stop if we encounter a sequence
	* number greater than the highest claimed sequence number.
	*/
	lrbuf = zio_buf_alloc(SPA_OLD_MAXBLOCKSIZE);
	zil_bp_tree_init(zilog);

	for (blk = zh->zh_log; !BP_IS_HOLE(&blk); blk = next_blk) {
	uint64_t blk_seq = blk.blk_cksum.zc_word[ZIL_ZC_SEQ];
	int reclen;
	char *end = NULL;

	if (blk_seq > claim_blk_seq)
	break;

	error = parse_blk_func(zilog, &blk, arg, txg);
	if (error != 0)
	break;
	ASSERT3U(max_blk_seq, <, blk_seq);
	max_blk_seq = blk_seq;
	blk_count++;

	if (max_lr_seq == claim_lr_seq && max_blk_seq == claim_blk_seq)
	break;

	error = zil_read_log_block(zilog, decrypt, &blk, &next_blk,
	lrbuf, &end);
	if (error != 0)
	break;

	for (lrp = lrbuf; lrp < end; lrp += reclen) {
	lr_t lr = (lr_t )lrp;
	reclen = lr->lrc_reclen;
	ASSERT3U(reclen, >=, sizeof (lr_t));
	if (lr->lrc_seq > claim_lr_seq)
	goto done;

	error = parse_lr_func(zilog, lr, arg, txg);
	if (error != 0)
	goto done;
	ASSERT3U(max_lr_seq, <, lr->lrc_seq);
	max_lr_seq = lr->lrc_seq;
	lr_count++;
	}
	}
	done:
	zilog->zl_parse_error = error;
	zilog->zl_parse_blk_seq = max_blk_seq;
	zilog->zl_parse_lr_seq = max_lr_seq;
	zilog->zl_parse_blk_count = blk_count;
	zilog->zl_parse_lr_count = lr_count;

	ASSERT(!claimed \|\| !(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID) \|\|
	(max_blk_seq == claim_blk_seq && max_lr_seq == claim_lr_seq) \|\|
	(decrypt && error == EIO));

	zil_bp_tree_fini(zilog);
	zio_buf_free(lrbuf, SPA_OLD_MAXBLOCKSIZE);

	return (error);
	}

	static int
	zil_clear_log_block(zilog_t zilog, const blkptr_t bp, void *tx,
	uint64_t first_txg)
	{
	(void) tx;
	ASSERT(!BP_IS_HOLE(bp));

	/*
	* As we call this function from the context of a rewind to a
	* checkpoint, each ZIL block whose txg is later than the txg
	* that we rewind to is invalid. Thus, we return -1 so
	* zil_parse() doesn't attempt to read it.
	*/
	if (bp->blk_birth >= first_txg)
	return (-1);

	if (zil_bp_tree_add(zilog, bp) != 0)
	return (0);

	zio_free(zilog->zl_spa, first_txg, bp);
	return (0);
	}

	static int
	zil_noop_log_record(zilog_t zilog, const lr_t lrc, void *tx,
	uint64_t first_txg)
	{
	(void) zilog, (void) lrc, (void) tx, (void) first_txg;
	return (0);
	}

	static int
	zil_claim_log_block(zilog_t zilog, const blkptr_t bp, void *tx,
	uint64_t first_txg)
	{
	/*
	* Claim log block if not already committed and not already claimed.
	* If tx == NULL, just verify that the block is claimable.
	*/
	if (BP_IS_HOLE(bp) \|\| bp->blk_birth < first_txg \|\|
	zil_bp_tree_add(zilog, bp) != 0)
	return (0);

	return (zio_wait(zio_claim(NULL, zilog->zl_spa,
	tx == NULL ? 0 : first_txg, bp, spa_claim_notify, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_SCRUB)));
	}

	static int
	zil_claim_log_record(zilog_t zilog, const lr_t lrc, void *tx,
	uint64_t first_txg)
	{
	lr_write_t lr = (lr_write_t )lrc;
	int error;

	if (lrc->lrc_txtype != TX_WRITE)
	return (0);

	/*
	* If the block is not readable, don't claim it. This can happen
	* in normal operation when a log block is written to disk before
	* some of the dmu_sync() blocks it points to. In this case, the
	* transaction cannot have been committed to anyone (we would have
	* waited for all writes to be stable first), so it is semantically
	* correct to declare this the end of the log.
	*/
	if (lr->lr_blkptr.blk_birth >= first_txg) {
	error = zil_read_log_data(zilog, lr, NULL);
	if (error != 0)
	return (error);
	}

	return (zil_claim_log_block(zilog, &lr->lr_blkptr, tx, first_txg));
	}

	static int
	zil_free_log_block(zilog_t zilog, const blkptr_t bp, void *tx,
	uint64_t claim_txg)
	{
	(void) claim_txg;

	zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp);

	return (0);
	}

	static int
	zil_free_log_record(zilog_t zilog, const lr_t lrc, void *tx,
	uint64_t claim_txg)
	{
	lr_write_t lr = (lr_write_t )lrc;
	blkptr_t *bp = &lr->lr_blkptr;

	/*
	* If we previously claimed it, we need to free it.
	*/
	if (claim_txg != 0 && lrc->lrc_txtype == TX_WRITE &&
	bp->blk_birth >= claim_txg && zil_bp_tree_add(zilog, bp) == 0 &&
	!BP_IS_HOLE(bp))
	zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp);

	return (0);
	}

	static int
	zil_lwb_vdev_compare(const void x1, const void x2)
	{
	const uint64_t v1 = ((zil_vdev_node_t *)x1)->zv_vdev;
	const uint64_t v2 = ((zil_vdev_node_t *)x2)->zv_vdev;

	return (TREE_CMP(v1, v2));
	}

	static lwb_t *
	zil_alloc_lwb(zilog_t zilog, blkptr_t bp, boolean_t slog, uint64_t txg,
	boolean_t fastwrite)
	{
	lwb_t *lwb;

	lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
	lwb->lwb_zilog = zilog;
	lwb->lwb_blk = *bp;
	lwb->lwb_fastwrite = fastwrite;
	lwb->lwb_slog = slog;
	lwb->lwb_state = LWB_STATE_CLOSED;
	lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp));
	lwb->lwb_max_txg = txg;
	lwb->lwb_write_zio = NULL;
	lwb->lwb_root_zio = NULL;
	lwb->lwb_issued_timestamp = 0;
	lwb->lwb_issued_txg = 0;
	if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) {
	lwb->lwb_nused = sizeof (zil_chain_t);
	lwb->lwb_sz = BP_GET_LSIZE(bp);
	} else {
	lwb->lwb_nused = 0;
	lwb->lwb_sz = BP_GET_LSIZE(bp) - sizeof (zil_chain_t);
	}

	mutex_enter(&zilog->zl_lock);
	list_insert_tail(&zilog->zl_lwb_list, lwb);
	mutex_exit(&zilog->zl_lock);

	ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock));
	ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
	VERIFY(list_is_empty(&lwb->lwb_waiters));
	VERIFY(list_is_empty(&lwb->lwb_itxs));

	return (lwb);
	}

	static void
	zil_free_lwb(zilog_t zilog, lwb_t lwb)
	{
	ASSERT(MUTEX_HELD(&zilog->zl_lock));
	ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock));
	VERIFY(list_is_empty(&lwb->lwb_waiters));
	VERIFY(list_is_empty(&lwb->lwb_itxs));
	ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
	ASSERT3P(lwb->lwb_write_zio, ==, NULL);
	ASSERT3P(lwb->lwb_root_zio, ==, NULL);
	ASSERT3U(lwb->lwb_max_txg, <=, spa_syncing_txg(zilog->zl_spa));
	ASSERT(lwb->lwb_state == LWB_STATE_CLOSED \|\|
	lwb->lwb_state == LWB_STATE_FLUSH_DONE);

	/*
	* Clear the zilog's field to indicate this lwb is no longer
	* valid, and prevent use-after-free errors.
	*/
	if (zilog->zl_last_lwb_opened == lwb)
	zilog->zl_last_lwb_opened = NULL;

	kmem_cache_free(zil_lwb_cache, lwb);
	}

	/*
	* Called when we create in-memory log transactions so that we know
	* to cleanup the itxs at the end of spa_sync().
	*/
	static void
	zilog_dirty(zilog_t *zilog, uint64_t txg)
	{
	dsl_pool_t *dp = zilog->zl_dmu_pool;
	dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);

	ASSERT(spa_writeable(zilog->zl_spa));

	if (ds->ds_is_snapshot)
	panic("dirtying snapshot!");

	if (txg_list_add(&dp->dp_dirty_zilogs, zilog, txg)) {
	/* up the hold count until we can be written out */
	dmu_buf_add_ref(ds->ds_dbuf, zilog);

	zilog->zl_dirty_max_txg = MAX(txg, zilog->zl_dirty_max_txg);
	}
	}

	/*
	* Determine if the zil is dirty in the specified txg. Callers wanting to
	* ensure that the dirty state does not change must hold the itxg_lock for
	* the specified txg. Holding the lock will ensure that the zil cannot be
	* dirtied (zil_itx_assign) or cleaned (zil_clean) while we check its current
	* state.
	*/
	static boolean_t __maybe_unused
	zilog_is_dirty_in_txg(zilog_t *zilog, uint64_t txg)
	{
	dsl_pool_t *dp = zilog->zl_dmu_pool;

	if (txg_list_member(&dp->dp_dirty_zilogs, zilog, txg & TXG_MASK))
	return (B_TRUE);
	return (B_FALSE);
	}

	/*
	* Determine if the zil is dirty. The zil is considered dirty if it has
	* any pending itx records that have not been cleaned by zil_clean().
	*/
	static boolean_t
	zilog_is_dirty(zilog_t *zilog)
	{
	dsl_pool_t *dp = zilog->zl_dmu_pool;

	for (int t = 0; t < TXG_SIZE; t++) {
	if (txg_list_member(&dp->dp_dirty_zilogs, zilog, t))
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Its called in zil_commit context (zil_process_commit_list()/zil_create()).
	* It activates SPA_FEATURE_ZILSAXATTR feature, if its enabled.
	* Check dsl_dataset_feature_is_active to avoid txg_wait_synced() on every
	* zil_commit.
	*/
	static void
	zil_commit_activate_saxattr_feature(zilog_t *zilog)
	{
	dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);
	uint64_t txg = 0;
	dmu_tx_t *tx = NULL;

	if (spa_feature_is_enabled(zilog->zl_spa,
	SPA_FEATURE_ZILSAXATTR) &&
	dmu_objset_type(zilog->zl_os) != DMU_OST_ZVOL &&
	!dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_ZILSAXATTR)) {
	tx = dmu_tx_create(zilog->zl_os);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	dsl_dataset_dirty(ds, tx);
	txg = dmu_tx_get_txg(tx);

	mutex_enter(&ds->ds_lock);
	ds->ds_feature_activation[SPA_FEATURE_ZILSAXATTR] =
	(void *)B_TRUE;
	mutex_exit(&ds->ds_lock);
	dmu_tx_commit(tx);
	txg_wait_synced(zilog->zl_dmu_pool, txg);
	}
	}

	/*
	* Create an on-disk intent log.
	*/
	static lwb_t *
	zil_create(zilog_t *zilog)
	{
	const zil_header_t *zh = zilog->zl_header;
	lwb_t *lwb = NULL;
	uint64_t txg = 0;
	dmu_tx_t *tx = NULL;
	blkptr_t blk;
	int error = 0;
	boolean_t fastwrite = FALSE;
	boolean_t slog = FALSE;
	dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);


	/*
	* Wait for any previous destroy to complete.
	*/
	txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);

	ASSERT(zh->zh_claim_txg == 0);
	ASSERT(zh->zh_replay_seq == 0);

	blk = zh->zh_log;

	/*
	* Allocate an initial log block if:
	* - there isn't one already
	* - the existing block is the wrong endianness
	*/
	if (BP_IS_HOLE(&blk) \|\| BP_SHOULD_BYTESWAP(&blk)) {
	tx = dmu_tx_create(zilog->zl_os);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
	txg = dmu_tx_get_txg(tx);

	if (!BP_IS_HOLE(&blk)) {
	zio_free(zilog->zl_spa, txg, &blk);
	BP_ZERO(&blk);
	}

	error = zio_alloc_zil(zilog->zl_spa, zilog->zl_os, txg, &blk,
	ZIL_MIN_BLKSZ, &slog);
	fastwrite = TRUE;

	if (error == 0)
	zil_init_log_chain(zilog, &blk);
	}

	/*
	* Allocate a log write block (lwb) for the first log block.
	*/
	if (error == 0)
	lwb = zil_alloc_lwb(zilog, &blk, slog, txg, fastwrite);

	/*
	* If we just allocated the first log block, commit our transaction
	* and wait for zil_sync() to stuff the block pointer into zh_log.
	* (zh is part of the MOS, so we cannot modify it in open context.)
	*/
	if (tx != NULL) {
	/*
	* If "zilsaxattr" feature is enabled on zpool, then activate
	* it now when we're creating the ZIL chain. We can't wait with
	* this until we write the first xattr log record because we
	* need to wait for the feature activation to sync out.
	*/
	if (spa_feature_is_enabled(zilog->zl_spa,
	SPA_FEATURE_ZILSAXATTR) && dmu_objset_type(zilog->zl_os) !=
	DMU_OST_ZVOL) {
	mutex_enter(&ds->ds_lock);
	ds->ds_feature_activation[SPA_FEATURE_ZILSAXATTR] =
	(void *)B_TRUE;
	mutex_exit(&ds->ds_lock);
	}

	dmu_tx_commit(tx);
	txg_wait_synced(zilog->zl_dmu_pool, txg);
	} else {
	/*
	* This branch covers the case where we enable the feature on a
	* zpool that has existing ZIL headers.
	*/
	zil_commit_activate_saxattr_feature(zilog);
	}
	IMPLY(spa_feature_is_enabled(zilog->zl_spa, SPA_FEATURE_ZILSAXATTR) &&
	dmu_objset_type(zilog->zl_os) != DMU_OST_ZVOL,
	dsl_dataset_feature_is_active(ds, SPA_FEATURE_ZILSAXATTR));

	ASSERT(error != 0 \|\| memcmp(&blk, &zh->zh_log, sizeof (blk)) == 0);
	IMPLY(error == 0, lwb != NULL);

	return (lwb);
	}

	/*
	* In one tx, free all log blocks and clear the log header. If keep_first
	* is set, then we're replaying a log with no content. We want to keep the
	* first block, however, so that the first synchronous transaction doesn't
	* require a txg_wait_synced() in zil_create(). We don't need to
	* txg_wait_synced() here either when keep_first is set, because both
	* zil_create() and zil_destroy() will wait for any in-progress destroys
	* to complete.
	*/
	void
	zil_destroy(zilog_t *zilog, boolean_t keep_first)
	{
	const zil_header_t *zh = zilog->zl_header;
	lwb_t *lwb;
	dmu_tx_t *tx;
	uint64_t txg;

	/*
	* Wait for any previous destroy to complete.
	*/
	txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);

	zilog->zl_old_header = zh; / debugging aid */

	if (BP_IS_HOLE(&zh->zh_log))
	return;

	tx = dmu_tx_create(zilog->zl_os);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
	txg = dmu_tx_get_txg(tx);

	mutex_enter(&zilog->zl_lock);

	ASSERT3U(zilog->zl_destroy_txg, <, txg);
	zilog->zl_destroy_txg = txg;
	zilog->zl_keep_first = keep_first;

	if (!list_is_empty(&zilog->zl_lwb_list)) {
	ASSERT(zh->zh_claim_txg == 0);
	VERIFY(!keep_first);
	while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
	if (lwb->lwb_fastwrite)
	metaslab_fastwrite_unmark(zilog->zl_spa,
	&lwb->lwb_blk);

	list_remove(&zilog->zl_lwb_list, lwb);
	if (lwb->lwb_buf != NULL)
	zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
	zio_free(zilog->zl_spa, txg, &lwb->lwb_blk);
	zil_free_lwb(zilog, lwb);
	}
	} else if (!keep_first) {
	zil_destroy_sync(zilog, tx);
	}
	mutex_exit(&zilog->zl_lock);

	dmu_tx_commit(tx);
	}

	void
	zil_destroy_sync(zilog_t zilog, dmu_tx_t tx)
	{
	ASSERT(list_is_empty(&zilog->zl_lwb_list));
	(void) zil_parse(zilog, zil_free_log_block,
	zil_free_log_record, tx, zilog->zl_header->zh_claim_txg, B_FALSE);
	}

	int
	zil_claim(dsl_pool_t dp, dsl_dataset_t ds, void *txarg)
	{
	dmu_tx_t *tx = txarg;
	zilog_t *zilog;
	uint64_t first_txg;
	zil_header_t *zh;
	objset_t *os;
	int error;

	error = dmu_objset_own_obj(dp, ds->ds_object,
	DMU_OST_ANY, B_FALSE, B_FALSE, FTAG, &os);
	if (error != 0) {
	/*
	* EBUSY indicates that the objset is inconsistent, in which
	* case it can not have a ZIL.
	*/
	if (error != EBUSY) {
	cmn_err(CE_WARN, "can't open objset for %llu, error %u",
	(unsigned long long)ds->ds_object, error);
	}

	return (0);
	}

	zilog = dmu_objset_zil(os);
	zh = zil_header_in_syncing_context(zilog);
	ASSERT3U(tx->tx_txg, ==, spa_first_txg(zilog->zl_spa));
	first_txg = spa_min_claim_txg(zilog->zl_spa);

	/*
	* If the spa_log_state is not set to be cleared, check whether
	* the current uberblock is a checkpoint one and if the current
	* header has been claimed before moving on.
	*
	* If the current uberblock is a checkpointed uberblock then
	* one of the following scenarios took place:
	*
	* 1] We are currently rewinding to the checkpoint of the pool.
	* 2] We crashed in the middle of a checkpoint rewind but we
	* did manage to write the checkpointed uberblock to the
	* vdev labels, so when we tried to import the pool again
	* the checkpointed uberblock was selected from the import
	* procedure.
	*
	* In both cases we want to zero out all the ZIL blocks, except
	* the ones that have been claimed at the time of the checkpoint
	* (their zh_claim_txg != 0). The reason is that these blocks
	* may be corrupted since we may have reused their locations on
	* disk after we took the checkpoint.
	*
	* We could try to set spa_log_state to SPA_LOG_CLEAR earlier
	* when we first figure out whether the current uberblock is
	* checkpointed or not. Unfortunately, that would discard all
	* the logs, including the ones that are claimed, and we would
	* leak space.
	*/
	if (spa_get_log_state(zilog->zl_spa) == SPA_LOG_CLEAR \|\|
	(zilog->zl_spa->spa_uberblock.ub_checkpoint_txg != 0 &&
	zh->zh_claim_txg == 0)) {
	if (!BP_IS_HOLE(&zh->zh_log)) {
	(void) zil_parse(zilog, zil_clear_log_block,
	zil_noop_log_record, tx, first_txg, B_FALSE);
	}
	BP_ZERO(&zh->zh_log);
	if (os->os_encrypted)
	os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;
	dsl_dataset_dirty(dmu_objset_ds(os), tx);
	dmu_objset_disown(os, B_FALSE, FTAG);
	return (0);
	}

	/*
	* If we are not rewinding and opening the pool normally, then
	* the min_claim_txg should be equal to the first txg of the pool.
	*/
	ASSERT3U(first_txg, ==, spa_first_txg(zilog->zl_spa));

	/*
	* Claim all log blocks if we haven't already done so, and remember
	* the highest claimed sequence number. This ensures that if we can
	* read only part of the log now (e.g. due to a missing device),
	* but we can read the entire log later, we will not try to replay
	* or destroy beyond the last block we successfully claimed.
	*/
	ASSERT3U(zh->zh_claim_txg, <=, first_txg);
	if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) {
	(void) zil_parse(zilog, zil_claim_log_block,
	zil_claim_log_record, tx, first_txg, B_FALSE);
	zh->zh_claim_txg = first_txg;
	zh->zh_claim_blk_seq = zilog->zl_parse_blk_seq;
	zh->zh_claim_lr_seq = zilog->zl_parse_lr_seq;
	if (zilog->zl_parse_lr_count \|\| zilog->zl_parse_blk_count > 1)
	zh->zh_flags \|= ZIL_REPLAY_NEEDED;
	zh->zh_flags \|= ZIL_CLAIM_LR_SEQ_VALID;
	if (os->os_encrypted)
	os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE;
	dsl_dataset_dirty(dmu_objset_ds(os), tx);
	}

	ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1));
	dmu_objset_disown(os, B_FALSE, FTAG);
	return (0);
	}

	/*
	* Check the log by walking the log chain.
	* Checksum errors are ok as they indicate the end of the chain.
	* Any other error (no device or read failure) returns an error.
	*/
	int
	zil_check_log_chain(dsl_pool_t dp, dsl_dataset_t ds, void *tx)
	{
	(void) dp;
	zilog_t *zilog;
	objset_t *os;
	blkptr_t *bp;
	int error;

	ASSERT(tx == NULL);

	error = dmu_objset_from_ds(ds, &os);
	if (error != 0) {
	cmn_err(CE_WARN, "can't open objset %llu, error %d",
	(unsigned long long)ds->ds_object, error);
	return (0);
	}

	zilog = dmu_objset_zil(os);
	bp = (blkptr_t *)&zilog->zl_header->zh_log;

	if (!BP_IS_HOLE(bp)) {
	vdev_t *vd;
	boolean_t valid = B_TRUE;

	/*
	* Check the first block and determine if it's on a log device
	* which may have been removed or faulted prior to loading this
	* pool. If so, there's no point in checking the rest of the
	* log as its content should have already been synced to the
	* pool.
	*/
	spa_config_enter(os->os_spa, SCL_STATE, FTAG, RW_READER);
	vd = vdev_lookup_top(os->os_spa, DVA_GET_VDEV(&bp->blk_dva[0]));
	if (vd->vdev_islog && vdev_is_dead(vd))
	valid = vdev_log_state_valid(vd);
	spa_config_exit(os->os_spa, SCL_STATE, FTAG);

	if (!valid)
	return (0);

	/*
	* Check whether the current uberblock is checkpointed (e.g.
	* we are rewinding) and whether the current header has been
	* claimed or not. If it hasn't then skip verifying it. We
	* do this because its ZIL blocks may be part of the pool's
	* state before the rewind, which is no longer valid.
	*/
	zil_header_t *zh = zil_header_in_syncing_context(zilog);
	if (zilog->zl_spa->spa_uberblock.ub_checkpoint_txg != 0 &&
	zh->zh_claim_txg == 0)
	return (0);
	}

	/*
	* Because tx == NULL, zil_claim_log_block() will not actually claim
	* any blocks, but just determine whether it is possible to do so.
	* In addition to checking the log chain, zil_claim_log_block()
	* will invoke zio_claim() with a done func of spa_claim_notify(),
	* which will update spa_max_claim_txg. See spa_load() for details.
	*/
	error = zil_parse(zilog, zil_claim_log_block, zil_claim_log_record, tx,
	zilog->zl_header->zh_claim_txg ? -1ULL :
	spa_min_claim_txg(os->os_spa), B_FALSE);

	return ((error == ECKSUM \|\| error == ENOENT) ? 0 : error);
	}

	/*
	* When an itx is "skipped", this function is used to properly mark the
	* waiter as "done, and signal any thread(s) waiting on it. An itx can
	* be skipped (and not committed to an lwb) for a variety of reasons,
	* one of them being that the itx was committed via spa_sync(), prior to
	* it being committed to an lwb; this can happen if a thread calling
	* zil_commit() is racing with spa_sync().
	*/
	static void
	zil_commit_waiter_skip(zil_commit_waiter_t *zcw)
	{
	mutex_enter(&zcw->zcw_lock);
	ASSERT3B(zcw->zcw_done, ==, B_FALSE);
	zcw->zcw_done = B_TRUE;
	cv_broadcast(&zcw->zcw_cv);
	mutex_exit(&zcw->zcw_lock);
	}

	/*
	* This function is used when the given waiter is to be linked into an
	* lwb's "lwb_waiter" list; i.e. when the itx is committed to the lwb.
	* At this point, the waiter will no longer be referenced by the itx,
	* and instead, will be referenced by the lwb.
	*/
	static void
	zil_commit_waiter_link_lwb(zil_commit_waiter_t zcw, lwb_t lwb)
	{
	/*
	* The lwb_waiters field of the lwb is protected by the zilog's
	* zl_lock, thus it must be held when calling this function.
	*/
	ASSERT(MUTEX_HELD(&lwb->lwb_zilog->zl_lock));

	mutex_enter(&zcw->zcw_lock);
	ASSERT(!list_link_active(&zcw->zcw_node));
	ASSERT3P(zcw->zcw_lwb, ==, NULL);
	ASSERT3P(lwb, !=, NULL);
	ASSERT(lwb->lwb_state == LWB_STATE_OPENED \|\|
	lwb->lwb_state == LWB_STATE_ISSUED \|\|
	lwb->lwb_state == LWB_STATE_WRITE_DONE);

	list_insert_tail(&lwb->lwb_waiters, zcw);
	zcw->zcw_lwb = lwb;
	mutex_exit(&zcw->zcw_lock);
	}

	/*
	* This function is used when zio_alloc_zil() fails to allocate a ZIL
	* block, and the given waiter must be linked to the "nolwb waiters"
	* list inside of zil_process_commit_list().
	*/
	static void
	zil_commit_waiter_link_nolwb(zil_commit_waiter_t zcw, list_t nolwb)
	{
	mutex_enter(&zcw->zcw_lock);
	ASSERT(!list_link_active(&zcw->zcw_node));
	ASSERT3P(zcw->zcw_lwb, ==, NULL);
	list_insert_tail(nolwb, zcw);
	mutex_exit(&zcw->zcw_lock);
	}

	void
	zil_lwb_add_block(lwb_t lwb, const blkptr_t bp)
	{
	avl_tree_t *t = &lwb->lwb_vdev_tree;
	avl_index_t where;
	zil_vdev_node_t *zv, zvsearch;
	int ndvas = BP_GET_NDVAS(bp);
	int i;

	if (zil_nocacheflush)
	return;

	mutex_enter(&lwb->lwb_vdev_lock);
	for (i = 0; i < ndvas; i++) {
	zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
	if (avl_find(t, &zvsearch, &where) == NULL) {
	zv = kmem_alloc(sizeof (*zv), KM_SLEEP);
	zv->zv_vdev = zvsearch.zv_vdev;
	avl_insert(t, zv, where);
	}
	}
	mutex_exit(&lwb->lwb_vdev_lock);
	}

	static void
	zil_lwb_flush_defer(lwb_t lwb, lwb_t nlwb)
	{
	avl_tree_t *src = &lwb->lwb_vdev_tree;
	avl_tree_t *dst = &nlwb->lwb_vdev_tree;
	void *cookie = NULL;
	zil_vdev_node_t *zv;

	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_WRITE_DONE);
	ASSERT3S(nlwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
	ASSERT3S(nlwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);

	/*
	* While 'lwb' is at a point in its lifetime where lwb_vdev_tree does
	* not need the protection of lwb_vdev_lock (it will only be modified
	* while holding zilog->zl_lock) as its writes and those of its
	* children have all completed. The younger 'nlwb' may be waiting on
	* future writes to additional vdevs.
	*/
	mutex_enter(&nlwb->lwb_vdev_lock);
	/*
	* Tear down the 'lwb' vdev tree, ensuring that entries which do not
	* exist in 'nlwb' are moved to it, freeing any would-be duplicates.
	*/
	while ((zv = avl_destroy_nodes(src, &cookie)) != NULL) {
	avl_index_t where;

	if (avl_find(dst, zv, &where) == NULL) {
	avl_insert(dst, zv, where);
	} else {
	kmem_free(zv, sizeof (*zv));
	}
	}
	mutex_exit(&nlwb->lwb_vdev_lock);
	}

	void
	zil_lwb_add_txg(lwb_t *lwb, uint64_t txg)
	{
	lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg);
	}

	/*
	* This function is a called after all vdevs associated with a given lwb
	* write have completed their DKIOCFLUSHWRITECACHE command; or as soon
	* as the lwb write completes, if "zil_nocacheflush" is set. Further,
	* all "previous" lwb's will have completed before this function is
	* called; i.e. this function is called for all previous lwbs before
	* it's called for "this" lwb (enforced via zio the dependencies
	* configured in zil_lwb_set_zio_dependency()).
	*
	* The intention is for this function to be called as soon as the
	* contents of an lwb are considered "stable" on disk, and will survive
	* any sudden loss of power. At this point, any threads waiting for the
	* lwb to reach this state are signalled, and the "waiter" structures
	* are marked "done".
	*/
	static void
	zil_lwb_flush_vdevs_done(zio_t *zio)
	{
	lwb_t *lwb = zio->io_private;
	zilog_t *zilog = lwb->lwb_zilog;
	zil_commit_waiter_t *zcw;
	itx_t *itx;
	uint64_t txg;

	spa_config_exit(zilog->zl_spa, SCL_STATE, lwb);

	zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);

	mutex_enter(&zilog->zl_lock);

	/*
	* If we have had an allocation failure and the txg is
	* waiting to sync then we want zil_sync() to remove the lwb so
	* that it's not picked up as the next new one in
	* zil_process_commit_list(). zil_sync() will only remove the
	* lwb if lwb_buf is null.
	*/
	lwb->lwb_buf = NULL;

	ASSERT3U(lwb->lwb_issued_timestamp, >, 0);
	zilog->zl_last_lwb_latency = gethrtime() - lwb->lwb_issued_timestamp;

	lwb->lwb_root_zio = NULL;

	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_WRITE_DONE);
	lwb->lwb_state = LWB_STATE_FLUSH_DONE;

	if (zilog->zl_last_lwb_opened == lwb) {
	/*
	* Remember the highest committed log sequence number
	* for ztest. We only update this value when all the log
	* writes succeeded, because ztest wants to ASSERT that
	* it got the whole log chain.
	*/
	zilog->zl_commit_lr_seq = zilog->zl_lr_seq;
	}

	while ((itx = list_head(&lwb->lwb_itxs)) != NULL) {
	list_remove(&lwb->lwb_itxs, itx);
	zil_itx_destroy(itx);
	}

	while ((zcw = list_head(&lwb->lwb_waiters)) != NULL) {
	mutex_enter(&zcw->zcw_lock);

	ASSERT(list_link_active(&zcw->zcw_node));
	list_remove(&lwb->lwb_waiters, zcw);

	ASSERT3P(zcw->zcw_lwb, ==, lwb);
	zcw->zcw_lwb = NULL;
	/*
	* We expect any ZIO errors from child ZIOs to have been
	* propagated "up" to this specific LWB's root ZIO, in
	* order for this error handling to work correctly. This
	* includes ZIO errors from either this LWB's write or
	* flush, as well as any errors from other dependent LWBs
	* (e.g. a root LWB ZIO that might be a child of this LWB).
	*
	* With that said, it's important to note that LWB flush
	* errors are not propagated up to the LWB root ZIO.
	* This is incorrect behavior, and results in VDEV flush
	* errors not being handled correctly here. See the
	* comment above the call to "zio_flush" for details.
	*/

	zcw->zcw_zio_error = zio->io_error;

	ASSERT3B(zcw->zcw_done, ==, B_FALSE);
	zcw->zcw_done = B_TRUE;
	cv_broadcast(&zcw->zcw_cv);

	mutex_exit(&zcw->zcw_lock);
	}

	mutex_exit(&zilog->zl_lock);

	mutex_enter(&zilog->zl_lwb_io_lock);
	txg = lwb->lwb_issued_txg;
	ASSERT3U(zilog->zl_lwb_inflight[txg & TXG_MASK], >, 0);
	zilog->zl_lwb_inflight[txg & TXG_MASK]--;
	if (zilog->zl_lwb_inflight[txg & TXG_MASK] == 0)
	cv_broadcast(&zilog->zl_lwb_io_cv);
	mutex_exit(&zilog->zl_lwb_io_lock);
	}

	/*
	* Wait for the completion of all issued write/flush of that txg provided.
	* It guarantees zil_lwb_flush_vdevs_done() is called and returned.
	*/
	static void
	zil_lwb_flush_wait_all(zilog_t *zilog, uint64_t txg)
	{
	ASSERT3U(txg, ==, spa_syncing_txg(zilog->zl_spa));

	mutex_enter(&zilog->zl_lwb_io_lock);
	while (zilog->zl_lwb_inflight[txg & TXG_MASK] > 0)
	cv_wait(&zilog->zl_lwb_io_cv, &zilog->zl_lwb_io_lock);
	mutex_exit(&zilog->zl_lwb_io_lock);

	#ifdef ZFS_DEBUG
	mutex_enter(&zilog->zl_lock);
	mutex_enter(&zilog->zl_lwb_io_lock);
	lwb_t *lwb = list_head(&zilog->zl_lwb_list);
	while (lwb != NULL && lwb->lwb_max_txg <= txg) {
	if (lwb->lwb_issued_txg <= txg) {
	ASSERT(lwb->lwb_state != LWB_STATE_ISSUED);
	ASSERT(lwb->lwb_state != LWB_STATE_WRITE_DONE);
	IMPLY(lwb->lwb_issued_txg > 0,
	lwb->lwb_state == LWB_STATE_FLUSH_DONE);
	}
	IMPLY(lwb->lwb_state == LWB_STATE_FLUSH_DONE,
	lwb->lwb_buf == NULL);
	lwb = list_next(&zilog->zl_lwb_list, lwb);
	}
	mutex_exit(&zilog->zl_lwb_io_lock);
	mutex_exit(&zilog->zl_lock);
	#endif
	}

	/*
	* This is called when an lwb's write zio completes. The callback's
	* purpose is to issue the DKIOCFLUSHWRITECACHE commands for the vdevs
	* in the lwb's lwb_vdev_tree. The tree will contain the vdevs involved
	* in writing out this specific lwb's data, and in the case that cache
	* flushes have been deferred, vdevs involved in writing the data for
	* previous lwbs. The writes corresponding to all the vdevs in the
	* lwb_vdev_tree will have completed by the time this is called, due to
	* the zio dependencies configured in zil_lwb_set_zio_dependency(),
	* which takes deferred flushes into account. The lwb will be "done"
	* once zil_lwb_flush_vdevs_done() is called, which occurs in the zio
	* completion callback for the lwb's root zio.
	*/
	static void
	zil_lwb_write_done(zio_t *zio)
	{
	lwb_t *lwb = zio->io_private;
	spa_t *spa = zio->io_spa;
	zilog_t *zilog = lwb->lwb_zilog;
	avl_tree_t *t = &lwb->lwb_vdev_tree;
	void *cookie = NULL;
	zil_vdev_node_t *zv;
	lwb_t *nlwb;

	ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), !=, 0);

	ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
	ASSERT(BP_GET_TYPE(zio->io_bp) == DMU_OT_INTENT_LOG);
	ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
	ASSERT(BP_GET_BYTEORDER(zio->io_bp) == ZFS_HOST_BYTEORDER);
	ASSERT(!BP_IS_GANG(zio->io_bp));
	ASSERT(!BP_IS_HOLE(zio->io_bp));
	ASSERT(BP_GET_FILL(zio->io_bp) == 0);

	abd_free(zio->io_abd);

	mutex_enter(&zilog->zl_lock);
	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_ISSUED);
	lwb->lwb_state = LWB_STATE_WRITE_DONE;
	lwb->lwb_write_zio = NULL;
	lwb->lwb_fastwrite = FALSE;
	nlwb = list_next(&zilog->zl_lwb_list, lwb);
	mutex_exit(&zilog->zl_lock);

	if (avl_numnodes(t) == 0)
	return;

	/*
	* If there was an IO error, we're not going to call zio_flush()
	* on these vdevs, so we simply empty the tree and free the
	* nodes. We avoid calling zio_flush() since there isn't any
	* good reason for doing so, after the lwb block failed to be
	* written out.
	*
	* Additionally, we don't perform any further error handling at
	* this point (e.g. setting "zcw_zio_error" appropriately), as
	* we expect that to occur in "zil_lwb_flush_vdevs_done" (thus,
	* we expect any error seen here, to have been propagated to
	* that function).
	*/
	if (zio->io_error != 0) {
	while ((zv = avl_destroy_nodes(t, &cookie)) != NULL)
	kmem_free(zv, sizeof (*zv));
	return;
	}

	/*
	* If this lwb does not have any threads waiting for it to
	* complete, we want to defer issuing the DKIOCFLUSHWRITECACHE
	* command to the vdevs written to by "this" lwb, and instead
	* rely on the "next" lwb to handle the DKIOCFLUSHWRITECACHE
	* command for those vdevs. Thus, we merge the vdev tree of
	* "this" lwb with the vdev tree of the "next" lwb in the list,
	* and assume the "next" lwb will handle flushing the vdevs (or
	* deferring the flush(s) again).
	*
	* This is a useful performance optimization, especially for
	* workloads with lots of async write activity and few sync
	* write and/or fsync activity, as it has the potential to
	* coalesce multiple flush commands to a vdev into one.
	*/
	if (list_head(&lwb->lwb_waiters) == NULL && nlwb != NULL) {
	zil_lwb_flush_defer(lwb, nlwb);
	ASSERT(avl_is_empty(&lwb->lwb_vdev_tree));
	return;
	}

	while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) {
	vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev);
	if (vd != NULL) {
	/*
	* The "ZIO_FLAG_DONT_PROPAGATE" is currently
	* always used within "zio_flush". This means,
	* any errors when flushing the vdev(s), will
	* (unfortunately) not be handled correctly,
	* since these "zio_flush" errors will not be
	* propagated up to "zil_lwb_flush_vdevs_done".
	*/
	zio_flush(lwb->lwb_root_zio, vd);
	}
	kmem_free(zv, sizeof (*zv));
	}
	}

	static void
	zil_lwb_set_zio_dependency(zilog_t zilog, lwb_t lwb)
	{
	lwb_t *last_lwb_opened = zilog->zl_last_lwb_opened;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT(MUTEX_HELD(&zilog->zl_lock));

	/*
	* The zilog's "zl_last_lwb_opened" field is used to build the
	* lwb/zio dependency chain, which is used to preserve the
	* ordering of lwb completions that is required by the semantics
	* of the ZIL. Each new lwb zio becomes a parent of the
	* "previous" lwb zio, such that the new lwb's zio cannot
	* complete until the "previous" lwb's zio completes.
	*
	* This is required by the semantics of zil_commit(); the commit
	* waiters attached to the lwbs will be woken in the lwb zio's
	* completion callback, so this zio dependency graph ensures the
	* waiters are woken in the correct order (the same order the
	* lwbs were created).
	*/
	if (last_lwb_opened != NULL &&
	last_lwb_opened->lwb_state != LWB_STATE_FLUSH_DONE) {
	ASSERT(last_lwb_opened->lwb_state == LWB_STATE_OPENED \|\|
	last_lwb_opened->lwb_state == LWB_STATE_ISSUED \|\|
	last_lwb_opened->lwb_state == LWB_STATE_WRITE_DONE);

	ASSERT3P(last_lwb_opened->lwb_root_zio, !=, NULL);
	zio_add_child(lwb->lwb_root_zio,
	last_lwb_opened->lwb_root_zio);

	/*
	* If the previous lwb's write hasn't already completed,
	* we also want to order the completion of the lwb write
	* zios (above, we only order the completion of the lwb
	* root zios). This is required because of how we can
	* defer the DKIOCFLUSHWRITECACHE commands for each lwb.
	*
	* When the DKIOCFLUSHWRITECACHE commands are deferred,
	* the previous lwb will rely on this lwb to flush the
	* vdevs written to by that previous lwb. Thus, we need
	* to ensure this lwb doesn't issue the flush until
	* after the previous lwb's write completes. We ensure
	* this ordering by setting the zio parent/child
	* relationship here.
	*
	* Without this relationship on the lwb's write zio,
	* it's possible for this lwb's write to complete prior
	* to the previous lwb's write completing; and thus, the
	* vdevs for the previous lwb would be flushed prior to
	* that lwb's data being written to those vdevs (the
	* vdevs are flushed in the lwb write zio's completion
	* handler, zil_lwb_write_done()).
	*/
	if (last_lwb_opened->lwb_state != LWB_STATE_WRITE_DONE) {
	ASSERT(last_lwb_opened->lwb_state == LWB_STATE_OPENED \|\|
	last_lwb_opened->lwb_state == LWB_STATE_ISSUED);

	ASSERT3P(last_lwb_opened->lwb_write_zio, !=, NULL);
	zio_add_child(lwb->lwb_write_zio,
	last_lwb_opened->lwb_write_zio);
	}
	}
	}


	/*
	* This function's purpose is to "open" an lwb such that it is ready to
	* accept new itxs being committed to it. To do this, the lwb's zio
	* structures are created, and linked to the lwb. This function is
	* idempotent; if the passed in lwb has already been opened, this
	* function is essentially a no-op.
	*/
	static void
	zil_lwb_write_open(zilog_t zilog, lwb_t lwb)
	{
	zbookmark_phys_t zb;
	zio_priority_t prio;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT3P(lwb, !=, NULL);
	EQUIV(lwb->lwb_root_zio == NULL, lwb->lwb_state == LWB_STATE_CLOSED);
	EQUIV(lwb->lwb_root_zio != NULL, lwb->lwb_state == LWB_STATE_OPENED);

	SET_BOOKMARK(&zb, lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET],
	ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
	lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ]);

	/* Lock so zil_sync() doesn't fastwrite_unmark after zio is created */
	mutex_enter(&zilog->zl_lock);
	if (lwb->lwb_root_zio == NULL) {
	abd_t *lwb_abd = abd_get_from_buf(lwb->lwb_buf,
	BP_GET_LSIZE(&lwb->lwb_blk));

	if (!lwb->lwb_fastwrite) {
	metaslab_fastwrite_mark(zilog->zl_spa, &lwb->lwb_blk);
	lwb->lwb_fastwrite = 1;
	}

	if (!lwb->lwb_slog \|\| zilog->zl_cur_used <= zil_slog_bulk)
	prio = ZIO_PRIORITY_SYNC_WRITE;
	else
	prio = ZIO_PRIORITY_ASYNC_WRITE;

	lwb->lwb_root_zio = zio_root(zilog->zl_spa,
	zil_lwb_flush_vdevs_done, lwb, ZIO_FLAG_CANFAIL);
	ASSERT3P(lwb->lwb_root_zio, !=, NULL);

	lwb->lwb_write_zio = zio_rewrite(lwb->lwb_root_zio,
	zilog->zl_spa, 0, &lwb->lwb_blk, lwb_abd,
	BP_GET_LSIZE(&lwb->lwb_blk), zil_lwb_write_done, lwb,
	prio, ZIO_FLAG_CANFAIL \| ZIO_FLAG_FASTWRITE, &zb);
	ASSERT3P(lwb->lwb_write_zio, !=, NULL);

	lwb->lwb_state = LWB_STATE_OPENED;

	zil_lwb_set_zio_dependency(zilog, lwb);
	zilog->zl_last_lwb_opened = lwb;
	}
	mutex_exit(&zilog->zl_lock);

	ASSERT3P(lwb->lwb_root_zio, !=, NULL);
	ASSERT3P(lwb->lwb_write_zio, !=, NULL);
	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);
	}

	/*
	* Define a limited set of intent log block sizes.
	*
	* These must be a multiple of 4KB. Note only the amount used (again
	* aligned to 4KB) actually gets written. However, we can't always just
	* allocate SPA_OLD_MAXBLOCKSIZE as the slog space could be exhausted.
	*/
	static const struct {
	uint64_t limit;
	uint64_t blksz;
	} zil_block_buckets[] = {
	{ 4096, 4096 }, /* non TX_WRITE */
	{ 8192 + 4096, 8192 + 4096 }, /* database */
	{ 32768 + 4096, 32768 + 4096 }, /* NFS writes */
	{ 65536 + 4096, 65536 + 4096 }, /* 64KB writes */
	{ 131072, 131072 }, /* < 128KB writes */
	{ 131072 +4096, 65536 + 4096 }, /* 128KB writes */
	{ UINT64_MAX, SPA_OLD_MAXBLOCKSIZE}, /* > 128KB writes */
	};

	/*
	* Maximum block size used by the ZIL. This is picked up when the ZIL is
	* initialized. Otherwise this should not be used directly; see
	* zl_max_block_size instead.
	*/
	-static int zil_maxblocksize = SPA_OLD_MAXBLOCKSIZE;
	+static uint_t zil_maxblocksize = SPA_OLD_MAXBLOCKSIZE;

	/*
	* Start a log block write and advance to the next log block.
	* Calls are serialized.
	*/
	static lwb_t *
	zil_lwb_write_issue(zilog_t zilog, lwb_t lwb)
	{
	lwb_t *nlwb = NULL;
	zil_chain_t *zilc;
	spa_t *spa = zilog->zl_spa;
	blkptr_t *bp;
	dmu_tx_t *tx;
	uint64_t txg;
	uint64_t zil_blksz, wsz;
	int i, error;
	boolean_t slog;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT3P(lwb->lwb_root_zio, !=, NULL);
	ASSERT3P(lwb->lwb_write_zio, !=, NULL);
	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);

	if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) {
	zilc = (zil_chain_t *)lwb->lwb_buf;
	bp = &zilc->zc_next_blk;
	} else {
	zilc = (zil_chain_t *)(lwb->lwb_buf + lwb->lwb_sz);
	bp = &zilc->zc_next_blk;
	}

	ASSERT(lwb->lwb_nused <= lwb->lwb_sz);

	/*
	* Allocate the next block and save its address in this block
	* before writing it in order to establish the log chain.
	*/

	tx = dmu_tx_create(zilog->zl_os);

	/*
	* Since we are not going to create any new dirty data, and we
	* can even help with clearing the existing dirty data, we
	* should not be subject to the dirty data based delays. We
	* use TXG_NOTHROTTLE to bypass the delay mechanism.
	*/
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT \| TXG_NOTHROTTLE));

	dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
	txg = dmu_tx_get_txg(tx);

	mutex_enter(&zilog->zl_lwb_io_lock);
	lwb->lwb_issued_txg = txg;
	zilog->zl_lwb_inflight[txg & TXG_MASK]++;
	zilog->zl_lwb_max_issued_txg = MAX(txg, zilog->zl_lwb_max_issued_txg);
	mutex_exit(&zilog->zl_lwb_io_lock);

	/*
	* Log blocks are pre-allocated. Here we select the size of the next
	* block, based on size used in the last block.
	* - first find the smallest bucket that will fit the block from a
	* limited set of block sizes. This is because it's faster to write
	* blocks allocated from the same metaslab as they are adjacent or
	* close.
	* - next find the maximum from the new suggested size and an array of
	* previous sizes. This lessens a picket fence effect of wrongly
	* guessing the size if we have a stream of say 2k, 64k, 2k, 64k
	* requests.
	*
	* Note we only write what is used, but we can't just allocate
	* the maximum block size because we can exhaust the available
	* pool log space.
	*/
	zil_blksz = zilog->zl_cur_used + sizeof (zil_chain_t);
	for (i = 0; zil_blksz > zil_block_buckets[i].limit; i++)
	continue;
	zil_blksz = MIN(zil_block_buckets[i].blksz, zilog->zl_max_block_size);
	zilog->zl_prev_blks[zilog->zl_prev_rotor] = zil_blksz;
	for (i = 0; i < ZIL_PREV_BLKS; i++)
	zil_blksz = MAX(zil_blksz, zilog->zl_prev_blks[i]);
	zilog->zl_prev_rotor = (zilog->zl_prev_rotor + 1) & (ZIL_PREV_BLKS - 1);

	BP_ZERO(bp);
	error = zio_alloc_zil(spa, zilog->zl_os, txg, bp, zil_blksz, &slog);
	if (slog) {
	ZIL_STAT_BUMP(zilog, zil_itx_metaslab_slog_count);
	ZIL_STAT_INCR(zilog, zil_itx_metaslab_slog_bytes,
	lwb->lwb_nused);
	} else {
	ZIL_STAT_BUMP(zilog, zil_itx_metaslab_normal_count);
	ZIL_STAT_INCR(zilog, zil_itx_metaslab_normal_bytes,
	lwb->lwb_nused);
	}
	if (error == 0) {
	ASSERT3U(bp->blk_birth, ==, txg);
	bp->blk_cksum = lwb->lwb_blk.blk_cksum;
	bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++;

	/*
	* Allocate a new log write block (lwb).
	*/
	nlwb = zil_alloc_lwb(zilog, bp, slog, txg, TRUE);
	}

	if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) {
	/* For Slim ZIL only write what is used. */
	wsz = P2ROUNDUP_TYPED(lwb->lwb_nused, ZIL_MIN_BLKSZ, uint64_t);
	ASSERT3U(wsz, <=, lwb->lwb_sz);
	zio_shrink(lwb->lwb_write_zio, wsz);

	} else {
	wsz = lwb->lwb_sz;
	}

	zilc->zc_pad = 0;
	zilc->zc_nused = lwb->lwb_nused;
	zilc->zc_eck.zec_cksum = lwb->lwb_blk.blk_cksum;

	/*
	* clear unused data for security
	*/
	memset(lwb->lwb_buf + lwb->lwb_nused, 0, wsz - lwb->lwb_nused);

	spa_config_enter(zilog->zl_spa, SCL_STATE, lwb, RW_READER);

	zil_lwb_add_block(lwb, &lwb->lwb_blk);
	lwb->lwb_issued_timestamp = gethrtime();
	lwb->lwb_state = LWB_STATE_ISSUED;

	zio_nowait(lwb->lwb_root_zio);
	zio_nowait(lwb->lwb_write_zio);

	dmu_tx_commit(tx);

	/*
	* If there was an allocation failure then nlwb will be null which
	* forces a txg_wait_synced().
	*/
	return (nlwb);
	}

	/*
	* Maximum amount of write data that can be put into single log block.
	*/
	uint64_t
	zil_max_log_data(zilog_t *zilog)
	{
	return (zilog->zl_max_block_size -
	sizeof (zil_chain_t) - sizeof (lr_write_t));
	}

	/*
	* Maximum amount of log space we agree to waste to reduce number of
	* WR_NEED_COPY chunks to reduce zl_get_data() overhead (~12%).
	*/
	static inline uint64_t
	zil_max_waste_space(zilog_t *zilog)
	{
	return (zil_max_log_data(zilog) / 8);
	}

	/*
	* Maximum amount of write data for WR_COPIED. For correctness, consumers
	* must fall back to WR_NEED_COPY if we can't fit the entire record into one
	* maximum sized log block, because each WR_COPIED record must fit in a
	* single log block. For space efficiency, we want to fit two records into a
	* max-sized log block.
	*/
	uint64_t
	zil_max_copied_data(zilog_t *zilog)
	{
	return ((zilog->zl_max_block_size - sizeof (zil_chain_t)) / 2 -
	sizeof (lr_write_t));
	}

	static lwb_t *
	zil_lwb_commit(zilog_t zilog, itx_t itx, lwb_t *lwb)
	{
	lr_t lrcb, lrc;
	lr_write_t lrwb, lrw;
	char *lr_buf;
	uint64_t dlen, dnow, dpad, lwb_sp, reclen, txg, max_log_data;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT3P(lwb, !=, NULL);
	ASSERT3P(lwb->lwb_buf, !=, NULL);

	zil_lwb_write_open(zilog, lwb);

	lrc = &itx->itx_lr;
	lrw = (lr_write_t *)lrc;

	/*
	* A commit itx doesn't represent any on-disk state; instead
	* it's simply used as a place holder on the commit list, and
	* provides a mechanism for attaching a "commit waiter" onto the
	* correct lwb (such that the waiter can be signalled upon
	* completion of that lwb). Thus, we don't process this itx's
	* log record if it's a commit itx (these itx's don't have log
	* records), and instead link the itx's waiter onto the lwb's
	* list of waiters.
	*
	* For more details, see the comment above zil_commit().
	*/
	if (lrc->lrc_txtype == TX_COMMIT) {
	mutex_enter(&zilog->zl_lock);
	zil_commit_waiter_link_lwb(itx->itx_private, lwb);
	itx->itx_private = NULL;
	mutex_exit(&zilog->zl_lock);
	return (lwb);
	}

	if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) {
	dlen = P2ROUNDUP_TYPED(
	lrw->lr_length, sizeof (uint64_t), uint64_t);
	dpad = dlen - lrw->lr_length;
	} else {
	dlen = dpad = 0;
	}
	reclen = lrc->lrc_reclen;
	zilog->zl_cur_used += (reclen + dlen);
	txg = lrc->lrc_txg;

	ASSERT3U(zilog->zl_cur_used, <, UINT64_MAX - (reclen + dlen));

	cont:
	/*
	* If this record won't fit in the current log block, start a new one.
	* For WR_NEED_COPY optimize layout for minimal number of chunks.
	*/
	lwb_sp = lwb->lwb_sz - lwb->lwb_nused;
	max_log_data = zil_max_log_data(zilog);
	if (reclen > lwb_sp \|\| (reclen + dlen > lwb_sp &&
	lwb_sp < zil_max_waste_space(zilog) &&
	(dlen % max_log_data == 0 \|\|
	lwb_sp < reclen + dlen % max_log_data))) {
	lwb = zil_lwb_write_issue(zilog, lwb);
	if (lwb == NULL)
	return (NULL);
	zil_lwb_write_open(zilog, lwb);
	ASSERT(LWB_EMPTY(lwb));
	lwb_sp = lwb->lwb_sz - lwb->lwb_nused;

	/*
	* There must be enough space in the new, empty log block to
	* hold reclen. For WR_COPIED, we need to fit the whole
	* record in one block, and reclen is the header size + the
	* data size. For WR_NEED_COPY, we can create multiple
	* records, splitting the data into multiple blocks, so we
	* only need to fit one word of data per block; in this case
	* reclen is just the header size (no data).
	*/
	ASSERT3U(reclen + MIN(dlen, sizeof (uint64_t)), <=, lwb_sp);
	}

	dnow = MIN(dlen, lwb_sp - reclen);
	lr_buf = lwb->lwb_buf + lwb->lwb_nused;
	memcpy(lr_buf, lrc, reclen);
	lrcb = (lr_t )lr_buf; / Like lrc, but inside lwb. */
	lrwb = (lr_write_t )lrcb; / Like lrw, but inside lwb. */

	ZIL_STAT_BUMP(zilog, zil_itx_count);

	/*
	* If it's a write, fetch the data or get its blkptr as appropriate.
	*/
	if (lrc->lrc_txtype == TX_WRITE) {
	if (txg > spa_freeze_txg(zilog->zl_spa))
	txg_wait_synced(zilog->zl_dmu_pool, txg);
	if (itx->itx_wr_state == WR_COPIED) {
	ZIL_STAT_BUMP(zilog, zil_itx_copied_count);
	ZIL_STAT_INCR(zilog, zil_itx_copied_bytes,
	lrw->lr_length);
	} else {
	char *dbuf;
	int error;

	if (itx->itx_wr_state == WR_NEED_COPY) {
	dbuf = lr_buf + reclen;
	lrcb->lrc_reclen += dnow;
	if (lrwb->lr_length > dnow)
	lrwb->lr_length = dnow;
	lrw->lr_offset += dnow;
	lrw->lr_length -= dnow;
	ZIL_STAT_BUMP(zilog, zil_itx_needcopy_count);
	ZIL_STAT_INCR(zilog, zil_itx_needcopy_bytes,
	dnow);
	} else {
	ASSERT3S(itx->itx_wr_state, ==, WR_INDIRECT);
	dbuf = NULL;
	ZIL_STAT_BUMP(zilog, zil_itx_indirect_count);
	ZIL_STAT_INCR(zilog, zil_itx_indirect_bytes,
	lrw->lr_length);
	}

	/*
	* We pass in the "lwb_write_zio" rather than
	* "lwb_root_zio" so that the "lwb_write_zio"
	* becomes the parent of any zio's created by
	* the "zl_get_data" callback. The vdevs are
	* flushed after the "lwb_write_zio" completes,
	* so we want to make sure that completion
	* callback waits for these additional zio's,
	* such that the vdevs used by those zio's will
	* be included in the lwb's vdev tree, and those
	* vdevs will be properly flushed. If we passed
	* in "lwb_root_zio" here, then these additional
	* vdevs may not be flushed; e.g. if these zio's
	* completed after "lwb_write_zio" completed.
	*/
	error = zilog->zl_get_data(itx->itx_private,
	itx->itx_gen, lrwb, dbuf, lwb,
	lwb->lwb_write_zio);
	if (dbuf != NULL && error == 0 && dnow == dlen)
	/* Zero any padding bytes in the last block. */
	memset((char *)dbuf + lrwb->lr_length, 0, dpad);

	if (error == EIO) {
	txg_wait_synced(zilog->zl_dmu_pool, txg);
	return (lwb);
	}
	if (error != 0) {
	ASSERT(error == ENOENT \|\| error == EEXIST \|\|
	error == EALREADY);
	return (lwb);
	}
	}
	}

	/*
	* We're actually making an entry, so update lrc_seq to be the
	* log record sequence number. Note that this is generally not
	* equal to the itx sequence number because not all transactions
	* are synchronous, and sometimes spa_sync() gets there first.
	*/
	lrcb->lrc_seq = ++zilog->zl_lr_seq;
	lwb->lwb_nused += reclen + dnow;

	zil_lwb_add_txg(lwb, txg);

	ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_sz);
	ASSERT0(P2PHASE(lwb->lwb_nused, sizeof (uint64_t)));

	dlen -= dnow;
	if (dlen > 0) {
	zilog->zl_cur_used += reclen;
	goto cont;
	}

	return (lwb);
	}

	itx_t *
	zil_itx_create(uint64_t txtype, size_t olrsize)
	{
	size_t itxsize, lrsize;
	itx_t *itx;

	lrsize = P2ROUNDUP_TYPED(olrsize, sizeof (uint64_t), size_t);
	itxsize = offsetof(itx_t, itx_lr) + lrsize;

	itx = zio_data_buf_alloc(itxsize);
	itx->itx_lr.lrc_txtype = txtype;
	itx->itx_lr.lrc_reclen = lrsize;
	itx->itx_lr.lrc_seq = 0; /* defensive */
	memset((char *)&itx->itx_lr + olrsize, 0, lrsize - olrsize);
	itx->itx_sync = B_TRUE; /* default is synchronous */
	itx->itx_callback = NULL;
	itx->itx_callback_data = NULL;
	itx->itx_size = itxsize;

	return (itx);
	}

	void
	zil_itx_destroy(itx_t *itx)
	{
	IMPLY(itx->itx_lr.lrc_txtype == TX_COMMIT, itx->itx_callback == NULL);
	IMPLY(itx->itx_callback != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT);

	if (itx->itx_callback != NULL)
	itx->itx_callback(itx->itx_callback_data);

	zio_data_buf_free(itx, itx->itx_size);
	}

	/*
	* Free up the sync and async itxs. The itxs_t has already been detached
	* so no locks are needed.
	*/
	static void
	zil_itxg_clean(void *arg)
	{
	itx_t *itx;
	list_t *list;
	avl_tree_t *t;
	void *cookie;
	itxs_t *itxs = arg;
	itx_async_node_t *ian;

	list = &itxs->i_sync_list;
	while ((itx = list_head(list)) != NULL) {
	/*
	* In the general case, commit itxs will not be found
	* here, as they'll be committed to an lwb via
	* zil_lwb_commit(), and free'd in that function. Having
	* said that, it is still possible for commit itxs to be
	* found here, due to the following race:
	*
	* - a thread calls zil_commit() which assigns the
	* commit itx to a per-txg i_sync_list
	* - zil_itxg_clean() is called (e.g. via spa_sync())
	* while the waiter is still on the i_sync_list
	*
	* There's nothing to prevent syncing the txg while the
	* waiter is on the i_sync_list. This normally doesn't
	* happen because spa_sync() is slower than zil_commit(),
	* but if zil_commit() calls txg_wait_synced() (e.g.
	* because zil_create() or zil_commit_writer_stall() is
	* called) we will hit this case.
	*/
	if (itx->itx_lr.lrc_txtype == TX_COMMIT)
	zil_commit_waiter_skip(itx->itx_private);

	list_remove(list, itx);
	zil_itx_destroy(itx);
	}

	cookie = NULL;
	t = &itxs->i_async_tree;
	while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) {
	list = &ian->ia_list;
	while ((itx = list_head(list)) != NULL) {
	list_remove(list, itx);
	/* commit itxs should never be on the async lists. */
	ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT);
	zil_itx_destroy(itx);
	}
	list_destroy(list);
	kmem_free(ian, sizeof (itx_async_node_t));
	}
	avl_destroy(t);

	kmem_free(itxs, sizeof (itxs_t));
	}

	static int
	zil_aitx_compare(const void x1, const void x2)
	{
	const uint64_t o1 = ((itx_async_node_t *)x1)->ia_foid;
	const uint64_t o2 = ((itx_async_node_t *)x2)->ia_foid;

	return (TREE_CMP(o1, o2));
	}

	/*
	* Remove all async itx with the given oid.
	*/
	void
	zil_remove_async(zilog_t *zilog, uint64_t oid)
	{
	uint64_t otxg, txg;
	itx_async_node_t *ian;
	avl_tree_t *t;
	avl_index_t where;
	list_t clean_list;
	itx_t *itx;

	ASSERT(oid != 0);
	list_create(&clean_list, sizeof (itx_t), offsetof(itx_t, itx_node));

	if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
	otxg = ZILTEST_TXG;
	else
	otxg = spa_last_synced_txg(zilog->zl_spa) + 1;

	for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
	itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];

	mutex_enter(&itxg->itxg_lock);
	if (itxg->itxg_txg != txg) {
	mutex_exit(&itxg->itxg_lock);
	continue;
	}

	/*
	* Locate the object node and append its list.
	*/
	t = &itxg->itxg_itxs->i_async_tree;
	ian = avl_find(t, &oid, &where);
	if (ian != NULL)
	list_move_tail(&clean_list, &ian->ia_list);
	mutex_exit(&itxg->itxg_lock);
	}
	while ((itx = list_head(&clean_list)) != NULL) {
	list_remove(&clean_list, itx);
	/* commit itxs should never be on the async lists. */
	ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT);
	zil_itx_destroy(itx);
	}
	list_destroy(&clean_list);
	}

	void
	zil_itx_assign(zilog_t zilog, itx_t itx, dmu_tx_t *tx)
	{
	uint64_t txg;
	itxg_t *itxg;
	itxs_t itxs, clean = NULL;

	/*
	* Ensure the data of a renamed file is committed before the rename.
	*/
	if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_RENAME)
	zil_async_to_sync(zilog, itx->itx_oid);

	if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX)
	txg = ZILTEST_TXG;
	else
	txg = dmu_tx_get_txg(tx);

	itxg = &zilog->zl_itxg[txg & TXG_MASK];
	mutex_enter(&itxg->itxg_lock);
	itxs = itxg->itxg_itxs;
	if (itxg->itxg_txg != txg) {
	if (itxs != NULL) {
	/*
	* The zil_clean callback hasn't got around to cleaning
	* this itxg. Save the itxs for release below.
	* This should be rare.
	*/
	zfs_dbgmsg("zil_itx_assign: missed itx cleanup for "
	"txg %llu", (u_longlong_t)itxg->itxg_txg);
	clean = itxg->itxg_itxs;
	}
	itxg->itxg_txg = txg;
	itxs = itxg->itxg_itxs = kmem_zalloc(sizeof (itxs_t),
	KM_SLEEP);

	list_create(&itxs->i_sync_list, sizeof (itx_t),
	offsetof(itx_t, itx_node));
	avl_create(&itxs->i_async_tree, zil_aitx_compare,
	sizeof (itx_async_node_t),
	offsetof(itx_async_node_t, ia_node));
	}
	if (itx->itx_sync) {
	list_insert_tail(&itxs->i_sync_list, itx);
	} else {
	avl_tree_t *t = &itxs->i_async_tree;
	uint64_t foid =
	LR_FOID_GET_OBJ(((lr_ooo_t *)&itx->itx_lr)->lr_foid);
	itx_async_node_t *ian;
	avl_index_t where;

	ian = avl_find(t, &foid, &where);
	if (ian == NULL) {
	ian = kmem_alloc(sizeof (itx_async_node_t),
	KM_SLEEP);
	list_create(&ian->ia_list, sizeof (itx_t),
	offsetof(itx_t, itx_node));
	ian->ia_foid = foid;
	avl_insert(t, ian, where);
	}
	list_insert_tail(&ian->ia_list, itx);
	}

	itx->itx_lr.lrc_txg = dmu_tx_get_txg(tx);

	/*
	* We don't want to dirty the ZIL using ZILTEST_TXG, because
	* zil_clean() will never be called using ZILTEST_TXG. Thus, we
	* need to be careful to always dirty the ZIL using the "real"
	* TXG (not itxg_txg) even when the SPA is frozen.
	*/
	zilog_dirty(zilog, dmu_tx_get_txg(tx));
	mutex_exit(&itxg->itxg_lock);

	/* Release the old itxs now we've dropped the lock */
	if (clean != NULL)
	zil_itxg_clean(clean);
	}

	/*
	* If there are any in-memory intent log transactions which have now been
	* synced then start up a taskq to free them. We should only do this after we
	* have written out the uberblocks (i.e. txg has been committed) so that
	* don't inadvertently clean out in-memory log records that would be required
	* by zil_commit().
	*/
	void
	zil_clean(zilog_t *zilog, uint64_t synced_txg)
	{
	itxg_t *itxg = &zilog->zl_itxg[synced_txg & TXG_MASK];
	itxs_t *clean_me;

	ASSERT3U(synced_txg, <, ZILTEST_TXG);

	mutex_enter(&itxg->itxg_lock);
	if (itxg->itxg_itxs == NULL \|\| itxg->itxg_txg == ZILTEST_TXG) {
	mutex_exit(&itxg->itxg_lock);
	return;
	}
	ASSERT3U(itxg->itxg_txg, <=, synced_txg);
	ASSERT3U(itxg->itxg_txg, !=, 0);
	clean_me = itxg->itxg_itxs;
	itxg->itxg_itxs = NULL;
	itxg->itxg_txg = 0;
	mutex_exit(&itxg->itxg_lock);
	/*
	* Preferably start a task queue to free up the old itxs but
	* if taskq_dispatch can't allocate resources to do that then
	* free it in-line. This should be rare. Note, using TQ_SLEEP
	* created a bad performance problem.
	*/
	ASSERT3P(zilog->zl_dmu_pool, !=, NULL);
	ASSERT3P(zilog->zl_dmu_pool->dp_zil_clean_taskq, !=, NULL);
	taskqid_t id = taskq_dispatch(zilog->zl_dmu_pool->dp_zil_clean_taskq,
	zil_itxg_clean, clean_me, TQ_NOSLEEP);
	if (id == TASKQID_INVALID)
	zil_itxg_clean(clean_me);
	}

	/*
	* This function will traverse the queue of itxs that need to be
	* committed, and move them onto the ZIL's zl_itx_commit_list.
	*/
	static void
	zil_get_commit_list(zilog_t *zilog)
	{
	uint64_t otxg, txg;
	list_t *commit_list = &zilog->zl_itx_commit_list;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));

	if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
	otxg = ZILTEST_TXG;
	else
	otxg = spa_last_synced_txg(zilog->zl_spa) + 1;

	/*
	* This is inherently racy, since there is nothing to prevent
	* the last synced txg from changing. That's okay since we'll
	* only commit things in the future.
	*/
	for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
	itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];

	mutex_enter(&itxg->itxg_lock);
	if (itxg->itxg_txg != txg) {
	mutex_exit(&itxg->itxg_lock);
	continue;
	}

	/*
	* If we're adding itx records to the zl_itx_commit_list,
	* then the zil better be dirty in this "txg". We can assert
	* that here since we're holding the itxg_lock which will
	* prevent spa_sync from cleaning it. Once we add the itxs
	* to the zl_itx_commit_list we must commit it to disk even
	* if it's unnecessary (i.e. the txg was synced).
	*/
	ASSERT(zilog_is_dirty_in_txg(zilog, txg) \|\|
	spa_freeze_txg(zilog->zl_spa) != UINT64_MAX);
	list_move_tail(commit_list, &itxg->itxg_itxs->i_sync_list);

	mutex_exit(&itxg->itxg_lock);
	}
	}

	/*
	* Move the async itxs for a specified object to commit into sync lists.
	*/
	void
	zil_async_to_sync(zilog_t *zilog, uint64_t foid)
	{
	uint64_t otxg, txg;
	itx_async_node_t *ian;
	avl_tree_t *t;
	avl_index_t where;

	if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
	otxg = ZILTEST_TXG;
	else
	otxg = spa_last_synced_txg(zilog->zl_spa) + 1;

	/*
	* This is inherently racy, since there is nothing to prevent
	* the last synced txg from changing.
	*/
	for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
	itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];

	mutex_enter(&itxg->itxg_lock);
	if (itxg->itxg_txg != txg) {
	mutex_exit(&itxg->itxg_lock);
	continue;
	}

	/*
	* If a foid is specified then find that node and append its
	* list. Otherwise walk the tree appending all the lists
	* to the sync list. We add to the end rather than the
	* beginning to ensure the create has happened.
	*/
	t = &itxg->itxg_itxs->i_async_tree;
	if (foid != 0) {
	ian = avl_find(t, &foid, &where);
	if (ian != NULL) {
	list_move_tail(&itxg->itxg_itxs->i_sync_list,
	&ian->ia_list);
	}
	} else {
	void *cookie = NULL;

	while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) {
	list_move_tail(&itxg->itxg_itxs->i_sync_list,
	&ian->ia_list);
	list_destroy(&ian->ia_list);
	kmem_free(ian, sizeof (itx_async_node_t));
	}
	}
	mutex_exit(&itxg->itxg_lock);
	}
	}

	/*
	* This function will prune commit itxs that are at the head of the
	* commit list (it won't prune past the first non-commit itx), and
	* either: a) attach them to the last lwb that's still pending
	* completion, or b) skip them altogether.
	*
	* This is used as a performance optimization to prevent commit itxs
	* from generating new lwbs when it's unnecessary to do so.
	*/
	static void
	zil_prune_commit_list(zilog_t *zilog)
	{
	itx_t *itx;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));

	while ((itx = list_head(&zilog->zl_itx_commit_list)) != NULL) {
	lr_t *lrc = &itx->itx_lr;
	if (lrc->lrc_txtype != TX_COMMIT)
	break;

	mutex_enter(&zilog->zl_lock);

	lwb_t *last_lwb = zilog->zl_last_lwb_opened;
	if (last_lwb == NULL \|\|
	last_lwb->lwb_state == LWB_STATE_FLUSH_DONE) {
	/*
	* All of the itxs this waiter was waiting on
	* must have already completed (or there were
	* never any itx's for it to wait on), so it's
	* safe to skip this waiter and mark it done.
	*/
	zil_commit_waiter_skip(itx->itx_private);
	} else {
	zil_commit_waiter_link_lwb(itx->itx_private, last_lwb);
	itx->itx_private = NULL;
	}

	mutex_exit(&zilog->zl_lock);

	list_remove(&zilog->zl_itx_commit_list, itx);
	zil_itx_destroy(itx);
	}

	IMPLY(itx != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT);
	}

	static void
	zil_commit_writer_stall(zilog_t *zilog)
	{
	/*
	* When zio_alloc_zil() fails to allocate the next lwb block on
	* disk, we must call txg_wait_synced() to ensure all of the
	* lwbs in the zilog's zl_lwb_list are synced and then freed (in
	* zil_sync()), such that any subsequent ZIL writer (i.e. a call
	* to zil_process_commit_list()) will have to call zil_create(),
	* and start a new ZIL chain.
	*
	* Since zil_alloc_zil() failed, the lwb that was previously
	* issued does not have a pointer to the "next" lwb on disk.
	* Thus, if another ZIL writer thread was to allocate the "next"
	* on-disk lwb, that block could be leaked in the event of a
	* crash (because the previous lwb on-disk would not point to
	* it).
	*
	* We must hold the zilog's zl_issuer_lock while we do this, to
	* ensure no new threads enter zil_process_commit_list() until
	* all lwb's in the zl_lwb_list have been synced and freed
	* (which is achieved via the txg_wait_synced() call).
	*/
	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));
	txg_wait_synced(zilog->zl_dmu_pool, 0);
	ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL);
	}

	/*
	* This function will traverse the commit list, creating new lwbs as
	* needed, and committing the itxs from the commit list to these newly
	* created lwbs. Additionally, as a new lwb is created, the previous
	* lwb will be issued to the zio layer to be written to disk.
	*/
	static void
	zil_process_commit_list(zilog_t *zilog)
	{
	spa_t *spa = zilog->zl_spa;
	list_t nolwb_itxs;
	list_t nolwb_waiters;
	lwb_t *lwb;
	itx_t *itx;

	ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock));

	/*
	* Return if there's nothing to commit before we dirty the fs by
	* calling zil_create().
	*/
	if (list_head(&zilog->zl_itx_commit_list) == NULL)
	return;

	list_create(&nolwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node));
	list_create(&nolwb_waiters, sizeof (zil_commit_waiter_t),
	offsetof(zil_commit_waiter_t, zcw_node));

	lwb = list_tail(&zilog->zl_lwb_list);
	if (lwb == NULL) {
	lwb = zil_create(zilog);
	} else {
	/*
	* Activate SPA_FEATURE_ZILSAXATTR for the cases where ZIL will
	* have already been created (zl_lwb_list not empty).
	*/
	zil_commit_activate_saxattr_feature(zilog);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);
	}

	while ((itx = list_head(&zilog->zl_itx_commit_list)) != NULL) {
	lr_t *lrc = &itx->itx_lr;
	uint64_t txg = lrc->lrc_txg;

	ASSERT3U(txg, !=, 0);

	if (lrc->lrc_txtype == TX_COMMIT) {
	DTRACE_PROBE2(zil__process__commit__itx,
	zilog_t , zilog, itx_t , itx);
	} else {
	DTRACE_PROBE2(zil__process__normal__itx,
	zilog_t , zilog, itx_t , itx);
	}

	list_remove(&zilog->zl_itx_commit_list, itx);

	boolean_t synced = txg <= spa_last_synced_txg(spa);
	boolean_t frozen = txg > spa_freeze_txg(spa);

	/*
	* If the txg of this itx has already been synced out, then
	* we don't need to commit this itx to an lwb. This is
	* because the data of this itx will have already been
	* written to the main pool. This is inherently racy, and
	* it's still ok to commit an itx whose txg has already
	* been synced; this will result in a write that's
	* unnecessary, but will do no harm.
	*
	* With that said, we always want to commit TX_COMMIT itxs
	* to an lwb, regardless of whether or not that itx's txg
	* has been synced out. We do this to ensure any OPENED lwb
	* will always have at least one zil_commit_waiter_t linked
	* to the lwb.
	*
	* As a counter-example, if we skipped TX_COMMIT itx's
	* whose txg had already been synced, the following
	* situation could occur if we happened to be racing with
	* spa_sync:
	*
	* 1. We commit a non-TX_COMMIT itx to an lwb, where the
	* itx's txg is 10 and the last synced txg is 9.
	* 2. spa_sync finishes syncing out txg 10.
	* 3. We move to the next itx in the list, it's a TX_COMMIT
	* whose txg is 10, so we skip it rather than committing
	* it to the lwb used in (1).
	*
	* If the itx that is skipped in (3) is the last TX_COMMIT
	* itx in the commit list, than it's possible for the lwb
	* used in (1) to remain in the OPENED state indefinitely.
	*
	* To prevent the above scenario from occurring, ensuring
	* that once an lwb is OPENED it will transition to ISSUED
	* and eventually DONE, we always commit TX_COMMIT itx's to
	* an lwb here, even if that itx's txg has already been
	* synced.
	*
	* Finally, if the pool is frozen, we _always_ commit the
	* itx. The point of freezing the pool is to prevent data
	* from being written to the main pool via spa_sync, and
	* instead rely solely on the ZIL to persistently store the
	* data; i.e. when the pool is frozen, the last synced txg
	* value can't be trusted.
	*/
	if (frozen \|\| !synced \|\| lrc->lrc_txtype == TX_COMMIT) {
	if (lwb != NULL) {
	lwb = zil_lwb_commit(zilog, itx, lwb);

	if (lwb == NULL)
	list_insert_tail(&nolwb_itxs, itx);
	else
	list_insert_tail(&lwb->lwb_itxs, itx);
	} else {
	if (lrc->lrc_txtype == TX_COMMIT) {
	zil_commit_waiter_link_nolwb(
	itx->itx_private, &nolwb_waiters);
	}

	list_insert_tail(&nolwb_itxs, itx);
	}
	} else {
	ASSERT3S(lrc->lrc_txtype, !=, TX_COMMIT);
	zil_itx_destroy(itx);
	}
	}

	if (lwb == NULL) {
	/*
	* This indicates zio_alloc_zil() failed to allocate the
	* "next" lwb on-disk. When this happens, we must stall
	* the ZIL write pipeline; see the comment within
	* zil_commit_writer_stall() for more details.
	*/
	zil_commit_writer_stall(zilog);

	/*
	* Additionally, we have to signal and mark the "nolwb"
	* waiters as "done" here, since without an lwb, we
	* can't do this via zil_lwb_flush_vdevs_done() like
	* normal.
	*/
	zil_commit_waiter_t *zcw;
	while ((zcw = list_head(&nolwb_waiters)) != NULL) {
	zil_commit_waiter_skip(zcw);
	list_remove(&nolwb_waiters, zcw);
	}

	/*
	* And finally, we have to destroy the itx's that
	* couldn't be committed to an lwb; this will also call
	* the itx's callback if one exists for the itx.
	*/
	while ((itx = list_head(&nolwb_itxs)) != NULL) {
	list_remove(&nolwb_itxs, itx);
	zil_itx_destroy(itx);
	}
	} else {
	ASSERT(list_is_empty(&nolwb_waiters));
	ASSERT3P(lwb, !=, NULL);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_WRITE_DONE);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_FLUSH_DONE);

	/*
	* At this point, the ZIL block pointed at by the "lwb"
	* variable is in one of the following states: "closed"
	* or "open".
	*
	* If it's "closed", then no itxs have been committed to
	* it, so there's no point in issuing its zio (i.e. it's
	* "empty").
	*
	* If it's "open", then it contains one or more itxs that
	* eventually need to be committed to stable storage. In
	* this case we intentionally do not issue the lwb's zio
	* to disk yet, and instead rely on one of the following
	* two mechanisms for issuing the zio:
	*
	* 1. Ideally, there will be more ZIL activity occurring
	* on the system, such that this function will be
	* immediately called again (not necessarily by the same
	* thread) and this lwb's zio will be issued via
	* zil_lwb_commit(). This way, the lwb is guaranteed to
	* be "full" when it is issued to disk, and we'll make
	* use of the lwb's size the best we can.
	*
	* 2. If there isn't sufficient ZIL activity occurring on
	* the system, such that this lwb's zio isn't issued via
	* zil_lwb_commit(), zil_commit_waiter() will issue the
	* lwb's zio. If this occurs, the lwb is not guaranteed
	* to be "full" by the time its zio is issued, and means
	* the size of the lwb was "too large" given the amount
	* of ZIL activity occurring on the system at that time.
	*
	* We do this for a couple of reasons:
	*
	* 1. To try and reduce the number of IOPs needed to
	* write the same number of itxs. If an lwb has space
	* available in its buffer for more itxs, and more itxs
	* will be committed relatively soon (relative to the
	* latency of performing a write), then it's beneficial
	* to wait for these "next" itxs. This way, more itxs
	* can be committed to stable storage with fewer writes.
	*
	* 2. To try and use the largest lwb block size that the
	* incoming rate of itxs can support. Again, this is to
	* try and pack as many itxs into as few lwbs as
	* possible, without significantly impacting the latency
	* of each individual itx.
	*/
	}
	}

	/*
	* This function is responsible for ensuring the passed in commit waiter
	* (and associated commit itx) is committed to an lwb. If the waiter is
	* not already committed to an lwb, all itxs in the zilog's queue of
	* itxs will be processed. The assumption is the passed in waiter's
	* commit itx will found in the queue just like the other non-commit
	* itxs, such that when the entire queue is processed, the waiter will
	* have been committed to an lwb.
	*
	* The lwb associated with the passed in waiter is not guaranteed to
	* have been issued by the time this function completes. If the lwb is
	* not issued, we rely on future calls to zil_commit_writer() to issue
	* the lwb, or the timeout mechanism found in zil_commit_waiter().
	*/
	static void
	zil_commit_writer(zilog_t zilog, zil_commit_waiter_t zcw)
	{
	ASSERT(!MUTEX_HELD(&zilog->zl_lock));
	ASSERT(spa_writeable(zilog->zl_spa));

	mutex_enter(&zilog->zl_issuer_lock);

	if (zcw->zcw_lwb != NULL \|\| zcw->zcw_done) {
	/*
	* It's possible that, while we were waiting to acquire
	* the "zl_issuer_lock", another thread committed this
	* waiter to an lwb. If that occurs, we bail out early,
	* without processing any of the zilog's queue of itxs.
	*
	* On certain workloads and system configurations, the
	* "zl_issuer_lock" can become highly contended. In an
	* attempt to reduce this contention, we immediately drop
	* the lock if the waiter has already been processed.
	*
	* We've measured this optimization to reduce CPU spent
	* contending on this lock by up to 5%, using a system
	* with 32 CPUs, low latency storage (~50 usec writes),
	* and 1024 threads performing sync writes.
	*/
	goto out;
	}

	ZIL_STAT_BUMP(zilog, zil_commit_writer_count);

	zil_get_commit_list(zilog);
	zil_prune_commit_list(zilog);
	zil_process_commit_list(zilog);

	out:
	mutex_exit(&zilog->zl_issuer_lock);
	}

	static void
	zil_commit_waiter_timeout(zilog_t zilog, zil_commit_waiter_t zcw)
	{
	ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT(MUTEX_HELD(&zcw->zcw_lock));
	ASSERT3B(zcw->zcw_done, ==, B_FALSE);

	lwb_t *lwb = zcw->zcw_lwb;
	ASSERT3P(lwb, !=, NULL);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_CLOSED);

	/*
	* If the lwb has already been issued by another thread, we can
	* immediately return since there's no work to be done (the
	* point of this function is to issue the lwb). Additionally, we
	* do this prior to acquiring the zl_issuer_lock, to avoid
	* acquiring it when it's not necessary to do so.
	*/
	if (lwb->lwb_state == LWB_STATE_ISSUED \|\|
	lwb->lwb_state == LWB_STATE_WRITE_DONE \|\|
	lwb->lwb_state == LWB_STATE_FLUSH_DONE)
	return;

	/*
	* In order to call zil_lwb_write_issue() we must hold the
	* zilog's "zl_issuer_lock". We can't simply acquire that lock,
	* since we're already holding the commit waiter's "zcw_lock",
	* and those two locks are acquired in the opposite order
	* elsewhere.
	*/
	mutex_exit(&zcw->zcw_lock);
	mutex_enter(&zilog->zl_issuer_lock);
	mutex_enter(&zcw->zcw_lock);

	/*
	* Since we just dropped and re-acquired the commit waiter's
	* lock, we have to re-check to see if the waiter was marked
	* "done" during that process. If the waiter was marked "done",
	* the "lwb" pointer is no longer valid (it can be free'd after
	* the waiter is marked "done"), so without this check we could
	* wind up with a use-after-free error below.
	*/
	if (zcw->zcw_done)
	goto out;

	ASSERT3P(lwb, ==, zcw->zcw_lwb);

	/*
	* We've already checked this above, but since we hadn't acquired
	* the zilog's zl_issuer_lock, we have to perform this check a
	* second time while holding the lock.
	*
	* We don't need to hold the zl_lock since the lwb cannot transition
	* from OPENED to ISSUED while we hold the zl_issuer_lock. The lwb
	* _can_ transition from ISSUED to DONE, but it's OK to race with
	* that transition since we treat the lwb the same, whether it's in
	* the ISSUED or DONE states.
	*
	* The important thing, is we treat the lwb differently depending on
	* if it's ISSUED or OPENED, and block any other threads that might
	* attempt to issue this lwb. For that reason we hold the
	* zl_issuer_lock when checking the lwb_state; we must not call
	* zil_lwb_write_issue() if the lwb had already been issued.
	*
	* See the comment above the lwb_state_t structure definition for
	* more details on the lwb states, and locking requirements.
	*/
	if (lwb->lwb_state == LWB_STATE_ISSUED \|\|
	lwb->lwb_state == LWB_STATE_WRITE_DONE \|\|
	lwb->lwb_state == LWB_STATE_FLUSH_DONE)
	goto out;

	ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED);

	/*
	* As described in the comments above zil_commit_waiter() and
	* zil_process_commit_list(), we need to issue this lwb's zio
	* since we've reached the commit waiter's timeout and it still
	* hasn't been issued.
	*/
	lwb_t *nlwb = zil_lwb_write_issue(zilog, lwb);

	IMPLY(nlwb != NULL, lwb->lwb_state != LWB_STATE_OPENED);

	/*
	* Since the lwb's zio hadn't been issued by the time this thread
	* reached its timeout, we reset the zilog's "zl_cur_used" field
	* to influence the zil block size selection algorithm.
	*
	* By having to issue the lwb's zio here, it means the size of the
	* lwb was too large, given the incoming throughput of itxs. By
	* setting "zl_cur_used" to zero, we communicate this fact to the
	* block size selection algorithm, so it can take this information
	* into account, and potentially select a smaller size for the
	* next lwb block that is allocated.
	*/
	zilog->zl_cur_used = 0;

	if (nlwb == NULL) {
	/*
	* When zil_lwb_write_issue() returns NULL, this
	* indicates zio_alloc_zil() failed to allocate the
	* "next" lwb on-disk. When this occurs, the ZIL write
	* pipeline must be stalled; see the comment within the
	* zil_commit_writer_stall() function for more details.
	*
	* We must drop the commit waiter's lock prior to
	* calling zil_commit_writer_stall() or else we can wind
	* up with the following deadlock:
	*
	* - This thread is waiting for the txg to sync while
	* holding the waiter's lock; txg_wait_synced() is
	* used within txg_commit_writer_stall().
	*
	* - The txg can't sync because it is waiting for this
	* lwb's zio callback to call dmu_tx_commit().
	*
	* - The lwb's zio callback can't call dmu_tx_commit()
	* because it's blocked trying to acquire the waiter's
	* lock, which occurs prior to calling dmu_tx_commit()
	*/
	mutex_exit(&zcw->zcw_lock);
	zil_commit_writer_stall(zilog);
	mutex_enter(&zcw->zcw_lock);
	}

	out:
	mutex_exit(&zilog->zl_issuer_lock);
	ASSERT(MUTEX_HELD(&zcw->zcw_lock));
	}

	/*
	* This function is responsible for performing the following two tasks:
	*
	* 1. its primary responsibility is to block until the given "commit
	* waiter" is considered "done".
	*
	* 2. its secondary responsibility is to issue the zio for the lwb that
	* the given "commit waiter" is waiting on, if this function has
	* waited "long enough" and the lwb is still in the "open" state.
	*
	* Given a sufficient amount of itxs being generated and written using
	* the ZIL, the lwb's zio will be issued via the zil_lwb_commit()
	* function. If this does not occur, this secondary responsibility will
	* ensure the lwb is issued even if there is not other synchronous
	* activity on the system.
	*
	* For more details, see zil_process_commit_list(); more specifically,
	* the comment at the bottom of that function.
	*/
	static void
	zil_commit_waiter(zilog_t zilog, zil_commit_waiter_t zcw)
	{
	ASSERT(!MUTEX_HELD(&zilog->zl_lock));
	ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock));
	ASSERT(spa_writeable(zilog->zl_spa));

	mutex_enter(&zcw->zcw_lock);

	/*
	* The timeout is scaled based on the lwb latency to avoid
	* significantly impacting the latency of each individual itx.
	* For more details, see the comment at the bottom of the
	* zil_process_commit_list() function.
	*/
	int pct = MAX(zfs_commit_timeout_pct, 1);
	hrtime_t sleep = (zilog->zl_last_lwb_latency * pct) / 100;
	hrtime_t wakeup = gethrtime() + sleep;
	boolean_t timedout = B_FALSE;

	while (!zcw->zcw_done) {
	ASSERT(MUTEX_HELD(&zcw->zcw_lock));

	lwb_t *lwb = zcw->zcw_lwb;

	/*
	* Usually, the waiter will have a non-NULL lwb field here,
	* but it's possible for it to be NULL as a result of
	* zil_commit() racing with spa_sync().
	*
	* When zil_clean() is called, it's possible for the itxg
	* list (which may be cleaned via a taskq) to contain
	* commit itxs. When this occurs, the commit waiters linked
	* off of these commit itxs will not be committed to an
	* lwb. Additionally, these commit waiters will not be
	* marked done until zil_commit_waiter_skip() is called via
	* zil_itxg_clean().
	*
	* Thus, it's possible for this commit waiter (i.e. the
	* "zcw" variable) to be found in this "in between" state;
	* where it's "zcw_lwb" field is NULL, and it hasn't yet
	* been skipped, so it's "zcw_done" field is still B_FALSE.
	*/
	IMPLY(lwb != NULL, lwb->lwb_state != LWB_STATE_CLOSED);

	if (lwb != NULL && lwb->lwb_state == LWB_STATE_OPENED) {
	ASSERT3B(timedout, ==, B_FALSE);

	/*
	* If the lwb hasn't been issued yet, then we
	* need to wait with a timeout, in case this
	* function needs to issue the lwb after the
	* timeout is reached; responsibility (2) from
	* the comment above this function.
	*/
	int rc = cv_timedwait_hires(&zcw->zcw_cv,
	&zcw->zcw_lock, wakeup, USEC2NSEC(1),
	CALLOUT_FLAG_ABSOLUTE);

	if (rc != -1 \|\| zcw->zcw_done)
	continue;

	timedout = B_TRUE;
	zil_commit_waiter_timeout(zilog, zcw);

	if (!zcw->zcw_done) {
	/*
	* If the commit waiter has already been
	* marked "done", it's possible for the
	* waiter's lwb structure to have already
	* been freed. Thus, we can only reliably
	* make these assertions if the waiter
	* isn't done.
	*/
	ASSERT3P(lwb, ==, zcw->zcw_lwb);
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_OPENED);
	}
	} else {
	/*
	* If the lwb isn't open, then it must have already
	* been issued. In that case, there's no need to
	* use a timeout when waiting for the lwb to
	* complete.
	*
	* Additionally, if the lwb is NULL, the waiter
	* will soon be signaled and marked done via
	* zil_clean() and zil_itxg_clean(), so no timeout
	* is required.
	*/

	IMPLY(lwb != NULL,
	lwb->lwb_state == LWB_STATE_ISSUED \|\|
	lwb->lwb_state == LWB_STATE_WRITE_DONE \|\|
	lwb->lwb_state == LWB_STATE_FLUSH_DONE);
	cv_wait(&zcw->zcw_cv, &zcw->zcw_lock);
	}
	}

	mutex_exit(&zcw->zcw_lock);
	}

	static zil_commit_waiter_t *
	zil_alloc_commit_waiter(void)
	{
	zil_commit_waiter_t *zcw = kmem_cache_alloc(zil_zcw_cache, KM_SLEEP);

	cv_init(&zcw->zcw_cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&zcw->zcw_lock, NULL, MUTEX_DEFAULT, NULL);
	list_link_init(&zcw->zcw_node);
	zcw->zcw_lwb = NULL;
	zcw->zcw_done = B_FALSE;
	zcw->zcw_zio_error = 0;

	return (zcw);
	}

	static void
	zil_free_commit_waiter(zil_commit_waiter_t *zcw)
	{
	ASSERT(!list_link_active(&zcw->zcw_node));
	ASSERT3P(zcw->zcw_lwb, ==, NULL);
	ASSERT3B(zcw->zcw_done, ==, B_TRUE);
	mutex_destroy(&zcw->zcw_lock);
	cv_destroy(&zcw->zcw_cv);
	kmem_cache_free(zil_zcw_cache, zcw);
	}

	/*
	* This function is used to create a TX_COMMIT itx and assign it. This
	* way, it will be linked into the ZIL's list of synchronous itxs, and
	* then later committed to an lwb (or skipped) when
	* zil_process_commit_list() is called.
	*/
	static void
	zil_commit_itx_assign(zilog_t zilog, zil_commit_waiter_t zcw)
	{
	dmu_tx_t *tx = dmu_tx_create(zilog->zl_os);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	itx_t *itx = zil_itx_create(TX_COMMIT, sizeof (lr_t));
	itx->itx_sync = B_TRUE;
	itx->itx_private = zcw;

	zil_itx_assign(zilog, itx, tx);

	dmu_tx_commit(tx);
	}

	/*
	* Commit ZFS Intent Log transactions (itxs) to stable storage.
	*
	* When writing ZIL transactions to the on-disk representation of the
	* ZIL, the itxs are committed to a Log Write Block (lwb). Multiple
	* itxs can be committed to a single lwb. Once a lwb is written and
	* committed to stable storage (i.e. the lwb is written, and vdevs have
	* been flushed), each itx that was committed to that lwb is also
	* considered to be committed to stable storage.
	*
	* When an itx is committed to an lwb, the log record (lr_t) contained
	* by the itx is copied into the lwb's zio buffer, and once this buffer
	* is written to disk, it becomes an on-disk ZIL block.
	*
	* As itxs are generated, they're inserted into the ZIL's queue of
	* uncommitted itxs. The semantics of zil_commit() are such that it will
	* block until all itxs that were in the queue when it was called, are
	* committed to stable storage.
	*
	* If "foid" is zero, this means all "synchronous" and "asynchronous"
	* itxs, for all objects in the dataset, will be committed to stable
	* storage prior to zil_commit() returning. If "foid" is non-zero, all
	* "synchronous" itxs for all objects, but only "asynchronous" itxs
	* that correspond to the foid passed in, will be committed to stable
	* storage prior to zil_commit() returning.
	*
	* Generally speaking, when zil_commit() is called, the consumer doesn't
	* actually care about _all_ of the uncommitted itxs. Instead, they're
	* simply trying to waiting for a specific itx to be committed to disk,
	* but the interface(s) for interacting with the ZIL don't allow such
	* fine-grained communication. A better interface would allow a consumer
	* to create and assign an itx, and then pass a reference to this itx to
	* zil_commit(); such that zil_commit() would return as soon as that
	* specific itx was committed to disk (instead of waiting for _all_
	* itxs to be committed).
	*
	* When a thread calls zil_commit() a special "commit itx" will be
	* generated, along with a corresponding "waiter" for this commit itx.
	* zil_commit() will wait on this waiter's CV, such that when the waiter
	* is marked done, and signaled, zil_commit() will return.
	*
	* This commit itx is inserted into the queue of uncommitted itxs. This
	* provides an easy mechanism for determining which itxs were in the
	* queue prior to zil_commit() having been called, and which itxs were
	* added after zil_commit() was called.
	*
	* The commit itx is special; it doesn't have any on-disk representation.
	* When a commit itx is "committed" to an lwb, the waiter associated
	* with it is linked onto the lwb's list of waiters. Then, when that lwb
	* completes, each waiter on the lwb's list is marked done and signaled
	* -- allowing the thread waiting on the waiter to return from zil_commit().
	*
	* It's important to point out a few critical factors that allow us
	* to make use of the commit itxs, commit waiters, per-lwb lists of
	* commit waiters, and zio completion callbacks like we're doing:
	*
	* 1. The list of waiters for each lwb is traversed, and each commit
	* waiter is marked "done" and signaled, in the zio completion
	* callback of the lwb's zio[*].
	*
	* * Actually, the waiters are signaled in the zio completion
	* callback of the root zio for the DKIOCFLUSHWRITECACHE commands
	* that are sent to the vdevs upon completion of the lwb zio.
	*
	* 2. When the itxs are inserted into the ZIL's queue of uncommitted
	* itxs, the order in which they are inserted is preserved[*]; as
	* itxs are added to the queue, they are added to the tail of
	* in-memory linked lists.
	*
	* When committing the itxs to lwbs (to be written to disk), they
	* are committed in the same order in which the itxs were added to
	* the uncommitted queue's linked list(s); i.e. the linked list of
	* itxs to commit is traversed from head to tail, and each itx is
	* committed to an lwb in that order.
	*
	* * To clarify:
	*
	* - the order of "sync" itxs is preserved w.r.t. other
	* "sync" itxs, regardless of the corresponding objects.
	* - the order of "async" itxs is preserved w.r.t. other
	* "async" itxs corresponding to the same object.
	* - the order of "async" itxs is not preserved w.r.t. other
	* "async" itxs corresponding to different objects.
	* - the order of "sync" itxs w.r.t. "async" itxs (or vice
	* versa) is not preserved, even for itxs that correspond
	* to the same object.
	*
	* For more details, see: zil_itx_assign(), zil_async_to_sync(),
	* zil_get_commit_list(), and zil_process_commit_list().
	*
	* 3. The lwbs represent a linked list of blocks on disk. Thus, any
	* lwb cannot be considered committed to stable storage, until its
	* "previous" lwb is also committed to stable storage. This fact,
	* coupled with the fact described above, means that itxs are
	* committed in (roughly) the order in which they were generated.
	* This is essential because itxs are dependent on prior itxs.
	* Thus, we must not deem an itx as being committed to stable
	* storage, until all prior itxs have also been committed to
	* stable storage.
	*
	* To enforce this ordering of lwb zio's, while still leveraging as
	* much of the underlying storage performance as possible, we rely
	* on two fundamental concepts:
	*
	* 1. The creation and issuance of lwb zio's is protected by
	* the zilog's "zl_issuer_lock", which ensures only a single
	* thread is creating and/or issuing lwb's at a time
	* 2. The "previous" lwb is a child of the "current" lwb
	* (leveraging the zio parent-child dependency graph)
	*
	* By relying on this parent-child zio relationship, we can have
	* many lwb zio's concurrently issued to the underlying storage,
	* but the order in which they complete will be the same order in
	* which they were created.
	*/
	void
	zil_commit(zilog_t *zilog, uint64_t foid)
	{
	/*
	* We should never attempt to call zil_commit on a snapshot for
	* a couple of reasons:
	*
	* 1. A snapshot may never be modified, thus it cannot have any
	* in-flight itxs that would have modified the dataset.
	*
	* 2. By design, when zil_commit() is called, a commit itx will
	* be assigned to this zilog; as a result, the zilog will be
	* dirtied. We must not dirty the zilog of a snapshot; there's
	* checks in the code that enforce this invariant, and will
	* cause a panic if it's not upheld.
	*/
	ASSERT3B(dmu_objset_is_snapshot(zilog->zl_os), ==, B_FALSE);

	if (zilog->zl_sync == ZFS_SYNC_DISABLED)
	return;

	if (!spa_writeable(zilog->zl_spa)) {
	/*
	* If the SPA is not writable, there should never be any
	* pending itxs waiting to be committed to disk. If that
	* weren't true, we'd skip writing those itxs out, and
	* would break the semantics of zil_commit(); thus, we're
	* verifying that truth before we return to the caller.
	*/
	ASSERT(list_is_empty(&zilog->zl_lwb_list));
	ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL);
	for (int i = 0; i < TXG_SIZE; i++)
	ASSERT3P(zilog->zl_itxg[i].itxg_itxs, ==, NULL);
	return;
	}

	/*
	* If the ZIL is suspended, we don't want to dirty it by calling
	* zil_commit_itx_assign() below, nor can we write out
	* lwbs like would be done in zil_commit_write(). Thus, we
	* simply rely on txg_wait_synced() to maintain the necessary
	* semantics, and avoid calling those functions altogether.
	*/
	if (zilog->zl_suspend > 0) {
	txg_wait_synced(zilog->zl_dmu_pool, 0);
	return;
	}

	zil_commit_impl(zilog, foid);
	}

	void
	zil_commit_impl(zilog_t *zilog, uint64_t foid)
	{
	ZIL_STAT_BUMP(zilog, zil_commit_count);

	/*
	* Move the "async" itxs for the specified foid to the "sync"
	* queues, such that they will be later committed (or skipped)
	* to an lwb when zil_process_commit_list() is called.
	*
	* Since these "async" itxs must be committed prior to this
	* call to zil_commit returning, we must perform this operation
	* before we call zil_commit_itx_assign().
	*/
	zil_async_to_sync(zilog, foid);

	/*
	* We allocate a new "waiter" structure which will initially be
	* linked to the commit itx using the itx's "itx_private" field.
	* Since the commit itx doesn't represent any on-disk state,
	* when it's committed to an lwb, rather than copying the its
	* lr_t into the lwb's buffer, the commit itx's "waiter" will be
	* added to the lwb's list of waiters. Then, when the lwb is
	* committed to stable storage, each waiter in the lwb's list of
	* waiters will be marked "done", and signalled.
	*
	* We must create the waiter and assign the commit itx prior to
	* calling zil_commit_writer(), or else our specific commit itx
	* is not guaranteed to be committed to an lwb prior to calling
	* zil_commit_waiter().
	*/
	zil_commit_waiter_t *zcw = zil_alloc_commit_waiter();
	zil_commit_itx_assign(zilog, zcw);

	zil_commit_writer(zilog, zcw);
	zil_commit_waiter(zilog, zcw);

	if (zcw->zcw_zio_error != 0) {
	/*
	* If there was an error writing out the ZIL blocks that
	* this thread is waiting on, then we fallback to
	* relying on spa_sync() to write out the data this
	* thread is waiting on. Obviously this has performance
	* implications, but the expectation is for this to be
	* an exceptional case, and shouldn't occur often.
	*/
	DTRACE_PROBE2(zil__commit__io__error,
	zilog_t , zilog, zil_commit_waiter_t , zcw);
	txg_wait_synced(zilog->zl_dmu_pool, 0);
	}

	zil_free_commit_waiter(zcw);
	}

	/*
	* Called in syncing context to free committed log blocks and update log header.
	*/
	void
	zil_sync(zilog_t zilog, dmu_tx_t tx)
	{
	zil_header_t *zh = zil_header_in_syncing_context(zilog);
	uint64_t txg = dmu_tx_get_txg(tx);
	spa_t *spa = zilog->zl_spa;
	uint64_t *replayed_seq = &zilog->zl_replayed_seq[txg & TXG_MASK];
	lwb_t *lwb;

	/*
	* We don't zero out zl_destroy_txg, so make sure we don't try
	* to destroy it twice.
	*/
	if (spa_sync_pass(spa) != 1)
	return;

	zil_lwb_flush_wait_all(zilog, txg);

	mutex_enter(&zilog->zl_lock);

	ASSERT(zilog->zl_stop_sync == 0);

	if (*replayed_seq != 0) {
	ASSERT(zh->zh_replay_seq < *replayed_seq);
	zh->zh_replay_seq = *replayed_seq;
	*replayed_seq = 0;
	}

	if (zilog->zl_destroy_txg == txg) {
	blkptr_t blk = zh->zh_log;
	dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os);

	ASSERT(list_head(&zilog->zl_lwb_list) == NULL);

	memset(zh, 0, sizeof (zil_header_t));
	memset(zilog->zl_replayed_seq, 0,
	sizeof (zilog->zl_replayed_seq));

	if (zilog->zl_keep_first) {
	/*
	* If this block was part of log chain that couldn't
	* be claimed because a device was missing during
	* zil_claim(), but that device later returns,
	* then this block could erroneously appear valid.
	* To guard against this, assign a new GUID to the new
	* log chain so it doesn't matter what blk points to.
	*/
	zil_init_log_chain(zilog, &blk);
	zh->zh_log = blk;
	} else {
	/*
	* A destroyed ZIL chain can't contain any TX_SETSAXATTR
	* records. So, deactivate the feature for this dataset.
	* We activate it again when we start a new ZIL chain.
	*/
	if (dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_ZILSAXATTR))
	dsl_dataset_deactivate_feature(ds,
	SPA_FEATURE_ZILSAXATTR, tx);
	}
	}

	while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
	zh->zh_log = lwb->lwb_blk;
	if (lwb->lwb_buf != NULL \|\| lwb->lwb_max_txg > txg)
	break;
	list_remove(&zilog->zl_lwb_list, lwb);
	zio_free(spa, txg, &lwb->lwb_blk);
	zil_free_lwb(zilog, lwb);

	/*
	* If we don't have anything left in the lwb list then
	* we've had an allocation failure and we need to zero
	* out the zil_header blkptr so that we don't end
	* up freeing the same block twice.
	*/
	if (list_head(&zilog->zl_lwb_list) == NULL)
	BP_ZERO(&zh->zh_log);
	}

	/*
	* Remove fastwrite on any blocks that have been pre-allocated for
	* the next commit. This prevents fastwrite counter pollution by
	* unused, long-lived LWBs.
	*/
	for (; lwb != NULL; lwb = list_next(&zilog->zl_lwb_list, lwb)) {
	if (lwb->lwb_fastwrite && !lwb->lwb_write_zio) {
	metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);
	lwb->lwb_fastwrite = 0;
	}
	}

	mutex_exit(&zilog->zl_lock);
	}

	static int
	zil_lwb_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	lwb_t *lwb = vbuf;
	list_create(&lwb->lwb_itxs, sizeof (itx_t), offsetof(itx_t, itx_node));
	list_create(&lwb->lwb_waiters, sizeof (zil_commit_waiter_t),
	offsetof(zil_commit_waiter_t, zcw_node));
	avl_create(&lwb->lwb_vdev_tree, zil_lwb_vdev_compare,
	sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node));
	mutex_init(&lwb->lwb_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	return (0);
	}

	static void
	zil_lwb_dest(void vbuf, void unused)
	{
	(void) unused;
	lwb_t *lwb = vbuf;
	mutex_destroy(&lwb->lwb_vdev_lock);
	avl_destroy(&lwb->lwb_vdev_tree);
	list_destroy(&lwb->lwb_waiters);
	list_destroy(&lwb->lwb_itxs);
	}

	void
	zil_init(void)
	{
	zil_lwb_cache = kmem_cache_create("zil_lwb_cache",
	sizeof (lwb_t), 0, zil_lwb_cons, zil_lwb_dest, NULL, NULL, NULL, 0);

	zil_zcw_cache = kmem_cache_create("zil_zcw_cache",
	sizeof (zil_commit_waiter_t), 0, NULL, NULL, NULL, NULL, NULL, 0);

	zil_sums_init(&zil_sums_global);
	zil_kstats_global = kstat_create("zfs", 0, "zil", "misc",
	KSTAT_TYPE_NAMED, sizeof (zil_stats) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);

	if (zil_kstats_global != NULL) {
	zil_kstats_global->ks_data = &zil_stats;
	zil_kstats_global->ks_update = zil_kstats_global_update;
	zil_kstats_global->ks_private = NULL;
	kstat_install(zil_kstats_global);
	}
	}

	void
	zil_fini(void)
	{
	kmem_cache_destroy(zil_zcw_cache);
	kmem_cache_destroy(zil_lwb_cache);

	if (zil_kstats_global != NULL) {
	kstat_delete(zil_kstats_global);
	zil_kstats_global = NULL;
	}

	zil_sums_fini(&zil_sums_global);
	}

	void
	zil_set_sync(zilog_t *zilog, uint64_t sync)
	{
	zilog->zl_sync = sync;
	}

	void
	zil_set_logbias(zilog_t *zilog, uint64_t logbias)
	{
	zilog->zl_logbias = logbias;
	}

	zilog_t *
	zil_alloc(objset_t os, zil_header_t zh_phys)
	{
	zilog_t *zilog;

	zilog = kmem_zalloc(sizeof (zilog_t), KM_SLEEP);

	zilog->zl_header = zh_phys;
	zilog->zl_os = os;
	zilog->zl_spa = dmu_objset_spa(os);
	zilog->zl_dmu_pool = dmu_objset_pool(os);
	zilog->zl_destroy_txg = TXG_INITIAL - 1;
	zilog->zl_logbias = dmu_objset_logbias(os);
	zilog->zl_sync = dmu_objset_syncprop(os);
	zilog->zl_dirty_max_txg = 0;
	zilog->zl_last_lwb_opened = NULL;
	zilog->zl_last_lwb_latency = 0;
	zilog->zl_max_block_size = zil_maxblocksize;

	mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&zilog->zl_issuer_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&zilog->zl_lwb_io_lock, NULL, MUTEX_DEFAULT, NULL);

	for (int i = 0; i < TXG_SIZE; i++) {
	mutex_init(&zilog->zl_itxg[i].itxg_lock, NULL,
	MUTEX_DEFAULT, NULL);
	}

	list_create(&zilog->zl_lwb_list, sizeof (lwb_t),
	offsetof(lwb_t, lwb_node));

	list_create(&zilog->zl_itx_commit_list, sizeof (itx_t),
	offsetof(itx_t, itx_node));

	cv_init(&zilog->zl_cv_suspend, NULL, CV_DEFAULT, NULL);
	cv_init(&zilog->zl_lwb_io_cv, NULL, CV_DEFAULT, NULL);

	return (zilog);
	}

	void
	zil_free(zilog_t *zilog)
	{
	int i;

	zilog->zl_stop_sync = 1;

	ASSERT0(zilog->zl_suspend);
	ASSERT0(zilog->zl_suspending);

	ASSERT(list_is_empty(&zilog->zl_lwb_list));
	list_destroy(&zilog->zl_lwb_list);

	ASSERT(list_is_empty(&zilog->zl_itx_commit_list));
	list_destroy(&zilog->zl_itx_commit_list);

	for (i = 0; i < TXG_SIZE; i++) {
	/*
	* It's possible for an itx to be generated that doesn't dirty
	* a txg (e.g. ztest TX_TRUNCATE). So there's no zil_clean()
	* callback to remove the entry. We remove those here.
	*
	* Also free up the ziltest itxs.
	*/
	if (zilog->zl_itxg[i].itxg_itxs)
	zil_itxg_clean(zilog->zl_itxg[i].itxg_itxs);
	mutex_destroy(&zilog->zl_itxg[i].itxg_lock);
	}

	mutex_destroy(&zilog->zl_issuer_lock);
	mutex_destroy(&zilog->zl_lock);
	mutex_destroy(&zilog->zl_lwb_io_lock);

	cv_destroy(&zilog->zl_cv_suspend);
	cv_destroy(&zilog->zl_lwb_io_cv);

	kmem_free(zilog, sizeof (zilog_t));
	}

	/*
	* Open an intent log.
	*/
	zilog_t *
	zil_open(objset_t os, zil_get_data_t get_data, zil_sums_t *zil_sums)
	{
	zilog_t *zilog = dmu_objset_zil(os);

	ASSERT3P(zilog->zl_get_data, ==, NULL);
	ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL);
	ASSERT(list_is_empty(&zilog->zl_lwb_list));

	zilog->zl_get_data = get_data;
	zilog->zl_sums = zil_sums;

	return (zilog);
	}

	/*
	* Close an intent log.
	*/
	void
	zil_close(zilog_t *zilog)
	{
	lwb_t *lwb;
	uint64_t txg;

	if (!dmu_objset_is_snapshot(zilog->zl_os)) {
	zil_commit(zilog, 0);
	} else {
	ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL);
	ASSERT0(zilog->zl_dirty_max_txg);
	ASSERT3B(zilog_is_dirty(zilog), ==, B_FALSE);
	}

	mutex_enter(&zilog->zl_lock);
	lwb = list_tail(&zilog->zl_lwb_list);
	if (lwb == NULL)
	txg = zilog->zl_dirty_max_txg;
	else
	txg = MAX(zilog->zl_dirty_max_txg, lwb->lwb_max_txg);
	mutex_exit(&zilog->zl_lock);

	/*
	* zl_lwb_max_issued_txg may be larger than lwb_max_txg. It depends
	* on the time when the dmu_tx transaction is assigned in
	* zil_lwb_write_issue().
	*/
	mutex_enter(&zilog->zl_lwb_io_lock);
	txg = MAX(zilog->zl_lwb_max_issued_txg, txg);
	mutex_exit(&zilog->zl_lwb_io_lock);

	/*
	* We need to use txg_wait_synced() to wait until that txg is synced.
	* zil_sync() will guarantee all lwbs up to that txg have been
	* written out, flushed, and cleaned.
	*/
	if (txg != 0)
	txg_wait_synced(zilog->zl_dmu_pool, txg);

	if (zilog_is_dirty(zilog))
	zfs_dbgmsg("zil (%px) is dirty, txg %llu", zilog,
	(u_longlong_t)txg);
	if (txg < spa_freeze_txg(zilog->zl_spa))
	VERIFY(!zilog_is_dirty(zilog));

	zilog->zl_get_data = NULL;

	/*
	* We should have only one lwb left on the list; remove it now.
	*/
	mutex_enter(&zilog->zl_lock);
	lwb = list_head(&zilog->zl_lwb_list);
	if (lwb != NULL) {
	ASSERT3P(lwb, ==, list_tail(&zilog->zl_lwb_list));
	ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED);

	if (lwb->lwb_fastwrite)
	metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);

	list_remove(&zilog->zl_lwb_list, lwb);
	zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
	zil_free_lwb(zilog, lwb);
	}
	mutex_exit(&zilog->zl_lock);
	}

	static const char *suspend_tag = "zil suspending";

	/*
	* Suspend an intent log. While in suspended mode, we still honor
	* synchronous semantics, but we rely on txg_wait_synced() to do it.
	* On old version pools, we suspend the log briefly when taking a
	* snapshot so that it will have an empty intent log.
	*
	* Long holds are not really intended to be used the way we do here --
	* held for such a short time. A concurrent caller of dsl_dataset_long_held()
	* could fail. Therefore we take pains to only put a long hold if it is
	* actually necessary. Fortunately, it will only be necessary if the
	* objset is currently mounted (or the ZVOL equivalent). In that case it
	* will already have a long hold, so we are not really making things any worse.
	*
	* Ideally, we would locate the existing long-holder (i.e. the zfsvfs_t or
	* zvol_state_t), and use their mechanism to prevent their hold from being
	* dropped (e.g. VFS_HOLD()). However, that would be even more pain for
	* very little gain.
	*
	* if cookiep == NULL, this does both the suspend & resume.
	* Otherwise, it returns with the dataset "long held", and the cookie
	* should be passed into zil_resume().
	*/
	int
	zil_suspend(const char osname, void *cookiep)
	{
	objset_t *os;
	zilog_t *zilog;
	const zil_header_t *zh;
	int error;

	error = dmu_objset_hold(osname, suspend_tag, &os);
	if (error != 0)
	return (error);
	zilog = dmu_objset_zil(os);

	mutex_enter(&zilog->zl_lock);
	zh = zilog->zl_header;

	if (zh->zh_flags & ZIL_REPLAY_NEEDED) { /* unplayed log */
	mutex_exit(&zilog->zl_lock);
	dmu_objset_rele(os, suspend_tag);
	return (SET_ERROR(EBUSY));
	}

	/*
	* Don't put a long hold in the cases where we can avoid it. This
	* is when there is no cookie so we are doing a suspend & resume
	* (i.e. called from zil_vdev_offline()), and there's nothing to do
	* for the suspend because it's already suspended, or there's no ZIL.
	*/
	if (cookiep == NULL && !zilog->zl_suspending &&
	(zilog->zl_suspend > 0 \|\| BP_IS_HOLE(&zh->zh_log))) {
	mutex_exit(&zilog->zl_lock);
	dmu_objset_rele(os, suspend_tag);
	return (0);
	}

	dsl_dataset_long_hold(dmu_objset_ds(os), suspend_tag);
	dsl_pool_rele(dmu_objset_pool(os), suspend_tag);

	zilog->zl_suspend++;

	if (zilog->zl_suspend > 1) {
	/*
	* Someone else is already suspending it.
	* Just wait for them to finish.
	*/

	while (zilog->zl_suspending)
	cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock);
	mutex_exit(&zilog->zl_lock);

	if (cookiep == NULL)
	zil_resume(os);
	else
	*cookiep = os;
	return (0);
	}

	/*
	* If there is no pointer to an on-disk block, this ZIL must not
	* be active (e.g. filesystem not mounted), so there's nothing
	* to clean up.
	*/
	if (BP_IS_HOLE(&zh->zh_log)) {
	ASSERT(cookiep != NULL); /* fast path already handled */

	*cookiep = os;
	mutex_exit(&zilog->zl_lock);
	return (0);
	}

	/*
	* The ZIL has work to do. Ensure that the associated encryption
	* key will remain mapped while we are committing the log by
	* grabbing a reference to it. If the key isn't loaded we have no
	* choice but to return an error until the wrapping key is loaded.
	*/
	if (os->os_encrypted &&
	dsl_dataset_create_key_mapping(dmu_objset_ds(os)) != 0) {
	zilog->zl_suspend--;
	mutex_exit(&zilog->zl_lock);
	dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag);
	dsl_dataset_rele(dmu_objset_ds(os), suspend_tag);
	return (SET_ERROR(EACCES));
	}

	zilog->zl_suspending = B_TRUE;
	mutex_exit(&zilog->zl_lock);

	/*
	* We need to use zil_commit_impl to ensure we wait for all
	* LWB_STATE_OPENED and LWB_STATE_ISSUED lwbs to be committed
	* to disk before proceeding. If we used zil_commit instead, it
	* would just call txg_wait_synced(), because zl_suspend is set.
	* txg_wait_synced() doesn't wait for these lwb's to be
	* LWB_STATE_FLUSH_DONE before returning.
	*/
	zil_commit_impl(zilog, 0);

	/*
	* Now that we've ensured all lwb's are LWB_STATE_FLUSH_DONE, we
	* use txg_wait_synced() to ensure the data from the zilog has
	* migrated to the main pool before calling zil_destroy().
	*/
	txg_wait_synced(zilog->zl_dmu_pool, 0);

	zil_destroy(zilog, B_FALSE);

	mutex_enter(&zilog->zl_lock);
	zilog->zl_suspending = B_FALSE;
	cv_broadcast(&zilog->zl_cv_suspend);
	mutex_exit(&zilog->zl_lock);

	if (os->os_encrypted)
	dsl_dataset_remove_key_mapping(dmu_objset_ds(os));

	if (cookiep == NULL)
	zil_resume(os);
	else
	*cookiep = os;
	return (0);
	}

	void
	zil_resume(void *cookie)
	{
	objset_t *os = cookie;
	zilog_t *zilog = dmu_objset_zil(os);

	mutex_enter(&zilog->zl_lock);
	ASSERT(zilog->zl_suspend != 0);
	zilog->zl_suspend--;
	mutex_exit(&zilog->zl_lock);
	dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag);
	dsl_dataset_rele(dmu_objset_ds(os), suspend_tag);
	}

	typedef struct zil_replay_arg {
	zil_replay_func_t const zr_replay;
	void *zr_arg;
	boolean_t zr_byteswap;
	char *zr_lr;
	} zil_replay_arg_t;

	static int
	zil_replay_error(zilog_t zilog, const lr_t lr, int error)
	{
	char name[ZFS_MAX_DATASET_NAME_LEN];

	zilog->zl_replaying_seq--; /* didn't actually replay this one */

	dmu_objset_name(zilog->zl_os, name);

	cmn_err(CE_WARN, "ZFS replay transaction error %d, "
	"dataset %s, seq 0x%llx, txtype %llu %s\n", error, name,
	(u_longlong_t)lr->lrc_seq,
	(u_longlong_t)(lr->lrc_txtype & ~TX_CI),
	(lr->lrc_txtype & TX_CI) ? "CI" : "");

	return (error);
	}

	static int
	zil_replay_log_record(zilog_t zilog, const lr_t lr, void *zra,
	uint64_t claim_txg)
	{
	zil_replay_arg_t *zr = zra;
	const zil_header_t *zh = zilog->zl_header;
	uint64_t reclen = lr->lrc_reclen;
	uint64_t txtype = lr->lrc_txtype;
	int error = 0;

	zilog->zl_replaying_seq = lr->lrc_seq;

	if (lr->lrc_seq <= zh->zh_replay_seq) /* already replayed */
	return (0);

	if (lr->lrc_txg < claim_txg) /* already committed */
	return (0);

	/* Strip case-insensitive bit, still present in log record */
	txtype &= ~TX_CI;

	if (txtype == 0 \|\| txtype >= TX_MAX_TYPE)
	return (zil_replay_error(zilog, lr, EINVAL));

	/*
	* If this record type can be logged out of order, the object
	* (lr_foid) may no longer exist. That's legitimate, not an error.
	*/
	if (TX_OOO(txtype)) {
	error = dmu_object_info(zilog->zl_os,
	LR_FOID_GET_OBJ(((lr_ooo_t *)lr)->lr_foid), NULL);
	if (error == ENOENT \|\| error == EEXIST)
	return (0);
	}

	/*
	* Make a copy of the data so we can revise and extend it.
	*/
	memcpy(zr->zr_lr, lr, reclen);

	/*
	* If this is a TX_WRITE with a blkptr, suck in the data.
	*/
	if (txtype == TX_WRITE && reclen == sizeof (lr_write_t)) {
	error = zil_read_log_data(zilog, (lr_write_t *)lr,
	zr->zr_lr + reclen);
	if (error != 0)
	return (zil_replay_error(zilog, lr, error));
	}

	/*
	* The log block containing this lr may have been byteswapped
	* so that we can easily examine common fields like lrc_txtype.
	* However, the log is a mix of different record types, and only the
	* replay vectors know how to byteswap their records. Therefore, if
	* the lr was byteswapped, undo it before invoking the replay vector.
	*/
	if (zr->zr_byteswap)
	byteswap_uint64_array(zr->zr_lr, reclen);

	/*
	* We must now do two things atomically: replay this log record,
	* and update the log header sequence number to reflect the fact that
	* we did so. At the end of each replay function the sequence number
	* is updated if we are in replay mode.
	*/
	error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, zr->zr_byteswap);
	if (error != 0) {
	/*
	* The DMU's dnode layer doesn't see removes until the txg
	* commits, so a subsequent claim can spuriously fail with
	* EEXIST. So if we receive any error we try syncing out
	* any removes then retry the transaction. Note that we
	* specify B_FALSE for byteswap now, so we don't do it twice.
	*/
	txg_wait_synced(spa_get_dsl(zilog->zl_spa), 0);
	error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, B_FALSE);
	if (error != 0)
	return (zil_replay_error(zilog, lr, error));
	}
	return (0);
	}

	static int
	zil_incr_blks(zilog_t zilog, const blkptr_t bp, void *arg, uint64_t claim_txg)
	{
	(void) bp, (void) arg, (void) claim_txg;

	zilog->zl_replay_blks++;

	return (0);
	}

	/*
	* If this dataset has a non-empty intent log, replay it and destroy it.
	*/
	void
	zil_replay(objset_t os, void arg,
	zil_replay_func_t *const replay_func[TX_MAX_TYPE])
	{
	zilog_t *zilog = dmu_objset_zil(os);
	const zil_header_t *zh = zilog->zl_header;
	zil_replay_arg_t zr;

	if ((zh->zh_flags & ZIL_REPLAY_NEEDED) == 0) {
	zil_destroy(zilog, B_TRUE);
	return;
	}

	zr.zr_replay = replay_func;
	zr.zr_arg = arg;
	zr.zr_byteswap = BP_SHOULD_BYTESWAP(&zh->zh_log);
	zr.zr_lr = vmem_alloc(2 * SPA_MAXBLOCKSIZE, KM_SLEEP);

	/*
	* Wait for in-progress removes to sync before starting replay.
	*/
	txg_wait_synced(zilog->zl_dmu_pool, 0);

	zilog->zl_replay = B_TRUE;
	zilog->zl_replay_time = ddi_get_lbolt();
	ASSERT(zilog->zl_replay_blks == 0);
	(void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr,
	zh->zh_claim_txg, B_TRUE);
	vmem_free(zr.zr_lr, 2 * SPA_MAXBLOCKSIZE);

	zil_destroy(zilog, B_FALSE);
	txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
	zilog->zl_replay = B_FALSE;
	}

	boolean_t
	zil_replaying(zilog_t zilog, dmu_tx_t tx)
	{
	if (zilog->zl_sync == ZFS_SYNC_DISABLED)
	return (B_TRUE);

	if (zilog->zl_replay) {
	dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
	zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] =
	zilog->zl_replaying_seq;
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	int
	zil_reset(const char osname, void arg)
	{
	(void) arg;

	int error = zil_suspend(osname, NULL);
	/* EACCES means crypto key not loaded */
	if ((error == EACCES) \|\| (error == EBUSY))
	return (SET_ERROR(error));
	if (error != 0)
	return (SET_ERROR(EEXIST));
	return (0);
	}

	EXPORT_SYMBOL(zil_alloc);
	EXPORT_SYMBOL(zil_free);
	EXPORT_SYMBOL(zil_open);
	EXPORT_SYMBOL(zil_close);
	EXPORT_SYMBOL(zil_replay);
	EXPORT_SYMBOL(zil_replaying);
	EXPORT_SYMBOL(zil_destroy);
	EXPORT_SYMBOL(zil_destroy_sync);
	EXPORT_SYMBOL(zil_itx_create);
	EXPORT_SYMBOL(zil_itx_destroy);
	EXPORT_SYMBOL(zil_itx_assign);
	EXPORT_SYMBOL(zil_commit);
	EXPORT_SYMBOL(zil_claim);
	EXPORT_SYMBOL(zil_check_log_chain);
	EXPORT_SYMBOL(zil_sync);
	EXPORT_SYMBOL(zil_clean);
	EXPORT_SYMBOL(zil_suspend);
	EXPORT_SYMBOL(zil_resume);
	EXPORT_SYMBOL(zil_lwb_add_block);
	EXPORT_SYMBOL(zil_bp_tree_add);
	EXPORT_SYMBOL(zil_set_sync);
	EXPORT_SYMBOL(zil_set_logbias);
	EXPORT_SYMBOL(zil_sums_init);
	EXPORT_SYMBOL(zil_sums_fini);
	EXPORT_SYMBOL(zil_kstat_values_update);

	-ZFS_MODULE_PARAM(zfs, zfs_, commit_timeout_pct, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, commit_timeout_pct, UINT, ZMOD_RW,
	"ZIL block open timeout percentage");

	ZFS_MODULE_PARAM(zfs_zil, zil_, replay_disable, INT, ZMOD_RW,
	"Disable intent logging replay");

	ZFS_MODULE_PARAM(zfs_zil, zil_, nocacheflush, INT, ZMOD_RW,
	"Disable ZIL cache flushes");

	ZFS_MODULE_PARAM(zfs_zil, zil_, slog_bulk, ULONG, ZMOD_RW,
	"Limit in bytes slog sync writes per commit");

	-ZFS_MODULE_PARAM(zfs_zil, zil_, maxblocksize, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs_zil, zil_, maxblocksize, UINT, ZMOD_RW,
	"Limit in bytes of ZIL log block size");
	diff --git a/sys/contrib/openzfs/module/zfs/zio.c b/sys/contrib/openzfs/module/zfs/zio.c
	index 7b55450ca906..c2e3c6169fa3 100644
	--- a/sys/contrib/openzfs/module/zfs/zio.c
	+++ b/sys/contrib/openzfs/module/zfs/zio.c
	@@ -1,5067 +1,5073 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2021, Datto, Inc.
	*/

	#include <sys/sysmacros.h>
	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_trim.h>
	#include <sys/zio_impl.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/dmu_objset.h>
	#include <sys/arc.h>
	#include <sys/ddt.h>
	#include <sys/blkptr.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_scan.h>
	#include <sys/metaslab_impl.h>
	#include <sys/time.h>
	#include <sys/trace_zfs.h>
	#include <sys/abd.h>
	#include <sys/dsl_crypt.h>
	#include <cityhash.h>

	/*
	* ==========================================================================
	* I/O type descriptions
	* ==========================================================================
	*/
	const char *const zio_type_name[ZIO_TYPES] = {
	/*
	* Note: Linux kernel thread name length is limited
	* so these names will differ from upstream open zfs.
	*/
	"z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_ioctl", "z_trim"
	};

	int zio_dva_throttle_enabled = B_TRUE;
	static int zio_deadman_log_all = B_FALSE;

	/*
	* ==========================================================================
	* I/O kmem caches
	* ==========================================================================
	*/
	static kmem_cache_t *zio_cache;
	static kmem_cache_t *zio_link_cache;
	kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	static uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	static uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	#endif

	/* Mark IOs as "slow" if they take longer than 30 seconds */
	-static int zio_slow_io_ms = (30 * MILLISEC);
	+static uint_t zio_slow_io_ms = (30 * MILLISEC);

	#define BP_SPANB(indblkshift, level) \
	(((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT)))
	#define COMPARE_META_LEVEL 0x80000000ul
	/*
	* The following actions directly effect the spa's sync-to-convergence logic.
	* The values below define the sync pass when we start performing the action.
	* Care should be taken when changing these values as they directly impact
	* spa_sync() performance. Tuning these values may introduce subtle performance
	* pathologies and should only be done in the context of performance analysis.
	* These tunables will eventually be removed and replaced with #defines once
	* enough analysis has been done to determine optimal values.
	*
	* The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that
	* regular blocks are not deferred.
	*
	* Starting in sync pass 8 (zfs_sync_pass_dont_compress), we disable
	* compression (including of metadata). In practice, we don't have this
	* many sync passes, so this has no effect.
	*
	* The original intent was that disabling compression would help the sync
	* passes to converge. However, in practice disabling compression increases
	* the average number of sync passes, because when we turn compression off, a
	* lot of block's size will change and thus we have to re-allocate (not
	* overwrite) them. It also increases the number of 128KB allocations (e.g.
	* for indirect blocks and spacemaps) because these will not be compressed.
	* The 128K allocations are especially detrimental to performance on highly
	* fragmented systems, which may have very few free segments of this size,
	* and may need to load new metaslabs to satisfy 128K allocations.
	*/
	-int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */
	-static int zfs_sync_pass_dont_compress = 8; /* don't compress s. i. t. p. */
	-static int zfs_sync_pass_rewrite = 2; /* rewrite new bps s. i. t. p. */
	+
	+/* defer frees starting in this pass */
	+uint_t zfs_sync_pass_deferred_free = 2;
	+
	+/* don't compress starting in this pass */
	+static uint_t zfs_sync_pass_dont_compress = 8;
	+
	+/* rewrite new bps starting in this pass */
	+static uint_t zfs_sync_pass_rewrite = 2;

	/*
	* An allocating zio is one that either currently has the DVA allocate
	* stage set or will have it later in its lifetime.
	*/
	#define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE)

	/*
	* Enable smaller cores by excluding metadata
	* allocations as well.
	*/
	int zio_exclude_metadata = 0;
	static int zio_requeue_io_start_cut_in_line = 1;

	#ifdef ZFS_DEBUG
	static const int zio_buf_debug_limit = 16384;
	#else
	static const int zio_buf_debug_limit = 0;
	#endif

	static inline void __zio_execute(zio_t *zio);

	static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t);

	void
	zio_init(void)
	{
	size_t c;

	zio_cache = kmem_cache_create("zio_cache",
	sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
	zio_link_cache = kmem_cache_create("zio_link_cache",
	sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0);

	/*
	* For small buffers, we want a cache for each multiple of
	* SPA_MINBLOCKSIZE. For larger buffers, we want a cache
	* for each quarter-power of 2.
	*/
	for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
	size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
	size_t p2 = size;
	size_t align = 0;
	size_t data_cflags, cflags;

	data_cflags = KMC_NODEBUG;
	cflags = (zio_exclude_metadata \|\| size > zio_buf_debug_limit) ?
	KMC_NODEBUG : 0;

	while (!ISP2(p2))
	p2 &= p2 - 1;

	#ifndef _KERNEL
	/*
	* If we are using watchpoints, put each buffer on its own page,
	* to eliminate the performance overhead of trapping to the
	* kernel when modifying a non-watched buffer that shares the
	* page with a watched buffer.
	*/
	if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
	continue;
	/*
	* Here's the problem - on 4K native devices in userland on
	* Linux using O_DIRECT, buffers must be 4K aligned or I/O
	* will fail with EINVAL, causing zdb (and others) to coredump.
	* Since userland probably doesn't need optimized buffer caches,
	* we just force 4K alignment on everything.
	*/
	align = 8 * SPA_MINBLOCKSIZE;
	#else
	if (size < PAGESIZE) {
	align = SPA_MINBLOCKSIZE;
	} else if (IS_P2ALIGNED(size, p2 >> 2)) {
	align = PAGESIZE;
	}
	#endif

	if (align != 0) {
	char name[36];
	if (cflags == data_cflags) {
	/*
	* Resulting kmem caches would be identical.
	* Save memory by creating only one.
	*/
	(void) snprintf(name, sizeof (name),
	"zio_buf_comb_%lu", (ulong_t)size);
	zio_buf_cache[c] = kmem_cache_create(name,
	size, align, NULL, NULL, NULL, NULL, NULL,
	cflags);
	zio_data_buf_cache[c] = zio_buf_cache[c];
	continue;
	}
	(void) snprintf(name, sizeof (name), "zio_buf_%lu",
	(ulong_t)size);
	zio_buf_cache[c] = kmem_cache_create(name, size,
	align, NULL, NULL, NULL, NULL, NULL, cflags);

	(void) snprintf(name, sizeof (name), "zio_data_buf_%lu",
	(ulong_t)size);
	zio_data_buf_cache[c] = kmem_cache_create(name, size,
	align, NULL, NULL, NULL, NULL, NULL, data_cflags);
	}
	}

	while (--c != 0) {
	ASSERT(zio_buf_cache[c] != NULL);
	if (zio_buf_cache[c - 1] == NULL)
	zio_buf_cache[c - 1] = zio_buf_cache[c];

	ASSERT(zio_data_buf_cache[c] != NULL);
	if (zio_data_buf_cache[c - 1] == NULL)
	zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
	}

	zio_inject_init();

	lz4_init();
	}

	void
	zio_fini(void)
	{
	size_t n = SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT;

	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	for (size_t i = 0; i < n; i++) {
	if (zio_buf_cache_allocs[i] != zio_buf_cache_frees[i])
	(void) printf("zio_fini: [%d] %llu != %llu\n",
	(int)((i + 1) << SPA_MINBLOCKSHIFT),
	(long long unsigned)zio_buf_cache_allocs[i],
	(long long unsigned)zio_buf_cache_frees[i]);
	}
	#endif

	/*
	* The same kmem cache can show up multiple times in both zio_buf_cache
	* and zio_data_buf_cache. Do a wasteful but trivially correct scan to
	* sort it out.
	*/
	for (size_t i = 0; i < n; i++) {
	kmem_cache_t *cache = zio_buf_cache[i];
	if (cache == NULL)
	continue;
	for (size_t j = i; j < n; j++) {
	if (cache == zio_buf_cache[j])
	zio_buf_cache[j] = NULL;
	if (cache == zio_data_buf_cache[j])
	zio_data_buf_cache[j] = NULL;
	}
	kmem_cache_destroy(cache);
	}

	for (size_t i = 0; i < n; i++) {
	kmem_cache_t *cache = zio_data_buf_cache[i];
	if (cache == NULL)
	continue;
	for (size_t j = i; j < n; j++) {
	if (cache == zio_data_buf_cache[j])
	zio_data_buf_cache[j] = NULL;
	}
	kmem_cache_destroy(cache);
	}

	for (size_t i = 0; i < n; i++) {
	VERIFY3P(zio_buf_cache[i], ==, NULL);
	VERIFY3P(zio_data_buf_cache[i], ==, NULL);
	}

	kmem_cache_destroy(zio_link_cache);
	kmem_cache_destroy(zio_cache);

	zio_inject_fini();

	lz4_fini();
	}

	/*
	* ==========================================================================
	* Allocate and free I/O buffers
	* ==========================================================================
	*/

	/*
	* Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a
	* crashdump if the kernel panics, so use it judiciously. Obviously, it's
	* useful to inspect ZFS metadata, but if possible, we should avoid keeping
	* excess / transient data in-core during a crashdump.
	*/
	void *
	zio_buf_alloc(size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_allocs[c], 1);
	#endif

	return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
	}

	/*
	* Use zio_data_buf_alloc to allocate data. The data will not appear in a
	* crashdump if the kernel panics. This exists so that we will limit the amount
	* of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount
	* of kernel heap dumped to disk when the kernel panics)
	*/
	void *
	zio_data_buf_alloc(size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
	}

	void
	zio_buf_free(void *buf, size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_frees[c], 1);
	#endif

	kmem_cache_free(zio_buf_cache[c], buf);
	}

	void
	zio_data_buf_free(void *buf, size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	kmem_cache_free(zio_data_buf_cache[c], buf);
	}

	static void
	zio_abd_free(void *abd, size_t size)
	{
	(void) size;
	abd_free((abd_t *)abd);
	}

	/*
	* ==========================================================================
	* Push and pop I/O transform buffers
	* ==========================================================================
	*/
	void
	zio_push_transform(zio_t zio, abd_t data, uint64_t size, uint64_t bufsize,
	zio_transform_func_t *transform)
	{
	zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP);

	zt->zt_orig_abd = zio->io_abd;
	zt->zt_orig_size = zio->io_size;
	zt->zt_bufsize = bufsize;
	zt->zt_transform = transform;

	zt->zt_next = zio->io_transform_stack;
	zio->io_transform_stack = zt;

	zio->io_abd = data;
	zio->io_size = size;
	}

	void
	zio_pop_transforms(zio_t *zio)
	{
	zio_transform_t *zt;

	while ((zt = zio->io_transform_stack) != NULL) {
	if (zt->zt_transform != NULL)
	zt->zt_transform(zio,
	zt->zt_orig_abd, zt->zt_orig_size);

	if (zt->zt_bufsize != 0)
	abd_free(zio->io_abd);

	zio->io_abd = zt->zt_orig_abd;
	zio->io_size = zt->zt_orig_size;
	zio->io_transform_stack = zt->zt_next;

	kmem_free(zt, sizeof (zio_transform_t));
	}
	}

	/*
	* ==========================================================================
	* I/O transform callbacks for subblocks, decompression, and decryption
	* ==========================================================================
	*/
	static void
	zio_subblock(zio_t zio, abd_t data, uint64_t size)
	{
	ASSERT(zio->io_size > size);

	if (zio->io_type == ZIO_TYPE_READ)
	abd_copy(data, zio->io_abd, size);
	}

	static void
	zio_decompress(zio_t zio, abd_t data, uint64_t size)
	{
	if (zio->io_error == 0) {
	void *tmp = abd_borrow_buf(data, size);
	int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
	zio->io_abd, tmp, zio->io_size, size,
	&zio->io_prop.zp_complevel);
	abd_return_buf_copy(data, tmp, size);

	if (zio_injection_enabled && ret == 0)
	ret = zio_handle_fault_injection(zio, EINVAL);

	if (ret != 0)
	zio->io_error = SET_ERROR(EIO);
	}
	}

	static void
	zio_decrypt(zio_t zio, abd_t data, uint64_t size)
	{
	int ret;
	void *tmp;
	blkptr_t *bp = zio->io_bp;
	spa_t *spa = zio->io_spa;
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	uint64_t lsize = BP_GET_LSIZE(bp);
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	ASSERT(BP_USES_CRYPT(bp));
	ASSERT3U(size, !=, 0);

	if (zio->io_error != 0)
	return;

	/*
	* Verify the cksum of MACs stored in an indirect bp. It will always
	* be possible to verify this since it does not require an encryption
	* key.
	*/
	if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {
	zio_crypt_decode_mac_bp(bp, mac);

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
	/*
	* We haven't decompressed the data yet, but
	* zio_crypt_do_indirect_mac_checksum() requires
	* decompressed data to be able to parse out the MACs
	* from the indirect block. We decompress it now and
	* throw away the result after we are finished.
	*/
	tmp = zio_buf_alloc(lsize);
	ret = zio_decompress_data(BP_GET_COMPRESS(bp),
	zio->io_abd, tmp, zio->io_size, lsize,
	&zio->io_prop.zp_complevel);
	if (ret != 0) {
	ret = SET_ERROR(EIO);
	goto error;
	}
	ret = zio_crypt_do_indirect_mac_checksum(B_FALSE,
	tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac);
	zio_buf_free(tmp, lsize);
	} else {
	ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE,
	zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac);
	}
	abd_copy(data, zio->io_abd, size);

	if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) {
	ret = zio_handle_decrypt_injection(spa,
	&zio->io_bookmark, ot, ECKSUM);
	}
	if (ret != 0)
	goto error;

	return;
	}

	/*
	* If this is an authenticated block, just check the MAC. It would be
	* nice to separate this out into its own flag, but for the moment
	* enum zio_flag is out of bits.
	*/
	if (BP_IS_AUTHENTICATED(bp)) {
	if (ot == DMU_OT_OBJSET) {
	ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa,
	dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp));
	} else {
	zio_crypt_decode_mac_bp(bp, mac);
	ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj,
	zio->io_abd, size, mac);
	if (zio_injection_enabled && ret == 0) {
	ret = zio_handle_decrypt_injection(spa,
	&zio->io_bookmark, ot, ECKSUM);
	}
	}
	abd_copy(data, zio->io_abd, size);

	if (ret != 0)
	goto error;

	return;
	}

	zio_crypt_decode_params_bp(bp, salt, iv);

	if (ot == DMU_OT_INTENT_LOG) {
	tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t));
	zio_crypt_decode_mac_zil(tmp, mac);
	abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t));
	} else {
	zio_crypt_decode_mac_bp(bp, mac);
	}

	ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp),
	BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data,
	zio->io_abd, &no_crypt);
	if (no_crypt)
	abd_copy(data, zio->io_abd, size);

	if (ret != 0)
	goto error;

	return;

	error:
	/* assert that the key was found unless this was speculative */
	ASSERT(ret != EACCES \|\| (zio->io_flags & ZIO_FLAG_SPECULATIVE));

	/*
	* If there was a decryption / authentication error return EIO as
	* the io_error. If this was not a speculative zio, create an ereport.
	*/
	if (ret == ECKSUM) {
	zio->io_error = SET_ERROR(EIO);
	if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(spa, &zio->io_bookmark);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, &zio->io_bookmark, zio, 0);
	}
	} else {
	zio->io_error = ret;
	}
	}

	/*
	* ==========================================================================
	* I/O parent/child relationships and pipeline interlocks
	* ==========================================================================
	*/
	zio_t *
	zio_walk_parents(zio_t cio, zio_link_t *zl)
	{
	list_t *pl = &cio->io_parent_list;

	zl = (zl == NULL) ? list_head(pl) : list_next(pl, *zl);
	if (*zl == NULL)
	return (NULL);

	ASSERT((*zl)->zl_child == cio);
	return ((*zl)->zl_parent);
	}

	zio_t *
	zio_walk_children(zio_t pio, zio_link_t *zl)
	{
	list_t *cl = &pio->io_child_list;

	ASSERT(MUTEX_HELD(&pio->io_lock));

	zl = (zl == NULL) ? list_head(cl) : list_next(cl, *zl);
	if (*zl == NULL)
	return (NULL);

	ASSERT((*zl)->zl_parent == pio);
	return ((*zl)->zl_child);
	}

	zio_t *
	zio_unique_parent(zio_t *cio)
	{
	zio_link_t *zl = NULL;
	zio_t *pio = zio_walk_parents(cio, &zl);

	VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL);
	return (pio);
	}

	void
	zio_add_child(zio_t pio, zio_t cio)
	{
	zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP);

	/*
	* Logical I/Os can have logical, gang, or vdev children.
	* Gang I/Os can have gang or vdev children.
	* Vdev I/Os can only have vdev children.
	* The following ASSERT captures all of these constraints.
	*/
	ASSERT3S(cio->io_child_type, <=, pio->io_child_type);

	zl->zl_parent = pio;
	zl->zl_child = cio;

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0);

	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	pio->io_children[cio->io_child_type][w] += !cio->io_state[w];

	list_insert_head(&pio->io_child_list, zl);
	list_insert_head(&cio->io_parent_list, zl);

	pio->io_child_count++;
	cio->io_parent_count++;

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
	}

	static void
	zio_remove_child(zio_t pio, zio_t cio, zio_link_t *zl)
	{
	ASSERT(zl->zl_parent == pio);
	ASSERT(zl->zl_child == cio);

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	list_remove(&pio->io_child_list, zl);
	list_remove(&cio->io_parent_list, zl);

	pio->io_child_count--;
	cio->io_parent_count--;

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
	kmem_cache_free(zio_link_cache, zl);
	}

	static boolean_t
	zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait)
	{
	boolean_t waiting = B_FALSE;

	mutex_enter(&zio->io_lock);
	ASSERT(zio->io_stall == NULL);
	for (int c = 0; c < ZIO_CHILD_TYPES; c++) {
	if (!(ZIO_CHILD_BIT_IS_SET(childbits, c)))
	continue;

	uint64_t *countp = &zio->io_children[c][wait];
	if (*countp != 0) {
	zio->io_stage >>= 1;
	ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN);
	zio->io_stall = countp;
	waiting = B_TRUE;
	break;
	}
	}
	mutex_exit(&zio->io_lock);
	return (waiting);
	}

	__attribute__((always_inline))
	static inline void
	zio_notify_parent(zio_t pio, zio_t zio, enum zio_wait_type wait,
	zio_t **next_to_executep)
	{
	uint64_t *countp = &pio->io_children[zio->io_child_type][wait];
	int *errorp = &pio->io_child_error[zio->io_child_type];

	mutex_enter(&pio->io_lock);
	if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
	errorp = zio_worst_error(errorp, zio->io_error);
	pio->io_reexecute \|= zio->io_reexecute;
	ASSERT3U(*countp, >, 0);

	(*countp)--;

	if (*countp == 0 && pio->io_stall == countp) {
	zio_taskq_type_t type =
	pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE :
	ZIO_TASKQ_INTERRUPT;
	pio->io_stall = NULL;
	mutex_exit(&pio->io_lock);

	/*
	* If we can tell the caller to execute this parent next, do
	* so. Otherwise dispatch the parent zio as its own task.
	*
	* Having the caller execute the parent when possible reduces
	* locking on the zio taskq's, reduces context switch
	* overhead, and has no recursion penalty. Note that one
	* read from disk typically causes at least 3 zio's: a
	* zio_null(), the logical zio_read(), and then a physical
	* zio. When the physical ZIO completes, we are able to call
	* zio_done() on all 3 of these zio's from one invocation of
	* zio_execute() by returning the parent back to
	* zio_execute(). Since the parent isn't executed until this
	* thread returns back to zio_execute(), the caller should do
	* so promptly.
	*
	* In other cases, dispatching the parent prevents
	* overflowing the stack when we have deeply nested
	* parent-child relationships, as we do with the "mega zio"
	* of writes for spa_sync(), and the chain of ZIL blocks.
	*/
	if (next_to_executep != NULL && *next_to_executep == NULL) {
	*next_to_executep = pio;
	} else {
	zio_taskq_dispatch(pio, type, B_FALSE);
	}
	} else {
	mutex_exit(&pio->io_lock);
	}
	}

	static void
	zio_inherit_child_errors(zio_t *zio, enum zio_child c)
	{
	if (zio->io_child_error[c] != 0 && zio->io_error == 0)
	zio->io_error = zio->io_child_error[c];
	}

	int
	zio_bookmark_compare(const void x1, const void x2)
	{
	const zio_t *z1 = x1;
	const zio_t *z2 = x2;

	if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset)
	return (-1);
	if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset)
	return (1);

	if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object)
	return (-1);
	if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object)
	return (1);

	if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level)
	return (-1);
	if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level)
	return (1);

	if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid)
	return (-1);
	if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid)
	return (1);

	if (z1 < z2)
	return (-1);
	if (z1 > z2)
	return (1);

	return (0);
	}

	/*
	* ==========================================================================
	* Create the various types of I/O (read, write, free, etc)
	* ==========================================================================
	*/
	static zio_t *
	zio_create(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	abd_t data, uint64_t lsize, uint64_t psize, zio_done_func_t done,
	void *private, zio_type_t type, zio_priority_t priority,
	enum zio_flag flags, vdev_t *vd, uint64_t offset,
	const zbookmark_phys_t *zb, enum zio_stage stage,
	enum zio_stage pipeline)
	{
	zio_t *zio;

	IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE);
	ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0);
	ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0);

	ASSERT(!vd \|\| spa_config_held(spa, SCL_STATE_ALL, RW_READER));
	ASSERT(!bp \|\| !(flags & ZIO_FLAG_CONFIG_WRITER));
	ASSERT(vd \|\| stage == ZIO_STAGE_OPEN);

	IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0);

	zio = kmem_cache_alloc(zio_cache, KM_SLEEP);
	memset(zio, 0, sizeof (zio_t));

	mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL);

	list_create(&zio->io_parent_list, sizeof (zio_link_t),
	offsetof(zio_link_t, zl_parent_node));
	list_create(&zio->io_child_list, sizeof (zio_link_t),
	offsetof(zio_link_t, zl_child_node));
	metaslab_trace_init(&zio->io_alloc_list);

	if (vd != NULL)
	zio->io_child_type = ZIO_CHILD_VDEV;
	else if (flags & ZIO_FLAG_GANG_CHILD)
	zio->io_child_type = ZIO_CHILD_GANG;
	else if (flags & ZIO_FLAG_DDT_CHILD)
	zio->io_child_type = ZIO_CHILD_DDT;
	else
	zio->io_child_type = ZIO_CHILD_LOGICAL;

	if (bp != NULL) {
	zio->io_bp = (blkptr_t *)bp;
	zio->io_bp_copy = *bp;
	zio->io_bp_orig = *bp;
	if (type != ZIO_TYPE_WRITE \|\|
	zio->io_child_type == ZIO_CHILD_DDT)
	zio->io_bp = &zio->io_bp_copy; /* so caller can free */
	if (zio->io_child_type == ZIO_CHILD_LOGICAL)
	zio->io_logical = zio;
	if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp))
	pipeline \|= ZIO_GANG_STAGES;
	}

	zio->io_spa = spa;
	zio->io_txg = txg;
	zio->io_done = done;
	zio->io_private = private;
	zio->io_type = type;
	zio->io_priority = priority;
	zio->io_vd = vd;
	zio->io_offset = offset;
	zio->io_orig_abd = zio->io_abd = data;
	zio->io_orig_size = zio->io_size = psize;
	zio->io_lsize = lsize;
	zio->io_orig_flags = zio->io_flags = flags;
	zio->io_orig_stage = zio->io_stage = stage;
	zio->io_orig_pipeline = zio->io_pipeline = pipeline;
	zio->io_pipeline_trace = ZIO_STAGE_OPEN;

	zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY);
	zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE);

	if (zb != NULL)
	zio->io_bookmark = *zb;

	if (pio != NULL) {
	zio->io_metaslab_class = pio->io_metaslab_class;
	if (zio->io_logical == NULL)
	zio->io_logical = pio->io_logical;
	if (zio->io_child_type == ZIO_CHILD_GANG)
	zio->io_gang_leader = pio->io_gang_leader;
	zio_add_child(pio, zio);
	}

	taskq_init_ent(&zio->io_tqent);

	return (zio);
	}

	void
	zio_destroy(zio_t *zio)
	{
	metaslab_trace_fini(&zio->io_alloc_list);
	list_destroy(&zio->io_parent_list);
	list_destroy(&zio->io_child_list);
	mutex_destroy(&zio->io_lock);
	cv_destroy(&zio->io_cv);
	kmem_cache_free(zio_cache, zio);
	}

	zio_t *
	zio_null(zio_t pio, spa_t spa, vdev_t vd, zio_done_func_t done,
	void *private, enum zio_flag flags)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
	ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
	ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE);

	return (zio);
	}

	zio_t *
	zio_root(spa_t spa, zio_done_func_t done, void *private, enum zio_flag flags)
	{
	return (zio_null(NULL, spa, NULL, done, private, flags));
	}

	static int
	zfs_blkptr_verify_log(spa_t spa, const blkptr_t bp,
	enum blk_verify_flag blk_verify, const char *fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	switch (blk_verify) {
	case BLK_VERIFY_HALT:
	dprintf_bp(bp, "blkptr at %p dprintf_bp():", bp);
	zfs_panic_recover("%s: %s", spa_name(spa), buf);
	break;
	case BLK_VERIFY_LOG:
	zfs_dbgmsg("%s: %s", spa_name(spa), buf);
	break;
	case BLK_VERIFY_ONLY:
	break;
	}

	return (1);
	}

	/*
	* Verify the block pointer fields contain reasonable values. This means
	* it only contains known object types, checksum/compression identifiers,
	* block sizes within the maximum allowed limits, valid DVAs, etc.
	*
	* If everything checks out B_TRUE is returned. The zfs_blkptr_verify
	* argument controls the behavior when an invalid field is detected.
	*
	* Modes for zfs_blkptr_verify:
	* 1) BLK_VERIFY_ONLY (evaluate the block)
	* 2) BLK_VERIFY_LOG (evaluate the block and log problems)
	* 3) BLK_VERIFY_HALT (call zfs_panic_recover on error)
	*/
	boolean_t
	zfs_blkptr_verify(spa_t spa, const blkptr_t bp, boolean_t config_held,
	enum blk_verify_flag blk_verify)
	{
	int errors = 0;

	if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid TYPE %llu",
	bp, (longlong_t)BP_GET_TYPE(bp));
	}
	if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid CHECKSUM %llu",
	bp, (longlong_t)BP_GET_CHECKSUM(bp));
	}
	if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid COMPRESS %llu",
	bp, (longlong_t)BP_GET_COMPRESS(bp));
	}
	if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid LSIZE %llu",
	bp, (longlong_t)BP_GET_LSIZE(bp));
	}
	if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid PSIZE %llu",
	bp, (longlong_t)BP_GET_PSIZE(bp));
	}

	if (BP_IS_EMBEDDED(bp)) {
	if (BPE_GET_ETYPE(bp) >= NUM_BP_EMBEDDED_TYPES) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p has invalid ETYPE %llu",
	bp, (longlong_t)BPE_GET_ETYPE(bp));
	}
	}

	/*
	* Do not verify individual DVAs if the config is not trusted. This
	* will be done once the zio is executed in vdev_mirror_map_alloc.
	*/
	if (!spa->spa_trust_config)
	return (errors == 0);

	if (!config_held)
	spa_config_enter(spa, SCL_VDEV, bp, RW_READER);
	else
	ASSERT(spa_config_held(spa, SCL_VDEV, RW_WRITER));
	/*
	* Pool-specific checks.
	*
	* Note: it would be nice to verify that the blk_birth and
	* BP_PHYSICAL_BIRTH() are not too large. However, spa_freeze()
	* allows the birth time of log blocks (and dmu_sync()-ed blocks
	* that are in the log) to be arbitrarily large.
	*/
	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
	const dva_t *dva = &bp->blk_dva[i];
	uint64_t vdevid = DVA_GET_VDEV(dva);

	if (vdevid >= spa->spa_root_vdev->vdev_children) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p DVA %u has invalid VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
	if (vd == NULL) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p DVA %u has invalid VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	if (vd->vdev_ops == &vdev_hole_ops) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p DVA %u has hole VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	if (vd->vdev_ops == &vdev_missing_ops) {
	/*
	* "missing" vdevs are valid during import, but we
	* don't have their detailed info (e.g. asize), so
	* we can't perform any more checks on them.
	*/
	continue;
	}
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t asize = DVA_GET_ASIZE(dva);
	if (DVA_GET_GANG(dva))
	asize = vdev_gang_header_asize(vd);
	if (offset + asize > vd->vdev_asize) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %p DVA %u has invalid OFFSET %llu",
	bp, i, (longlong_t)offset);
	}
	}
	if (errors > 0)
	dprintf_bp(bp, "blkptr at %p dprintf_bp():", bp);
	if (!config_held)
	spa_config_exit(spa, SCL_VDEV, bp);

	return (errors == 0);
	}

	boolean_t
	zfs_dva_valid(spa_t spa, const dva_t dva, const blkptr_t *bp)
	{
	(void) bp;
	uint64_t vdevid = DVA_GET_VDEV(dva);

	if (vdevid >= spa->spa_root_vdev->vdev_children)
	return (B_FALSE);

	vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
	if (vd == NULL)
	return (B_FALSE);

	if (vd->vdev_ops == &vdev_hole_ops)
	return (B_FALSE);

	if (vd->vdev_ops == &vdev_missing_ops) {
	return (B_FALSE);
	}

	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t asize = DVA_GET_ASIZE(dva);

	if (DVA_GET_GANG(dva))
	asize = vdev_gang_header_asize(vd);
	if (offset + asize > vd->vdev_asize)
	return (B_FALSE);

	return (B_TRUE);
	}

	zio_t *
	zio_read(zio_t pio, spa_t spa, const blkptr_t *bp,
	abd_t data, uint64_t size, zio_done_func_t done, void *private,
	zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp,
	data, size, size, done, private,
	ZIO_TYPE_READ, priority, flags, NULL, 0, zb,
	ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE);

	return (zio);
	}

	zio_t *
	zio_write(zio_t pio, spa_t spa, uint64_t txg, blkptr_t *bp,
	abd_t data, uint64_t lsize, uint64_t psize, const zio_prop_t zp,
	zio_done_func_t ready, zio_done_func_t children_ready,
	zio_done_func_t physdone, zio_done_func_t done,
	void *private, zio_priority_t priority, enum zio_flag flags,
	const zbookmark_phys_t *zb)
	{
	zio_t *zio;

	ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF &&
	zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
	zp->zp_compress >= ZIO_COMPRESS_OFF &&
	zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
	DMU_OT_IS_VALID(zp->zp_type) &&
	zp->zp_level < 32 &&
	zp->zp_copies > 0 &&
	zp->zp_copies <= spa_max_replication(spa));

	zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private,
	ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
	ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE);

	zio->io_ready = ready;
	zio->io_children_ready = children_ready;
	zio->io_physdone = physdone;
	zio->io_prop = *zp;

	/*
	* Data can be NULL if we are going to call zio_write_override() to
	* provide the already-allocated BP. But we may need the data to
	* verify a dedup hit (if requested). In this case, don't try to
	* dedup (just take the already-allocated BP verbatim). Encrypted
	* dedup blocks need data as well so we also disable dedup in this
	* case.
	*/
	if (data == NULL &&
	(zio->io_prop.zp_dedup_verify \|\| zio->io_prop.zp_encrypt)) {
	zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE;
	}

	return (zio);
	}

	zio_t *
	zio_rewrite(zio_t pio, spa_t spa, uint64_t txg, blkptr_t bp, abd_t data,
	uint64_t size, zio_done_func_t done, void private,
	zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, txg, bp, data, size, size, done, private,
	ZIO_TYPE_WRITE, priority, flags \| ZIO_FLAG_IO_REWRITE, NULL, 0, zb,
	ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE);

	return (zio);
	}

	void
	zio_write_override(zio_t zio, blkptr_t bp, int copies, boolean_t nopwrite)
	{
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
	ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa));

	/*
	* We must reset the io_prop to match the values that existed
	* when the bp was first written by dmu_sync() keeping in mind
	* that nopwrite and dedup are mutually exclusive.
	*/
	zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup;
	zio->io_prop.zp_nopwrite = nopwrite;
	zio->io_prop.zp_copies = copies;
	zio->io_bp_override = bp;
	}

	void
	zio_free(spa_t spa, uint64_t txg, const blkptr_t bp)
	{

	(void) zfs_blkptr_verify(spa, bp, B_FALSE, BLK_VERIFY_HALT);

	/*
	* The check for EMBEDDED is a performance optimization. We
	* process the free here (by ignoring it) rather than
	* putting it on the list and then processing it in zio_free_sync().
	*/
	if (BP_IS_EMBEDDED(bp))
	return;
	metaslab_check_free(spa, bp);

	/*
	* Frees that are for the currently-syncing txg, are not going to be
	* deferred, and which will not need to do a read (i.e. not GANG or
	* DEDUP), can be processed immediately. Otherwise, put them on the
	* in-memory list for later processing.
	*
	* Note that we only defer frees after zfs_sync_pass_deferred_free
	* when the log space map feature is disabled. [see relevant comment
	* in spa_sync_iterate_to_convergence()]
	*/
	if (BP_IS_GANG(bp) \|\|
	BP_GET_DEDUP(bp) \|\|
	txg != spa->spa_syncing_txg \|\|
	(spa_sync_pass(spa) >= zfs_sync_pass_deferred_free &&
	!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))) {
	bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
	} else {
	VERIFY3P(zio_free_sync(NULL, spa, txg, bp, 0), ==, NULL);
	}
	}

	/*
	* To improve performance, this function may return NULL if we were able
	* to do the free immediately. This avoids the cost of creating a zio
	* (and linking it to the parent, etc).
	*/
	zio_t *
	zio_free_sync(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	enum zio_flag flags)
	{
	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(spa_syncing_txg(spa) == txg);

	if (BP_IS_EMBEDDED(bp))
	return (NULL);

	metaslab_check_free(spa, bp);
	arc_freed(spa, bp);
	dsl_scan_freed(spa, bp);

	if (BP_IS_GANG(bp) \|\| BP_GET_DEDUP(bp)) {
	/*
	* GANG and DEDUP blocks can induce a read (for the gang block
	* header, or the DDT), so issue them asynchronously so that
	* this thread is not tied up.
	*/
	enum zio_stage stage =
	ZIO_FREE_PIPELINE \| ZIO_STAGE_ISSUE_ASYNC;

	return (zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	BP_GET_PSIZE(bp), NULL, NULL,
	ZIO_TYPE_FREE, ZIO_PRIORITY_NOW,
	flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage));
	} else {
	metaslab_free(spa, bp, txg, B_FALSE);
	return (NULL);
	}
	}

	zio_t *
	zio_claim(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	zio_done_func_t done, void private, enum zio_flag flags)
	{
	zio_t *zio;

	(void) zfs_blkptr_verify(spa, bp, flags & ZIO_FLAG_CONFIG_WRITER,
	BLK_VERIFY_HALT);

	if (BP_IS_EMBEDDED(bp))
	return (zio_null(pio, spa, NULL, NULL, NULL, 0));

	/*
	* A claim is an allocation of a specific block. Claims are needed
	* to support immediate writes in the intent log. The issue is that
	* immediate writes contain committed data, but in a txg that was
	* not committed. Upon opening the pool after an unclean shutdown,
	* the intent log claims all blocks that contain immediate write data
	* so that the SPA knows they're in use.
	*
	* All claims must be resolved in the first txg -- before the SPA
	* starts allocating blocks -- so that nothing is allocated twice.
	* If txg == 0 we just verify that the block is claimable.
	*/
	ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <,
	spa_min_claim_txg(spa));
	ASSERT(txg == spa_min_claim_txg(spa) \|\| txg == 0);
	ASSERT(!BP_GET_DEDUP(bp) \|\| !spa_writeable(spa)); /* zdb(8) */

	zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW,
	flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE);
	ASSERT0(zio->io_queued_timestamp);

	return (zio);
	}

	zio_t *
	zio_ioctl(zio_t pio, spa_t spa, vdev_t *vd, int cmd,
	zio_done_func_t done, void private, enum zio_flag flags)
	{
	zio_t *zio;
	int c;

	if (vd->vdev_children == 0) {
	zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
	ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
	ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE);

	zio->io_cmd = cmd;
	} else {
	zio = zio_null(pio, spa, NULL, NULL, NULL, flags);

	for (c = 0; c < vd->vdev_children; c++)
	zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd,
	done, private, flags));
	}

	return (zio);
	}

	zio_t *
	zio_trim(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	zio_done_func_t done, void private, zio_priority_t priority,
	enum zio_flag flags, enum trim_flag trim_flags)
	{
	zio_t *zio;

	ASSERT0(vd->vdev_children);
	ASSERT0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	ASSERT0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	ASSERT3U(size, !=, 0);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, NULL, size, size, done,
	private, ZIO_TYPE_TRIM, priority, flags \| ZIO_FLAG_PHYSICAL,
	vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_TRIM_PIPELINE);
	zio->io_trim_flags = trim_flags;

	return (zio);
	}

	zio_t *
	zio_read_phys(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	abd_t data, int checksum, zio_done_func_t done, void *private,
	zio_priority_t priority, enum zio_flag flags, boolean_t labels)
	{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels \|\| offset + size <= VDEV_LABEL_START_SIZE \|\|
	offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	private, ZIO_TYPE_READ, priority, flags \| ZIO_FLAG_PHYSICAL, vd,
	offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	return (zio);
	}

	zio_t *
	zio_write_phys(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	abd_t data, int checksum, zio_done_func_t done, void *private,
	zio_priority_t priority, enum zio_flag flags, boolean_t labels)
	{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels \|\| offset + size <= VDEV_LABEL_START_SIZE \|\|
	offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	private, ZIO_TYPE_WRITE, priority, flags \| ZIO_FLAG_PHYSICAL, vd,
	offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
	/*
	* zec checksums are necessarily destructive -- they modify
	* the end of the write buffer to hold the verifier/checksum.
	* Therefore, we must make a local copy in case the data is
	* being written to multiple places in parallel.
	*/
	abd_t *wbuf = abd_alloc_sametype(data, size);
	abd_copy(wbuf, data, size);

	zio_push_transform(zio, wbuf, size, size, NULL);
	}

	return (zio);
	}

	/*
	* Create a child I/O to do some work for us.
	*/
	zio_t *
	zio_vdev_child_io(zio_t pio, blkptr_t bp, vdev_t *vd, uint64_t offset,
	abd_t *data, uint64_t size, int type, zio_priority_t priority,
	enum zio_flag flags, zio_done_func_t done, void private)
	{
	enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE;
	zio_t *zio;

	/*
	* vdev child I/Os do not propagate their error to the parent.
	* Therefore, for correct operation the caller must check for
	* and handle the error in the child i/o's done callback.
	* The only exceptions are i/os that we don't care about
	* (OPTIONAL or REPAIR).
	*/
	ASSERT((flags & ZIO_FLAG_OPTIONAL) \|\| (flags & ZIO_FLAG_IO_REPAIR) \|\|
	done != NULL);

	if (type == ZIO_TYPE_READ && bp != NULL) {
	/*
	* If we have the bp, then the child should perform the
	* checksum and the parent need not. This pushes error
	* detection as close to the leaves as possible and
	* eliminates redundant checksums in the interior nodes.
	*/
	pipeline \|= ZIO_STAGE_CHECKSUM_VERIFY;
	pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
	}

	if (vd->vdev_ops->vdev_op_leaf) {
	ASSERT0(vd->vdev_children);
	offset += VDEV_LABEL_START_SIZE;
	}

	flags \|= ZIO_VDEV_CHILD_FLAGS(pio);

	/*
	* If we've decided to do a repair, the write is not speculative --
	* even if the original read was.
	*/
	if (flags & ZIO_FLAG_IO_REPAIR)
	flags &= ~ZIO_FLAG_SPECULATIVE;

	/*
	* If we're creating a child I/O that is not associated with a
	* top-level vdev, then the child zio is not an allocating I/O.
	* If this is a retried I/O then we ignore it since we will
	* have already processed the original allocating I/O.
	*/
	if (flags & ZIO_FLAG_IO_ALLOCATING &&
	(vd != vd->vdev_top \|\| (flags & ZIO_FLAG_IO_RETRY))) {
	ASSERT(pio->io_metaslab_class != NULL);
	ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled);
	ASSERT(type == ZIO_TYPE_WRITE);
	ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(!(flags & ZIO_FLAG_IO_REPAIR));
	ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) \|\|
	pio->io_child_type == ZIO_CHILD_GANG);

	flags &= ~ZIO_FLAG_IO_ALLOCATING;
	}


	zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size,
	done, private, type, priority, flags, vd, offset, &pio->io_bookmark,
	ZIO_STAGE_VDEV_IO_START >> 1, pipeline);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);

	zio->io_physdone = pio->io_physdone;
	if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL)
	zio->io_logical->io_phys_children++;

	return (zio);
	}

	zio_t *
	zio_vdev_delegated_io(vdev_t vd, uint64_t offset, abd_t data, uint64_t size,
	zio_type_t type, zio_priority_t priority, enum zio_flag flags,
	zio_done_func_t done, void private)
	{
	zio_t *zio;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
	data, size, size, done, private, type, priority,
	flags \| ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_DELEGATED,
	vd, offset, NULL,
	ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE);

	return (zio);
	}

	void
	zio_flush(zio_t zio, vdev_t vd)
	{
	zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE,
	NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY));
	}

	void
	zio_shrink(zio_t *zio, uint64_t size)
	{
	ASSERT3P(zio->io_executor, ==, NULL);
	ASSERT3U(zio->io_orig_size, ==, zio->io_size);
	ASSERT3U(size, <=, zio->io_size);

	/*
	* We don't shrink for raidz because of problems with the
	* reconstruction when reading back less than the block size.
	* Note, BP_IS_RAIDZ() assumes no compression.
	*/
	ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
	if (!BP_IS_RAIDZ(zio->io_bp)) {
	/* we are not doing a raw write */
	ASSERT3U(zio->io_size, ==, zio->io_lsize);
	zio->io_orig_size = zio->io_size = zio->io_lsize = size;
	}
	}

	/*
	* ==========================================================================
	* Prepare to read and write logical blocks
	* ==========================================================================
	*/

	static zio_t *
	zio_read_bp_init(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	uint64_t psize =
	BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp);

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
	zio->io_child_type == ZIO_CHILD_LOGICAL &&
	!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
	zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
	psize, psize, zio_decompress);
	}

	if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) \|\|
	BP_HAS_INDIRECT_MAC_CKSUM(bp)) &&
	zio->io_child_type == ZIO_CHILD_LOGICAL) {
	zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
	psize, psize, zio_decrypt);
	}

	if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) {
	int psize = BPE_GET_PSIZE(bp);
	void *data = abd_borrow_buf(zio->io_abd, psize);

	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	decode_embedded_bp_compressed(bp, data);
	abd_return_buf_copy(zio->io_abd, data, psize);
	} else {
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);
	}

	if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0)
	zio->io_flags \|= ZIO_FLAG_DONT_CACHE;

	if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP)
	zio->io_flags \|= ZIO_FLAG_DONT_CACHE;

	if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL)
	zio->io_pipeline = ZIO_DDT_READ_PIPELINE;

	return (zio);
	}

	static zio_t *
	zio_write_bp_init(zio_t *zio)
	{
	if (!IO_IS_ALLOCATING(zio))
	return (zio);

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);

	if (zio->io_bp_override) {
	blkptr_t *bp = zio->io_bp;
	zio_prop_t *zp = &zio->io_prop;

	ASSERT(bp->blk_birth != zio->io_txg);
	ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0);

	bp = zio->io_bp_override;
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (BP_IS_EMBEDDED(bp))
	return (zio);

	/*
	* If we've been overridden and nopwrite is set then
	* set the flag accordingly to indicate that a nopwrite
	* has already occurred.
	*/
	if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) {
	ASSERT(!zp->zp_dedup);
	ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum);
	zio->io_flags \|= ZIO_FLAG_NOPWRITE;
	return (zio);
	}

	ASSERT(!zp->zp_nopwrite);

	if (BP_IS_HOLE(bp) \|\| !zp->zp_dedup)
	return (zio);

	ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP) \|\| zp->zp_dedup_verify);

	if (BP_GET_CHECKSUM(bp) == zp->zp_checksum &&
	!zp->zp_encrypt) {
	BP_SET_DEDUP(bp, 1);
	zio->io_pipeline \|= ZIO_STAGE_DDT_WRITE;
	return (zio);
	}

	/*
	* We were unable to handle this as an override bp, treat
	* it as a regular write I/O.
	*/
	zio->io_bp_override = NULL;
	*bp = zio->io_bp_orig;
	zio->io_pipeline = zio->io_orig_pipeline;
	}

	return (zio);
	}

	static zio_t *
	zio_write_compress(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	zio_prop_t *zp = &zio->io_prop;
	enum zio_compress compress = zp->zp_compress;
	blkptr_t *bp = zio->io_bp;
	uint64_t lsize = zio->io_lsize;
	uint64_t psize = zio->io_size;
	- int pass = 1;
	+ uint32_t pass = 1;

	/*
	* If our children haven't all reached the ready stage,
	* wait for them and then repeat this pipeline stage.
	*/
	if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT \|
	ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) {
	return (NULL);
	}

	if (!IO_IS_ALLOCATING(zio))
	return (zio);

	if (zio->io_children_ready != NULL) {
	/*
	* Now that all our children are ready, run the callback
	* associated with this zio in case it wants to modify the
	* data to be written.
	*/
	ASSERT3U(zp->zp_level, >, 0);
	zio->io_children_ready(zio);
	}

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) {
	/*
	* We're rewriting an existing block, which means we're
	* working on behalf of spa_sync(). For spa_sync() to
	* converge, it must eventually be the case that we don't
	* have to allocate new blocks. But compression changes
	* the blocksize, which forces a reallocate, and makes
	* convergence take longer. Therefore, after the first
	* few passes, stop compressing to ensure convergence.
	*/
	pass = spa_sync_pass(spa);

	ASSERT(zio->io_txg == spa_syncing_txg(spa));
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!BP_GET_DEDUP(bp));

	if (pass >= zfs_sync_pass_dont_compress)
	compress = ZIO_COMPRESS_OFF;

	/* Make sure someone doesn't change their mind on overwrites */
	ASSERT(BP_IS_EMBEDDED(bp) \|\| MIN(zp->zp_copies + BP_IS_GANG(bp),
	spa_max_replication(spa)) == BP_GET_NDVAS(bp));
	}

	/* If it's a compressed write that is not raw, compress the buffer. */
	if (compress != ZIO_COMPRESS_OFF &&
	!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
	void *cbuf = zio_buf_alloc(lsize);
	psize = zio_compress_data(compress, zio->io_abd, cbuf, lsize,
	zp->zp_complevel);
	if (psize == 0 \|\| psize >= lsize) {
	compress = ZIO_COMPRESS_OFF;
	zio_buf_free(cbuf, lsize);
	} else if (!zp->zp_dedup && !zp->zp_encrypt &&
	psize <= BPE_PAYLOAD_SIZE &&
	zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) &&
	spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) {
	encode_embedded_bp_compressed(bp,
	cbuf, compress, lsize, psize);
	BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA);
	BP_SET_TYPE(bp, zio->io_prop.zp_type);
	BP_SET_LEVEL(bp, zio->io_prop.zp_level);
	zio_buf_free(cbuf, lsize);
	bp->blk_birth = zio->io_txg;
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	ASSERT(spa_feature_is_active(spa,
	SPA_FEATURE_EMBEDDED_DATA));
	return (zio);
	} else {
	/*
	* Round compressed size up to the minimum allocation
	* size of the smallest-ashift device, and zero the
	* tail. This ensures that the compressed size of the
	* BP (and thus compressratio property) are correct,
	* in that we charge for the padding used to fill out
	* the last sector.
	*/
	ASSERT3U(spa->spa_min_alloc, >=, SPA_MINBLOCKSHIFT);
	size_t rounded = (size_t)roundup(psize,
	spa->spa_min_alloc);
	if (rounded >= lsize) {
	compress = ZIO_COMPRESS_OFF;
	zio_buf_free(cbuf, lsize);
	psize = lsize;
	} else {
	abd_t *cdata = abd_get_from_buf(cbuf, lsize);
	abd_take_ownership_of_buf(cdata, B_TRUE);
	abd_zero_off(cdata, psize, rounded - psize);
	psize = rounded;
	zio_push_transform(zio, cdata,
	psize, lsize, NULL);
	}
	}

	/*
	* We were unable to handle this as an override bp, treat
	* it as a regular write I/O.
	*/
	zio->io_bp_override = NULL;
	*bp = zio->io_bp_orig;
	zio->io_pipeline = zio->io_orig_pipeline;

	} else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 &&
	zp->zp_type == DMU_OT_DNODE) {
	/*
	* The DMU actually relies on the zio layer's compression
	* to free metadnode blocks that have had all contained
	* dnodes freed. As a result, even when doing a raw
	* receive, we must check whether the block can be compressed
	* to a hole.
	*/
	psize = zio_compress_data(ZIO_COMPRESS_EMPTY,
	zio->io_abd, NULL, lsize, zp->zp_complevel);
	if (psize == 0 \|\| psize >= lsize)
	compress = ZIO_COMPRESS_OFF;
	} else if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS &&
	!(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) {
	/*
	* If we are raw receiving an encrypted dataset we should not
	* take this codepath because it will change the on-disk block
	* and decryption will fail.
	*/
	size_t rounded = MIN((size_t)roundup(psize,
	spa->spa_min_alloc), lsize);

	if (rounded != psize) {
	abd_t *cdata = abd_alloc_linear(rounded, B_TRUE);
	abd_zero_off(cdata, psize, rounded - psize);
	abd_copy_off(cdata, zio->io_abd, 0, 0, psize);
	psize = rounded;
	zio_push_transform(zio, cdata,
	psize, rounded, NULL);
	}
	} else {
	ASSERT3U(psize, !=, 0);
	}

	/*
	* The final pass of spa_sync() must be all rewrites, but the first
	* few passes offer a trade-off: allocating blocks defers convergence,
	* but newly allocated blocks are sequential, so they can be written
	* to disk faster. Therefore, we allow the first few passes of
	* spa_sync() to allocate new blocks, but force rewrites after that.
	* There should only be a handful of blocks after pass 1 in any case.
	*/
	if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg &&
	BP_GET_PSIZE(bp) == psize &&
	pass >= zfs_sync_pass_rewrite) {
	VERIFY3U(psize, !=, 0);
	enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;

	zio->io_pipeline = ZIO_REWRITE_PIPELINE \| gang_stages;
	zio->io_flags \|= ZIO_FLAG_IO_REWRITE;
	} else {
	BP_ZERO(bp);
	zio->io_pipeline = ZIO_WRITE_PIPELINE;
	}

	if (psize == 0) {
	if (zio->io_bp_orig.blk_birth != 0 &&
	spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
	BP_SET_LSIZE(bp, lsize);
	BP_SET_TYPE(bp, zp->zp_type);
	BP_SET_LEVEL(bp, zp->zp_level);
	BP_SET_BIRTH(bp, zio->io_txg, 0);
	}
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	} else {
	ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER);
	BP_SET_LSIZE(bp, lsize);
	BP_SET_TYPE(bp, zp->zp_type);
	BP_SET_LEVEL(bp, zp->zp_level);
	BP_SET_PSIZE(bp, psize);
	BP_SET_COMPRESS(bp, compress);
	BP_SET_CHECKSUM(bp, zp->zp_checksum);
	BP_SET_DEDUP(bp, zp->zp_dedup);
	BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
	if (zp->zp_dedup) {
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(!zp->zp_encrypt \|\|
	DMU_OT_IS_ENCRYPTED(zp->zp_type));
	zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE;
	}
	if (zp->zp_nopwrite) {
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	zio->io_pipeline \|= ZIO_STAGE_NOP_WRITE;
	}
	}
	return (zio);
	}

	static zio_t *
	zio_free_bp_init(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio->io_child_type == ZIO_CHILD_LOGICAL) {
	if (BP_GET_DEDUP(bp))
	zio->io_pipeline = ZIO_DDT_FREE_PIPELINE;
	}

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	return (zio);
	}

	/*
	* ==========================================================================
	* Execute the I/O pipeline
	* ==========================================================================
	*/

	static void
	zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline)
	{
	spa_t *spa = zio->io_spa;
	zio_type_t t = zio->io_type;
	int flags = (cutinline ? TQ_FRONT : 0);

	/*
	* If we're a config writer or a probe, the normal issue and
	* interrupt threads may all be blocked waiting for the config lock.
	* In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL.
	*/
	if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER \| ZIO_FLAG_PROBE))
	t = ZIO_TYPE_NULL;

	/*
	* A similar issue exists for the L2ARC write thread until L2ARC 2.0.
	*/
	if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux)
	t = ZIO_TYPE_NULL;

	/*
	* If this is a high priority I/O, then use the high priority taskq if
	* available.
	*/
	if ((zio->io_priority == ZIO_PRIORITY_NOW \|\|
	zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) &&
	spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
	q++;

	ASSERT3U(q, <, ZIO_TASKQ_TYPES);

	/*
	* NB: We are assuming that the zio can only be dispatched
	* to a single taskq at a time. It would be a grievous error
	* to dispatch the zio to another taskq at the same time.
	*/
	ASSERT(taskq_empty_ent(&zio->io_tqent));
	spa_taskq_dispatch_ent(spa, t, q, zio_execute, zio, flags,
	&zio->io_tqent);
	}

	static boolean_t
	zio_taskq_member(zio_t *zio, zio_taskq_type_t q)
	{
	spa_t *spa = zio->io_spa;

	taskq_t *tq = taskq_of_curthread();

	for (zio_type_t t = 0; t < ZIO_TYPES; t++) {
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	uint_t i;
	for (i = 0; i < tqs->stqs_count; i++) {
	if (tqs->stqs_taskq[i] == tq)
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}

	static zio_t *
	zio_issue_async(zio_t *zio)
	{
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);

	return (NULL);
	}

	void
	zio_interrupt(void *zio)
	{
	zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE);
	}

	void
	zio_delay_interrupt(zio_t *zio)
	{
	/*
	* The timeout_generic() function isn't defined in userspace, so
	* rather than trying to implement the function, the zio delay
	* functionality has been disabled for userspace builds.
	*/

	#ifdef _KERNEL
	/*
	* If io_target_timestamp is zero, then no delay has been registered
	* for this IO, thus jump to the end of this function and "skip" the
	* delay; issuing it directly to the zio layer.
	*/
	if (zio->io_target_timestamp != 0) {
	hrtime_t now = gethrtime();

	if (now >= zio->io_target_timestamp) {
	/*
	* This IO has already taken longer than the target
	* delay to complete, so we don't want to delay it
	* any longer; we "miss" the delay and issue it
	* directly to the zio layer. This is likely due to
	* the target latency being set to a value less than
	* the underlying hardware can satisfy (e.g. delay
	* set to 1ms, but the disks take 10ms to complete an
	* IO request).
	*/

	DTRACE_PROBE2(zio__delay__miss, zio_t *, zio,
	hrtime_t, now);

	zio_interrupt(zio);
	} else {
	taskqid_t tid;
	hrtime_t diff = zio->io_target_timestamp - now;
	clock_t expire_at_tick = ddi_get_lbolt() +
	NSEC_TO_TICK(diff);

	DTRACE_PROBE3(zio__delay__hit, zio_t *, zio,
	hrtime_t, now, hrtime_t, diff);

	if (NSEC_TO_TICK(diff) == 0) {
	/* Our delay is less than a jiffy - just spin */
	zfs_sleep_until(zio->io_target_timestamp);
	zio_interrupt(zio);
	} else {
	/*
	* Use taskq_dispatch_delay() in the place of
	* OpenZFS's timeout_generic().
	*/
	tid = taskq_dispatch_delay(system_taskq,
	zio_interrupt, zio, TQ_NOSLEEP,
	expire_at_tick);
	if (tid == TASKQID_INVALID) {
	/*
	* Couldn't allocate a task. Just
	* finish the zio without a delay.
	*/
	zio_interrupt(zio);
	}
	}
	}
	return;
	}
	#endif
	DTRACE_PROBE1(zio__delay__skip, zio_t *, zio);
	zio_interrupt(zio);
	}

	static void
	zio_deadman_impl(zio_t *pio, int ziodepth)
	{
	zio_t cio, cio_next;
	zio_link_t *zl = NULL;
	vdev_t *vd = pio->io_vd;

	if (zio_deadman_log_all \|\| (vd != NULL && vd->vdev_ops->vdev_op_leaf)) {
	vdev_queue_t *vq = vd ? &vd->vdev_queue : NULL;
	zbookmark_phys_t *zb = &pio->io_bookmark;
	uint64_t delta = gethrtime() - pio->io_timestamp;
	uint64_t failmode = spa_get_deadman_failmode(pio->io_spa);

	zfs_dbgmsg("slow zio[%d]: zio=%px timestamp=%llu "
	"delta=%llu queued=%llu io=%llu "
	"path=%s "
	"last=%llu type=%d "
	"priority=%d flags=0x%x stage=0x%x "
	"pipeline=0x%x pipeline-trace=0x%x "
	"objset=%llu object=%llu "
	"level=%llu blkid=%llu "
	"offset=%llu size=%llu "
	"error=%d",
	ziodepth, pio, pio->io_timestamp,
	(u_longlong_t)delta, pio->io_delta, pio->io_delay,
	vd ? vd->vdev_path : "NULL",
	vq ? vq->vq_io_complete_ts : 0, pio->io_type,
	pio->io_priority, pio->io_flags, pio->io_stage,
	pio->io_pipeline, pio->io_pipeline_trace,
	(u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object,
	(u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid,
	(u_longlong_t)pio->io_offset, (u_longlong_t)pio->io_size,
	pio->io_error);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN,
	pio->io_spa, vd, zb, pio, 0);

	if (failmode == ZIO_FAILURE_MODE_CONTINUE &&
	taskq_empty_ent(&pio->io_tqent)) {
	zio_interrupt(pio);
	}
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	zio_deadman_impl(cio, ziodepth + 1);
	}
	mutex_exit(&pio->io_lock);
	}

	/*
	* Log the critical information describing this zio and all of its children
	* using the zfs_dbgmsg() interface then post deadman event for the ZED.
	*/
	void
	zio_deadman(zio_t pio, const char tag)
	{
	spa_t *spa = pio->io_spa;
	char *name = spa_name(spa);

	if (!zfs_deadman_enabled \|\| spa_suspended(spa))
	return;

	zio_deadman_impl(pio, 0);

	switch (spa_get_deadman_failmode(spa)) {
	case ZIO_FAILURE_MODE_WAIT:
	zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name);
	break;

	case ZIO_FAILURE_MODE_CONTINUE:
	zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name);
	break;

	case ZIO_FAILURE_MODE_PANIC:
	fm_panic("%s determined I/O to pool '%s' is hung.", tag, name);
	break;
	}
	}

	/*
	* Execute the I/O pipeline until one of the following occurs:
	* (1) the I/O completes; (2) the pipeline stalls waiting for
	* dependent child I/Os; (3) the I/O issues, so we're waiting
	* for an I/O completion interrupt; (4) the I/O is delegated by
	* vdev-level caching or aggregation; (5) the I/O is deferred
	* due to vdev-level queueing; (6) the I/O is handed off to
	* another thread. In all cases, the pipeline stops whenever
	* there's no CPU work; it never burns a thread in cv_wait_io().
	*
	* There's no locking on io_stage because there's no legitimate way
	* for multiple threads to be attempting to process the same I/O.
	*/
	static zio_pipe_stage_t *zio_pipeline[];

	/*
	* zio_execute() is a wrapper around the static function
	* __zio_execute() so that we can force __zio_execute() to be
	* inlined. This reduces stack overhead which is important
	* because __zio_execute() is called recursively in several zio
	* code paths. zio_execute() itself cannot be inlined because
	* it is externally visible.
	*/
	void
	zio_execute(void *zio)
	{
	fstrans_cookie_t cookie;

	cookie = spl_fstrans_mark();
	__zio_execute(zio);
	spl_fstrans_unmark(cookie);
	}

	/*
	* Used to determine if in the current context the stack is sized large
	* enough to allow zio_execute() to be called recursively. A minimum
	* stack size of 16K is required to avoid needing to re-dispatch the zio.
	*/
	static boolean_t
	zio_execute_stack_check(zio_t *zio)
	{
	#if !defined(HAVE_LARGE_STACKS)
	dsl_pool_t *dp = spa_get_dsl(zio->io_spa);

	/* Executing in txg_sync_thread() context. */
	if (dp && curthread == dp->dp_tx.tx_sync_thread)
	return (B_TRUE);

	/* Pool initialization outside of zio_taskq context. */
	if (dp && spa_is_initializing(dp->dp_spa) &&
	!zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
	!zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))
	return (B_TRUE);
	#else
	(void) zio;
	#endif /* HAVE_LARGE_STACKS */

	return (B_FALSE);
	}

	__attribute__((always_inline))
	static inline void
	__zio_execute(zio_t *zio)
	{
	ASSERT3U(zio->io_queued_timestamp, >, 0);

	while (zio->io_stage < ZIO_STAGE_DONE) {
	enum zio_stage pipeline = zio->io_pipeline;
	enum zio_stage stage = zio->io_stage;

	zio->io_executor = curthread;

	ASSERT(!MUTEX_HELD(&zio->io_lock));
	ASSERT(ISP2(stage));
	ASSERT(zio->io_stall == NULL);

	do {
	stage <<= 1;
	} while ((stage & pipeline) == 0);

	ASSERT(stage <= ZIO_STAGE_DONE);

	/*
	* If we are in interrupt context and this pipeline stage
	* will grab a config lock that is held across I/O,
	* or may wait for an I/O that needs an interrupt thread
	* to complete, issue async to avoid deadlock.
	*
	* For VDEV_IO_START, we cut in line so that the io will
	* be sent to disk promptly.
	*/
	if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
	zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
	boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
	zio_requeue_io_start_cut_in_line : B_FALSE;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
	return;
	}

	/*
	* If the current context doesn't have large enough stacks
	* the zio must be issued asynchronously to prevent overflow.
	*/
	if (zio_execute_stack_check(zio)) {
	boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
	zio_requeue_io_start_cut_in_line : B_FALSE;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
	return;
	}

	zio->io_stage = stage;
	zio->io_pipeline_trace \|= zio->io_stage;

	/*
	* The zio pipeline stage returns the next zio to execute
	* (typically the same as this one), or NULL if we should
	* stop.
	*/
	zio = zio_pipeline[highbit64(stage) - 1](zio);

	if (zio == NULL)
	return;
	}
	}


	/*
	* ==========================================================================
	* Initiate I/O, either sync or async
	* ==========================================================================
	*/
	int
	zio_wait(zio_t *zio)
	{
	/*
	* Some routines, like zio_free_sync(), may return a NULL zio
	* to avoid the performance overhead of creating and then destroying
	* an unneeded zio. For the callers' simplicity, we accept a NULL
	* zio and ignore it.
	*/
	if (zio == NULL)
	return (0);

	long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms);
	int error;

	ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN);
	ASSERT3P(zio->io_executor, ==, NULL);

	zio->io_waiter = curthread;
	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();

	__zio_execute(zio);

	mutex_enter(&zio->io_lock);
	while (zio->io_executor != NULL) {
	error = cv_timedwait_io(&zio->io_cv, &zio->io_lock,
	ddi_get_lbolt() + timeout);

	if (zfs_deadman_enabled && error == -1 &&
	gethrtime() - zio->io_queued_timestamp >
	spa_deadman_ziotime(zio->io_spa)) {
	mutex_exit(&zio->io_lock);
	timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms);
	zio_deadman(zio, FTAG);
	mutex_enter(&zio->io_lock);
	}
	}
	mutex_exit(&zio->io_lock);

	error = zio->io_error;
	zio_destroy(zio);

	return (error);
	}

	void
	zio_nowait(zio_t *zio)
	{
	/*
	* See comment in zio_wait().
	*/
	if (zio == NULL)
	return;

	ASSERT3P(zio->io_executor, ==, NULL);

	if (zio->io_child_type == ZIO_CHILD_LOGICAL &&
	zio_unique_parent(zio) == NULL) {
	zio_t *pio;

	/*
	* This is a logical async I/O with no parent to wait for it.
	* We add it to the spa_async_root_zio "Godfather" I/O which
	* will ensure they complete prior to unloading the pool.
	*/
	spa_t *spa = zio->io_spa;
	pio = spa->spa_async_zio_root[CPU_SEQID_UNSTABLE];

	zio_add_child(pio, zio);
	}

	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();
	__zio_execute(zio);
	}

	/*
	* ==========================================================================
	* Reexecute, cancel, or suspend/resume failed I/O
	* ==========================================================================
	*/

	static void
	zio_reexecute(void *arg)
	{
	zio_t *pio = arg;
	zio_t cio, cio_next;

	ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN);
	ASSERT(pio->io_gang_leader == NULL);
	ASSERT(pio->io_gang_tree == NULL);

	pio->io_flags = pio->io_orig_flags;
	pio->io_stage = pio->io_orig_stage;
	pio->io_pipeline = pio->io_orig_pipeline;
	pio->io_reexecute = 0;
	pio->io_flags \|= ZIO_FLAG_REEXECUTED;
	pio->io_pipeline_trace = 0;
	pio->io_error = 0;
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	pio->io_state[w] = 0;
	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
	pio->io_child_error[c] = 0;

	if (IO_IS_ALLOCATING(pio))
	BP_ZERO(pio->io_bp);

	/*
	* As we reexecute pio's children, new children could be created.
	* New children go to the head of pio's io_child_list, however,
	* so we will (correctly) not reexecute them. The key is that
	* the remainder of pio's io_child_list, from 'cio_next' onward,
	* cannot be affected by any side effects of reexecuting 'cio'.
	*/
	zio_link_t *zl = NULL;
	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	pio->io_children[cio->io_child_type][w]++;
	mutex_exit(&pio->io_lock);
	zio_reexecute(cio);
	mutex_enter(&pio->io_lock);
	}
	mutex_exit(&pio->io_lock);

	/*
	* Now that all children have been reexecuted, execute the parent.
	* We don't reexecute "The Godfather" I/O here as it's the
	* responsibility of the caller to wait on it.
	*/
	if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) {
	pio->io_queued_timestamp = gethrtime();
	__zio_execute(pio);
	}
	}

	void
	zio_suspend(spa_t spa, zio_t zio, zio_suspend_reason_t reason)
	{
	if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC)
	fm_panic("Pool '%s' has encountered an uncorrectable I/O "
	"failure and the failure mode property for this pool "
	"is set to panic.", spa_name(spa));

	cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O "
	"failure and has been suspended.\n", spa_name(spa));

	(void) zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL,
	NULL, NULL, 0);

	mutex_enter(&spa->spa_suspend_lock);

	if (spa->spa_suspend_zio_root == NULL)
	spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);

	spa->spa_suspended = reason;

	if (zio != NULL) {
	ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
	ASSERT(zio != spa->spa_suspend_zio_root);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(zio_unique_parent(zio) == NULL);
	ASSERT(zio->io_stage == ZIO_STAGE_DONE);
	zio_add_child(spa->spa_suspend_zio_root, zio);
	}

	mutex_exit(&spa->spa_suspend_lock);
	}

	int
	zio_resume(spa_t *spa)
	{
	zio_t *pio;

	/*
	* Reexecute all previously suspended i/o.
	*/
	mutex_enter(&spa->spa_suspend_lock);
	spa->spa_suspended = ZIO_SUSPEND_NONE;
	cv_broadcast(&spa->spa_suspend_cv);
	pio = spa->spa_suspend_zio_root;
	spa->spa_suspend_zio_root = NULL;
	mutex_exit(&spa->spa_suspend_lock);

	if (pio == NULL)
	return (0);

	zio_reexecute(pio);
	return (zio_wait(pio));
	}

	void
	zio_resume_wait(spa_t *spa)
	{
	mutex_enter(&spa->spa_suspend_lock);
	while (spa_suspended(spa))
	cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock);
	mutex_exit(&spa->spa_suspend_lock);
	}

	/*
	* ==========================================================================
	* Gang blocks.
	*
	* A gang block is a collection of small blocks that looks to the DMU
	* like one large block. When zio_dva_allocate() cannot find a block
	* of the requested size, due to either severe fragmentation or the pool
	* being nearly full, it calls zio_write_gang_block() to construct the
	* block from smaller fragments.
	*
	* A gang block consists of a gang header (zio_gbh_phys_t) and up to
	* three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like
	* an indirect block: it's an array of block pointers. It consumes
	* only one sector and hence is allocatable regardless of fragmentation.
	* The gang header's bps point to its gang members, which hold the data.
	*
	* Gang blocks are self-checksumming, using the bp's <vdev, offset, txg>
	* as the verifier to ensure uniqueness of the SHA256 checksum.
	* Critically, the gang block bp's blk_cksum is the checksum of the data,
	* not the gang header. This ensures that data block signatures (needed for
	* deduplication) are independent of how the block is physically stored.
	*
	* Gang blocks can be nested: a gang member may itself be a gang block.
	* Thus every gang block is a tree in which root and all interior nodes are
	* gang headers, and the leaves are normal blocks that contain user data.
	* The root of the gang tree is called the gang leader.
	*
	* To perform any operation (read, rewrite, free, claim) on a gang block,
	* zio_gang_assemble() first assembles the gang tree (minus data leaves)
	* in the io_gang_tree field of the original logical i/o by recursively
	* reading the gang leader and all gang headers below it. This yields
	* an in-core tree containing the contents of every gang header and the
	* bps for every constituent of the gang block.
	*
	* With the gang tree now assembled, zio_gang_issue() just walks the gang tree
	* and invokes a callback on each bp. To free a gang block, zio_gang_issue()
	* calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp.
	* zio_claim_gang() provides a similarly trivial wrapper for zio_claim().
	* zio_read_gang() is a wrapper around zio_read() that omits reading gang
	* headers, since we already have those in io_gang_tree. zio_rewrite_gang()
	* performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite()
	* of the gang header plus zio_checksum_compute() of the data to update the
	* gang header's blk_cksum as described above.
	*
	* The two-phase assemble/issue model solves the problem of partial failure --
	* what if you'd freed part of a gang block but then couldn't read the
	* gang header for another part? Assembling the entire gang tree first
	* ensures that all the necessary gang header I/O has succeeded before
	* starting the actual work of free, claim, or write. Once the gang tree
	* is assembled, free and claim are in-memory operations that cannot fail.
	*
	* In the event that a gang write fails, zio_dva_unallocate() walks the
	* gang tree to immediately free (i.e. insert back into the space map)
	* everything we've allocated. This ensures that we don't get ENOSPC
	* errors during repeated suspend/resume cycles due to a flaky device.
	*
	* Gang rewrites only happen during sync-to-convergence. If we can't assemble
	* the gang tree, we won't modify the block, so we can safely defer the free
	* (knowing that the block is still intact). If we can assemble the gang
	* tree, then even if some of the rewrites fail, zio_dva_unallocate() will free
	* each constituent bp and we can allocate a new block on the next sync pass.
	*
	* In all cases, the gang tree allows complete recovery from partial failure.
	* ==========================================================================
	*/

	static void
	zio_gang_issue_func_done(zio_t *zio)
	{
	abd_free(zio->io_abd);
	}

	static zio_t *
	zio_read_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	if (gn != NULL)
	return (pio);

	return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset),
	BP_GET_PSIZE(bp), zio_gang_issue_func_done,
	NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
	&pio->io_bookmark));
	}

	static zio_t *
	zio_rewrite_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	zio_t *zio;

	if (gn != NULL) {
	abd_t *gbh_abd =
	abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);
	zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
	gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL,
	pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
	&pio->io_bookmark);
	/*
	* As we rewrite each gang header, the pipeline will compute
	* a new gang block header checksum for it; but no one will
	* compute a new data checksum, so we do that here. The one
	* exception is the gang leader: the pipeline already computed
	* its data checksum because that stage precedes gang assembly.
	* (Presently, nothing actually uses interior data checksums;
	* this is just good hygiene.)
	*/
	if (gn != pio->io_gang_leader->io_gang_tree) {
	abd_t *buf = abd_get_offset(data, offset);

	zio_checksum_compute(zio, BP_GET_CHECKSUM(bp),
	buf, BP_GET_PSIZE(bp));

	abd_free(buf);
	}
	/*
	* If we are here to damage data for testing purposes,
	* leave the GBH alone so that we can detect the damage.
	*/
	if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE)
	zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	} else {
	zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
	abd_get_offset(data, offset), BP_GET_PSIZE(bp),
	zio_gang_issue_func_done, NULL, pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
	}

	return (zio);
	}

	static zio_t *
	zio_free_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	(void) gn, (void) data, (void) offset;

	zio_t *zio = zio_free_sync(pio, pio->io_spa, pio->io_txg, bp,
	ZIO_GANG_CHILD_FLAGS(pio));
	if (zio == NULL) {
	zio = zio_null(pio, pio->io_spa,
	NULL, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio));
	}
	return (zio);
	}

	static zio_t *
	zio_claim_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	(void) gn, (void) data, (void) offset;
	return (zio_claim(pio, pio->io_spa, pio->io_txg, bp,
	NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
	}

	static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = {
	NULL,
	zio_read_gang,
	zio_rewrite_gang,
	zio_free_gang,
	zio_claim_gang,
	NULL
	};

	static void zio_gang_tree_assemble_done(zio_t *zio);

	static zio_gang_node_t *
	zio_gang_node_alloc(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t *gn;

	ASSERT(*gnpp == NULL);

	gn = kmem_zalloc(sizeof (*gn), KM_SLEEP);
	gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE);
	*gnpp = gn;

	return (gn);
	}

	static void
	zio_gang_node_free(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t gn = gnpp;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
	ASSERT(gn->gn_child[g] == NULL);

	zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE);
	kmem_free(gn, sizeof (*gn));
	*gnpp = NULL;
	}

	static void
	zio_gang_tree_free(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t gn = gnpp;

	if (gn == NULL)
	return;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
	zio_gang_tree_free(&gn->gn_child[g]);

	zio_gang_node_free(gnpp);
	}

	static void
	zio_gang_tree_assemble(zio_t gio, blkptr_t bp, zio_gang_node_t **gnpp)
	{
	zio_gang_node_t *gn = zio_gang_node_alloc(gnpp);
	abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);

	ASSERT(gio->io_gang_leader == gio);
	ASSERT(BP_IS_GANG(bp));

	zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	zio_gang_tree_assemble_done, gn, gio->io_priority,
	ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark));
	}

	static void
	zio_gang_tree_assemble_done(zio_t *zio)
	{
	zio_t *gio = zio->io_gang_leader;
	zio_gang_node_t *gn = zio->io_private;
	blkptr_t *bp = zio->io_bp;

	ASSERT(gio == zio_unique_parent(zio));
	ASSERT(zio->io_child_count == 0);

	if (zio->io_error)
	return;

	/* this ABD was created from a linear buf in zio_gang_tree_assemble */
	if (BP_SHOULD_BYTESWAP(bp))
	byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size);

	ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh);
	ASSERT(zio->io_size == SPA_GANGBLOCKSIZE);
	ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

	abd_free(zio->io_abd);

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
	if (!BP_IS_GANG(gbp))
	continue;
	zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]);
	}
	}

	static void
	zio_gang_tree_issue(zio_t pio, zio_gang_node_t gn, blkptr_t bp, abd_t data,
	uint64_t offset)
	{
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;

	ASSERT(BP_IS_GANG(bp) == !!gn);
	ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp));
	ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) \|\| gn == gio->io_gang_tree);

	/*
	* If you're a gang header, your data is in gn->gn_gbh.
	* If you're a gang member, your data is in 'data' and gn == NULL.
	*/
	zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset);

	if (gn != NULL) {
	ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
	if (BP_IS_HOLE(gbp))
	continue;
	zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data,
	offset);
	offset += BP_GET_PSIZE(gbp);
	}
	}

	if (gn == gio->io_gang_tree)
	ASSERT3U(gio->io_size, ==, offset);

	if (zio != pio)
	zio_nowait(zio);
	}

	static zio_t *
	zio_gang_assemble(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	zio->io_gang_leader = zio;

	zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree);

	return (zio);
	}

	static zio_t *
	zio_gang_issue(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	if (zio->io_child_error[ZIO_CHILD_GANG] == 0)
	zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd,
	0);
	else
	zio_gang_tree_free(&zio->io_gang_tree);

	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	return (zio);
	}

	static void
	zio_write_gang_member_ready(zio_t *zio)
	{
	zio_t *pio = zio_unique_parent(zio);
	dva_t *cdva = zio->io_bp->blk_dva;
	dva_t *pdva = pio->io_bp->blk_dva;
	uint64_t asize;
	zio_t *gio __maybe_unused = zio->io_gang_leader;

	if (BP_IS_HOLE(zio->io_bp))
	return;

	ASSERT(BP_IS_HOLE(&zio->io_bp_orig));

	ASSERT(zio->io_child_type == ZIO_CHILD_GANG);
	ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies);
	ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp));
	ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp));
	ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp));

	mutex_enter(&pio->io_lock);
	for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) {
	ASSERT(DVA_GET_GANG(&pdva[d]));
	asize = DVA_GET_ASIZE(&pdva[d]);
	asize += DVA_GET_ASIZE(&cdva[d]);
	DVA_SET_ASIZE(&pdva[d], asize);
	}
	mutex_exit(&pio->io_lock);
	}

	static void
	zio_write_gang_done(zio_t *zio)
	{
	/*
	* The io_abd field will be NULL for a zio with no data. The io_flags
	* will initially have the ZIO_FLAG_NODATA bit flag set, but we can't
	* check for it here as it is cleared in zio_ready.
	*/
	if (zio->io_abd != NULL)
	abd_free(zio->io_abd);
	}

	static zio_t *
	zio_write_gang_block(zio_t pio, metaslab_class_t mc)
	{
	spa_t *spa = pio->io_spa;
	blkptr_t *bp = pio->io_bp;
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;
	zio_gang_node_t gn, *gnpp;
	zio_gbh_phys_t *gbh;
	abd_t *gbh_abd;
	uint64_t txg = pio->io_txg;
	uint64_t resid = pio->io_size;
	uint64_t lsize;
	int copies = gio->io_prop.zp_copies;
	int gbh_copies;
	zio_prop_t zp;
	int error;
	boolean_t has_data = !(pio->io_flags & ZIO_FLAG_NODATA);

	/*
	* encrypted blocks need DVA[2] free so encrypted gang headers can't
	* have a third copy.
	*/
	gbh_copies = MIN(copies + 1, spa_max_replication(spa));
	if (gio->io_prop.zp_encrypt && gbh_copies >= SPA_DVAS_PER_BP)
	gbh_copies = SPA_DVAS_PER_BP - 1;

	int flags = METASLAB_HINTBP_FAVOR \| METASLAB_GANG_HEADER;
	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	flags \|= METASLAB_ASYNC_ALLOC;
	VERIFY(zfs_refcount_held(&mc->mc_allocator[pio->io_allocator].
	mca_alloc_slots, pio));

	/*
	* The logical zio has already placed a reservation for
	* 'copies' allocation slots but gang blocks may require
	* additional copies. These additional copies
	* (i.e. gbh_copies - copies) are guaranteed to succeed
	* since metaslab_class_throttle_reserve() always allows
	* additional reservations for gang blocks.
	*/
	VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies,
	pio->io_allocator, pio, flags));
	}

	error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE,
	bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags,
	&pio->io_alloc_list, pio, pio->io_allocator);
	if (error) {
	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	/*
	* If we failed to allocate the gang block header then
	* we remove any additional allocation reservations that
	* we placed here. The original reservation will
	* be removed when the logical I/O goes to the ready
	* stage.
	*/
	metaslab_class_throttle_unreserve(mc,
	gbh_copies - copies, pio->io_allocator, pio);
	}

	pio->io_error = error;
	return (pio);
	}

	if (pio == gio) {
	gnpp = &gio->io_gang_tree;
	} else {
	gnpp = pio->io_private;
	ASSERT(pio->io_ready == zio_write_gang_member_ready);
	}

	gn = zio_gang_node_alloc(gnpp);
	gbh = gn->gn_gbh;
	memset(gbh, 0, SPA_GANGBLOCKSIZE);
	gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE);

	/*
	* Create the gang header.
	*/
	zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	zio_write_gang_done, NULL, pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

	/*
	* Create and nowait the gang children.
	*/
	for (int g = 0; resid != 0; resid -= lsize, g++) {
	lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g),
	SPA_MINBLOCKSIZE);
	ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid);

	zp.zp_checksum = gio->io_prop.zp_checksum;
	zp.zp_compress = ZIO_COMPRESS_OFF;
	zp.zp_complevel = gio->io_prop.zp_complevel;
	zp.zp_type = DMU_OT_NONE;
	zp.zp_level = 0;
	zp.zp_copies = gio->io_prop.zp_copies;
	zp.zp_dedup = B_FALSE;
	zp.zp_dedup_verify = B_FALSE;
	zp.zp_nopwrite = B_FALSE;
	zp.zp_encrypt = gio->io_prop.zp_encrypt;
	zp.zp_byteorder = gio->io_prop.zp_byteorder;
	memset(zp.zp_salt, 0, ZIO_DATA_SALT_LEN);
	memset(zp.zp_iv, 0, ZIO_DATA_IV_LEN);
	memset(zp.zp_mac, 0, ZIO_DATA_MAC_LEN);

	zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
	has_data ? abd_get_offset(pio->io_abd, pio->io_size -
	resid) : NULL, lsize, lsize, &zp,
	zio_write_gang_member_ready, NULL, NULL,
	zio_write_gang_done, &gn->gn_child[g], pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	/*
	* Gang children won't throttle but we should
	* account for their work, so reserve an allocation
	* slot for them here.
	*/
	VERIFY(metaslab_class_throttle_reserve(mc,
	zp.zp_copies, cio->io_allocator, cio, flags));
	}
	zio_nowait(cio);
	}

	/*
	* Set pio's pipeline to just wait for zio to finish.
	*/
	pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	/*
	* We didn't allocate this bp, so make sure it doesn't get unmarked.
	*/
	pio->io_flags &= ~ZIO_FLAG_FASTWRITE;

	zio_nowait(zio);

	return (pio);
	}

	/*
	* The zio_nop_write stage in the pipeline determines if allocating a
	* new bp is necessary. The nopwrite feature can handle writes in
	* either syncing or open context (i.e. zil writes) and as a result is
	* mutually exclusive with dedup.
	*
	* By leveraging a cryptographically secure checksum, such as SHA256, we
	* can compare the checksums of the new data and the old to determine if
	* allocating a new block is required. Note that our requirements for
	* cryptographic strength are fairly weak: there can't be any accidental
	* hash collisions, but we don't need to be secure against intentional
	* (malicious) collisions. To trigger a nopwrite, you have to be able
	* to write the file to begin with, and triggering an incorrect (hash
	* collision) nopwrite is no worse than simply writing to the file.
	* That said, there are no known attacks against the checksum algorithms
	* used for nopwrite, assuming that the salt and the checksums
	* themselves remain secret.
	*/
	static zio_t *
	zio_nop_write(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	zio_prop_t *zp = &zio->io_prop;

	ASSERT(BP_GET_LEVEL(bp) == 0);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(zp->zp_nopwrite);
	ASSERT(!zp->zp_dedup);
	ASSERT(zio->io_bp_override == NULL);
	ASSERT(IO_IS_ALLOCATING(zio));

	/*
	* Check to see if the original bp and the new bp have matching
	* characteristics (i.e. same checksum, compression algorithms, etc).
	* If they don't then just continue with the pipeline which will
	* allocate a new bp.
	*/
	if (BP_IS_HOLE(bp_orig) \|\|
	!(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE) \|\|
	BP_IS_ENCRYPTED(bp) \|\| BP_IS_ENCRYPTED(bp_orig) \|\|
	BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) \|\|
	BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) \|\|
	BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) \|\|
	zp->zp_copies != BP_GET_NDVAS(bp_orig))
	return (zio);

	/*
	* If the checksums match then reset the pipeline so that we
	* avoid allocating a new bp and issuing any I/O.
	*/
	if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) {
	ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE);
	ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig));
	ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig));
	ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF);
	ASSERT(memcmp(&bp->blk_prop, &bp_orig->blk_prop,
	sizeof (uint64_t)) == 0);

	/*
	* If we're overwriting a block that is currently on an
	* indirect vdev, then ignore the nopwrite request and
	* allow a new block to be allocated on a concrete vdev.
	*/
	spa_config_enter(zio->io_spa, SCL_VDEV, FTAG, RW_READER);
	vdev_t *tvd = vdev_lookup_top(zio->io_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]));
	if (tvd->vdev_ops == &vdev_indirect_ops) {
	spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);
	return (zio);
	}
	spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);

	bp = bp_orig;
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	zio->io_flags \|= ZIO_FLAG_NOPWRITE;
	}

	return (zio);
	}

	/*
	* ==========================================================================
	* Dedup
	* ==========================================================================
	*/
	static void
	zio_ddt_child_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp;
	zio_t *pio = zio_unique_parent(zio);

	mutex_enter(&pio->io_lock);
	ddp = ddt_phys_select(dde, bp);
	if (zio->io_error == 0)
	ddt_phys_clear(ddp); /* this ddp doesn't need repair */

	if (zio->io_error == 0 && dde->dde_repair_abd == NULL)
	dde->dde_repair_abd = zio->io_abd;
	else
	abd_free(zio->io_abd);
	mutex_exit(&pio->io_lock);
	}

	static zio_t *
	zio_ddt_read_start(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
	ddt_t *ddt = ddt_select(zio->io_spa, bp);
	ddt_entry_t *dde = ddt_repair_start(ddt, bp);
	ddt_phys_t *ddp = dde->dde_phys;
	ddt_phys_t *ddp_self = ddt_phys_select(dde, bp);
	blkptr_t blk;

	ASSERT(zio->io_vsd == NULL);
	zio->io_vsd = dde;

	if (ddp_self == NULL)
	return (zio);

	for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0 \|\| ddp == ddp_self)
	continue;
	ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp,
	&blk);
	zio_nowait(zio_read(zio, zio->io_spa, &blk,
	abd_alloc_for_io(zio->io_size, B_TRUE),
	zio->io_size, zio_ddt_child_read_done, dde,
	zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) \|
	ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark));
	}
	return (zio);
	}

	zio_nowait(zio_read(zio, zio->io_spa, bp,
	zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority,
	ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark));

	return (zio);
	}

	static zio_t *
	zio_ddt_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
	ddt_t *ddt = ddt_select(zio->io_spa, bp);
	ddt_entry_t *dde = zio->io_vsd;
	if (ddt == NULL) {
	ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE);
	return (zio);
	}
	if (dde == NULL) {
	zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
	return (NULL);
	}
	if (dde->dde_repair_abd != NULL) {
	abd_copy(zio->io_abd, dde->dde_repair_abd,
	zio->io_size);
	zio->io_child_error[ZIO_CHILD_DDT] = 0;
	}
	ddt_repair_done(ddt, dde);
	zio->io_vsd = NULL;
	}

	ASSERT(zio->io_vsd == NULL);

	return (zio);
	}

	static boolean_t
	zio_ddt_collision(zio_t zio, ddt_t ddt, ddt_entry_t *dde)
	{
	spa_t *spa = zio->io_spa;
	boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW);

	ASSERT(!(zio->io_bp_override && do_raw));

	/*
	* Note: we compare the original data, not the transformed data,
	* because when zio->io_bp is an override bp, we will not have
	* pushed the I/O transforms. That's an important optimization
	* because otherwise we'd compress/encrypt all dmu_sync() data twice.
	* However, we should never get a raw, override zio so in these
	* cases we can compare the io_abd directly. This is useful because
	* it allows us to do dedup verification even if we don't have access
	* to the original data (for instance, if the encryption keys aren't
	* loaded).
	*/

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
	zio_t *lio = dde->dde_lead_zio[p];

	if (lio != NULL && do_raw) {
	return (lio->io_size != zio->io_size \|\|
	abd_cmp(zio->io_abd, lio->io_abd) != 0);
	} else if (lio != NULL) {
	return (lio->io_orig_size != zio->io_orig_size \|\|
	abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0);
	}
	}

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
	ddt_phys_t *ddp = &dde->dde_phys[p];

	if (ddp->ddp_phys_birth != 0 && do_raw) {
	blkptr_t blk = *zio->io_bp;
	uint64_t psize;
	abd_t *tmpabd;
	int error;

	ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);
	psize = BP_GET_PSIZE(&blk);

	if (psize != zio->io_size)
	return (B_TRUE);

	ddt_exit(ddt);

	tmpabd = abd_alloc_for_io(psize, B_TRUE);

	error = zio_wait(zio_read(NULL, spa, &blk, tmpabd,
	psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_RAW, &zio->io_bookmark));

	if (error == 0) {
	if (abd_cmp(tmpabd, zio->io_abd) != 0)
	error = SET_ERROR(ENOENT);
	}

	abd_free(tmpabd);
	ddt_enter(ddt);
	return (error != 0);
	} else if (ddp->ddp_phys_birth != 0) {
	arc_buf_t *abuf = NULL;
	arc_flags_t aflags = ARC_FLAG_WAIT;
	blkptr_t blk = *zio->io_bp;
	int error;

	ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);

	if (BP_GET_LSIZE(&blk) != zio->io_orig_size)
	return (B_TRUE);

	ddt_exit(ddt);

	error = arc_read(NULL, spa, &blk,
	arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&aflags, &zio->io_bookmark);

	if (error == 0) {
	if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data,
	zio->io_orig_size) != 0)
	error = SET_ERROR(ENOENT);
	arc_buf_destroy(abuf, &abuf);
	}

	ddt_enter(ddt);
	return (error != 0);
	}
	}

	return (B_FALSE);
	}

	static void
	zio_ddt_child_write_ready(zio_t *zio)
	{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];
	zio_t *pio;

	if (zio->io_error)
	return;

	ddt_enter(ddt);

	ASSERT(dde->dde_lead_zio[p] == zio);

	ddt_phys_fill(ddp, zio->io_bp);

	zio_link_t *zl = NULL;
	while ((pio = zio_walk_parents(zio, &zl)) != NULL)
	ddt_bp_fill(ddp, pio->io_bp, zio->io_txg);

	ddt_exit(ddt);
	}

	static void
	zio_ddt_child_write_done(zio_t *zio)
	{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];

	ddt_enter(ddt);

	ASSERT(ddp->ddp_refcnt == 0);
	ASSERT(dde->dde_lead_zio[p] == zio);
	dde->dde_lead_zio[p] = NULL;

	if (zio->io_error == 0) {
	zio_link_t *zl = NULL;
	while (zio_walk_parents(zio, &zl) != NULL)
	ddt_phys_addref(ddp);
	} else {
	ddt_phys_clear(ddp);
	}

	ddt_exit(ddt);
	}

	static zio_t *
	zio_ddt_write(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t txg = zio->io_txg;
	zio_prop_t *zp = &zio->io_prop;
	int p = zp->zp_copies;
	zio_t *cio = NULL;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum);
	ASSERT(BP_IS_HOLE(bp) \|\| zio->io_bp_override);
	ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW)));

	ddt_enter(ddt);
	dde = ddt_lookup(ddt, bp, B_TRUE);
	ddp = &dde->dde_phys[p];

	if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) {
	/*
	* If we're using a weak checksum, upgrade to a strong checksum
	* and try again. If we're already using a strong checksum,
	* we can't resolve it, so just convert to an ordinary write.
	* (And automatically e-mail a paper to Nature?)
	*/
	if (!(zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP)) {
	zp->zp_checksum = spa_dedup_checksum(spa);
	zio_pop_transforms(zio);
	zio->io_stage = ZIO_STAGE_OPEN;
	BP_ZERO(bp);
	} else {
	zp->zp_dedup = B_FALSE;
	BP_SET_DEDUP(bp, B_FALSE);
	}
	ASSERT(!BP_GET_DEDUP(bp));
	zio->io_pipeline = ZIO_WRITE_PIPELINE;
	ddt_exit(ddt);
	return (zio);
	}

	if (ddp->ddp_phys_birth != 0 \|\| dde->dde_lead_zio[p] != NULL) {
	if (ddp->ddp_phys_birth != 0)
	ddt_bp_fill(ddp, bp, txg);
	if (dde->dde_lead_zio[p] != NULL)
	zio_add_child(zio, dde->dde_lead_zio[p]);
	else
	ddt_phys_addref(ddp);
	} else if (zio->io_bp_override) {
	ASSERT(bp->blk_birth == txg);
	ASSERT(BP_EQUAL(bp, zio->io_bp_override));
	ddt_phys_fill(ddp, bp);
	ddt_phys_addref(ddp);
	} else {
	cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd,
	zio->io_orig_size, zio->io_orig_size, zp,
	zio_ddt_child_write_ready, NULL, NULL,
	zio_ddt_child_write_done, dde, zio->io_priority,
	ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);

	zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL);
	dde->dde_lead_zio[p] = cio;
	}

	ddt_exit(ddt);

	zio_nowait(cio);

	return (zio);
	}

	ddt_entry_t freedde; / for debugging */

	static zio_t *
	zio_ddt_free(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	ddt_enter(ddt);
	freedde = dde = ddt_lookup(ddt, bp, B_TRUE);
	if (dde) {
	ddp = ddt_phys_select(dde, bp);
	if (ddp)
	ddt_phys_decref(ddp);
	}
	ddt_exit(ddt);

	return (zio);
	}

	/*
	* ==========================================================================
	* Allocate and free blocks
	* ==========================================================================
	*/

	static zio_t *
	zio_io_to_allocate(spa_t *spa, int allocator)
	{
	zio_t *zio;

	ASSERT(MUTEX_HELD(&spa->spa_allocs[allocator].spaa_lock));

	zio = avl_first(&spa->spa_allocs[allocator].spaa_tree);
	if (zio == NULL)
	return (NULL);

	ASSERT(IO_IS_ALLOCATING(zio));

	/*
	* Try to place a reservation for this zio. If we're unable to
	* reserve then we throttle.
	*/
	ASSERT3U(zio->io_allocator, ==, allocator);
	if (!metaslab_class_throttle_reserve(zio->io_metaslab_class,
	zio->io_prop.zp_copies, allocator, zio, 0)) {
	return (NULL);
	}

	avl_remove(&spa->spa_allocs[allocator].spaa_tree, zio);
	ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE);

	return (zio);
	}

	static zio_t *
	zio_dva_throttle(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	zio_t *nio;
	metaslab_class_t *mc;

	/* locate an appropriate allocation class */
	mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type,
	zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk);

	if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE \|\|
	!mc->mc_alloc_throttle_enabled \|\|
	zio->io_child_type == ZIO_CHILD_GANG \|\|
	zio->io_flags & ZIO_FLAG_NODATA) {
	return (zio);
	}

	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
	ASSERT3U(zio->io_queued_timestamp, >, 0);
	ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE);

	zbookmark_phys_t *bm = &zio->io_bookmark;
	/*
	* We want to try to use as many allocators as possible to help improve
	* performance, but we also want logically adjacent IOs to be physically
	* adjacent to improve sequential read performance. We chunk each object
	* into 2^20 block regions, and then hash based on the objset, object,
	* level, and region to accomplish both of these goals.
	*/
	int allocator = (uint_t)cityhash4(bm->zb_objset, bm->zb_object,
	bm->zb_level, bm->zb_blkid >> 20) % spa->spa_alloc_count;
	zio->io_allocator = allocator;
	zio->io_metaslab_class = mc;
	mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
	avl_add(&spa->spa_allocs[allocator].spaa_tree, zio);
	nio = zio_io_to_allocate(spa, allocator);
	mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
	return (nio);
	}

	static void
	zio_allocate_dispatch(spa_t *spa, int allocator)
	{
	zio_t *zio;

	mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
	zio = zio_io_to_allocate(spa, allocator);
	mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
	if (zio == NULL)
	return;

	ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE);
	ASSERT0(zio->io_error);
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE);
	}

	static zio_t *
	zio_dva_allocate(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	metaslab_class_t *mc;
	blkptr_t *bp = zio->io_bp;
	int error;
	int flags = 0;

	if (zio->io_gang_leader == NULL) {
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
	zio->io_gang_leader = zio;
	}

	ASSERT(BP_IS_HOLE(bp));
	ASSERT0(BP_GET_NDVAS(bp));
	ASSERT3U(zio->io_prop.zp_copies, >, 0);
	ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
	ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));

	flags \|= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
	if (zio->io_flags & ZIO_FLAG_NODATA)
	flags \|= METASLAB_DONT_THROTTLE;
	if (zio->io_flags & ZIO_FLAG_GANG_CHILD)
	flags \|= METASLAB_GANG_CHILD;
	if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE)
	flags \|= METASLAB_ASYNC_ALLOC;

	/*
	* if not already chosen, locate an appropriate allocation class
	*/
	mc = zio->io_metaslab_class;
	if (mc == NULL) {
	mc = spa_preferred_class(spa, zio->io_size,
	zio->io_prop.zp_type, zio->io_prop.zp_level,
	zio->io_prop.zp_zpl_smallblk);
	zio->io_metaslab_class = mc;
	}

	/*
	* Try allocating the block in the usual metaslab class.
	* If that's full, allocate it in the normal class.
	* If that's full, allocate as a gang block,
	* and if all are full, the allocation fails (which shouldn't happen).
	*
	* Note that we do not fall back on embedded slog (ZIL) space, to
	* preserve unfragmented slog space, which is critical for decent
	* sync write performance. If a log allocation fails, we will fall
	* back to spa_sync() which is abysmal for performance.
	*/
	error = metaslab_alloc(spa, mc, zio->io_size, bp,
	zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
	&zio->io_alloc_list, zio, zio->io_allocator);

	/*
	* Fallback to normal class when an alloc class is full
	*/
	if (error == ENOSPC && mc != spa_normal_class(spa)) {
	/*
	* If throttling, transfer reservation over to normal class.
	* The io_allocator slot can remain the same even though we
	* are switching classes.
	*/
	if (mc->mc_alloc_throttle_enabled &&
	(zio->io_flags & ZIO_FLAG_IO_ALLOCATING)) {
	metaslab_class_throttle_unreserve(mc,
	zio->io_prop.zp_copies, zio->io_allocator, zio);
	zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING;

	VERIFY(metaslab_class_throttle_reserve(
	spa_normal_class(spa),
	zio->io_prop.zp_copies, zio->io_allocator, zio,
	flags \| METASLAB_MUST_RESERVE));
	}
	zio->io_metaslab_class = mc = spa_normal_class(spa);
	if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
	zfs_dbgmsg("%s: metaslab allocation failure, "
	"trying normal class: zio %px, size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}

	error = metaslab_alloc(spa, mc, zio->io_size, bp,
	zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
	&zio->io_alloc_list, zio, zio->io_allocator);
	}

	if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) {
	if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
	zfs_dbgmsg("%s: metaslab allocation failure, "
	"trying ganging: zio %px, size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}
	return (zio_write_gang_block(zio, mc));
	}
	if (error != 0) {
	if (error != ENOSPC \|\|
	(zfs_flags & ZFS_DEBUG_METASLAB_ALLOC)) {
	zfs_dbgmsg("%s: metaslab allocation failure: zio %px, "
	"size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}
	zio->io_error = error;
	}

	return (zio);
	}

	static zio_t *
	zio_dva_free(zio_t *zio)
	{
	metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE);

	return (zio);
	}

	static zio_t *
	zio_dva_claim(zio_t *zio)
	{
	int error;

	error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg);
	if (error)
	zio->io_error = error;

	return (zio);
	}

	/*
	* Undo an allocation. This is used by zio_done() when an I/O fails
	* and we want to give back the block we just allocated.
	* This handles both normal blocks and gang blocks.
	*/
	static void
	zio_dva_unallocate(zio_t zio, zio_gang_node_t gn, blkptr_t *bp)
	{
	ASSERT(bp->blk_birth == zio->io_txg \|\| BP_IS_HOLE(bp));
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp))
	metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE);

	if (gn != NULL) {
	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	zio_dva_unallocate(zio, gn->gn_child[g],
	&gn->gn_gbh->zg_blkptr[g]);
	}
	}
	}

	/*
	* Try to allocate an intent log block. Return 0 on success, errno on failure.
	*/
	int
	zio_alloc_zil(spa_t spa, objset_t os, uint64_t txg, blkptr_t *new_bp,
	uint64_t size, boolean_t *slog)
	{
	int error = 1;
	zio_alloc_list_t io_alloc_list;

	ASSERT(txg > spa_syncing_txg(spa));

	metaslab_trace_init(&io_alloc_list);

	/*
	* Block pointer fields are useful to metaslabs for stats and debugging.
	* Fill in the obvious ones before calling into metaslab_alloc().
	*/
	BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
	BP_SET_PSIZE(new_bp, size);
	BP_SET_LEVEL(new_bp, 0);

	/*
	* When allocating a zil block, we don't have information about
	* the final destination of the block except the objset it's part
	* of, so we just hash the objset ID to pick the allocator to get
	* some parallelism.
	*/
	int flags = METASLAB_FASTWRITE \| METASLAB_ZIL;
	int allocator = (uint_t)cityhash4(0, 0, 0,
	os->os_dsl_dataset->ds_object) % spa->spa_alloc_count;
	error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
	txg, NULL, flags, &io_alloc_list, NULL, allocator);
	*slog = (error == 0);
	if (error != 0) {
	error = metaslab_alloc(spa, spa_embedded_log_class(spa), size,
	new_bp, 1, txg, NULL, flags,
	&io_alloc_list, NULL, allocator);
	}
	if (error != 0) {
	error = metaslab_alloc(spa, spa_normal_class(spa), size,
	new_bp, 1, txg, NULL, flags,
	&io_alloc_list, NULL, allocator);
	}
	metaslab_trace_fini(&io_alloc_list);

	if (error == 0) {
	BP_SET_LSIZE(new_bp, size);
	BP_SET_PSIZE(new_bp, size);
	BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(new_bp,
	spa_version(spa) >= SPA_VERSION_SLIM_ZIL
	? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG);
	BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
	BP_SET_LEVEL(new_bp, 0);
	BP_SET_DEDUP(new_bp, 0);
	BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER);

	/*
	* encrypted blocks will require an IV and salt. We generate
	* these now since we will not be rewriting the bp at
	* rewrite time.
	*/
	if (os->os_encrypted) {
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t salt[ZIO_DATA_SALT_LEN];

	BP_SET_CRYPT(new_bp, B_TRUE);
	VERIFY0(spa_crypt_get_salt(spa,
	dmu_objset_id(os), salt));
	VERIFY0(zio_crypt_generate_iv(iv));

	zio_crypt_encode_params_bp(new_bp, salt, iv);
	}
	} else {
	zfs_dbgmsg("%s: zil block allocation failure: "
	"size %llu, error %d", spa_name(spa), (u_longlong_t)size,
	error);
	}

	return (error);
	}

	/*
	* ==========================================================================
	* Read and write to physical devices
	* ==========================================================================
	*/

	/*
	* Issue an I/O to the underlying vdev. Typically the issue pipeline
	* stops after this stage and will resume upon I/O completion.
	* However, there are instances where the vdev layer may need to
	* continue the pipeline when an I/O was not issued. Since the I/O
	* that was sent to the vdev layer might be different than the one
	* currently active in the pipeline (see vdev_queue_io()), we explicitly
	* force the underlying vdev layers to call either zio_execute() or
	* zio_interrupt() to ensure that the pipeline continues with the correct I/O.
	*/
	static zio_t *
	zio_vdev_io_start(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	uint64_t align;
	spa_t *spa = zio->io_spa;

	zio->io_delay = 0;

	ASSERT(zio->io_error == 0);
	ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0);

	if (vd == NULL) {
	if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
	spa_config_enter(spa, SCL_ZIO, zio, RW_READER);

	/*
	* The mirror_ops handle multiple DVAs in a single BP.
	*/
	vdev_mirror_ops.vdev_op_io_start(zio);
	return (NULL);
	}

	ASSERT3P(zio->io_logical, !=, zio);
	if (zio->io_type == ZIO_TYPE_WRITE) {
	ASSERT(spa->spa_trust_config);

	/*
	* Note: the code can handle other kinds of writes,
	* but we don't expect them.
	*/
	if (zio->io_vd->vdev_noalloc) {
	ASSERT(zio->io_flags &
	(ZIO_FLAG_PHYSICAL \| ZIO_FLAG_SELF_HEAL \|
	ZIO_FLAG_RESILVER \| ZIO_FLAG_INDUCE_DAMAGE));
	}
	}

	align = 1ULL << vd->vdev_top->vdev_ashift;

	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) &&
	P2PHASE(zio->io_size, align) != 0) {
	/* Transform logical writes to be a full physical block size. */
	uint64_t asize = P2ROUNDUP(zio->io_size, align);
	abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize);
	ASSERT(vd == vd->vdev_top);
	if (zio->io_type == ZIO_TYPE_WRITE) {
	abd_copy(abuf, zio->io_abd, zio->io_size);
	abd_zero_off(abuf, zio->io_size, asize - zio->io_size);
	}
	zio_push_transform(zio, abuf, asize, asize, zio_subblock);
	}

	/*
	* If this is not a physical io, make sure that it is properly aligned
	* before proceeding.
	*/
	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) {
	ASSERT0(P2PHASE(zio->io_offset, align));
	ASSERT0(P2PHASE(zio->io_size, align));
	} else {
	/*
	* For physical writes, we allow 512b aligned writes and assume
	* the device will perform a read-modify-write as necessary.
	*/
	ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE));
	ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE));
	}

	VERIFY(zio->io_type != ZIO_TYPE_WRITE \|\| spa_writeable(spa));

	/*
	* If this is a repair I/O, and there's no self-healing involved --
	* that is, we're just resilvering what we expect to resilver --
	* then don't do the I/O unless zio's txg is actually in vd's DTL.
	* This prevents spurious resilvering.
	*
	* There are a few ways that we can end up creating these spurious
	* resilver i/os:
	*
	* 1. A resilver i/o will be issued if any DVA in the BP has a
	* dirty DTL. The mirror code will issue resilver writes to
	* each DVA, including the one(s) that are not on vdevs with dirty
	* DTLs.
	*
	* 2. With nested replication, which happens when we have a
	* "replacing" or "spare" vdev that's a child of a mirror or raidz.
	* For example, given mirror(replacing(A+B), C), it's likely that
	* only A is out of date (it's the new device). In this case, we'll
	* read from C, then use the data to resilver A+B -- but we don't
	* actually want to resilver B, just A. The top-level mirror has no
	* way to know this, so instead we just discard unnecessary repairs
	* as we work our way down the vdev tree.
	*
	* 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc.
	* The same logic applies to any form of nested replication: ditto
	* + mirror, RAID-Z + replacing, etc.
	*
	* However, indirect vdevs point off to other vdevs which may have
	* DTL's, so we never bypass them. The child i/os on concrete vdevs
	* will be properly bypassed instead.
	*
	* Leaf DTL_PARTIAL can be empty when a legitimate write comes from
	* a dRAID spare vdev. For example, when a dRAID spare is first
	* used, its spare blocks need to be written to but the leaf vdev's
	* of such blocks can have empty DTL_PARTIAL.
	*
	* There seemed no clean way to allow such writes while bypassing
	* spurious ones. At this point, just avoid all bypassing for dRAID
	* for correctness.
	*/
	if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) &&
	!(zio->io_flags & ZIO_FLAG_SELF_HEAL) &&
	zio->io_txg != 0 && /* not a delegated i/o */
	vd->vdev_ops != &vdev_indirect_ops &&
	vd->vdev_top->vdev_ops != &vdev_draid_ops &&
	!vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) {
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	zio_vdev_io_bypass(zio);
	return (zio);
	}

	/*
	* Select the next best leaf I/O to process. Distributed spares are
	* excluded since they dispatch the I/O directly to a leaf vdev after
	* applying the dRAID mapping.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops &&
	(zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_WRITE \|\|
	zio->io_type == ZIO_TYPE_TRIM)) {

	if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio))
	return (zio);

	if ((zio = vdev_queue_io(zio)) == NULL)
	return (NULL);

	if (!vdev_accessible(vd, zio)) {
	zio->io_error = SET_ERROR(ENXIO);
	zio_interrupt(zio);
	return (NULL);
	}
	zio->io_delay = gethrtime();
	}

	vd->vdev_ops->vdev_op_io_start(zio);
	return (NULL);
	}

	static zio_t *
	zio_vdev_io_done(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops;
	boolean_t unexpected_error = B_FALSE;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_WRITE \|\| zio->io_type == ZIO_TYPE_TRIM);

	if (zio->io_delay)
	zio->io_delay = gethrtime() - zio->io_delay;

	if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops) {
	vdev_queue_io_done(zio);

	if (zio->io_type == ZIO_TYPE_WRITE)
	vdev_cache_write(zio);

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_device_injections(vd, zio,
	EIO, EILSEQ);

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_label_injection(zio, EIO);

	if (zio->io_error && zio->io_type != ZIO_TYPE_TRIM) {
	if (!vdev_accessible(vd, zio)) {
	zio->io_error = SET_ERROR(ENXIO);
	} else {
	unexpected_error = B_TRUE;
	}
	}
	}

	ops->vdev_op_io_done(zio);

	- if (unexpected_error)
	+ if (unexpected_error && vd->vdev_remove_wanted == B_FALSE)
	VERIFY(vdev_probe(vd, zio) == NULL);

	return (zio);
	}

	/*
	* This function is used to change the priority of an existing zio that is
	* currently in-flight. This is used by the arc to upgrade priority in the
	* event that a demand read is made for a block that is currently queued
	* as a scrub or async read IO. Otherwise, the high priority read request
	* would end up having to wait for the lower priority IO.
	*/
	void
	zio_change_priority(zio_t *pio, zio_priority_t priority)
	{
	zio_t cio, cio_next;
	zio_link_t *zl = NULL;

	ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);

	if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) {
	vdev_queue_change_io_priority(pio, priority);
	} else {
	pio->io_priority = priority;
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	zio_change_priority(cio, priority);
	}
	mutex_exit(&pio->io_lock);
	}

	/*
	* For non-raidz ZIOs, we can just copy aside the bad data read from the
	* disk, and use that to finish the checksum ereport later.
	*/
	static void
	zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr,
	const abd_t *good_buf)
	{
	/* no processing needed */
	zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE);
	}

	void
	zio_vsd_default_cksum_report(zio_t zio, zio_cksum_report_t zcr)
	{
	void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size);

	abd_copy(abd, zio->io_abd, zio->io_size);

	zcr->zcr_cbinfo = zio->io_size;
	zcr->zcr_cbdata = abd;
	zcr->zcr_finish = zio_vsd_default_cksum_finish;
	zcr->zcr_free = zio_abd_free;
	}

	static zio_t *
	zio_vdev_io_assess(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
	spa_config_exit(zio->io_spa, SCL_ZIO, zio);

	if (zio->io_vsd != NULL) {
	zio->io_vsd_ops->vsd_free(zio);
	zio->io_vsd = NULL;
	}

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_fault_injection(zio, EIO);

	/*
	* If the I/O failed, determine whether we should attempt to retry it.
	*
	* On retry, we cut in line in the issue queue, since we don't want
	* compression/checksumming/etc. work to prevent our (cheap) IO reissue.
	*/
	if (zio->io_error && vd == NULL &&
	!(zio->io_flags & (ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_IO_RETRY))) {
	ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */
	zio->io_error = 0;
	zio->io_flags \|= ZIO_FLAG_IO_RETRY \|
	ZIO_FLAG_DONT_CACHE \| ZIO_FLAG_DONT_AGGREGATE;
	zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE,
	zio_requeue_io_start_cut_in_line);
	return (NULL);
	}

	/*
	* If we got an error on a leaf device, convert it to ENXIO
	* if the device is not accessible at all.
	*/
	if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	!vdev_accessible(vd, zio))
	zio->io_error = SET_ERROR(ENXIO);

	/*
	* If we can't write to an interior vdev (mirror or RAID-Z),
	* set vdev_cant_write so that we stop trying to allocate from it.
	*/
	if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
	vd != NULL && !vd->vdev_ops->vdev_op_leaf) {
	vdev_dbgmsg(vd, "zio_vdev_io_assess(zio=%px) setting "
	"cant_write=TRUE due to write failure with ENXIO",
	zio);
	vd->vdev_cant_write = B_TRUE;
	}

	/*
	* If a cache flush returns ENOTSUP or ENOTTY, we know that no future
	* attempts will ever succeed. In this case we set a persistent
	* boolean flag so that we don't bother with it in the future.
	*/
	if ((zio->io_error == ENOTSUP \|\| zio->io_error == ENOTTY) &&
	zio->io_type == ZIO_TYPE_IOCTL &&
	zio->io_cmd == DKIOCFLUSHWRITECACHE && vd != NULL)
	vd->vdev_nowritecache = B_TRUE;

	if (zio->io_error)
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	zio->io_physdone != NULL) {
	ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED));
	ASSERT(zio->io_child_type == ZIO_CHILD_VDEV);
	zio->io_physdone(zio->io_logical);
	}

	return (zio);
	}

	void
	zio_vdev_io_reissue(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_stage >>= 1;
	}

	void
	zio_vdev_io_redone(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE);

	zio->io_stage >>= 1;
	}

	void
	zio_vdev_io_bypass(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_flags \|= ZIO_FLAG_IO_BYPASS;
	zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1;
	}

	/*
	* ==========================================================================
	* Encrypt and store encryption parameters
	* ==========================================================================
	*/


	/*
	* This function is used for ZIO_STAGE_ENCRYPT. It is responsible for
	* managing the storage of encryption parameters and passing them to the
	* lower-level encryption functions.
	*/
	static zio_t *
	zio_encrypt(zio_t *zio)
	{
	zio_prop_t *zp = &zio->io_prop;
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t psize = BP_GET_PSIZE(bp);
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	void *enc_buf = NULL;
	abd_t *eabd = NULL;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	/* the root zio already encrypted the data */
	if (zio->io_child_type == ZIO_CHILD_GANG)
	return (zio);

	/* only ZIL blocks are re-encrypted on rewrite */
	if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG)
	return (zio);

	if (!(zp->zp_encrypt \|\| BP_IS_ENCRYPTED(bp))) {
	BP_SET_CRYPT(bp, B_FALSE);
	return (zio);
	}

	/* if we are doing raw encryption set the provided encryption params */
	if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) {
	ASSERT0(BP_GET_LEVEL(bp));
	BP_SET_CRYPT(bp, B_TRUE);
	BP_SET_BYTEORDER(bp, zp->zp_byteorder);
	if (ot != DMU_OT_OBJSET)
	zio_crypt_encode_mac_bp(bp, zp->zp_mac);

	/* dnode blocks must be written out in the provided byteorder */
	if (zp->zp_byteorder != ZFS_HOST_BYTEORDER &&
	ot == DMU_OT_DNODE) {
	void *bswap_buf = zio_buf_alloc(psize);
	abd_t *babd = abd_get_from_buf(bswap_buf, psize);

	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	abd_copy_to_buf(bswap_buf, zio->io_abd, psize);
	dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf,
	psize);

	abd_take_ownership_of_buf(babd, B_TRUE);
	zio_push_transform(zio, babd, psize, psize, NULL);
	}

	if (DMU_OT_IS_ENCRYPTED(ot))
	zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv);
	return (zio);
	}

	/* indirect blocks only maintain a cksum of the lower level MACs */
	if (BP_GET_LEVEL(bp) > 0) {
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE,
	zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp),
	mac));
	zio_crypt_encode_mac_bp(bp, mac);
	return (zio);
	}

	/*
	* Objset blocks are a special case since they have 2 256-bit MACs
	* embedded within them.
	*/
	if (ot == DMU_OT_OBJSET) {
	ASSERT0(DMU_OT_IS_ENCRYPTED(ot));
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj,
	zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp)));
	return (zio);
	}

	/* unencrypted object types are only authenticated with a MAC */
	if (!DMU_OT_IS_ENCRYPTED(ot)) {
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj,
	zio->io_abd, psize, mac));
	zio_crypt_encode_mac_bp(bp, mac);
	return (zio);
	}

	/*
	* Later passes of sync-to-convergence may decide to rewrite data
	* in place to avoid more disk reallocations. This presents a problem
	* for encryption because this constitutes rewriting the new data with
	* the same encryption key and IV. However, this only applies to blocks
	* in the MOS (particularly the spacemaps) and we do not encrypt the
	* MOS. We assert that the zio is allocating or an intent log write
	* to enforce this.
	*/
	ASSERT(IO_IS_ALLOCATING(zio) \|\| ot == DMU_OT_INTENT_LOG);
	ASSERT(BP_GET_LEVEL(bp) == 0 \|\| ot == DMU_OT_INTENT_LOG);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION));
	ASSERT3U(psize, !=, 0);

	enc_buf = zio_buf_alloc(psize);
	eabd = abd_get_from_buf(enc_buf, psize);
	abd_take_ownership_of_buf(eabd, B_TRUE);

	/*
	* For an explanation of what encryption parameters are stored
	* where, see the block comment in zio_crypt.c.
	*/
	if (ot == DMU_OT_INTENT_LOG) {
	zio_crypt_decode_params_bp(bp, salt, iv);
	} else {
	BP_SET_CRYPT(bp, B_TRUE);
	}

	/* Perform the encryption. This should not fail */
	VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark,
	BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
	salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt));

	/* encode encryption metadata into the bp */
	if (ot == DMU_OT_INTENT_LOG) {
	/*
	* ZIL blocks store the MAC in the embedded checksum, so the
	* transform must always be applied.
	*/
	zio_crypt_encode_mac_zil(enc_buf, mac);
	zio_push_transform(zio, eabd, psize, psize, NULL);
	} else {
	BP_SET_CRYPT(bp, B_TRUE);
	zio_crypt_encode_params_bp(bp, salt, iv);
	zio_crypt_encode_mac_bp(bp, mac);

	if (no_crypt) {
	ASSERT3U(ot, ==, DMU_OT_DNODE);
	abd_free(eabd);
	} else {
	zio_push_transform(zio, eabd, psize, psize, NULL);
	}
	}

	return (zio);
	}

	/*
	* ==========================================================================
	* Generate and verify checksums
	* ==========================================================================
	*/
	static zio_t *
	zio_checksum_generate(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	enum zio_checksum checksum;

	if (bp == NULL) {
	/*
	* This is zio_write_phys().
	* We're either generating a label checksum, or none at all.
	*/
	checksum = zio->io_prop.zp_checksum;

	if (checksum == ZIO_CHECKSUM_OFF)
	return (zio);

	ASSERT(checksum == ZIO_CHECKSUM_LABEL);
	} else {
	if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) {
	ASSERT(!IO_IS_ALLOCATING(zio));
	checksum = ZIO_CHECKSUM_GANG_HEADER;
	} else {
	checksum = BP_GET_CHECKSUM(bp);
	}
	}

	zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size);

	return (zio);
	}

	static zio_t *
	zio_checksum_verify(zio_t *zio)
	{
	zio_bad_cksum_t info;
	blkptr_t *bp = zio->io_bp;
	int error;

	ASSERT(zio->io_vd != NULL);

	if (bp == NULL) {
	/*
	* This is zio_read_phys().
	* We're either verifying a label checksum, or nothing at all.
	*/
	if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF)
	return (zio);

	ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL);
	}

	if ((error = zio_checksum_error(zio, &info)) != 0) {
	zio->io_error = error;
	if (error == ECKSUM &&
	!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
	(void) zfs_ereport_start_checksum(zio->io_spa,
	zio->io_vd, &zio->io_bookmark, zio,
	zio->io_offset, zio->io_size, &info);
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	zio->io_vd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);
	}
	}

	return (zio);
	}

	/*
	* Called by RAID-Z to ensure we don't compute the checksum twice.
	*/
	void
	zio_checksum_verified(zio_t *zio)
	{
	zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
	}

	/*
	* ==========================================================================
	* Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
	* An error of 0 indicates success. ENXIO indicates whole-device failure,
	* which may be transient (e.g. unplugged) or permanent. ECKSUM and EIO
	* indicate errors that are specific to one I/O, and most likely permanent.
	* Any other error is presumed to be worse because we weren't expecting it.
	* ==========================================================================
	*/
	int
	zio_worst_error(int e1, int e2)
	{
	static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO };
	int r1, r2;

	for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++)
	if (e1 == zio_error_rank[r1])
	break;

	for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++)
	if (e2 == zio_error_rank[r2])
	break;

	return (r1 > r2 ? e1 : e2);
	}

	/*
	* ==========================================================================
	* I/O completion
	* ==========================================================================
	*/
	static zio_t *
	zio_ready(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	zio_t pio, pio_next;
	zio_link_t *zl = NULL;

	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT \| ZIO_CHILD_DDT_BIT,
	ZIO_WAIT_READY)) {
	return (NULL);
	}

	if (zio->io_ready) {
	ASSERT(IO_IS_ALLOCATING(zio));
	ASSERT(bp->blk_birth == zio->io_txg \|\| BP_IS_HOLE(bp) \|\|
	(zio->io_flags & ZIO_FLAG_NOPWRITE));
	ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);

	zio->io_ready(zio);
	}

	if (bp != NULL && bp != &zio->io_bp_copy)
	zio->io_bp_copy = *bp;

	if (zio->io_error != 0) {
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(IO_IS_ALLOCATING(zio));
	ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(zio->io_metaslab_class != NULL);

	/*
	* We were unable to allocate anything, unreserve and
	* issue the next I/O to allocate.
	*/
	metaslab_class_throttle_unreserve(
	zio->io_metaslab_class, zio->io_prop.zp_copies,
	zio->io_allocator, zio);
	zio_allocate_dispatch(zio->io_spa, zio->io_allocator);
	}
	}

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_READY] = 1;
	pio = zio_walk_parents(zio, &zl);
	mutex_exit(&zio->io_lock);

	/*
	* As we notify zio's parents, new parents could be added.
	* New parents go to the head of zio's io_parent_list, however,
	* so we will (correctly) not notify them. The remainder of zio's
	* io_parent_list, from 'pio_next' onward, cannot change because
	* all parents must wait for us to be done before they can be done.
	*/
	for (; pio != NULL; pio = pio_next) {
	pio_next = zio_walk_parents(zio, &zl);
	zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL);
	}

	if (zio->io_flags & ZIO_FLAG_NODATA) {
	if (BP_IS_GANG(bp)) {
	zio->io_flags &= ~ZIO_FLAG_NODATA;
	} else {
	ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE);
	zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	}
	}

	if (zio_injection_enabled &&
	zio->io_spa->spa_syncing_txg == zio->io_txg)
	zio_handle_ignored_writes(zio);

	return (zio);
	}

	/*
	* Update the allocation throttle accounting.
	*/
	static void
	zio_dva_throttle_done(zio_t *zio)
	{
	zio_t *lio __maybe_unused = zio->io_logical;
	zio_t *pio = zio_unique_parent(zio);
	vdev_t *vd = zio->io_vd;
	int flags = METASLAB_ASYNC_ALLOC;

	ASSERT3P(zio->io_bp, !=, NULL);
	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
	ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);
	ASSERT(vd != NULL);
	ASSERT3P(vd, ==, vd->vdev_top);
	ASSERT(zio_injection_enabled \|\| !(zio->io_flags & ZIO_FLAG_IO_RETRY));
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING);
	ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA));

	/*
	* Parents of gang children can have two flavors -- ones that
	* allocated the gang header (will have ZIO_FLAG_IO_REWRITE set)
	* and ones that allocated the constituent blocks. The allocation
	* throttle needs to know the allocating parent zio so we must find
	* it here.
	*/
	if (pio->io_child_type == ZIO_CHILD_GANG) {
	/*
	* If our parent is a rewrite gang child then our grandparent
	* would have been the one that performed the allocation.
	*/
	if (pio->io_flags & ZIO_FLAG_IO_REWRITE)
	pio = zio_unique_parent(pio);
	flags \|= METASLAB_GANG_CHILD;
	}

	ASSERT(IO_IS_ALLOCATING(pio));
	ASSERT3P(zio, !=, zio->io_logical);
	ASSERT(zio->io_logical != NULL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE);
	ASSERT(zio->io_metaslab_class != NULL);

	mutex_enter(&pio->io_lock);
	metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags,
	pio->io_allocator, B_TRUE);
	mutex_exit(&pio->io_lock);

	metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1,
	pio->io_allocator, pio);

	/*
	* Call into the pipeline to see if there is more work that
	* needs to be done. If there is work to be done it will be
	* dispatched to another taskq thread.
	*/
	zio_allocate_dispatch(zio->io_spa, pio->io_allocator);
	}

	static zio_t *
	zio_done(zio_t *zio)
	{
	/*
	* Always attempt to keep stack usage minimal here since
	* we can be called recursively up to 19 levels deep.
	*/
	const uint64_t psize = zio->io_size;
	zio_t pio, pio_next;
	zio_link_t *zl = NULL;

	/*
	* If our children haven't all completed,
	* wait for them and then repeat this pipeline stage.
	*/
	if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	/*
	* If the allocation throttle is enabled, then update the accounting.
	* We only track child I/Os that are part of an allocating async
	* write. We must do this since the allocation is performed
	* by the logical I/O but the actual write is done by child I/Os.
	*/
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING &&
	zio->io_child_type == ZIO_CHILD_VDEV) {
	ASSERT(zio->io_metaslab_class != NULL);
	ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled);
	zio_dva_throttle_done(zio);
	}

	/*
	* If the allocation throttle is enabled, verify that
	* we have decremented the refcounts for every I/O that was throttled.
	*/
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(zio->io_bp != NULL);

	metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio,
	zio->io_allocator);
	VERIFY(zfs_refcount_not_held(&zio->io_metaslab_class->
	mc_allocator[zio->io_allocator].mca_alloc_slots, zio));
	}


	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	ASSERT(zio->io_children[c][w] == 0);

	if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) {
	ASSERT(zio->io_bp->blk_pad[0] == 0);
	ASSERT(zio->io_bp->blk_pad[1] == 0);
	ASSERT(memcmp(zio->io_bp, &zio->io_bp_copy,
	sizeof (blkptr_t)) == 0 \|\|
	(zio->io_bp == zio_unique_parent(zio)->io_bp));
	if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) &&
	zio->io_bp_override == NULL &&
	!(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
	ASSERT3U(zio->io_prop.zp_copies, <=,
	BP_GET_NDVAS(zio->io_bp));
	ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 \|\|
	(BP_COUNT_GANG(zio->io_bp) ==
	BP_GET_NDVAS(zio->io_bp)));
	}
	if (zio->io_flags & ZIO_FLAG_NOPWRITE)
	VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
	}

	/*
	* If there were child vdev/gang/ddt errors, they apply to us now.
	*/
	zio_inherit_child_errors(zio, ZIO_CHILD_VDEV);
	zio_inherit_child_errors(zio, ZIO_CHILD_GANG);
	zio_inherit_child_errors(zio, ZIO_CHILD_DDT);

	/*
	* If the I/O on the transformed data was successful, generate any
	* checksum reports now while we still have the transformed data.
	*/
	if (zio->io_error == 0) {
	while (zio->io_cksum_report != NULL) {
	zio_cksum_report_t *zcr = zio->io_cksum_report;
	uint64_t align = zcr->zcr_align;
	uint64_t asize = P2ROUNDUP(psize, align);
	abd_t *adata = zio->io_abd;

	if (adata != NULL && asize != psize) {
	adata = abd_alloc(asize, B_TRUE);
	abd_copy(adata, zio->io_abd, psize);
	abd_zero_off(adata, psize, asize - psize);
	}

	zio->io_cksum_report = zcr->zcr_next;
	zcr->zcr_next = NULL;
	zcr->zcr_finish(zcr, adata);
	zfs_ereport_free_checksum(zcr);

	if (adata != NULL && asize != psize)
	abd_free(adata);
	}
	}

	zio_pop_transforms(zio); /* note: may set zio->io_error */

	vdev_stat_update(zio, psize);

	/*
	* If this I/O is attached to a particular vdev is slow, exceeding
	* 30 seconds to complete, post an error described the I/O delay.
	* We ignore these errors if the device is currently unavailable.
	*/
	if (zio->io_delay >= MSEC2NSEC(zio_slow_io_ms)) {
	if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) {
	/*
	* We want to only increment our slow IO counters if
	* the IO is valid (i.e. not if the drive is removed).
	*
	* zfs_ereport_post() will also do these checks, but
	* it can also ratelimit and have other failures, so we
	* need to increment the slow_io counters independent
	* of it.
	*/
	if (zfs_ereport_is_valid(FM_EREPORT_ZFS_DELAY,
	zio->io_spa, zio->io_vd, zio)) {
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	zio->io_vd->vdev_stat.vs_slow_ios++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);

	(void) zfs_ereport_post(FM_EREPORT_ZFS_DELAY,
	zio->io_spa, zio->io_vd, &zio->io_bookmark,
	zio, 0);
	}
	}
	}

	if (zio->io_error) {
	/*
	* If this I/O is attached to a particular vdev,
	* generate an error message describing the I/O failure
	* at the block level. We ignore these errors if the
	* device is currently unavailable.
	*/
	if (zio->io_error != ECKSUM && zio->io_vd != NULL &&
	!vdev_is_dead(zio->io_vd)) {
	int ret = zfs_ereport_post(FM_EREPORT_ZFS_IO,
	zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0);
	if (ret != EALREADY) {
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	if (zio->io_type == ZIO_TYPE_READ)
	zio->io_vd->vdev_stat.vs_read_errors++;
	else if (zio->io_type == ZIO_TYPE_WRITE)
	zio->io_vd->vdev_stat.vs_write_errors++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);
	}
	}

	if ((zio->io_error == EIO \|\| !(zio->io_flags &
	(ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_DONT_PROPAGATE))) &&
	zio == zio->io_logical) {
	/*
	* For logical I/O requests, tell the SPA to log the
	* error and generate a logical data ereport.
	*/
	spa_log_error(zio->io_spa, &zio->io_bookmark);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_DATA,
	zio->io_spa, NULL, &zio->io_bookmark, zio, 0);
	}
	}

	if (zio->io_error && zio == zio->io_logical) {
	/*
	* Determine whether zio should be reexecuted. This will
	* propagate all the way to the root via zio_notify_parent().
	*/
	ASSERT(zio->io_vd == NULL && zio->io_bp != NULL);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (IO_IS_ALLOCATING(zio) &&
	!(zio->io_flags & ZIO_FLAG_CANFAIL)) {
	if (zio->io_error != ENOSPC)
	zio->io_reexecute \|= ZIO_REEXECUTE_NOW;
	else
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;
	}

	if ((zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_FREE) &&
	!(zio->io_flags & ZIO_FLAG_SCAN_THREAD) &&
	zio->io_error == ENXIO &&
	spa_load_state(zio->io_spa) == SPA_LOAD_NONE &&
	spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE)
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;

	if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute)
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;

	/*
	* Here is a possibly good place to attempt to do
	* either combinatorial reconstruction or error correction
	* based on checksums. It also might be a good place
	* to send out preliminary ereports before we suspend
	* processing.
	*/
	}

	/*
	* If there were logical child errors, they apply to us now.
	* We defer this until now to avoid conflating logical child
	* errors with errors that happened to the zio itself when
	* updating vdev stats and reporting FMA events above.
	*/
	zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL);

	if ((zio->io_error \|\| zio->io_reexecute) &&
	IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio &&
	!(zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)))
	zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp);

	zio_gang_tree_free(&zio->io_gang_tree);

	/*
	* Godfather I/Os should never suspend.
	*/
	if ((zio->io_flags & ZIO_FLAG_GODFATHER) &&
	(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND))
	zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND;

	if (zio->io_reexecute) {
	/*
	* This is a logical I/O that wants to reexecute.
	*
	* Reexecute is top-down. When an i/o fails, if it's not
	* the root, it simply notifies its parent and sticks around.
	* The parent, seeing that it still has children in zio_done(),
	* does the same. This percolates all the way up to the root.
	* The root i/o will reexecute or suspend the entire tree.
	*
	* This approach ensures that zio_reexecute() honors
	* all the original i/o dependency relationships, e.g.
	* parents not executing until children are ready.
	*/
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	zio->io_gang_leader = NULL;

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_DONE] = 1;
	mutex_exit(&zio->io_lock);

	/*
	* "The Godfather" I/O monitors its children but is
	* not a true parent to them. It will track them through
	* the pipeline but severs its ties whenever they get into
	* trouble (e.g. suspended). This allows "The Godfather"
	* I/O to return status without blocking.
	*/
	zl = NULL;
	for (pio = zio_walk_parents(zio, &zl); pio != NULL;
	pio = pio_next) {
	zio_link_t *remove_zl = zl;
	pio_next = zio_walk_parents(zio, &zl);

	if ((pio->io_flags & ZIO_FLAG_GODFATHER) &&
	(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) {
	zio_remove_child(pio, zio, remove_zl);
	/*
	* This is a rare code path, so we don't
	* bother with "next_to_execute".
	*/
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE,
	NULL);
	}
	}

	if ((pio = zio_unique_parent(zio)) != NULL) {
	/*
	* We're not a root i/o, so there's nothing to do
	* but notify our parent. Don't propagate errors
	* upward since we haven't permanently failed yet.
	*/
	ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
	zio->io_flags \|= ZIO_FLAG_DONT_PROPAGATE;
	/*
	* This is a rare code path, so we don't bother with
	* "next_to_execute".
	*/
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL);
	} else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) {
	/*
	* We'd fail again if we reexecuted now, so suspend
	* until conditions improve (e.g. device comes online).
	*/
	zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR);
	} else {
	/*
	* Reexecution is potentially a huge amount of work.
	* Hand it off to the otherwise-unused claim taskq.
	*/
	ASSERT(taskq_empty_ent(&zio->io_tqent));
	spa_taskq_dispatch_ent(zio->io_spa,
	ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE,
	zio_reexecute, zio, 0, &zio->io_tqent);
	}
	return (NULL);
	}

	ASSERT(zio->io_child_count == 0);
	ASSERT(zio->io_reexecute == 0);
	ASSERT(zio->io_error == 0 \|\| (zio->io_flags & ZIO_FLAG_CANFAIL));

	/*
	* Report any checksum errors, since the I/O is complete.
	*/
	while (zio->io_cksum_report != NULL) {
	zio_cksum_report_t *zcr = zio->io_cksum_report;
	zio->io_cksum_report = zcr->zcr_next;
	zcr->zcr_next = NULL;
	zcr->zcr_finish(zcr, NULL);
	zfs_ereport_free_checksum(zcr);
	}

	if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
	!BP_IS_HOLE(zio->io_bp) && !BP_IS_EMBEDDED(zio->io_bp) &&
	!(zio->io_flags & ZIO_FLAG_NOPWRITE)) {
	metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
	}

	/*
	* It is the responsibility of the done callback to ensure that this
	* particular zio is no longer discoverable for adoption, and as
	* such, cannot acquire any new parents.
	*/
	if (zio->io_done)
	zio->io_done(zio);

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_DONE] = 1;
	mutex_exit(&zio->io_lock);

	/*
	* We are done executing this zio. We may want to execute a parent
	* next. See the comment in zio_notify_parent().
	*/
	zio_t *next_to_execute = NULL;
	zl = NULL;
	for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) {
	zio_link_t *remove_zl = zl;
	pio_next = zio_walk_parents(zio, &zl);
	zio_remove_child(pio, zio, remove_zl);
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute);
	}

	if (zio->io_waiter != NULL) {
	mutex_enter(&zio->io_lock);
	zio->io_executor = NULL;
	cv_broadcast(&zio->io_cv);
	mutex_exit(&zio->io_lock);
	} else {
	zio_destroy(zio);
	}

	return (next_to_execute);
	}

	/*
	* ==========================================================================
	* I/O pipeline definition
	* ==========================================================================
	*/
	static zio_pipe_stage_t *zio_pipeline[] = {
	NULL,
	zio_read_bp_init,
	zio_write_bp_init,
	zio_free_bp_init,
	zio_issue_async,
	zio_write_compress,
	zio_encrypt,
	zio_checksum_generate,
	zio_nop_write,
	zio_ddt_read_start,
	zio_ddt_read_done,
	zio_ddt_write,
	zio_ddt_free,
	zio_gang_assemble,
	zio_gang_issue,
	zio_dva_throttle,
	zio_dva_allocate,
	zio_dva_free,
	zio_dva_claim,
	zio_ready,
	zio_vdev_io_start,
	zio_vdev_io_done,
	zio_vdev_io_assess,
	zio_checksum_verify,
	zio_done
	};




	/*
	* Compare two zbookmark_phys_t's to see which we would reach first in a
	* pre-order traversal of the object tree.
	*
	* This is simple in every case aside from the meta-dnode object. For all other
	* objects, we traverse them in order (object 1 before object 2, and so on).
	* However, all of these objects are traversed while traversing object 0, since
	* the data it points to is the list of objects. Thus, we need to convert to a
	* canonical representation so we can compare meta-dnode bookmarks to
	* non-meta-dnode bookmarks.
	*
	* We do this by calculating "equivalents" for each field of the zbookmark.
	* zbookmarks outside of the meta-dnode use their own object and level, and
	* calculate the level 0 equivalent (the first L0 blkid that is contained in the
	* blocks this bookmark refers to) by multiplying their blkid by their span
	* (the number of L0 blocks contained within one block at their level).
	* zbookmarks inside the meta-dnode calculate their object equivalent
	* (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use
	* level + 1<<31 (any value larger than a level could ever be) for their level.
	* This causes them to always compare before a bookmark in their object
	* equivalent, compare appropriately to bookmarks in other objects, and to
	* compare appropriately to other bookmarks in the meta-dnode.
	*/
	int
	zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2,
	const zbookmark_phys_t zb1, const zbookmark_phys_t zb2)
	{
	/*
	* These variables represent the "equivalent" values for the zbookmark,
	* after converting zbookmarks inside the meta dnode to their
	* normal-object equivalents.
	*/
	uint64_t zb1obj, zb2obj;
	uint64_t zb1L0, zb2L0;
	uint64_t zb1level, zb2level;

	if (zb1->zb_object == zb2->zb_object &&
	zb1->zb_level == zb2->zb_level &&
	zb1->zb_blkid == zb2->zb_blkid)
	return (0);

	IMPLY(zb1->zb_level > 0, ibs1 >= SPA_MINBLOCKSHIFT);
	IMPLY(zb2->zb_level > 0, ibs2 >= SPA_MINBLOCKSHIFT);

	/*
	* BP_SPANB calculates the span in blocks.
	*/
	zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level);
	zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level);

	if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
	zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
	zb1L0 = 0;
	zb1level = zb1->zb_level + COMPARE_META_LEVEL;
	} else {
	zb1obj = zb1->zb_object;
	zb1level = zb1->zb_level;
	}

	if (zb2->zb_object == DMU_META_DNODE_OBJECT) {
	zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
	zb2L0 = 0;
	zb2level = zb2->zb_level + COMPARE_META_LEVEL;
	} else {
	zb2obj = zb2->zb_object;
	zb2level = zb2->zb_level;
	}

	/* Now that we have a canonical representation, do the comparison. */
	if (zb1obj != zb2obj)
	return (zb1obj < zb2obj ? -1 : 1);
	else if (zb1L0 != zb2L0)
	return (zb1L0 < zb2L0 ? -1 : 1);
	else if (zb1level != zb2level)
	return (zb1level > zb2level ? -1 : 1);
	/*
	* This can (theoretically) happen if the bookmarks have the same object
	* and level, but different blkids, if the block sizes are not the same.
	* There is presently no way to change the indirect block sizes
	*/
	return (0);
	}

	/*
	* This function checks the following: given that last_block is the place that
	* our traversal stopped last time, does that guarantee that we've visited
	* every node under subtree_root? Therefore, we can't just use the raw output
	* of zbookmark_compare. We have to pass in a modified version of
	* subtree_root; by incrementing the block id, and then checking whether
	* last_block is before or equal to that, we can tell whether or not having
	* visited last_block implies that all of subtree_root's children have been
	* visited.
	*/
	boolean_t
	zbookmark_subtree_completed(const dnode_phys_t *dnp,
	const zbookmark_phys_t subtree_root, const zbookmark_phys_t last_block)
	{
	zbookmark_phys_t mod_zb = *subtree_root;
	mod_zb.zb_blkid++;
	ASSERT0(last_block->zb_level);

	/* The objset_phys_t isn't before anything. */
	if (dnp == NULL)
	return (B_FALSE);

	/*
	* We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the
	* data block size in sectors, because that variable is only used if
	* the bookmark refers to a block in the meta-dnode. Since we don't
	* know without examining it what object it refers to, and there's no
	* harm in passing in this value in other cases, we always pass it in.
	*
	* We pass in 0 for the indirect block size shift because zb2 must be
	* level 0. The indirect block size is only used to calculate the span
	* of the bookmark, but since the bookmark must be level 0, the span is
	* always 1, so the math works out.
	*
	* If you make changes to how the zbookmark_compare code works, be sure
	* to make sure that this code still works afterwards.
	*/
	return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
	1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb,
	last_block) <= 0);
	}

	/*
	* This function is similar to zbookmark_subtree_completed(), but returns true
	* if subtree_root is equal or ahead of last_block, i.e. still to be done.
	*/
	boolean_t
	zbookmark_subtree_tbd(const dnode_phys_t *dnp,
	const zbookmark_phys_t subtree_root, const zbookmark_phys_t last_block)
	{
	ASSERT0(last_block->zb_level);
	if (dnp == NULL)
	return (B_FALSE);
	return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
	1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, subtree_root,
	last_block) >= 0);
	}

	EXPORT_SYMBOL(zio_type_name);
	EXPORT_SYMBOL(zio_buf_alloc);
	EXPORT_SYMBOL(zio_data_buf_alloc);
	EXPORT_SYMBOL(zio_buf_free);
	EXPORT_SYMBOL(zio_data_buf_free);

	ZFS_MODULE_PARAM(zfs_zio, zio_, slow_io_ms, INT, ZMOD_RW,
	"Max I/O completion time (milliseconds) before marking it as slow");

	ZFS_MODULE_PARAM(zfs_zio, zio_, requeue_io_start_cut_in_line, INT, ZMOD_RW,
	"Prioritize requeued I/O");

	-ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_deferred_free, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_deferred_free, UINT, ZMOD_RW,
	"Defer frees starting in this pass");

	-ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_dont_compress, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_dont_compress, UINT, ZMOD_RW,
	"Don't compress starting in this pass");

	-ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_rewrite, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_rewrite, UINT, ZMOD_RW,
	"Rewrite new bps starting in this pass");

	ZFS_MODULE_PARAM(zfs_zio, zio_, dva_throttle_enabled, INT, ZMOD_RW,
	"Throttle block allocations in the ZIO pipeline");

	ZFS_MODULE_PARAM(zfs_zio, zio_, deadman_log_all, INT, ZMOD_RW,
	"Log all slow ZIOs, not just those with vdevs");
	diff --git a/sys/contrib/openzfs/module/zfs/zio_checksum.c b/sys/contrib/openzfs/module/zfs/zio_checksum.c
	index b3c5fbbd8bba..37cd35bc847f 100644
	--- a/sys/contrib/openzfs/module/zfs/zio_checksum.c
	+++ b/sys/contrib/openzfs/module/zfs/zio_checksum.c
	@@ -1,572 +1,572 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2016 by Delphix. All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/zil.h>
	#include <sys/abd.h>
	#include <zfs_fletcher.h>

	/*
	* Checksum vectors.
	*
	* In the SPA, everything is checksummed. We support checksum vectors
	* for three distinct reasons:
	*
	* 1. Different kinds of data need different levels of protection.
	* For SPA metadata, we always want a very strong checksum.
	* For user data, we let users make the trade-off between speed
	* and checksum strength.
	*
	* 2. Cryptographic hash and MAC algorithms are an area of active research.
	* It is likely that in future hash functions will be at least as strong
	* as current best-of-breed, and may be substantially faster as well.
	* We want the ability to take advantage of these new hashes as soon as
	* they become available.
	*
	* 3. If someone develops hardware that can compute a strong hash quickly,
	* we want the ability to take advantage of that hardware.
	*
	* Of course, we don't want a checksum upgrade to invalidate existing
	* data, so we store the checksum function in eight bits of the bp.
	* This gives us room for up to 256 different checksum functions.
	*
	* When writing a block, we always checksum it with the latest-and-greatest
	* checksum function of the appropriate strength. When reading a block,
	* we compare the expected checksum against the actual checksum, which we
	* compute via the checksum function specified by BP_GET_CHECKSUM(bp).
	*
	* SALTED CHECKSUMS
	*
	* To enable the use of less secure hash algorithms with dedup, we
	* introduce the notion of salted checksums (MACs, really). A salted
	* checksum is fed both a random 256-bit value (the salt) and the data
	* to be checksummed. This salt is kept secret (stored on the pool, but
	* never shown to the user). Thus even if an attacker knew of collision
	* weaknesses in the hash algorithm, they won't be able to mount a known
	* plaintext attack on the DDT, since the actual hash value cannot be
	* known ahead of time. How the salt is used is algorithm-specific
	* (some might simply prefix it to the data block, others might need to
	* utilize a full-blown HMAC). On disk the salt is stored in a ZAP
	* object in the MOS (DMU_POOL_CHECKSUM_SALT).
	*
	* CONTEXT TEMPLATES
	*
	* Some hashing algorithms need to perform a substantial amount of
	* initialization work (e.g. salted checksums above may need to pre-hash
	* the salt) before being able to process data. Performing this
	* redundant work for each block would be wasteful, so we instead allow
	* a checksum algorithm to do the work once (the first time it's used)
	* and then keep this pre-initialized context as a template inside the
	* spa_t (spa_cksum_tmpls). If the zio_checksum_info_t contains
	* non-NULL ci_tmpl_init and ci_tmpl_free callbacks, they are used to
	* construct and destruct the pre-initialized checksum context. The
	* pre-initialized context is then reused during each checksum
	* invocation and passed to the checksum function.
	*/

	static void
	abd_checksum_off(abd_t *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) abd, (void) size, (void) ctx_template;
	ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
	}

	static void
	abd_fletcher_2_native(abd_t *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_init(zcp);
	(void) abd_iterate_func(abd, 0, size,
	fletcher_2_incremental_native, zcp);
	}

	static void
	abd_fletcher_2_byteswap(abd_t *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_init(zcp);
	(void) abd_iterate_func(abd, 0, size,
	fletcher_2_incremental_byteswap, zcp);
	}

	static inline void
	abd_fletcher_4_impl(abd_t abd, uint64_t size, zio_abd_checksum_data_t acdp)
	{
	fletcher_4_abd_ops.acf_init(acdp);
	abd_iterate_func(abd, 0, size, fletcher_4_abd_ops.acf_iter, acdp);
	fletcher_4_abd_ops.acf_fini(acdp);
	}

	void
	abd_fletcher_4_native(abd_t *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_4_ctx_t ctx;

	zio_abd_checksum_data_t acd = {
	.acd_byteorder = ZIO_CHECKSUM_NATIVE,
	.acd_zcp = zcp,
	.acd_ctx = &ctx
	};

	abd_fletcher_4_impl(abd, size, &acd);

	}

	void
	abd_fletcher_4_byteswap(abd_t *abd, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_4_ctx_t ctx;

	zio_abd_checksum_data_t acd = {
	.acd_byteorder = ZIO_CHECKSUM_BYTESWAP,
	.acd_zcp = zcp,
	.acd_ctx = &ctx
	};

	abd_fletcher_4_impl(abd, size, &acd);
	}

	-const zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = {
	+zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = {
	{{NULL, NULL}, NULL, NULL, 0, "inherit"},
	{{NULL, NULL}, NULL, NULL, 0, "on"},
	{{abd_checksum_off, abd_checksum_off},
	NULL, NULL, 0, "off"},
	{{abd_checksum_SHA256, abd_checksum_SHA256},
	NULL, NULL, ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_EMBEDDED,
	"label"},
	{{abd_checksum_SHA256, abd_checksum_SHA256},
	NULL, NULL, ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_EMBEDDED,
	"gang_header"},
	{{abd_fletcher_2_native, abd_fletcher_2_byteswap},
	NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog"},
	{{abd_fletcher_2_native, abd_fletcher_2_byteswap},
	NULL, NULL, 0, "fletcher2"},
	{{abd_fletcher_4_native, abd_fletcher_4_byteswap},
	NULL, NULL, ZCHECKSUM_FLAG_METADATA, "fletcher4"},
	{{abd_checksum_SHA256, abd_checksum_SHA256},
	NULL, NULL, ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_DEDUP \|
	ZCHECKSUM_FLAG_NOPWRITE, "sha256"},
	{{abd_fletcher_4_native, abd_fletcher_4_byteswap},
	NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog2"},
	{{abd_checksum_off, abd_checksum_off},
	NULL, NULL, 0, "noparity"},
	{{abd_checksum_SHA512_native, abd_checksum_SHA512_byteswap},
	NULL, NULL, ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_DEDUP \|
	ZCHECKSUM_FLAG_NOPWRITE, "sha512"},
	{{abd_checksum_skein_native, abd_checksum_skein_byteswap},
	abd_checksum_skein_tmpl_init, abd_checksum_skein_tmpl_free,
	ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_DEDUP \|
	ZCHECKSUM_FLAG_SALTED \| ZCHECKSUM_FLAG_NOPWRITE, "skein"},
	{{abd_checksum_edonr_native, abd_checksum_edonr_byteswap},
	abd_checksum_edonr_tmpl_init, abd_checksum_edonr_tmpl_free,
	ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_SALTED \|
	ZCHECKSUM_FLAG_NOPWRITE, "edonr"},
	{{abd_checksum_blake3_native, abd_checksum_blake3_byteswap},
	abd_checksum_blake3_tmpl_init, abd_checksum_blake3_tmpl_free,
	ZCHECKSUM_FLAG_METADATA \| ZCHECKSUM_FLAG_DEDUP \|
	ZCHECKSUM_FLAG_SALTED \| ZCHECKSUM_FLAG_NOPWRITE, "blake3"},
	};

	/*
	* The flag corresponding to the "verify" in dedup=[checksum,]verify
	* must be cleared first, so callers should use ZIO_CHECKSUM_MASK.
	*/
	spa_feature_t
	zio_checksum_to_feature(enum zio_checksum cksum)
	{
	VERIFY((cksum & ~ZIO_CHECKSUM_MASK) == 0);

	switch (cksum) {
	case ZIO_CHECKSUM_BLAKE3:
	return (SPA_FEATURE_BLAKE3);
	case ZIO_CHECKSUM_SHA512:
	return (SPA_FEATURE_SHA512);
	case ZIO_CHECKSUM_SKEIN:
	return (SPA_FEATURE_SKEIN);
	case ZIO_CHECKSUM_EDONR:
	return (SPA_FEATURE_EDONR);
	default:
	return (SPA_FEATURE_NONE);
	}
	}

	enum zio_checksum
	zio_checksum_select(enum zio_checksum child, enum zio_checksum parent)
	{
	ASSERT(child < ZIO_CHECKSUM_FUNCTIONS);
	ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS);
	ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);

	if (child == ZIO_CHECKSUM_INHERIT)
	return (parent);

	if (child == ZIO_CHECKSUM_ON)
	return (ZIO_CHECKSUM_ON_VALUE);

	return (child);
	}

	enum zio_checksum
	zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child,
	enum zio_checksum parent)
	{
	ASSERT((child & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
	ASSERT((parent & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
	ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);

	if (child == ZIO_CHECKSUM_INHERIT)
	return (parent);

	if (child == ZIO_CHECKSUM_ON)
	return (spa_dedup_checksum(spa));

	if (child == (ZIO_CHECKSUM_ON \| ZIO_CHECKSUM_VERIFY))
	return (spa_dedup_checksum(spa) \| ZIO_CHECKSUM_VERIFY);

	ASSERT((zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_flags &
	ZCHECKSUM_FLAG_DEDUP) \|\|
	(child & ZIO_CHECKSUM_VERIFY) \|\| child == ZIO_CHECKSUM_OFF);

	return (child);
	}

	/*
	* Set the external verifier for a gang block based on <vdev, offset, txg>,
	* a tuple which is guaranteed to be unique for the life of the pool.
	*/
	static void
	zio_checksum_gang_verifier(zio_cksum_t zcp, const blkptr_t bp)
	{
	const dva_t *dva = BP_IDENTITY(bp);
	uint64_t txg = BP_PHYSICAL_BIRTH(bp);

	ASSERT(BP_IS_GANG(bp));

	ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0);
	}

	/*
	* Set the external verifier for a label block based on its offset.
	* The vdev is implicit, and the txg is unknowable at pool open time --
	* hence the logic in vdev_uberblock_load() to find the most recent copy.
	*/
	static void
	zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset)
	{
	ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0);
	}

	/*
	* Calls the template init function of a checksum which supports context
	* templates and installs the template into the spa_t.
	*/
	static void
	zio_checksum_template_init(enum zio_checksum checksum, spa_t *spa)
	{
	zio_checksum_info_t *ci = &zio_checksum_table[checksum];

	if (ci->ci_tmpl_init == NULL)
	return;
	if (spa->spa_cksum_tmpls[checksum] != NULL)
	return;

	VERIFY(ci->ci_tmpl_free != NULL);
	mutex_enter(&spa->spa_cksum_tmpls_lock);
	if (spa->spa_cksum_tmpls[checksum] == NULL) {
	spa->spa_cksum_tmpls[checksum] =
	ci->ci_tmpl_init(&spa->spa_cksum_salt);
	VERIFY(spa->spa_cksum_tmpls[checksum] != NULL);
	}
	mutex_exit(&spa->spa_cksum_tmpls_lock);
	}

	/* convenience function to update a checksum to accommodate an encryption MAC */
	static void
	zio_checksum_handle_crypt(zio_cksum_t cksum, zio_cksum_t saved, boolean_t xor)
	{
	/*
	* Weak checksums do not have their entropy spread evenly
	* across the bits of the checksum. Therefore, when truncating
	* a weak checksum we XOR the first 2 words with the last 2 so
	* that we don't "lose" any entropy unnecessarily.
	*/
	if (xor) {
	cksum->zc_word[0] ^= cksum->zc_word[2];
	cksum->zc_word[1] ^= cksum->zc_word[3];
	}

	cksum->zc_word[2] = saved->zc_word[2];
	cksum->zc_word[3] = saved->zc_word[3];
	}

	/*
	* Generate the checksum.
	*/
	void
	zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
	abd_t *abd, uint64_t size)
	{
	static const uint64_t zec_magic = ZEC_MAGIC;
	blkptr_t *bp = zio->io_bp;
	uint64_t offset = zio->io_offset;
	zio_checksum_info_t *ci = &zio_checksum_table[checksum];
	zio_cksum_t cksum, saved;
	spa_t *spa = zio->io_spa;
	boolean_t insecure = (ci->ci_flags & ZCHECKSUM_FLAG_DEDUP) == 0;

	ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS);
	ASSERT(ci->ci_func[0] != NULL);

	zio_checksum_template_init(checksum, spa);

	if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
	zio_eck_t eck;
	size_t eck_offset;

	memset(&saved, 0, sizeof (zio_cksum_t));

	if (checksum == ZIO_CHECKSUM_ZILOG2) {
	zil_chain_t zilc;
	abd_copy_to_buf(&zilc, abd, sizeof (zil_chain_t));

	size = P2ROUNDUP_TYPED(zilc.zc_nused, ZIL_MIN_BLKSZ,
	uint64_t);
	eck = zilc.zc_eck;
	eck_offset = offsetof(zil_chain_t, zc_eck);
	} else {
	eck_offset = size - sizeof (zio_eck_t);
	abd_copy_to_buf_off(&eck, abd, eck_offset,
	sizeof (zio_eck_t));
	}

	if (checksum == ZIO_CHECKSUM_GANG_HEADER) {
	zio_checksum_gang_verifier(&eck.zec_cksum, bp);
	} else if (checksum == ZIO_CHECKSUM_LABEL) {
	zio_checksum_label_verifier(&eck.zec_cksum, offset);
	} else {
	saved = eck.zec_cksum;
	eck.zec_cksum = bp->blk_cksum;
	}

	abd_copy_from_buf_off(abd, &zec_magic,
	eck_offset + offsetof(zio_eck_t, zec_magic),
	sizeof (zec_magic));
	abd_copy_from_buf_off(abd, &eck.zec_cksum,
	eck_offset + offsetof(zio_eck_t, zec_cksum),
	sizeof (zio_cksum_t));

	ci->ci_func[0](abd, size, spa->spa_cksum_tmpls[checksum],
	&cksum);
	if (bp != NULL && BP_USES_CRYPT(bp) &&
	BP_GET_TYPE(bp) != DMU_OT_OBJSET)
	zio_checksum_handle_crypt(&cksum, &saved, insecure);

	abd_copy_from_buf_off(abd, &cksum,
	eck_offset + offsetof(zio_eck_t, zec_cksum),
	sizeof (zio_cksum_t));
	} else {
	saved = bp->blk_cksum;
	ci->ci_func[0](abd, size, spa->spa_cksum_tmpls[checksum],
	&cksum);
	if (BP_USES_CRYPT(bp) && BP_GET_TYPE(bp) != DMU_OT_OBJSET)
	zio_checksum_handle_crypt(&cksum, &saved, insecure);
	bp->blk_cksum = cksum;
	}
	}

	int
	zio_checksum_error_impl(spa_t spa, const blkptr_t bp,
	enum zio_checksum checksum, abd_t *abd, uint64_t size, uint64_t offset,
	zio_bad_cksum_t *info)
	{
	zio_checksum_info_t *ci = &zio_checksum_table[checksum];
	zio_cksum_t actual_cksum, expected_cksum;
	zio_eck_t eck;
	int byteswap;

	if (checksum >= ZIO_CHECKSUM_FUNCTIONS \|\| ci->ci_func[0] == NULL)
	return (SET_ERROR(EINVAL));

	zio_checksum_template_init(checksum, spa);

	if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
	zio_cksum_t verifier;
	size_t eck_offset;

	if (checksum == ZIO_CHECKSUM_ZILOG2) {
	zil_chain_t zilc;
	uint64_t nused;

	abd_copy_to_buf(&zilc, abd, sizeof (zil_chain_t));

	eck = zilc.zc_eck;
	eck_offset = offsetof(zil_chain_t, zc_eck) +
	offsetof(zio_eck_t, zec_cksum);

	if (eck.zec_magic == ZEC_MAGIC) {
	nused = zilc.zc_nused;
	} else if (eck.zec_magic == BSWAP_64(ZEC_MAGIC)) {
	nused = BSWAP_64(zilc.zc_nused);
	} else {
	return (SET_ERROR(ECKSUM));
	}

	if (nused > size) {
	return (SET_ERROR(ECKSUM));
	}

	size = P2ROUNDUP_TYPED(nused, ZIL_MIN_BLKSZ, uint64_t);
	} else {
	eck_offset = size - sizeof (zio_eck_t);
	abd_copy_to_buf_off(&eck, abd, eck_offset,
	sizeof (zio_eck_t));
	eck_offset += offsetof(zio_eck_t, zec_cksum);
	}

	if (checksum == ZIO_CHECKSUM_GANG_HEADER)
	zio_checksum_gang_verifier(&verifier, bp);
	else if (checksum == ZIO_CHECKSUM_LABEL)
	zio_checksum_label_verifier(&verifier, offset);
	else
	verifier = bp->blk_cksum;

	byteswap = (eck.zec_magic == BSWAP_64(ZEC_MAGIC));

	if (byteswap)
	byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));

	expected_cksum = eck.zec_cksum;

	abd_copy_from_buf_off(abd, &verifier, eck_offset,
	sizeof (zio_cksum_t));

	ci->ci_func[byteswap](abd, size,
	spa->spa_cksum_tmpls[checksum], &actual_cksum);

	abd_copy_from_buf_off(abd, &expected_cksum, eck_offset,
	sizeof (zio_cksum_t));

	if (byteswap) {
	byteswap_uint64_array(&expected_cksum,
	sizeof (zio_cksum_t));
	}
	} else {
	byteswap = BP_SHOULD_BYTESWAP(bp);
	expected_cksum = bp->blk_cksum;
	ci->ci_func[byteswap](abd, size,
	spa->spa_cksum_tmpls[checksum], &actual_cksum);
	}

	/*
	* MAC checksums are a special case since half of this checksum will
	* actually be the encryption MAC. This will be verified by the
	* decryption process, so we just check the truncated checksum now.
	* Objset blocks use embedded MACs so we don't truncate the checksum
	* for them.
	*/
	if (bp != NULL && BP_USES_CRYPT(bp) &&
	BP_GET_TYPE(bp) != DMU_OT_OBJSET) {
	if (!(ci->ci_flags & ZCHECKSUM_FLAG_DEDUP)) {
	actual_cksum.zc_word[0] ^= actual_cksum.zc_word[2];
	actual_cksum.zc_word[1] ^= actual_cksum.zc_word[3];
	}

	actual_cksum.zc_word[2] = 0;
	actual_cksum.zc_word[3] = 0;
	expected_cksum.zc_word[2] = 0;
	expected_cksum.zc_word[3] = 0;
	}

	if (info != NULL) {
	info->zbc_expected = expected_cksum;
	info->zbc_actual = actual_cksum;
	info->zbc_checksum_name = ci->ci_name;
	info->zbc_byteswapped = byteswap;
	info->zbc_injected = 0;
	info->zbc_has_cksum = 1;
	}

	if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
	return (SET_ERROR(ECKSUM));

	return (0);
	}

	int
	zio_checksum_error(zio_t zio, zio_bad_cksum_t info)
	{
	blkptr_t *bp = zio->io_bp;
	uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum :
	(BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp)));
	int error;
	uint64_t size = (bp == NULL ? zio->io_size :
	(BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp)));
	uint64_t offset = zio->io_offset;
	abd_t *data = zio->io_abd;
	spa_t *spa = zio->io_spa;

	error = zio_checksum_error_impl(spa, bp, checksum, data, size,
	offset, info);

	if (zio_injection_enabled && error == 0 && zio->io_error == 0) {
	error = zio_handle_fault_injection(zio, ECKSUM);
	if (error != 0)
	info->zbc_injected = 1;
	}

	return (error);
	}

	/*
	* Called by a spa_t that's about to be deallocated. This steps through
	* all of the checksum context templates and deallocates any that were
	* initialized using the algorithm-specific template init function.
	*/
	void
	zio_checksum_templates_free(spa_t *spa)
	{
	for (enum zio_checksum checksum = 0;
	checksum < ZIO_CHECKSUM_FUNCTIONS; checksum++) {
	if (spa->spa_cksum_tmpls[checksum] != NULL) {
	zio_checksum_info_t *ci = &zio_checksum_table[checksum];

	VERIFY(ci->ci_tmpl_free != NULL);
	ci->ci_tmpl_free(spa->spa_cksum_tmpls[checksum]);
	spa->spa_cksum_tmpls[checksum] = NULL;
	}
	}
	}
	diff --git a/sys/contrib/openzfs/module/zfs/zio_compress.c b/sys/contrib/openzfs/module/zfs/zio_compress.c
	index 6527e33ca37d..4c9cbc962093 100644
	--- a/sys/contrib/openzfs/module/zfs/zio_compress.c
	+++ b/sys/contrib/openzfs/module/zfs/zio_compress.c
	@@ -1,222 +1,222 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/
	/*
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	*/

	/*
	* Copyright (c) 2013, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/zfeature.h>
	#include <sys/zio.h>
	#include <sys/zio_compress.h>
	#include <sys/zstd/zstd.h>

	/*
	* If nonzero, every 1/X decompression attempts will fail, simulating
	* an undetected memory error.
	*/
	unsigned long zio_decompress_fail_fraction = 0;

	/*
	* Compression vectors.
	*/
	-const zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS] = {
	+zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS] = {
	{"inherit", 0, NULL, NULL, NULL},
	{"on", 0, NULL, NULL, NULL},
	{"uncompressed", 0, NULL, NULL, NULL},
	{"lzjb", 0, lzjb_compress, lzjb_decompress, NULL},
	{"empty", 0, NULL, NULL, NULL},
	{"gzip-1", 1, gzip_compress, gzip_decompress, NULL},
	{"gzip-2", 2, gzip_compress, gzip_decompress, NULL},
	{"gzip-3", 3, gzip_compress, gzip_decompress, NULL},
	{"gzip-4", 4, gzip_compress, gzip_decompress, NULL},
	{"gzip-5", 5, gzip_compress, gzip_decompress, NULL},
	{"gzip-6", 6, gzip_compress, gzip_decompress, NULL},
	{"gzip-7", 7, gzip_compress, gzip_decompress, NULL},
	{"gzip-8", 8, gzip_compress, gzip_decompress, NULL},
	{"gzip-9", 9, gzip_compress, gzip_decompress, NULL},
	{"zle", 64, zle_compress, zle_decompress, NULL},
	{"lz4", 0, lz4_compress_zfs, lz4_decompress_zfs, NULL},
	{"zstd", ZIO_ZSTD_LEVEL_DEFAULT, zfs_zstd_compress_wrap,
	zfs_zstd_decompress, zfs_zstd_decompress_level},
	};

	uint8_t
	zio_complevel_select(spa_t *spa, enum zio_compress compress, uint8_t child,
	uint8_t parent)
	{
	(void) spa;
	uint8_t result;

	if (!ZIO_COMPRESS_HASLEVEL(compress))
	return (0);

	result = child;
	if (result == ZIO_COMPLEVEL_INHERIT)
	result = parent;

	return (result);
	}

	enum zio_compress
	zio_compress_select(spa_t *spa, enum zio_compress child,
	enum zio_compress parent)
	{
	enum zio_compress result;

	ASSERT(child < ZIO_COMPRESS_FUNCTIONS);
	ASSERT(parent < ZIO_COMPRESS_FUNCTIONS);
	ASSERT(parent != ZIO_COMPRESS_INHERIT);

	result = child;
	if (result == ZIO_COMPRESS_INHERIT)
	result = parent;

	if (result == ZIO_COMPRESS_ON) {
	if (spa_feature_is_active(spa, SPA_FEATURE_LZ4_COMPRESS))
	result = ZIO_COMPRESS_LZ4_ON_VALUE;
	else
	result = ZIO_COMPRESS_LEGACY_ON_VALUE;
	}

	return (result);
	}

	static int
	zio_compress_zeroed_cb(void data, size_t len, void private)
	{
	(void) private;

	uint64_t end = (uint64_t )((char *)data + len);
	for (uint64_t word = (uint64_t )data; word < end; word++)
	if (*word != 0)
	return (1);

	return (0);
	}

	size_t
	zio_compress_data(enum zio_compress c, abd_t src, void dst, size_t s_len,
	uint8_t level)
	{
	size_t c_len, d_len;
	uint8_t complevel;
	zio_compress_info_t *ci = &zio_compress_table[c];

	ASSERT((uint_t)c < ZIO_COMPRESS_FUNCTIONS);
	ASSERT((uint_t)c == ZIO_COMPRESS_EMPTY \|\| ci->ci_compress != NULL);

	/*
	* If the data is all zeroes, we don't even need to allocate
	* a block for it. We indicate this by returning zero size.
	*/
	if (abd_iterate_func(src, 0, s_len, zio_compress_zeroed_cb, NULL) == 0)
	return (0);

	if (c == ZIO_COMPRESS_EMPTY)
	return (s_len);

	/* Compress at least 12.5% */
	d_len = s_len - (s_len >> 3);

	complevel = ci->ci_level;

	if (c == ZIO_COMPRESS_ZSTD) {
	/* If we don't know the level, we can't compress it */
	if (level == ZIO_COMPLEVEL_INHERIT)
	return (s_len);

	if (level == ZIO_COMPLEVEL_DEFAULT)
	complevel = ZIO_ZSTD_LEVEL_DEFAULT;
	else
	complevel = level;

	ASSERT3U(complevel, !=, ZIO_COMPLEVEL_INHERIT);
	}

	/* No compression algorithms can read from ABDs directly */
	void *tmp = abd_borrow_buf_copy(src, s_len);
	c_len = ci->ci_compress(tmp, dst, s_len, d_len, complevel);
	abd_return_buf(src, tmp, s_len);

	if (c_len > d_len)
	return (s_len);

	ASSERT3U(c_len, <=, d_len);
	return (c_len);
	}

	int
	zio_decompress_data_buf(enum zio_compress c, void src, void dst,
	size_t s_len, size_t d_len, uint8_t *level)
	{
	zio_compress_info_t *ci = &zio_compress_table[c];
	if ((uint_t)c >= ZIO_COMPRESS_FUNCTIONS \|\| ci->ci_decompress == NULL)
	return (SET_ERROR(EINVAL));

	if (ci->ci_decompress_level != NULL && level != NULL)
	return (ci->ci_decompress_level(src, dst, s_len, d_len, level));

	return (ci->ci_decompress(src, dst, s_len, d_len, ci->ci_level));
	}

	int
	zio_decompress_data(enum zio_compress c, abd_t src, void dst,
	size_t s_len, size_t d_len, uint8_t *level)
	{
	void *tmp = abd_borrow_buf_copy(src, s_len);
	int ret = zio_decompress_data_buf(c, tmp, dst, s_len, d_len, level);
	abd_return_buf(src, tmp, s_len);

	/*
	* Decompression shouldn't fail, because we've already verified
	* the checksum. However, for extra protection (e.g. against bitflips
	* in non-ECC RAM), we handle this error (and test it).
	*/
	if (zio_decompress_fail_fraction != 0 &&
	random_in_range(zio_decompress_fail_fraction) == 0)
	ret = SET_ERROR(EINVAL);

	return (ret);
	}

	int
	zio_compress_to_feature(enum zio_compress comp)
	{
	switch (comp) {
	case ZIO_COMPRESS_ZSTD:
	return (SPA_FEATURE_ZSTD_COMPRESS);
	default:
	break;
	}
	return (SPA_FEATURE_NONE);
	}
	diff --git a/sys/contrib/openzfs/module/zfs/zio_inject.c b/sys/contrib/openzfs/module/zfs/zio_inject.c
	index 4f7cb8430d3e..3598351c499d 100644
	--- a/sys/contrib/openzfs/module/zfs/zio_inject.c
	+++ b/sys/contrib/openzfs/module/zfs/zio_inject.c
	@@ -1,972 +1,972 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	*/

	/*
	* ZFS fault injection
	*
	* To handle fault injection, we keep track of a series of zinject_record_t
	* structures which describe which logical block(s) should be injected with a
	* fault. These are kept in a global list. Each record corresponds to a given
	* spa_t and maintains a special hold on the spa_t so that it cannot be deleted
	* or exported while the injection record exists.
	*
	* Device level injection is done using the 'zi_guid' field. If this is set, it
	* means that the error is destined for a particular device, not a piece of
	* data.
	*
	* This is a rather poor data structure and algorithm, but we don't expect more
	* than a few faults at any one time, so it should be sufficient for our needs.
	*/

	#include <sys/arc.h>
	#include <sys/zio.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/vdev_impl.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dataset.h>
	#include <sys/fs/zfs.h>

	uint32_t zio_injection_enabled = 0;

	/*
	* Data describing each zinject handler registered on the system, and
	* contains the list node linking the handler in the global zinject
	* handler list.
	*/
	typedef struct inject_handler {
	int zi_id;
	spa_t *zi_spa;
	zinject_record_t zi_record;
	uint64_t *zi_lanes;
	int zi_next_lane;
	list_node_t zi_link;
	} inject_handler_t;

	/*
	* List of all zinject handlers registered on the system, protected by
	* the inject_lock defined below.
	*/
	static list_t inject_handlers;

	/*
	* This protects insertion into, and traversal of, the inject handler
	* list defined above; as well as the inject_delay_count. Any time a
	* handler is inserted or removed from the list, this lock should be
	* taken as a RW_WRITER; and any time traversal is done over the list
	* (without modification to it) this lock should be taken as a RW_READER.
	*/
	static krwlock_t inject_lock;

	/*
	* This holds the number of zinject delay handlers that have been
	* registered on the system. It is protected by the inject_lock defined
	* above. Thus modifications to this count must be a RW_WRITER of the
	* inject_lock, and reads of this count must be (at least) a RW_READER
	* of the lock.
	*/
	static int inject_delay_count = 0;

	/*
	* This lock is used only in zio_handle_io_delay(), refer to the comment
	* in that function for more details.
	*/
	static kmutex_t inject_delay_mtx;

	/*
	* Used to assign unique identifying numbers to each new zinject handler.
	*/
	static int inject_next_id = 1;

	/*
	* Test if the requested frequency was triggered
	*/
	static boolean_t
	freq_triggered(uint32_t frequency)
	{
	/*
	* zero implies always (100%)
	*/
	if (frequency == 0)
	return (B_TRUE);

	/*
	* Note: we still handle legacy (unscaled) frequency values
	*/
	uint32_t maximum = (frequency <= 100) ? 100 : ZI_PERCENTAGE_MAX;

	return (random_in_range(maximum) < frequency);
	}

	/*
	* Returns true if the given record matches the I/O in progress.
	*/
	static boolean_t
	zio_match_handler(const zbookmark_phys_t *zb, uint64_t type, int dva,
	zinject_record_t *record, int error)
	{
	/*
	* Check for a match against the MOS, which is based on type
	*/
	if (zb->zb_objset == DMU_META_OBJSET &&
	record->zi_objset == DMU_META_OBJSET &&
	record->zi_object == DMU_META_DNODE_OBJECT) {
	if (record->zi_type == DMU_OT_NONE \|\|
	type == record->zi_type)
	return (freq_triggered(record->zi_freq));
	else
	return (B_FALSE);
	}

	/*
	* Check for an exact match.
	*/
	if (zb->zb_objset == record->zi_objset &&
	zb->zb_object == record->zi_object &&
	zb->zb_level == record->zi_level &&
	zb->zb_blkid >= record->zi_start &&
	zb->zb_blkid <= record->zi_end &&
	(record->zi_dvas == 0 \|\|
	(dva != ZI_NO_DVA && (record->zi_dvas & (1ULL << dva)))) &&
	error == record->zi_error) {
	return (freq_triggered(record->zi_freq));
	}

	return (B_FALSE);
	}

	/*
	* Panic the system when a config change happens in the function
	* specified by tag.
	*/
	void
	zio_handle_panic_injection(spa_t spa, const char tag, uint64_t type)
	{
	inject_handler_t *handler;

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {

	if (spa != handler->zi_spa)
	continue;

	if (handler->zi_record.zi_type == type &&
	strcmp(tag, handler->zi_record.zi_func) == 0)
	panic("Panic requested in function %s\n", tag);
	}

	rw_exit(&inject_lock);
	}

	/*
	* Inject a decryption failure. Decryption failures can occur in
	* both the ARC and the ZIO layers.
	*/
	int
	zio_handle_decrypt_injection(spa_t spa, const zbookmark_phys_t zb,
	uint64_t type, int error)
	{
	int ret = 0;
	inject_handler_t *handler;

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {

	if (spa != handler->zi_spa \|\|
	handler->zi_record.zi_cmd != ZINJECT_DECRYPT_FAULT)
	continue;

	if (zio_match_handler(zb, type, ZI_NO_DVA,
	&handler->zi_record, error)) {
	ret = error;
	break;
	}
	}

	rw_exit(&inject_lock);
	return (ret);
	}

	/*
	* If this is a physical I/O for a vdev child determine which DVA it is
	* for. We iterate backwards through the DVAs matching on the offset so
	* that we end up with ZI_NO_DVA (-1) if we don't find a match.
	*/
	static int
	zio_match_dva(zio_t *zio)
	{
	int i = ZI_NO_DVA;

	if (zio->io_bp != NULL && zio->io_vd != NULL &&
	zio->io_child_type == ZIO_CHILD_VDEV) {
	for (i = BP_GET_NDVAS(zio->io_bp) - 1; i >= 0; i--) {
	dva_t *dva = &zio->io_bp->blk_dva[i];
	uint64_t off = DVA_GET_OFFSET(dva);
	vdev_t *vd = vdev_lookup_top(zio->io_spa,
	DVA_GET_VDEV(dva));

	/* Compensate for vdev label added to leaves */
	if (zio->io_vd->vdev_ops->vdev_op_leaf)
	off += VDEV_LABEL_START_SIZE;

	if (zio->io_vd == vd && zio->io_offset == off)
	break;
	}
	}

	return (i);
	}


	/*
	* Determine if the I/O in question should return failure. Returns the errno
	* to be returned to the caller.
	*/
	int
	zio_handle_fault_injection(zio_t *zio, int error)
	{
	int ret = 0;
	inject_handler_t *handler;

	/*
	* Ignore I/O not associated with any logical data.
	*/
	if (zio->io_logical == NULL)
	return (0);

	/*
	* Currently, we only support fault injection on reads.
	*/
	if (zio->io_type != ZIO_TYPE_READ)
	return (0);

	/*
	* A rebuild I/O has no checksum to verify.
	*/
	if (zio->io_priority == ZIO_PRIORITY_REBUILD && error == ECKSUM)
	return (0);

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {
	if (zio->io_spa != handler->zi_spa \|\|
	handler->zi_record.zi_cmd != ZINJECT_DATA_FAULT)
	continue;

	/* If this handler matches, return the specified error */
	if (zio_match_handler(&zio->io_logical->io_bookmark,
	zio->io_bp ? BP_GET_TYPE(zio->io_bp) : DMU_OT_NONE,
	zio_match_dva(zio), &handler->zi_record, error)) {
	ret = error;
	break;
	}
	}

	rw_exit(&inject_lock);

	return (ret);
	}

	/*
	* Determine if the zio is part of a label update and has an injection
	* handler associated with that portion of the label. Currently, we
	* allow error injection in either the nvlist or the uberblock region of
	* of the vdev label.
	*/
	int
	zio_handle_label_injection(zio_t *zio, int error)
	{
	inject_handler_t *handler;
	vdev_t *vd = zio->io_vd;
	uint64_t offset = zio->io_offset;
	int label;
	int ret = 0;

	if (offset >= VDEV_LABEL_START_SIZE &&
	offset < vd->vdev_psize - VDEV_LABEL_END_SIZE)
	return (0);

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {
	uint64_t start = handler->zi_record.zi_start;
	uint64_t end = handler->zi_record.zi_end;

	if (handler->zi_record.zi_cmd != ZINJECT_LABEL_FAULT)
	continue;

	/*
	* The injection region is the relative offsets within a
	* vdev label. We must determine the label which is being
	* updated and adjust our region accordingly.
	*/
	label = vdev_label_number(vd->vdev_psize, offset);
	start = vdev_label_offset(vd->vdev_psize, label, start);
	end = vdev_label_offset(vd->vdev_psize, label, end);

	if (zio->io_vd->vdev_guid == handler->zi_record.zi_guid &&
	(offset >= start && offset <= end)) {
	ret = error;
	break;
	}
	}
	rw_exit(&inject_lock);
	return (ret);
	}

	static int
	zio_inject_bitflip_cb(void data, size_t len, void private)
	{
	zio_t *zio = private;
	uint8_t *buffer = data;
	uint_t byte = random_in_range(len);

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);

	/* flip a single random bit in an abd data buffer */
	buffer[byte] ^= 1 << random_in_range(8);

	return (1); /* stop after first flip */
	}

	static int
	zio_handle_device_injection_impl(vdev_t vd, zio_t zio, int err1, int err2)
	{
	inject_handler_t *handler;
	int ret = 0;

	/*
	* We skip over faults in the labels unless it's during
	* device open (i.e. zio == NULL).
	*/
	if (zio != NULL) {
	uint64_t offset = zio->io_offset;

	if (offset < VDEV_LABEL_START_SIZE \|\|
	offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE)
	return (0);
	}

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {

	if (handler->zi_record.zi_cmd != ZINJECT_DEVICE_FAULT)
	continue;

	if (vd->vdev_guid == handler->zi_record.zi_guid) {
	if (handler->zi_record.zi_failfast &&
	(zio == NULL \|\| (zio->io_flags &
	(ZIO_FLAG_IO_RETRY \| ZIO_FLAG_TRYHARD)))) {
	continue;
	}

	/* Handle type specific I/O failures */
	if (zio != NULL &&
	handler->zi_record.zi_iotype != ZIO_TYPES &&
	handler->zi_record.zi_iotype != zio->io_type)
	continue;

	if (handler->zi_record.zi_error == err1 \|\|
	handler->zi_record.zi_error == err2) {
	/*
	* limit error injection if requested
	*/
	if (!freq_triggered(handler->zi_record.zi_freq))
	continue;

	/*
	* For a failed open, pretend like the device
	* has gone away.
	*/
	if (err1 == ENXIO)
	vd->vdev_stat.vs_aux =
	VDEV_AUX_OPEN_FAILED;

	/*
	* Treat these errors as if they had been
	* retried so that all the appropriate stats
	* and FMA events are generated.
	*/
	if (!handler->zi_record.zi_failfast &&
	zio != NULL)
	zio->io_flags \|= ZIO_FLAG_IO_RETRY;

	/*
	* EILSEQ means flip a bit after a read
	*/
	if (handler->zi_record.zi_error == EILSEQ) {
	if (zio == NULL)
	break;

	/* locate buffer data and flip a bit */
	(void) abd_iterate_func(zio->io_abd, 0,
	zio->io_size, zio_inject_bitflip_cb,
	zio);
	break;
	}

	ret = handler->zi_record.zi_error;
	break;
	}
	if (handler->zi_record.zi_error == ENXIO) {
	ret = SET_ERROR(EIO);
	break;
	}
	}
	}

	rw_exit(&inject_lock);

	return (ret);
	}

	int
	zio_handle_device_injection(vdev_t vd, zio_t zio, int error)
	{
	return (zio_handle_device_injection_impl(vd, zio, error, INT_MAX));
	}

	int
	zio_handle_device_injections(vdev_t vd, zio_t zio, int err1, int err2)
	{
	return (zio_handle_device_injection_impl(vd, zio, err1, err2));
	}

	/*
	* Simulate hardware that ignores cache flushes. For requested number
	* of seconds nix the actual writing to disk.
	*/
	void
	zio_handle_ignored_writes(zio_t *zio)
	{
	inject_handler_t *handler;

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {

	/* Ignore errors not destined for this pool */
	if (zio->io_spa != handler->zi_spa \|\|
	handler->zi_record.zi_cmd != ZINJECT_IGNORED_WRITES)
	continue;

	/*
	* Positive duration implies # of seconds, negative
	* a number of txgs
	*/
	if (handler->zi_record.zi_timer == 0) {
	if (handler->zi_record.zi_duration > 0)
	handler->zi_record.zi_timer = ddi_get_lbolt64();
	else
	handler->zi_record.zi_timer = zio->io_txg;
	}

	/* Have a "problem" writing 60% of the time */
	if (random_in_range(100) < 60)
	zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	break;
	}

	rw_exit(&inject_lock);
	}

	void
	spa_handle_ignored_writes(spa_t *spa)
	{
	inject_handler_t *handler;

	if (zio_injection_enabled == 0)
	return;

	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler)) {

	if (spa != handler->zi_spa \|\|
	handler->zi_record.zi_cmd != ZINJECT_IGNORED_WRITES)
	continue;

	if (handler->zi_record.zi_duration > 0) {
	VERIFY(handler->zi_record.zi_timer == 0 \|\|
	ddi_time_after64(
	(int64_t)handler->zi_record.zi_timer +
	handler->zi_record.zi_duration * hz,
	ddi_get_lbolt64()));
	} else {
	/* duration is negative so the subtraction here adds */
	VERIFY(handler->zi_record.zi_timer == 0 \|\|
	handler->zi_record.zi_timer -
	handler->zi_record.zi_duration >=
	spa_syncing_txg(spa));
	}
	}

	rw_exit(&inject_lock);
	}

	hrtime_t
	zio_handle_io_delay(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	inject_handler_t *min_handler = NULL;
	hrtime_t min_target = 0;

	rw_enter(&inject_lock, RW_READER);

	/*
	* inject_delay_count is a subset of zio_injection_enabled that
	* is only incremented for delay handlers. These checks are
	* mainly added to remind the reader why we're not explicitly
	* checking zio_injection_enabled like the other functions.
	*/
	IMPLY(inject_delay_count > 0, zio_injection_enabled > 0);
	IMPLY(zio_injection_enabled == 0, inject_delay_count == 0);

	/*
	* If there aren't any inject delay handlers registered, then we
	* can short circuit and simply return 0 here. A value of zero
	* informs zio_delay_interrupt() that this request should not be
	* delayed. This short circuit keeps us from acquiring the
	* inject_delay_mutex unnecessarily.
	*/
	if (inject_delay_count == 0) {
	rw_exit(&inject_lock);
	return (0);
	}

	/*
	* Each inject handler has a number of "lanes" associated with
	* it. Each lane is able to handle requests independently of one
	* another, and at a latency defined by the inject handler
	* record's zi_timer field. Thus if a handler in configured with
	* a single lane with a 10ms latency, it will delay requests
	* such that only a single request is completed every 10ms. So,
	* if more than one request is attempted per each 10ms interval,
	* the average latency of the requests will be greater than
	* 10ms; but if only a single request is submitted each 10ms
	* interval the average latency will be 10ms.
	*
	* We need to acquire this mutex to prevent multiple concurrent
	* threads being assigned to the same lane of a given inject
	* handler. The mutex allows us to perform the following two
	* operations atomically:
	*
	* 1. determine the minimum handler and minimum target
	* value of all the possible handlers
	* 2. update that minimum handler's lane array
	*
	* Without atomicity, two (or more) threads could pick the same
	* lane in step (1), and then conflict with each other in step
	* (2). This could allow a single lane handler to process
	* multiple requests simultaneously, which shouldn't be possible.
	*/
	mutex_enter(&inject_delay_mtx);

	for (inject_handler_t *handler = list_head(&inject_handlers);
	handler != NULL; handler = list_next(&inject_handlers, handler)) {
	if (handler->zi_record.zi_cmd != ZINJECT_DELAY_IO)
	continue;

	if (!freq_triggered(handler->zi_record.zi_freq))
	continue;

	if (vd->vdev_guid != handler->zi_record.zi_guid)
	continue;

	/*
	* Defensive; should never happen as the array allocation
	* occurs prior to inserting this handler on the list.
	*/
	ASSERT3P(handler->zi_lanes, !=, NULL);

	/*
	* This should never happen, the zinject command should
	* prevent a user from setting an IO delay with zero lanes.
	*/
	ASSERT3U(handler->zi_record.zi_nlanes, !=, 0);

	ASSERT3U(handler->zi_record.zi_nlanes, >,
	handler->zi_next_lane);

	/*
	* We want to issue this IO to the lane that will become
	* idle the soonest, so we compare the soonest this
	* specific handler can complete the IO with all other
	* handlers, to find the lowest value of all possible
	* lanes. We then use this lane to submit the request.
	*
	* Since each handler has a constant value for its
	* delay, we can just use the "next" lane for that
	* handler; as it will always be the lane with the
	* lowest value for that particular handler (i.e. the
	* lane that will become idle the soonest). This saves a
	* scan of each handler's lanes array.
	*
	* There's two cases to consider when determining when
	* this specific IO request should complete. If this
	* lane is idle, we want to "submit" the request now so
	* it will complete after zi_timer milliseconds. Thus,
	* we set the target to now + zi_timer.
	*
	* If the lane is busy, we want this request to complete
	* zi_timer milliseconds after the lane becomes idle.
	* Since the 'zi_lanes' array holds the time at which
	* each lane will become idle, we use that value to
	* determine when this request should complete.
	*/
	hrtime_t idle = handler->zi_record.zi_timer + gethrtime();
	hrtime_t busy = handler->zi_record.zi_timer +
	handler->zi_lanes[handler->zi_next_lane];
	hrtime_t target = MAX(idle, busy);

	if (min_handler == NULL) {
	min_handler = handler;
	min_target = target;
	continue;
	}

	ASSERT3P(min_handler, !=, NULL);
	ASSERT3U(min_target, !=, 0);

	/*
	* We don't yet increment the "next lane" variable since
	* we still might find a lower value lane in another
	* handler during any remaining iterations. Once we're
	* sure we've selected the absolute minimum, we'll claim
	* the lane and increment the handler's "next lane"
	* field below.
	*/

	if (target < min_target) {
	min_handler = handler;
	min_target = target;
	}
	}

	/*
	* 'min_handler' will be NULL if no IO delays are registered for
	* this vdev, otherwise it will point to the handler containing
	* the lane that will become idle the soonest.
	*/
	if (min_handler != NULL) {
	ASSERT3U(min_target, !=, 0);
	min_handler->zi_lanes[min_handler->zi_next_lane] = min_target;

	/*
	* If we've used all possible lanes for this handler,
	* loop back and start using the first lane again;
	* otherwise, just increment the lane index.
	*/
	min_handler->zi_next_lane = (min_handler->zi_next_lane + 1) %
	min_handler->zi_record.zi_nlanes;
	}

	mutex_exit(&inject_delay_mtx);
	rw_exit(&inject_lock);

	return (min_target);
	}

	static int
	zio_calculate_range(const char pool, zinject_record_t record)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	objset_t *os = NULL;
	dnode_t *dn = NULL;
	int error;

	/*
	* Obtain the dnode for object using pool, objset, and object
	*/
	error = dsl_pool_hold(pool, FTAG, &dp);
	if (error)
	return (error);

	error = dsl_dataset_hold_obj(dp, record->zi_objset, FTAG, &ds);
	dsl_pool_rele(dp, FTAG);
	if (error)
	return (error);

	error = dmu_objset_from_ds(ds, &os);
	dsl_dataset_rele(ds, FTAG);
	if (error)
	return (error);

	error = dnode_hold(os, record->zi_object, FTAG, &dn);
	if (error)
	return (error);

	/*
	* Translate the range into block IDs
	*/
	if (record->zi_start != 0 \|\| record->zi_end != -1ULL) {
	record->zi_start >>= dn->dn_datablkshift;
	record->zi_end >>= dn->dn_datablkshift;
	}
	if (record->zi_level > 0) {
	if (record->zi_level >= dn->dn_nlevels) {
	dnode_rele(dn, FTAG);
	return (SET_ERROR(EDOM));
	}

	if (record->zi_start != 0 \|\| record->zi_end != 0) {
	int shift = dn->dn_indblkshift - SPA_BLKPTRSHIFT;

	for (int level = record->zi_level; level > 0; level--) {
	record->zi_start >>= shift;
	record->zi_end >>= shift;
	}
	}
	}

	dnode_rele(dn, FTAG);
	return (0);
	}

	/*
	* Create a new handler for the given record. We add it to the list, adding
	* a reference to the spa_t in the process. We increment zio_injection_enabled,
	* which is the switch to trigger all fault injection.
	*/
	int
	zio_inject_fault(char name, int flags, int id, zinject_record_t *record)
	{
	inject_handler_t *handler;
	int error;
	spa_t *spa;

	/*
	* If this is pool-wide metadata, make sure we unload the corresponding
	* spa_t, so that the next attempt to load it will trigger the fault.
	* We call spa_reset() to unload the pool appropriately.
	*/
	if (flags & ZINJECT_UNLOAD_SPA)
	if ((error = spa_reset(name)) != 0)
	return (error);

	if (record->zi_cmd == ZINJECT_DELAY_IO) {
	/*
	* A value of zero for the number of lanes or for the
	* delay time doesn't make sense.
	*/
	if (record->zi_timer == 0 \|\| record->zi_nlanes == 0)
	return (SET_ERROR(EINVAL));

	/*
	* The number of lanes is directly mapped to the size of
	* an array used by the handler. Thus, to ensure the
	* user doesn't trigger an allocation that's "too large"
	* we cap the number of lanes here.
	*/
	if (record->zi_nlanes >= UINT16_MAX)
	return (SET_ERROR(EINVAL));
	}

	/*
	* If the supplied range was in bytes -- calculate the actual blkid
	*/
	if (flags & ZINJECT_CALC_RANGE) {
	error = zio_calculate_range(name, record);
	if (error != 0)
	return (error);
	}

	if (!(flags & ZINJECT_NULL)) {
	/*
	* spa_inject_ref() will add an injection reference, which will
	* prevent the pool from being removed from the namespace while
	* still allowing it to be unloaded.
	*/
	if ((spa = spa_inject_addref(name)) == NULL)
	return (SET_ERROR(ENOENT));

	handler = kmem_alloc(sizeof (inject_handler_t), KM_SLEEP);

	handler->zi_spa = spa;
	handler->zi_record = *record;

	if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) {
	handler->zi_lanes = kmem_zalloc(
	sizeof (handler->zi_lanes)
	handler->zi_record.zi_nlanes, KM_SLEEP);
	handler->zi_next_lane = 0;
	} else {
	handler->zi_lanes = NULL;
	handler->zi_next_lane = 0;
	}

	rw_enter(&inject_lock, RW_WRITER);

	/*
	* We can't move this increment into the conditional
	* above because we need to hold the RW_WRITER lock of
	* inject_lock, and we don't want to hold that while
	* allocating the handler's zi_lanes array.
	*/
	if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) {
	ASSERT3S(inject_delay_count, >=, 0);
	inject_delay_count++;
	ASSERT3S(inject_delay_count, >, 0);
	}

	*id = handler->zi_id = inject_next_id++;
	list_insert_tail(&inject_handlers, handler);
	atomic_inc_32(&zio_injection_enabled);

	rw_exit(&inject_lock);
	}

	/*
	* Flush the ARC, so that any attempts to read this data will end up
	* going to the ZIO layer. Note that this is a little overkill, but
	* we don't have the necessary ARC interfaces to do anything else, and
	* fault injection isn't a performance critical path.
	*/
	if (flags & ZINJECT_FLUSH_ARC)
	/*
	* We must use FALSE to ensure arc_flush returns, since
	* we're not preventing concurrent ARC insertions.
	*/
	arc_flush(NULL, FALSE);

	return (0);
	}

	/*
	* Returns the next record with an ID greater than that supplied to the
	* function. Used to iterate over all handlers in the system.
	*/
	int
	zio_inject_list_next(int id, char name, size_t buflen,
	zinject_record_t *record)
	{
	inject_handler_t *handler;
	int ret;

	mutex_enter(&spa_namespace_lock);
	rw_enter(&inject_lock, RW_READER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler))
	if (handler->zi_id > *id)
	break;

	if (handler) {
	*record = handler->zi_record;
	*id = handler->zi_id;
	- (void) strncpy(name, spa_name(handler->zi_spa), buflen);
	+ (void) strlcpy(name, spa_name(handler->zi_spa), buflen);
	ret = 0;
	} else {
	ret = SET_ERROR(ENOENT);
	}

	rw_exit(&inject_lock);
	mutex_exit(&spa_namespace_lock);

	return (ret);
	}

	/*
	* Clear the fault handler with the given identifier, or return ENOENT if none
	* exists.
	*/
	int
	zio_clear_fault(int id)
	{
	inject_handler_t *handler;

	rw_enter(&inject_lock, RW_WRITER);

	for (handler = list_head(&inject_handlers); handler != NULL;
	handler = list_next(&inject_handlers, handler))
	if (handler->zi_id == id)
	break;

	if (handler == NULL) {
	rw_exit(&inject_lock);
	return (SET_ERROR(ENOENT));
	}

	if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) {
	ASSERT3S(inject_delay_count, >, 0);
	inject_delay_count--;
	ASSERT3S(inject_delay_count, >=, 0);
	}

	list_remove(&inject_handlers, handler);
	rw_exit(&inject_lock);

	if (handler->zi_record.zi_cmd == ZINJECT_DELAY_IO) {
	ASSERT3P(handler->zi_lanes, !=, NULL);
	kmem_free(handler->zi_lanes, sizeof (handler->zi_lanes)
	handler->zi_record.zi_nlanes);
	} else {
	ASSERT3P(handler->zi_lanes, ==, NULL);
	}

	spa_inject_delref(handler->zi_spa);
	kmem_free(handler, sizeof (inject_handler_t));
	atomic_dec_32(&zio_injection_enabled);

	return (0);
	}

	void
	zio_inject_init(void)
	{
	rw_init(&inject_lock, NULL, RW_DEFAULT, NULL);
	mutex_init(&inject_delay_mtx, NULL, MUTEX_DEFAULT, NULL);
	list_create(&inject_handlers, sizeof (inject_handler_t),
	offsetof(inject_handler_t, zi_link));
	}

	void
	zio_inject_fini(void)
	{
	list_destroy(&inject_handlers);
	mutex_destroy(&inject_delay_mtx);
	rw_destroy(&inject_lock);
	}

	#if defined(_KERNEL)
	EXPORT_SYMBOL(zio_injection_enabled);
	EXPORT_SYMBOL(zio_inject_fault);
	EXPORT_SYMBOL(zio_inject_list_next);
	EXPORT_SYMBOL(zio_clear_fault);
	EXPORT_SYMBOL(zio_handle_fault_injection);
	EXPORT_SYMBOL(zio_handle_device_injection);
	EXPORT_SYMBOL(zio_handle_label_injection);
	#endif
	diff --git a/sys/contrib/openzfs/module/zfs/zvol.c b/sys/contrib/openzfs/module/zfs/zvol.c
	index 3a1870f568b7..2e2860ff0212 100644
	--- a/sys/contrib/openzfs/module/zfs/zvol.c
	+++ b/sys/contrib/openzfs/module/zfs/zvol.c
	@@ -1,1738 +1,1738 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
	* LLNL-CODE-403049.
	*
	* ZFS volume emulation driver.
	*
	* Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes.
	* Volumes are accessed through the symbolic links named:
	*
	* /dev/<pool_name>/<dataset_name>
	*
	* Volumes are persistent through reboot and module load. No user command
	* needs to be run before opening and using a device.
	*
	* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	*/

	/*
	* Note on locking of zvol state structures.
	*
	* These structures are used to maintain internal state used to emulate block
	* devices on top of zvols. In particular, management of device minor number
	* operations - create, remove, rename, and set_snapdev - involves access to
	* these structures. The zvol_state_lock is primarily used to protect the
	* zvol_state_list. The zv->zv_state_lock is used to protect the contents
	* of the zvol_state_t structures, as well as to make sure that when the
	* time comes to remove the structure from the list, it is not in use, and
	* therefore, it can be taken off zvol_state_list and freed.
	*
	* The zv_suspend_lock was introduced to allow for suspending I/O to a zvol,
	* e.g. for the duration of receive and rollback operations. This lock can be
	* held for significant periods of time. Given that it is undesirable to hold
	* mutexes for long periods of time, the following lock ordering applies:
	* - take zvol_state_lock if necessary, to protect zvol_state_list
	* - take zv_suspend_lock if necessary, by the code path in question
	* - take zv_state_lock to protect zvol_state_t
	*
	* The minor operations are issued to spa->spa_zvol_taskq queues, that are
	* single-threaded (to preserve order of minor operations), and are executed
	* through the zvol_task_cb that dispatches the specific operations. Therefore,
	* these operations are serialized per pool. Consequently, we can be certain
	* that for a given zvol, there is only one operation at a time in progress.
	* That is why one can be sure that first, zvol_state_t for a given zvol is
	* allocated and placed on zvol_state_list, and then other minor operations
	* for this zvol are going to proceed in the order of issue.
	*
	*/

	#include <sys/dataset_kstats.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dir.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/zil_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/zio.h>
	#include <sys/zfs_rlock.h>
	#include <sys/spa_impl.h>
	#include <sys/zvol.h>
	#include <sys/zvol_impl.h>

	unsigned int zvol_inhibit_dev = 0;
	unsigned int zvol_volmode = ZFS_VOLMODE_GEOM;

	struct hlist_head *zvol_htable;
	static list_t zvol_state_list;
	krwlock_t zvol_state_lock;

	typedef enum {
	ZVOL_ASYNC_REMOVE_MINORS,
	ZVOL_ASYNC_RENAME_MINORS,
	ZVOL_ASYNC_SET_SNAPDEV,
	ZVOL_ASYNC_SET_VOLMODE,
	ZVOL_ASYNC_MAX
	} zvol_async_op_t;

	typedef struct {
	zvol_async_op_t op;
	char name1[MAXNAMELEN];
	char name2[MAXNAMELEN];
	uint64_t value;
	} zvol_task_t;

	uint64_t
	zvol_name_hash(const char *name)
	{
	int i;
	uint64_t crc = -1ULL;
	const uint8_t p = (const uint8_t )name;
	ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
	for (i = 0; i < MAXNAMELEN - 1 && *p; i++, p++) {
	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (*p)) & 0xFF];
	}
	return (crc);
	}

	/*
	* Find a zvol_state_t given the name and hash generated by zvol_name_hash.
	* If found, return with zv_suspend_lock and zv_state_lock taken, otherwise,
	* return (NULL) without the taking locks. The zv_suspend_lock is always taken
	* before zv_state_lock. The mode argument indicates the mode (including none)
	* for zv_suspend_lock to be taken.
	*/
	zvol_state_t *
	zvol_find_by_name_hash(const char *name, uint64_t hash, int mode)
	{
	zvol_state_t *zv;
	struct hlist_node *p = NULL;

	rw_enter(&zvol_state_lock, RW_READER);
	hlist_for_each(p, ZVOL_HT_HEAD(hash)) {
	zv = hlist_entry(p, zvol_state_t, zv_hlink);
	mutex_enter(&zv->zv_state_lock);
	if (zv->zv_hash == hash &&
	strncmp(zv->zv_name, name, MAXNAMELEN) == 0) {
	/*
	* this is the right zvol, take the locks in the
	* right order
	*/
	if (mode != RW_NONE &&
	!rw_tryenter(&zv->zv_suspend_lock, mode)) {
	mutex_exit(&zv->zv_state_lock);
	rw_enter(&zv->zv_suspend_lock, mode);
	mutex_enter(&zv->zv_state_lock);
	/*
	* zvol cannot be renamed as we continue
	* to hold zvol_state_lock
	*/
	ASSERT(zv->zv_hash == hash &&
	strncmp(zv->zv_name, name, MAXNAMELEN)
	== 0);
	}
	rw_exit(&zvol_state_lock);
	return (zv);
	}
	mutex_exit(&zv->zv_state_lock);
	}
	rw_exit(&zvol_state_lock);

	return (NULL);
	}

	/*
	* Find a zvol_state_t given the name.
	* If found, return with zv_suspend_lock and zv_state_lock taken, otherwise,
	* return (NULL) without the taking locks. The zv_suspend_lock is always taken
	* before zv_state_lock. The mode argument indicates the mode (including none)
	* for zv_suspend_lock to be taken.
	*/
	static zvol_state_t *
	zvol_find_by_name(const char *name, int mode)
	{
	return (zvol_find_by_name_hash(name, zvol_name_hash(name), mode));
	}

	/*
	* ZFS_IOC_CREATE callback handles dmu zvol and zap object creation.
	*/
	void
	zvol_create_cb(objset_t os, void arg, cred_t cr, dmu_tx_t tx)
	{
	zfs_creat_t *zct = arg;
	nvlist_t *nvprops = zct->zct_props;
	int error;
	uint64_t volblocksize, volsize;

	VERIFY(nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0);
	if (nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0)
	volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);

	/*
	* These properties must be removed from the list so the generic
	* property setting step won't apply to them.
	*/
	VERIFY(nvlist_remove_all(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0);
	(void) nvlist_remove_all(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE));

	error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize,
	DMU_OT_NONE, 0, tx);
	ASSERT(error == 0);

	error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
	DMU_OT_NONE, 0, tx);
	ASSERT(error == 0);

	error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx);
	ASSERT(error == 0);
	}

	/*
	* ZFS_IOC_OBJSET_STATS entry point.
	*/
	int
	zvol_get_stats(objset_t os, nvlist_t nv)
	{
	int error;
	dmu_object_info_t *doi;
	uint64_t val;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val);
	if (error)
	return (SET_ERROR(error));

	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val);
	doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
	error = dmu_object_info(os, ZVOL_OBJ, doi);

	if (error == 0) {
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE,
	doi->doi_data_block_size);
	}

	kmem_free(doi, sizeof (dmu_object_info_t));

	return (SET_ERROR(error));
	}

	/*
	* Sanity check volume size.
	*/
	int
	zvol_check_volsize(uint64_t volsize, uint64_t blocksize)
	{
	if (volsize == 0)
	return (SET_ERROR(EINVAL));

	if (volsize % blocksize != 0)
	return (SET_ERROR(EINVAL));

	#ifdef _ILP32
	if (volsize - 1 > SPEC_MAXOFFSET_T)
	return (SET_ERROR(EOVERFLOW));
	#endif
	return (0);
	}

	/*
	* Ensure the zap is flushed then inform the VFS of the capacity change.
	*/
	static int
	zvol_update_volsize(uint64_t volsize, objset_t *os)
	{
	dmu_tx_t *tx;
	int error;
	uint64_t txg;

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	return (SET_ERROR(error));
	}
	txg = dmu_tx_get_txg(tx);

	error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
	&volsize, tx);
	dmu_tx_commit(tx);

	txg_wait_synced(dmu_objset_pool(os), txg);

	if (error == 0)
	error = dmu_free_long_range(os,
	ZVOL_OBJ, volsize, DMU_OBJECT_END);

	return (error);
	}

	/*
	* Set ZFS_PROP_VOLSIZE set entry point. Note that modifying the volume
	* size will result in a udev "change" event being generated.
	*/
	int
	zvol_set_volsize(const char *name, uint64_t volsize)
	{
	objset_t *os = NULL;
	uint64_t readonly;
	int error;
	boolean_t owned = B_FALSE;

	error = dsl_prop_get_integer(name,
	zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL);
	if (error != 0)
	return (SET_ERROR(error));
	if (readonly)
	return (SET_ERROR(EROFS));

	zvol_state_t *zv = zvol_find_by_name(name, RW_READER);

	ASSERT(zv == NULL \|\| (MUTEX_HELD(&zv->zv_state_lock) &&
	RW_READ_HELD(&zv->zv_suspend_lock)));

	if (zv == NULL \|\| zv->zv_objset == NULL) {
	if (zv != NULL)
	rw_exit(&zv->zv_suspend_lock);
	if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE, B_TRUE,
	FTAG, &os)) != 0) {
	if (zv != NULL)
	mutex_exit(&zv->zv_state_lock);
	return (SET_ERROR(error));
	}
	owned = B_TRUE;
	if (zv != NULL)
	zv->zv_objset = os;
	} else {
	os = zv->zv_objset;
	}

	dmu_object_info_t doi = kmem_alloc(sizeof (doi), KM_SLEEP);

	if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) \|\|
	(error = zvol_check_volsize(volsize, doi->doi_data_block_size)))
	goto out;

	error = zvol_update_volsize(volsize, os);
	if (error == 0 && zv != NULL) {
	zv->zv_volsize = volsize;
	zv->zv_changed = 1;
	}
	out:
	kmem_free(doi, sizeof (dmu_object_info_t));

	if (owned) {
	dmu_objset_disown(os, B_TRUE, FTAG);
	if (zv != NULL)
	zv->zv_objset = NULL;
	} else {
	rw_exit(&zv->zv_suspend_lock);
	}

	if (zv != NULL)
	mutex_exit(&zv->zv_state_lock);

	if (error == 0 && zv != NULL)
	zvol_os_update_volsize(zv, volsize);

	return (SET_ERROR(error));
	}

	/*
	* Sanity check volume block size.
	*/
	int
	zvol_check_volblocksize(const char *name, uint64_t volblocksize)
	{
	/* Record sizes above 128k need the feature to be enabled */
	if (volblocksize > SPA_OLD_MAXBLOCKSIZE) {
	spa_t *spa;
	int error;

	if ((error = spa_open(name, &spa, FTAG)) != 0)
	return (error);

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	/*
	* We don't allow setting the property above 1MB,
	* unless the tunable has been changed.
	*/
	if (volblocksize > zfs_max_recordsize)
	return (SET_ERROR(EDOM));

	spa_close(spa, FTAG);
	}

	if (volblocksize < SPA_MINBLOCKSIZE \|\|
	volblocksize > SPA_MAXBLOCKSIZE \|\|
	!ISP2(volblocksize))
	return (SET_ERROR(EDOM));

	return (0);
	}

	/*
	* Replay a TX_TRUNCATE ZIL transaction if asked. TX_TRUNCATE is how we
	* implement DKIOCFREE/free-long-range.
	*/
	static int
	zvol_replay_truncate(void arg1, void arg2, boolean_t byteswap)
	{
	zvol_state_t *zv = arg1;
	lr_truncate_t *lr = arg2;
	uint64_t offset, length;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	offset = lr->lr_offset;
	length = lr->lr_length;

	dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);
	dmu_tx_mark_netfree(tx);
	int error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	} else {
	- zil_replaying(zv->zv_zilog, tx);
	+ (void) zil_replaying(zv->zv_zilog, tx);
	dmu_tx_commit(tx);
	error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset,
	length);
	}

	return (error);
	}

	/*
	* Replay a TX_WRITE ZIL transaction that didn't get committed
	* after a system failure
	*/
	static int
	zvol_replay_write(void arg1, void arg2, boolean_t byteswap)
	{
	zvol_state_t *zv = arg1;
	lr_write_t *lr = arg2;
	objset_t *os = zv->zv_objset;
	char data = (char )(lr + 1); /* data follows lr_write_t */
	uint64_t offset, length;
	dmu_tx_t *tx;
	int error;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	offset = lr->lr_offset;
	length = lr->lr_length;

	/* If it's a dmu_sync() block, write the whole block */
	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
	uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
	if (length < blocksize) {
	offset -= offset % blocksize;
	length = blocksize;
	}
	}

	tx = dmu_tx_create(os);
	dmu_tx_hold_write(tx, ZVOL_OBJ, offset, length);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	} else {
	dmu_write(os, ZVOL_OBJ, offset, length, data, tx);
	- zil_replaying(zv->zv_zilog, tx);
	+ (void) zil_replaying(zv->zv_zilog, tx);
	dmu_tx_commit(tx);
	}

	return (error);
	}

	static int
	zvol_replay_err(void arg1, void arg2, boolean_t byteswap)
	{
	(void) arg1, (void) arg2, (void) byteswap;
	return (SET_ERROR(ENOTSUP));
	}

	/*
	* Callback vectors for replaying records.
	* Only TX_WRITE and TX_TRUNCATE are needed for zvol.
	*/
	zil_replay_func_t *const zvol_replay_vector[TX_MAX_TYPE] = {
	zvol_replay_err, /* no such transaction type */
	zvol_replay_err, /* TX_CREATE */
	zvol_replay_err, /* TX_MKDIR */
	zvol_replay_err, /* TX_MKXATTR */
	zvol_replay_err, /* TX_SYMLINK */
	zvol_replay_err, /* TX_REMOVE */
	zvol_replay_err, /* TX_RMDIR */
	zvol_replay_err, /* TX_LINK */
	zvol_replay_err, /* TX_RENAME */
	zvol_replay_write, /* TX_WRITE */
	zvol_replay_truncate, /* TX_TRUNCATE */
	zvol_replay_err, /* TX_SETATTR */
	zvol_replay_err, /* TX_ACL */
	zvol_replay_err, /* TX_CREATE_ATTR */
	zvol_replay_err, /* TX_CREATE_ACL_ATTR */
	zvol_replay_err, /* TX_MKDIR_ACL */
	zvol_replay_err, /* TX_MKDIR_ATTR */
	zvol_replay_err, /* TX_MKDIR_ACL_ATTR */
	zvol_replay_err, /* TX_WRITE2 */
	zvol_replay_err, /* TX_SETSAXATTR */
	};

	/*
	* zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions.
	*
	* We store data in the log buffers if it's small enough.
	* Otherwise we will later flush the data out via dmu_sync().
	*/
	static const ssize_t zvol_immediate_write_sz = 32768;

	void
	zvol_log_write(zvol_state_t zv, dmu_tx_t tx, uint64_t offset,
	uint64_t size, int sync)
	{
	uint32_t blocksize = zv->zv_volblocksize;
	zilog_t *zilog = zv->zv_zilog;
	itx_wr_state_t write_state;
	uint64_t sz = size;

	if (zil_replaying(zilog, tx))
	return;

	if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
	write_state = WR_INDIRECT;
	else if (!spa_has_slogs(zilog->zl_spa) &&
	size >= blocksize && blocksize > zvol_immediate_write_sz)
	write_state = WR_INDIRECT;
	else if (sync)
	write_state = WR_COPIED;
	else
	write_state = WR_NEED_COPY;

	while (size) {
	itx_t *itx;
	lr_write_t *lr;
	itx_wr_state_t wr_state = write_state;
	ssize_t len = size;

	if (wr_state == WR_COPIED && size > zil_max_copied_data(zilog))
	wr_state = WR_NEED_COPY;
	else if (wr_state == WR_INDIRECT)
	len = MIN(blocksize - P2PHASE(offset, blocksize), size);

	itx = zil_itx_create(TX_WRITE, sizeof (*lr) +
	(wr_state == WR_COPIED ? len : 0));
	lr = (lr_write_t *)&itx->itx_lr;
	if (wr_state == WR_COPIED && dmu_read_by_dnode(zv->zv_dn,
	offset, len, lr+1, DMU_READ_NO_PREFETCH) != 0) {
	zil_itx_destroy(itx);
	itx = zil_itx_create(TX_WRITE, sizeof (*lr));
	lr = (lr_write_t *)&itx->itx_lr;
	wr_state = WR_NEED_COPY;
	}

	itx->itx_wr_state = wr_state;
	lr->lr_foid = ZVOL_OBJ;
	lr->lr_offset = offset;
	lr->lr_length = len;
	lr->lr_blkoff = 0;
	BP_ZERO(&lr->lr_blkptr);

	itx->itx_private = zv;
	itx->itx_sync = sync;

	(void) zil_itx_assign(zilog, itx, tx);

	offset += len;
	size -= len;
	}

	if (write_state == WR_COPIED \|\| write_state == WR_NEED_COPY) {
	dsl_pool_wrlog_count(zilog->zl_dmu_pool, sz, tx->tx_txg);
	}
	}

	/*
	* Log a DKIOCFREE/free-long-range to the ZIL with TX_TRUNCATE.
	*/
	void
	zvol_log_truncate(zvol_state_t zv, dmu_tx_t tx, uint64_t off, uint64_t len,
	boolean_t sync)
	{
	itx_t *itx;
	lr_truncate_t *lr;
	zilog_t *zilog = zv->zv_zilog;

	if (zil_replaying(zilog, tx))
	return;

	itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
	lr = (lr_truncate_t *)&itx->itx_lr;
	lr->lr_foid = ZVOL_OBJ;
	lr->lr_offset = off;
	lr->lr_length = len;

	itx->itx_sync = sync;
	zil_itx_assign(zilog, itx, tx);
	}


	static void
	zvol_get_done(zgd_t *zgd, int error)
	{
	(void) error;
	if (zgd->zgd_db)
	dmu_buf_rele(zgd->zgd_db, zgd);

	zfs_rangelock_exit(zgd->zgd_lr);

	kmem_free(zgd, sizeof (zgd_t));
	}

	/*
	* Get data to generate a TX_WRITE intent log record.
	*/
	int
	zvol_get_data(void arg, uint64_t arg2, lr_write_t lr, char *buf,
	struct lwb lwb, zio_t zio)
	{
	zvol_state_t *zv = arg;
	uint64_t offset = lr->lr_offset;
	uint64_t size = lr->lr_length;
	dmu_buf_t *db;
	zgd_t *zgd;
	int error;

	ASSERT3P(lwb, !=, NULL);
	ASSERT3P(zio, !=, NULL);
	ASSERT3U(size, !=, 0);

	zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
	zgd->zgd_lwb = lwb;

	/*
	* Write records come in two flavors: immediate and indirect.
	* For small writes it's cheaper to store the data with the
	* log record (immediate); for large writes it's cheaper to
	* sync the data and get a pointer to it (indirect) so that
	* we don't have to write the data twice.
	*/
	if (buf != NULL) { /* immediate write */
	zgd->zgd_lr = zfs_rangelock_enter(&zv->zv_rangelock, offset,
	size, RL_READER);
	error = dmu_read_by_dnode(zv->zv_dn, offset, size, buf,
	DMU_READ_NO_PREFETCH);
	} else { /* indirect write */
	/*
	* Have to lock the whole block to ensure when it's written out
	* and its checksum is being calculated that no one can change
	* the data. Contrarily to zfs_get_data we need not re-check
	* blocksize after we get the lock because it cannot be changed.
	*/
	size = zv->zv_volblocksize;
	offset = P2ALIGN_TYPED(offset, size, uint64_t);
	zgd->zgd_lr = zfs_rangelock_enter(&zv->zv_rangelock, offset,
	size, RL_READER);
	error = dmu_buf_hold_by_dnode(zv->zv_dn, offset, zgd, &db,
	DMU_READ_NO_PREFETCH);
	if (error == 0) {
	blkptr_t *bp = &lr->lr_blkptr;

	zgd->zgd_db = db;
	zgd->zgd_bp = bp;

	ASSERT(db != NULL);
	ASSERT(db->db_offset == offset);
	ASSERT(db->db_size == size);

	error = dmu_sync(zio, lr->lr_common.lrc_txg,
	zvol_get_done, zgd);

	if (error == 0)
	return (0);
	}
	}

	zvol_get_done(zgd, error);

	return (SET_ERROR(error));
	}

	/*
	* The zvol_state_t's are inserted into zvol_state_list and zvol_htable.
	*/

	void
	zvol_insert(zvol_state_t *zv)
	{
	ASSERT(RW_WRITE_HELD(&zvol_state_lock));
	list_insert_head(&zvol_state_list, zv);
	hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash));
	}

	/*
	* Simply remove the zvol from to list of zvols.
	*/
	static void
	zvol_remove(zvol_state_t *zv)
	{
	ASSERT(RW_WRITE_HELD(&zvol_state_lock));
	list_remove(&zvol_state_list, zv);
	hlist_del(&zv->zv_hlink);
	}

	/*
	* Setup zv after we just own the zv->objset
	*/
	static int
	zvol_setup_zv(zvol_state_t *zv)
	{
	uint64_t volsize;
	int error;
	uint64_t ro;
	objset_t *os = zv->zv_objset;

	ASSERT(MUTEX_HELD(&zv->zv_state_lock));
	ASSERT(RW_LOCK_HELD(&zv->zv_suspend_lock));

	zv->zv_zilog = NULL;
	zv->zv_flags &= ~ZVOL_WRITTEN_TO;

	error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
	if (error)
	return (SET_ERROR(error));

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
	return (SET_ERROR(error));

	error = dnode_hold(os, ZVOL_OBJ, zv, &zv->zv_dn);
	if (error)
	return (SET_ERROR(error));

	zvol_os_set_capacity(zv, volsize >> 9);
	zv->zv_volsize = volsize;

	if (ro \|\| dmu_objset_is_snapshot(os) \|\|
	!spa_writeable(dmu_objset_spa(os))) {
	zvol_os_set_disk_ro(zv, 1);
	zv->zv_flags \|= ZVOL_RDONLY;
	} else {
	zvol_os_set_disk_ro(zv, 0);
	zv->zv_flags &= ~ZVOL_RDONLY;
	}
	return (0);
	}

	/*
	* Shutdown every zv_objset related stuff except zv_objset itself.
	* The is the reverse of zvol_setup_zv.
	*/
	static void
	zvol_shutdown_zv(zvol_state_t *zv)
	{
	ASSERT(MUTEX_HELD(&zv->zv_state_lock) &&
	RW_LOCK_HELD(&zv->zv_suspend_lock));

	if (zv->zv_flags & ZVOL_WRITTEN_TO) {
	ASSERT(zv->zv_zilog != NULL);
	zil_close(zv->zv_zilog);
	}

	zv->zv_zilog = NULL;

	dnode_rele(zv->zv_dn, zv);
	zv->zv_dn = NULL;

	/*
	* Evict cached data. We must write out any dirty data before
	* disowning the dataset.
	*/
	if (zv->zv_flags & ZVOL_WRITTEN_TO)
	txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
	(void) dmu_objset_evict_dbufs(zv->zv_objset);
	}

	/*
	* return the proper tag for rollback and recv
	*/
	void *
	zvol_tag(zvol_state_t *zv)
	{
	ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));
	return (zv->zv_open_count > 0 ? zv : NULL);
	}

	/*
	* Suspend the zvol for recv and rollback.
	*/
	zvol_state_t *
	zvol_suspend(const char *name)
	{
	zvol_state_t *zv;

	zv = zvol_find_by_name(name, RW_WRITER);

	if (zv == NULL)
	return (NULL);

	/* block all I/O, release in zvol_resume. */
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));
	ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));

	atomic_inc(&zv->zv_suspend_ref);

	if (zv->zv_open_count > 0)
	zvol_shutdown_zv(zv);

	/*
	* do not hold zv_state_lock across suspend/resume to
	* avoid locking up zvol lookups
	*/
	mutex_exit(&zv->zv_state_lock);

	/* zv_suspend_lock is released in zvol_resume() */
	return (zv);
	}

	int
	zvol_resume(zvol_state_t *zv)
	{
	int error = 0;

	ASSERT(RW_WRITE_HELD(&zv->zv_suspend_lock));

	mutex_enter(&zv->zv_state_lock);

	if (zv->zv_open_count > 0) {
	VERIFY0(dmu_objset_hold(zv->zv_name, zv, &zv->zv_objset));
	VERIFY3P(zv->zv_objset->os_dsl_dataset->ds_owner, ==, zv);
	VERIFY(dsl_dataset_long_held(zv->zv_objset->os_dsl_dataset));
	dmu_objset_rele(zv->zv_objset, zv);

	error = zvol_setup_zv(zv);
	}

	mutex_exit(&zv->zv_state_lock);

	rw_exit(&zv->zv_suspend_lock);
	/*
	* We need this because we don't hold zvol_state_lock while releasing
	* zv_suspend_lock. zvol_remove_minors_impl thus cannot check
	* zv_suspend_lock to determine it is safe to free because rwlock is
	* not inherent atomic.
	*/
	atomic_dec(&zv->zv_suspend_ref);

	return (SET_ERROR(error));
	}

	int
	zvol_first_open(zvol_state_t *zv, boolean_t readonly)
	{
	objset_t *os;
	int error;

	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));
	ASSERT(mutex_owned(&spa_namespace_lock));

	boolean_t ro = (readonly \|\| (strchr(zv->zv_name, '@') != NULL));
	error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, ro, B_TRUE, zv, &os);
	if (error)
	return (SET_ERROR(error));

	zv->zv_objset = os;

	error = zvol_setup_zv(zv);
	if (error) {
	dmu_objset_disown(os, 1, zv);
	zv->zv_objset = NULL;
	}

	return (error);
	}

	void
	zvol_last_close(zvol_state_t *zv)
	{
	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	zvol_shutdown_zv(zv);

	dmu_objset_disown(zv->zv_objset, 1, zv);
	zv->zv_objset = NULL;
	}

	typedef struct minors_job {
	list_t *list;
	list_node_t link;
	/* input */
	char *name;
	/* output */
	int error;
	} minors_job_t;

	/*
	* Prefetch zvol dnodes for the minors_job
	*/
	static void
	zvol_prefetch_minors_impl(void *arg)
	{
	minors_job_t *job = arg;
	char *dsname = job->name;
	objset_t *os = NULL;

	job->error = dmu_objset_own(dsname, DMU_OST_ZVOL, B_TRUE, B_TRUE,
	FTAG, &os);
	if (job->error == 0) {
	dmu_prefetch(os, ZVOL_OBJ, 0, 0, 0, ZIO_PRIORITY_SYNC_READ);
	dmu_objset_disown(os, B_TRUE, FTAG);
	}
	}

	/*
	* Mask errors to continue dmu_objset_find() traversal
	*/
	static int
	zvol_create_snap_minor_cb(const char dsname, void arg)
	{
	minors_job_t *j = arg;
	list_t *minors_list = j->list;
	const char *name = j->name;

	ASSERT0(MUTEX_HELD(&spa_namespace_lock));

	/* skip the designated dataset */
	if (name && strcmp(dsname, name) == 0)
	return (0);

	/* at this point, the dsname should name a snapshot */
	if (strchr(dsname, '@') == 0) {
	dprintf("zvol_create_snap_minor_cb(): "
	"%s is not a snapshot name\n", dsname);
	} else {
	minors_job_t *job;
	char *n = kmem_strdup(dsname);
	if (n == NULL)
	return (0);

	job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
	job->name = n;
	job->list = minors_list;
	job->error = 0;
	list_insert_tail(minors_list, job);
	/* don't care if dispatch fails, because job->error is 0 */
	taskq_dispatch(system_taskq, zvol_prefetch_minors_impl, job,
	TQ_SLEEP);
	}

	return (0);
	}

	/*
	* If spa_keystore_load_wkey() is called for an encrypted zvol,
	* we need to look for any clones also using the key. This function
	* is "best effort" - so we just skip over it if there are failures.
	*/
	static void
	zvol_add_clones(const char dsname, list_t minors_list)
	{
	/* Also check if it has clones */
	dsl_dir_t *dd = NULL;
	dsl_pool_t *dp = NULL;

	if (dsl_pool_hold(dsname, FTAG, &dp) != 0)
	return;

	if (!spa_feature_is_enabled(dp->dp_spa,
	SPA_FEATURE_ENCRYPTION))
	goto out;

	if (dsl_dir_hold(dp, dsname, FTAG, &dd, NULL) != 0)
	goto out;

	if (dsl_dir_phys(dd)->dd_clones == 0)
	goto out;

	zap_cursor_t *zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP);
	zap_attribute_t *za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
	objset_t *mos = dd->dd_pool->dp_meta_objset;

	for (zap_cursor_init(zc, mos, dsl_dir_phys(dd)->dd_clones);
	zap_cursor_retrieve(zc, za) == 0;
	zap_cursor_advance(zc)) {
	dsl_dataset_t *clone;
	minors_job_t *job;

	if (dsl_dataset_hold_obj(dd->dd_pool,
	za->za_first_integer, FTAG, &clone) == 0) {

	char name[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(clone, name);

	char *n = kmem_strdup(name);
	job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
	job->name = n;
	job->list = minors_list;
	job->error = 0;
	list_insert_tail(minors_list, job);

	dsl_dataset_rele(clone, FTAG);
	}
	}
	zap_cursor_fini(zc);
	kmem_free(za, sizeof (zap_attribute_t));
	kmem_free(zc, sizeof (zap_cursor_t));

	out:
	if (dd != NULL)
	dsl_dir_rele(dd, FTAG);
	if (dp != NULL)
	dsl_pool_rele(dp, FTAG);
	}

	/*
	* Mask errors to continue dmu_objset_find() traversal
	*/
	static int
	zvol_create_minors_cb(const char dsname, void arg)
	{
	uint64_t snapdev;
	int error;
	list_t *minors_list = arg;

	ASSERT0(MUTEX_HELD(&spa_namespace_lock));

	error = dsl_prop_get_integer(dsname, "snapdev", &snapdev, NULL);
	if (error)
	return (0);

	/*
	* Given the name and the 'snapdev' property, create device minor nodes
	* with the linkages to zvols/snapshots as needed.
	* If the name represents a zvol, create a minor node for the zvol, then
	* check if its snapshots are 'visible', and if so, iterate over the
	* snapshots and create device minor nodes for those.
	*/
	if (strchr(dsname, '@') == 0) {
	minors_job_t *job;
	char *n = kmem_strdup(dsname);
	if (n == NULL)
	return (0);

	job = kmem_alloc(sizeof (minors_job_t), KM_SLEEP);
	job->name = n;
	job->list = minors_list;
	job->error = 0;
	list_insert_tail(minors_list, job);
	/* don't care if dispatch fails, because job->error is 0 */
	taskq_dispatch(system_taskq, zvol_prefetch_minors_impl, job,
	TQ_SLEEP);

	zvol_add_clones(dsname, minors_list);

	if (snapdev == ZFS_SNAPDEV_VISIBLE) {
	/*
	* traverse snapshots only, do not traverse children,
	* and skip the 'dsname'
	*/
	error = dmu_objset_find(dsname,
	zvol_create_snap_minor_cb, (void *)job,
	DS_FIND_SNAPSHOTS);
	}
	} else {
	dprintf("zvol_create_minors_cb(): %s is not a zvol name\n",
	dsname);
	}

	return (0);
	}

	/*
	* Create minors for the specified dataset, including children and snapshots.
	* Pay attention to the 'snapdev' property and iterate over the snapshots
	* only if they are 'visible'. This approach allows one to assure that the
	* snapshot metadata is read from disk only if it is needed.
	*
	* The name can represent a dataset to be recursively scanned for zvols and
	* their snapshots, or a single zvol snapshot. If the name represents a
	* dataset, the scan is performed in two nested stages:
	* - scan the dataset for zvols, and
	* - for each zvol, create a minor node, then check if the zvol's snapshots
	* are 'visible', and only then iterate over the snapshots if needed
	*
	* If the name represents a snapshot, a check is performed if the snapshot is
	* 'visible' (which also verifies that the parent is a zvol), and if so,
	* a minor node for that snapshot is created.
	*/
	void
	zvol_create_minors_recursive(const char *name)
	{
	list_t minors_list;
	minors_job_t *job;

	if (zvol_inhibit_dev)
	return;

	/*
	* This is the list for prefetch jobs. Whenever we found a match
	* during dmu_objset_find, we insert a minors_job to the list and do
	* taskq_dispatch to parallel prefetch zvol dnodes. Note we don't need
	* any lock because all list operation is done on the current thread.
	*
	* We will use this list to do zvol_os_create_minor after prefetch
	* so we don't have to traverse using dmu_objset_find again.
	*/
	list_create(&minors_list, sizeof (minors_job_t),
	offsetof(minors_job_t, link));


	if (strchr(name, '@') != NULL) {
	uint64_t snapdev;

	int error = dsl_prop_get_integer(name, "snapdev",
	&snapdev, NULL);

	if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
	(void) zvol_os_create_minor(name);
	} else {
	fstrans_cookie_t cookie = spl_fstrans_mark();
	(void) dmu_objset_find(name, zvol_create_minors_cb,
	&minors_list, DS_FIND_CHILDREN);
	spl_fstrans_unmark(cookie);
	}

	taskq_wait_outstanding(system_taskq, 0);

	/*
	* Prefetch is completed, we can do zvol_os_create_minor
	* sequentially.
	*/
	while ((job = list_head(&minors_list)) != NULL) {
	list_remove(&minors_list, job);
	if (!job->error)
	(void) zvol_os_create_minor(job->name);
	kmem_strfree(job->name);
	kmem_free(job, sizeof (minors_job_t));
	}

	list_destroy(&minors_list);
	}

	void
	zvol_create_minor(const char *name)
	{
	/*
	* Note: the dsl_pool_config_lock must not be held.
	* Minor node creation needs to obtain the zvol_state_lock.
	* zvol_open() obtains the zvol_state_lock and then the dsl pool
	* config lock. Therefore, we can't have the config lock now if
	* we are going to wait for the zvol_state_lock, because it
	* would be a lock order inversion which could lead to deadlock.
	*/

	if (zvol_inhibit_dev)
	return;

	if (strchr(name, '@') != NULL) {
	uint64_t snapdev;

	int error = dsl_prop_get_integer(name,
	"snapdev", &snapdev, NULL);

	if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
	(void) zvol_os_create_minor(name);
	} else {
	(void) zvol_os_create_minor(name);
	}
	}

	/*
	* Remove minors for specified dataset including children and snapshots.
	*/

	static void
	zvol_free_task(void *arg)
	{
	zvol_os_free(arg);
	}

	void
	zvol_remove_minors_impl(const char *name)
	{
	zvol_state_t zv, zv_next;
	int namelen = ((name) ? strlen(name) : 0);
	taskqid_t t;
	list_t free_list;

	if (zvol_inhibit_dev)
	return;

	list_create(&free_list, sizeof (zvol_state_t),
	offsetof(zvol_state_t, zv_next));

	rw_enter(&zvol_state_lock, RW_WRITER);

	for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
	zv_next = list_next(&zvol_state_list, zv);

	mutex_enter(&zv->zv_state_lock);
	if (name == NULL \|\| strcmp(zv->zv_name, name) == 0 \|\|
	(strncmp(zv->zv_name, name, namelen) == 0 &&
	(zv->zv_name[namelen] == '/' \|\|
	zv->zv_name[namelen] == '@'))) {
	/*
	* By holding zv_state_lock here, we guarantee that no
	* one is currently using this zv
	*/

	/* If in use, leave alone */
	if (zv->zv_open_count > 0 \|\|
	atomic_read(&zv->zv_suspend_ref)) {
	mutex_exit(&zv->zv_state_lock);
	continue;
	}

	zvol_remove(zv);

	/*
	* Cleared while holding zvol_state_lock as a writer
	* which will prevent zvol_open() from opening it.
	*/
	zvol_os_clear_private(zv);

	/* Drop zv_state_lock before zvol_free() */
	mutex_exit(&zv->zv_state_lock);

	/* Try parallel zv_free, if failed do it in place */
	t = taskq_dispatch(system_taskq, zvol_free_task, zv,
	TQ_SLEEP);
	if (t == TASKQID_INVALID)
	list_insert_head(&free_list, zv);
	} else {
	mutex_exit(&zv->zv_state_lock);
	}
	}
	rw_exit(&zvol_state_lock);

	/* Drop zvol_state_lock before calling zvol_free() */
	while ((zv = list_head(&free_list)) != NULL) {
	list_remove(&free_list, zv);
	zvol_os_free(zv);
	}
	}

	/* Remove minor for this specific volume only */
	static void
	zvol_remove_minor_impl(const char *name)
	{
	zvol_state_t zv = NULL, zv_next;

	if (zvol_inhibit_dev)
	return;

	rw_enter(&zvol_state_lock, RW_WRITER);

	for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
	zv_next = list_next(&zvol_state_list, zv);

	mutex_enter(&zv->zv_state_lock);
	if (strcmp(zv->zv_name, name) == 0) {
	/*
	* By holding zv_state_lock here, we guarantee that no
	* one is currently using this zv
	*/

	/* If in use, leave alone */
	if (zv->zv_open_count > 0 \|\|
	atomic_read(&zv->zv_suspend_ref)) {
	mutex_exit(&zv->zv_state_lock);
	continue;
	}
	zvol_remove(zv);

	zvol_os_clear_private(zv);
	mutex_exit(&zv->zv_state_lock);
	break;
	} else {
	mutex_exit(&zv->zv_state_lock);
	}
	}

	/* Drop zvol_state_lock before calling zvol_free() */
	rw_exit(&zvol_state_lock);

	if (zv != NULL)
	zvol_os_free(zv);
	}

	/*
	* Rename minors for specified dataset including children and snapshots.
	*/
	static void
	zvol_rename_minors_impl(const char oldname, const char newname)
	{
	zvol_state_t zv, zv_next;
	int oldnamelen;

	if (zvol_inhibit_dev)
	return;

	oldnamelen = strlen(oldname);

	rw_enter(&zvol_state_lock, RW_READER);

	for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
	zv_next = list_next(&zvol_state_list, zv);

	mutex_enter(&zv->zv_state_lock);

	if (strcmp(zv->zv_name, oldname) == 0) {
	zvol_os_rename_minor(zv, newname);
	} else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 &&
	(zv->zv_name[oldnamelen] == '/' \|\|
	zv->zv_name[oldnamelen] == '@')) {
	char *name = kmem_asprintf("%s%c%s", newname,
	zv->zv_name[oldnamelen],
	zv->zv_name + oldnamelen + 1);
	zvol_os_rename_minor(zv, name);
	kmem_strfree(name);
	}

	mutex_exit(&zv->zv_state_lock);
	}

	rw_exit(&zvol_state_lock);
	}

	typedef struct zvol_snapdev_cb_arg {
	uint64_t snapdev;
	} zvol_snapdev_cb_arg_t;

	static int
	zvol_set_snapdev_cb(const char dsname, void param)
	{
	zvol_snapdev_cb_arg_t *arg = param;

	if (strchr(dsname, '@') == NULL)
	return (0);

	switch (arg->snapdev) {
	case ZFS_SNAPDEV_VISIBLE:
	(void) zvol_os_create_minor(dsname);
	break;
	case ZFS_SNAPDEV_HIDDEN:
	(void) zvol_remove_minor_impl(dsname);
	break;
	}

	return (0);
	}

	static void
	zvol_set_snapdev_impl(char *name, uint64_t snapdev)
	{
	zvol_snapdev_cb_arg_t arg = {snapdev};
	fstrans_cookie_t cookie = spl_fstrans_mark();
	/*
	* The zvol_set_snapdev_sync() sets snapdev appropriately
	* in the dataset hierarchy. Here, we only scan snapshots.
	*/
	dmu_objset_find(name, zvol_set_snapdev_cb, &arg, DS_FIND_SNAPSHOTS);
	spl_fstrans_unmark(cookie);
	}

	static void
	zvol_set_volmode_impl(char *name, uint64_t volmode)
	{
	fstrans_cookie_t cookie;
	uint64_t old_volmode;
	zvol_state_t *zv;

	if (strchr(name, '@') != NULL)
	return;

	/*
	* It's unfortunate we need to remove minors before we create new ones:
	* this is necessary because our backing gendisk (zvol_state->zv_disk)
	* could be different when we set, for instance, volmode from "geom"
	* to "dev" (or vice versa).
	*/
	zv = zvol_find_by_name(name, RW_NONE);
	if (zv == NULL && volmode == ZFS_VOLMODE_NONE)
	return;
	if (zv != NULL) {
	old_volmode = zv->zv_volmode;
	mutex_exit(&zv->zv_state_lock);
	if (old_volmode == volmode)
	return;
	zvol_wait_close(zv);
	}
	cookie = spl_fstrans_mark();
	switch (volmode) {
	case ZFS_VOLMODE_NONE:
	(void) zvol_remove_minor_impl(name);
	break;
	case ZFS_VOLMODE_GEOM:
	case ZFS_VOLMODE_DEV:
	(void) zvol_remove_minor_impl(name);
	(void) zvol_os_create_minor(name);
	break;
	case ZFS_VOLMODE_DEFAULT:
	(void) zvol_remove_minor_impl(name);
	if (zvol_volmode == ZFS_VOLMODE_NONE)
	break;
	else /* if zvol_volmode is invalid defaults to "geom" */
	(void) zvol_os_create_minor(name);
	break;
	}
	spl_fstrans_unmark(cookie);
	}

	static zvol_task_t *
	zvol_task_alloc(zvol_async_op_t op, const char name1, const char name2,
	uint64_t value)
	{
	zvol_task_t *task;

	/* Never allow tasks on hidden names. */
	if (name1[0] == '$')
	return (NULL);

	task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP);
	task->op = op;
	task->value = value;

	strlcpy(task->name1, name1, MAXNAMELEN);
	if (name2 != NULL)
	strlcpy(task->name2, name2, MAXNAMELEN);

	return (task);
	}

	static void
	zvol_task_free(zvol_task_t *task)
	{
	kmem_free(task, sizeof (zvol_task_t));
	}

	/*
	* The worker thread function performed asynchronously.
	*/
	static void
	zvol_task_cb(void *arg)
	{
	zvol_task_t *task = arg;

	switch (task->op) {
	case ZVOL_ASYNC_REMOVE_MINORS:
	zvol_remove_minors_impl(task->name1);
	break;
	case ZVOL_ASYNC_RENAME_MINORS:
	zvol_rename_minors_impl(task->name1, task->name2);
	break;
	case ZVOL_ASYNC_SET_SNAPDEV:
	zvol_set_snapdev_impl(task->name1, task->value);
	break;
	case ZVOL_ASYNC_SET_VOLMODE:
	zvol_set_volmode_impl(task->name1, task->value);
	break;
	default:
	VERIFY(0);
	break;
	}

	zvol_task_free(task);
	}

	typedef struct zvol_set_prop_int_arg {
	const char *zsda_name;
	uint64_t zsda_value;
	zprop_source_t zsda_source;
	dmu_tx_t *zsda_tx;
	} zvol_set_prop_int_arg_t;

	/*
	* Sanity check the dataset for safe use by the sync task. No additional
	* conditions are imposed.
	*/
	static int
	zvol_set_snapdev_check(void arg, dmu_tx_t tx)
	{
	zvol_set_prop_int_arg_t *zsda = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *dd;
	int error;

	error = dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL);
	if (error != 0)
	return (error);

	dsl_dir_rele(dd, FTAG);

	return (error);
	}

	static int
	zvol_set_snapdev_sync_cb(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) arg;
	char dsname[MAXNAMELEN];
	zvol_task_t *task;
	uint64_t snapdev;

	dsl_dataset_name(ds, dsname);
	if (dsl_prop_get_int_ds(ds, "snapdev", &snapdev) != 0)
	return (0);
	task = zvol_task_alloc(ZVOL_ASYNC_SET_SNAPDEV, dsname, NULL, snapdev);
	if (task == NULL)
	return (0);

	(void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb,
	task, TQ_SLEEP);
	return (0);
	}

	/*
	* Traverse all child datasets and apply snapdev appropriately.
	* We call dsl_prop_set_sync_impl() here to set the value only on the toplevel
	* dataset and read the effective "snapdev" on every child in the callback
	* function: this is because the value is not guaranteed to be the same in the
	* whole dataset hierarchy.
	*/
	static void
	zvol_set_snapdev_sync(void arg, dmu_tx_t tx)
	{
	zvol_set_prop_int_arg_t *zsda = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *dd;
	dsl_dataset_t *ds;
	int error;

	VERIFY0(dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL));
	zsda->zsda_tx = tx;

	error = dsl_dataset_hold(dp, zsda->zsda_name, FTAG, &ds);
	if (error == 0) {
	dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_SNAPDEV),
	zsda->zsda_source, sizeof (zsda->zsda_value), 1,
	&zsda->zsda_value, zsda->zsda_tx);
	dsl_dataset_rele(ds, FTAG);
	}
	dmu_objset_find_dp(dp, dd->dd_object, zvol_set_snapdev_sync_cb,
	zsda, DS_FIND_CHILDREN);

	dsl_dir_rele(dd, FTAG);
	}

	int
	zvol_set_snapdev(const char *ddname, zprop_source_t source, uint64_t snapdev)
	{
	zvol_set_prop_int_arg_t zsda;

	zsda.zsda_name = ddname;
	zsda.zsda_source = source;
	zsda.zsda_value = snapdev;

	return (dsl_sync_task(ddname, zvol_set_snapdev_check,
	zvol_set_snapdev_sync, &zsda, 0, ZFS_SPACE_CHECK_NONE));
	}

	/*
	* Sanity check the dataset for safe use by the sync task. No additional
	* conditions are imposed.
	*/
	static int
	zvol_set_volmode_check(void arg, dmu_tx_t tx)
	{
	zvol_set_prop_int_arg_t *zsda = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *dd;
	int error;

	error = dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL);
	if (error != 0)
	return (error);

	dsl_dir_rele(dd, FTAG);

	return (error);
	}

	static int
	zvol_set_volmode_sync_cb(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) arg;
	char dsname[MAXNAMELEN];
	zvol_task_t *task;
	uint64_t volmode;

	dsl_dataset_name(ds, dsname);
	if (dsl_prop_get_int_ds(ds, "volmode", &volmode) != 0)
	return (0);
	task = zvol_task_alloc(ZVOL_ASYNC_SET_VOLMODE, dsname, NULL, volmode);
	if (task == NULL)
	return (0);

	(void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb,
	task, TQ_SLEEP);
	return (0);
	}

	/*
	* Traverse all child datasets and apply volmode appropriately.
	* We call dsl_prop_set_sync_impl() here to set the value only on the toplevel
	* dataset and read the effective "volmode" on every child in the callback
	* function: this is because the value is not guaranteed to be the same in the
	* whole dataset hierarchy.
	*/
	static void
	zvol_set_volmode_sync(void arg, dmu_tx_t tx)
	{
	zvol_set_prop_int_arg_t *zsda = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *dd;
	dsl_dataset_t *ds;
	int error;

	VERIFY0(dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL));
	zsda->zsda_tx = tx;

	error = dsl_dataset_hold(dp, zsda->zsda_name, FTAG, &ds);
	if (error == 0) {
	dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_VOLMODE),
	zsda->zsda_source, sizeof (zsda->zsda_value), 1,
	&zsda->zsda_value, zsda->zsda_tx);
	dsl_dataset_rele(ds, FTAG);
	}

	dmu_objset_find_dp(dp, dd->dd_object, zvol_set_volmode_sync_cb,
	zsda, DS_FIND_CHILDREN);

	dsl_dir_rele(dd, FTAG);
	}

	int
	zvol_set_volmode(const char *ddname, zprop_source_t source, uint64_t volmode)
	{
	zvol_set_prop_int_arg_t zsda;

	zsda.zsda_name = ddname;
	zsda.zsda_source = source;
	zsda.zsda_value = volmode;

	return (dsl_sync_task(ddname, zvol_set_volmode_check,
	zvol_set_volmode_sync, &zsda, 0, ZFS_SPACE_CHECK_NONE));
	}

	void
	zvol_remove_minors(spa_t spa, const char name, boolean_t async)
	{
	zvol_task_t *task;
	taskqid_t id;

	task = zvol_task_alloc(ZVOL_ASYNC_REMOVE_MINORS, name, NULL, ~0ULL);
	if (task == NULL)
	return;

	id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
	if ((async == B_FALSE) && (id != TASKQID_INVALID))
	taskq_wait_id(spa->spa_zvol_taskq, id);
	}

	void
	zvol_rename_minors(spa_t spa, const char name1, const char *name2,
	boolean_t async)
	{
	zvol_task_t *task;
	taskqid_t id;

	task = zvol_task_alloc(ZVOL_ASYNC_RENAME_MINORS, name1, name2, ~0ULL);
	if (task == NULL)
	return;

	id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
	if ((async == B_FALSE) && (id != TASKQID_INVALID))
	taskq_wait_id(spa->spa_zvol_taskq, id);
	}

	boolean_t
	zvol_is_zvol(const char *name)
	{

	return (zvol_os_is_zvol(name));
	}

	int
	zvol_init_impl(void)
	{
	int i;

	list_create(&zvol_state_list, sizeof (zvol_state_t),
	offsetof(zvol_state_t, zv_next));
	rw_init(&zvol_state_lock, NULL, RW_DEFAULT, NULL);

	zvol_htable = kmem_alloc(ZVOL_HT_SIZE * sizeof (struct hlist_head),
	KM_SLEEP);
	for (i = 0; i < ZVOL_HT_SIZE; i++)
	INIT_HLIST_HEAD(&zvol_htable[i]);

	return (0);
	}

	void
	zvol_fini_impl(void)
	{
	zvol_remove_minors_impl(NULL);

	/*
	* The call to "zvol_remove_minors_impl" may dispatch entries to
	* the system_taskq, but it doesn't wait for those entries to
	* complete before it returns. Thus, we must wait for all of the
	* removals to finish, before we can continue.
	*/
	taskq_wait_outstanding(system_taskq, 0);

	kmem_free(zvol_htable, ZVOL_HT_SIZE * sizeof (struct hlist_head));
	list_destroy(&zvol_state_list);
	rw_destroy(&zvol_state_lock);
	}
	diff --git a/sys/contrib/openzfs/module/zstd/zfs_zstd.c b/sys/contrib/openzfs/module/zstd/zfs_zstd.c
	index 48b41ff7532c..7d031bbdbe31 100644
	--- a/sys/contrib/openzfs/module/zstd/zfs_zstd.c
	+++ b/sys/contrib/openzfs/module/zstd/zfs_zstd.c
	@@ -1,920 +1,920 @@
	/*
	* BSD 3-Clause New License (https://spdx.org/licenses/BSD-3-Clause.html)
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions are met:
	*
	* 1. Redistributions of source code must retain the above copyright notice,
	* this list of conditions and the following disclaimer.
	*
	* 2. Redistributions in binary form must reproduce the above copyright notice,
	* this list of conditions and the following disclaimer in the documentation
	* and/or other materials provided with the distribution.
	*
	* 3. Neither the name of the copyright holder nor the names of its
	* contributors may be used to endorse or promote products derived from this
	* software without specific prior written permission.
	*
	* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
	* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
	* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
	* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
	* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
	* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
	* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
	* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
	* POSSIBILITY OF SUCH DAMAGE.
	*/

	/*
	* Copyright (c) 2016-2018, Klara Inc.
	* Copyright (c) 2016-2018, Allan Jude
	* Copyright (c) 2018-2020, Sebastian Gottschall
	* Copyright (c) 2019-2020, Michael Niewöhner
	* Copyright (c) 2020, The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	*/

	#include <sys/param.h>
	#include <sys/sysmacros.h>
	#include <sys/zfs_context.h>
	#include <sys/zio_compress.h>
	#include <sys/spa.h>
	#include <sys/zstd/zstd.h>

	#define ZSTD_STATIC_LINKING_ONLY
	#include "lib/zstd.h"
	#include "lib/common/zstd_errors.h"

	#ifndef IN_LIBSA
	-static int zstd_earlyabort_pass = 1;
	+static uint_t zstd_earlyabort_pass = 1;
	static int zstd_cutoff_level = ZIO_ZSTD_LEVEL_3;
	static unsigned int zstd_abort_size = (128 * 1024);
	#endif

	static kstat_t *zstd_ksp = NULL;

	typedef struct zstd_stats {
	kstat_named_t zstd_stat_alloc_fail;
	kstat_named_t zstd_stat_alloc_fallback;
	kstat_named_t zstd_stat_com_alloc_fail;
	kstat_named_t zstd_stat_dec_alloc_fail;
	kstat_named_t zstd_stat_com_inval;
	kstat_named_t zstd_stat_dec_inval;
	kstat_named_t zstd_stat_dec_header_inval;
	kstat_named_t zstd_stat_com_fail;
	kstat_named_t zstd_stat_dec_fail;
	/*
	* LZ4 first-pass early abort verdict
	*/
	kstat_named_t zstd_stat_lz4pass_allowed;
	kstat_named_t zstd_stat_lz4pass_rejected;
	/*
	* zstd-1 second-pass early abort verdict
	*/
	kstat_named_t zstd_stat_zstdpass_allowed;
	kstat_named_t zstd_stat_zstdpass_rejected;
	/*
	* We excluded this from early abort for some reason
	*/
	kstat_named_t zstd_stat_passignored;
	kstat_named_t zstd_stat_passignored_size;
	kstat_named_t zstd_stat_buffers;
	kstat_named_t zstd_stat_size;
	} zstd_stats_t;

	static zstd_stats_t zstd_stats = {
	{ "alloc_fail", KSTAT_DATA_UINT64 },
	{ "alloc_fallback", KSTAT_DATA_UINT64 },
	{ "compress_alloc_fail", KSTAT_DATA_UINT64 },
	{ "decompress_alloc_fail", KSTAT_DATA_UINT64 },
	{ "compress_level_invalid", KSTAT_DATA_UINT64 },
	{ "decompress_level_invalid", KSTAT_DATA_UINT64 },
	{ "decompress_header_invalid", KSTAT_DATA_UINT64 },
	{ "compress_failed", KSTAT_DATA_UINT64 },
	{ "decompress_failed", KSTAT_DATA_UINT64 },
	{ "lz4pass_allowed", KSTAT_DATA_UINT64 },
	{ "lz4pass_rejected", KSTAT_DATA_UINT64 },
	{ "zstdpass_allowed", KSTAT_DATA_UINT64 },
	{ "zstdpass_rejected", KSTAT_DATA_UINT64 },
	{ "passignored", KSTAT_DATA_UINT64 },
	{ "passignored_size", KSTAT_DATA_UINT64 },
	{ "buffers", KSTAT_DATA_UINT64 },
	{ "size", KSTAT_DATA_UINT64 },
	};

	#ifdef _KERNEL
	static int
	kstat_zstd_update(kstat_t *ksp, int rw)
	{
	ASSERT(ksp != NULL);

	if (rw == KSTAT_WRITE && ksp == zstd_ksp) {
	ZSTDSTAT_ZERO(zstd_stat_alloc_fail);
	ZSTDSTAT_ZERO(zstd_stat_alloc_fallback);
	ZSTDSTAT_ZERO(zstd_stat_com_alloc_fail);
	ZSTDSTAT_ZERO(zstd_stat_dec_alloc_fail);
	ZSTDSTAT_ZERO(zstd_stat_com_inval);
	ZSTDSTAT_ZERO(zstd_stat_dec_inval);
	ZSTDSTAT_ZERO(zstd_stat_dec_header_inval);
	ZSTDSTAT_ZERO(zstd_stat_com_fail);
	ZSTDSTAT_ZERO(zstd_stat_dec_fail);
	ZSTDSTAT_ZERO(zstd_stat_lz4pass_allowed);
	ZSTDSTAT_ZERO(zstd_stat_lz4pass_rejected);
	ZSTDSTAT_ZERO(zstd_stat_zstdpass_allowed);
	ZSTDSTAT_ZERO(zstd_stat_zstdpass_rejected);
	ZSTDSTAT_ZERO(zstd_stat_passignored);
	ZSTDSTAT_ZERO(zstd_stat_passignored_size);
	}

	return (0);
	}
	#endif

	/* Enums describing the allocator type specified by kmem_type in zstd_kmem */
	enum zstd_kmem_type {
	ZSTD_KMEM_UNKNOWN = 0,
	/* Allocation type using kmem_vmalloc */
	ZSTD_KMEM_DEFAULT,
	/* Pool based allocation using mempool_alloc */
	ZSTD_KMEM_POOL,
	/* Reserved fallback memory for decompression only */
	ZSTD_KMEM_DCTX,
	ZSTD_KMEM_COUNT,
	};

	/* Structure for pooled memory objects */
	struct zstd_pool {
	void *mem;
	size_t size;
	kmutex_t barrier;
	hrtime_t timeout;
	};

	/* Global structure for handling memory allocations */
	struct zstd_kmem {
	enum zstd_kmem_type kmem_type;
	size_t kmem_size;
	struct zstd_pool *pool;
	};

	/* Fallback memory structure used for decompression only if memory runs out */
	struct zstd_fallback_mem {
	size_t mem_size;
	void *mem;
	kmutex_t barrier;
	};

	struct zstd_levelmap {
	int16_t zstd_level;
	enum zio_zstd_levels level;
	};

	/*
	* ZSTD memory handlers
	*
	* For decompression we use a different handler which also provides fallback
	* memory allocation in case memory runs out.
	*
	* The ZSTD handlers were split up for the most simplified implementation.
	*/
	#ifndef IN_LIBSA
	static void zstd_alloc(void opaque, size_t size);
	#endif
	static void zstd_dctx_alloc(void opaque, size_t size);
	static void zstd_free(void opaque, void ptr);

	#ifndef IN_LIBSA
	/* Compression memory handler */
	static const ZSTD_customMem zstd_malloc = {
	zstd_alloc,
	zstd_free,
	NULL,
	};
	#endif

	/* Decompression memory handler */
	static const ZSTD_customMem zstd_dctx_malloc = {
	zstd_dctx_alloc,
	zstd_free,
	NULL,
	};

	/* Level map for converting ZFS internal levels to ZSTD levels and vice versa */
	static struct zstd_levelmap zstd_levels[] = {
	{ZIO_ZSTD_LEVEL_1, ZIO_ZSTD_LEVEL_1},
	{ZIO_ZSTD_LEVEL_2, ZIO_ZSTD_LEVEL_2},
	{ZIO_ZSTD_LEVEL_3, ZIO_ZSTD_LEVEL_3},
	{ZIO_ZSTD_LEVEL_4, ZIO_ZSTD_LEVEL_4},
	{ZIO_ZSTD_LEVEL_5, ZIO_ZSTD_LEVEL_5},
	{ZIO_ZSTD_LEVEL_6, ZIO_ZSTD_LEVEL_6},
	{ZIO_ZSTD_LEVEL_7, ZIO_ZSTD_LEVEL_7},
	{ZIO_ZSTD_LEVEL_8, ZIO_ZSTD_LEVEL_8},
	{ZIO_ZSTD_LEVEL_9, ZIO_ZSTD_LEVEL_9},
	{ZIO_ZSTD_LEVEL_10, ZIO_ZSTD_LEVEL_10},
	{ZIO_ZSTD_LEVEL_11, ZIO_ZSTD_LEVEL_11},
	{ZIO_ZSTD_LEVEL_12, ZIO_ZSTD_LEVEL_12},
	{ZIO_ZSTD_LEVEL_13, ZIO_ZSTD_LEVEL_13},
	{ZIO_ZSTD_LEVEL_14, ZIO_ZSTD_LEVEL_14},
	{ZIO_ZSTD_LEVEL_15, ZIO_ZSTD_LEVEL_15},
	{ZIO_ZSTD_LEVEL_16, ZIO_ZSTD_LEVEL_16},
	{ZIO_ZSTD_LEVEL_17, ZIO_ZSTD_LEVEL_17},
	{ZIO_ZSTD_LEVEL_18, ZIO_ZSTD_LEVEL_18},
	{ZIO_ZSTD_LEVEL_19, ZIO_ZSTD_LEVEL_19},
	{-1, ZIO_ZSTD_LEVEL_FAST_1},
	{-2, ZIO_ZSTD_LEVEL_FAST_2},
	{-3, ZIO_ZSTD_LEVEL_FAST_3},
	{-4, ZIO_ZSTD_LEVEL_FAST_4},
	{-5, ZIO_ZSTD_LEVEL_FAST_5},
	{-6, ZIO_ZSTD_LEVEL_FAST_6},
	{-7, ZIO_ZSTD_LEVEL_FAST_7},
	{-8, ZIO_ZSTD_LEVEL_FAST_8},
	{-9, ZIO_ZSTD_LEVEL_FAST_9},
	{-10, ZIO_ZSTD_LEVEL_FAST_10},
	{-20, ZIO_ZSTD_LEVEL_FAST_20},
	{-30, ZIO_ZSTD_LEVEL_FAST_30},
	{-40, ZIO_ZSTD_LEVEL_FAST_40},
	{-50, ZIO_ZSTD_LEVEL_FAST_50},
	{-60, ZIO_ZSTD_LEVEL_FAST_60},
	{-70, ZIO_ZSTD_LEVEL_FAST_70},
	{-80, ZIO_ZSTD_LEVEL_FAST_80},
	{-90, ZIO_ZSTD_LEVEL_FAST_90},
	{-100, ZIO_ZSTD_LEVEL_FAST_100},
	{-500, ZIO_ZSTD_LEVEL_FAST_500},
	{-1000, ZIO_ZSTD_LEVEL_FAST_1000},
	};

	/*
	* This variable represents the maximum count of the pool based on the number
	* of CPUs plus some buffer. We default to cpu count * 4, see init_zstd.
	*/
	static int pool_count = 16;

	#define ZSTD_POOL_MAX pool_count
	#define ZSTD_POOL_TIMEOUT 60 * 2

	static struct zstd_fallback_mem zstd_dctx_fallback;
	static struct zstd_pool *zstd_mempool_cctx;
	static struct zstd_pool *zstd_mempool_dctx;

	/*
	* The library zstd code expects these if ADDRESS_SANITIZER gets defined,
	* and while ASAN does this, KASAN defines that and does not. So to avoid
	* changing the external code, we do this.
	*/
	#if defined(ZFS_ASAN_ENABLED)
	#define ADDRESS_SANITIZER 1
	#endif
	#if defined(_KERNEL) && defined(ADDRESS_SANITIZER)
	void __asan_unpoison_memory_region(void const volatile *addr, size_t size);
	void __asan_poison_memory_region(void const volatile *addr, size_t size);
	void __asan_unpoison_memory_region(void const volatile *addr, size_t size) {};
	void __asan_poison_memory_region(void const volatile *addr, size_t size) {};
	#endif


	static void
	zstd_mempool_reap(struct zstd_pool *zstd_mempool)
	{
	struct zstd_pool *pool;

	if (!zstd_mempool \|\| !ZSTDSTAT(zstd_stat_buffers)) {
	return;
	}

	/* free obsolete slots */
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
	pool = &zstd_mempool[i];
	if (pool->mem && mutex_tryenter(&pool->barrier)) {
	/* Free memory if unused object older than 2 minutes */
	if (pool->mem && gethrestime_sec() > pool->timeout) {
	vmem_free(pool->mem, pool->size);
	ZSTDSTAT_SUB(zstd_stat_buffers, 1);
	ZSTDSTAT_SUB(zstd_stat_size, pool->size);
	pool->mem = NULL;
	pool->size = 0;
	pool->timeout = 0;
	}
	mutex_exit(&pool->barrier);
	}
	}
	}

	/*
	* Try to get a cached allocated buffer from memory pool or allocate a new one
	* if necessary. If a object is older than 2 minutes and does not fit the
	* requested size, it will be released and a new cached entry will be allocated.
	* If other pooled objects are detected without being used for 2 minutes, they
	* will be released, too.
	*
	* The concept is that high frequency memory allocations of bigger objects are
	* expensive. So if a lot of work is going on, allocations will be kept for a
	* while and can be reused in that time frame.
	*
	* The scheduled release will be updated every time a object is reused.
	*/

	static void *
	zstd_mempool_alloc(struct zstd_pool *zstd_mempool, size_t size)
	{
	struct zstd_pool *pool;
	struct zstd_kmem *mem = NULL;

	if (!zstd_mempool) {
	return (NULL);
	}

	/* Seek for preallocated memory slot and free obsolete slots */
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
	pool = &zstd_mempool[i];
	/*
	* This lock is simply a marker for a pool object being in use.
	* If it's already hold, it will be skipped.
	*
	* We need to create it before checking it to avoid race
	* conditions caused by running in a threaded context.
	*
	* The lock is later released by zstd_mempool_free.
	*/
	if (mutex_tryenter(&pool->barrier)) {
	/*
	* Check if objects fits the size, if so we take it and
	* update the timestamp.
	*/
	if (pool->mem && size <= pool->size) {
	pool->timeout = gethrestime_sec() +
	ZSTD_POOL_TIMEOUT;
	mem = pool->mem;
	return (mem);
	}
	mutex_exit(&pool->barrier);
	}
	}

	/*
	* If no preallocated slot was found, try to fill in a new one.
	*
	* We run a similar algorithm twice here to avoid pool fragmentation.
	* The first one may generate holes in the list if objects get released.
	* We always make sure that these holes get filled instead of adding new
	* allocations constantly at the end.
	*/
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
	pool = &zstd_mempool[i];
	if (mutex_tryenter(&pool->barrier)) {
	/* Object is free, try to allocate new one */
	if (!pool->mem) {
	mem = vmem_alloc(size, KM_SLEEP);
	if (mem) {
	ZSTDSTAT_ADD(zstd_stat_buffers, 1);
	ZSTDSTAT_ADD(zstd_stat_size, size);
	pool->mem = mem;
	pool->size = size;
	/* Keep track for later release */
	mem->pool = pool;
	mem->kmem_type = ZSTD_KMEM_POOL;
	mem->kmem_size = size;
	}
	}

	if (size <= pool->size) {
	/* Update timestamp */
	pool->timeout = gethrestime_sec() +
	ZSTD_POOL_TIMEOUT;

	return (pool->mem);
	}

	mutex_exit(&pool->barrier);
	}
	}

	/*
	* If the pool is full or the allocation failed, try lazy allocation
	* instead.
	*/
	if (!mem) {
	mem = vmem_alloc(size, KM_NOSLEEP);
	if (mem) {
	mem->pool = NULL;
	mem->kmem_type = ZSTD_KMEM_DEFAULT;
	mem->kmem_size = size;
	}
	}

	return (mem);
	}

	/* Mark object as released by releasing the barrier mutex */
	static void
	zstd_mempool_free(struct zstd_kmem *z)
	{
	mutex_exit(&z->pool->barrier);
	}

	/* Convert ZFS internal enum to ZSTD level */
	static int
	zstd_enum_to_level(enum zio_zstd_levels level, int16_t *zstd_level)
	{
	if (level > 0 && level <= ZIO_ZSTD_LEVEL_19) {
	*zstd_level = zstd_levels[level - 1].zstd_level;
	return (0);
	}
	if (level >= ZIO_ZSTD_LEVEL_FAST_1 &&
	level <= ZIO_ZSTD_LEVEL_FAST_1000) {
	*zstd_level = zstd_levels[level - ZIO_ZSTD_LEVEL_FAST_1
	+ ZIO_ZSTD_LEVEL_19].zstd_level;
	return (0);
	}

	/* Invalid/unknown zfs compression enum - this should never happen. */
	return (1);
	}

	#ifndef IN_LIBSA
	size_t
	zfs_zstd_compress_wrap(void s_start, void d_start, size_t s_len, size_t d_len,
	int level)
	{
	int16_t zstd_level;
	if (zstd_enum_to_level(level, &zstd_level)) {
	ZSTDSTAT_BUMP(zstd_stat_com_inval);
	return (s_len);
	}
	/*
	* A zstd early abort heuristic.
	*
	* - Zeroth, if this is <= zstd-3, or < zstd_abort_size (currently
	* 128k), don't try any of this, just go.
	* (because experimentally that was a reasonable cutoff for a perf win
	* with tiny ratio change)
	* - First, we try LZ4 compression, and if it doesn't early abort, we
	* jump directly to whatever compression level we intended to try.
	* - Second, we try zstd-1 - if that errors out (usually, but not
	* exclusively, if it would overflow), we give up early.
	*
	* If it works, instead we go on and compress anyway.
	*
	* Why two passes? LZ4 alone gets you a lot of the way, but on highly
	* compressible data, it was losing up to 8.5% of the compressed
	* savings versus no early abort, and all the zstd-fast levels are
	* worse indications on their own than LZ4, and don't improve the LZ4
	* pass noticably if stacked like this.
	*/
	size_t actual_abort_size = zstd_abort_size;
	if (zstd_earlyabort_pass > 0 && zstd_level >= zstd_cutoff_level &&
	s_len >= actual_abort_size) {
	int pass_len = 1;
	pass_len = lz4_compress_zfs(s_start, d_start, s_len, d_len, 0);
	if (pass_len < d_len) {
	ZSTDSTAT_BUMP(zstd_stat_lz4pass_allowed);
	goto keep_trying;
	}
	ZSTDSTAT_BUMP(zstd_stat_lz4pass_rejected);

	pass_len = zfs_zstd_compress(s_start, d_start, s_len, d_len,
	ZIO_ZSTD_LEVEL_1);
	if (pass_len == s_len \|\| pass_len <= 0 \|\| pass_len > d_len) {
	ZSTDSTAT_BUMP(zstd_stat_zstdpass_rejected);
	return (s_len);
	}
	ZSTDSTAT_BUMP(zstd_stat_zstdpass_allowed);
	} else {
	ZSTDSTAT_BUMP(zstd_stat_passignored);
	if (s_len < actual_abort_size) {
	ZSTDSTAT_BUMP(zstd_stat_passignored_size);
	}
	}
	keep_trying:
	return (zfs_zstd_compress(s_start, d_start, s_len, d_len, level));

	}

	/* Compress block using zstd */
	size_t
	zfs_zstd_compress(void s_start, void d_start, size_t s_len, size_t d_len,
	int level)
	{
	size_t c_len;
	int16_t zstd_level;
	zfs_zstdhdr_t *hdr;
	ZSTD_CCtx *cctx;

	hdr = (zfs_zstdhdr_t *)d_start;

	/* Skip compression if the specified level is invalid */
	if (zstd_enum_to_level(level, &zstd_level)) {
	ZSTDSTAT_BUMP(zstd_stat_com_inval);
	return (s_len);
	}

	ASSERT3U(d_len, >=, sizeof (*hdr));
	ASSERT3U(d_len, <=, s_len);
	ASSERT3U(zstd_level, !=, 0);

	cctx = ZSTD_createCCtx_advanced(zstd_malloc);

	/*
	* Out of kernel memory, gently fall through - this will disable
	* compression in zio_compress_data
	*/
	if (!cctx) {
	ZSTDSTAT_BUMP(zstd_stat_com_alloc_fail);
	return (s_len);
	}

	/* Set the compression level */
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, zstd_level);

	/* Use the "magicless" zstd header which saves us 4 header bytes */
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless);

	/*
	* Disable redundant checksum calculation and content size storage since
	* this is already done by ZFS itself.
	*/
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 0);
	ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0);

	c_len = ZSTD_compress2(cctx,
	hdr->data,
	d_len - sizeof (*hdr),
	s_start, s_len);

	ZSTD_freeCCtx(cctx);

	/* Error in the compression routine, disable compression. */
	if (ZSTD_isError(c_len)) {
	/*
	* If we are aborting the compression because the saves are
	* too small, that is not a failure. Everything else is a
	* failure, so increment the compression failure counter.
	*/
	int err = ZSTD_getErrorCode(c_len);
	if (err != ZSTD_error_dstSize_tooSmall) {
	ZSTDSTAT_BUMP(zstd_stat_com_fail);
	dprintf("Error: %s", ZSTD_getErrorString(err));
	}
	return (s_len);
	}

	/*
	* Encode the compressed buffer size at the start. We'll need this in
	* decompression to counter the effects of padding which might be added
	* to the compressed buffer and which, if unhandled, would confuse the
	* hell out of our decompression function.
	*/
	hdr->c_len = BE_32(c_len);

	/*
	* Check version for overflow.
	* The limit of 24 bits must not be exceeded. This allows a maximum
	* version 1677.72.15 which we don't expect to be ever reached.
	*/
	ASSERT3U(ZSTD_VERSION_NUMBER, <=, 0xFFFFFF);

	/*
	* Encode the compression level as well. We may need to know the
	* original compression level if compressed_arc is disabled, to match
	* the compression settings to write this block to the L2ARC.
	*
	* Encode the actual level, so if the enum changes in the future, we
	* will be compatible.
	*
	* The upper 24 bits store the ZSTD version to be able to provide
	* future compatibility, since new versions might enhance the
	* compression algorithm in a way, where the compressed data will
	* change.
	*
	* As soon as such incompatibility occurs, handling code needs to be
	* added, differentiating between the versions.
	*/
	zfs_set_hdrversion(hdr, ZSTD_VERSION_NUMBER);
	zfs_set_hdrlevel(hdr, level);
	hdr->raw_version_level = BE_32(hdr->raw_version_level);

	return (c_len + sizeof (*hdr));
	}
	#endif

	/* Decompress block using zstd and return its stored level */
	int
	zfs_zstd_decompress_level(void s_start, void d_start, size_t s_len,
	size_t d_len, uint8_t *level)
	{
	ZSTD_DCtx *dctx;
	size_t result;
	int16_t zstd_level;
	uint32_t c_len;
	const zfs_zstdhdr_t *hdr;
	zfs_zstdhdr_t hdr_copy;

	hdr = (const zfs_zstdhdr_t *)s_start;
	c_len = BE_32(hdr->c_len);

	/*
	* Make a copy instead of directly converting the header, since we must
	* not modify the original data that may be used again later.
	*/
	hdr_copy.raw_version_level = BE_32(hdr->raw_version_level);
	uint8_t curlevel = zfs_get_hdrlevel(&hdr_copy);

	/*
	* NOTE: We ignore the ZSTD version for now. As soon as any
	* incompatibility occurs, it has to be handled accordingly.
	* The version can be accessed via `hdr_copy.version`.
	*/

	/*
	* Convert and check the level
	* An invalid level is a strong indicator for data corruption! In such
	* case return an error so the upper layers can try to fix it.
	*/
	if (zstd_enum_to_level(curlevel, &zstd_level)) {
	ZSTDSTAT_BUMP(zstd_stat_dec_inval);
	return (1);
	}

	ASSERT3U(d_len, >=, s_len);
	ASSERT3U(curlevel, !=, ZIO_COMPLEVEL_INHERIT);

	/* Invalid compressed buffer size encoded at start */
	if (c_len + sizeof (*hdr) > s_len) {
	ZSTDSTAT_BUMP(zstd_stat_dec_header_inval);
	return (1);
	}

	dctx = ZSTD_createDCtx_advanced(zstd_dctx_malloc);
	if (!dctx) {
	ZSTDSTAT_BUMP(zstd_stat_dec_alloc_fail);
	return (1);
	}

	/* Set header type to "magicless" */
	ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless);

	/* Decompress the data and release the context */
	result = ZSTD_decompressDCtx(dctx, d_start, d_len, hdr->data, c_len);
	ZSTD_freeDCtx(dctx);

	/*
	* Returns 0 on success (decompression function returned non-negative)
	* and non-zero on failure (decompression function returned negative.
	*/
	if (ZSTD_isError(result)) {
	ZSTDSTAT_BUMP(zstd_stat_dec_fail);
	return (1);
	}

	if (level) {
	*level = curlevel;
	}

	return (0);
	}

	/* Decompress datablock using zstd */
	int
	zfs_zstd_decompress(void s_start, void d_start, size_t s_len, size_t d_len,
	int level __maybe_unused)
	{

	return (zfs_zstd_decompress_level(s_start, d_start, s_len, d_len,
	NULL));
	}

	#ifndef IN_LIBSA
	/* Allocator for zstd compression context using mempool_allocator */
	static void *
	zstd_alloc(void *opaque __maybe_unused, size_t size)
	{
	size_t nbytes = sizeof (struct zstd_kmem) + size;
	struct zstd_kmem *z = NULL;

	z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_cctx, nbytes);

	if (!z) {
	ZSTDSTAT_BUMP(zstd_stat_alloc_fail);
	return (NULL);
	}

	return ((void*)z + (sizeof (struct zstd_kmem)));
	}
	#endif

	/*
	* Allocator for zstd decompression context using mempool_allocator with
	* fallback to reserved memory if allocation fails
	*/
	static void *
	zstd_dctx_alloc(void *opaque __maybe_unused, size_t size)
	{
	size_t nbytes = sizeof (struct zstd_kmem) + size;
	struct zstd_kmem *z = NULL;
	enum zstd_kmem_type type = ZSTD_KMEM_DEFAULT;

	z = (struct zstd_kmem *)zstd_mempool_alloc(zstd_mempool_dctx, nbytes);
	if (!z) {
	/* Try harder, decompression shall not fail */
	z = vmem_alloc(nbytes, KM_SLEEP);
	if (z) {
	z->pool = NULL;
	}
	ZSTDSTAT_BUMP(zstd_stat_alloc_fail);
	} else {
	return ((void*)z + (sizeof (struct zstd_kmem)));
	}

	/* Fallback if everything fails */
	if (!z) {
	/*
	* Barrier since we only can handle it in a single thread. All
	* other following threads need to wait here until decompression
	* is completed. zstd_free will release this barrier later.
	*/
	mutex_enter(&zstd_dctx_fallback.barrier);

	z = zstd_dctx_fallback.mem;
	type = ZSTD_KMEM_DCTX;
	ZSTDSTAT_BUMP(zstd_stat_alloc_fallback);
	}

	/* Allocation should always be successful */
	if (!z) {
	return (NULL);
	}

	z->kmem_type = type;
	z->kmem_size = nbytes;

	return ((void*)z + (sizeof (struct zstd_kmem)));
	}

	/* Free allocated memory by its specific type */
	static void
	zstd_free(void opaque __maybe_unused, void ptr)
	{
	struct zstd_kmem *z = (ptr - sizeof (struct zstd_kmem));
	enum zstd_kmem_type type;

	ASSERT3U(z->kmem_type, <, ZSTD_KMEM_COUNT);
	ASSERT3U(z->kmem_type, >, ZSTD_KMEM_UNKNOWN);

	type = z->kmem_type;
	switch (type) {
	case ZSTD_KMEM_DEFAULT:
	vmem_free(z, z->kmem_size);
	break;
	case ZSTD_KMEM_POOL:
	zstd_mempool_free(z);
	break;
	case ZSTD_KMEM_DCTX:
	mutex_exit(&zstd_dctx_fallback.barrier);
	break;
	default:
	break;
	}
	}

	/* Allocate fallback memory to ensure safe decompression */
	static void __init
	create_fallback_mem(struct zstd_fallback_mem *mem, size_t size)
	{
	mem->mem_size = size;
	mem->mem = vmem_zalloc(mem->mem_size, KM_SLEEP);
	mutex_init(&mem->barrier, NULL, MUTEX_DEFAULT, NULL);
	}

	/* Initialize memory pool barrier mutexes */
	static void __init
	zstd_mempool_init(void)
	{
	zstd_mempool_cctx = (struct zstd_pool *)
	kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);
	zstd_mempool_dctx = (struct zstd_pool *)
	kmem_zalloc(ZSTD_POOL_MAX * sizeof (struct zstd_pool), KM_SLEEP);

	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
	mutex_init(&zstd_mempool_cctx[i].barrier, NULL,
	MUTEX_DEFAULT, NULL);
	mutex_init(&zstd_mempool_dctx[i].barrier, NULL,
	MUTEX_DEFAULT, NULL);
	}
	}

	/* Initialize zstd-related memory handling */
	static int __init
	zstd_meminit(void)
	{
	zstd_mempool_init();

	/*
	* Estimate the size of the fallback decompression context.
	* The expected size on x64 with current ZSTD should be about 160 KB.
	*/
	create_fallback_mem(&zstd_dctx_fallback,
	P2ROUNDUP(ZSTD_estimateDCtxSize() + sizeof (struct zstd_kmem),
	PAGESIZE));

	return (0);
	}

	/* Release object from pool and free memory */
	static void
	release_pool(struct zstd_pool *pool)
	{
	mutex_destroy(&pool->barrier);
	vmem_free(pool->mem, pool->size);
	pool->mem = NULL;
	pool->size = 0;
	}

	/* Release memory pool objects */
	static void
	zstd_mempool_deinit(void)
	{
	for (int i = 0; i < ZSTD_POOL_MAX; i++) {
	release_pool(&zstd_mempool_cctx[i]);
	release_pool(&zstd_mempool_dctx[i]);
	}

	kmem_free(zstd_mempool_dctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
	kmem_free(zstd_mempool_cctx, ZSTD_POOL_MAX * sizeof (struct zstd_pool));
	zstd_mempool_dctx = NULL;
	zstd_mempool_cctx = NULL;
	}

	/* release unused memory from pool */

	void
	zfs_zstd_cache_reap_now(void)
	{

	/*
	* Short-circuit if there are no buffers to begin with.
	*/
	if (ZSTDSTAT(zstd_stat_buffers) == 0)
	return;

	/*
	* calling alloc with zero size seeks
	* and releases old unused objects
	*/
	zstd_mempool_reap(zstd_mempool_cctx);
	zstd_mempool_reap(zstd_mempool_dctx);
	}

	extern int __init
	zstd_init(void)
	{
	/* Set pool size by using maximum sane thread count * 4 */
	pool_count = (boot_ncpus * 4);
	zstd_meminit();

	/* Initialize kstat */
	zstd_ksp = kstat_create("zfs", 0, "zstd", "misc",
	KSTAT_TYPE_NAMED, sizeof (zstd_stats) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);
	if (zstd_ksp != NULL) {
	zstd_ksp->ks_data = &zstd_stats;
	kstat_install(zstd_ksp);
	#ifdef _KERNEL
	zstd_ksp->ks_update = kstat_zstd_update;
	#endif
	}

	return (0);
	}

	extern void
	zstd_fini(void)
	{
	/* Deinitialize kstat */
	if (zstd_ksp != NULL) {
	kstat_delete(zstd_ksp);
	zstd_ksp = NULL;
	}

	/* Release fallback memory */
	vmem_free(zstd_dctx_fallback.mem, zstd_dctx_fallback.mem_size);
	mutex_destroy(&zstd_dctx_fallback.barrier);

	/* Deinit memory pool */
	zstd_mempool_deinit();
	}

	#if defined(_KERNEL)
	#ifdef __FreeBSD__
	module_init(zstd_init);
	module_exit(zstd_fini);
	#endif

	-ZFS_MODULE_PARAM(zfs, zstd_, earlyabort_pass, INT, ZMOD_RW,
	+ZFS_MODULE_PARAM(zfs, zstd_, earlyabort_pass, UINT, ZMOD_RW,
	"Enable early abort attempts when using zstd");
	ZFS_MODULE_PARAM(zfs, zstd_, abort_size, UINT, ZMOD_RW,
	"Minimal size of block to attempt early abort");
	#endif
	diff --git a/sys/contrib/openzfs/tests/runfiles/common.run b/sys/contrib/openzfs/tests/runfiles/common.run
	index e8443ffabcfa..65b64f4fa8cd 100644
	--- a/sys/contrib/openzfs/tests/runfiles/common.run
	+++ b/sys/contrib/openzfs/tests/runfiles/common.run
	@@ -1,972 +1,972 @@
	#
	# This file and its contents are supplied under the terms of the
	# Common Development and Distribution License ("CDDL"), version 1.0.
	# You may only use this file in accordance with the terms of version
	# 1.0 of the CDDL.
	#
	# A full copy of the text of the CDDL should have accompanied this
	# source. A copy of the CDDL is also available via the Internet at
	# http://www.illumos.org/license/CDDL.
	#
	# This run file contains all of the common functional tests. When
	# adding a new test consider also adding it to the sanity.run file
	# if the new test runs to completion in only a few seconds.
	#
	# Approximate run time: 4-5 hours
	#

	[DEFAULT]
	pre = setup
	quiet = False
	pre_user = root
	user = root
	timeout = 600
	post_user = root
	post = cleanup
	failsafe_user = root
	failsafe = callbacks/zfs_failsafe
	outputdir = /var/tmp/test_results
	tags = ['functional']

	[tests/functional/acl/off]
	tests = ['dosmode', 'posixmode']
	tags = ['functional', 'acl']

	[tests/functional/alloc_class]
	tests = ['alloc_class_001_pos', 'alloc_class_002_neg', 'alloc_class_003_pos',
	'alloc_class_004_pos', 'alloc_class_005_pos', 'alloc_class_006_pos',
	'alloc_class_007_pos', 'alloc_class_008_pos', 'alloc_class_009_pos',
	'alloc_class_010_pos', 'alloc_class_011_neg', 'alloc_class_012_pos',
	'alloc_class_013_pos']
	tags = ['functional', 'alloc_class']

	[tests/functional/append]
	tests = ['file_append', 'threadsappend_001_pos']
	tags = ['functional', 'append']

	[tests/functional/arc]
	tests = ['dbufstats_001_pos', 'dbufstats_002_pos', 'dbufstats_003_pos',
	'arcstats_runtime_tuning']
	tags = ['functional', 'arc']

	[tests/functional/atime]
	tests = ['atime_001_pos', 'atime_002_neg', 'root_atime_off', 'root_atime_on']
	tags = ['functional', 'atime']

	[tests/functional/bootfs]
	tests = ['bootfs_001_pos', 'bootfs_002_neg', 'bootfs_003_pos',
	'bootfs_004_neg', 'bootfs_005_neg', 'bootfs_006_pos', 'bootfs_007_pos',
	'bootfs_008_pos']
	tags = ['functional', 'bootfs']

	[tests/functional/btree]
	tests = ['btree_positive', 'btree_negative']
	tags = ['functional', 'btree']
	pre =
	post =

	[tests/functional/cache]
	tests = ['cache_001_pos', 'cache_002_pos', 'cache_003_pos', 'cache_004_neg',
	'cache_005_neg', 'cache_006_pos', 'cache_007_neg', 'cache_008_neg',
	'cache_009_pos', 'cache_010_pos', 'cache_011_pos', 'cache_012_pos']
	tags = ['functional', 'cache']

	[tests/functional/cachefile]
	tests = ['cachefile_001_pos', 'cachefile_002_pos', 'cachefile_003_pos',
	'cachefile_004_pos']
	tags = ['functional', 'cachefile']

	[tests/functional/casenorm]
	tests = ['case_all_values', 'norm_all_values', 'mixed_create_failure',
	'sensitive_none_lookup', 'sensitive_none_delete',
	'sensitive_formd_lookup', 'sensitive_formd_delete',
	'insensitive_none_lookup', 'insensitive_none_delete',
	'insensitive_formd_lookup', 'insensitive_formd_delete',
	'mixed_none_lookup', 'mixed_none_lookup_ci', 'mixed_none_delete',
	'mixed_formd_lookup', 'mixed_formd_lookup_ci', 'mixed_formd_delete']
	tags = ['functional', 'casenorm']

	[tests/functional/channel_program/lua_core]
	tests = ['tst.args_to_lua', 'tst.divide_by_zero', 'tst.exists',
	'tst.integer_illegal', 'tst.integer_overflow', 'tst.language_functions_neg',
	'tst.language_functions_pos', 'tst.large_prog', 'tst.libraries',
	'tst.memory_limit', 'tst.nested_neg', 'tst.nested_pos', 'tst.nvlist_to_lua',
	'tst.recursive_neg', 'tst.recursive_pos', 'tst.return_large',
	'tst.return_nvlist_neg', 'tst.return_nvlist_pos',
	'tst.return_recursive_table', 'tst.stack_gsub', 'tst.timeout']
	tags = ['functional', 'channel_program', 'lua_core']

	[tests/functional/channel_program/synctask_core]
	tests = ['tst.destroy_fs', 'tst.destroy_snap', 'tst.get_count_and_limit',
	'tst.get_index_props', 'tst.get_mountpoint', 'tst.get_neg',
	'tst.get_number_props', 'tst.get_string_props', 'tst.get_type',
	'tst.get_userquota', 'tst.get_written', 'tst.inherit', 'tst.list_bookmarks',
	'tst.list_children', 'tst.list_clones', 'tst.list_holds',
	'tst.list_snapshots', 'tst.list_system_props',
	'tst.list_user_props', 'tst.parse_args_neg','tst.promote_conflict',
	'tst.promote_multiple', 'tst.promote_simple', 'tst.rollback_mult',
	'tst.rollback_one', 'tst.set_props', 'tst.snapshot_destroy', 'tst.snapshot_neg',
	'tst.snapshot_recursive', 'tst.snapshot_simple',
	'tst.bookmark.create', 'tst.bookmark.copy',
	'tst.terminate_by_signal'
	]
	tags = ['functional', 'channel_program', 'synctask_core']

	[tests/functional/checksum]
	tests = ['run_edonr_test', 'run_sha2_test', 'run_skein_test', 'run_blake3_test',
	'filetest_001_pos', 'filetest_002_pos']
	tags = ['functional', 'checksum']

	[tests/functional/clean_mirror]
	tests = [ 'clean_mirror_001_pos', 'clean_mirror_002_pos',
	'clean_mirror_003_pos', 'clean_mirror_004_pos']
	tags = ['functional', 'clean_mirror']

	[tests/functional/cli_root/zdb]
	tests = ['zdb_002_pos', 'zdb_003_pos', 'zdb_004_pos', 'zdb_005_pos',
	'zdb_006_pos', 'zdb_args_neg', 'zdb_args_pos',
	'zdb_block_size_histogram', 'zdb_checksum', 'zdb_decompress',
	'zdb_display_block', 'zdb_label_checksum', 'zdb_object_range_neg',
	'zdb_object_range_pos', 'zdb_objset_id', 'zdb_decompress_zstd',
	'zdb_recover', 'zdb_recover_2']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zdb']

	[tests/functional/cli_root/zfs]
	tests = ['zfs_001_neg', 'zfs_002_pos']
	tags = ['functional', 'cli_root', 'zfs']

	[tests/functional/cli_root/zfs_bookmark]
	tests = ['zfs_bookmark_cliargs']
	tags = ['functional', 'cli_root', 'zfs_bookmark']

	[tests/functional/cli_root/zfs_change-key]
	tests = ['zfs_change-key', 'zfs_change-key_child', 'zfs_change-key_format',
	'zfs_change-key_inherit', 'zfs_change-key_load', 'zfs_change-key_location',
	'zfs_change-key_pbkdf2iters', 'zfs_change-key_clones']
	tags = ['functional', 'cli_root', 'zfs_change-key']

	[tests/functional/cli_root/zfs_clone]
	tests = ['zfs_clone_001_neg', 'zfs_clone_002_pos', 'zfs_clone_003_pos',
	'zfs_clone_004_pos', 'zfs_clone_005_pos', 'zfs_clone_006_pos',
	'zfs_clone_007_pos', 'zfs_clone_008_neg', 'zfs_clone_009_neg',
	'zfs_clone_010_pos', 'zfs_clone_encrypted', 'zfs_clone_deeply_nested']
	tags = ['functional', 'cli_root', 'zfs_clone']

	[tests/functional/cli_root/zfs_copies]
	tests = ['zfs_copies_001_pos', 'zfs_copies_002_pos', 'zfs_copies_003_pos',
	'zfs_copies_004_neg', 'zfs_copies_005_neg', 'zfs_copies_006_pos']
	tags = ['functional', 'cli_root', 'zfs_copies']

	[tests/functional/cli_root/zfs_create]
	tests = ['zfs_create_001_pos', 'zfs_create_002_pos', 'zfs_create_003_pos',
	'zfs_create_004_pos', 'zfs_create_005_pos', 'zfs_create_006_pos',
	'zfs_create_007_pos', 'zfs_create_008_neg', 'zfs_create_009_neg',
	'zfs_create_010_neg', 'zfs_create_011_pos', 'zfs_create_012_pos',
	'zfs_create_013_pos', 'zfs_create_014_pos', 'zfs_create_encrypted',
	'zfs_create_crypt_combos', 'zfs_create_dryrun', 'zfs_create_nomount',
	'zfs_create_verbose']
	tags = ['functional', 'cli_root', 'zfs_create']

	[tests/functional/cli_root/zfs_destroy]
	tests = ['zfs_clone_livelist_condense_and_disable',
	'zfs_clone_livelist_condense_races', 'zfs_clone_livelist_dedup',
	'zfs_destroy_001_pos', 'zfs_destroy_002_pos', 'zfs_destroy_003_pos',
	'zfs_destroy_004_pos', 'zfs_destroy_005_neg', 'zfs_destroy_006_neg',
	'zfs_destroy_007_neg', 'zfs_destroy_008_pos', 'zfs_destroy_009_pos',
	'zfs_destroy_010_pos', 'zfs_destroy_011_pos', 'zfs_destroy_012_pos',
	'zfs_destroy_013_neg', 'zfs_destroy_014_pos', 'zfs_destroy_015_pos',
	'zfs_destroy_016_pos', 'zfs_destroy_clone_livelist',
	'zfs_destroy_dev_removal', 'zfs_destroy_dev_removal_condense']
	tags = ['functional', 'cli_root', 'zfs_destroy']

	[tests/functional/cli_root/zfs_diff]
	tests = ['zfs_diff_changes', 'zfs_diff_cliargs', 'zfs_diff_timestamp',
	'zfs_diff_types', 'zfs_diff_encrypted', 'zfs_diff_mangle']
	tags = ['functional', 'cli_root', 'zfs_diff']

	[tests/functional/cli_root/zfs_get]
	tests = ['zfs_get_001_pos', 'zfs_get_002_pos', 'zfs_get_003_pos',
	'zfs_get_004_pos', 'zfs_get_005_neg', 'zfs_get_006_neg', 'zfs_get_007_neg',
	'zfs_get_008_pos', 'zfs_get_009_pos', 'zfs_get_010_neg']
	tags = ['functional', 'cli_root', 'zfs_get']

	[tests/functional/cli_root/zfs_ids_to_path]
	tests = ['zfs_ids_to_path_001_pos']
	tags = ['functional', 'cli_root', 'zfs_ids_to_path']

	[tests/functional/cli_root/zfs_inherit]
	tests = ['zfs_inherit_001_neg', 'zfs_inherit_002_neg', 'zfs_inherit_003_pos',
	'zfs_inherit_mountpoint']
	tags = ['functional', 'cli_root', 'zfs_inherit']

	[tests/functional/cli_root/zfs_load-key]
	tests = ['zfs_load-key', 'zfs_load-key_all', 'zfs_load-key_file',
	'zfs_load-key_https', 'zfs_load-key_location', 'zfs_load-key_noop',
	'zfs_load-key_recursive']
	tags = ['functional', 'cli_root', 'zfs_load-key']

	[tests/functional/cli_root/zfs_mount]
	tests = ['zfs_mount_001_pos', 'zfs_mount_002_pos', 'zfs_mount_003_pos',
	'zfs_mount_004_pos', 'zfs_mount_005_pos', 'zfs_mount_007_pos',
	'zfs_mount_009_neg', 'zfs_mount_010_neg', 'zfs_mount_011_neg',
	'zfs_mount_012_pos', 'zfs_mount_all_001_pos', 'zfs_mount_encrypted',
	'zfs_mount_remount', 'zfs_mount_all_fail', 'zfs_mount_all_mountpoints',
	'zfs_mount_test_race']
	tags = ['functional', 'cli_root', 'zfs_mount']

	[tests/functional/cli_root/zfs_program]
	tests = ['zfs_program_json']
	tags = ['functional', 'cli_root', 'zfs_program']

	[tests/functional/cli_root/zfs_promote]
	tests = ['zfs_promote_001_pos', 'zfs_promote_002_pos', 'zfs_promote_003_pos',
	'zfs_promote_004_pos', 'zfs_promote_005_pos', 'zfs_promote_006_neg',
	'zfs_promote_007_neg', 'zfs_promote_008_pos', 'zfs_promote_encryptionroot']
	tags = ['functional', 'cli_root', 'zfs_promote']

	[tests/functional/cli_root/zfs_property]
	tests = ['zfs_written_property_001_pos']
	tags = ['functional', 'cli_root', 'zfs_property']

	[tests/functional/cli_root/zfs_receive]
	tests = ['zfs_receive_001_pos', 'zfs_receive_002_pos', 'zfs_receive_003_pos',
	'zfs_receive_004_neg', 'zfs_receive_005_neg', 'zfs_receive_006_pos',
	'zfs_receive_007_neg', 'zfs_receive_008_pos', 'zfs_receive_009_neg',
	'zfs_receive_010_pos', 'zfs_receive_011_pos', 'zfs_receive_012_pos',
	'zfs_receive_013_pos', 'zfs_receive_014_pos', 'zfs_receive_015_pos',
	'zfs_receive_016_pos', 'receive-o-x_props_override',
	'receive-o-x_props_aliases',
	'zfs_receive_from_encrypted', 'zfs_receive_to_encrypted',
	'zfs_receive_raw', 'zfs_receive_raw_incremental', 'zfs_receive_-e',
	'zfs_receive_raw_-d', 'zfs_receive_from_zstd', 'zfs_receive_new_props',
	'zfs_receive_-wR-encrypted-mix', 'zfs_receive_corrective',
	'zfs_receive_compressed_corrective']
	tags = ['functional', 'cli_root', 'zfs_receive']

	[tests/functional/cli_root/zfs_rename]
	tests = ['zfs_rename_001_pos', 'zfs_rename_002_pos', 'zfs_rename_003_pos',
	'zfs_rename_004_neg', 'zfs_rename_005_neg', 'zfs_rename_006_pos',
	'zfs_rename_007_pos', 'zfs_rename_008_pos', 'zfs_rename_009_neg',
	'zfs_rename_010_neg', 'zfs_rename_011_pos', 'zfs_rename_012_neg',
	'zfs_rename_013_pos', 'zfs_rename_014_neg', 'zfs_rename_encrypted_child',
	'zfs_rename_to_encrypted', 'zfs_rename_mountpoint', 'zfs_rename_nounmount']
	tags = ['functional', 'cli_root', 'zfs_rename']

	[tests/functional/cli_root/zfs_reservation]
	tests = ['zfs_reservation_001_pos', 'zfs_reservation_002_pos']
	tags = ['functional', 'cli_root', 'zfs_reservation']

	[tests/functional/cli_root/zfs_rollback]
	tests = ['zfs_rollback_001_pos', 'zfs_rollback_002_pos',
	'zfs_rollback_003_neg', 'zfs_rollback_004_neg']
	tags = ['functional', 'cli_root', 'zfs_rollback']

	[tests/functional/cli_root/zfs_send]
	tests = ['zfs_send_001_pos', 'zfs_send_002_pos', 'zfs_send_003_pos',
	'zfs_send_004_neg', 'zfs_send_005_pos', 'zfs_send_006_pos',
	'zfs_send_007_pos', 'zfs_send_encrypted', 'zfs_send_raw',
	'zfs_send_sparse', 'zfs_send-b', 'zfs_send_skip_missing']
	tags = ['functional', 'cli_root', 'zfs_send']

	[tests/functional/cli_root/zfs_set]
	tests = ['cache_001_pos', 'cache_002_neg', 'canmount_001_pos',
	'canmount_002_pos', 'canmount_003_pos', 'canmount_004_pos',
	'checksum_001_pos', 'compression_001_pos', 'mountpoint_001_pos',
	'mountpoint_002_pos', 'reservation_001_neg', 'user_property_002_pos',
	'share_mount_001_neg', 'snapdir_001_pos', 'onoffs_001_pos',
	'user_property_001_pos', 'user_property_003_neg', 'readonly_001_pos',
	'user_property_004_pos', 'version_001_neg', 'zfs_set_001_neg',
	'zfs_set_002_neg', 'zfs_set_003_neg', 'property_alias_001_pos',
	'mountpoint_003_pos', 'ro_props_001_pos', 'zfs_set_keylocation',
	'zfs_set_feature_activation']
	tags = ['functional', 'cli_root', 'zfs_set']

	[tests/functional/cli_root/zfs_share]
	tests = ['zfs_share_001_pos', 'zfs_share_002_pos', 'zfs_share_003_pos',
	'zfs_share_004_pos', 'zfs_share_006_pos', 'zfs_share_008_neg',
	'zfs_share_010_neg', 'zfs_share_011_pos', 'zfs_share_concurrent_shares']
	tags = ['functional', 'cli_root', 'zfs_share']

	[tests/functional/cli_root/zfs_snapshot]
	tests = ['zfs_snapshot_001_neg', 'zfs_snapshot_002_neg',
	'zfs_snapshot_003_neg', 'zfs_snapshot_004_neg', 'zfs_snapshot_005_neg',
	'zfs_snapshot_006_pos', 'zfs_snapshot_007_neg', 'zfs_snapshot_008_neg',
	'zfs_snapshot_009_pos']
	tags = ['functional', 'cli_root', 'zfs_snapshot']

	[tests/functional/cli_root/zfs_unload-key]
	tests = ['zfs_unload-key', 'zfs_unload-key_all', 'zfs_unload-key_recursive']
	tags = ['functional', 'cli_root', 'zfs_unload-key']

	[tests/functional/cli_root/zfs_unmount]
	tests = ['zfs_unmount_001_pos', 'zfs_unmount_002_pos', 'zfs_unmount_003_pos',
	'zfs_unmount_004_pos', 'zfs_unmount_005_pos', 'zfs_unmount_006_pos',
	'zfs_unmount_007_neg', 'zfs_unmount_008_neg', 'zfs_unmount_009_pos',
	'zfs_unmount_all_001_pos', 'zfs_unmount_nested', 'zfs_unmount_unload_keys']
	tags = ['functional', 'cli_root', 'zfs_unmount']

	[tests/functional/cli_root/zfs_unshare]
	tests = ['zfs_unshare_001_pos', 'zfs_unshare_002_pos', 'zfs_unshare_003_pos',
	'zfs_unshare_004_neg', 'zfs_unshare_005_neg', 'zfs_unshare_006_pos',
	'zfs_unshare_007_pos', 'zfs_unshare_008_pos']
	tags = ['functional', 'cli_root', 'zfs_unshare']

	[tests/functional/cli_root/zfs_upgrade]
	tests = ['zfs_upgrade_001_pos', 'zfs_upgrade_002_pos', 'zfs_upgrade_003_pos',
	'zfs_upgrade_004_pos', 'zfs_upgrade_005_pos', 'zfs_upgrade_006_neg',
	'zfs_upgrade_007_neg']
	tags = ['functional', 'cli_root', 'zfs_upgrade']

	[tests/functional/cli_root/zfs_wait]
	tests = ['zfs_wait_deleteq', 'zfs_wait_getsubopt']
	tags = ['functional', 'cli_root', 'zfs_wait']

	[tests/functional/cli_root/zhack]
	tests = ['zhack_label_checksum']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zhack']

	[tests/functional/cli_root/zpool]
	tests = ['zpool_001_neg', 'zpool_002_pos', 'zpool_003_pos', 'zpool_colors']
	tags = ['functional', 'cli_root', 'zpool']

	[tests/functional/cli_root/zpool_add]
	tests = ['zpool_add_001_pos', 'zpool_add_002_pos', 'zpool_add_003_pos',
	'zpool_add_004_pos', 'zpool_add_006_pos', 'zpool_add_007_neg',
	'zpool_add_008_neg', 'zpool_add_009_neg', 'zpool_add_010_pos',
	'add-o_ashift', 'add_prop_ashift', 'zpool_add_dryrun_output']
	tags = ['functional', 'cli_root', 'zpool_add']

	[tests/functional/cli_root/zpool_attach]
	tests = ['zpool_attach_001_neg', 'attach-o_ashift']
	tags = ['functional', 'cli_root', 'zpool_attach']

	[tests/functional/cli_root/zpool_clear]
	tests = ['zpool_clear_001_pos', 'zpool_clear_002_neg', 'zpool_clear_003_neg',
	'zpool_clear_readonly']
	tags = ['functional', 'cli_root', 'zpool_clear']

	[tests/functional/cli_root/zpool_create]
	tests = ['zpool_create_001_pos', 'zpool_create_002_pos',
	'zpool_create_003_pos', 'zpool_create_004_pos', 'zpool_create_005_pos',
	'zpool_create_006_pos', 'zpool_create_007_neg', 'zpool_create_008_pos',
	'zpool_create_009_neg', 'zpool_create_010_neg', 'zpool_create_011_neg',
	'zpool_create_012_neg', 'zpool_create_014_neg', 'zpool_create_015_neg',
	'zpool_create_017_neg', 'zpool_create_018_pos', 'zpool_create_019_pos',
	'zpool_create_020_pos', 'zpool_create_021_pos', 'zpool_create_022_pos',
	'zpool_create_023_neg', 'zpool_create_024_pos',
	'zpool_create_encrypted', 'zpool_create_crypt_combos',
	'zpool_create_draid_001_pos', 'zpool_create_draid_002_pos',
	'zpool_create_draid_003_pos', 'zpool_create_draid_004_pos',
	'zpool_create_features_001_pos', 'zpool_create_features_002_pos',
	'zpool_create_features_003_pos', 'zpool_create_features_004_neg',
	'zpool_create_features_005_pos', 'zpool_create_features_006_pos',
	'zpool_create_features_007_pos', 'zpool_create_features_008_pos',
	'zpool_create_features_009_pos', 'create-o_ashift',
	'zpool_create_tempname', 'zpool_create_dryrun_output']
	tags = ['functional', 'cli_root', 'zpool_create']

	[tests/functional/cli_root/zpool_destroy]
	tests = ['zpool_destroy_001_pos', 'zpool_destroy_002_pos',
	'zpool_destroy_003_neg']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zpool_destroy']

	[tests/functional/cli_root/zpool_detach]
	tests = ['zpool_detach_001_neg']
	tags = ['functional', 'cli_root', 'zpool_detach']

	[tests/functional/cli_root/zpool_events]
	tests = ['zpool_events_clear', 'zpool_events_cliargs', 'zpool_events_follow',
	'zpool_events_poolname', 'zpool_events_errors', 'zpool_events_duplicates',
	'zpool_events_clear_retained']
	tags = ['functional', 'cli_root', 'zpool_events']

	[tests/functional/cli_root/zpool_export]
	tests = ['zpool_export_001_pos', 'zpool_export_002_pos',
	'zpool_export_003_neg', 'zpool_export_004_pos']
	tags = ['functional', 'cli_root', 'zpool_export']

	[tests/functional/cli_root/zpool_get]
	tests = ['zpool_get_001_pos', 'zpool_get_002_pos', 'zpool_get_003_pos',
	'zpool_get_004_neg', 'zpool_get_005_pos']
	tags = ['functional', 'cli_root', 'zpool_get']

	[tests/functional/cli_root/zpool_history]
	tests = ['zpool_history_001_neg', 'zpool_history_002_pos']
	tags = ['functional', 'cli_root', 'zpool_history']

	[tests/functional/cli_root/zpool_import]
	tests = ['zpool_import_001_pos', 'zpool_import_002_pos',
	'zpool_import_003_pos', 'zpool_import_004_pos', 'zpool_import_005_pos',
	'zpool_import_006_pos', 'zpool_import_007_pos', 'zpool_import_008_pos',
	'zpool_import_009_neg', 'zpool_import_010_pos', 'zpool_import_011_neg',
	'zpool_import_012_pos', 'zpool_import_013_neg', 'zpool_import_014_pos',
	'zpool_import_015_pos', 'zpool_import_016_pos', 'zpool_import_017_pos',
	'zpool_import_features_001_pos', 'zpool_import_features_002_neg',
	'zpool_import_features_003_pos', 'zpool_import_missing_001_pos',
	'zpool_import_missing_002_pos', 'zpool_import_missing_003_pos',
	'zpool_import_rename_001_pos', 'zpool_import_all_001_pos',
	'zpool_import_encrypted', 'zpool_import_encrypted_load',
	'zpool_import_errata3', 'zpool_import_errata4',
	'import_cachefile_device_added',
	'import_cachefile_device_removed',
	'import_cachefile_device_replaced',
	'import_cachefile_mirror_attached',
	'import_cachefile_mirror_detached',
	'import_cachefile_paths_changed',
	'import_cachefile_shared_device',
	'import_devices_missing',
	'import_paths_changed',
	'import_rewind_config_changed',
	'import_rewind_device_replaced']
	tags = ['functional', 'cli_root', 'zpool_import']
	timeout = 1200

	[tests/functional/cli_root/zpool_labelclear]
	tests = ['zpool_labelclear_active', 'zpool_labelclear_exported',
	'zpool_labelclear_removed', 'zpool_labelclear_valid']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zpool_labelclear']

	[tests/functional/cli_root/zpool_initialize]
	tests = ['zpool_initialize_attach_detach_add_remove',
	'zpool_initialize_fault_export_import_online',
	'zpool_initialize_import_export',
	'zpool_initialize_offline_export_import_online',
	'zpool_initialize_online_offline',
	'zpool_initialize_split',
	'zpool_initialize_start_and_cancel_neg',
	'zpool_initialize_start_and_cancel_pos',
	'zpool_initialize_suspend_resume',
	'zpool_initialize_unsupported_vdevs',
	'zpool_initialize_verify_checksums',
	'zpool_initialize_verify_initialized']
	pre =
	tags = ['functional', 'cli_root', 'zpool_initialize']

	[tests/functional/cli_root/zpool_offline]
	tests = ['zpool_offline_001_pos', 'zpool_offline_002_neg',
	'zpool_offline_003_pos']
	tags = ['functional', 'cli_root', 'zpool_offline']

	[tests/functional/cli_root/zpool_online]
	tests = ['zpool_online_001_pos', 'zpool_online_002_neg']
	tags = ['functional', 'cli_root', 'zpool_online']

	[tests/functional/cli_root/zpool_remove]
	tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos',
	'zpool_remove_003_pos']
	tags = ['functional', 'cli_root', 'zpool_remove']

	[tests/functional/cli_root/zpool_replace]
	tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift']
	tags = ['functional', 'cli_root', 'zpool_replace']

	[tests/functional/cli_root/zpool_resilver]
	tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart']
	tags = ['functional', 'cli_root', 'zpool_resilver']

	[tests/functional/cli_root/zpool_scrub]
	tests = ['zpool_scrub_001_neg', 'zpool_scrub_002_pos', 'zpool_scrub_003_pos',
	'zpool_scrub_004_pos', 'zpool_scrub_005_pos',
	'zpool_scrub_encrypted_unloaded', 'zpool_scrub_print_repairing',
	'zpool_scrub_offline_device', 'zpool_scrub_multiple_copies']
	tags = ['functional', 'cli_root', 'zpool_scrub']

	[tests/functional/cli_root/zpool_set]
	tests = ['zpool_set_001_pos', 'zpool_set_002_neg', 'zpool_set_003_neg',
	'zpool_set_ashift', 'zpool_set_features']
	tags = ['functional', 'cli_root', 'zpool_set']

	[tests/functional/cli_root/zpool_split]
	tests = ['zpool_split_cliargs', 'zpool_split_devices',
	'zpool_split_encryption', 'zpool_split_props', 'zpool_split_vdevs',
	'zpool_split_resilver', 'zpool_split_indirect',
	'zpool_split_dryrun_output']
	tags = ['functional', 'cli_root', 'zpool_split']

	[tests/functional/cli_root/zpool_status]
	tests = ['zpool_status_001_pos', 'zpool_status_002_pos',
	'zpool_status_003_pos', 'zpool_status_004_pos',
	'zpool_status_005_pos', 'zpool_status_features_001_pos']
	tags = ['functional', 'cli_root', 'zpool_status']

	[tests/functional/cli_root/zpool_sync]
	tests = ['zpool_sync_001_pos', 'zpool_sync_002_neg']
	tags = ['functional', 'cli_root', 'zpool_sync']

	[tests/functional/cli_root/zpool_trim]
	tests = ['zpool_trim_attach_detach_add_remove',
	'zpool_trim_fault_export_import_online',
	'zpool_trim_import_export', 'zpool_trim_multiple', 'zpool_trim_neg',
	'zpool_trim_offline_export_import_online', 'zpool_trim_online_offline',
	'zpool_trim_partial', 'zpool_trim_rate', 'zpool_trim_rate_neg',
	'zpool_trim_secure', 'zpool_trim_split', 'zpool_trim_start_and_cancel_neg',
	'zpool_trim_start_and_cancel_pos', 'zpool_trim_suspend_resume',
	'zpool_trim_unsupported_vdevs', 'zpool_trim_verify_checksums',
	'zpool_trim_verify_trimmed']
	tags = ['functional', 'zpool_trim']

	[tests/functional/cli_root/zpool_upgrade]
	tests = ['zpool_upgrade_001_pos', 'zpool_upgrade_002_pos',
	'zpool_upgrade_003_pos', 'zpool_upgrade_004_pos',
	'zpool_upgrade_005_neg', 'zpool_upgrade_006_neg',
	'zpool_upgrade_007_pos', 'zpool_upgrade_008_pos',
	'zpool_upgrade_009_neg', 'zpool_upgrade_features_001_pos']
	tags = ['functional', 'cli_root', 'zpool_upgrade']

	[tests/functional/cli_root/zpool_wait]
	tests = ['zpool_wait_discard', 'zpool_wait_freeing',
	'zpool_wait_initialize_basic', 'zpool_wait_initialize_cancel',
	'zpool_wait_initialize_flag', 'zpool_wait_multiple',
	'zpool_wait_no_activity', 'zpool_wait_remove', 'zpool_wait_remove_cancel',
	'zpool_wait_trim_basic', 'zpool_wait_trim_cancel', 'zpool_wait_trim_flag',
	'zpool_wait_usage']
	tags = ['functional', 'cli_root', 'zpool_wait']

	[tests/functional/cli_root/zpool_wait/scan]
	tests = ['zpool_wait_replace_cancel', 'zpool_wait_rebuild',
	'zpool_wait_resilver', 'zpool_wait_scrub_cancel',
	'zpool_wait_replace', 'zpool_wait_scrub_basic', 'zpool_wait_scrub_flag']
	tags = ['functional', 'cli_root', 'zpool_wait']

	[tests/functional/cli_user/misc]
	tests = ['zdb_001_neg', 'zfs_001_neg', 'zfs_allow_001_neg',
	'zfs_clone_001_neg', 'zfs_create_001_neg', 'zfs_destroy_001_neg',
	'zfs_get_001_neg', 'zfs_inherit_001_neg', 'zfs_mount_001_neg',
	'zfs_promote_001_neg', 'zfs_receive_001_neg', 'zfs_rename_001_neg',
	'zfs_rollback_001_neg', 'zfs_send_001_neg', 'zfs_set_001_neg',
	'zfs_share_001_neg', 'zfs_snapshot_001_neg', 'zfs_unallow_001_neg',
	'zfs_unmount_001_neg', 'zfs_unshare_001_neg', 'zfs_upgrade_001_neg',
	'zpool_001_neg', 'zpool_add_001_neg', 'zpool_attach_001_neg',
	'zpool_clear_001_neg', 'zpool_create_001_neg', 'zpool_destroy_001_neg',
	'zpool_detach_001_neg', 'zpool_export_001_neg', 'zpool_get_001_neg',
	'zpool_history_001_neg', 'zpool_import_001_neg', 'zpool_import_002_neg',
	'zpool_offline_001_neg', 'zpool_online_001_neg', 'zpool_remove_001_neg',
	'zpool_replace_001_neg', 'zpool_scrub_001_neg', 'zpool_set_001_neg',
	'zpool_status_001_neg', 'zpool_upgrade_001_neg', 'arcstat_001_pos',
	'arc_summary_001_pos', 'arc_summary_002_neg', 'zpool_wait_privilege',
	'zilstat_001_pos']
	user =
	tags = ['functional', 'cli_user', 'misc']

	[tests/functional/cli_user/zfs_list]
	tests = ['zfs_list_001_pos', 'zfs_list_002_pos', 'zfs_list_003_pos',
	'zfs_list_004_neg', 'zfs_list_005_neg', 'zfs_list_007_pos',
	'zfs_list_008_neg']
	user =
	tags = ['functional', 'cli_user', 'zfs_list']

	[tests/functional/cli_user/zpool_iostat]
	tests = ['zpool_iostat_001_neg', 'zpool_iostat_002_pos',
	'zpool_iostat_003_neg', 'zpool_iostat_004_pos',
	'zpool_iostat_005_pos', 'zpool_iostat_-c_disable',
	'zpool_iostat_-c_homedir', 'zpool_iostat_-c_searchpath']
	user =
	tags = ['functional', 'cli_user', 'zpool_iostat']

	[tests/functional/cli_user/zpool_list]
	tests = ['zpool_list_001_pos', 'zpool_list_002_neg']
	user =
	tags = ['functional', 'cli_user', 'zpool_list']

	[tests/functional/cli_user/zpool_status]
	tests = ['zpool_status_003_pos', 'zpool_status_-c_disable',
	'zpool_status_-c_homedir', 'zpool_status_-c_searchpath']
	user =
	tags = ['functional', 'cli_user', 'zpool_status']

	[tests/functional/compression]
	tests = ['compress_001_pos', 'compress_002_pos', 'compress_003_pos',
	'l2arc_compressed_arc', 'l2arc_compressed_arc_disabled',
	'l2arc_encrypted', 'l2arc_encrypted_no_compressed_arc']
	tags = ['functional', 'compression']

	[tests/functional/cp_files]
	tests = ['cp_files_001_pos']
	tags = ['functional', 'cp_files']

	[tests/functional/crtime]
	tests = ['crtime_001_pos' ]
	tags = ['functional', 'crtime']

	[tests/functional/ctime]
	tests = ['ctime_001_pos' ]
	tags = ['functional', 'ctime']

	[tests/functional/deadman]
	tests = ['deadman_ratelimit', 'deadman_sync', 'deadman_zio']
	pre =
	post =
	tags = ['functional', 'deadman']

	[tests/functional/delegate]
	tests = ['zfs_allow_001_pos', 'zfs_allow_002_pos', 'zfs_allow_003_pos',
	'zfs_allow_004_pos', 'zfs_allow_005_pos', 'zfs_allow_006_pos',
	'zfs_allow_007_pos', 'zfs_allow_008_pos', 'zfs_allow_009_neg',
	'zfs_allow_010_pos', 'zfs_allow_011_neg', 'zfs_allow_012_neg',
	'zfs_unallow_001_pos', 'zfs_unallow_002_pos', 'zfs_unallow_003_pos',
	'zfs_unallow_004_pos', 'zfs_unallow_005_pos', 'zfs_unallow_006_pos',
	'zfs_unallow_007_neg', 'zfs_unallow_008_neg']
	tags = ['functional', 'delegate']

	[tests/functional/exec]
	tests = ['exec_001_pos', 'exec_002_neg']
	tags = ['functional', 'exec']

	[tests/functional/fallocate]
	tests = ['fallocate_punch-hole']
	tags = ['functional', 'fallocate']

	[tests/functional/features/async_destroy]
	tests = ['async_destroy_001_pos']
	tags = ['functional', 'features', 'async_destroy']

	[tests/functional/features/large_dnode]
	tests = ['large_dnode_001_pos', 'large_dnode_003_pos', 'large_dnode_004_neg',
	'large_dnode_005_pos', 'large_dnode_007_neg', 'large_dnode_009_pos']
	tags = ['functional', 'features', 'large_dnode']

	[tests/functional/grow]
	pre =
	post =
	tests = ['grow_pool_001_pos', 'grow_replicas_001_pos']
	tags = ['functional', 'grow']

	[tests/functional/history]
	tests = ['history_001_pos', 'history_002_pos', 'history_003_pos',
	'history_004_pos', 'history_005_neg', 'history_006_neg',
	'history_007_pos', 'history_008_pos', 'history_009_pos',
	'history_010_pos']
	tags = ['functional', 'history']

	[tests/functional/hkdf]
	pre =
	post =
	tests = ['hkdf_test']
	tags = ['functional', 'hkdf']

	[tests/functional/inheritance]
	tests = ['inherit_001_pos']
	pre =
	tags = ['functional', 'inheritance']

	[tests/functional/io]
	tests = ['sync', 'psync', 'posixaio', 'mmap']
	tags = ['functional', 'io']

	[tests/functional/inuse]
	tests = ['inuse_004_pos', 'inuse_005_pos', 'inuse_008_pos', 'inuse_009_pos']
	post =
	tags = ['functional', 'inuse']

	[tests/functional/large_files]
	tests = ['large_files_001_pos', 'large_files_002_pos']
	tags = ['functional', 'large_files']

	[tests/functional/limits]
	tests = ['filesystem_count', 'filesystem_limit', 'snapshot_count',
	'snapshot_limit']
	tags = ['functional', 'limits']

	[tests/functional/link_count]
	tests = ['link_count_001', 'link_count_root_inode']
	tags = ['functional', 'link_count']

	[tests/functional/migration]
	tests = ['migration_001_pos', 'migration_002_pos', 'migration_003_pos',
	'migration_004_pos', 'migration_005_pos', 'migration_006_pos',
	'migration_007_pos', 'migration_008_pos', 'migration_009_pos',
	'migration_010_pos', 'migration_011_pos', 'migration_012_pos']
	tags = ['functional', 'migration']

	[tests/functional/mmap]
	tests = ['mmap_write_001_pos', 'mmap_read_001_pos', 'mmap_seek_001_pos', 'mmap_sync_001_pos']
	tags = ['functional', 'mmap']

	[tests/functional/mount]
	tests = ['umount_001', 'umountall_001']
	tags = ['functional', 'mount']

	[tests/functional/mv_files]
	tests = ['mv_files_001_pos', 'mv_files_002_pos', 'random_creation']
	tags = ['functional', 'mv_files']

	[tests/functional/nestedfs]
	tests = ['nestedfs_001_pos']
	tags = ['functional', 'nestedfs']

	[tests/functional/no_space]
	tests = ['enospc_001_pos', 'enospc_002_pos', 'enospc_003_pos',
	'enospc_df', 'enospc_rm']
	tags = ['functional', 'no_space']

	[tests/functional/nopwrite]
	tests = ['nopwrite_copies', 'nopwrite_mtime', 'nopwrite_negative',
	'nopwrite_promoted_clone', 'nopwrite_recsize', 'nopwrite_sync',
	'nopwrite_varying_compression', 'nopwrite_volume']
	tags = ['functional', 'nopwrite']

	[tests/functional/online_offline]
	tests = ['online_offline_001_pos', 'online_offline_002_neg',
	'online_offline_003_neg']
	tags = ['functional', 'online_offline']

	[tests/functional/pool_checkpoint]
	tests = ['checkpoint_after_rewind', 'checkpoint_big_rewind',
	'checkpoint_capacity', 'checkpoint_conf_change', 'checkpoint_discard',
	'checkpoint_discard_busy', 'checkpoint_discard_many',
	'checkpoint_indirect', 'checkpoint_invalid', 'checkpoint_lun_expsz',
	'checkpoint_open', 'checkpoint_removal', 'checkpoint_rewind',
	'checkpoint_ro_rewind', 'checkpoint_sm_scale', 'checkpoint_twice',
	'checkpoint_vdev_add', 'checkpoint_zdb', 'checkpoint_zhack_feat']
	tags = ['functional', 'pool_checkpoint']
	timeout = 1800

	[tests/functional/pool_names]
	tests = ['pool_names_001_pos', 'pool_names_002_neg']
	pre =
	post =
	tags = ['functional', 'pool_names']

	[tests/functional/poolversion]
	tests = ['poolversion_001_pos', 'poolversion_002_pos']
	tags = ['functional', 'poolversion']

	[tests/functional/pyzfs]
	tests = ['pyzfs_unittest']
	pre =
	post =
	tags = ['functional', 'pyzfs']

	[tests/functional/quota]
	tests = ['quota_001_pos', 'quota_002_pos', 'quota_003_pos',
	'quota_004_pos', 'quota_005_pos', 'quota_006_neg']
	tags = ['functional', 'quota']

	[tests/functional/redacted_send]
	tests = ['redacted_compressed', 'redacted_contents', 'redacted_deleted',
	'redacted_disabled_feature', 'redacted_embedded', 'redacted_holes',
	'redacted_incrementals', 'redacted_largeblocks', 'redacted_many_clones',
	'redacted_mixed_recsize', 'redacted_mounts', 'redacted_negative',
	'redacted_origin', 'redacted_panic', 'redacted_props', 'redacted_resume',
	'redacted_size', 'redacted_volume']
	tags = ['functional', 'redacted_send']

	[tests/functional/raidz]
	tests = ['raidz_001_neg', 'raidz_002_pos', 'raidz_003_pos', 'raidz_004_pos']
	tags = ['functional', 'raidz']

	[tests/functional/redundancy]
	tests = ['redundancy_draid', 'redundancy_draid1', 'redundancy_draid2',
	'redundancy_draid3', 'redundancy_draid_damaged1',
	'redundancy_draid_damaged2', 'redundancy_draid_spare1',
	'redundancy_draid_spare2', 'redundancy_draid_spare3', 'redundancy_mirror',
	'redundancy_raidz', 'redundancy_raidz1', 'redundancy_raidz2',
	'redundancy_raidz3', 'redundancy_stripe']
	tags = ['functional', 'redundancy']
	timeout = 1200

	[tests/functional/refquota]
	tests = ['refquota_001_pos', 'refquota_002_pos', 'refquota_003_pos',
	'refquota_004_pos', 'refquota_005_pos', 'refquota_006_neg',
	'refquota_007_neg', 'refquota_008_neg']
	tags = ['functional', 'refquota']

	[tests/functional/refreserv]
	tests = ['refreserv_001_pos', 'refreserv_002_pos', 'refreserv_003_pos',
	'refreserv_004_pos', 'refreserv_005_pos', 'refreserv_multi_raidz',
	'refreserv_raidz']
	tags = ['functional', 'refreserv']

	[tests/functional/removal]
	pre =
	tests = ['removal_all_vdev', 'removal_cancel', 'removal_check_space',
	'removal_condense_export', 'removal_multiple_indirection',
	'removal_nopwrite', 'removal_remap_deadlists',
	'removal_resume_export', 'removal_sanity', 'removal_with_add',
	'removal_with_create_fs', 'removal_with_dedup',
	'removal_with_errors', 'removal_with_export',
	'removal_with_ganging', 'removal_with_faulted',
	'removal_with_remove', 'removal_with_scrub', 'removal_with_send',
	'removal_with_send_recv', 'removal_with_snapshot',
	'removal_with_write', 'removal_with_zdb', 'remove_expanded',
	'remove_mirror', 'remove_mirror_sanity', 'remove_raidz',
	'remove_indirect', 'remove_attach_mirror']
	tags = ['functional', 'removal']

	[tests/functional/rename_dirs]
	tests = ['rename_dirs_001_pos']
	tags = ['functional', 'rename_dirs']

	[tests/functional/replacement]
	tests = ['attach_import', 'attach_multiple', 'attach_rebuild',
	'attach_resilver', 'detach', 'rebuild_disabled_feature',
	'rebuild_multiple', 'rebuild_raidz', 'replace_import', 'replace_rebuild',
	'replace_resilver', 'resilver_restart_001', 'resilver_restart_002',
	'scrub_cancel']
	tags = ['functional', 'replacement']

	[tests/functional/reservation]
	tests = ['reservation_001_pos', 'reservation_002_pos', 'reservation_003_pos',
	'reservation_004_pos', 'reservation_005_pos', 'reservation_006_pos',
	'reservation_007_pos', 'reservation_008_pos', 'reservation_009_pos',
	'reservation_010_pos', 'reservation_011_pos', 'reservation_012_pos',
	'reservation_013_pos', 'reservation_014_pos', 'reservation_015_pos',
	'reservation_016_pos', 'reservation_017_pos', 'reservation_018_pos',
	'reservation_019_pos', 'reservation_020_pos', 'reservation_021_neg',
	'reservation_022_pos']
	tags = ['functional', 'reservation']

	[tests/functional/rootpool]
	tests = ['rootpool_002_neg', 'rootpool_003_neg', 'rootpool_007_pos']
	tags = ['functional', 'rootpool']

	[tests/functional/rsend]
	tests = ['recv_dedup', 'recv_dedup_encrypted_zvol', 'rsend_001_pos',
	'rsend_002_pos', 'rsend_003_pos', 'rsend_004_pos', 'rsend_005_pos',
	'rsend_006_pos', 'rsend_007_pos', 'rsend_008_pos', 'rsend_009_pos',
	'rsend_010_pos', 'rsend_011_pos', 'rsend_012_pos', 'rsend_013_pos',
	'rsend_014_pos', 'rsend_016_neg', 'rsend_019_pos', 'rsend_020_pos',
	'rsend_021_pos', 'rsend_022_pos', 'rsend_024_pos', 'rsend_025_pos',
	'rsend_026_neg', 'rsend_027_pos', 'rsend_028_neg', 'rsend_029_neg',
	- 'send-c_verify_ratio', 'send-c_verify_contents', 'send-c_props',
	- 'send-c_incremental', 'send-c_volume', 'send-c_zstreamdump',
	- 'send-c_lz4_disabled', 'send-c_recv_lz4_disabled',
	+ 'rsend_030_pos', 'send-c_verify_ratio', 'send-c_verify_contents',
	+ 'send-c_props', 'send-c_incremental', 'send-c_volume',
	+ 'send-c_zstreamdump', 'send-c_lz4_disabled', 'send-c_recv_lz4_disabled',
	'send-c_mixed_compression', 'send-c_stream_size_estimate',
	'send-c_embedded_blocks', 'send-c_resume', 'send-cpL_varied_recsize',
	'send-c_recv_dedup', 'send-L_toggle', 'send_encrypted_hierarchy',
	'send_encrypted_props', 'send_encrypted_truncated_files',
	'send_freeobjects', 'send_realloc_files',
	'send_realloc_encrypted_files', 'send_spill_block', 'send_holds',
	'send_hole_birth', 'send_mixed_raw', 'send-wR_encrypted_zvol',
	'send_partial_dataset', 'send_invalid', 'send_doall',
	'send_raw_spill_block', 'send_raw_ashift']
	tags = ['functional', 'rsend']

	[tests/functional/scrub_mirror]
	tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
	'scrub_mirror_003_pos', 'scrub_mirror_004_pos']
	tags = ['functional', 'scrub_mirror']

	[tests/functional/slog]
	tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
	'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
	'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
	'slog_013_pos', 'slog_014_pos', 'slog_015_neg', 'slog_replay_fs_001',
	'slog_replay_fs_002', 'slog_replay_volume', 'slog_016_pos']
	tags = ['functional', 'slog']

	[tests/functional/snapshot]
	tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
	'rollback_003_pos', 'snapshot_001_pos', 'snapshot_002_pos',
	'snapshot_003_pos', 'snapshot_004_pos', 'snapshot_005_pos',
	'snapshot_006_pos', 'snapshot_007_pos', 'snapshot_008_pos',
	'snapshot_009_pos', 'snapshot_010_pos', 'snapshot_011_pos',
	'snapshot_012_pos', 'snapshot_013_pos', 'snapshot_014_pos',
	'snapshot_017_pos', 'snapshot_018_pos']
	tags = ['functional', 'snapshot']

	[tests/functional/snapused]
	tests = ['snapused_001_pos', 'snapused_002_pos', 'snapused_003_pos',
	'snapused_004_pos', 'snapused_005_pos']
	tags = ['functional', 'snapused']

	[tests/functional/sparse]
	tests = ['sparse_001_pos']
	tags = ['functional', 'sparse']

	[tests/functional/stat]
	tests = ['stat_001_pos']
	tags = ['functional', 'stat']

	[tests/functional/suid]
	tests = ['suid_write_to_suid', 'suid_write_to_sgid', 'suid_write_to_suid_sgid',
	'suid_write_to_none', 'suid_write_zil_replay']
	tags = ['functional', 'suid']

	[tests/functional/trim]
	tests = ['autotrim_integrity', 'autotrim_config', 'autotrim_trim_integrity',
	'trim_integrity', 'trim_config', 'trim_l2arc']
	tags = ['functional', 'trim']

	[tests/functional/truncate]
	tests = ['truncate_001_pos', 'truncate_002_pos', 'truncate_timestamps']
	tags = ['functional', 'truncate']

	[tests/functional/upgrade]
	tests = ['upgrade_userobj_001_pos', 'upgrade_readonly_pool']
	tags = ['functional', 'upgrade']

	[tests/functional/userquota]
	tests = [
	'userquota_001_pos', 'userquota_002_pos', 'userquota_003_pos',
	'userquota_004_pos', 'userquota_005_neg', 'userquota_006_pos',
	'userquota_007_pos', 'userquota_008_pos', 'userquota_009_pos',
	'userquota_010_pos', 'userquota_011_pos', 'userquota_012_neg',
	'userspace_001_pos', 'userspace_002_pos', 'userspace_encrypted',
	'userspace_send_encrypted']
	tags = ['functional', 'userquota']

	[tests/functional/vdev_zaps]
	tests = ['vdev_zaps_001_pos', 'vdev_zaps_002_pos', 'vdev_zaps_003_pos',
	'vdev_zaps_004_pos', 'vdev_zaps_005_pos', 'vdev_zaps_006_pos',
	'vdev_zaps_007_pos']
	tags = ['functional', 'vdev_zaps']

	[tests/functional/write_dirs]
	tests = ['write_dirs_001_pos', 'write_dirs_002_pos']
	tags = ['functional', 'write_dirs']

	[tests/functional/xattr]
	tests = ['xattr_001_pos', 'xattr_002_neg', 'xattr_003_neg', 'xattr_004_pos',
	'xattr_005_pos', 'xattr_006_pos', 'xattr_007_neg',
	'xattr_011_pos', 'xattr_012_pos', 'xattr_013_pos', 'xattr_compat']
	tags = ['functional', 'xattr']

	[tests/functional/zvol/zvol_ENOSPC]
	tests = ['zvol_ENOSPC_001_pos']
	tags = ['functional', 'zvol', 'zvol_ENOSPC']

	[tests/functional/zvol/zvol_cli]
	tests = ['zvol_cli_001_pos', 'zvol_cli_002_pos', 'zvol_cli_003_neg']
	tags = ['functional', 'zvol', 'zvol_cli']

	[tests/functional/zvol/zvol_misc]
	tests = ['zvol_misc_002_pos', 'zvol_misc_hierarchy', 'zvol_misc_rename_inuse',
	'zvol_misc_snapdev', 'zvol_misc_trim', 'zvol_misc_volmode', 'zvol_misc_zil']
	tags = ['functional', 'zvol', 'zvol_misc']

	[tests/functional/zvol/zvol_stress]
	tests = ['zvol_stress']
	tags = ['functional', 'zvol', 'zvol_stress']

	[tests/functional/zvol/zvol_swap]
	tests = ['zvol_swap_001_pos', 'zvol_swap_002_pos', 'zvol_swap_004_pos']
	tags = ['functional', 'zvol', 'zvol_swap']

	[tests/functional/libzfs]
	tests = ['many_fds', 'libzfs_input']
	tags = ['functional', 'libzfs']

	[tests/functional/log_spacemap]
	tests = ['log_spacemap_import_logs']
	pre =
	post =
	tags = ['functional', 'log_spacemap']

	[tests/functional/l2arc]
	tests = ['l2arc_arcstats_pos', 'l2arc_mfuonly_pos', 'l2arc_l2miss_pos',
	'persist_l2arc_001_pos', 'persist_l2arc_002_pos',
	'persist_l2arc_003_neg', 'persist_l2arc_004_pos', 'persist_l2arc_005_pos']
	tags = ['functional', 'l2arc']

	[tests/functional/zpool_influxdb]
	tests = ['zpool_influxdb']
	tags = ['functional', 'zpool_influxdb']
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/btree_test.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/btree_test.c
	index aaad4e47e464..4e2023003b0e 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/btree_test.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/btree_test.c
	@@ -1,555 +1,561 @@
	/*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*/

	/*
	* Copyright (c) 2019 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	+#include <string.h>
	#include <sys/avl.h>
	#include <sys/btree.h>
	#include <sys/time.h>
	#include <sys/resource.h>

	#define BUFSIZE 256

	int seed = 0;
	int stress_timeout = 180;
	int contents_frequency = 100;
	int tree_limit = 64 * 1024;
	boolean_t stress_only = B_FALSE;

	static void
	usage(int exit_value)
	{
	(void) fprintf(stderr, "Usage:\tbtree_test -n <test_name>\n");
	(void) fprintf(stderr, "\tbtree_test -s [-r <seed>] [-l <limit>] "
	"[-t timeout>] [-c check_contents]\n");
	(void) fprintf(stderr, "\tbtree_test [-r <seed>] [-l <limit>] "
	"[-t timeout>] [-c check_contents]\n");
	(void) fprintf(stderr, "\n With the -n option, run the named "
	"negative test. With the -s option,\n");
	(void) fprintf(stderr, " run the stress test according to the "
	"other options passed. With\n");
	(void) fprintf(stderr, " neither, run all the positive tests, "
	"including the stress test with\n");
	(void) fprintf(stderr, " the default options.\n");
	(void) fprintf(stderr, "\n Options that control the stress test\n");
	(void) fprintf(stderr, "\t-c stress iterations after which to compare "
	"tree contents [default: 100]\n");
	(void) fprintf(stderr, "\t-l the largest value to allow in the tree "
	"[default: 1M]\n");
	(void) fprintf(stderr, "\t-r random seed [default: from "
	"gettimeofday()]\n");
	(void) fprintf(stderr, "\t-t seconds to let the stress test run "
	"[default: 180]\n");
	exit(exit_value);
	}

	typedef struct int_node {
	avl_node_t node;
	uint64_t data;
	} int_node_t;

	/*
	* Utility functions
	*/

	static int
	avl_compare(const void v1, const void v2)
	{
	const int_node_t *n1 = v1;
	const int_node_t *n2 = v2;
	uint64_t a = n1->data;
	uint64_t b = n2->data;

	return (TREE_CMP(a, b));
	}

	static int
	zfs_btree_compare(const void v1, const void v2)
	{
	const uint64_t *a = v1;
	const uint64_t *b = v2;

	return (TREE_CMP(a, b));
	}

	static void
	verify_contents(avl_tree_t avl, zfs_btree_t bt)
	{
	static int count = 0;
	zfs_btree_index_t bt_idx = {0};
	int_node_t *node;
	uint64_t *data;

	boolean_t forward = count % 2 == 0 ? B_TRUE : B_FALSE;
	count++;

	ASSERT3U(avl_numnodes(avl), ==, zfs_btree_numnodes(bt));
	if (forward == B_TRUE) {
	node = avl_first(avl);
	data = zfs_btree_first(bt, &bt_idx);
	} else {
	node = avl_last(avl);
	data = zfs_btree_last(bt, &bt_idx);
	}

	while (node != NULL) {
	ASSERT3U(*data, ==, node->data);
	if (forward == B_TRUE) {
	data = zfs_btree_next(bt, &bt_idx, &bt_idx);
	node = AVL_NEXT(avl, node);
	} else {
	data = zfs_btree_prev(bt, &bt_idx, &bt_idx);
	node = AVL_PREV(avl, node);
	}
	}
	}

	static void
	verify_node(avl_tree_t avl, zfs_btree_t bt, int_node_t *node)
	{
	zfs_btree_index_t bt_idx = {0};
	zfs_btree_index_t bt_idx2 = {0};
	int_node_t *inp;
	uint64_t data = node->data;
	uint64_t *rv = NULL;

	ASSERT3U(avl_numnodes(avl), ==, zfs_btree_numnodes(bt));
	ASSERT3P((rv = (uint64_t *)zfs_btree_find(bt, &data, &bt_idx)), !=,
	NULL);
	ASSERT3S(*rv, ==, data);
	ASSERT3P(zfs_btree_get(bt, &bt_idx), !=, NULL);
	ASSERT3S(data, ==, (uint64_t )zfs_btree_get(bt, &bt_idx));

	if ((inp = AVL_NEXT(avl, node)) != NULL) {
	ASSERT3P((rv = zfs_btree_next(bt, &bt_idx, &bt_idx2)), !=,
	NULL);
	ASSERT3P(rv, ==, zfs_btree_get(bt, &bt_idx2));
	ASSERT3S(inp->data, ==, *rv);
	} else {
	ASSERT3U(data, ==, (uint64_t )zfs_btree_last(bt, &bt_idx));
	}

	if ((inp = AVL_PREV(avl, node)) != NULL) {
	ASSERT3P((rv = zfs_btree_prev(bt, &bt_idx, &bt_idx2)), !=,
	NULL);
	ASSERT3P(rv, ==, zfs_btree_get(bt, &bt_idx2));
	ASSERT3S(inp->data, ==, *rv);
	} else {
	ASSERT3U(data, ==, (uint64_t )zfs_btree_first(bt, &bt_idx));
	}
	}

	/*
	* Tests
	*/

	/* Verify that zfs_btree_find works correctly with a NULL index. */
	static int
	find_without_index(zfs_btree_t bt, char why)
	{
	u_longlong_t *p, i = 12345;

	zfs_btree_add(bt, &i);
	if ((p = (u_longlong_t *)zfs_btree_find(bt, &i, NULL)) == NULL \|\|
	*p != i) {
	- snprintf(why, BUFSIZE, "Unexpectedly found %llu\n",
	+ (void) snprintf(why, BUFSIZE, "Unexpectedly found %llu\n",
	p == NULL ? 0 : *p);
	return (1);
	}

	i++;

	if ((p = (u_longlong_t *)zfs_btree_find(bt, &i, NULL)) != NULL) {
	- snprintf(why, BUFSIZE, "Found bad value: %llu\n", *p);
	+ (void) snprintf(why, BUFSIZE, "Found bad value: %llu\n", *p);
	return (1);
	}

	return (0);
	}

	/* Verify simple insertion and removal from the tree. */
	static int
	insert_find_remove(zfs_btree_t bt, char why)
	{
	u_longlong_t *p, i = 12345;
	zfs_btree_index_t bt_idx = {0};

	/* Insert 'i' into the tree, and attempt to find it again. */
	zfs_btree_add(bt, &i);
	if ((p = (u_longlong_t *)zfs_btree_find(bt, &i, &bt_idx)) == NULL) {
	- snprintf(why, BUFSIZE, "Didn't find value in tree\n");
	+ (void) snprintf(why, BUFSIZE, "Didn't find value in tree\n");
	return (1);
	} else if (*p != i) {
	- snprintf(why, BUFSIZE, "Found (%llu) in tree\n", *p);
	+ (void) snprintf(why, BUFSIZE, "Found (%llu) in tree\n", *p);
	return (1);
	}
	ASSERT3S(zfs_btree_numnodes(bt), ==, 1);
	zfs_btree_verify(bt);

	/* Remove 'i' from the tree, and verify it is not found. */
	zfs_btree_remove(bt, &i);
	if ((p = (u_longlong_t *)zfs_btree_find(bt, &i, &bt_idx)) != NULL) {
	- snprintf(why, BUFSIZE, "Found removed value (%llu)\n", *p);
	+ (void) snprintf(why, BUFSIZE,
	+ "Found removed value (%llu)\n", *p);
	return (1);
	}
	ASSERT3S(zfs_btree_numnodes(bt), ==, 0);
	zfs_btree_verify(bt);

	return (0);
	}

	/*
	* Add a number of random entries into a btree and avl tree. Then walk them
	* backwards and forwards while emptying the tree, verifying the trees look
	* the same.
	*/
	static int
	drain_tree(zfs_btree_t bt, char why)
	{
	uint64_t *p;
	avl_tree_t avl;
	int i = 0;
	int_node_t *node;
	avl_index_t avl_idx = {0};
	zfs_btree_index_t bt_idx = {0};

	avl_create(&avl, avl_compare, sizeof (int_node_t),
	offsetof(int_node_t, node));

	/* Fill both trees with the same data */
	for (i = 0; i < 64 * 1024; i++) {
	void *ret;

	u_longlong_t randval = random();
	- node = malloc(sizeof (int_node_t));
	if ((p = (uint64_t *)zfs_btree_find(bt, &randval, &bt_idx)) !=
	NULL) {
	continue;
	}
	zfs_btree_add_idx(bt, &randval, &bt_idx);

	+ node = malloc(sizeof (int_node_t));
	+ ASSERT3P(node, !=, NULL);
	+
	node->data = randval;
	if ((ret = avl_find(&avl, node, &avl_idx)) != NULL) {
	- snprintf(why, BUFSIZE, "Found in avl: %llu\n", randval);
	+ (void) snprintf(why, BUFSIZE,
	+ "Found in avl: %llu\n", randval);
	return (1);
	}
	avl_insert(&avl, node, avl_idx);
	}

	/* Remove data from either side of the trees, comparing the data */
	while (avl_numnodes(&avl) != 0) {
	uint64_t *data;

	ASSERT3U(avl_numnodes(&avl), ==, zfs_btree_numnodes(bt));
	if (avl_numnodes(&avl) % 2 == 0) {
	node = avl_first(&avl);
	data = zfs_btree_first(bt, &bt_idx);
	} else {
	node = avl_last(&avl);
	data = zfs_btree_last(bt, &bt_idx);
	}
	ASSERT3U(node->data, ==, *data);
	zfs_btree_remove_idx(bt, &bt_idx);
	avl_remove(&avl, node);

	if (avl_numnodes(&avl) == 0) {
	break;
	}

	node = avl_first(&avl);
	ASSERT3U(node->data, ==,
	(uint64_t )zfs_btree_first(bt, NULL));
	node = avl_last(&avl);
	ASSERT3U(node->data, ==, (uint64_t )zfs_btree_last(bt, NULL));
	}
	ASSERT3S(zfs_btree_numnodes(bt), ==, 0);

	void *avl_cookie = NULL;
	while ((node = avl_destroy_nodes(&avl, &avl_cookie)) != NULL)
	free(node);
	avl_destroy(&avl);

	return (0);
	}

	/*
	* This test uses an avl and btree, and continually processes new random
	* values. Each value is either removed or inserted, depending on whether
	* or not it is found in the tree. The test periodically checks that both
	* trees have the same data and does consistency checks. This stress
	* option can also be run on its own from the command line.
	*/
	static int
	stress_tree(zfs_btree_t bt, char why)
	{
	(void) why;
	avl_tree_t avl;
	int_node_t *node;
	struct timeval tp;
	time_t t0;
	int insertions = 0, removals = 0, iterations = 0;
	u_longlong_t max = 0, min = UINT64_MAX;

	(void) gettimeofday(&tp, NULL);
	t0 = tp.tv_sec;

	avl_create(&avl, avl_compare, sizeof (int_node_t),
	offsetof(int_node_t, node));

	while (1) {
	zfs_btree_index_t bt_idx = {0};
	avl_index_t avl_idx = {0};

	uint64_t randval = random() % tree_limit;
	node = malloc(sizeof (*node));
	node->data = randval;

	max = randval > max ? randval : max;
	min = randval < min ? randval : min;

	void *ret = avl_find(&avl, node, &avl_idx);
	if (ret == NULL) {
	insertions++;
	avl_insert(&avl, node, avl_idx);
	ASSERT3P(zfs_btree_find(bt, &randval, &bt_idx), ==,
	NULL);
	zfs_btree_add_idx(bt, &randval, &bt_idx);
	verify_node(&avl, bt, node);
	} else {
	removals++;
	verify_node(&avl, bt, ret);
	zfs_btree_remove(bt, &randval);
	avl_remove(&avl, ret);
	free(ret);
	free(node);
	}

	zfs_btree_verify(bt);

	iterations++;
	if (iterations % contents_frequency == 0) {
	verify_contents(&avl, bt);
	}

	zfs_btree_verify(bt);

	(void) gettimeofday(&tp, NULL);
	if (tp.tv_sec > t0 + stress_timeout) {
	fprintf(stderr, "insertions/removals: %u/%u\nmax/min: "
	"%llu/%llu\n", insertions, removals, max, min);
	break;
	}
	}

	void *avl_cookie = NULL;
	while ((node = avl_destroy_nodes(&avl, &avl_cookie)) != NULL)
	free(node);
	avl_destroy(&avl);

	if (stress_only) {
	zfs_btree_index_t *idx = NULL;
	uint64_t *rv;

	while ((rv = zfs_btree_destroy_nodes(bt, &idx)) != NULL)
	;
	zfs_btree_verify(bt);
	}

	return (0);
	}

	/*
	* Verify inserting a duplicate value will cause a crash.
	* Note: negative test; return of 0 is a failure.
	*/
	static int
	insert_duplicate(zfs_btree_t *bt)
	{
	uint64_t *p, i = 23456;
	zfs_btree_index_t bt_idx = {0};

	if ((p = (uint64_t *)zfs_btree_find(bt, &i, &bt_idx)) != NULL) {
	fprintf(stderr, "Found value in empty tree.\n");
	return (0);
	}
	zfs_btree_add_idx(bt, &i, &bt_idx);
	if ((p = (uint64_t *)zfs_btree_find(bt, &i, &bt_idx)) == NULL) {
	fprintf(stderr, "Did not find expected value.\n");
	return (0);
	}

	/* Crash on inserting a duplicate */
	zfs_btree_add_idx(bt, &i, NULL);

	return (0);
	}

	/*
	* Verify removing a non-existent value will cause a crash.
	* Note: negative test; return of 0 is a failure.
	*/
	static int
	remove_missing(zfs_btree_t *bt)
	{
	uint64_t *p, i = 23456;
	zfs_btree_index_t bt_idx = {0};

	if ((p = (uint64_t *)zfs_btree_find(bt, &i, &bt_idx)) != NULL) {
	fprintf(stderr, "Found value in empty tree.\n");
	return (0);
	}

	/* Crash removing a nonexistent entry */
	zfs_btree_remove(bt, &i);

	return (0);
	}

	static int
	do_negative_test(zfs_btree_t bt, char test_name)
	{
	int rval = 0;
	struct rlimit rlim = {0};
	- setrlimit(RLIMIT_CORE, &rlim);
	+
	+ (void) setrlimit(RLIMIT_CORE, &rlim);

	if (strcmp(test_name, "insert_duplicate") == 0) {
	rval = insert_duplicate(bt);
	} else if (strcmp(test_name, "remove_missing") == 0) {
	rval = remove_missing(bt);
	}

	/*
	* Return 0, since callers will expect non-zero return values for
	* these tests, and we should have crashed before getting here anyway.
	*/
	(void) fprintf(stderr, "Test: %s returned %d.\n", test_name, rval);
	return (0);
	}

	typedef struct btree_test {
	const char *name;
	int (func)(zfs_btree_t , char *);
	} btree_test_t;

	static btree_test_t test_table[] = {
	{ "insert_find_remove", insert_find_remove },
	{ "find_without_index", find_without_index },
	{ "drain_tree", drain_tree },
	{ "stress_tree", stress_tree },
	{ NULL, NULL }
	};

	int
	main(int argc, char *argv[])
	{
	char *negative_test = NULL;
	int failed_tests = 0;
	struct timeval tp;
	zfs_btree_t bt;
	int c;

	while ((c = getopt(argc, argv, "c:l:n:r:st:")) != -1) {
	switch (c) {
	case 'c':
	contents_frequency = atoi(optarg);
	break;
	case 'l':
	tree_limit = atoi(optarg);
	break;
	case 'n':
	negative_test = optarg;
	break;
	case 'r':
	seed = atoi(optarg);
	break;
	case 's':
	stress_only = B_TRUE;
	break;
	case 't':
	stress_timeout = atoi(optarg);
	break;
	case 'h':
	default:
	usage(1);
	break;
	}
	}
	argc -= optind;
	argv += optind;
	optind = 1;


	if (seed == 0) {
	(void) gettimeofday(&tp, NULL);
	seed = tp.tv_sec;
	}
	srandom(seed);

	zfs_btree_init();
	zfs_btree_create(&bt, zfs_btree_compare, sizeof (uint64_t));

	/*
	* This runs the named negative test. None of them should
	* return, as they both cause crashes.
	*/
	if (negative_test) {
	return (do_negative_test(&bt, negative_test));
	}

	fprintf(stderr, "Seed: %u\n", seed);

	/*
	* This is a stress test that does operations on a btree over the
	* requested timeout period, verifying them against identical
	* operations in an avl tree.
	*/
	if (stress_only != 0) {
	return (stress_tree(&bt, NULL));
	}

	/* Do the positive tests */
	btree_test_t *test = &test_table[0];
	while (test->name) {
	int retval;
	uint64_t *rv;
	char why[BUFSIZE] = {0};
	zfs_btree_index_t *idx = NULL;

	(void) fprintf(stdout, "%-20s", test->name);
	retval = test->func(&bt, why);

	if (retval == 0) {
	(void) fprintf(stdout, "ok\n");
	} else {
	(void) fprintf(stdout, "failed with %d\n", retval);
	if (strlen(why) != 0)
	(void) fprintf(stdout, "\t%s\n", why);
	why[0] = '\0';
	failed_tests++;
	}

	/* Remove all the elements and re-verify the tree */
	while ((rv = zfs_btree_destroy_nodes(&bt, &idx)) != NULL)
	;
	zfs_btree_verify(&bt);

	test++;
	}

	zfs_btree_verify(&bt);
	zfs_btree_fini();

	return (failed_tests);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/dir_rd_update.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/dir_rd_update.c
	index 80c395cc62d8..c0d283105a92 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/dir_rd_update.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/dir_rd_update.c
	@@ -1,136 +1,135 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Assertion:
	*
	* A read operation and directory update operation performed
	* concurrently on the same directory can lead to deadlock
	* on a UFS logging file system, but not on a ZFS file system.
	*/

	#include <sys/types.h>
	#include <sys/stat.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <string.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#define TMP_DIR /tmp

	static char dirpath[256];

	int
	main(int argc, char **argv)
	{
	const char *cp1 = "";
	int i = 0;
	int ret = 0;
	int testdd = 0;
	pid_t pid;
	static const int op_num = 5;

	if (argc == 1) {
	(void) printf("Usage: %s <mount point>\n", argv[0]);
	exit(-1);
	}
	for (i = 0; i < 256; i++) {
	dirpath[i] = 0;
	}

	cp1 = argv[1];
	- if (strlen(cp1) >= (sizeof (dirpath) - strlen("TMP_DIR"))) {
	+ if (strlen(cp1) >= (sizeof (dirpath) - strlen("/TMP_DIR"))) {
	(void) printf("The string length of mount point is "
	"too large\n");
	exit(-1);
	}
	- (void) strcpy(&dirpath[0], (const char *)cp1);
	- (void) strcat(&dirpath[strlen(dirpath)], "TMP_DIR");
	+ (void) snprintf(dirpath, sizeof (dirpath), "%s/TMP_DIR", cp1);

	ret = mkdir(dirpath, 0777);
	if (ret != 0) {
	if (errno != EEXIST) {
	(void) printf("%s: mkdir(<%s>, 0777) failed: errno "
	"(decimal)=%d\n", argv[0], dirpath, errno);
	exit(-1);
	}
	}
	testdd = open(dirpath, O_RDONLY\|O_RSYNC\|O_SYNC\|O_DSYNC);
	if (testdd < 0) {
	(void) printf("%s: open(<%s>, O_RDONLY\|O_RSYNC\|O_SYNC\|O_DSYNC)"
	" failed: errno (decimal)=%d\n", argv[0], dirpath, errno);
	exit(-1);
	} else {
	(void) close(testdd);
	}
	pid = fork();
	if (pid > 0) {
	int fd = open(dirpath, O_RDONLY\|O_RSYNC\|O_SYNC\|O_DSYNC);
	char buf[16];
	int rdret;
	int j = 0;

	if (fd < 0) {
	(void) printf("%s: open <%s> again failed:"
	" errno = %d\n", argv[0], dirpath, errno);
	exit(-1);
	}

	while (j < op_num) {
	(void) sleep(1);
	rdret = read(fd, buf, 16);
	if (rdret == -1) {
	(void) printf("readdir failed");
	}
	j++;
	}
	(void) close(fd);
	} else if (pid == 0) {
	int fd = open(dirpath, O_RDONLY);
	int chownret;
	int k = 0;

	if (fd < 0) {
	(void) printf("%s: open(<%s>, O_RDONLY) again failed:"
	" errno (decimal)=%d\n", argv[0], dirpath, errno);
	exit(-1);
	}

	while (k < op_num) {
	(void) sleep(1);
	chownret = fchown(fd, 0, 0);
	if (chownret == -1) {
	(void) printf("chown failed");
	}

	k++;
	}
	(void) close(fd);
	}

	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/draid.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/draid.c
	index b995be46a373..46d7b4dcc69d 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/draid.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/draid.c
	@@ -1,1403 +1,1409 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2018 Intel Corporation.
	* Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
	*/

	#include <stdio.h>
	#include <zlib.h>
	#include <zfs_fletcher.h>
	#include <sys/vdev_draid.h>
	#include <sys/nvpair.h>
	#include <sys/stat.h>

	/*
	* The number of rows to generate for new permutation maps.
	*/
	#define MAP_ROWS_DEFAULT 256

	/*
	* Key values for dRAID maps when stored as nvlists.
	*/
	#define MAP_SEED "seed"
	#define MAP_CHECKSUM "checksum"
	#define MAP_WORST_RATIO "worst_ratio"
	#define MAP_AVG_RATIO "avg_ratio"
	#define MAP_CHILDREN "children"
	#define MAP_NPERMS "nperms"
	#define MAP_PERMS "perms"

	static void
	draid_usage(void)
	{
	(void) fprintf(stderr,
	"usage: draid command args ...\n"
	"Available commands are:\n"
	"\n"
	"\tdraid generate [-cv] [-m min] [-n max] [-p passes] FILE\n"
	"\tdraid verify [-rv] FILE\n"
	"\tdraid dump [-v] [-m min] [-n max] FILE\n"
	"\tdraid table FILE\n"
	"\tdraid merge FILE SRC SRC...\n");
	exit(1);
	}

	static int
	read_map(const char filename, nvlist_t *allcfgs)
	{
	int block_size = 131072;
	int buf_size = 131072;
	int tmp_size, error;
	char *tmp_buf;

	struct stat64 stat;
	if (lstat64(filename, &stat) != 0)
	return (errno);

	if (stat.st_size == 0 \|\|
	!(S_ISREG(stat.st_mode) \|\| S_ISLNK(stat.st_mode))) {
	return (EINVAL);
	}

	gzFile fp = gzopen(filename, "rb");
	if (fp == Z_NULL)
	return (errno);

	char *buf = malloc(buf_size);
	if (buf == NULL) {
	(void) gzclose(fp);
	return (ENOMEM);
	}

	ssize_t rc, bytes = 0;
	while (!gzeof(fp)) {
	rc = gzread(fp, buf + bytes, block_size);
	if ((rc < 0) \|\| (rc == 0 && !gzeof(fp))) {
	free(buf);
	- (void) gzclose(fp);
	(void) gzerror(fp, &error);
	+ (void) gzclose(fp);
	return (error);
	} else {
	bytes += rc;

	if (bytes + block_size >= buf_size) {
	tmp_size = 2 * buf_size;
	tmp_buf = malloc(tmp_size);
	if (tmp_buf == NULL) {
	free(buf);
	(void) gzclose(fp);
	return (ENOMEM);
	}

	memcpy(tmp_buf, buf, bytes);
	free(buf);
	buf = tmp_buf;
	buf_size = tmp_size;
	}
	}
	}

	(void) gzclose(fp);

	error = nvlist_unpack(buf, bytes, allcfgs, 0);
	free(buf);

	return (error);
	}

	/*
	* Read a map from the specified filename. A file contains multiple maps
	* which are indexed by the number of children. The caller is responsible
	* for freeing the configuration returned.
	*/
	static int
	read_map_key(const char filename, const char key, nvlist_t **cfg)
	{
	nvlist_t allcfgs, foundcfg = NULL;
	int error;

	error = read_map(filename, &allcfgs);
	if (error != 0)
	return (error);

	- nvlist_lookup_nvlist(allcfgs, key, &foundcfg);
	+ (void) nvlist_lookup_nvlist(allcfgs, key, &foundcfg);
	if (foundcfg != NULL) {
	nvlist_dup(foundcfg, cfg, KM_SLEEP);
	error = 0;
	} else {
	error = ENOENT;
	}

	nvlist_free(allcfgs);

	return (error);
	}

	/*
	* Write all mappings to the map file.
	*/
	static int
	write_map(const char filename, nvlist_t allcfgs)
	{
	size_t buflen = 0;
	int error;

	error = nvlist_size(allcfgs, &buflen, NV_ENCODE_XDR);
	if (error)
	return (error);

	char *buf = malloc(buflen);
	if (buf == NULL)
	return (ENOMEM);

	error = nvlist_pack(allcfgs, &buf, &buflen, NV_ENCODE_XDR, KM_SLEEP);
	if (error) {
	free(buf);
	return (error);
	}

	/*
	* Atomically update the file using a temporary file and the
	* traditional unlink then rename steps. This code provides
	* no locking, it only guarantees the packed nvlist on disk
	* is updated atomically and is internally consistent.
	*/
	char *tmpname = calloc(1, MAXPATHLEN);
	if (tmpname == NULL) {
	free(buf);
	return (ENOMEM);
	}

	snprintf(tmpname, MAXPATHLEN - 1, "%s.XXXXXX", filename);

	int fd = mkstemp(tmpname);
	if (fd < 0) {
	error = errno;
	free(buf);
	free(tmpname);
	return (error);
	}
	(void) close(fd);

	gzFile fp = gzopen(tmpname, "w9b");
	if (fp == Z_NULL) {
	error = errno;
	free(buf);
	free(tmpname);
	return (errno);
	}

	ssize_t rc, bytes = 0;
	while (bytes < buflen) {
	size_t size = MIN(buflen - bytes, 131072);
	rc = gzwrite(fp, buf + bytes, size);
	if (rc < 0) {
	free(buf);
	(void) gzerror(fp, &error);
	(void) gzclose(fp);
	(void) unlink(tmpname);
	free(tmpname);
	return (error);
	} else if (rc == 0) {
	break;
	} else {
	bytes += rc;
	}
	}

	free(buf);
	(void) gzclose(fp);

	if (bytes != buflen) {
	(void) unlink(tmpname);
	free(tmpname);
	return (EIO);
	}

	/*
	* Unlink the previous config file and replace it with the updated
	* version. If we're able to unlink the file then directory is
	* writable by us and the subsequent rename should never fail.
	*/
	error = unlink(filename);
	if (error != 0 && errno != ENOENT) {
	error = errno;
	(void) unlink(tmpname);
	free(tmpname);
	return (error);
	}

	error = rename(tmpname, filename);
	if (error != 0) {
	error = errno;
	(void) unlink(tmpname);
	free(tmpname);
	return (error);
	}

	free(tmpname);

	return (0);
	}

	/*
	* Add the dRAID map to the file and write it out.
	*/
	static int
	write_map_key(const char filename, char key, draid_map_t *map,
	double worst_ratio, double avg_ratio)
	{
	nvlist_t nv_cfg, allcfgs;
	int error;

	/*
	* Add the configuration to an existing or new file. The new
	* configuration will replace an existing configuration with the
	* same key if it has a lower ratio and is therefore better.
	*/
	error = read_map(filename, &allcfgs);
	if (error == ENOENT) {
	allcfgs = fnvlist_alloc();
	} else if (error != 0) {
	return (error);
	}

	error = nvlist_lookup_nvlist(allcfgs, key, &nv_cfg);
	if (error == 0) {
	uint64_t nv_cfg_worst_ratio = fnvlist_lookup_uint64(nv_cfg,
	MAP_WORST_RATIO);
	double nv_worst_ratio = (double)nv_cfg_worst_ratio / 1000.0;

	if (worst_ratio < nv_worst_ratio) {
	/* Replace old map with the more balanced new map. */
	fnvlist_remove(allcfgs, key);
	} else {
	/* The old map is preferable, keep it. */
	nvlist_free(allcfgs);
	return (EEXIST);
	}
	}

	nvlist_t *cfg = fnvlist_alloc();
	fnvlist_add_uint64(cfg, MAP_SEED, map->dm_seed);
	fnvlist_add_uint64(cfg, MAP_CHECKSUM, map->dm_checksum);
	fnvlist_add_uint64(cfg, MAP_CHILDREN, map->dm_children);
	fnvlist_add_uint64(cfg, MAP_NPERMS, map->dm_nperms);
	fnvlist_add_uint8_array(cfg, MAP_PERMS, map->dm_perms,
	map->dm_children * map->dm_nperms * sizeof (uint8_t));

	fnvlist_add_uint64(cfg, MAP_WORST_RATIO,
	(uint64_t)(worst_ratio * 1000.0));
	fnvlist_add_uint64(cfg, MAP_AVG_RATIO,
	(uint64_t)(avg_ratio * 1000.0));

	error = nvlist_add_nvlist(allcfgs, key, cfg);
	if (error == 0)
	error = write_map(filename, allcfgs);

	nvlist_free(cfg);
	nvlist_free(allcfgs);
	return (error);
	}

	static void
	dump_map(draid_map_t map, const char key, double worst_ratio,
	double avg_ratio, int verbose)
	{
	if (verbose == 0) {
	return;
	} else if (verbose == 1) {
	printf(" \"%s\": seed: 0x%016llx worst_ratio: %2.03f "
	"avg_ratio: %2.03f\n", key, (u_longlong_t)map->dm_seed,
	worst_ratio, avg_ratio);
	return;
	} else {
	printf(" \"%s\":\n"
	" seed: 0x%016llx\n"
	" checksum: 0x%016llx\n"
	" worst_ratio: %2.03f\n"
	" avg_ratio: %2.03f\n"
	" children: %llu\n"
	" nperms: %llu\n",
	key, (u_longlong_t)map->dm_seed,
	(u_longlong_t)map->dm_checksum, worst_ratio, avg_ratio,
	(u_longlong_t)map->dm_children,
	(u_longlong_t)map->dm_nperms);

	if (verbose > 2) {
	printf(" perms = {\n");
	for (int i = 0; i < map->dm_nperms; i++) {
	printf(" { ");
	for (int j = 0; j < map->dm_children; j++) {
	printf("%3d%s ", map->dm_perms[
	i * map->dm_children + j],
	j < map->dm_children - 1 ?
	"," : "");
	}
	printf(" },\n");
	}
	printf(" }\n");
	} else if (verbose == 2) {
	printf(" draid_perms = <omitted>\n");
	}
	}
	}

	static void
	dump_map_nv(const char key, nvlist_t cfg, int verbose)
	{
	draid_map_t map;
	uint_t c;

	uint64_t worst_ratio = fnvlist_lookup_uint64(cfg, MAP_WORST_RATIO);
	uint64_t avg_ratio = fnvlist_lookup_uint64(cfg, MAP_AVG_RATIO);

	map.dm_seed = fnvlist_lookup_uint64(cfg, MAP_SEED);
	map.dm_checksum = fnvlist_lookup_uint64(cfg, MAP_CHECKSUM);
	map.dm_children = fnvlist_lookup_uint64(cfg, MAP_CHILDREN);
	map.dm_nperms = fnvlist_lookup_uint64(cfg, MAP_NPERMS);
	- nvlist_lookup_uint8_array(cfg, MAP_PERMS, &map.dm_perms, &c);
	+ map.dm_perms = fnvlist_lookup_uint8_array(cfg, MAP_PERMS, &c);

	dump_map(&map, key, (double)worst_ratio / 1000.0,
	avg_ratio / 1000.0, verbose);
	}

	/*
	* Print a summary of the mapping.
	*/
	static int
	dump_map_key(const char filename, const char key, int verbose)
	{
	nvlist_t *cfg;
	int error;

	error = read_map_key(filename, key, &cfg);
	if (error != 0)
	return (error);

	dump_map_nv(key, cfg, verbose);

	return (0);
	}

	/*
	* Allocate a new permutation map for evaluation.
	*/
	static int
	alloc_new_map(uint64_t children, uint64_t nperms, uint64_t seed,
	draid_map_t **mapp)
	{
	draid_map_t *map;
	int error;

	map = malloc(sizeof (draid_map_t));
	if (map == NULL)
	return (ENOMEM);

	map->dm_children = children;
	map->dm_nperms = nperms;
	map->dm_seed = seed;
	map->dm_checksum = 0;

	error = vdev_draid_generate_perms(map, &map->dm_perms);
	if (error) {
	free(map);
	return (error);
	}

	*mapp = map;

	return (0);
	}

	/*
	* Allocate the fixed permutation map for N children.
	*/
	static int
	alloc_fixed_map(uint64_t children, draid_map_t **mapp)
	{
	const draid_map_t *fixed_map;
	draid_map_t *map;
	int error;

	error = vdev_draid_lookup_map(children, &fixed_map);
	if (error)
	return (error);

	map = malloc(sizeof (draid_map_t));
	if (map == NULL)
	return (ENOMEM);

	memcpy(map, fixed_map, sizeof (draid_map_t));
	VERIFY3U(map->dm_checksum, !=, 0);

	error = vdev_draid_generate_perms(map, &map->dm_perms);
	if (error) {
	free(map);
	return (error);
	}

	*mapp = map;

	return (0);
	}

	/*
	* Free a permutation map.
	*/
	static void
	free_map(draid_map_t *map)
	{
	free(map->dm_perms);
	free(map);
	}

	/*
	* Check if dev is in the provided list of faulted devices.
	*/
	static inline boolean_t
	is_faulted(int *faulted_devs, int nfaulted, int dev)
	{
	for (int i = 0; i < nfaulted; i++)
	if (faulted_devs[i] == dev)
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Evaluate how resilvering I/O will be distributed given a list of faulted
	* vdevs. As a simplification we assume one IO is sufficient to repair each
	* damaged device in a group.
	*/
	static double
	eval_resilver(draid_map_t *map, uint64_t groupwidth, uint64_t nspares,
	int faulted_devs, int nfaulted, int min_child_ios, int *max_child_ios)
	{
	uint64_t children = map->dm_children;
	uint64_t ngroups = 1;
	uint64_t ndisks = children - nspares;

	/*
	* Calculate the minimum number of groups required to fill a slice.
	*/
	while (ngroups * (groupwidth) % (children - nspares) != 0)
	ngroups++;

	int *ios = calloc(map->dm_children, sizeof (uint64_t));

	+ ASSERT3P(ios, !=, NULL);
	+
	/* Resilver all rows */
	for (int i = 0; i < map->dm_nperms; i++) {
	uint8_t row = &map->dm_perms[i map->dm_children];

	/* Resilver all groups with faulted drives */
	for (int j = 0; j < ngroups; j++) {
	uint64_t spareidx = map->dm_children - nspares;
	boolean_t repair_needed = B_FALSE;

	/* See if any devices in this group are faulted */
	uint64_t groupstart = (j * groupwidth) % ndisks;

	for (int k = 0; k < groupwidth; k++) {
	uint64_t groupidx = (groupstart + k) % ndisks;

	repair_needed = is_faulted(faulted_devs,
	nfaulted, row[groupidx]);
	if (repair_needed)
	break;
	}

	if (repair_needed == B_FALSE)
	continue;

	/*
	* This group is degraded. Calculate the number of
	* reads the non-faulted drives require and the number
	* of writes to the distributed hot spare for this row.
	*/
	for (int k = 0; k < groupwidth; k++) {
	uint64_t groupidx = (groupstart + k) % ndisks;

	if (!is_faulted(faulted_devs, nfaulted,
	row[groupidx])) {
	ios[row[groupidx]]++;
	} else if (nspares > 0) {
	while (is_faulted(faulted_devs,
	nfaulted, row[spareidx])) {
	spareidx++;
	}

	ASSERT3U(spareidx, <, map->dm_children);
	ios[row[spareidx]]++;
	spareidx++;
	}
	}
	}
	}

	*min_child_ios = INT_MAX;
	*max_child_ios = 0;

	/*
	* Find the drives with fewest and most required I/O. These values
	* are used to calculate the imbalance ratio. To avoid returning an
	* infinite value for permutations which have children that perform
	* no IO a floor of 1 IO per child is set. This ensures a meaningful
	* ratio is returned for comparison and it is not an uncommon when
	* there are a large number of children.
	*/
	for (int i = 0; i < map->dm_children; i++) {

	if (is_faulted(faulted_devs, nfaulted, i)) {
	ASSERT0(ios[i]);
	continue;
	}

	if (ios[i] == 0)
	ios[i] = 1;

	if (ios[i] < *min_child_ios)
	*min_child_ios = ios[i];

	if (ios[i] > *max_child_ios)
	*max_child_ios = ios[i];
	}

	ASSERT3S(*min_child_ios, !=, INT_MAX);
	ASSERT3S(*max_child_ios, !=, 0);

	double ratio = (double)(max_child_ios) / (double)(min_child_ios);

	free(ios);

	return (ratio);
	}

	/*
	* Evaluate the quality of the permutation mapping by considering possible
	* device failures. Returns the imbalance ratio for the worst mapping which
	* is defined to be the largest number of child IOs over the fewest number
	* child IOs. A value of 1.0 indicates the mapping is perfectly balance and
	* all children perform an equal amount of work during reconstruction.
	*/
	static void
	eval_decluster(draid_map_t map, double worst_ratiop, double *avg_ratiop)
	{
	uint64_t children = map->dm_children;
	double worst_ratio = 1.0;
	double sum = 0;
	int worst_min_ios = 0, worst_max_ios = 0;
	int n = 0;

	/*
	* When there are only 2 children there can be no distributed
	* spare and no resilver to evaluate. Default to a ratio of 1.0
	* for this degenerate case.
	*/
	if (children == VDEV_DRAID_MIN_CHILDREN) {
	*worst_ratiop = 1.0;
	*avg_ratiop = 1.0;
	return;
	}

	/*
	* Score the mapping as if it had either 1 or 2 distributed spares.
	*/
	for (int nspares = 1; nspares <= 2; nspares++) {
	uint64_t faults = nspares;

	/*
	* Score groupwidths up to 19. This value was chosen as the
	* largest reasonable width (16d+3p). dRAID pools may be still
	* be created with wider stripes but they are not considered in
	* this analysis in order to optimize for the most common cases.
	*/
	for (uint64_t groupwidth = 2;
	groupwidth <= MIN(children - nspares, 19);
	groupwidth++) {
	int faulted_devs[2];
	int min_ios, max_ios;

	/*
	* Score possible devices faults. This is limited
	* to exactly one fault per distributed spare for
	* the purposes of this similation.
	*/
	for (int f1 = 0; f1 < children; f1++) {
	faulted_devs[0] = f1;
	double ratio;

	if (faults == 1) {
	ratio = eval_resilver(map, groupwidth,
	nspares, faulted_devs, faults,
	&min_ios, &max_ios);

	if (ratio > worst_ratio) {
	worst_ratio = ratio;
	worst_min_ios = min_ios;
	worst_max_ios = max_ios;
	}

	sum += ratio;
	n++;
	} else if (faults == 2) {
	for (int f2 = f1 + 1; f2 < children;
	f2++) {
	faulted_devs[1] = f2;

	ratio = eval_resilver(map,
	groupwidth, nspares,
	faulted_devs, faults,
	&min_ios, &max_ios);

	if (ratio > worst_ratio) {
	worst_ratio = ratio;
	worst_min_ios = min_ios;
	worst_max_ios = max_ios;
	}

	sum += ratio;
	n++;
	}
	}
	}
	}
	}

	*worst_ratiop = worst_ratio;
	*avg_ratiop = sum / n;

	/*
	* Log the min/max io values for particularly unbalanced maps.
	* Since the maps are generated entirely randomly these are possible
	* be exceedingly unlikely. We log it for possible investigation.
	*/
	if (worst_ratio > 100.0) {
	dump_map(map, "DEBUG", worst_ratio, *avg_ratiop, 2);
	printf("worst_min_ios=%d worst_max_ios=%d\n",
	worst_min_ios, worst_max_ios);
	}
	}

	static int
	eval_maps(uint64_t children, int passes, uint64_t *map_seed,
	draid_map_t *best_mapp, double best_ratiop, double *avg_ratiop)
	{
	draid_map_t *best_map = NULL;
	double best_worst_ratio = 1000.0;
	double best_avg_ratio = 1000.0;

	/*
	* Perform the requested number of passes evaluating randomly
	* generated permutation maps. Only the best version is kept.
	*/
	for (int i = 0; i < passes; i++) {
	double worst_ratio, avg_ratio;
	draid_map_t *map;
	int error;

	/*
	* Calculate the next seed and generate a new candidate map.
	*/
	error = alloc_new_map(children, MAP_ROWS_DEFAULT,
	vdev_draid_rand(map_seed), &map);
	- if (error)
	+ if (error) {
	+ if (best_map != NULL)
	+ free_map(best_map);
	return (error);
	+ }

	/*
	* Consider maps with a lower worst_ratio to be of higher
	* quality. Some maps may have a lower avg_ratio but they
	* are discarded since they might include some particularly
	* imbalanced permutations. The average is tracked to in
	* order to get a sense of the average permutation quality.
	*/
	eval_decluster(map, &worst_ratio, &avg_ratio);

	if (best_map == NULL \|\| worst_ratio < best_worst_ratio) {

	if (best_map != NULL)
	free_map(best_map);

	best_map = map;
	best_worst_ratio = worst_ratio;
	best_avg_ratio = avg_ratio;
	} else {
	free_map(map);
	}
	}

	/*
	* After determining the best map generate a checksum over the full
	* permutation array. This checksum is verified when opening a dRAID
	* pool to ensure the generated in memory permutations are correct.
	*/
	zio_cksum_t cksum;
	fletcher_4_native_varsize(best_map->dm_perms,
	sizeof (uint8_t) * best_map->dm_children * best_map->dm_nperms,
	&cksum);
	best_map->dm_checksum = cksum.zc_word[0];

	*best_mapp = best_map;
	*best_ratiop = best_worst_ratio;
	*avg_ratiop = best_avg_ratio;

	return (0);
	}

	static int
	draid_generate(int argc, char *argv[])
	{
	char filename[MAXPATHLEN] = {0};
	- uint64_t map_seed;
	+ uint64_t map_seed[2];
	int c, fd, error, verbose = 0, passes = 1, continuous = 0;
	int min_children = VDEV_DRAID_MIN_CHILDREN;
	int max_children = VDEV_DRAID_MAX_CHILDREN;
	int restarts = 0;

	while ((c = getopt(argc, argv, ":cm:n:p:v")) != -1) {
	switch (c) {
	case 'c':
	continuous++;
	break;
	case 'm':
	min_children = (int)strtol(optarg, NULL, 0);
	if (min_children < VDEV_DRAID_MIN_CHILDREN) {
	(void) fprintf(stderr, "A minimum of 2 "
	"children are required.\n");
	return (1);
	}

	break;
	case 'n':
	max_children = (int)strtol(optarg, NULL, 0);
	if (max_children > VDEV_DRAID_MAX_CHILDREN) {
	(void) fprintf(stderr, "A maximum of %d "
	"children are allowed.\n",
	VDEV_DRAID_MAX_CHILDREN);
	return (1);
	}
	break;
	case 'p':
	passes = (int)strtol(optarg, NULL, 0);
	break;
	case 'v':
	/*
	* 0 - Only log when a better map is added to the file.
	* 1 - Log the current best map for each child count.
	* Minimal output on a single summary line.
	* 2 - Log the current best map for each child count.
	* More verbose includes most map fields.
	* 3 - Log the current best map for each child count.
	* Very verbose all fields including the full map.
	*/
	verbose++;
	break;
	case ':':
	(void) fprintf(stderr,
	"missing argument for '%c' option\n", optopt);
	draid_usage();
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	draid_usage();
	break;
	}
	}

	if (argc > optind)
	- strncpy(filename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(filename, argv[optind], sizeof (filename));
	else {
	(void) fprintf(stderr, "A FILE must be specified.\n");
	return (1);
	}

	restart:
	/*
	* Start with a fresh seed from /dev/urandom.
	*/
	fd = open("/dev/urandom", O_RDONLY);
	if (fd < 0) {
	printf("Unable to open /dev/urandom: %s\n:", strerror(errno));
	return (1);
	} else {
	ssize_t bytes = sizeof (map_seed);
	ssize_t bytes_read = 0;

	while (bytes_read < bytes) {
	- ssize_t rc = read(fd, ((char *)&map_seed) + bytes_read,
	+ ssize_t rc = read(fd, ((char *)map_seed) + bytes_read,
	bytes - bytes_read);
	if (rc < 0) {
	printf("Unable to read /dev/urandom: %s\n:",
	strerror(errno));
	+ close(fd);
	return (1);
	}
	bytes_read += rc;
	}

	(void) close(fd);
	}

	if (restarts == 0)
	printf("Writing generated mappings to '%s':\n", filename);

	/*
	* Generate maps for all requested child counts. The best map for
	* each child count is written out to the specified file. If the file
	* already contains a better mapping this map will not be added.
	*/
	for (uint64_t children = min_children;
	children <= max_children; children++) {
	char key[8] = { 0 };
	draid_map_t *map;
	double worst_ratio = 1000.0;
	double avg_ratio = 1000.0;

	- error = eval_maps(children, passes, &map_seed, &map,
	+ error = eval_maps(children, passes, map_seed, &map,
	&worst_ratio, &avg_ratio);
	if (error) {
	printf("Error eval_maps(): %s\n", strerror(error));
	return (1);
	}

	if (worst_ratio < 1.0 \|\| avg_ratio < 1.0) {
	printf("Error ratio < 1.0: worst_ratio = %2.03f "
	"avg_ratio = %2.03f\n", worst_ratio, avg_ratio);
	return (1);
	}

	snprintf(key, 7, "%llu", (u_longlong_t)children);
	error = write_map_key(filename, key, map, worst_ratio,
	avg_ratio);
	if (error == 0) {
	/* The new map was added to the file. */
	dump_map(map, key, worst_ratio, avg_ratio,
	MAX(verbose, 1));
	} else if (error == EEXIST) {
	/* The existing map was preferable and kept. */
	if (verbose > 0)
	dump_map_key(filename, key, verbose);
	} else {
	printf("Error write_map_key(): %s\n", strerror(error));
	return (1);
	}

	free_map(map);
	}

	/*
	* When the continuous option is set restart at the minimum number of
	* children instead of exiting. This option is useful as a mechanism
	* to continuous try and refine the discovered permutations.
	*/
	if (continuous) {
	restarts++;
	printf("Restarting by request (-c): %d\n", restarts);
	goto restart;
	}

	return (0);
	}

	/*
	* Verify each map in the file by generating its in-memory permutation array
	* and comfirming its checksum is correct.
	*/
	static int
	draid_verify(int argc, char *argv[])
	{
	char filename[MAXPATHLEN] = {0};
	int n = 0, c, error, verbose = 1;
	int check_ratios = 0;

	while ((c = getopt(argc, argv, ":rv")) != -1) {
	switch (c) {
	case 'r':
	check_ratios++;
	break;
	case 'v':
	verbose++;
	break;
	case ':':
	(void) fprintf(stderr,
	"missing argument for '%c' option\n", optopt);
	draid_usage();
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	draid_usage();
	break;
	}
	}

	if (argc > optind) {
	char *abspath = malloc(MAXPATHLEN);
	if (abspath == NULL)
	return (ENOMEM);

	if (realpath(argv[optind], abspath) != NULL)
	- strncpy(filename, abspath, MAXPATHLEN - 1);
	+ strlcpy(filename, abspath, sizeof (filename));
	else
	- strncpy(filename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(filename, argv[optind], sizeof (filename));

	free(abspath);
	} else {
	(void) fprintf(stderr, "A FILE must be specified.\n");
	return (1);
	}

	printf("Verifying permutation maps: '%s'\n", filename);

	/*
	* Lookup hardcoded permutation map for each valid number of children
	* and verify a generated map has the correct checksum. Then compare
	* the generated map values with the nvlist map values read from the
	* reference file to cross-check the permutation.
	*/
	for (uint64_t children = VDEV_DRAID_MIN_CHILDREN;
	children <= VDEV_DRAID_MAX_CHILDREN;
	children++) {
	draid_map_t *map;
	char key[8] = {0};

	snprintf(key, 8, "%llu", (u_longlong_t)children);

	error = alloc_fixed_map(children, &map);
	if (error) {
	printf("Error alloc_fixed_map() failed: %s\n",
	error == ECKSUM ? "Invalid checksum" :
	strerror(error));
	return (1);
	}

	uint64_t nv_seed, nv_checksum, nv_children, nv_nperms;
	uint8_t *nv_perms;
	nvlist_t *cfg;
	uint_t c;

	error = read_map_key(filename, key, &cfg);
	if (error != 0) {
	printf("Error read_map_key() failed: %s\n",
	strerror(error));
	free_map(map);
	return (1);
	}

	nv_seed = fnvlist_lookup_uint64(cfg, MAP_SEED);
	nv_checksum = fnvlist_lookup_uint64(cfg, MAP_CHECKSUM);
	nv_children = fnvlist_lookup_uint64(cfg, MAP_CHILDREN);
	nv_nperms = fnvlist_lookup_uint64(cfg, MAP_NPERMS);
	nvlist_lookup_uint8_array(cfg, MAP_PERMS, &nv_perms, &c);

	/*
	* Compare draid_map_t and nvlist reference values.
	*/
	if (map->dm_seed != nv_seed) {
	printf("Error different seeds: 0x%016llx != "
	"0x%016llx\n", (u_longlong_t)map->dm_seed,
	(u_longlong_t)nv_seed);
	error = EINVAL;
	}

	if (map->dm_checksum != nv_checksum) {
	printf("Error different checksums: 0x%016llx "
	"!= 0x%016llx\n",
	(u_longlong_t)map->dm_checksum,
	(u_longlong_t)nv_checksum);
	error = EINVAL;
	}

	if (map->dm_children != nv_children) {
	printf("Error different children: %llu "
	"!= %llu\n", (u_longlong_t)map->dm_children,
	(u_longlong_t)nv_children);
	error = EINVAL;
	}

	if (map->dm_nperms != nv_nperms) {
	printf("Error different nperms: %llu "
	"!= %llu\n", (u_longlong_t)map->dm_nperms,
	(u_longlong_t)nv_nperms);
	error = EINVAL;
	}

	for (uint64_t i = 0; i < nv_children * nv_nperms; i++) {
	if (map->dm_perms[i] != nv_perms[i]) {
	printf("Error different perms[%llu]: "
	"%d != %d\n", (u_longlong_t)i,
	(int)map->dm_perms[i],
	(int)nv_perms[i]);
	error = EINVAL;
	break;
	}
	}

	/*
	* For good measure recalculate the worst and average
	* ratios and confirm they match the nvlist values.
	*/
	if (check_ratios) {
	uint64_t nv_worst_ratio, nv_avg_ratio;
	double worst_ratio, avg_ratio;

	eval_decluster(map, &worst_ratio, &avg_ratio);

	nv_worst_ratio = fnvlist_lookup_uint64(cfg,
	MAP_WORST_RATIO);
	nv_avg_ratio = fnvlist_lookup_uint64(cfg,
	MAP_AVG_RATIO);

	if (worst_ratio < 1.0 \|\| avg_ratio < 1.0) {
	printf("Error ratio out of range %2.03f, "
	"%2.03f\n", worst_ratio, avg_ratio);
	error = EINVAL;
	}

	if ((uint64_t)(worst_ratio * 1000.0) !=
	nv_worst_ratio) {
	printf("Error different worst_ratio %2.03f "
	"!= %2.03f\n", (double)nv_worst_ratio /
	1000.0, worst_ratio);
	error = EINVAL;
	}

	if ((uint64_t)(avg_ratio * 1000.0) != nv_avg_ratio) {
	printf("Error different average_ratio %2.03f "
	"!= %2.03f\n", (double)nv_avg_ratio /
	1000.0, avg_ratio);
	error = EINVAL;
	}
	}

	if (error) {
	free_map(map);
	nvlist_free(cfg);
	return (1);
	}

	if (verbose > 0) {
	printf("- %llu children: good\n",
	(u_longlong_t)children);
	}
	n++;

	free_map(map);
	nvlist_free(cfg);
	}

	if (n != (VDEV_DRAID_MAX_CHILDREN - 1)) {
	printf("Error permutation maps missing: %d / %d checked\n",
	n, VDEV_DRAID_MAX_CHILDREN - 1);
	return (1);
	}

	printf("Successfully verified %d / %d permutation maps\n",
	n, VDEV_DRAID_MAX_CHILDREN - 1);

	return (0);
	}

	/*
	* Dump the contents of the specified mapping(s) for inspection.
	*/
	static int
	draid_dump(int argc, char *argv[])
	{
	char filename[MAXPATHLEN] = {0};
	int c, error, verbose = 1;
	int min_children = VDEV_DRAID_MIN_CHILDREN;
	int max_children = VDEV_DRAID_MAX_CHILDREN;

	while ((c = getopt(argc, argv, ":vm:n:")) != -1) {
	switch (c) {
	case 'm':
	min_children = (int)strtol(optarg, NULL, 0);
	if (min_children < 2) {
	(void) fprintf(stderr, "A minimum of 2 "
	"children are required.\n");
	return (1);
	}

	break;
	case 'n':
	max_children = (int)strtol(optarg, NULL, 0);
	if (max_children > VDEV_DRAID_MAX_CHILDREN) {
	(void) fprintf(stderr, "A maximum of %d "
	"children are allowed.\n",
	VDEV_DRAID_MAX_CHILDREN);
	return (1);
	}
	break;
	case 'v':
	verbose++;
	break;
	case ':':
	(void) fprintf(stderr,
	"missing argument for '%c' option\n", optopt);
	draid_usage();
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	draid_usage();
	break;
	}
	}

	if (argc > optind)
	- strncpy(filename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(filename, argv[optind], sizeof (filename));
	else {
	(void) fprintf(stderr, "A FILE must be specified.\n");
	return (1);
	}

	/*
	* Dump maps for the requested child counts.
	*/
	for (uint64_t children = min_children;
	children <= max_children; children++) {
	char key[8] = { 0 };

	snprintf(key, 7, "%llu", (u_longlong_t)children);
	error = dump_map_key(filename, key, verbose);
	if (error) {
	printf("Error dump_map_key(): %s\n", strerror(error));
	return (1);
	}
	}

	return (0);
	}

	/*
	* Print all of the mappings as a C formatted draid_map_t array. This table
	* is found in the module/zcommon/zfs_draid.c file and is the definitive
	* source for all mapping used by dRAID. It cannot be updated without
	* changing the dRAID on disk format.
	*/
	static int
	draid_table(int argc, char *argv[])
	{
	char filename[MAXPATHLEN] = {0};
	int error;

	if (argc > optind)
	- strncpy(filename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(filename, argv[optind], sizeof (filename));
	else {
	(void) fprintf(stderr, "A FILE must be specified.\n");
	return (1);
	}

	printf("static const draid_map_t "
	"draid_maps[VDEV_DRAID_MAX_MAPS] = {\n");

	for (uint64_t children = VDEV_DRAID_MIN_CHILDREN;
	children <= VDEV_DRAID_MAX_CHILDREN;
	children++) {
	uint64_t seed, checksum, nperms, avg_ratio;
	nvlist_t *cfg;
	char key[8] = {0};

	snprintf(key, 8, "%llu", (u_longlong_t)children);

	error = read_map_key(filename, key, &cfg);
	if (error != 0) {
	printf("Error read_map_key() failed: %s\n",
	strerror(error));
	return (1);
	}

	seed = fnvlist_lookup_uint64(cfg, MAP_SEED);
	checksum = fnvlist_lookup_uint64(cfg, MAP_CHECKSUM);
	children = fnvlist_lookup_uint64(cfg, MAP_CHILDREN);
	nperms = fnvlist_lookup_uint64(cfg, MAP_NPERMS);
	avg_ratio = fnvlist_lookup_uint64(cfg, MAP_AVG_RATIO);

	printf("\t{ %3llu, %3llu, 0x%016llx, 0x%016llx },\t"
	"/* %2.03f */\n", (u_longlong_t)children,
	(u_longlong_t)nperms, (u_longlong_t)seed,
	(u_longlong_t)checksum, (double)avg_ratio / 1000.0);

	nvlist_free(cfg);
	}

	printf("};\n");

	return (0);
	}

	static int
	draid_merge_impl(nvlist_t allcfgs, const char srcfilename, int *mergedp)
	{
	nvlist_t *srccfgs;
	nvpair_t *elem = NULL;
	int error, merged = 0;

	error = read_map(srcfilename, &srccfgs);
	if (error != 0)
	return (error);

	while ((elem = nvlist_next_nvpair(srccfgs, elem)) != NULL) {
	uint64_t nv_worst_ratio;
	uint64_t allcfg_worst_ratio;
	nvlist_t cfg, allcfg;
	char *key;

	switch (nvpair_type(elem)) {
	case DATA_TYPE_NVLIST:

	(void) nvpair_value_nvlist(elem, &cfg);
	key = nvpair_name(elem);

	nv_worst_ratio = fnvlist_lookup_uint64(cfg,
	MAP_WORST_RATIO);

	error = nvlist_lookup_nvlist(allcfgs, key, &allcfg);
	if (error == 0) {
	allcfg_worst_ratio = fnvlist_lookup_uint64(
	allcfg, MAP_WORST_RATIO);

	if (nv_worst_ratio < allcfg_worst_ratio) {
	fnvlist_remove(allcfgs, key);
	error = nvlist_add_nvlist(allcfgs,
	key, cfg);
	merged++;
	}
	} else if (error == ENOENT) {
	error = nvlist_add_nvlist(allcfgs, key, cfg);
	merged++;
	} else {
	return (error);
	}

	break;
	default:
	continue;
	}
	}

	nvlist_free(srccfgs);

	*mergedp = merged;

	return (0);
	}

	/*
	* Merge the best map for each child count found in the listed files into
	* a new file. This allows 'draid generate' to be run in parallel and for
	* the results maps to be combined.
	*/
	static int
	draid_merge(int argc, char *argv[])
	{
	char filename[MAXPATHLEN] = {0};
	int c, error, total_merged = 0;
	nvlist_t *allcfgs;

	while ((c = getopt(argc, argv, ":")) != -1) {
	switch (c) {
	case ':':
	(void) fprintf(stderr,
	"missing argument for '%c' option\n", optopt);
	draid_usage();
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	draid_usage();
	break;
	}
	}

	if (argc < 4) {
	(void) fprintf(stderr,
	"A FILE and multiple SRCs must be specified.\n");
	return (1);
	}

	- strncpy(filename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(filename, argv[optind], sizeof (filename));
	optind++;

	error = read_map(filename, &allcfgs);
	if (error == ENOENT) {
	allcfgs = fnvlist_alloc();
	} else if (error != 0) {
	printf("Error read_map(): %s\n", strerror(error));
	return (error);
	}

	while (optind < argc) {
	char srcfilename[MAXPATHLEN] = {0};
	int merged = 0;

	- strncpy(srcfilename, argv[optind], MAXPATHLEN - 1);
	+ strlcpy(srcfilename, argv[optind], sizeof (srcfilename));

	error = draid_merge_impl(allcfgs, srcfilename, &merged);
	if (error) {
	printf("Error draid_merge_impl(): %s\n",
	strerror(error));
	nvlist_free(allcfgs);
	return (1);
	}

	total_merged += merged;
	printf("Merged %d key(s) from '%s' into '%s'\n", merged,
	srcfilename, filename);

	optind++;
	}

	if (total_merged > 0)
	write_map(filename, allcfgs);

	printf("Merged a total of %d key(s) into '%s'\n", total_merged,
	filename);

	nvlist_free(allcfgs);

	return (0);
	}

	int
	main(int argc, char *argv[])
	{
	if (argc < 2)
	draid_usage();

	char *subcommand = argv[1];

	if (strcmp(subcommand, "generate") == 0) {
	return (draid_generate(argc - 1, argv + 1));
	} else if (strcmp(subcommand, "verify") == 0) {
	return (draid_verify(argc - 1, argv + 1));
	} else if (strcmp(subcommand, "dump") == 0) {
	return (draid_dump(argc - 1, argv + 1));
	} else if (strcmp(subcommand, "table") == 0) {
	return (draid_table(argc - 1, argv + 1));
	} else if (strcmp(subcommand, "merge") == 0) {
	return (draid_merge(argc - 1, argv + 1));
	} else {
	draid_usage();
	}
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_fadvise.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_fadvise.c
	index e1afb6d0a11c..d64e2dea3696 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_fadvise.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_fadvise.c
	@@ -1,97 +1,101 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Copyright (c) 2022 by Information2 Software, Inc. All rights reserved.
	*/

	#include "file_common.h"
	#include <sys/types.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <string.h>

	/*
	* Call fadvise to prefetch data
	*/
	static const char *execname = "file_fadvise";

	static void
	usage(void)
	{
	(void) fprintf(stderr,
	"usage: %s -f filename -a advise \n", execname);
	}

	int
	main(int argc, char *argv[])
	{
	char *filename = NULL;
	int advise = 0;
	int fd, ch;
	int err = 0;

	while ((ch = getopt(argc, argv, "a:f:")) != EOF) {
	switch (ch) {
	case 'a':
	advise = atoll(optarg);
	break;
	case 'f':
	filename = optarg;
	break;
	case '?':
	(void) printf("unknown arg %c\n", optopt);
	usage();
	break;
	}
	}

	if (!filename) {
	(void) printf("Filename not specified (-f <file>)\n");
	err++;
	}

	if (advise < POSIX_FADV_NORMAL \|\| advise > POSIX_FADV_NOREUSE) {
	(void) printf("advise is invalid\n");
	err++;
	}

	if (err) {
	usage(); /* no return */
	return (1);
	}

	if ((fd = open(filename, O_RDWR, 0666)) < 0) {
	perror("open");
	return (1);
	}

	- posix_fadvise(fd, 0, 0, advise);
	+ if (posix_fadvise(fd, 0, 0, advise) != 0) {
	+ perror("posix_fadvise");
	+ close(fd);
	+ return (1);
	+ }

	close(fd);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_trunc.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_trunc.c
	index 2085f5955ea1..9e3bba24afdf 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_trunc.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/file_trunc.c
	@@ -1,240 +1,240 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <limits.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <sys/types.h>
	#include <sys/stat.h>
	#include <sys/statvfs.h>
	#include <sys/time.h>
	#include <sys/ioctl.h>
	#include <sys/wait.h>
	#include <sys/param.h>
	#include <string.h>
	#include <time.h>
	#include <inttypes.h>

	#define FSIZE 25610241024
	#define BSIZE 512

	/* Initialize Globals */
	static long fsize = FSIZE;
	static size_t bsize = BSIZE;
	static int count = 0;
	static int rflag = 0;
	-static int seed = 0;
	+static uint_t seed = 0;
	static int vflag = 0;
	static int errflag = 0;
	static off_t offset = 0;
	static char *filename = NULL;

	static void usage(char *execname);
	static void parse_options(int argc, char *argv[]);
	static void do_write(int fd);
	static void do_trunc(int fd);

	static void
	usage(char *execname)
	{
	(void) fprintf(stderr,
	"usage: %s [-b blocksize] [-c count] [-f filesize]"
	" [-o offset] [-s seed] [-r] [-v] filename\n", execname);
	(void) exit(1);
	}

	int
	main(int argc, char *argv[])
	{
	int i = 0;
	int fd = -1;

	parse_options(argc, argv);

	fd = open(filename, O_RDWR\|O_CREAT\|O_TRUNC, 0666);
	if (fd < 0) {
	perror("open");
	exit(3);
	}

	for (i = 0; count == 0 \|\| i < count; i++) {
	(void) do_write(fd);
	(void) do_trunc(fd);
	}

	(void) close(fd);
	return (0);
	}

	static void
	parse_options(int argc, char *argv[])
	{
	int c;

	extern char *optarg;
	extern int optind, optopt;

	count = fsize / bsize;
	- seed = time(NULL);
	+ seed = (uint_t)time(NULL);
	while ((c = getopt(argc, argv, "b:c:f:o:rs:v")) != -1) {
	switch (c) {
	case 'b':
	bsize = atoi(optarg);
	break;

	case 'c':
	count = atoi(optarg);
	break;

	case 'f':
	fsize = atoi(optarg);
	break;

	case 'o':
	offset = atoi(optarg);
	break;

	case 'r':
	rflag++;
	break;

	case 's':
	seed = atoi(optarg);
	break;

	case 'v':
	vflag++;
	break;

	case ':':
	(void) fprintf(stderr,
	"Option -%c requires an operand\n", optopt);
	errflag++;
	break;

	case '?':
	(void) fprintf(stderr,
	"Unrecognized option: -%c\n", optopt);
	errflag++;
	break;
	}

	if (errflag) {
	(void) usage(argv[0]);
	}
	}
	if (argc <= optind) {
	(void) fprintf(stderr,
	"No filename specified\n");
	usage(argv[0]);
	}
	filename = argv[optind];

	if (vflag) {
	(void) fprintf(stderr, "Seed = %d\n", seed);
	}
	srandom(seed);
	}

	static void
	do_write(int fd)
	{
	off_t roffset = 0;
	char *buf = NULL;
	char *rbuf = NULL;

	buf = (char *)calloc(1, bsize);
	rbuf = (char *)calloc(1, bsize);
	if (buf == NULL \|\| rbuf == NULL) {
	perror("malloc");
	exit(4);
	}

	roffset = random() % fsize;
	if (lseek64(fd, (offset + roffset), SEEK_SET) < 0) {
	perror("lseek");
	exit(5);
	}

	(void) strcpy(buf, "ZFS Test Suite Truncation Test");
	if (write(fd, buf, bsize) < bsize) {
	perror("write");
	exit(6);
	}

	if (rflag) {
	if (lseek64(fd, (offset + roffset), SEEK_SET) < 0) {
	perror("lseek");
	exit(7);
	}

	if (read(fd, rbuf, bsize) < bsize) {
	perror("read");
	exit(8);
	}

	if (memcmp(buf, rbuf, bsize) != 0) {
	perror("memcmp");
	exit(9);
	}
	}
	if (vflag) {
	(void) fprintf(stderr,
	"Wrote to offset %" PRId64 "\n", (offset + roffset));
	if (rflag) {
	(void) fprintf(stderr,
	"Read back from offset %" PRId64 "\n",
	(offset + roffset));
	}
	}

	(void) free(buf);
	(void) free(rbuf);
	}

	static void
	do_trunc(int fd)
	{
	off_t roffset = 0;

	roffset = random() % fsize;
	if (ftruncate64(fd, (offset + roffset)) < 0) {
	perror("truncate");
	exit(7);
	}

	if (vflag) {
	(void) fprintf(stderr, "Truncated at offset %" PRId64 "\n",
	(offset + roffset));
	}
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/randwritecomp.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/randwritecomp.c
	index cc70d1212f84..2d5c0ec8643c 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/file/randwritecomp.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/file/randwritecomp.c
	@@ -1,187 +1,192 @@
	/*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*/

	/*
	* Copyright (c) 2017 by Delphix. All rights reserved.
	*/

	#include <stdint.h>
	#include <string.h>
	#include "file_common.h"

	/*
	* The following sample was derived from real-world data
	* of a production Oracle database.
	*/
	static const uint64_t size_distribution[] = {
	0,
	1499018,
	352084,
	1503485,
	4206227,
	5626657,
	5387001,
	3733756,
	2233094,
	874652,
	238635,
	81434,
	33357,
	13106,
	2009,
	1,
	23660,
	};


	static uint64_t distribution_n;

	static uint8_t randbuf[BLOCKSZ];

	static void
	rwc_pwrite(int fd, const void *buf, size_t nbytes, off_t offset)
	{
	size_t nleft = nbytes;
	ssize_t nwrite = 0;

	nwrite = pwrite(fd, buf, nbytes, offset);
	if (nwrite < 0) {
	perror("pwrite");
	exit(EXIT_FAILURE);
	}

	nleft -= nwrite;
	if (nleft != 0) {
	(void) fprintf(stderr, "warning: pwrite: "
	"wrote %zu out of %zu bytes\n",
	(nbytes - nleft), nbytes);
	}
	}

	static void
	fillbuf(char *buf)
	{
	uint64_t rv = lrand48() % distribution_n;
	uint64_t sum = 0;

	uint64_t i;
	for (i = 0;
	i < sizeof (size_distribution) / sizeof (size_distribution[0]);
	i++) {
	sum += size_distribution[i];
	if (rv < sum)
	break;
	}

	memcpy(buf, randbuf, BLOCKSZ);
	if (i == 0)
	memset(buf, 0, BLOCKSZ - 10);
	else if (i < 16)
	memset(buf, 0, BLOCKSZ - i * 512 + 256);
	/LINTED: E_BAD_PTR_CAST_ALIGN/
	((uint32_t *)buf)[0] = lrand48();
	}

	static void
	exit_usage(void)
	{
	(void) puts("usage: randwritecomp [-s] file [nwrites]");
	exit(EXIT_FAILURE);
	}

	static void
	sequential_writes(int fd, char *buf, uint64_t nblocks, int64_t n)
	{
	for (int64_t i = 0; n == -1 \|\| i < n; i++) {
	fillbuf(buf);

	static uint64_t j = 0;
	if (j == 0)
	j = lrand48() % nblocks;
	rwc_pwrite(fd, buf, BLOCKSZ, j * BLOCKSZ);
	j++;
	if (j >= nblocks)
	j = 0;
	}
	}

	static void
	random_writes(int fd, char *buf, uint64_t nblocks, int64_t n)
	{
	for (int64_t i = 0; n == -1 \|\| i < n; i++) {
	fillbuf(buf);
	rwc_pwrite(fd, buf, BLOCKSZ, (lrand48() % nblocks) * BLOCKSZ);
	}
	}

	int
	main(int argc, char *argv[])
	{
	int fd, err;
	char *filename = NULL;
	char buf[BLOCKSZ];
	struct stat ss;
	uint64_t nblocks;
	int64_t n = -1;
	int sequential = 0;

	if (argc < 2)
	exit_usage();

	argv++;
	if (strcmp("-s", argv[0]) == 0) {
	sequential = 1;
	argv++;
	}

	if (argv[0] == NULL)
	exit_usage();
	else
	filename = argv[0];

	argv++;
	if (argv[0] != NULL)
	n = strtoull(argv[0], NULL, 0);

	fd = open(filename, O_RDWR\|O_CREAT, 0666);
	+ if (fd == -1) {
	+ (void) fprintf(stderr, "open(%s) failed: %s\n", filename,
	+ strerror(errno));
	+ exit(EXIT_FAILURE);
	+ }
	err = fstat(fd, &ss);
	if (err != 0) {
	(void) fprintf(stderr,
	"error: fstat returned error code %d\n", err);
	exit(EXIT_FAILURE);
	}

	nblocks = ss.st_size / BLOCKSZ;
	if (nblocks == 0) {
	(void) fprintf(stderr, "error: "
	"file is too small (min allowed size is %d bytes)\n",
	BLOCKSZ);
	exit(EXIT_FAILURE);
	}

	srand48(getpid());
	for (int i = 0; i < BLOCKSZ; i++)
	randbuf[i] = lrand48();

	distribution_n = 0;
	for (uint64_t i = 0;
	i < sizeof (size_distribution) / sizeof (size_distribution[0]);
	i++) {
	distribution_n += size_distribution[i];
	}

	if (sequential)
	sequential_writes(fd, buf, nblocks, n);
	else
	random_writes(fd, buf, nblocks, n);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/get_diff.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/get_diff.c
	index 3f8fe787f7b9..61467f25014c 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/get_diff.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/get_diff.c
	@@ -1,108 +1,108 @@
	/*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*/

	/*
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <fcntl.h>
	#include <unistd.h>
	#include <stdlib.h>
	#include <string.h>
	#include <assert.h>
	#include <sys/param.h>
	#include <sys/types.h>
	#include <sys/stat.h>
	#include <errno.h>

	static void
	usage(const char *msg, int exit_value)
	{
	(void) fprintf(stderr, "usage: get_diff file redacted_file\n%s\n", msg);
	exit(exit_value);
	}

	/*
	* This utility compares two files, an original and its redacted counterpart
	* (in that order). It compares the files 512 bytes at a time, printing out
	* any ranges (as offset and length) where the redacted file does not match
	* the original. This output is used to verify that the expected ranges of
	* a redacted file do not contain the original data.
	*/
	int
	main(int argc, char *argv[])
	{
	off_t diff_off = 0, diff_len = 0, off = 0;
	int fd1, fd2;
	char fname1, fname2;
	char buf1[DEV_BSIZE], buf2[DEV_BSIZE];
	ssize_t bytes;

	if (argc != 3)
	usage("Incorrect number of arguments.", 1);

	if ((fname1 = argv[1]) == NULL)
	usage("Filename missing.", 1);
	if ((fd1 = open(fname1, O_LARGEFILE \| O_RDONLY)) < 0) {
	perror("open1 failed");
	exit(1);
	}

	if ((fname2 = argv[2]) == NULL)
	usage("Redacted filename missing.", 1);
	if ((fd2 = open(fname2, O_LARGEFILE \| O_RDONLY)) < 0) {
	perror("open2 failed");
	exit(1);
	}

	while ((bytes = pread(fd1, buf1, DEV_BSIZE, off)) > 0) {
	if (pread(fd2, buf2, DEV_BSIZE, off) < 0) {
	if (errno == EIO) {
	/*
	* A read in a redacted section of a file will
	* fail with EIO. If we get EIO, continue on
	* but ensure that a comparison of buf1 and
	* buf2 will fail, indicating a redacted block.
	*/
	buf2[0] = ~buf1[0];
	} else {
	perror("pread failed");
	exit(1);
	}
	}
	if (memcmp(buf1, buf2, bytes) == 0) {
	if (diff_len != 0) {
	(void) fprintf(stdout, "%lld,%lld\n",
	(long long)diff_off, (long long)diff_len);
	assert(off == diff_off + diff_len);
	diff_len = 0;
	}
	diff_off = 0;
	} else {
	if (diff_len == 0)
	diff_off = off;
	assert(off == diff_off + diff_len);
	diff_len += bytes;
	}
	off += bytes;
	}

	- if (diff_len != 0 && diff_len != 0) {
	+ if (diff_len != 0) {
	(void) fprintf(stdout, "%lld,%lld\n", (long long)diff_off,
	(long long)diff_len);
	}

	(void) close(fd1);
	(void) close(fd2);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	index 434cc863f36c..2e1859b1eef0 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	@@ -1,1063 +1,1067 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <libzfs_core.h>
	#include <libzutil.h>

	#include <sys/nvpair.h>
	#include <sys/vdev_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_bootenv.h>

	/*
	* Test the nvpair inputs for the non-legacy zfs ioctl commands.
	*/

	static boolean_t unexpected_failures;
	static int zfs_fd;
	static const char *active_test;

	/*
	* Tracks which zfs_ioc_t commands were tested
	*/
	static boolean_t ioc_tested[ZFS_IOC_LAST - ZFS_IOC_FIRST];

	/*
	* Legacy ioctls that are skipped (for now)
	*/
	static const zfs_ioc_t ioc_skip[] = {
	ZFS_IOC_POOL_CREATE,
	ZFS_IOC_POOL_DESTROY,
	ZFS_IOC_POOL_IMPORT,
	ZFS_IOC_POOL_EXPORT,
	ZFS_IOC_POOL_CONFIGS,
	ZFS_IOC_POOL_STATS,
	ZFS_IOC_POOL_TRYIMPORT,
	ZFS_IOC_POOL_SCAN,
	ZFS_IOC_POOL_FREEZE,
	ZFS_IOC_POOL_UPGRADE,
	ZFS_IOC_POOL_GET_HISTORY,

	ZFS_IOC_VDEV_ADD,
	ZFS_IOC_VDEV_REMOVE,
	ZFS_IOC_VDEV_SET_STATE,
	ZFS_IOC_VDEV_ATTACH,
	ZFS_IOC_VDEV_DETACH,
	ZFS_IOC_VDEV_SETPATH,
	ZFS_IOC_VDEV_SETFRU,

	ZFS_IOC_OBJSET_STATS,
	ZFS_IOC_OBJSET_ZPLPROPS,
	ZFS_IOC_DATASET_LIST_NEXT,
	ZFS_IOC_SNAPSHOT_LIST_NEXT,
	ZFS_IOC_SET_PROP,
	ZFS_IOC_DESTROY,
	ZFS_IOC_RENAME,
	ZFS_IOC_RECV,
	ZFS_IOC_SEND,
	ZFS_IOC_INJECT_FAULT,
	ZFS_IOC_CLEAR_FAULT,
	ZFS_IOC_INJECT_LIST_NEXT,
	ZFS_IOC_ERROR_LOG,
	ZFS_IOC_CLEAR,
	ZFS_IOC_PROMOTE,
	ZFS_IOC_DSOBJ_TO_DSNAME,
	ZFS_IOC_OBJ_TO_PATH,
	ZFS_IOC_POOL_SET_PROPS,
	ZFS_IOC_POOL_GET_PROPS,
	ZFS_IOC_SET_FSACL,
	ZFS_IOC_GET_FSACL,
	ZFS_IOC_SHARE,
	ZFS_IOC_INHERIT_PROP,
	ZFS_IOC_SMB_ACL,
	ZFS_IOC_USERSPACE_ONE,
	ZFS_IOC_USERSPACE_MANY,
	ZFS_IOC_USERSPACE_UPGRADE,
	ZFS_IOC_OBJSET_RECVD_PROPS,
	ZFS_IOC_VDEV_SPLIT,
	ZFS_IOC_NEXT_OBJ,
	ZFS_IOC_DIFF,
	ZFS_IOC_TMP_SNAPSHOT,
	ZFS_IOC_OBJ_TO_STATS,
	ZFS_IOC_SPACE_WRITTEN,
	ZFS_IOC_POOL_REGUID,
	ZFS_IOC_SEND_PROGRESS,
	ZFS_IOC_EVENTS_NEXT,
	ZFS_IOC_EVENTS_CLEAR,
	ZFS_IOC_EVENTS_SEEK,
	ZFS_IOC_NEXTBOOT,
	ZFS_IOC_JAIL,
	ZFS_IOC_UNJAIL,
	};


	#define IOC_INPUT_TEST(ioc, name, req, opt, err) \
	IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, B_FALSE)

	#define IOC_INPUT_TEST_WILD(ioc, name, req, opt, err) \
	IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, B_TRUE)

	#define IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, wild) \
	do { \
	active_test = __func__ + 5; \
	ioc_tested[ioc - ZFS_IOC_FIRST] = B_TRUE; \
	lzc_ioctl_test(ioc, name, req, opt, err, wild); \
	} while (0)

	/*
	* run a zfs ioctl command, verify expected results and log failures
	*/
	static void
	lzc_ioctl_run(zfs_ioc_t ioc, const char name, nvlist_t innvl, int expected)
	{
	zfs_cmd_t zc = {"\0"};
	char *packed = NULL;
	const char *variant;
	size_t size = 0;
	int error = 0;

	switch (expected) {
	case ZFS_ERR_IOC_ARG_UNAVAIL:
	variant = "unsupported input";
	break;
	case ZFS_ERR_IOC_ARG_REQUIRED:
	variant = "missing input";
	break;
	case ZFS_ERR_IOC_ARG_BADTYPE:
	variant = "invalid input type";
	break;
	default:
	variant = "valid input";
	break;
	}

	packed = fnvlist_pack(innvl, &size);
	(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';
	zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
	zc.zc_nvlist_src_size = size;
	zc.zc_nvlist_dst_size = MAX(size * 2, 128 * 1024);
	zc.zc_nvlist_dst = (uint64_t)(uintptr_t)malloc(zc.zc_nvlist_dst_size);

	if (lzc_ioctl_fd(zfs_fd, ioc, &zc) != 0)
	error = errno;

	if (error != expected) {
	unexpected_failures = B_TRUE;
	(void) fprintf(stderr, "%s: Unexpected result with %s, "
	"error %d (expecting %d)\n",
	active_test, variant, error, expected);
	}

	fnvlist_pack_free(packed, size);
	free((void *)(uintptr_t)zc.zc_nvlist_dst);
	}

	/*
	* Test each ioc for the following ioctl input errors:
	* ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
	* ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
	* ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
	*/
	static int
	lzc_ioctl_test(zfs_ioc_t ioc, const char name, nvlist_t required,
	nvlist_t *optional, int expected_error, boolean_t wildcard)
	{
	nvlist_t *input = fnvlist_alloc();
	nvlist_t *future = fnvlist_alloc();
	int error = 0;

	if (required != NULL) {
	for (nvpair_t *pair = nvlist_next_nvpair(required, NULL);
	pair != NULL; pair = nvlist_next_nvpair(required, pair)) {
	fnvlist_add_nvpair(input, pair);
	}
	}
	if (optional != NULL) {
	for (nvpair_t *pair = nvlist_next_nvpair(optional, NULL);
	pair != NULL; pair = nvlist_next_nvpair(optional, pair)) {
	fnvlist_add_nvpair(input, pair);
	}
	}

	/*
	* Generic input run with 'optional' nvlist pair
	*/
	if (!wildcard)
	fnvlist_add_nvlist(input, "optional", future);
	lzc_ioctl_run(ioc, name, input, expected_error);
	if (!wildcard)
	fnvlist_remove(input, "optional");

	/*
	* Bogus input value
	*/
	if (!wildcard) {
	fnvlist_add_string(input, "bogus_input", "bogus");
	lzc_ioctl_run(ioc, name, input, ZFS_ERR_IOC_ARG_UNAVAIL);
	fnvlist_remove(input, "bogus_input");
	}

	/*
	* Missing required inputs
	*/
	if (required != NULL) {
	nvlist_t *empty = fnvlist_alloc();
	lzc_ioctl_run(ioc, name, empty, ZFS_ERR_IOC_ARG_REQUIRED);
	nvlist_free(empty);
	}

	/*
	* Wrong nvpair type
	*/
	if (required != NULL \|\| optional != NULL) {
	/*
	* switch the type of one of the input pairs
	*/
	for (nvpair_t *pair = nvlist_next_nvpair(input, NULL);
	pair != NULL; pair = nvlist_next_nvpair(input, pair)) {
	char pname[MAXNAMELEN];
	data_type_t ptype;

	strlcpy(pname, nvpair_name(pair), sizeof (pname));
	pname[sizeof (pname) - 1] = '\0';
	ptype = nvpair_type(pair);
	fnvlist_remove_nvpair(input, pair);

	switch (ptype) {
	case DATA_TYPE_STRING:
	fnvlist_add_uint64(input, pname, 42);
	break;
	default:
	fnvlist_add_string(input, pname, "bogus");
	break;
	}
	}
	lzc_ioctl_run(ioc, name, input, ZFS_ERR_IOC_ARG_BADTYPE);
	}

	nvlist_free(future);
	nvlist_free(input);

	return (error);
	}

	static void
	test_pool_sync(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_boolean_value(required, "force", B_TRUE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_SYNC, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_pool_reopen(const char *pool)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_boolean_value(optional, "scrub_restart", B_FALSE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_REOPEN, pool, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_pool_checkpoint(const char *pool)
	{
	IOC_INPUT_TEST(ZFS_IOC_POOL_CHECKPOINT, pool, NULL, NULL, 0);
	}

	static void
	test_pool_discard_checkpoint(const char *pool)
	{
	int err = lzc_pool_checkpoint(pool);
	if (err == 0 \|\| err == ZFS_ERR_CHECKPOINT_EXISTS)
	IOC_INPUT_TEST(ZFS_IOC_POOL_DISCARD_CHECKPOINT, pool, NULL,
	NULL, 0);
	}

	static void
	test_log_history(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_string(required, "message", "input check");

	IOC_INPUT_TEST(ZFS_IOC_LOG_HISTORY, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_create(const char *pool)
	{
	char dataset[MAXNAMELEN + 32];

	(void) snprintf(dataset, sizeof (dataset), "%s/create-fs", pool);

	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_int32(required, "type", DMU_OST_ZFS);
	fnvlist_add_uint64(props, "recordsize", 8192);
	fnvlist_add_nvlist(optional, "props", props);

	IOC_INPUT_TEST(ZFS_IOC_CREATE, dataset, required, optional, 0);

	nvlist_free(required);
	nvlist_free(optional);
	}

	static void
	test_snapshot(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *snaps = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_boolean(snaps, snapshot);
	fnvlist_add_nvlist(required, "snaps", snaps);

	fnvlist_add_string(props, "org.openzfs:launch", "September 17th, 2013");
	fnvlist_add_nvlist(optional, "props", props);

	IOC_INPUT_TEST(ZFS_IOC_SNAPSHOT, pool, required, optional, 0);

	nvlist_free(props);
	nvlist_free(snaps);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_space_snaps(const char *snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	fnvlist_add_string(required, "firstsnap", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_SPACE_SNAPS, snapshot, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_destroy_snaps(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *snaps = fnvlist_alloc();

	fnvlist_add_boolean(snaps, snapshot);
	fnvlist_add_nvlist(required, "snaps", snaps);

	IOC_INPUT_TEST(ZFS_IOC_DESTROY_SNAPS, pool, required, NULL, 0);

	nvlist_free(snaps);
	nvlist_free(required);
	}


	static void
	test_bookmark(const char pool, const char snapshot, const char *bookmark)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_string(required, bookmark, snapshot);

	IOC_INPUT_TEST_WILD(ZFS_IOC_BOOKMARK, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_get_bookmarks(const char *dataset)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_boolean(optional, "guid");
	fnvlist_add_boolean(optional, "createtxg");
	fnvlist_add_boolean(optional, "creation");

	IOC_INPUT_TEST_WILD(ZFS_IOC_GET_BOOKMARKS, dataset, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_destroy_bookmarks(const char pool, const char bookmark)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_boolean(required, bookmark);

	IOC_INPUT_TEST_WILD(ZFS_IOC_DESTROY_BOOKMARKS, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_clone(const char snapshot, const char clone)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_string(required, "origin", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_CLONE, clone, required, NULL, 0);

	nvlist_free(props);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_rollback(const char dataset, const char snapshot)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_string(optional, "target", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_ROLLBACK, dataset, NULL, optional, B_FALSE);

	nvlist_free(optional);
	}

	static void
	test_hold(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *holds = fnvlist_alloc();

	fnvlist_add_string(holds, snapshot, "libzfs_check_hold");
	fnvlist_add_nvlist(required, "holds", holds);
	fnvlist_add_int32(optional, "cleanup_fd", zfs_fd);

	IOC_INPUT_TEST(ZFS_IOC_HOLD, pool, required, optional, 0);

	nvlist_free(holds);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_get_holds(const char *snapshot)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_HOLDS, snapshot, NULL, NULL, 0);
	}

	static void
	test_release(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *release = fnvlist_alloc();

	fnvlist_add_boolean(release, "libzfs_check_hold");
	fnvlist_add_nvlist(required, snapshot, release);

	IOC_INPUT_TEST_WILD(ZFS_IOC_RELEASE, pool, required, NULL, 0);

	nvlist_free(release);
	nvlist_free(required);
	}


	static void
	test_send_new(const char *snapshot, int fd)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_int32(required, "fd", fd);

	fnvlist_add_boolean(optional, "largeblockok");
	fnvlist_add_boolean(optional, "embedok");
	fnvlist_add_boolean(optional, "compressok");
	fnvlist_add_boolean(optional, "rawok");

	/*
	* TODO - Resumable send is harder to set up. So we currently
	* ignore testing for that variant.
	*/
	#if 0
	fnvlist_add_string(optional, "fromsnap", from);
	fnvlist_add_uint64(optional, "resume_object", resumeobj);
	fnvlist_add_uint64(optional, "resume_offset", offset);
	fnvlist_add_boolean(optional, "savedok");
	#endif
	IOC_INPUT_TEST(ZFS_IOC_SEND_NEW, snapshot, required, optional, 0);

	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_recv_new(const char *dataset, int fd)
	{
	dmu_replay_record_t drr = { 0 };
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();
	char snapshot[MAXNAMELEN + 32];
	ssize_t count;

	int cleanup_fd = open(ZFS_DEV, O_RDWR);
	-
	+ if (cleanup_fd == -1) {
	+ (void) fprintf(stderr, "open(%s) failed: %s\n", ZFS_DEV,
	+ strerror(errno));
	+ exit(EXIT_FAILURE);
	+ }
	(void) snprintf(snapshot, sizeof (snapshot), "%s@replicant", dataset);

	count = pread(fd, &drr, sizeof (drr), 0);
	if (count != sizeof (drr)) {
	(void) fprintf(stderr, "could not read stream: %s\n",
	strerror(errno));
	}

	fnvlist_add_string(required, "snapname", snapshot);
	fnvlist_add_byte_array(required, "begin_record", (uchar_t *)&drr,
	sizeof (drr));
	fnvlist_add_int32(required, "input_fd", fd);

	fnvlist_add_string(props, "org.openzfs:launch", "September 17th, 2013");
	fnvlist_add_nvlist(optional, "localprops", props);
	fnvlist_add_boolean(optional, "force");
	fnvlist_add_boolean(optional, "heal");
	fnvlist_add_int32(optional, "cleanup_fd", cleanup_fd);

	/*
	* TODO - Resumable receive is harder to set up. So we currently
	* ignore testing for one.
	*/
	#if 0
	fnvlist_add_nvlist(optional, "props", recvdprops);
	fnvlist_add_string(optional, "origin", origin);
	fnvlist_add_boolean(optional, "resumable");
	fnvlist_add_uint64(optional, "action_handle", *action_handle);
	#endif
	IOC_INPUT_TEST(ZFS_IOC_RECV_NEW, dataset, required, optional,
	ZFS_ERR_STREAM_TRUNCATED);

	nvlist_free(props);
	nvlist_free(optional);
	nvlist_free(required);

	(void) close(cleanup_fd);
	}

	static void
	test_send_space(const char snapshot1, const char snapshot2)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_string(optional, "from", snapshot1);
	fnvlist_add_boolean(optional, "largeblockok");
	fnvlist_add_boolean(optional, "embedok");
	fnvlist_add_boolean(optional, "compressok");
	fnvlist_add_boolean(optional, "rawok");

	IOC_INPUT_TEST(ZFS_IOC_SEND_SPACE, snapshot2, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_remap(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_REMAP, dataset, NULL, NULL, 0);
	}

	static void
	test_channel_program(const char *pool)
	{
	const char *program =
	"arg = ...\n"
	"argv = arg[\"argv\"]\n"
	"return argv[1]";
	const char *const argv[1] = { "Hello World!" };
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *args = fnvlist_alloc();

	fnvlist_add_string(required, "program", program);
	fnvlist_add_string_array(args, "argv", argv, 1);
	fnvlist_add_nvlist(required, "arg", args);

	fnvlist_add_boolean_value(optional, "sync", B_TRUE);
	fnvlist_add_uint64(optional, "instrlimit", 1000 * 1000);
	fnvlist_add_uint64(optional, "memlimit", 8192 * 1024);

	IOC_INPUT_TEST(ZFS_IOC_CHANNEL_PROGRAM, pool, required, optional, 0);

	nvlist_free(args);
	nvlist_free(optional);
	nvlist_free(required);
	}

	#define WRAPPING_KEY_LEN 32

	static void
	test_load_key(const char *dataset)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *hidden = fnvlist_alloc();
	uint8_t keydata[WRAPPING_KEY_LEN] = {0};

	fnvlist_add_uint8_array(hidden, "wkeydata", keydata, sizeof (keydata));
	fnvlist_add_nvlist(required, "hidden_args", hidden);
	fnvlist_add_boolean(optional, "noop");

	IOC_INPUT_TEST(ZFS_IOC_LOAD_KEY, dataset, required, optional, EINVAL);
	nvlist_free(hidden);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_change_key(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_CHANGE_KEY, dataset, NULL, NULL, EINVAL);
	}

	static void
	test_unload_key(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_UNLOAD_KEY, dataset, NULL, NULL, EACCES);
	}

	static void
	test_vdev_initialize(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *vdev_guids = fnvlist_alloc();

	fnvlist_add_uint64(vdev_guids, "path", 0xdeadbeefdeadbeef);
	fnvlist_add_uint64(required, ZPOOL_INITIALIZE_COMMAND,
	POOL_INITIALIZE_START);
	fnvlist_add_nvlist(required, ZPOOL_INITIALIZE_VDEVS, vdev_guids);

	IOC_INPUT_TEST(ZFS_IOC_POOL_INITIALIZE, pool, required, NULL, EINVAL);
	nvlist_free(vdev_guids);
	nvlist_free(required);
	}

	static void
	test_vdev_trim(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *vdev_guids = fnvlist_alloc();

	fnvlist_add_uint64(vdev_guids, "path", 0xdeadbeefdeadbeef);
	fnvlist_add_uint64(required, ZPOOL_TRIM_COMMAND, POOL_TRIM_START);
	fnvlist_add_nvlist(required, ZPOOL_TRIM_VDEVS, vdev_guids);
	fnvlist_add_uint64(optional, ZPOOL_TRIM_RATE, 1ULL << 30);
	fnvlist_add_boolean_value(optional, ZPOOL_TRIM_SECURE, B_TRUE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_TRIM, pool, required, optional, EINVAL);
	nvlist_free(vdev_guids);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static int
	zfs_destroy(const char *dataset)
	{
	zfs_cmd_t zc = {"\0"};
	int err;

	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';
	err = lzc_ioctl_fd(zfs_fd, ZFS_IOC_DESTROY, &zc);

	return (err == 0 ? 0 : errno);
	}

	static void
	test_redact(const char snapshot1, const char snapshot2)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *snapnv = fnvlist_alloc();
	char bookmark[MAXNAMELEN + 32];

	fnvlist_add_string(required, "bookname", "testbookmark");
	fnvlist_add_boolean(snapnv, snapshot2);
	fnvlist_add_nvlist(required, "snapnv", snapnv);

	IOC_INPUT_TEST(ZFS_IOC_REDACT, snapshot1, required, NULL, 0);

	nvlist_free(snapnv);
	nvlist_free(required);

	strlcpy(bookmark, snapshot1, sizeof (bookmark));
	*strchr(bookmark, '@') = '\0';
	strlcat(bookmark, "#testbookmark", sizeof (bookmark) -
	strlen(bookmark));
	zfs_destroy(bookmark);
	}

	static void
	test_get_bookmark_props(const char *bookmark)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_BOOKMARK_PROPS, bookmark, NULL, NULL, 0);
	}

	static void
	test_wait(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_int32(required, "wait_activity", 2);
	fnvlist_add_uint64(optional, "wait_tag", 0xdeadbeefdeadbeef);

	IOC_INPUT_TEST(ZFS_IOC_WAIT, pool, required, optional, EINVAL);

	nvlist_free(required);
	nvlist_free(optional);
	}

	static void
	test_wait_fs(const char *dataset)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_int32(required, "wait_activity", 2);

	IOC_INPUT_TEST(ZFS_IOC_WAIT_FS, dataset, required, NULL, EINVAL);

	nvlist_free(required);
	}

	static void
	test_get_bootenv(const char *pool)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_BOOTENV, pool, NULL, NULL, 0);
	}

	static void
	test_set_bootenv(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_uint64(required, "version", VB_RAW);
	fnvlist_add_string(required, GRUB_ENVMAP, "test");

	IOC_INPUT_TEST_WILD(ZFS_IOC_SET_BOOTENV, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	zfs_ioc_input_tests(const char *pool)
	{
	char filepath[] = "/tmp/ioc_test_file_XXXXXX";
	char dataset[ZFS_MAX_DATASET_NAME_LEN];
	char snapbase[ZFS_MAX_DATASET_NAME_LEN + 32];
	char snapshot[ZFS_MAX_DATASET_NAME_LEN + 32];
	char bookmark[ZFS_MAX_DATASET_NAME_LEN + 32];
	char backup[ZFS_MAX_DATASET_NAME_LEN];
	char clone[ZFS_MAX_DATASET_NAME_LEN];
	char clonesnap[ZFS_MAX_DATASET_NAME_LEN + 32];
	int tmpfd, err;

	/*
	* Setup names and create a working dataset
	*/
	(void) snprintf(dataset, sizeof (dataset), "%s/test-fs", pool);
	(void) snprintf(snapbase, sizeof (snapbase), "%s@snapbase", dataset);
	(void) snprintf(snapshot, sizeof (snapshot), "%s@snapshot", dataset);
	(void) snprintf(bookmark, sizeof (bookmark), "%s#bookmark", dataset);
	(void) snprintf(clone, sizeof (clone), "%s/test-fs-clone", pool);
	(void) snprintf(clonesnap, sizeof (clonesnap), "%s@snap", clone);
	(void) snprintf(backup, sizeof (backup), "%s/backup", pool);

	err = lzc_create(dataset, LZC_DATSET_TYPE_ZFS, NULL, NULL, -1);
	if (err) {
	(void) fprintf(stderr, "could not create '%s': %s\n",
	dataset, strerror(errno));
	exit(2);
	}

	tmpfd = mkstemp(filepath);
	if (tmpfd < 0) {
	(void) fprintf(stderr, "could not create '%s': %s\n",
	filepath, strerror(errno));
	exit(2);
	}

	/*
	* run a test for each ioctl
	* Note that some test build on previous test operations
	*/
	test_pool_sync(pool);
	test_pool_reopen(pool);
	test_pool_checkpoint(pool);
	test_pool_discard_checkpoint(pool);
	test_log_history(pool);

	test_create(dataset);
	test_snapshot(pool, snapbase);
	test_snapshot(pool, snapshot);

	test_space_snaps(snapshot);
	test_send_space(snapbase, snapshot);
	test_send_new(snapshot, tmpfd);
	test_recv_new(backup, tmpfd);

	test_bookmark(pool, snapshot, bookmark);
	test_get_bookmarks(dataset);
	test_get_bookmark_props(bookmark);
	test_destroy_bookmarks(pool, bookmark);

	test_hold(pool, snapshot);
	test_get_holds(snapshot);
	test_release(pool, snapshot);

	test_clone(snapshot, clone);
	test_snapshot(pool, clonesnap);
	test_redact(snapshot, clonesnap);
	zfs_destroy(clonesnap);
	zfs_destroy(clone);

	test_rollback(dataset, snapshot);
	test_destroy_snaps(pool, snapshot);
	test_destroy_snaps(pool, snapbase);

	test_remap(dataset);
	test_channel_program(pool);

	test_load_key(dataset);
	test_change_key(dataset);
	test_unload_key(dataset);

	test_vdev_initialize(pool);
	test_vdev_trim(pool);

	test_wait(pool);
	test_wait_fs(dataset);

	test_set_bootenv(pool);
	test_get_bootenv(pool);

	/*
	* cleanup
	*/
	zfs_cmd_t zc = {"\0"};

	nvlist_t *snaps = fnvlist_alloc();
	fnvlist_add_boolean(snaps, snapshot);
	(void) lzc_destroy_snaps(snaps, B_FALSE, NULL);
	nvlist_free(snaps);

	(void) zfs_destroy(dataset);
	(void) zfs_destroy(backup);

	(void) close(tmpfd);
	(void) unlink(filepath);

	/*
	* All the unused slots should yield ZFS_ERR_IOC_CMD_UNAVAIL
	*/
	for (int i = 0; i < ARRAY_SIZE(ioc_skip); i++) {
	if (ioc_tested[ioc_skip[i] - ZFS_IOC_FIRST])
	(void) fprintf(stderr, "cmd %d tested, not skipped!\n",
	(int)(ioc_skip[i] - ZFS_IOC_FIRST));

	ioc_tested[ioc_skip[i] - ZFS_IOC_FIRST] = B_TRUE;
	}

	(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';

	for (unsigned ioc = ZFS_IOC_FIRST; ioc < ZFS_IOC_LAST; ioc++) {
	unsigned cmd = ioc - ZFS_IOC_FIRST;

	if (ioc_tested[cmd])
	continue;

	if (lzc_ioctl_fd(zfs_fd, ioc, &zc) != 0 &&
	errno != ZFS_ERR_IOC_CMD_UNAVAIL) {
	(void) fprintf(stderr, "cmd %d is missing a test case "
	"(%d)\n", cmd, errno);
	}
	}
	}

	enum zfs_ioc_ref {
	#ifdef __FreeBSD__
	ZFS_IOC_BASE = 0,
	#else
	ZFS_IOC_BASE = ('Z' << 8),
	#endif
	ZFS_IOC_PLATFORM_BASE = ZFS_IOC_BASE + 0x80,
	};

	/*
	* Canonical reference check of /dev/zfs ioctl numbers.
	* These cannot change and new ioctl numbers must be appended.
	*/
	static boolean_t
	validate_ioc_values(void)
	{
	boolean_t result = B_TRUE;

	#define CHECK(expr) do { \
	if (!(expr)) { \
	result = B_FALSE; \
	fprintf(stderr, "(%s) === FALSE\n", #expr); \
	} \
	} while (0)

	CHECK(ZFS_IOC_BASE + 0 == ZFS_IOC_POOL_CREATE);
	CHECK(ZFS_IOC_BASE + 1 == ZFS_IOC_POOL_DESTROY);
	CHECK(ZFS_IOC_BASE + 2 == ZFS_IOC_POOL_IMPORT);
	CHECK(ZFS_IOC_BASE + 3 == ZFS_IOC_POOL_EXPORT);
	CHECK(ZFS_IOC_BASE + 4 == ZFS_IOC_POOL_CONFIGS);
	CHECK(ZFS_IOC_BASE + 5 == ZFS_IOC_POOL_STATS);
	CHECK(ZFS_IOC_BASE + 6 == ZFS_IOC_POOL_TRYIMPORT);
	CHECK(ZFS_IOC_BASE + 7 == ZFS_IOC_POOL_SCAN);
	CHECK(ZFS_IOC_BASE + 8 == ZFS_IOC_POOL_FREEZE);
	CHECK(ZFS_IOC_BASE + 9 == ZFS_IOC_POOL_UPGRADE);
	CHECK(ZFS_IOC_BASE + 10 == ZFS_IOC_POOL_GET_HISTORY);
	CHECK(ZFS_IOC_BASE + 11 == ZFS_IOC_VDEV_ADD);
	CHECK(ZFS_IOC_BASE + 12 == ZFS_IOC_VDEV_REMOVE);
	CHECK(ZFS_IOC_BASE + 13 == ZFS_IOC_VDEV_SET_STATE);
	CHECK(ZFS_IOC_BASE + 14 == ZFS_IOC_VDEV_ATTACH);
	CHECK(ZFS_IOC_BASE + 15 == ZFS_IOC_VDEV_DETACH);
	CHECK(ZFS_IOC_BASE + 16 == ZFS_IOC_VDEV_SETPATH);
	CHECK(ZFS_IOC_BASE + 17 == ZFS_IOC_VDEV_SETFRU);
	CHECK(ZFS_IOC_BASE + 18 == ZFS_IOC_OBJSET_STATS);
	CHECK(ZFS_IOC_BASE + 19 == ZFS_IOC_OBJSET_ZPLPROPS);
	CHECK(ZFS_IOC_BASE + 20 == ZFS_IOC_DATASET_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 21 == ZFS_IOC_SNAPSHOT_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 22 == ZFS_IOC_SET_PROP);
	CHECK(ZFS_IOC_BASE + 23 == ZFS_IOC_CREATE);
	CHECK(ZFS_IOC_BASE + 24 == ZFS_IOC_DESTROY);
	CHECK(ZFS_IOC_BASE + 25 == ZFS_IOC_ROLLBACK);
	CHECK(ZFS_IOC_BASE + 26 == ZFS_IOC_RENAME);
	CHECK(ZFS_IOC_BASE + 27 == ZFS_IOC_RECV);
	CHECK(ZFS_IOC_BASE + 28 == ZFS_IOC_SEND);
	CHECK(ZFS_IOC_BASE + 29 == ZFS_IOC_INJECT_FAULT);
	CHECK(ZFS_IOC_BASE + 30 == ZFS_IOC_CLEAR_FAULT);
	CHECK(ZFS_IOC_BASE + 31 == ZFS_IOC_INJECT_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 32 == ZFS_IOC_ERROR_LOG);
	CHECK(ZFS_IOC_BASE + 33 == ZFS_IOC_CLEAR);
	CHECK(ZFS_IOC_BASE + 34 == ZFS_IOC_PROMOTE);
	CHECK(ZFS_IOC_BASE + 35 == ZFS_IOC_SNAPSHOT);
	CHECK(ZFS_IOC_BASE + 36 == ZFS_IOC_DSOBJ_TO_DSNAME);
	CHECK(ZFS_IOC_BASE + 37 == ZFS_IOC_OBJ_TO_PATH);
	CHECK(ZFS_IOC_BASE + 38 == ZFS_IOC_POOL_SET_PROPS);
	CHECK(ZFS_IOC_BASE + 39 == ZFS_IOC_POOL_GET_PROPS);
	CHECK(ZFS_IOC_BASE + 40 == ZFS_IOC_SET_FSACL);
	CHECK(ZFS_IOC_BASE + 41 == ZFS_IOC_GET_FSACL);
	CHECK(ZFS_IOC_BASE + 42 == ZFS_IOC_SHARE);
	CHECK(ZFS_IOC_BASE + 43 == ZFS_IOC_INHERIT_PROP);
	CHECK(ZFS_IOC_BASE + 44 == ZFS_IOC_SMB_ACL);
	CHECK(ZFS_IOC_BASE + 45 == ZFS_IOC_USERSPACE_ONE);
	CHECK(ZFS_IOC_BASE + 46 == ZFS_IOC_USERSPACE_MANY);
	CHECK(ZFS_IOC_BASE + 47 == ZFS_IOC_USERSPACE_UPGRADE);
	CHECK(ZFS_IOC_BASE + 48 == ZFS_IOC_HOLD);
	CHECK(ZFS_IOC_BASE + 49 == ZFS_IOC_RELEASE);
	CHECK(ZFS_IOC_BASE + 50 == ZFS_IOC_GET_HOLDS);
	CHECK(ZFS_IOC_BASE + 51 == ZFS_IOC_OBJSET_RECVD_PROPS);
	CHECK(ZFS_IOC_BASE + 52 == ZFS_IOC_VDEV_SPLIT);
	CHECK(ZFS_IOC_BASE + 53 == ZFS_IOC_NEXT_OBJ);
	CHECK(ZFS_IOC_BASE + 54 == ZFS_IOC_DIFF);
	CHECK(ZFS_IOC_BASE + 55 == ZFS_IOC_TMP_SNAPSHOT);
	CHECK(ZFS_IOC_BASE + 56 == ZFS_IOC_OBJ_TO_STATS);
	CHECK(ZFS_IOC_BASE + 57 == ZFS_IOC_SPACE_WRITTEN);
	CHECK(ZFS_IOC_BASE + 58 == ZFS_IOC_SPACE_SNAPS);
	CHECK(ZFS_IOC_BASE + 59 == ZFS_IOC_DESTROY_SNAPS);
	CHECK(ZFS_IOC_BASE + 60 == ZFS_IOC_POOL_REGUID);
	CHECK(ZFS_IOC_BASE + 61 == ZFS_IOC_POOL_REOPEN);
	CHECK(ZFS_IOC_BASE + 62 == ZFS_IOC_SEND_PROGRESS);
	CHECK(ZFS_IOC_BASE + 63 == ZFS_IOC_LOG_HISTORY);
	CHECK(ZFS_IOC_BASE + 64 == ZFS_IOC_SEND_NEW);
	CHECK(ZFS_IOC_BASE + 65 == ZFS_IOC_SEND_SPACE);
	CHECK(ZFS_IOC_BASE + 66 == ZFS_IOC_CLONE);
	CHECK(ZFS_IOC_BASE + 67 == ZFS_IOC_BOOKMARK);
	CHECK(ZFS_IOC_BASE + 68 == ZFS_IOC_GET_BOOKMARKS);
	CHECK(ZFS_IOC_BASE + 69 == ZFS_IOC_DESTROY_BOOKMARKS);
	CHECK(ZFS_IOC_BASE + 70 == ZFS_IOC_RECV_NEW);
	CHECK(ZFS_IOC_BASE + 71 == ZFS_IOC_POOL_SYNC);
	CHECK(ZFS_IOC_BASE + 72 == ZFS_IOC_CHANNEL_PROGRAM);
	CHECK(ZFS_IOC_BASE + 73 == ZFS_IOC_LOAD_KEY);
	CHECK(ZFS_IOC_BASE + 74 == ZFS_IOC_UNLOAD_KEY);
	CHECK(ZFS_IOC_BASE + 75 == ZFS_IOC_CHANGE_KEY);
	CHECK(ZFS_IOC_BASE + 76 == ZFS_IOC_REMAP);
	CHECK(ZFS_IOC_BASE + 77 == ZFS_IOC_POOL_CHECKPOINT);
	CHECK(ZFS_IOC_BASE + 78 == ZFS_IOC_POOL_DISCARD_CHECKPOINT);
	CHECK(ZFS_IOC_BASE + 79 == ZFS_IOC_POOL_INITIALIZE);
	CHECK(ZFS_IOC_BASE + 80 == ZFS_IOC_POOL_TRIM);
	CHECK(ZFS_IOC_BASE + 81 == ZFS_IOC_REDACT);
	CHECK(ZFS_IOC_BASE + 82 == ZFS_IOC_GET_BOOKMARK_PROPS);
	CHECK(ZFS_IOC_BASE + 83 == ZFS_IOC_WAIT);
	CHECK(ZFS_IOC_BASE + 84 == ZFS_IOC_WAIT_FS);
	CHECK(ZFS_IOC_PLATFORM_BASE + 1 == ZFS_IOC_EVENTS_NEXT);
	CHECK(ZFS_IOC_PLATFORM_BASE + 2 == ZFS_IOC_EVENTS_CLEAR);
	CHECK(ZFS_IOC_PLATFORM_BASE + 3 == ZFS_IOC_EVENTS_SEEK);
	CHECK(ZFS_IOC_PLATFORM_BASE + 4 == ZFS_IOC_NEXTBOOT);
	CHECK(ZFS_IOC_PLATFORM_BASE + 5 == ZFS_IOC_JAIL);
	CHECK(ZFS_IOC_PLATFORM_BASE + 6 == ZFS_IOC_UNJAIL);
	CHECK(ZFS_IOC_PLATFORM_BASE + 7 == ZFS_IOC_SET_BOOTENV);
	CHECK(ZFS_IOC_PLATFORM_BASE + 8 == ZFS_IOC_GET_BOOTENV);

	#undef CHECK

	return (result);
	}

	int
	main(int argc, const char *argv[])
	{
	if (argc != 2) {
	(void) fprintf(stderr, "usage: %s <pool>\n", argv[0]);
	exit(2);
	}

	if (!validate_ioc_values()) {
	(void) fprintf(stderr, "WARNING: zfs_ioc_t has binary "
	"incompatible command values\n");
	exit(3);
	}

	(void) libzfs_core_init();
	zfs_fd = open(ZFS_DEV, O_RDWR);
	if (zfs_fd < 0) {
	(void) fprintf(stderr, "open: %s\n", strerror(errno));
	libzfs_core_fini();
	exit(2);
	}

	zfs_ioc_input_tests(argv[1]);

	(void) close(zfs_fd);
	libzfs_core_fini();

	return (unexpected_failures);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/mkfile.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/mkfile.c
	index 7ce50e6a37c4..3b61deed6bf5 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/mkfile.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/mkfile.c
	@@ -1,280 +1,284 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2005 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include <stdio.h>
	#include <ctype.h>
	#include <unistd.h>
	#include <sys/stat.h>
	#include <fcntl.h>
	#include <stdlib.h>
	#include <string.h>
	#include <libintl.h>
	#include <errno.h>
	#include <sys/stdtypes.h>
	#include <sys/sysmacros.h>

	#define BLOCKSIZE 512 /* bytes */
	#define KILOBYTE 1024
	#define MEGABYTE (KILOBYTE * KILOBYTE)
	#define GIGABYTE (KILOBYTE * MEGABYTE)

	#define FILE_MODE (S_ISVTX + S_IRUSR + S_IWUSR)

	static __attribute__((noreturn)) void
	usage(void)
	{
	(void) fprintf(stderr, gettext(
	"Usage: mkfile [-nv] <size>[g\|k\|b\|m] <name1> [<name2>] ...\n"));
	exit(1);
	}

	int
	main(int argc, char **argv)
	{
	char *opts;
	off_t size;
	size_t len;
	size_t mult = 1;
	char *buf = NULL;
	size_t bufsz = 0;
	int errors = 0;
	int i;
	int verbose = 0; /* option variable */
	int nobytes = 0; /* option variable */
	int saverr;

	if (argc == 1)
	usage();

	while (argv[1] && argv[1][0] == '-') {
	opts = &argv[1][0];
	while (*(++opts)) {
	switch (*opts) {
	case 'v':
	verbose++;
	break;
	case 'n':
	nobytes++;
	break;
	default:
	usage();
	}
	}
	argc--;
	argv++;
	}
	if (argc < 3)
	usage();

	len = strlen(argv[1]);
	if (len && isalpha(argv[1][len-1])) {
	switch (argv[1][len-1]) {
	case 'k':
	case 'K':
	mult = KILOBYTE;
	break;
	case 'b':
	case 'B':
	mult = BLOCKSIZE;
	break;
	case 'm':
	case 'M':
	mult = MEGABYTE;
	break;
	case 'g':
	case 'G':
	mult = GIGABYTE;
	break;
	default:
	(void) fprintf(stderr,
	gettext("unknown size %s\n"), argv[1]);
	usage();
	}

	for (i = 0; i <= (len-2); i++) {
	if (!isdigit(argv[1][i])) {
	(void) fprintf(stderr,
	gettext("unknown size %s\n"), argv[1]);
	usage();
	}
	}
	argv[1][len-1] = '\0';
	}
	size = ((off_t)atoll(argv[1]) * (off_t)mult);

	argv++;
	argc--;

	while (argc > 1) {
	int fd;

	if (verbose)
	(void) fprintf(stdout, gettext("%s %lld bytes\n"),
	argv[1], (offset_t)size);
	fd = open(argv[1], O_CREAT\|O_TRUNC\|O_RDWR, FILE_MODE);
	if (fd < 0) {
	saverr = errno;
	(void) fprintf(stderr,
	gettext("Could not open %s: %s\n"),
	argv[1], strerror(saverr));
	errors++;
	argv++;
	argc--;
	continue;
	} else if (fchown(fd, getuid(), getgid()) < 0) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Could not set owner/group of %s: %s\n"),
	argv[1], strerror(saverr));
	(void) close(fd);
	errors++;
	argv++;
	argc--;
	continue;
	} else if (lseek(fd, (off_t)size-1, SEEK_SET) < 0) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Could not seek to offset %ld in %s: %s\n"),
	(unsigned long)size-1, argv[1], strerror(saverr));
	(void) close(fd);
	errors++;
	argv++;
	argc--;
	continue;
	} else if (write(fd, "", 1) != 1) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Could not set length of %s: %s\n"),
	argv[1], strerror(saverr));
	(void) close(fd);
	errors++;
	argv++;
	argc--;
	continue;
	}

	if (!nobytes) {
	off_t written = 0;
	struct stat64 st;

	if (lseek(fd, (off_t)0, SEEK_SET) < 0) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Could not seek to beginning of %s: %s\n"),
	argv[1], strerror(saverr));
	(void) close(fd);
	errors++;
	argv++;
	argc--;
	continue;
	}
	if (fstat64(fd, &st) < 0) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Could not fstat64 %s: %s\n"),
	argv[1], strerror(saverr));
	(void) close(fd);
	errors++;
	argv++;
	argc--;
	continue;
	}
	if (bufsz != st.st_blksize) {
	if (buf)
	free(buf);
	bufsz = (size_t)st.st_blksize;
	buf = calloc(1, bufsz);
	if (buf == NULL) {
	(void) fprintf(stderr, gettext(
	"Could not allocate buffer of"
	" size %d\n"), (int)bufsz);
	(void) close(fd);
	bufsz = 0;
	errors++;
	argv++;
	argc--;
	continue;
	}
	}
	while (written < size) {
	ssize_t result;
	size_t bytes = (size_t)MIN(bufsz, size-written);

	if ((result = write(fd, buf, bytes)) !=
	(ssize_t)bytes) {
	saverr = errno;
	if (result < 0)
	result = 0;
	written += result;
	(void) fprintf(stderr, gettext(
	"%s: initialized %lu of %lu bytes: %s\n"),
	argv[1], (unsigned long)written,
	(unsigned long)size,
	strerror(saverr));
	errors++;
	break;
	}
	written += bytes;
	}

	/*
	* A write(2) call in the above loop failed so
	* close out this file and go on (error was
	* already incremented when the write(2) failed).
	*/
	if (written < size) {
	(void) close(fd);
	argv++;
	argc--;
	continue;
	}
	}
	if (close(fd) < 0) {
	saverr = errno;
	(void) fprintf(stderr, gettext(
	"Error encountered when closing %s: %s\n"),
	argv[1], strerror(saverr));
	errors++;
	argv++;
	argc--;
	continue;
	}

	/*
	* Only set the modes (including the sticky bit) if we
	* had no problems. It is not an error for the chmod(2)
	* to fail, but do issue a warning.
	*/
	if (chmod(argv[1], FILE_MODE) < 0)
	(void) fprintf(stderr, gettext(
	"warning: couldn't set mode to %#o\n"), FILE_MODE);

	argv++;
	argc--;
	}
	+
	+ if (buf)
	+ free(buf);
	+
	return (errors);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/mmap_sync.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/mmap_sync.c
	index 0e4bba37d7be..e4d190aef1c6 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/mmap_sync.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/mmap_sync.c
	@@ -1,152 +1,152 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or http://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	#include <stdlib.h>
	#include <stdio.h>
	#include <string.h>
	#include <sys/mman.h>
	#include <sys/stat.h>
	#include <sys/time.h>
	#include <fcntl.h>
	#include <unistd.h>
	#include <time.h>

	static void
	cleanup(char *file)
	{
	- remove(file);
	+ (void) remove(file);
	}

	int
	main(int argc, char *argv[])
	{
	char *testdir = getenv("TESTDIR");
	if (!testdir) {
	fprintf(stderr, "environment variable TESTDIR not set\n");
	return (1);
	}

	struct stat st;
	umask(0);
	if (stat(testdir, &st) != 0 &&
	mkdir(testdir, 0777) != 0) {
	perror("mkdir");
	return (1);
	}

	if (argc > 3) {
	fprintf(stderr, "usage: %s "
	"[run time in mins] "
	"[max msync time in ms]\n", argv[0]);
	return (1);
	}

	int run_time_mins = 5;
	if (argc >= 2) {
	run_time_mins = atoi(argv[1]);
	}

	int max_msync_time_ms = 1000;
	if (argc >= 3) {
	max_msync_time_ms = atoi(argv[2]);
	}

	char filepath[512];
	filepath[0] = '\0';
	char *file = &filepath[0];

	- strcat(file, testdir);
	- strcat(file, "/msync_file");
	+ (void) snprintf(file, 512, "%s/msync_file", testdir);

	const int LEN = 8;
	cleanup(file);

	int fd = open(file, O_CREAT \| O_RDWR, S_IRUSR \| S_IWUSR \|
	S_IRGRP \| S_IROTH);

	if (fd == -1) {
	(void) fprintf(stderr, "%s: %s: ", argv[0], file);
	perror("open");
	return (1);
	}

	if (ftruncate(fd, LEN) != 0) {
	perror("ftruncate");
	cleanup(file);
	return (1);
	}

	void *ptr = mmap(NULL, LEN, PROT_READ \| PROT_WRITE, MAP_SHARED, fd, 0);

	if (ptr == MAP_FAILED) {
	perror("mmap");
	cleanup(file);
	return (1);
	}

	struct timeval tstart;
	gettimeofday(&tstart, NULL);

	long long x = 0LL;

	for (;;) {
	((long long )ptr) = x;
	x++;

	struct timeval t1, t2;
	gettimeofday(&t1, NULL);
	if (msync(ptr, LEN, MS_SYNC\|MS_INVALIDATE) != 0) {
	perror("msync");
	cleanup(file);
	return (1);
	}

	gettimeofday(&t2, NULL);

	double elapsed = (t2.tv_sec - t1.tv_sec) * 1000.0;
	elapsed += ((t2.tv_usec - t1.tv_usec) / 1000.0);
	if (elapsed > max_msync_time_ms) {
	fprintf(stderr, "slow msync: %f ms\n", elapsed);
	- munmap(ptr, LEN);
	+ if (munmap(ptr, LEN) != 0)
	+ perror("munmap");
	cleanup(file);
	return (1);
	}

	double elapsed_start = (t2.tv_sec - tstart.tv_sec) * 1000.0;
	elapsed_start += ((t2.tv_usec - tstart.tv_usec) / 1000.0);
	if (elapsed_start > run_time_mins * 60 * 1000) {
	break;
	}
	}

	if (munmap(ptr, LEN) != 0) {
	perror("munmap");
	cleanup(file);
	return (1);
	}

	if (close(fd) != 0) {
	perror("close");
	}

	cleanup(file);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/mmapwrite.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/mmapwrite.c
	index 0d277d6e3099..ca55d730fd34 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/mmapwrite.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/mmapwrite.c
	@@ -1,157 +1,161 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include <unistd.h>
	#include <fcntl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <sys/mman.h>
	#include <pthread.h>
	#include <errno.h>
	#include <err.h>

	/*
	* --------------------------------------------------------------------
	* Bug Issue Id: #7512
	* The bug time sequence:
	* 1. context #1, zfs_write assign a txg "n".
	* 2. In the same process, context #2, mmap page fault (which means the mm_sem
	* is hold) occurred, zfs_dirty_inode open a txg failed, and wait previous
	* txg "n" completed.
	* 3. context #1 call zfs_uiomove to write, however page fault is occurred in
	* zfs_uiomove, which means it needs mm_sem, but mm_sem is hold by
	* context #2, so it stuck and can't complete, then txg "n" will not
	* complete.
	*
	* So context #1 and context #2 trap into the "dead lock".
	* --------------------------------------------------------------------
	*/

	#define NORMAL_WRITE_TH_NUM 2

	static void *
	normal_writer(void *filename)
	{
	char *file_path = filename;
	int fd = -1;
	ssize_t write_num = 0;
	int page_size = getpagesize();

	fd = open(file_path, O_RDWR \| O_CREAT, 0777);
	if (fd == -1) {
	err(1, "failed to open %s", file_path);
	}

	- char buf;
	+ char buf = 'z';
	while (1) {
	write_num = write(fd, &buf, 1);
	if (write_num == 0) {
	err(1, "write failed!");
	break;
	}
	- lseek(fd, page_size, SEEK_CUR);
	+ if (lseek(fd, page_size, SEEK_CUR) == -1) {
	+ err(1, "lseek failed on %s: %s", file_path,
	+ strerror(errno));
	+ break;
	+ }
	}
	}

	static void *
	map_writer(void *filename)
	{
	int fd = -1;
	int ret = 0;
	char *buf = NULL;
	int page_size = getpagesize();
	int op_errno = 0;
	char *file_path = filename;

	while (1) {
	ret = access(file_path, F_OK);
	if (ret) {
	op_errno = errno;
	if (op_errno == ENOENT) {
	fd = open(file_path, O_RDWR \| O_CREAT, 0777);
	if (fd == -1) {
	err(1, "open file failed");
	}

	ret = ftruncate(fd, page_size);
	if (ret == -1) {
	err(1, "truncate file failed");
	}
	} else {
	err(1, "access file failed!");
	}
	} else {
	fd = open(file_path, O_RDWR, 0777);
	if (fd == -1) {
	err(1, "open file failed");
	}
	}

	if ((buf = mmap(NULL, page_size, PROT_READ \| PROT_WRITE,
	MAP_SHARED, fd, 0)) == MAP_FAILED) {
	err(1, "map file failed");
	}

	if (fd != -1)
	close(fd);

	char s[10] = {0, };
	memcpy(buf, s, 10);
	ret = munmap(buf, page_size);
	if (ret != 0) {
	err(1, "unmap file failed");
	}
	}
	}

	int
	main(int argc, char **argv)
	{
	pthread_t map_write_tid;
	pthread_t normal_write_tid[NORMAL_WRITE_TH_NUM];
	int i = 0;

	if (argc != 3) {
	(void) printf("usage: %s <normal write file name> "
	"<map write file name>\n", argv[0]);
	exit(1);
	}

	for (i = 0; i < NORMAL_WRITE_TH_NUM; i++) {
	if (pthread_create(&normal_write_tid[i], NULL, normal_writer,
	argv[1])) {
	err(1, "pthread_create normal_writer failed.");
	}
	}

	if (pthread_create(&map_write_tid, NULL, map_writer, argv[2])) {
	err(1, "pthread_create map_writer failed.");
	}

	pthread_join(map_write_tid, NULL);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/readmmap.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/readmmap.c
	index c4812b4a259d..704ffd55c8a5 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/readmmap.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/readmmap.c
	@@ -1,138 +1,138 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* --------------------------------------------------------------
	* BugId 5047993 : Getting bad read data.
	*
	* Usage: readmmap <filename>
	*
	* where:
	* filename is an absolute path to the file name.
	*
	* Return values:
	* 1 : error
	* 0 : no errors
	* --------------------------------------------------------------
	*/
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <errno.h>
	#include <sys/mman.h>
	#include <time.h>

	int
	main(int argc, char **argv)
	{
	const char *filename = "badfile";
	size_t size = 4395;
	size_t idx = 0;
	char *buf = NULL;
	char *map = NULL;
	int fd = -1, bytes, retval = 0;
	- unsigned seed;
	+ uint_t seed;

	if (argc < 2 \|\| optind == argc) {
	(void) fprintf(stderr,
	"usage: %s <file name>\n", argv[0]);
	exit(1);
	}

	if ((buf = calloc(1, size)) == NULL) {
	perror("calloc");
	exit(1);
	}

	filename = argv[optind];

	(void) remove(filename);

	fd = open(filename, O_RDWR\|O_CREAT\|O_TRUNC, 0666);
	if (fd == -1) {
	perror("open to create");
	retval = 1;
	goto end;
	}

	bytes = write(fd, buf, size);
	if (bytes != size) {
	(void) printf("short write: %d != %zd\n", bytes, size);
	retval = 1;
	goto end;
	}

	map = mmap(0, size, PROT_READ\|PROT_WRITE, MAP_SHARED, fd, 0);
	if (map == MAP_FAILED) {
	perror("mmap");
	retval = 1;
	goto end;
	}
	- seed = time(NULL);
	+ seed = (uint_t)time(NULL);
	srandom(seed);

	idx = random() % size;
	map[idx] = 1;

	if (msync(map, size, MS_SYNC) != 0) {
	perror("msync");
	retval = 1;
	goto end;
	}

	if (munmap(map, size) != 0) {
	perror("munmap");
	retval = 1;
	goto end;
	}

	bytes = pread(fd, buf, size, 0);
	if (bytes != size) {
	(void) printf("short read: %d != %zd\n", bytes, size);
	retval = 1;
	goto end;
	}

	if (buf[idx] != 1) {
	(void) printf(
	"bad data from read! got buf[%zd]=%d, expected 1\n",
	idx, buf[idx]);
	retval = 1;
	goto end;
	}

	(void) printf("good data from read: buf[%zd]=1\n", idx);
	end:
	if (fd != -1) {
	(void) close(fd);
	}
	if (buf != NULL) {
	free(buf);
	}

	return (retval);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/zed_fd_spill-zedlet.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/zed_fd_spill-zedlet.c
	index c072f906d23e..b248579abb7c 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/zed_fd_spill-zedlet.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/zed_fd_spill-zedlet.c
	@@ -1,36 +1,49 @@
	/*
	* Permission to use, copy, modify, and/or distribute this software for
	* any purpose with or without fee is hereby granted.
	*
	* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
	* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
	* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
	* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
	* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN
	* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
	* OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
	*/

	#include <sys/types.h>
	#include <sys/wait.h>
	+#include <errno.h>
	#include <fcntl.h>
	#include <unistd.h>
	#include <stdlib.h>
	#include <string.h>
	#include <stdio.h>

	int main(void) {
	if (fork()) {
	int err;
	wait(&err);
	return (err);
	}

	char buf[64];
	sprintf(buf, "/tmp/zts-zed_fd_spill-logdir/%d", getppid());
	- dup2(creat(buf, 0644), STDOUT_FILENO);
	+ int fd = creat(buf, 0644);
	+ if (fd == -1) {
	+ (void) fprintf(stderr, "creat(%s) failed: %s\n", buf,
	+ strerror(errno));
	+ exit(EXIT_FAILURE);
	+ }
	+
	+ if (dup2(fd, STDOUT_FILENO) == -1) {
	+ close(fd);
	+ (void) fprintf(stderr, "dup2(%s, STDOUT_FILENO) failed: %s\n",
	+ buf, strerror(errno));
	+ exit(EXIT_FAILURE);
	+ }

	snprintf(buf, sizeof (buf), "/proc/%d/fd", getppid());
	execlp("ls", "ls", buf, NULL);
	_exit(127);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/zfs_diff-socket.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/zfs_diff-socket.c
	index be4bf31dde9f..3ebc95799fe3 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/zfs_diff-socket.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/zfs_diff-socket.c
	@@ -1,57 +1,56 @@
	/*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*/

	/*
	* Copyright 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	*/

	#include <fcntl.h>
	#include <sys/un.h>
	#include <sys/socket.h>
	#include <sys/stat.h>
	#include <sys/types.h>
	#include <unistd.h>
	#include <errno.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>

	int
	main(int argc, char *argv[])
	{
	struct sockaddr_un sock;
	int fd;
	char *path;
	size_t size;
	if (argc != 2) {
	fprintf(stderr, "usage: %s /path/to/socket\n", argv[0]);
	exit(1);
	}
	path = argv[1];
	size = sizeof (sock.sun_path);
	- strncpy(sock.sun_path, (char *)path, size - 1);
	- sock.sun_path[size - 1] = '\0';
	+ (void) snprintf(sock.sun_path, size, "%s", path);

	sock.sun_family = AF_UNIX;
	if ((fd = socket(AF_UNIX, SOCK_DGRAM, 0)) == -1) {
	perror("socket");
	return (1);
	}
	if (bind(fd, (struct sockaddr *)&sock, sizeof (struct sockaddr_un))) {
	perror("bind");
	return (1);
	}
	if (close(fd)) {
	perror("close");
	return (1);
	}
	return (0);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/include/libtest.shlib b/sys/contrib/openzfs/tests/zfs-tests/include/libtest.shlib
	index d163fc7c8ccc..02e6a500a71a 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/include/libtest.shlib
	+++ b/sys/contrib/openzfs/tests/zfs-tests/include/libtest.shlib
	@@ -1,3849 +1,3849 @@
	#
	# CDDL HEADER START
	#
	# The contents of this file are subject to the terms of the
	# Common Development and Distribution License (the "License").
	# You may not use this file except in compliance with the License.
	#
	# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	# or https://opensource.org/licenses/CDDL-1.0.
	# See the License for the specific language governing permissions
	# and limitations under the License.
	#
	# When distributing Covered Code, include this CDDL HEADER in each
	# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	# If applicable, add the following below this CDDL HEADER, with the
	# fields enclosed by brackets "[]" replaced with your own identifying
	# information: Portions Copyright [yyyy] [name of copyright owner]
	#
	# CDDL HEADER END
	#

	#
	# Copyright (c) 2009, Sun Microsystems Inc. All rights reserved.
	# Copyright (c) 2012, 2020, Delphix. All rights reserved.
	# Copyright (c) 2017, Tim Chase. All rights reserved.
	# Copyright (c) 2017, Nexenta Systems Inc. All rights reserved.
	# Copyright (c) 2017, Lawrence Livermore National Security LLC.
	# Copyright (c) 2017, Datto Inc. All rights reserved.
	# Copyright (c) 2017, Open-E Inc. All rights reserved.
	# Copyright (c) 2021, The FreeBSD Foundation.
	# Use is subject to license terms.
	#

	. ${STF_SUITE}/include/tunables.cfg

	. ${STF_TOOLS}/include/logapi.shlib
	. ${STF_SUITE}/include/math.shlib
	. ${STF_SUITE}/include/blkdev.shlib

	#
	# Apply constrained path when available. This is required since the
	# PATH may have been modified by sudo's secure_path behavior.
	#
	if [ -n "$STF_PATH" ]; then
	export PATH="$STF_PATH"
	fi

	#
	# Generic dot version comparison function
	#
	# Returns success when version $1 is greater than or equal to $2.
	#
	function compare_version_gte
	{
	[ "$(printf "$1\n$2" \| sort -V \| tail -n1)" = "$1" ]
	}

	# Linux kernel version comparison function
	#
	# $1 Linux version ("4.10", "2.6.32") or blank for installed Linux version
	#
	# Used for comparison: if [ $(linux_version) -ge $(linux_version "2.6.32") ]
	#
	function linux_version
	{
	typeset ver="$1"

	[ -z "$ver" ] && ver=$(uname -r \| grep -Eo "^[0-9]+\.[0-9]+\.[0-9]+")

	typeset version major minor _
	IFS='.' read -r version major minor _ <<<"$ver"

	[ -z "$version" ] && version=0
	[ -z "$major" ] && major=0
	[ -z "$minor" ] && minor=0

	echo $((version * 100000 + major * 1000 + minor))
	}

	# Determine if this is a Linux test system
	#
	# Return 0 if platform Linux, 1 if otherwise

	function is_linux
	{
	[ "$UNAME" = "Linux" ]
	}

	# Determine if this is an illumos test system
	#
	# Return 0 if platform illumos, 1 if otherwise
	function is_illumos
	{
	[ "$UNAME" = "illumos" ]
	}

	# Determine if this is a FreeBSD test system
	#
	# Return 0 if platform FreeBSD, 1 if otherwise

	function is_freebsd
	{
	[ "$UNAME" = "FreeBSD" ]
	}

	# Determine if this is a 32-bit system
	#
	# Return 0 if platform is 32-bit, 1 if otherwise

	function is_32bit
	{
	[ $(getconf LONG_BIT) = "32" ]
	}

	# Determine if kmemleak is enabled
	#
	# Return 0 if kmemleak is enabled, 1 if otherwise

	function is_kmemleak
	{
	is_linux && [ -e /sys/kernel/debug/kmemleak ]
	}

	# Determine whether a dataset is mounted
	#
	# $1 dataset name
	# $2 filesystem type; optional - defaulted to zfs
	#
	# Return 0 if dataset is mounted; 1 if unmounted; 2 on error

	function ismounted
	{
	typeset fstype=$2
	[[ -z $fstype ]] && fstype=zfs
	typeset out dir name

	case $fstype in
	zfs)
	if [[ "$1" == "/"* ]] ; then
	! zfs mount \| awk -v fs="$1" '$2 == fs {exit 1}'
	else
	! zfs mount \| awk -v ds="$1" '$1 == ds {exit 1}'
	fi
	;;
	ufs\|nfs)
	if is_freebsd; then
	mount -pt $fstype \| while read dev dir _t _flags; do
	[[ "$1" == "$dev" \|\| "$1" == "$dir" ]] && return 0
	done
	else
	out=$(df -F $fstype $1 2>/dev/null) \|\| return

	dir=${out%%\(*}
	dir=${dir%% *}
	name=${out##*\(}
	name=${name%%\)*}
	name=${name%% *}

	[[ "$1" == "$dir" \|\| "$1" == "$name" ]] && return 0
	fi
	;;
	ext*)
	df -t $fstype $1 > /dev/null 2>&1
	;;
	zvol)
	if [[ -L "$ZVOL_DEVDIR/$1" ]]; then
	link=$(readlink -f $ZVOL_DEVDIR/$1)
	[[ -n "$link" ]] && \
	mount \| grep -q "^$link" && \
	return 0
	fi
	;;
	*)
	false
	;;
	esac
	}

	# Return 0 if a dataset is mounted; 1 otherwise
	#
	# $1 dataset name
	# $2 filesystem type; optional - defaulted to zfs

	function mounted
	{
	ismounted $1 $2
	}

	# Return 0 if a dataset is unmounted; 1 otherwise
	#
	# $1 dataset name
	# $2 filesystem type; optional - defaulted to zfs

	function unmounted
	{
	! ismounted $1 $2
	}

	function default_setup
	{
	default_setup_noexit "$@"

	log_pass
	}

	function default_setup_no_mountpoint
	{
	default_setup_noexit "$1" "$2" "$3" "yes"

	log_pass
	}

	#
	# Given a list of disks, setup storage pools and datasets.
	#
	function default_setup_noexit
	{
	typeset disklist=$1
	typeset container=$2
	typeset volume=$3
	typeset no_mountpoint=$4
	log_note begin default_setup_noexit

	if is_global_zone; then
	if poolexists $TESTPOOL ; then
	destroy_pool $TESTPOOL
	fi
	[[ -d /$TESTPOOL ]] && rm -rf /$TESTPOOL
	log_must zpool create -f $TESTPOOL $disklist
	else
	reexport_pool
	fi

	rm -rf $TESTDIR \|\| log_unresolved Could not remove $TESTDIR
	mkdir -p $TESTDIR \|\| log_unresolved Could not create $TESTDIR

	log_must zfs create $TESTPOOL/$TESTFS
	if [[ -z $no_mountpoint ]]; then
	log_must zfs set mountpoint=$TESTDIR $TESTPOOL/$TESTFS
	fi

	if [[ -n $container ]]; then
	rm -rf $TESTDIR1 \|\| \
	log_unresolved Could not remove $TESTDIR1
	mkdir -p $TESTDIR1 \|\| \
	log_unresolved Could not create $TESTDIR1

	log_must zfs create $TESTPOOL/$TESTCTR
	log_must zfs set canmount=off $TESTPOOL/$TESTCTR
	log_must zfs create $TESTPOOL/$TESTCTR/$TESTFS1
	if [[ -z $no_mountpoint ]]; then
	log_must zfs set mountpoint=$TESTDIR1 \
	$TESTPOOL/$TESTCTR/$TESTFS1
	fi
	fi

	if [[ -n $volume ]]; then
	if is_global_zone ; then
	log_must zfs create -V $VOLSIZE $TESTPOOL/$TESTVOL
	block_device_wait
	else
	log_must zfs create $TESTPOOL/$TESTVOL
	fi
	fi
	}

	#
	# Given a list of disks, setup a storage pool, file system and
	# a container.
	#
	function default_container_setup
	{
	typeset disklist=$1

	default_setup "$disklist" "true"
	}

	#
	# Given a list of disks, setup a storage pool,file system
	# and a volume.
	#
	function default_volume_setup
	{
	typeset disklist=$1

	default_setup "$disklist" "" "true"
	}

	#
	# Given a list of disks, setup a storage pool,file system,
	# a container and a volume.
	#
	function default_container_volume_setup
	{
	typeset disklist=$1

	default_setup "$disklist" "true" "true"
	}

	#
	# Create a snapshot on a filesystem or volume. Defaultly create a snapshot on
	# filesystem
	#
	# $1 Existing filesystem or volume name. Default, $TESTPOOL/$TESTFS
	# $2 snapshot name. Default, $TESTSNAP
	#
	function create_snapshot
	{
	typeset fs_vol=${1:-$TESTPOOL/$TESTFS}
	typeset snap=${2:-$TESTSNAP}

	[[ -z $fs_vol ]] && log_fail "Filesystem or volume's name is undefined."
	[[ -z $snap ]] && log_fail "Snapshot's name is undefined."

	if snapexists $fs_vol@$snap; then
	log_fail "$fs_vol@$snap already exists."
	fi
	datasetexists $fs_vol \|\| \
	log_fail "$fs_vol must exist."

	log_must zfs snapshot $fs_vol@$snap
	}

	#
	# Create a clone from a snapshot, default clone name is $TESTCLONE.
	#
	# $1 Existing snapshot, $TESTPOOL/$TESTFS@$TESTSNAP is default.
	# $2 Clone name, $TESTPOOL/$TESTCLONE is default.
	#
	function create_clone # snapshot clone
	{
	typeset snap=${1:-$TESTPOOL/$TESTFS@$TESTSNAP}
	typeset clone=${2:-$TESTPOOL/$TESTCLONE}

	[[ -z $snap ]] && \
	log_fail "Snapshot name is undefined."
	[[ -z $clone ]] && \
	log_fail "Clone name is undefined."

	log_must zfs clone $snap $clone
	}

	#
	# Create a bookmark of the given snapshot. Defaultly create a bookmark on
	# filesystem.
	#
	# $1 Existing filesystem or volume name. Default, $TESTFS
	# $2 Existing snapshot name. Default, $TESTSNAP
	# $3 bookmark name. Default, $TESTBKMARK
	#
	function create_bookmark
	{
	typeset fs_vol=${1:-$TESTFS}
	typeset snap=${2:-$TESTSNAP}
	typeset bkmark=${3:-$TESTBKMARK}

	[[ -z $fs_vol ]] && log_fail "Filesystem or volume's name is undefined."
	[[ -z $snap ]] && log_fail "Snapshot's name is undefined."
	[[ -z $bkmark ]] && log_fail "Bookmark's name is undefined."

	if bkmarkexists $fs_vol#$bkmark; then
	log_fail "$fs_vol#$bkmark already exists."
	fi
	datasetexists $fs_vol \|\| \
	log_fail "$fs_vol must exist."
	snapexists $fs_vol@$snap \|\| \
	log_fail "$fs_vol@$snap must exist."

	log_must zfs bookmark $fs_vol@$snap $fs_vol#$bkmark
	}

	#
	# Create a temporary clone result of an interrupted resumable 'zfs receive'
	# $1 Destination filesystem name. Must not exist, will be created as the result
	# of this function along with its %recv temporary clone
	# $2 Source filesystem name. Must not exist, will be created and destroyed
	#
	function create_recv_clone
	{
	typeset recvfs="$1"
	typeset sendfs="${2:-$TESTPOOL/create_recv_clone}"
	typeset snap="$sendfs@snap1"
	typeset incr="$sendfs@snap2"
	typeset mountpoint="$TESTDIR/create_recv_clone"
	typeset sendfile="$TESTDIR/create_recv_clone.zsnap"

	[[ -z $recvfs ]] && log_fail "Recv filesystem's name is undefined."

	datasetexists $recvfs && log_fail "Recv filesystem must not exist."
	datasetexists $sendfs && log_fail "Send filesystem must not exist."

	log_must zfs create -o compression=off -o mountpoint="$mountpoint" $sendfs
	log_must zfs snapshot $snap
	log_must eval "zfs send $snap \| zfs recv -u $recvfs"
	log_must mkfile 1m "$mountpoint/data"
	log_must zfs snapshot $incr
	log_must eval "zfs send -i $snap $incr \| dd bs=10K count=1 \
	iflag=fullblock > $sendfile"
	log_mustnot eval "zfs recv -su $recvfs < $sendfile"
	destroy_dataset "$sendfs" "-r"
	log_must rm -f "$sendfile"

	if [[ $(get_prop 'inconsistent' "$recvfs/%recv") -ne 1 ]]; then
	log_fail "Error creating temporary $recvfs/%recv clone"
	fi
	}

	function default_mirror_setup
	{
	default_mirror_setup_noexit $1 $2 $3

	log_pass
	}

	#
	# Given a pair of disks, set up a storage pool and dataset for the mirror
	# @parameters: $1 the primary side of the mirror
	# $2 the secondary side of the mirror
	# @uses: ZPOOL ZFS TESTPOOL TESTFS
	function default_mirror_setup_noexit
	{
	readonly func="default_mirror_setup_noexit"
	typeset primary=$1
	typeset secondary=$2

	[[ -z $primary ]] && \
	log_fail "$func: No parameters passed"
	[[ -z $secondary ]] && \
	log_fail "$func: No secondary partition passed"
	[[ -d /$TESTPOOL ]] && rm -rf /$TESTPOOL
	log_must zpool create -f $TESTPOOL mirror $@
	log_must zfs create $TESTPOOL/$TESTFS
	log_must zfs set mountpoint=$TESTDIR $TESTPOOL/$TESTFS
	}

	#
	# Destroy the configured testpool mirrors.
	# the mirrors are of the form ${TESTPOOL}{number}
	# @uses: ZPOOL ZFS TESTPOOL
	function destroy_mirrors
	{
	default_cleanup_noexit

	log_pass
	}

	function default_raidz_setup
	{
	default_raidz_setup_noexit "$*"

	log_pass
	}

	#
	# Given a minimum of two disks, set up a storage pool and dataset for the raid-z
	# $1 the list of disks
	#
	function default_raidz_setup_noexit
	{
	typeset disklist="$*"
	disks=(${disklist[*]})

	if [[ ${#disks[*]} -lt 2 ]]; then
	log_fail "A raid-z requires a minimum of two disks."
	fi

	[[ -d /$TESTPOOL ]] && rm -rf /$TESTPOOL
	log_must zpool create -f $TESTPOOL raidz $disklist
	log_must zfs create $TESTPOOL/$TESTFS
	log_must zfs set mountpoint=$TESTDIR $TESTPOOL/$TESTFS
	}

	#
	# Common function used to cleanup storage pools and datasets.
	#
	# Invoked at the start of the test suite to ensure the system
	# is in a known state, and also at the end of each set of
	# sub-tests to ensure errors from one set of tests doesn't
	# impact the execution of the next set.

	function default_cleanup
	{
	default_cleanup_noexit

	log_pass
	}

	#
	# Utility function used to list all available pool names.
	#
	# NOTE: $KEEP is a variable containing pool names, separated by a newline
	# character, that must be excluded from the returned list.
	#
	function get_all_pools
	{
	zpool list -H -o name \| grep -Fvx "$KEEP" \| grep -v "$NO_POOLS"
	}

	function default_cleanup_noexit
	{
	typeset pool=""
	#
	# Destroying the pool will also destroy any
	# filesystems it contains.
	#
	if is_global_zone; then
	zfs unmount -a > /dev/null 2>&1
	ALL_POOLS=$(get_all_pools)
	# Here, we loop through the pools we're allowed to
	# destroy, only destroying them if it's safe to do
	# so.
	while [ ! -z ${ALL_POOLS} ]
	do
	for pool in ${ALL_POOLS}
	do
	if safe_to_destroy_pool $pool ;
	then
	destroy_pool $pool
	fi
	done
	ALL_POOLS=$(get_all_pools)
	done

	zfs mount -a
	else
	typeset fs=""
	for fs in $(zfs list -H -o name \
	\| grep "^$ZONE_POOL/$ZONE_CTR[01234]/"); do
	destroy_dataset "$fs" "-Rf"
	done

	# Need cleanup here to avoid garbage dir left.
	for fs in $(zfs list -H -o name); do
	[[ $fs == /$ZONE_POOL ]] && continue
	[[ -d $fs ]] && log_must rm -rf $fs/*
	done

	#
	# Reset the $ZONE_POOL/$ZONE_CTR[01234] file systems property to
	# the default value
	#
	for fs in $(zfs list -H -o name); do
	if [[ $fs == $ZONE_POOL/$ZONE_CTR[01234] ]]; then
	log_must zfs set reservation=none $fs
	log_must zfs set recordsize=128K $fs
	log_must zfs set mountpoint=/$fs $fs
	typeset enc=$(get_prop encryption $fs)
	if [ -z "$enc" ] \|\| [ "$enc" = "off" ]; then
	log_must zfs set checksum=on $fs
	fi
	log_must zfs set compression=off $fs
	log_must zfs set atime=on $fs
	log_must zfs set devices=off $fs
	log_must zfs set exec=on $fs
	log_must zfs set setuid=on $fs
	log_must zfs set readonly=off $fs
	log_must zfs set snapdir=hidden $fs
	log_must zfs set aclmode=groupmask $fs
	log_must zfs set aclinherit=secure $fs
	fi
	done
	fi

	[[ -d $TESTDIR ]] && \
	log_must rm -rf $TESTDIR

	disk1=${DISKS%% *}
	if is_mpath_device $disk1; then
	delete_partitions
	fi

	rm -f $TEST_BASE_DIR/{err,out}
	}


	#
	# Common function used to cleanup storage pools, file systems
	# and containers.
	#
	function default_container_cleanup
	{
	if ! is_global_zone; then
	reexport_pool
	fi

	ismounted $TESTPOOL/$TESTCTR/$TESTFS1 &&
	log_must zfs unmount $TESTPOOL/$TESTCTR/$TESTFS1

	destroy_dataset "$TESTPOOL/$TESTCTR/$TESTFS1" "-R"
	destroy_dataset "$TESTPOOL/$TESTCTR" "-Rf"

	[[ -e $TESTDIR1 ]] && \
	log_must rm -rf $TESTDIR1

	default_cleanup
	}

	#
	# Common function used to cleanup snapshot of file system or volume. Default to
	# delete the file system's snapshot
	#
	# $1 snapshot name
	#
	function destroy_snapshot
	{
	typeset snap=${1:-$TESTPOOL/$TESTFS@$TESTSNAP}

	if ! snapexists $snap; then
	log_fail "'$snap' does not exist."
	fi

	#
	# For the sake of the value which come from 'get_prop' is not equal
	# to the really mountpoint when the snapshot is unmounted. So, firstly
	# check and make sure this snapshot's been mounted in current system.
	#
	typeset mtpt=""
	if ismounted $snap; then
	mtpt=$(get_prop mountpoint $snap)
	fi

	destroy_dataset "$snap"
	[[ $mtpt != "" && -d $mtpt ]] && \
	log_must rm -rf $mtpt
	}

	#
	# Common function used to cleanup clone.
	#
	# $1 clone name
	#
	function destroy_clone
	{
	typeset clone=${1:-$TESTPOOL/$TESTCLONE}

	if ! datasetexists $clone; then
	log_fail "'$clone' does not existed."
	fi

	# With the same reason in destroy_snapshot
	typeset mtpt=""
	if ismounted $clone; then
	mtpt=$(get_prop mountpoint $clone)
	fi

	destroy_dataset "$clone"
	[[ $mtpt != "" && -d $mtpt ]] && \
	log_must rm -rf $mtpt
	}

	#
	# Common function used to cleanup bookmark of file system or volume. Default
	# to delete the file system's bookmark.
	#
	# $1 bookmark name
	#
	function destroy_bookmark
	{
	typeset bkmark=${1:-$TESTPOOL/$TESTFS#$TESTBKMARK}

	if ! bkmarkexists $bkmark; then
	log_fail "'$bkmarkp' does not existed."
	fi

	destroy_dataset "$bkmark"
	}

	# Return 0 if a snapshot exists; $? otherwise
	#
	# $1 - snapshot name

	function snapexists
	{
	zfs list -H -t snapshot "$1" > /dev/null 2>&1
	}

	#
	# Return 0 if a bookmark exists; $? otherwise
	#
	# $1 - bookmark name
	#
	function bkmarkexists
	{
	zfs list -H -t bookmark "$1" > /dev/null 2>&1
	}

	#
	# Return 0 if a hold exists; $? otherwise
	#
	# $1 - hold tag
	# $2 - snapshot name
	#
	function holdexists
	{
	! zfs holds "$2" \| awk -v t="$1" '$2 ~ t { exit 1 }'
	}

	#
	# Set a property to a certain value on a dataset.
	# Sets a property of the dataset to the value as passed in.
	# @param:
	# $1 dataset who's property is being set
	# $2 property to set
	# $3 value to set property to
	# @return:
	# 0 if the property could be set.
	# non-zero otherwise.
	# @use: ZFS
	#
	function dataset_setprop
	{
	typeset fn=dataset_setprop

	if (($# < 3)); then
	log_note "$fn: Insufficient parameters (need 3, had $#)"
	return 1
	fi
	typeset output=
	output=$(zfs set $2=$3 $1 2>&1)
	typeset rv=$?
	if ((rv != 0)); then
	log_note "Setting property on $1 failed."
	log_note "property $2=$3"
	log_note "Return Code: $rv"
	log_note "Output: $output"
	return $rv
	fi
	return 0
	}

	#
	# Check a numeric assertion
	# @parameter: $@ the assertion to check
	# @output: big loud notice if assertion failed
	# @use: log_fail
	#
	function assert
	{
	(($@)) \|\| log_fail "$@"
	}

	#
	# Function to format partition size of a disk
	# Given a disk cxtxdx reduces all partitions
	# to 0 size
	#
	function zero_partitions #<whole_disk_name>
	{
	typeset diskname=$1
	typeset i

	if is_freebsd; then
	gpart destroy -F $diskname
	elif is_linux; then
	DSK=$DEV_DSKDIR/$diskname
	DSK=$(echo $DSK \| sed -e "s\|//\|/\|g")
	log_must parted $DSK -s -- mklabel gpt
	blockdev --rereadpt $DSK 2>/dev/null
	block_device_wait
	else
	for i in 0 1 3 4 5 6 7
	do
	log_must set_partition $i "" 0mb $diskname
	done
	fi

	return 0
	}

	#
	# Given a slice, size and disk, this function
	# formats the slice to the specified size.
	# Size should be specified with units as per
	# the `format` command requirements eg. 100mb 3gb
	#
	# NOTE: This entire interface is problematic for the Linux parted utility
	# which requires the end of the partition to be specified. It would be
	# best to retire this interface and replace it with something more flexible.
	# At the moment a best effort is made.
	#
	# arguments: <slice_num> <slice_start> <size_plus_units> <whole_disk_name>
	function set_partition
	{
	typeset -i slicenum=$1
	typeset start=$2
	typeset size=$3
	typeset disk=${4#$DEV_DSKDIR/}
	disk=${disk#$DEV_RDSKDIR/}

	case "$UNAME" in
	Linux)
	if [[ -z $size \|\| -z $disk ]]; then
	log_fail "The size or disk name is unspecified."
	fi
	disk=$DEV_DSKDIR/$disk
	typeset size_mb=${size%%[mMgG]}

	size_mb=${size_mb%%[mMgG][bB]}
	if [[ ${size:1:1} == 'g' ]]; then
	((size_mb = size_mb * 1024))
	fi

	# Create GPT partition table when setting slice 0 or
	# when the device doesn't already contain a GPT label.
	parted $disk -s -- print 1 >/dev/null
	typeset ret_val=$?
	if [[ $slicenum -eq 0 \|\| $ret_val -ne 0 ]]; then
	if ! parted $disk -s -- mklabel gpt; then
	log_note "Failed to create GPT partition table on $disk"
	return 1
	fi
	fi

	# When no start is given align on the first cylinder.
	if [[ -z "$start" ]]; then
	start=1
	fi

	# Determine the cylinder size for the device and using
	# that calculate the end offset in cylinders.
	typeset -i cly_size_kb=0
	cly_size_kb=$(parted -m $disk -s -- unit cyl print \|
	awk -F '[:k.]' 'NR == 3 {print $4}')
	((end = (size_mb * 1024 / cly_size_kb) + start))

	parted $disk -s -- \
	mkpart part$slicenum ${start}cyl ${end}cyl
	typeset ret_val=$?
	if [[ $ret_val -ne 0 ]]; then
	log_note "Failed to create partition $slicenum on $disk"
	return 1
	fi

	blockdev --rereadpt $disk 2>/dev/null
	block_device_wait $disk
	;;
	FreeBSD)
	if [[ -z $size \|\| -z $disk ]]; then
	log_fail "The size or disk name is unspecified."
	fi
	disk=$DEV_DSKDIR/$disk

	if [[ $slicenum -eq 0 ]] \|\| ! gpart show $disk >/dev/null 2>&1; then
	gpart destroy -F $disk >/dev/null 2>&1
	if ! gpart create -s GPT $disk; then
	log_note "Failed to create GPT partition table on $disk"
	return 1
	fi
	fi

	typeset index=$((slicenum + 1))

	if [[ -n $start ]]; then
	start="-b $start"
	fi
	gpart add -t freebsd-zfs $start -s $size -i $index $disk
	if [[ $ret_val -ne 0 ]]; then
	log_note "Failed to create partition $slicenum on $disk"
	return 1
	fi

	block_device_wait $disk
	;;
	*)
	if [[ -z $slicenum \|\| -z $size \|\| -z $disk ]]; then
	log_fail "The slice, size or disk name is unspecified."
	fi

	typeset format_file=/var/tmp/format_in.$$

	echo "partition" >$format_file
	echo "$slicenum" >> $format_file
	echo "" >> $format_file
	echo "" >> $format_file
	echo "$start" >> $format_file
	echo "$size" >> $format_file
	echo "label" >> $format_file
	echo "" >> $format_file
	echo "q" >> $format_file
	echo "q" >> $format_file

	format -e -s -d $disk -f $format_file
	typeset ret_val=$?
	rm -f $format_file
	;;
	esac

	if [[ $ret_val -ne 0 ]]; then
	log_note "Unable to format $disk slice $slicenum to $size"
	return 1
	fi
	return 0
	}

	#
	# Delete all partitions on all disks - this is specifically for the use of multipath
	# devices which currently can only be used in the test suite as raw/un-partitioned
	# devices (ie a zpool cannot be created on a whole mpath device that has partitions)
	#
	function delete_partitions
	{
	typeset disk

	if [[ -z $DISKSARRAY ]]; then
	DISKSARRAY=$DISKS
	fi

	if is_linux; then
	typeset -i part
	for disk in $DISKSARRAY; do
	for (( part = 1; part < MAX_PARTITIONS; part++ )); do
	typeset partition=${disk}${SLICE_PREFIX}${part}
	parted $DEV_DSKDIR/$disk -s rm $part > /dev/null 2>&1
	if lsblk \| grep -qF ${partition}; then
	log_fail "Partition ${partition} not deleted"
	else
	log_note "Partition ${partition} deleted"
	fi
	done
	done
	elif is_freebsd; then
	for disk in $DISKSARRAY; do
	if gpart destroy -F $disk; then
	log_note "Partitions for ${disk} deleted"
	else
	log_fail "Partitions for ${disk} not deleted"
	fi
	done
	fi
	}

	#
	# Get the end cyl of the given slice
	#
	function get_endslice #<disk> <slice>
	{
	typeset disk=$1
	typeset slice=$2
	if [[ -z $disk \|\| -z $slice ]] ; then
	log_fail "The disk name or slice number is unspecified."
	fi

	case "$UNAME" in
	Linux)
	endcyl=$(parted -s $DEV_DSKDIR/$disk -- unit cyl print \| \
	awk "/part${slice}/"' {sub(/cyl/, "", $3); print $3}')
	((endcyl = (endcyl + 1)))
	;;
	FreeBSD)
	disk=${disk#/dev/zvol/}
	disk=${disk%p*}
	slice=$((slice + 1))
	endcyl=$(gpart show $disk \| \
	awk -v slice=$slice '$3 == slice { print $1 + $2 }')
	;;
	*)
	disk=${disk#/dev/dsk/}
	disk=${disk#/dev/rdsk/}
	disk=${disk%s*}

	typeset -i ratio=0
	ratio=$(prtvtoc /dev/rdsk/${disk}s2 \| \
	awk '/sectors\/cylinder/ {print $2}')

	if ((ratio == 0)); then
	return
	fi

	typeset -i endcyl=$(prtvtoc -h /dev/rdsk/${disk}s2 \|
	awk -v token="$slice" '$1 == token {print $6}')

	((endcyl = (endcyl + 1) / ratio))
	;;
	esac

	echo $endcyl
	}


	#
	# Given a size,disk and total slice number, this function formats the
	# disk slices from 0 to the total slice number with the same specified
	# size.
	#
	function partition_disk #<slice_size> <whole_disk_name> <total_slices>
	{
	typeset -i i=0
	typeset slice_size=$1
	typeset disk_name=$2
	typeset total_slices=$3
	typeset cyl

	zero_partitions $disk_name
	while ((i < $total_slices)); do
	if ! is_linux; then
	if ((i == 2)); then
	((i = i + 1))
	continue
	fi
	fi
	log_must set_partition $i "$cyl" $slice_size $disk_name
	cyl=$(get_endslice $disk_name $i)
	((i = i+1))
	done
	}

	#
	# This function continues to write to a filenum number of files into dirnum
	# number of directories until either file_write returns an error or the
	# maximum number of files per directory have been written.
	#
	# Usage:
	# fill_fs [destdir] [dirnum] [filenum] [bytes] [num_writes] [data]
	#
	# Return value: 0 on success
	# non 0 on error
	#
	# Where :
	# destdir: is the directory where everything is to be created under
	# dirnum: the maximum number of subdirectories to use, -1 no limit
	# filenum: the maximum number of files per subdirectory
	# bytes: number of bytes to write
	# num_writes: number of types to write out bytes
	# data: the data that will be written
	#
	# E.g.
	# fill_fs /testdir 20 25 1024 256 0
	#
	# Note: bytes * num_writes equals the size of the testfile
	#
	function fill_fs # destdir dirnum filenum bytes num_writes data
	{
	typeset destdir=${1:-$TESTDIR}
	typeset -i dirnum=${2:-50}
	typeset -i filenum=${3:-50}
	typeset -i bytes=${4:-8192}
	typeset -i num_writes=${5:-10240}
	typeset data=${6:-0}

	mkdir -p $destdir/{1..$dirnum}
	for f in $destdir/{1..$dirnum}/$TESTFILE{1..$filenum}; do
	file_write -o create -f $f -b $bytes -c $num_writes -d $data \
	\|\| return
	done
	}

	# Get the specified dataset property in parsable format or fail
	function get_prop # property dataset
	{
	typeset prop=$1
	typeset dataset=$2

	zfs get -Hpo value "$prop" "$dataset" \|\| log_fail "zfs get $prop $dataset"
	}

	# Get the specified pool property in parsable format or fail
	function get_pool_prop # property pool
	{
	typeset prop=$1
	typeset pool=$2

	zpool get -Hpo value "$prop" "$pool" \|\| log_fail "zpool get $prop $pool"
	}

	# Return 0 if a pool exists; $? otherwise
	#
	# $1 - pool name

	function poolexists
	{
	typeset pool=$1

	if [[ -z $pool ]]; then
	log_note "No pool name given."
	return 1
	fi

	zpool get name "$pool" > /dev/null 2>&1
	}

	# Return 0 if all the specified datasets exist; $? otherwise
	#
	# $1-n dataset name
	function datasetexists
	{
	if (($# == 0)); then
	log_note "No dataset name given."
	return 1
	fi

	zfs get name "$@" > /dev/null 2>&1
	}

	# return 0 if none of the specified datasets exists, otherwise return 1.
	#
	# $1-n dataset name
	function datasetnonexists
	{
	if (($# == 0)); then
	log_note "No dataset name given."
	return 1
	fi

	while (($# > 0)); do
	zfs list -H -t filesystem,snapshot,volume $1 > /dev/null 2>&1 \
	&& return 1
	shift
	done

	return 0
	}

	# FreeBSD breaks exports(5) at whitespace and doesn't process escapes
	# Solaris just breaks
	#
	# cf. https://github.com/openzfs/zfs/pull/13165#issuecomment-1059845807
	#
	# Linux can have spaces (which are \OOO-escaped),
	# but can't have backslashes because they're parsed recursively
	function shares_can_have_whitespace
	{
	is_linux
	}

	function is_shared_freebsd
	{
	typeset fs=$1

	pgrep -q mountd && showmount -E \| grep -qx "$fs"
	}

	function is_shared_illumos
	{
	typeset fs=$1
	typeset mtpt

	for mtpt in `share \| awk '{print $2}'` ; do
	if [[ $mtpt == $fs ]] ; then
	return 0
	fi
	done

	typeset stat=$(svcs -H -o STA nfs/server:default)
	if [[ $stat != "ON" ]]; then
	log_note "Current nfs/server status: $stat"
	fi

	return 1
	}

	function is_shared_linux
	{
	typeset fs=$1
	! exportfs -s \| awk -v fs="${fs//\\/\\\\}" '/^\// && $1 == fs {exit 1}'
	}

	#
	# Given a mountpoint, or a dataset name, determine if it is shared via NFS.
	#
	# Returns 0 if shared, 1 otherwise.
	#
	function is_shared
	{
	typeset fs=$1
	typeset mtpt

	if [[ $fs != "/"* ]] ; then
	if datasetnonexists "$fs" ; then
	return 1
	else
	mtpt=$(get_prop mountpoint "$fs")
	case "$mtpt" in
	none\|legacy\|-) return 1
	;;
	*) fs=$mtpt
	;;
	esac
	fi
	fi

	case "$UNAME" in
	FreeBSD) is_shared_freebsd "$fs" ;;
	Linux) is_shared_linux "$fs" ;;
	*) is_shared_illumos "$fs" ;;
	esac
	}

	function is_exported_illumos
	{
	typeset fs=$1
	typeset mtpt _

	while read -r mtpt _; do
	[ "$mtpt" = "$fs" ] && return
	done < /etc/dfs/sharetab

	return 1
	}

	function is_exported_freebsd
	{
	typeset fs=$1
	typeset mtpt _

	while read -r mtpt _; do
	[ "$mtpt" = "$fs" ] && return
	done < /etc/zfs/exports

	return 1
	}

	function is_exported_linux
	{
	typeset fs=$1
	typeset mtpt _

	while read -r mtpt _; do
	[ "$(printf "$mtpt")" = "$fs" ] && return
	done < /etc/exports.d/zfs.exports

	return 1
	}

	#
	# Given a mountpoint, or a dataset name, determine if it is exported via
	# the os-specific NFS exports file.
	#
	# Returns 0 if exported, 1 otherwise.
	#
	function is_exported
	{
	typeset fs=$1
	typeset mtpt

	if [[ $fs != "/"* ]] ; then
	if datasetnonexists "$fs" ; then
	return 1
	else
	mtpt=$(get_prop mountpoint "$fs")
	case $mtpt in
	none\|legacy\|-) return 1
	;;
	*) fs=$mtpt
	;;
	esac
	fi
	fi

	case "$UNAME" in
	FreeBSD) is_exported_freebsd "$fs" ;;
	Linux) is_exported_linux "$fs" ;;
	*) is_exported_illumos "$fs" ;;
	esac
	}

	#
	# Given a dataset name determine if it is shared via SMB.
	#
	# Returns 0 if shared, 1 otherwise.
	#
	function is_shared_smb
	{
	typeset fs=$1

	datasetexists "$fs" \|\| return

	if is_linux; then
	net usershare list \| grep -xFq "${fs//[-\/]/_}"
	else
	log_note "SMB on $UNAME currently unsupported by the test framework"
	return 1
	fi
	}

	#
	# Given a mountpoint, determine if it is not shared via NFS.
	#
	# Returns 0 if not shared, 1 otherwise.
	#
	function not_shared
	{
	! is_shared $1
	}

	#
	# Given a dataset determine if it is not shared via SMB.
	#
	# Returns 0 if not shared, 1 otherwise.
	#
	function not_shared_smb
	{
	! is_shared_smb $1
	}

	#
	# Helper function to unshare a mountpoint.
	#
	function unshare_fs #fs
	{
	typeset fs=$1

	if is_shared $fs \|\| is_shared_smb $fs; then
	log_must zfs unshare $fs
	fi
	}

	#
	# Helper function to share a NFS mountpoint.
	#
	function share_nfs #fs
	{
	typeset fs=$1

	is_shared "$fs" && return

	case "$UNAME" in
	Linux)
	log_must exportfs "*:$fs"
	;;
	FreeBSD)
	typeset mountd
	read -r mountd < /var/run/mountd.pid
	log_must eval "printf '%s\t\n' \"$fs\" >> /etc/zfs/exports"
	log_must kill -s HUP "$mountd"
	;;
	*)
	log_must share -F nfs "$fs"
	;;
	esac

	return 0
	}

	#
	# Helper function to unshare a NFS mountpoint.
	#
	function unshare_nfs #fs
	{
	typeset fs=$1

	! is_shared "$fs" && return

	case "$UNAME" in
	Linux)
	log_must exportfs -u "*:$fs"
	;;
	FreeBSD)
	typeset mountd
	read -r mountd < /var/run/mountd.pid
	awk -v fs="${fs//\\/\\\\}" '$1 != fs' /etc/zfs/exports > /etc/zfs/exports.$$
	log_must mv /etc/zfs/exports.$$ /etc/zfs/exports
	log_must kill -s HUP "$mountd"
	;;
	*)
	log_must unshare -F nfs $fs
	;;
	esac

	return 0
	}

	#
	# Helper function to show NFS shares.
	#
	function showshares_nfs
	{
	case "$UNAME" in
	Linux)
	exportfs -v
	;;
	FreeBSD)
	showmount
	;;
	*)
	share -F nfs
	;;
	esac
	}

	function check_nfs
	{
	case "$UNAME" in
	Linux)
	exportfs -s
	;;
	FreeBSD)
	showmount -e
	;;
	*)
	log_unsupported "Unknown platform"
	;;
	esac \|\| log_unsupported "The NFS utilities are not installed"
	}

	#
	# Check NFS server status and trigger it online.
	#
	function setup_nfs_server
	{
	# Cannot share directory in non-global zone.
	#
	if ! is_global_zone; then
	log_note "Cannot trigger NFS server by sharing in LZ."
	return
	fi

	if is_linux; then
	#
	# Re-synchronize /var/lib/nfs/etab with /etc/exports and
	# /etc/exports.d./* to provide a clean test environment.
	#
	log_must exportfs -r

	log_note "NFS server must be started prior to running ZTS."
	return
	elif is_freebsd; then
	log_must kill -s HUP $(</var/run/mountd.pid)

	log_note "NFS server must be started prior to running ZTS."
	return
	fi

	typeset nfs_fmri="svc:/network/nfs/server:default"
	if [[ $(svcs -Ho STA $nfs_fmri) != "ON" ]]; then
	#
	# Only really sharing operation can enable NFS server
	# to online permanently.
	#
	typeset dummy=/tmp/dummy

	if [[ -d $dummy ]]; then
	log_must rm -rf $dummy
	fi

	log_must mkdir $dummy
	log_must share $dummy

	#
	# Waiting for fmri's status to be the final status.
	# Otherwise, in transition, an asterisk (*) is appended for
	# instances, unshare will reverse status to 'DIS' again.
	#
	# Waiting for 1's at least.
	#
	log_must sleep 1
	timeout=10
	while [[ timeout -ne 0 && $(svcs -Ho STA $nfs_fmri) == '' ]]
	do
	log_must sleep 1

	((timeout -= 1))
	done

	log_must unshare $dummy
	log_must rm -rf $dummy
	fi

	log_note "Current NFS status: '$(svcs -Ho STA,FMRI $nfs_fmri)'"
	}

	#
	# To verify whether calling process is in global zone
	#
	# Return 0 if in global zone, 1 in non-global zone
	#
	function is_global_zone
	{
	if is_linux \|\| is_freebsd; then
	return 0
	else
	typeset cur_zone=$(zonename 2>/dev/null)
	[ $cur_zone = "global" ]
	fi
	}

	#
	# Verify whether test is permitted to run from
	# global zone, local zone, or both
	#
	# $1 zone limit, could be "global", "local", or "both"(no limit)
	#
	# Return 0 if permitted, otherwise exit with log_unsupported
	#
	function verify_runnable # zone limit
	{
	typeset limit=$1

	[[ -z $limit ]] && return 0

	if is_global_zone ; then
	case $limit in
	global\|both)
	;;
	local) log_unsupported "Test is unable to run from "\
	"global zone."
	;;
	*) log_note "Warning: unknown limit $limit - " \
	"use both."
	;;
	esac
	else
	case $limit in
	local\|both)
	;;
	global) log_unsupported "Test is unable to run from "\
	"local zone."
	;;
	*) log_note "Warning: unknown limit $limit - " \
	"use both."
	;;
	esac

	reexport_pool
	fi

	return 0
	}

	# Return 0 if create successfully or the pool exists; $? otherwise
	# Note: In local zones, this function should return 0 silently.
	#
	# $1 - pool name
	# $2-n - [keyword] devs_list

	function create_pool #pool devs_list
	{
	typeset pool=${1%%/*}

	shift

	if [[ -z $pool ]]; then
	log_note "Missing pool name."
	return 1
	fi

	if poolexists $pool ; then
	destroy_pool $pool
	fi

	if is_global_zone ; then
	[[ -d /$pool ]] && rm -rf /$pool
	log_must zpool create -f $pool $@
	fi

	return 0
	}

	# Return 0 if destroy successfully or the pool exists; $? otherwise
	# Note: In local zones, this function should return 0 silently.
	#
	# $1 - pool name
	# Destroy pool with the given parameters.

	function destroy_pool #pool
	{
	typeset pool=${1%%/*}
	typeset mtpt

	if [[ -z $pool ]]; then
	log_note "No pool name given."
	return 1
	fi

	if is_global_zone ; then
	if poolexists "$pool" ; then
	mtpt=$(get_prop mountpoint "$pool")

	# At times, syseventd/udev activity can cause attempts
	# to destroy a pool to fail with EBUSY. We retry a few
	# times allowing failures before requiring the destroy
	# to succeed.
	log_must_busy zpool destroy -f $pool

	[[ -d $mtpt ]] && \
	log_must rm -rf $mtpt
	else
	log_note "Pool does not exist. ($pool)"
	return 1
	fi
	fi

	return 0
	}

	# Return 0 if created successfully; $? otherwise
	#
	# $1 - dataset name
	# $2-n - dataset options

	function create_dataset #dataset dataset_options
	{
	typeset dataset=$1

	shift

	if [[ -z $dataset ]]; then
	log_note "Missing dataset name."
	return 1
	fi

	if datasetexists $dataset ; then
	destroy_dataset $dataset
	fi

	log_must zfs create $@ $dataset

	return 0
	}

	# Return 0 if destroy successfully or the dataset exists; $? otherwise
	# Note: In local zones, this function should return 0 silently.
	#
	# $1 - dataset name
	# $2 - custom arguments for zfs destroy
	# Destroy dataset with the given parameters.

	function destroy_dataset # dataset [args]
	{
	typeset dataset=$1
	typeset mtpt
	typeset args=${2:-""}

	if [[ -z $dataset ]]; then
	log_note "No dataset name given."
	return 1
	fi

	if is_global_zone ; then
	if datasetexists "$dataset" ; then
	mtpt=$(get_prop mountpoint "$dataset")
	log_must_busy zfs destroy $args $dataset

	[ -d $mtpt ] && log_must rm -rf $mtpt
	else
	log_note "Dataset does not exist. ($dataset)"
	return 1
	fi
	fi

	return 0
	}

	#
	# Reexport TESTPOOL & TESTPOOL(1-4)
	#
	function reexport_pool
	{
	typeset -i cntctr=5
	typeset -i i=0

	while ((i < cntctr)); do
	if ((i == 0)); then
	TESTPOOL=$ZONE_POOL/$ZONE_CTR$i
	if ! ismounted $TESTPOOL; then
	log_must zfs mount $TESTPOOL
	fi
	else
	eval TESTPOOL$i=$ZONE_POOL/$ZONE_CTR$i
	if eval ! ismounted \$TESTPOOL$i; then
	log_must eval zfs mount \$TESTPOOL$i
	fi
	fi
	((i += 1))
	done
	}

	#
	# Verify a given disk or pool state
	#
	# Return 0 is pool/disk matches expected state, 1 otherwise
	#
	function check_state # pool disk state{online,offline,degraded}
	{
	typeset pool=$1
	typeset disk=${2#$DEV_DSKDIR/}
	typeset state=$3

	[[ -z $pool ]] \|\| [[ -z $state ]] \
	&& log_fail "Arguments invalid or missing"

	if [[ -z $disk ]]; then
	#check pool state only
	zpool get -H -o value health $pool \| grep -qi "$state"
	else
	zpool status -v $pool \| grep "$disk" \| grep -qi "$state"
	fi
	}

	#
	# Get the mountpoint of snapshot
	# For the snapshot use <mp_filesystem>/.zfs/snapshot/<snap>
	# as its mountpoint
	#
	function snapshot_mountpoint
	{
	typeset dataset=${1:-$TESTPOOL/$TESTFS@$TESTSNAP}

	if [[ $dataset != @ ]]; then
	log_fail "Error name of snapshot '$dataset'."
	fi

	typeset fs=${dataset%@*}
	typeset snap=${dataset#*@}

	if [[ -z $fs \|\| -z $snap ]]; then
	log_fail "Error name of snapshot '$dataset'."
	fi

	echo $(get_prop mountpoint $fs)/.zfs/snapshot/$snap
	}

	#
	# Given a device and 'ashift' value verify it's correctly set on every label
	#
	function verify_ashift # device ashift
	{
	typeset device="$1"
	typeset ashift="$2"

	zdb -e -lll $device \| awk -v ashift=$ashift '
	/ashift: / {
	if (ashift != $2)
	exit 1;
	else
	count++;
	}
	END {
	exit (count != 4);
	}'
	}

	#
	# Given a pool and file system, this function will verify the file system
	# using the zdb internal tool. Note that the pool is exported and imported
	# to ensure it has consistent state.
	#
	function verify_filesys # pool filesystem dir
	{
	typeset pool="$1"
	typeset filesys="$2"
	typeset zdbout="/tmp/zdbout.$$"

	shift
	shift
	typeset dirs=$@
	typeset search_path=""

	log_note "Calling zdb to verify filesystem '$filesys'"
	zfs unmount -a > /dev/null 2>&1
	log_must zpool export $pool

	if [[ -n $dirs ]] ; then
	for dir in $dirs ; do
	search_path="$search_path -d $dir"
	done
	fi

	log_must zpool import $search_path $pool

	if ! zdb -cudi $filesys > $zdbout 2>&1; then
	log_note "Output: zdb -cudi $filesys"
	cat $zdbout
	rm -f $zdbout
	log_fail "zdb detected errors with: '$filesys'"
	fi

	log_must zfs mount -a
	log_must rm -rf $zdbout
	}

	#
	# Given a pool issue a scrub and verify that no checksum errors are reported.
	#
	function verify_pool
	{
	typeset pool=${1:-$TESTPOOL}

	log_must zpool scrub $pool
	log_must wait_scrubbed $pool

	typeset -i cksum=$(zpool status $pool \| awk '
	!NF { isvdev = 0 }
	isvdev { errors += $NF }
	/CKSUM$/ { isvdev = 1 }
	END { print errors }
	')
	if [[ $cksum != 0 ]]; then
	log_must zpool status -v
	log_fail "Unexpected CKSUM errors found on $pool ($cksum)"
	fi
	}

	#
	# Given a pool, and this function list all disks in the pool
	#
	function get_disklist # pool
	{
	echo $(zpool iostat -v $1 \| awk '(NR > 4) {print $1}' \| \
	grep -vEe '^-----' -e "^(mirror\|raidz[1-3]\|draid[1-3]\|spare\|log\|cache\|special\|dedup)\|\-[0-9]$")
	}

	#
	# Given a pool, and this function list all disks in the pool with their full
	# path (like "/dev/sda" instead of "sda").
	#
	function get_disklist_fullpath # pool
	{
	get_disklist "-P $1"
	}



	# /**
	# This function kills a given list of processes after a time period. We use
	# this in the stress tests instead of STF_TIMEOUT so that we can have processes
	# run for a fixed amount of time, yet still pass. Tests that hit STF_TIMEOUT
	# would be listed as FAIL, which we don't want : we're happy with stress tests
	# running for a certain amount of time, then finishing.
	#
	# @param $1 the time in seconds after which we should terminate these processes
	# @param $2..$n the processes we wish to terminate.
	# */
	function stress_timeout
	{
	typeset -i TIMEOUT=$1
	shift
	typeset cpids="$@"

	log_note "Waiting for child processes($cpids). " \
	"It could last dozens of minutes, please be patient ..."
	log_must sleep $TIMEOUT

	log_note "Killing child processes after ${TIMEOUT} stress timeout."
	typeset pid
	for pid in $cpids; do
	ps -p $pid > /dev/null 2>&1 &&
	log_must kill -USR1 $pid
	done
	}

	#
	# Verify a given hotspare disk is inuse or avail
	#
	# Return 0 is pool/disk matches expected state, 1 otherwise
	#
	function check_hotspare_state # pool disk state{inuse,avail}
	{
	typeset pool=$1
	typeset disk=${2#$DEV_DSKDIR/}
	typeset state=$3

	cur_state=$(get_device_state $pool $disk "spares")

	[ $state = $cur_state ]
	}

	#
	# Wait until a hotspare transitions to a given state or times out.
	#
	# Return 0 when pool/disk matches expected state, 1 on timeout.
	#
	function wait_hotspare_state # pool disk state timeout
	{
	typeset pool=$1
	typeset disk=${2#*$DEV_DSKDIR/}
	typeset state=$3
	typeset timeout=${4:-60}
	typeset -i i=0

	while [[ $i -lt $timeout ]]; do
	if check_hotspare_state $pool $disk $state; then
	return 0
	fi

	i=$((i+1))
	sleep 1
	done

	return 1
	}

	#
	# Verify a given vdev disk is inuse or avail
	#
	# Return 0 is pool/disk matches expected state, 1 otherwise
	#
	-function check_vdev_state # pool disk state{online,offline,unavail}
	+function check_vdev_state # pool disk state{online,offline,unavail,removed}
	{
	typeset pool=$1
	typeset disk=${2#*$DEV_DSKDIR/}
	typeset state=$3

	cur_state=$(get_device_state $pool $disk)

	[ $state = $cur_state ]
	}

	#
	# Wait until a vdev transitions to a given state or times out.
	#
	# Return 0 when pool/disk matches expected state, 1 on timeout.
	#
	function wait_vdev_state # pool disk state timeout
	{
	typeset pool=$1
	typeset disk=${2#*$DEV_DSKDIR/}
	typeset state=$3
	typeset timeout=${4:-60}
	typeset -i i=0

	while [[ $i -lt $timeout ]]; do
	if check_vdev_state $pool $disk $state; then
	return 0
	fi

	i=$((i+1))
	sleep 1
	done

	return 1
	}

	#
	# Check the output of 'zpool status -v <pool>',
	# and to see if the content of <token> contain the <keyword> specified.
	#
	# Return 0 is contain, 1 otherwise
	#
	function check_pool_status # pool token keyword <verbose>
	{
	typeset pool=$1
	typeset token=$2
	typeset keyword=$3
	typeset verbose=${4:-false}

	scan=$(zpool status -v "$pool" 2>/dev/null \| awk -v token="$token:" '$1==token')
	if [[ $verbose == true ]]; then
	log_note $scan
	fi
	echo $scan \| grep -qi "$keyword"
	}

	#
	# The following functions are instance of check_pool_status()
	# is_pool_resilvering - to check if the pool resilver is in progress
	# is_pool_resilvered - to check if the pool resilver is completed
	# is_pool_scrubbing - to check if the pool scrub is in progress
	# is_pool_scrubbed - to check if the pool scrub is completed
	# is_pool_scrub_stopped - to check if the pool scrub is stopped
	# is_pool_scrub_paused - to check if the pool scrub has paused
	# is_pool_removing - to check if the pool removing is a vdev
	# is_pool_removed - to check if the pool remove is completed
	# is_pool_discarding - to check if the pool checkpoint is being discarded
	#
	function is_pool_resilvering #pool <verbose>
	{
	check_pool_status "$1" "scan" \
	"resilver[ ()0-9A-Za-z:_-]* in progress since" $2
	}

	function is_pool_resilvered #pool <verbose>
	{
	check_pool_status "$1" "scan" "resilvered " $2
	}

	function is_pool_scrubbing #pool <verbose>
	{
	check_pool_status "$1" "scan" "scrub in progress since " $2
	}

	function is_pool_scrubbed #pool <verbose>
	{
	check_pool_status "$1" "scan" "scrub repaired" $2
	}

	function is_pool_scrub_stopped #pool <verbose>
	{
	check_pool_status "$1" "scan" "scrub canceled" $2
	}

	function is_pool_scrub_paused #pool <verbose>
	{
	check_pool_status "$1" "scan" "scrub paused since " $2
	}

	function is_pool_removing #pool
	{
	check_pool_status "$1" "remove" "in progress since "
	}

	function is_pool_removed #pool
	{
	check_pool_status "$1" "remove" "completed on"
	}

	function is_pool_discarding #pool
	{
	check_pool_status "$1" "checkpoint" "discarding"
	}

	function wait_for_degraded
	{
	typeset pool=$1
	typeset timeout=${2:-30}
	typeset t0=$SECONDS

	while :; do
	[[ $(get_pool_prop health $pool) == "DEGRADED" ]] && break
	log_note "$pool is not yet degraded."
	sleep 1
	if ((SECONDS - t0 > $timeout)); then
	log_note "$pool not degraded after $timeout seconds."
	return 1
	fi
	done

	return 0
	}

	#
	# Use create_pool()/destroy_pool() to clean up the information in
	# in the given disk to avoid slice overlapping.
	#
	function cleanup_devices #vdevs
	{
	typeset pool="foopool$$"

	for vdev in $@; do
	zero_partitions $vdev
	done

	poolexists $pool && destroy_pool $pool
	create_pool $pool $@
	destroy_pool $pool

	return 0
	}

	#/**
	# A function to find and locate free disks on a system or from given
	# disks as the parameter. It works by locating disks that are in use
	# as swap devices and dump devices, and also disks listed in /etc/vfstab
	#
	# $@ given disks to find which are free, default is all disks in
	# the test system
	#
	# @return a string containing the list of available disks
	#*/
	function find_disks
	{
	# Trust provided list, no attempt is made to locate unused devices.
	if is_linux \|\| is_freebsd; then
	echo "$@"
	return
	fi


	sfi=/tmp/swaplist.$$
	dmpi=/tmp/dumpdev.$$
	max_finddisksnum=${MAX_FINDDISKSNUM:-6}

	swap -l > $sfi
	dumpadm > $dmpi 2>/dev/null

	disks=${@:-$(echo "" \| format -e 2>/dev/null \| awk '
	BEGIN { FS="."; }

	/^Specify disk/{
	searchdisks=0;
	}

	{
	if (searchdisks && $2 !~ "^$"){
	split($2,arr," ");
	print arr[1];
	}
	}

	/^AVAILABLE DISK SELECTIONS:/{
	searchdisks=1;
	}
	')}

	unused=""
	for disk in $disks; do
	# Check for mounted
	grep -q "${disk}[sp]" /etc/mnttab && continue
	# Check for swap
	grep -q "${disk}[sp]" $sfi && continue
	# check for dump device
	grep -q "${disk}[sp]" $dmpi && continue
	# check to see if this disk hasn't been explicitly excluded
	# by a user-set environment variable
	echo "${ZFS_HOST_DEVICES_IGNORE}" \| grep -q "${disk}" && continue
	unused_candidates="$unused_candidates $disk"
	done
	rm $sfi $dmpi

	# now just check to see if those disks do actually exist
	# by looking for a device pointing to the first slice in
	# each case. limit the number to max_finddisksnum
	count=0
	for disk in $unused_candidates; do
	if is_disk_device $DEV_DSKDIR/${disk}s0 && \
	[ $count -lt $max_finddisksnum ]; then
	unused="$unused $disk"
	# do not impose limit if $@ is provided
	[[ -z $@ ]] && ((count = count + 1))
	fi
	done

	# finally, return our disk list
	echo $unused
	}

	function add_user_freebsd #<group_name> <user_name> <basedir>
	{
	typeset group=$1
	typeset user=$2
	typeset basedir=$3

	# Check to see if the user exists.
	if id $user > /dev/null 2>&1; then
	return 0
	fi

	# Assign 1000 as the base uid
	typeset -i uid=1000
	while true; do
	pw useradd -u $uid -g $group -d $basedir/$user -m -n $user
	case $? in
	0) break ;;
	# The uid is not unique
	65) ((uid += 1)) ;;
	*) return 1 ;;
	esac
	if [[ $uid == 65000 ]]; then
	log_fail "No user id available under 65000 for $user"
	fi
	done

	# Silence MOTD
	touch $basedir/$user/.hushlogin

	return 0
	}

	#
	# Delete the specified user.
	#
	# $1 login name
	#
	function del_user_freebsd #<logname>
	{
	typeset user=$1

	if id $user > /dev/null 2>&1; then
	log_must pw userdel $user
	fi

	return 0
	}

	#
	# Select valid gid and create specified group.
	#
	# $1 group name
	#
	function add_group_freebsd #<group_name>
	{
	typeset group=$1

	# See if the group already exists.
	if pw groupshow $group >/dev/null 2>&1; then
	return 0
	fi

	# Assign 1000 as the base gid
	typeset -i gid=1000
	while true; do
	pw groupadd -g $gid -n $group > /dev/null 2>&1
	case $? in
	0) return 0 ;;
	# The gid is not unique
	65) ((gid += 1)) ;;
	*) return 1 ;;
	esac
	if [[ $gid == 65000 ]]; then
	log_fail "No user id available under 65000 for $group"
	fi
	done
	}

	#
	# Delete the specified group.
	#
	# $1 group name
	#
	function del_group_freebsd #<group_name>
	{
	typeset group=$1

	pw groupdel -n $group > /dev/null 2>&1
	case $? in
	# Group does not exist, or was deleted successfully.
	0\|6\|65) return 0 ;;
	# Name already exists as a group name
	9) log_must pw groupdel $group ;;
	*) return 1 ;;
	esac

	return 0
	}

	function add_user_illumos #<group_name> <user_name> <basedir>
	{
	typeset group=$1
	typeset user=$2
	typeset basedir=$3

	log_must useradd -g $group -d $basedir/$user -m $user

	return 0
	}

	function del_user_illumos #<user_name>
	{
	typeset user=$1

	if id $user > /dev/null 2>&1; then
	log_must_retry "currently used" 6 userdel $user
	fi

	return 0
	}

	function add_group_illumos #<group_name>
	{
	typeset group=$1

	typeset -i gid=100
	while true; do
	groupadd -g $gid $group > /dev/null 2>&1
	case $? in
	0) return 0 ;;
	# The gid is not unique
	4) ((gid += 1)) ;;
	*) return 1 ;;
	esac
	done
	}

	function del_group_illumos #<group_name>
	{
	typeset group=$1

	groupmod -n $grp $grp > /dev/null 2>&1
	case $? in
	# Group does not exist.
	6) return 0 ;;
	# Name already exists as a group name
	9) log_must groupdel $grp ;;
	*) return 1 ;;
	esac
	}

	function add_user_linux #<group_name> <user_name> <basedir>
	{
	typeset group=$1
	typeset user=$2
	typeset basedir=$3

	log_must useradd -g $group -d $basedir/$user -m $user

	# Add new users to the same group and the command line utils.
	# This allows them to be run out of the original users home
	# directory as long as it permissioned to be group readable.
	cmd_group=$(stat --format="%G" $(command -v zfs))
	log_must usermod -a -G $cmd_group $user

	return 0
	}

	function del_user_linux #<user_name>
	{
	typeset user=$1

	if id $user > /dev/null 2>&1; then
	log_must_retry "currently used" 6 userdel $user
	fi
	}

	function add_group_linux #<group_name>
	{
	typeset group=$1

	# Assign 100 as the base gid, a larger value is selected for
	# Linux because for many distributions 1000 and under are reserved.
	while true; do
	groupadd $group > /dev/null 2>&1
	case $? in
	0) return 0 ;;
	*) return 1 ;;
	esac
	done
	}

	function del_group_linux #<group_name>
	{
	typeset group=$1

	getent group $group > /dev/null 2>&1
	case $? in
	# Group does not exist.
	2) return 0 ;;
	# Name already exists as a group name
	0) log_must groupdel $group ;;
	*) return 1 ;;
	esac

	return 0
	}

	#
	# Add specified user to specified group
	#
	# $1 group name
	# $2 user name
	# $3 base of the homedir (optional)
	#
	function add_user #<group_name> <user_name> <basedir>
	{
	typeset group=$1
	typeset user=$2
	typeset basedir=${3:-"/var/tmp"}

	if ((${#group} == 0 \|\| ${#user} == 0)); then
	log_fail "group name or user name are not defined."
	fi

	case "$UNAME" in
	FreeBSD)
	add_user_freebsd "$group" "$user" "$basedir"
	;;
	Linux)
	add_user_linux "$group" "$user" "$basedir"
	;;
	*)
	add_user_illumos "$group" "$user" "$basedir"
	;;
	esac

	return 0
	}

	#
	# Delete the specified user.
	#
	# $1 login name
	# $2 base of the homedir (optional)
	#
	function del_user #<logname> <basedir>
	{
	typeset user=$1
	typeset basedir=${2:-"/var/tmp"}

	if ((${#user} == 0)); then
	log_fail "login name is necessary."
	fi

	case "$UNAME" in
	FreeBSD)
	del_user_freebsd "$user"
	;;
	Linux)
	del_user_linux "$user"
	;;
	*)
	del_user_illumos "$user"
	;;
	esac

	[[ -d $basedir/$user ]] && rm -fr $basedir/$user

	return 0
	}

	#
	# Select valid gid and create specified group.
	#
	# $1 group name
	#
	function add_group #<group_name>
	{
	typeset group=$1

	if ((${#group} == 0)); then
	log_fail "group name is necessary."
	fi

	case "$UNAME" in
	FreeBSD)
	add_group_freebsd "$group"
	;;
	Linux)
	add_group_linux "$group"
	;;
	*)
	add_group_illumos "$group"
	;;
	esac

	return 0
	}

	#
	# Delete the specified group.
	#
	# $1 group name
	#
	function del_group #<group_name>
	{
	typeset group=$1

	if ((${#group} == 0)); then
	log_fail "group name is necessary."
	fi

	case "$UNAME" in
	FreeBSD)
	del_group_freebsd "$group"
	;;
	Linux)
	del_group_linux "$group"
	;;
	*)
	del_group_illumos "$group"
	;;
	esac

	return 0
	}

	#
	# This function will return true if it's safe to destroy the pool passed
	# as argument 1. It checks for pools based on zvols and files, and also
	# files contained in a pool that may have a different mountpoint.
	#
	function safe_to_destroy_pool { # $1 the pool name

	typeset pool=""
	typeset DONT_DESTROY=""

	# We check that by deleting the $1 pool, we're not
	# going to pull the rug out from other pools. Do this
	# by looking at all other pools, ensuring that they
	# aren't built from files or zvols contained in this pool.

	for pool in $(zpool list -H -o name)
	do
	ALTMOUNTPOOL=""

	# this is a list of the top-level directories in each of the
	# files that make up the path to the files the pool is based on
	FILEPOOL=$(zpool status -v $pool \| awk -v pool="/$1/" '$0 ~ pool {print $1}')

	# this is a list of the zvols that make up the pool
	ZVOLPOOL=$(zpool status -v $pool \| awk -v zvols="$ZVOL_DEVDIR/$1$" '$0 ~ zvols {print $1}')

	# also want to determine if it's a file-based pool using an
	# alternate mountpoint...
	POOL_FILE_DIRS=$(zpool status -v $pool \| \
	awk '/\// {print $1}' \| \
	awk -F/ '!/dev/ {print $2}')

	for pooldir in $POOL_FILE_DIRS
	do
	OUTPUT=$(zfs list -H -r -o mountpoint $1 \| \
	awk -v pd="${pooldir}$" '$0 ~ pd {print $1}')

	ALTMOUNTPOOL="${ALTMOUNTPOOL}${OUTPUT}"
	done


	if [ ! -z "$ZVOLPOOL" ]
	then
	DONT_DESTROY="true"
	log_note "Pool $pool is built from $ZVOLPOOL on $1"
	fi

	if [ ! -z "$FILEPOOL" ]
	then
	DONT_DESTROY="true"
	log_note "Pool $pool is built from $FILEPOOL on $1"
	fi

	if [ ! -z "$ALTMOUNTPOOL" ]
	then
	DONT_DESTROY="true"
	log_note "Pool $pool is built from $ALTMOUNTPOOL on $1"
	fi
	done

	if [ -z "${DONT_DESTROY}" ]
	then
	return 0
	else
	log_note "Warning: it is not safe to destroy $1!"
	return 1
	fi
	}

	#
	# Verify zfs operation with -p option work as expected
	# $1 operation, value could be create, clone or rename
	# $2 dataset type, value could be fs or vol
	# $3 dataset name
	# $4 new dataset name
	#
	function verify_opt_p_ops
	{
	typeset ops=$1
	typeset datatype=$2
	typeset dataset=$3
	typeset newdataset=$4

	if [[ $datatype != "fs" && $datatype != "vol" ]]; then
	log_fail "$datatype is not supported."
	fi

	# check parameters accordingly
	case $ops in
	create)
	newdataset=$dataset
	dataset=""
	if [[ $datatype == "vol" ]]; then
	ops="create -V $VOLSIZE"
	fi
	;;
	clone)
	if [[ -z $newdataset ]]; then
	log_fail "newdataset should not be empty" \
	"when ops is $ops."
	fi
	log_must datasetexists $dataset
	log_must snapexists $dataset
	;;
	rename)
	if [[ -z $newdataset ]]; then
	log_fail "newdataset should not be empty" \
	"when ops is $ops."
	fi
	log_must datasetexists $dataset
	;;
	*)
	log_fail "$ops is not supported."
	;;
	esac

	# make sure the upper level filesystem does not exist
	destroy_dataset "${newdataset%/*}" "-rRf"

	# without -p option, operation will fail
	log_mustnot zfs $ops $dataset $newdataset
	log_mustnot datasetexists $newdataset ${newdataset%/*}

	# with -p option, operation should succeed
	log_must zfs $ops -p $dataset $newdataset
	block_device_wait

	if ! datasetexists $newdataset ; then
	log_fail "-p option does not work for $ops"
	fi

	# when $ops is create or clone, redo the operation still return zero
	if [[ $ops != "rename" ]]; then
	log_must zfs $ops -p $dataset $newdataset
	fi

	return 0
	}

	#
	# Get configuration of pool
	# $1 pool name
	# $2 config name
	#
	function get_config
	{
	typeset pool=$1
	typeset config=$2

	if ! poolexists "$pool" ; then
	return 1
	fi
	if [ "$(get_pool_prop cachefile "$pool")" = "none" ]; then
	zdb -e $pool
	else
	zdb -C $pool
	fi \| awk -F: -v cfg="$config:" '$0 ~ cfg {sub(/^'\''/, $2); sub(/'\''$/, $2); print $2}'
	}

	#
	# Privated function. Random select one of items from arguments.
	#
	# $1 count
	# $2-n string
	#
	function _random_get
	{
	typeset cnt=$1
	shift

	typeset str="$@"
	typeset -i ind
	((ind = RANDOM % cnt + 1))

	echo "$str" \| cut -f $ind -d ' '
	}

	#
	# Random select one of item from arguments which include NONE string
	#
	function random_get_with_non
	{
	typeset -i cnt=$#
	((cnt =+ 1))

	_random_get "$cnt" "$@"
	}

	#
	# Random select one of item from arguments which doesn't include NONE string
	#
	function random_get
	{
	_random_get "$#" "$@"
	}

	#
	# The function will generate a dataset name with specific length
	# $1, the length of the name
	# $2, the base string to construct the name
	#
	function gen_dataset_name
	{
	typeset -i len=$1
	typeset basestr="$2"
	typeset -i baselen=${#basestr}
	typeset -i iter=0
	typeset l_name=""

	if ((len % baselen == 0)); then
	((iter = len / baselen))
	else
	((iter = len / baselen + 1))
	fi
	while ((iter > 0)); do
	l_name="${l_name}$basestr"

	((iter -= 1))
	done

	echo $l_name
	}

	#
	# Get cksum tuple of dataset
	# $1 dataset name
	#
	# sample zdb output:
	# Dataset data/test [ZPL], ID 355, cr_txg 2413856, 31.0K, 7 objects, rootbp
	# DVA[0]=<0:803046400:200> DVA[1]=<0:81199000:200> [L0 DMU objset] fletcher4
	# lzjb LE contiguous unique double size=800L/200P birth=2413856L/2413856P
	# fill=7 cksum=11ce125712:643a9c18ee2:125e25238fca0:254a3f74b59744
	function datasetcksum
	{
	typeset cksum
	sync
	sync_all_pools
	zdb -vvv $1 \| awk -F= -v ds="^Dataset $1 "'\\[' '$0 ~ ds && /cksum/ {print $7}'
	}

	#
	# Get the given disk/slice state from the specific field of the pool
	#
	function get_device_state #pool disk field("", "spares","logs")
	{
	typeset pool=$1
	typeset disk=${2#$DEV_DSKDIR/}
	typeset field=${3:-$pool}

	zpool status -v "$pool" 2>/dev/null \| \
	awk -v device=$disk -v pool=$pool -v field=$field \
	'BEGIN {startconfig=0; startfield=0; }
	/config:/ {startconfig=1}
	(startconfig==1) && ($1==field) {startfield=1; next;}
	(startfield==1) && ($1==device) {print $2; exit;}
	(startfield==1) &&
	($1==field \|\| $1 ~ "^spares$" \|\| $1 ~ "^logs$") {startfield=0}'
	}

	#
	# get the root filesystem name if it's zfsroot system.
	#
	# return: root filesystem name
	function get_rootfs
	{
	typeset rootfs=""

	if is_freebsd; then
	rootfs=$(mount -p \| awk '$2 == "/" && $3 == "zfs" {print $1}')
	elif ! is_linux; then
	rootfs=$(awk '$2 == "/" && $3 == "zfs" {print $1}' \
	/etc/mnttab)
	fi
	if [[ -z "$rootfs" ]]; then
	log_fail "Can not get rootfs"
	fi
	if datasetexists $rootfs; then
	echo $rootfs
	else
	log_fail "This is not a zfsroot system."
	fi
	}

	#
	# get the rootfs's pool name
	# return:
	# rootpool name
	#
	function get_rootpool
	{
	typeset rootfs=$(get_rootfs)
	echo ${rootfs%%/*}
	}

	#
	# To verify if the require numbers of disks is given
	#
	function verify_disk_count
	{
	typeset -i min=${2:-1}

	typeset -i count=$(echo "$1" \| wc -w)

	if ((count < min)); then
	log_untested "A minimum of $min disks is required to run." \
	" You specified $count disk(s)"
	fi
	}

	function ds_is_volume
	{
	typeset type=$(get_prop type $1)
	[ $type = "volume" ]
	}

	function ds_is_filesystem
	{
	typeset type=$(get_prop type $1)
	[ $type = "filesystem" ]
	}

	#
	# Check if Trusted Extensions are installed and enabled
	#
	function is_te_enabled
	{
	svcs -H -o state labeld 2>/dev/null \| grep -q "enabled"
	}

	# Return the number of CPUs (cross-platform)
	function get_num_cpus
	{
	if is_linux ; then
	grep -c '^processor' /proc/cpuinfo
	elif is_freebsd; then
	sysctl -n kern.smp.cpus
	else
	psrinfo \| wc -l
	fi
	}

	# Utility function to determine if a system has multiple cpus.
	function is_mp
	{
	[[ $(get_num_cpus) -gt 1 ]]
	}

	function get_cpu_freq
	{
	if is_linux; then
	lscpu \| awk '/CPU MHz/ { print $3 }'
	elif is_freebsd; then
	sysctl -n hw.clockrate
	else
	psrinfo -v 0 \| awk '/processor operates at/ {print $6}'
	fi
	}

	# Run the given command as the user provided.
	function user_run
	{
	typeset user=$1
	shift

	log_note "user: $user"
	log_note "cmd: $*"

	typeset out=$TEST_BASE_DIR/out
	typeset err=$TEST_BASE_DIR/err

	sudo -Eu $user env PATH="$PATH" ksh <<<"$*" >$out 2>$err
	typeset res=$?
	log_note "out: $(<$out)"
	log_note "err: $(<$err)"
	return $res
	}

	#
	# Check if the pool contains the specified vdevs
	#
	# $1 pool
	# $2..n <vdev> ...
	#
	# Return 0 if the vdevs are contained in the pool, 1 if any of the specified
	# vdevs is not in the pool, and 2 if pool name is missing.
	#
	function vdevs_in_pool
	{
	typeset pool=$1
	typeset vdev

	if [[ -z $pool ]]; then
	log_note "Missing pool name."
	return 2
	fi

	shift

	# We could use 'zpool list' to only get the vdevs of the pool but we
	# can't reference a mirror/raidz vdev using its ID (i.e mirror-0),
	# therefore we use the 'zpool status' output.
	typeset tmpfile=$(mktemp)
	zpool status -v "$pool" \| grep -A 1000 "config:" >$tmpfile
	for vdev in "$@"; do
	grep -wq ${vdev##*/} $tmpfile \|\| return 1
	done

	rm -f $tmpfile
	return 0
	}

	function get_max
	{
	typeset -l i max=$1
	shift

	for i in "$@"; do
	max=$((max > i ? max : i))
	done

	echo $max
	}

	# Write data that can be compressed into a directory
	function write_compressible
	{
	typeset dir=$1
	typeset megs=$2
	typeset nfiles=${3:-1}
	typeset bs=${4:-1024k}
	typeset fname=${5:-file}

	[[ -d $dir ]] \|\| log_fail "No directory: $dir"

	# Under Linux fio is not currently used since its behavior can
	# differ significantly across versions. This includes missing
	# command line options and cases where the --buffer_compress_*
	# options fail to behave as expected.
	if is_linux; then
	typeset file_bytes=$(to_bytes $megs)
	typeset bs_bytes=4096
	typeset blocks=$(($file_bytes / $bs_bytes))

	for (( i = 0; i < $nfiles; i++ )); do
	truncate -s $file_bytes $dir/$fname.$i

	# Write every third block to get 66% compression.
	for (( j = 0; j < $blocks; j += 3 )); do
	dd if=/dev/urandom of=$dir/$fname.$i \
	seek=$j bs=$bs_bytes count=1 \
	conv=notrunc >/dev/null 2>&1
	done
	done
	else
	command -v fio > /dev/null \|\| log_unsupported "fio missing"
	log_must eval fio \
	--name=job \
	--fallocate=0 \
	--minimal \
	--randrepeat=0 \
	--buffer_compress_percentage=66 \
	--buffer_compress_chunk=4096 \
	--directory="$dir" \
	--numjobs="$nfiles" \
	--nrfiles="$nfiles" \
	--rw=write \
	--bs="$bs" \
	--filesize="$megs" \
	"--filename_format='$fname.\$jobnum' >/dev/null"
	fi
	}

	function get_objnum
	{
	typeset pathname=$1
	typeset objnum

	[[ -e $pathname ]] \|\| log_fail "No such file or directory: $pathname"
	if is_freebsd; then
	objnum=$(stat -f "%i" $pathname)
	else
	objnum=$(stat -c %i $pathname)
	fi
	echo $objnum
	}

	#
	# Sync data to the pool
	#
	# $1 pool name
	# $2 boolean to force uberblock (and config including zpool cache file) update
	#
	function sync_pool #pool <force>
	{
	typeset pool=${1:-$TESTPOOL}
	typeset force=${2:-false}

	if [[ $force == true ]]; then
	log_must zpool sync -f $pool
	else
	log_must zpool sync $pool
	fi

	return 0
	}

	#
	# Sync all pools
	#
	# $1 boolean to force uberblock (and config including zpool cache file) update
	#
	function sync_all_pools #<force>
	{
	typeset force=${1:-false}

	if [[ $force == true ]]; then
	log_must zpool sync -f
	else
	log_must zpool sync
	fi

	return 0
	}

	#
	# Wait for zpool 'freeing' property drops to zero.
	#
	# $1 pool name
	#
	function wait_freeing #pool
	{
	typeset pool=${1:-$TESTPOOL}
	while true; do
	[[ "0" == "$(zpool list -Ho freeing $pool)" ]] && break
	log_must sleep 1
	done
	}

	#
	# Wait for every device replace operation to complete
	#
	# $1 pool name
	#
	function wait_replacing #pool
	{
	typeset pool=${1:-$TESTPOOL}
	while zpool status $pool \| grep -qE 'replacing-[0-9]+'; do
	log_must sleep 1
	done
	}

	# Wait for a pool to be scrubbed
	#
	# $1 pool name
	# $2 timeout
	#
	function wait_scrubbed #pool timeout
	{
	typeset timeout=${2:-300}
	typeset pool=${1:-$TESTPOOL}
	for (( timer = 0; timer < $timeout; timer++ )); do
	is_pool_scrubbed $pool && break;
	sleep 1;
	done
	}

	# Backup the zed.rc in our test directory so that we can edit it for our test.
	#
	# Returns: Backup file name. You will need to pass this to zed_rc_restore().
	function zed_rc_backup
	{
	zedrc_backup="$(mktemp)"
	cp $ZEDLET_DIR/zed.rc $zedrc_backup
	echo $zedrc_backup
	}

	function zed_rc_restore
	{
	mv $1 $ZEDLET_DIR/zed.rc
	}

	#
	# Setup custom environment for the ZED.
	#
	# $@ Optional list of zedlets to run under zed.
	function zed_setup
	{
	if ! is_linux; then
	log_unsupported "No zed on $UNAME"
	fi

	if [[ ! -d $ZEDLET_DIR ]]; then
	log_must mkdir $ZEDLET_DIR
	fi

	if [[ ! -e $VDEVID_CONF ]]; then
	log_must touch $VDEVID_CONF
	fi

	if [[ -e $VDEVID_CONF_ETC ]]; then
	log_fail "Must not have $VDEVID_CONF_ETC file present on system"
	fi
	EXTRA_ZEDLETS=$@

	# Create a symlink for /etc/zfs/vdev_id.conf file.
	log_must ln -s $VDEVID_CONF $VDEVID_CONF_ETC

	# Setup minimal ZED configuration. Individual test cases should
	# add additional ZEDLETs as needed for their specific test.
	log_must cp ${ZEDLET_ETC_DIR}/zed.rc $ZEDLET_DIR
	log_must cp ${ZEDLET_ETC_DIR}/zed-functions.sh $ZEDLET_DIR

	# Scripts must only be user writable.
	if [[ -n "$EXTRA_ZEDLETS" ]] ; then
	saved_umask=$(umask)
	log_must umask 0022
	for i in $EXTRA_ZEDLETS ; do
	log_must cp ${ZEDLET_LIBEXEC_DIR}/$i $ZEDLET_DIR
	done
	log_must umask $saved_umask
	fi

	# Customize the zed.rc file to enable the full debug log.
	log_must sed -i '/\#ZED_DEBUG_LOG=.*/d' $ZEDLET_DIR/zed.rc
	echo "ZED_DEBUG_LOG=$ZED_DEBUG_LOG" >>$ZEDLET_DIR/zed.rc

	}

	#
	# Cleanup custom ZED environment.
	#
	# $@ Optional list of zedlets to remove from our test zed.d directory.
	function zed_cleanup
	{
	if ! is_linux; then
	return
	fi

	for extra_zedlet; do
	log_must rm -f ${ZEDLET_DIR}/$extra_zedlet
	done
	log_must rm -fd ${ZEDLET_DIR}/zed.rc ${ZEDLET_DIR}/zed-functions.sh ${ZEDLET_DIR}/all-syslog.sh ${ZEDLET_DIR}/all-debug.sh ${ZEDLET_DIR}/state \
	$ZED_LOG $ZED_DEBUG_LOG $VDEVID_CONF_ETC $VDEVID_CONF \
	$ZEDLET_DIR
	}

	#
	# Check if ZED is currently running; if so, returns PIDs
	#
	function zed_check
	{
	if ! is_linux; then
	return
	fi
	zedpids="$(pgrep -x zed)"
	zedpids2="$(pgrep -x lt-zed)"
	echo ${zedpids} ${zedpids2}
	}

	#
	# Check if ZED is currently running, if not start ZED.
	#
	function zed_start
	{
	if ! is_linux; then
	return
	fi

	# ZEDLET_DIR=/var/tmp/zed
	if [[ ! -d $ZEDLET_DIR ]]; then
	log_must mkdir $ZEDLET_DIR
	fi

	# Verify the ZED is not already running.
	zedpids=$(zed_check)
	if [ -n "$zedpids" ]; then
	# We never, ever, really want it to just keep going if zed
	# is already running - usually this implies our test cases
	# will break very strangely because whatever we wanted to
	# configure zed for won't be listening to our changes in the
	# tmpdir
	log_fail "ZED already running - ${zedpids}"
	else
	log_note "Starting ZED"
	# run ZED in the background and redirect foreground logging
	# output to $ZED_LOG.
	log_must truncate -s 0 $ZED_DEBUG_LOG
	log_must eval "zed -vF -d $ZEDLET_DIR -P $PATH" \
	"-s $ZEDLET_DIR/state -j 1 2>$ZED_LOG &"
	fi

	return 0
	}

	#
	# Kill ZED process
	#
	function zed_stop
	{
	if ! is_linux; then
	return ""
	fi

	log_note "Stopping ZED"
	while true; do
	zedpids=$(zed_check)
	[ ! -n "$zedpids" ] && break

	log_must kill $zedpids
	sleep 1
	done
	return 0
	}

	#
	# Drain all zevents
	#
	function zed_events_drain
	{
	while [ $(zpool events -H \| wc -l) -ne 0 ]; do
	sleep 1
	zpool events -c >/dev/null
	done
	}

	# Set a variable in zed.rc to something, un-commenting it in the process.
	#
	# $1 variable
	# $2 value
	function zed_rc_set
	{
	var="$1"
	val="$2"
	# Remove the line
	cmd="'/$var/d'"
	eval sed -i $cmd $ZEDLET_DIR/zed.rc

	# Add it at the end
	echo "$var=$val" >> $ZEDLET_DIR/zed.rc
	}


	#
	# Check is provided device is being active used as a swap device.
	#
	function is_swap_inuse
	{
	typeset device=$1

	if [[ -z $device ]] ; then
	log_note "No device specified."
	return 1
	fi

	case "$UNAME" in
	Linux)
	swapon -s \| grep -wq $(readlink -f $device)
	;;
	FreeBSD)
	swapctl -l \| grep -wq $device
	;;
	*)
	swap -l \| grep -wq $device
	;;
	esac
	}

	#
	# Setup a swap device using the provided device.
	#
	function swap_setup
	{
	typeset swapdev=$1

	case "$UNAME" in
	Linux)
	log_must eval "mkswap $swapdev > /dev/null 2>&1"
	log_must swapon $swapdev
	;;
	FreeBSD)
	log_must swapctl -a $swapdev
	;;
	*)
	log_must swap -a $swapdev
	;;
	esac

	return 0
	}

	#
	# Cleanup a swap device on the provided device.
	#
	function swap_cleanup
	{
	typeset swapdev=$1

	if is_swap_inuse $swapdev; then
	if is_linux; then
	log_must swapoff $swapdev
	elif is_freebsd; then
	log_must swapoff $swapdev
	else
	log_must swap -d $swapdev
	fi
	fi

	return 0
	}

	#
	# Set a global system tunable (64-bit value)
	#
	# $1 tunable name (use a NAME defined in tunables.cfg)
	# $2 tunable values
	#
	function set_tunable64
	{
	set_tunable_impl "$1" "$2" Z
	}

	#
	# Set a global system tunable (32-bit value)
	#
	# $1 tunable name (use a NAME defined in tunables.cfg)
	# $2 tunable values
	#
	function set_tunable32
	{
	set_tunable_impl "$1" "$2" W
	}

	function set_tunable_impl
	{
	typeset name="$1"
	typeset value="$2"
	typeset mdb_cmd="$3"

	eval "typeset tunable=\$$name"
	case "$tunable" in
	UNSUPPORTED)
	log_unsupported "Tunable '$name' is unsupported on $UNAME"
	;;
	"")
	log_fail "Tunable '$name' must be added to tunables.cfg"
	;;
	*)
	;;
	esac

	[[ -z "$value" ]] && return 1
	[[ -z "$mdb_cmd" ]] && return 1

	case "$UNAME" in
	Linux)
	typeset zfs_tunables="/sys/module/zfs/parameters"
	echo "$value" >"$zfs_tunables/$tunable"
	;;
	FreeBSD)
	sysctl vfs.zfs.$tunable=$value
	;;
	SunOS)
	echo "${tunable}/${mdb_cmd}0t${value}" \| mdb -kw
	;;
	esac
	}

	#
	# Get a global system tunable
	#
	# $1 tunable name (use a NAME defined in tunables.cfg)
	#
	function get_tunable
	{
	get_tunable_impl "$1"
	}

	function get_tunable_impl
	{
	typeset name="$1"
	typeset module="${2:-zfs}"
	typeset check_only="$3"

	eval "typeset tunable=\$$name"
	case "$tunable" in
	UNSUPPORTED)
	if [ -z "$check_only" ] ; then
	log_unsupported "Tunable '$name' is unsupported on $UNAME"
	else
	return 1
	fi
	;;
	"")
	if [ -z "$check_only" ] ; then
	log_fail "Tunable '$name' must be added to tunables.cfg"
	else
	return 1
	fi
	;;
	*)
	;;
	esac

	case "$UNAME" in
	Linux)
	typeset zfs_tunables="/sys/module/$module/parameters"
	cat $zfs_tunables/$tunable
	;;
	FreeBSD)
	sysctl -n vfs.zfs.$tunable
	;;
	SunOS)
	[[ "$module" -eq "zfs" ]] \|\| return 1
	;;
	esac
	}

	# Does a tunable exist?
	#
	# $1: Tunable name
	function tunable_exists
	{
	get_tunable_impl $1 "zfs" 1
	}

	#
	# Compute MD5 digest for given file or stdin if no file given.
	# Note: file path must not contain spaces
	#
	function md5digest
	{
	typeset file=$1

	case "$UNAME" in
	FreeBSD)
	md5 -q $file
	;;
	*)
	typeset sum _
	read -r sum _ < <(md5sum -b $file)
	echo $sum
	;;
	esac
	}

	#
	# Compute SHA256 digest for given file or stdin if no file given.
	# Note: file path must not contain spaces
	#
	function sha256digest
	{
	typeset file=$1

	case "$UNAME" in
	FreeBSD)
	sha256 -q $file
	;;
	*)
	typeset sum _
	read -r sum _ < <(sha256sum -b $file)
	echo $sum
	;;
	esac
	}

	function new_fs #<args>
	{
	case "$UNAME" in
	FreeBSD)
	newfs "$@"
	;;
	*)
	echo y \| newfs -v "$@"
	;;
	esac
	}

	function stat_size #<path>
	{
	typeset path=$1

	case "$UNAME" in
	FreeBSD)
	stat -f %z "$path"
	;;
	*)
	stat -c %s "$path"
	;;
	esac
	}

	function stat_mtime #<path>
	{
	typeset path=$1

	case "$UNAME" in
	FreeBSD)
	stat -f %m "$path"
	;;
	*)
	stat -c %Y "$path"
	;;
	esac
	}

	function stat_ctime #<path>
	{
	typeset path=$1

	case "$UNAME" in
	FreeBSD)
	stat -f %c "$path"
	;;
	*)
	stat -c %Z "$path"
	;;
	esac
	}

	function stat_crtime #<path>
	{
	typeset path=$1

	case "$UNAME" in
	FreeBSD)
	stat -f %B "$path"
	;;
	*)
	stat -c %W "$path"
	;;
	esac
	}

	function stat_generation #<path>
	{
	typeset path=$1

	case "$UNAME" in
	Linux)
	getversion "${path}"
	;;
	*)
	stat -f %v "${path}"
	;;
	esac
	}

	# Run a command as if it was being run in a TTY.
	#
	# Usage:
	#
	# faketty command
	#
	function faketty
	{
	if is_freebsd; then
	script -q /dev/null env "$@"
	else
	script --return --quiet -c "$*" /dev/null
	fi
	}

	#
	# Produce a random permutation of the integers in a given range (inclusive).
	#
	function range_shuffle # begin end
	{
	typeset -i begin=$1
	typeset -i end=$2

	seq ${begin} ${end} \| sort -R
	}

	#
	# Cross-platform xattr helpers
	#

	function get_xattr # name path
	{
	typeset name=$1
	typeset path=$2

	case "$UNAME" in
	FreeBSD)
	getextattr -qq user "${name}" "${path}"
	;;
	*)
	attr -qg "${name}" "${path}"
	;;
	esac
	}

	function set_xattr # name value path
	{
	typeset name=$1
	typeset value=$2
	typeset path=$3

	case "$UNAME" in
	FreeBSD)
	setextattr user "${name}" "${value}" "${path}"
	;;
	*)
	attr -qs "${name}" -V "${value}" "${path}"
	;;
	esac
	}

	function set_xattr_stdin # name value
	{
	typeset name=$1
	typeset path=$2

	case "$UNAME" in
	FreeBSD)
	setextattr -i user "${name}" "${path}"
	;;
	*)
	attr -qs "${name}" "${path}"
	;;
	esac
	}

	function rm_xattr # name path
	{
	typeset name=$1
	typeset path=$2

	case "$UNAME" in
	FreeBSD)
	rmextattr -q user "${name}" "${path}"
	;;
	*)
	attr -qr "${name}" "${path}"
	;;
	esac
	}

	function ls_xattr # path
	{
	typeset path=$1

	case "$UNAME" in
	FreeBSD)
	lsextattr -qq user "${path}"
	;;
	*)
	attr -ql "${path}"
	;;
	esac
	}

	function kstat # stat flags?
	{
	typeset stat=$1
	typeset flags=${2-"-n"}

	case "$UNAME" in
	FreeBSD)
	sysctl $flags kstat.zfs.misc.$stat
	;;
	Linux)
	cat "/proc/spl/kstat/zfs/$stat" 2>/dev/null
	;;
	*)
	false
	;;
	esac
	}

	function get_arcstat # stat
	{
	typeset stat=$1

	case "$UNAME" in
	FreeBSD)
	kstat arcstats.$stat
	;;
	Linux)
	kstat arcstats \| awk "/$stat/"' { print $3 }'
	;;
	*)
	false
	;;
	esac
	}

	function punch_hole # offset length file
	{
	typeset offset=$1
	typeset length=$2
	typeset file=$3

	case "$UNAME" in
	FreeBSD)
	truncate -d -o $offset -l $length "$file"
	;;
	Linux)
	fallocate --punch-hole --offset $offset --length $length "$file"
	;;
	*)
	false
	;;
	esac
	}

	function zero_range # offset length file
	{
	typeset offset=$1
	typeset length=$2
	typeset file=$3

	case "$UNAME" in
	Linux)
	fallocate --zero-range --offset $offset --length $length "$file"
	;;
	*)
	false
	;;
	esac
	}

	#
	# Wait for the specified arcstat to reach non-zero quiescence.
	# If echo is 1 echo the value after reaching quiescence, otherwise
	# if echo is 0 print the arcstat we are waiting on.
	#
	function arcstat_quiescence # stat echo
	{
	typeset stat=$1
	typeset echo=$2
	typeset do_once=true

	if [[ $echo -eq 0 ]]; then
	echo "Waiting for arcstat $1 quiescence."
	fi

	while $do_once \|\| [ $stat1 -ne $stat2 ] \|\| [ $stat2 -eq 0 ]; do
	typeset stat1=$(get_arcstat $stat)
	sleep 2
	typeset stat2=$(get_arcstat $stat)
	do_once=false
	done

	if [[ $echo -eq 1 ]]; then
	echo $stat2
	fi
	}

	function arcstat_quiescence_noecho # stat
	{
	typeset stat=$1
	arcstat_quiescence $stat 0
	}

	function arcstat_quiescence_echo # stat
	{
	typeset stat=$1
	arcstat_quiescence $stat 1
	}

	#
	# Given an array of pids, wait until all processes
	# have completed and check their return status.
	#
	function wait_for_children #children
	{
	rv=0
	children=("$@")
	for child in "${children[@]}"
	do
	child_exit=0
	wait ${child} \|\| child_exit=$?
	if [ $child_exit -ne 0 ]; then
	echo "child ${child} failed with ${child_exit}"
	rv=1
	fi
	done
	return $rv
	}

	#
	# Compare two directory trees recursively in a manner similar to diff(1), but
	# using rsync. If there are any discrepancies, a summary of the differences are
	# output and a non-zero error is returned.
	#
	# If you're comparing a directory after a ZIL replay, you should set
	# LIBTEST_DIFF_ZIL_REPLAY=1 or use replay_directory_diff which will cause
	# directory_diff to ignore mtime changes (the ZIL replay won't fix up mtime
	# information).
	#
	function directory_diff # dir_a dir_b
	{
	dir_a="$1"
	dir_b="$2"
	zil_replay="${LIBTEST_DIFF_ZIL_REPLAY:-0}"

	# If one of the directories doesn't exist, return 2. This is to match the
	# semantics of diff.
	if ! [ -d "$dir_a" -a -d "$dir_b" ]; then
	return 2
	fi

	# Run rsync with --dry-run --itemize-changes to get something akin to diff
	# output, but rsync is far more thorough in detecting differences (diff
	# doesn't compare file metadata, and cannot handle special files).
	#
	# Also make sure to filter out non-user.* xattrs when comparing. On
	# SELinux-enabled systems the copied tree will probably have different
	# SELinux labels.
	args=("-nicaAHX" '--filter=-x! user.*' "--delete")

	# NOTE: Quite a few rsync builds do not support --crtimes which would be
	# necessary to verify that creation times are being maintained properly.
	# Unfortunately because of this we cannot use it unconditionally but we can
	# check if this rsync build supports it and use it then. This check is
	# based on the same check in the rsync test suite (testsuite/crtimes.test).
	#
	# We check ctimes even with zil_replay=1 because the ZIL does store
	# creation times and we should make sure they match (if the creation times
	# do not match there is a "c" entry in one of the columns).
	if rsync --version \| grep -q "[, ] crtimes"; then
	args+=("--crtimes")
	else
	log_note "This rsync package does not support --crtimes (-N)."
	fi

	# If we are testing a ZIL replay, we need to ignore timestamp changes.
	# Unfortunately --no-times doesn't do what we want -- it will still tell
	# you if the timestamps don't match but rsync will set the timestamps to
	# the current time (leading to an itemised change entry). It's simpler to
	# just filter out those lines.
	if [ "$zil_replay" -eq 0 ]; then
	filter=("cat")
	else
	# Different rsync versions have different numbers of columns. So just
	# require that aside from the first two, all other columns must be
	# blank (literal ".") or a timestamp field ("[tT]").
	filter=("grep" "-v" '^\..[.Tt]\+ ')
	fi

	diff="$(rsync "${args[@]}" "$dir_a/" "$dir_b/" \| "${filter[@]}")"
	rv=0
	if [ -n "$diff" ]; then
	echo "$diff"
	rv=1
	fi
	return $rv
	}

	#
	# Compare two directory trees recursively, without checking whether the mtimes
	# match (creation times will be checked if the available rsync binary supports
	# it). This is necessary for ZIL replay checks (because the ZIL does not
	# contain mtimes and thus after a ZIL replay, mtimes won't match).
	#
	# This is shorthand for LIBTEST_DIFF_ZIL_REPLAY=1 directory_diff <...>.
	#
	function replay_directory_diff # dir_a dir_b
	{
	LIBTEST_DIFF_ZIL_REPLAY=1 directory_diff "$@"
	}

	#
	# Put coredumps into $1/core.{basename}
	#
	# Output must be saved and passed to pop_coredump_pattern on cleanup
	#
	function push_coredump_pattern # dir
	{
	ulimit -c unlimited
	case "$UNAME" in
	Linux)
	cat /proc/sys/kernel/core_pattern /proc/sys/kernel/core_uses_pid
	echo "$1/core.%e" >/proc/sys/kernel/core_pattern &&
	echo 0 >/proc/sys/kernel/core_uses_pid
	;;
	FreeBSD)
	sysctl -n kern.corefile
	sysctl kern.corefile="$1/core.%N" >/dev/null
	;;
	*)
	# Nothing to output – set only for this shell
	coreadm -p "$1/core.%f"
	;;
	esac
	}

	#
	# Put coredumps back into the default location
	#
	function pop_coredump_pattern
	{
	[ -s "$1" ] \|\| return 0
	case "$UNAME" in
	Linux)
	typeset pat pid
	{ read -r pat; read -r pid; } < "$1"
	echo "$pat" >/proc/sys/kernel/core_pattern &&
	echo "$pid" >/proc/sys/kernel/core_uses_pid
	;;
	FreeBSD)
	sysctl kern.corefile="$(<"$1")" >/dev/null
	;;
	esac
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	index d53316643bc5..6aeb862fbb85 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	@@ -1,2000 +1,2001 @@
	CLEANFILES =
	dist_noinst_DATA =
	include $(top_srcdir)/config/Substfiles.am


	datadir_zfs_tests_testsdir = $(datadir)/$(PACKAGE)/zfs-tests/tests
	nobase_dist_datadir_zfs_tests_tests_DATA = \
	perf/nfs-sample.cfg \
	perf/perf.shlib \
	\
	perf/fio/mkfiles.fio \
	perf/fio/random_reads.fio \
	perf/fio/random_readwrite.fio \
	perf/fio/random_readwrite_fixed.fio \
	perf/fio/random_writes.fio \
	perf/fio/sequential_reads.fio \
	perf/fio/sequential_readwrite.fio \
	perf/fio/sequential_writes.fio

	nobase_dist_datadir_zfs_tests_tests_SCRIPTS = \
	perf/regression/random_reads.ksh \
	perf/regression/random_readwrite.ksh \
	perf/regression/random_readwrite_fixed.ksh \
	perf/regression/random_writes.ksh \
	perf/regression/random_writes_zil.ksh \
	perf/regression/sequential_reads_arc_cached_clone.ksh \
	perf/regression/sequential_reads_arc_cached.ksh \
	perf/regression/sequential_reads_dbuf_cached.ksh \
	perf/regression/sequential_reads.ksh \
	perf/regression/sequential_writes.ksh \
	perf/regression/setup.ksh \
	\
	perf/scripts/prefetch_io.sh

	# These lists can be regenerated by running make regen-tests at the root, or, on a clean source:
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -executable -name '.in' \| sort \| sed 's/\.in$//;s/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' -executable -name '.in' \| sort \| sed 's/\.in$//;s/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -name '.in' ! -name '*.c' \| grep -Fe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -executable ! -name '.in' ! -name '*.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' -executable ! -name '.in' ! -name '*.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	#
	# simd and tmpfile are Linux-only and not installed elsewhere
	#
	# C programs are specced in ../Makefile.am above as part of the main Makefile

	find_common := find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po'
	regen:
	@$(MAKE) -C $(top_builddir) clean
	@$(MAKE) clean
	$(SED) $(ac_inplace) '/^# -- >8 --/q' Makefile.am
	echo >> Makefile.am
	echo 'nobase_nodist_datadir_zfs_tests_tests_DATA = \' >> Makefile.am
	$(find_common) ! -executable -name '*.in' \| sort \| sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo 'nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \' >> Makefile.am
	$(find_common) -executable -name '*.in' \| sort \| sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo >> Makefile.am
	echo 'SUBSTFILES += $$(nobase_nodist_datadir_zfs_tests_tests_DATA) $$(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS)' >> Makefile.am
	echo >> Makefile.am
	echo 'if BUILD_LINUX' >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am
	$(find_common) ! -name '.in' ! -name '.c' \| grep -Fe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo 'endif' >> Makefile.am
	echo >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_DATA += \' >> Makefile.am
	$(find_common) ! -executable ! -name '.in' ! -name '.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am
	$(find_common) -executable ! -name '.in' ! -name '.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am

	# -- >8 --

	nobase_nodist_datadir_zfs_tests_tests_DATA = \
	functional/pam/utilities.kshlib
	nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \
	functional/pyzfs/pyzfs_unittest.ksh

	SUBSTFILES += $(nobase_nodist_datadir_zfs_tests_tests_DATA) $(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS)

	if BUILD_LINUX
	nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
	functional/simd/simd_supported.ksh \
	functional/tmpfile/cleanup.ksh \
	functional/tmpfile/setup.ksh
	endif

	nobase_dist_datadir_zfs_tests_tests_DATA += \
	functional/acl/acl.cfg \
	functional/acl/acl_common.kshlib \
	functional/alloc_class/alloc_class.cfg \
	functional/alloc_class/alloc_class.kshlib \
	functional/atime/atime.cfg \
	functional/atime/atime_common.kshlib \
	functional/cache/cache.cfg \
	functional/cache/cache.kshlib \
	functional/cachefile/cachefile.cfg \
	functional/cachefile/cachefile.kshlib \
	functional/casenorm/casenorm.cfg \
	functional/casenorm/casenorm.kshlib \
	functional/channel_program/channel_common.kshlib \
	functional/channel_program/lua_core/tst.args_to_lua.out \
	functional/channel_program/lua_core/tst.args_to_lua.zcp \
	functional/channel_program/lua_core/tst.divide_by_zero.err \
	functional/channel_program/lua_core/tst.divide_by_zero.zcp \
	functional/channel_program/lua_core/tst.exists.zcp \
	functional/channel_program/lua_core/tst.large_prog.out \
	functional/channel_program/lua_core/tst.large_prog.zcp \
	functional/channel_program/lua_core/tst.lib_base.lua \
	functional/channel_program/lua_core/tst.lib_coroutine.lua \
	functional/channel_program/lua_core/tst.lib_strings.lua \
	functional/channel_program/lua_core/tst.lib_table.lua \
	functional/channel_program/lua_core/tst.nested_neg.zcp \
	functional/channel_program/lua_core/tst.nested_pos.zcp \
	functional/channel_program/lua_core/tst.recursive.zcp \
	functional/channel_program/lua_core/tst.return_large.zcp \
	functional/channel_program/lua_core/tst.return_recursive_table.zcp \
	functional/channel_program/lua_core/tst.stack_gsub.err \
	functional/channel_program/lua_core/tst.stack_gsub.zcp \
	functional/channel_program/lua_core/tst.timeout.zcp \
	functional/channel_program/synctask_core/tst.bookmark.copy.zcp \
	functional/channel_program/synctask_core/tst.bookmark.create.zcp \
	functional/channel_program/synctask_core/tst.get_index_props.out \
	functional/channel_program/synctask_core/tst.get_index_props.zcp \
	functional/channel_program/synctask_core/tst.get_number_props.out \
	functional/channel_program/synctask_core/tst.get_number_props.zcp \
	functional/channel_program/synctask_core/tst.get_string_props.out \
	functional/channel_program/synctask_core/tst.get_string_props.zcp \
	functional/channel_program/synctask_core/tst.promote_conflict.zcp \
	functional/channel_program/synctask_core/tst.set_props.zcp \
	functional/channel_program/synctask_core/tst.snapshot_destroy.zcp \
	functional/channel_program/synctask_core/tst.snapshot_neg.zcp \
	functional/channel_program/synctask_core/tst.snapshot_recursive.zcp \
	functional/channel_program/synctask_core/tst.snapshot_rename.zcp \
	functional/channel_program/synctask_core/tst.snapshot_simple.zcp \
	functional/checksum/default.cfg \
	functional/clean_mirror/clean_mirror_common.kshlib \
	functional/clean_mirror/default.cfg \
	functional/cli_root/cli_common.kshlib \
	functional/cli_root/zfs_copies/zfs_copies.cfg \
	functional/cli_root/zfs_copies/zfs_copies.kshlib \
	functional/cli_root/zfs_create/properties.kshlib \
	functional/cli_root/zfs_create/zfs_create.cfg \
	functional/cli_root/zfs_create/zfs_create_common.kshlib \
	functional/cli_root/zfs_destroy/zfs_destroy.cfg \
	functional/cli_root/zfs_destroy/zfs_destroy_common.kshlib \
	functional/cli_root/zfs_get/zfs_get_common.kshlib \
	functional/cli_root/zfs_get/zfs_get_list_d.kshlib \
	functional/cli_root/zfs_jail/jail.conf \
	functional/cli_root/zfs_load-key/HEXKEY \
	functional/cli_root/zfs_load-key/PASSPHRASE \
	functional/cli_root/zfs_load-key/RAWKEY \
	functional/cli_root/zfs_load-key/zfs_load-key.cfg \
	functional/cli_root/zfs_load-key/zfs_load-key_common.kshlib \
	functional/cli_root/zfs_mount/zfs_mount.cfg \
	functional/cli_root/zfs_mount/zfs_mount.kshlib \
	functional/cli_root/zfs_promote/zfs_promote.cfg \
	functional/cli_root/zfs_receive/zstd_test_data.txt \
	functional/cli_root/zfs_rename/zfs_rename.cfg \
	functional/cli_root/zfs_rename/zfs_rename.kshlib \
	functional/cli_root/zfs_rollback/zfs_rollback.cfg \
	functional/cli_root/zfs_rollback/zfs_rollback_common.kshlib \
	functional/cli_root/zfs_send/zfs_send.cfg \
	functional/cli_root/zfs_set/zfs_set_common.kshlib \
	functional/cli_root/zfs_share/zfs_share.cfg \
	functional/cli_root/zfs_snapshot/zfs_snapshot.cfg \
	functional/cli_root/zfs_unmount/zfs_unmount.cfg \
	functional/cli_root/zfs_unmount/zfs_unmount.kshlib \
	functional/cli_root/zfs_upgrade/zfs_upgrade.kshlib \
	functional/cli_root/zfs_wait/zfs_wait.kshlib \
	functional/cli_root/zpool_add/zpool_add.cfg \
	functional/cli_root/zpool_add/zpool_add.kshlib \
	functional/cli_root/zpool_clear/zpool_clear.cfg \
	functional/cli_root/zpool_create/draidcfg.gz \
	functional/cli_root/zpool_create/zpool_create.cfg \
	functional/cli_root/zpool_create/zpool_create.shlib \
	functional/cli_root/zpool_destroy/zpool_destroy.cfg \
	functional/cli_root/zpool_events/zpool_events.cfg \
	functional/cli_root/zpool_events/zpool_events.kshlib \
	functional/cli_root/zpool_expand/zpool_expand.cfg \
	functional/cli_root/zpool_export/zpool_export.cfg \
	functional/cli_root/zpool_export/zpool_export.kshlib \
	functional/cli_root/zpool_get/zpool_get.cfg \
	functional/cli_root/zpool_get/zpool_get_parsable.cfg \
	functional/cli_root/zpool_import/blockfiles/cryptv0.dat.bz2 \
	functional/cli_root/zpool_import/blockfiles/missing_ivset.dat.bz2 \
	functional/cli_root/zpool_import/blockfiles/unclean_export.dat.bz2 \
	functional/cli_root/zpool_import/zpool_import.cfg \
	functional/cli_root/zpool_import/zpool_import.kshlib \
	functional/cli_root/zpool_initialize/zpool_initialize.kshlib \
	functional/cli_root/zpool_labelclear/labelclear.cfg \
	functional/cli_root/zpool_remove/zpool_remove.cfg \
	functional/cli_root/zpool_reopen/zpool_reopen.cfg \
	functional/cli_root/zpool_reopen/zpool_reopen.shlib \
	functional/cli_root/zpool_resilver/zpool_resilver.cfg \
	functional/cli_root/zpool_scrub/zpool_scrub.cfg \
	functional/cli_root/zpool_split/zpool_split.cfg \
	functional/cli_root/zpool_trim/zpool_trim.kshlib \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v10.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v11.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v12.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v13.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v14.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v15.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz21.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz22.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz23.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v4.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v5.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v6.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v7.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v8.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v999.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v9.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-vBROKEN.dat.bz2 \
	functional/cli_root/zpool_upgrade/zpool_upgrade.cfg \
	functional/cli_root/zpool_upgrade/zpool_upgrade.kshlib \
	functional/cli_root/zpool_wait/zpool_wait.kshlib \
	functional/cli_user/misc/misc.cfg \
	functional/cli_user/zfs_list/zfs_list.cfg \
	functional/cli_user/zfs_list/zfs_list.kshlib \
	functional/compression/compress.cfg \
	functional/compression/testpool_zstd.tar.gz \
	functional/deadman/deadman.cfg \
	functional/delegate/delegate.cfg \
	functional/delegate/delegate_common.kshlib \
	functional/devices/devices.cfg \
	functional/devices/devices_common.kshlib \
	functional/events/events.cfg \
	functional/events/events_common.kshlib \
	functional/fault/fault.cfg \
	functional/grow/grow.cfg \
	functional/history/history.cfg \
	functional/history/history_common.kshlib \
	functional/history/i386.migratedpool.DAT.Z \
	functional/history/i386.orig_history.txt \
	functional/history/sparc.migratedpool.DAT.Z \
	functional/history/sparc.orig_history.txt \
	functional/history/zfs-pool-v4.dat.Z \
	functional/inheritance/config001.cfg \
	functional/inheritance/config002.cfg \
	functional/inheritance/config003.cfg \
	functional/inheritance/config004.cfg \
	functional/inheritance/config005.cfg \
	functional/inheritance/config006.cfg \
	functional/inheritance/config007.cfg \
	functional/inheritance/config008.cfg \
	functional/inheritance/config009.cfg \
	functional/inheritance/config010.cfg \
	functional/inheritance/config011.cfg \
	functional/inheritance/config012.cfg \
	functional/inheritance/config013.cfg \
	functional/inheritance/config014.cfg \
	functional/inheritance/config015.cfg \
	functional/inheritance/config016.cfg \
	functional/inheritance/config017.cfg \
	functional/inheritance/config018.cfg \
	functional/inheritance/config019.cfg \
	functional/inheritance/config020.cfg \
	functional/inheritance/config021.cfg \
	functional/inheritance/config022.cfg \
	functional/inheritance/config023.cfg \
	functional/inheritance/config024.cfg \
	functional/inheritance/inherit.kshlib \
	functional/inheritance/README.config \
	functional/inheritance/README.state \
	functional/inheritance/state001.cfg \
	functional/inheritance/state002.cfg \
	functional/inheritance/state003.cfg \
	functional/inheritance/state004.cfg \
	functional/inheritance/state005.cfg \
	functional/inheritance/state006.cfg \
	functional/inheritance/state007.cfg \
	functional/inheritance/state008.cfg \
	functional/inheritance/state009.cfg \
	functional/inheritance/state010.cfg \
	functional/inheritance/state011.cfg \
	functional/inheritance/state012.cfg \
	functional/inheritance/state013.cfg \
	functional/inheritance/state014.cfg \
	functional/inheritance/state015.cfg \
	functional/inheritance/state016.cfg \
	functional/inheritance/state017.cfg \
	functional/inheritance/state018.cfg \
	functional/inheritance/state019.cfg \
	functional/inheritance/state020.cfg \
	functional/inheritance/state021.cfg \
	functional/inheritance/state022.cfg \
	functional/inheritance/state023.cfg \
	functional/inheritance/state024.cfg \
	functional/inuse/inuse.cfg \
	functional/io/io.cfg \
	functional/l2arc/l2arc.cfg \
	functional/largest_pool/largest_pool.cfg \
	functional/migration/migration.cfg \
	functional/migration/migration.kshlib \
	functional/mmap/mmap.cfg \
	functional/mmp/mmp.cfg \
	functional/mmp/mmp.kshlib \
	functional/mv_files/mv_files.cfg \
	functional/mv_files/mv_files_common.kshlib \
	functional/nopwrite/nopwrite.shlib \
	functional/no_space/enospc.cfg \
	functional/online_offline/online_offline.cfg \
	functional/pool_checkpoint/pool_checkpoint.kshlib \
	functional/projectquota/projectquota.cfg \
	functional/projectquota/projectquota_common.kshlib \
	functional/quota/quota.cfg \
	functional/quota/quota.kshlib \
	functional/redacted_send/redacted.cfg \
	functional/redacted_send/redacted.kshlib \
	functional/redundancy/redundancy.cfg \
	functional/redundancy/redundancy.kshlib \
	functional/refreserv/refreserv.cfg \
	functional/removal/removal.kshlib \
	functional/replacement/replacement.cfg \
	functional/reservation/reservation.cfg \
	functional/reservation/reservation.shlib \
	functional/rsend/dedup_encrypted_zvol.bz2 \
	functional/rsend/dedup_encrypted_zvol.zsend.bz2 \
	functional/rsend/dedup.zsend.bz2 \
	functional/rsend/fs.tar.gz \
	functional/rsend/rsend.cfg \
	functional/rsend/rsend.kshlib \
	functional/scrub_mirror/default.cfg \
	functional/scrub_mirror/scrub_mirror_common.kshlib \
	functional/slog/slog.cfg \
	functional/slog/slog.kshlib \
	functional/snapshot/snapshot.cfg \
	functional/snapused/snapused.kshlib \
	functional/sparse/sparse.cfg \
	functional/trim/trim.cfg \
	functional/trim/trim.kshlib \
	functional/truncate/truncate.cfg \
	functional/upgrade/upgrade_common.kshlib \
	functional/user_namespace/user_namespace.cfg \
	functional/user_namespace/user_namespace_common.kshlib \
	functional/userquota/userquota.cfg \
	functional/userquota/userquota_common.kshlib \
	functional/vdev_zaps/vdev_zaps.kshlib \
	functional/xattr/xattr.cfg \
	functional/xattr/xattr_common.kshlib \
	functional/zvol/zvol.cfg \
	functional/zvol/zvol_cli/zvol_cli.cfg \
	functional/zvol/zvol_common.shlib \
	functional/zvol/zvol_ENOSPC/zvol_ENOSPC.cfg \
	functional/zvol/zvol_misc/zvol_misc_common.kshlib \
	functional/zvol/zvol_swap/zvol_swap.cfg

	nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
	functional/acl/off/cleanup.ksh \
	functional/acl/off/dosmode.ksh \
	functional/acl/off/posixmode.ksh \
	functional/acl/off/setup.ksh \
	functional/acl/posix/cleanup.ksh \
	functional/acl/posix/posix_001_pos.ksh \
	functional/acl/posix/posix_002_pos.ksh \
	functional/acl/posix/posix_003_pos.ksh \
	functional/acl/posix/posix_004_pos.ksh \
	functional/acl/posix-sa/cleanup.ksh \
	functional/acl/posix-sa/posix_001_pos.ksh \
	functional/acl/posix-sa/posix_002_pos.ksh \
	functional/acl/posix-sa/posix_003_pos.ksh \
	functional/acl/posix-sa/posix_004_pos.ksh \
	functional/acl/posix-sa/setup.ksh \
	functional/acl/posix/setup.ksh \
	functional/alloc_class/alloc_class_001_pos.ksh \
	functional/alloc_class/alloc_class_002_neg.ksh \
	functional/alloc_class/alloc_class_003_pos.ksh \
	functional/alloc_class/alloc_class_004_pos.ksh \
	functional/alloc_class/alloc_class_005_pos.ksh \
	functional/alloc_class/alloc_class_006_pos.ksh \
	functional/alloc_class/alloc_class_007_pos.ksh \
	functional/alloc_class/alloc_class_008_pos.ksh \
	functional/alloc_class/alloc_class_009_pos.ksh \
	functional/alloc_class/alloc_class_010_pos.ksh \
	functional/alloc_class/alloc_class_011_neg.ksh \
	functional/alloc_class/alloc_class_012_pos.ksh \
	functional/alloc_class/alloc_class_013_pos.ksh \
	functional/alloc_class/cleanup.ksh \
	functional/alloc_class/setup.ksh \
	functional/append/file_append.ksh \
	functional/append/threadsappend_001_pos.ksh \
	functional/append/cleanup.ksh \
	functional/append/setup.ksh \
	functional/arc/arcstats_runtime_tuning.ksh \
	functional/arc/cleanup.ksh \
	functional/arc/dbufstats_001_pos.ksh \
	functional/arc/dbufstats_002_pos.ksh \
	functional/arc/dbufstats_003_pos.ksh \
	functional/arc/setup.ksh \
	functional/atime/atime_001_pos.ksh \
	functional/atime/atime_002_neg.ksh \
	functional/atime/atime_003_pos.ksh \
	functional/atime/cleanup.ksh \
	functional/atime/root_atime_off.ksh \
	functional/atime/root_atime_on.ksh \
	functional/atime/root_relatime_on.ksh \
	functional/atime/setup.ksh \
	functional/bootfs/bootfs_001_pos.ksh \
	functional/bootfs/bootfs_002_neg.ksh \
	functional/bootfs/bootfs_003_pos.ksh \
	functional/bootfs/bootfs_004_neg.ksh \
	functional/bootfs/bootfs_005_neg.ksh \
	functional/bootfs/bootfs_006_pos.ksh \
	functional/bootfs/bootfs_007_pos.ksh \
	functional/bootfs/bootfs_008_pos.ksh \
	functional/bootfs/cleanup.ksh \
	functional/bootfs/setup.ksh \
	functional/btree/btree_negative.ksh \
	functional/btree/btree_positive.ksh \
	functional/cache/cache_001_pos.ksh \
	functional/cache/cache_002_pos.ksh \
	functional/cache/cache_003_pos.ksh \
	functional/cache/cache_004_neg.ksh \
	functional/cache/cache_005_neg.ksh \
	functional/cache/cache_006_pos.ksh \
	functional/cache/cache_007_neg.ksh \
	functional/cache/cache_008_neg.ksh \
	functional/cache/cache_009_pos.ksh \
	functional/cache/cache_010_pos.ksh \
	functional/cache/cache_011_pos.ksh \
	functional/cache/cache_012_pos.ksh \
	functional/cache/cleanup.ksh \
	functional/cachefile/cachefile_001_pos.ksh \
	functional/cachefile/cachefile_002_pos.ksh \
	functional/cachefile/cachefile_003_pos.ksh \
	functional/cachefile/cachefile_004_pos.ksh \
	functional/cachefile/cleanup.ksh \
	functional/cachefile/setup.ksh \
	functional/cache/setup.ksh \
	functional/casenorm/case_all_values.ksh \
	functional/casenorm/cleanup.ksh \
	functional/casenorm/insensitive_formd_delete.ksh \
	functional/casenorm/insensitive_formd_lookup.ksh \
	functional/casenorm/insensitive_none_delete.ksh \
	functional/casenorm/insensitive_none_lookup.ksh \
	functional/casenorm/mixed_create_failure.ksh \
	functional/casenorm/mixed_formd_delete.ksh \
	functional/casenorm/mixed_formd_lookup_ci.ksh \
	functional/casenorm/mixed_formd_lookup.ksh \
	functional/casenorm/mixed_none_delete.ksh \
	functional/casenorm/mixed_none_lookup_ci.ksh \
	functional/casenorm/mixed_none_lookup.ksh \
	functional/casenorm/norm_all_values.ksh \
	functional/casenorm/sensitive_formd_delete.ksh \
	functional/casenorm/sensitive_formd_lookup.ksh \
	functional/casenorm/sensitive_none_delete.ksh \
	functional/casenorm/sensitive_none_lookup.ksh \
	functional/casenorm/setup.ksh \
	functional/channel_program/lua_core/cleanup.ksh \
	functional/channel_program/lua_core/setup.ksh \
	functional/channel_program/lua_core/tst.args_to_lua.ksh \
	functional/channel_program/lua_core/tst.divide_by_zero.ksh \
	functional/channel_program/lua_core/tst.exists.ksh \
	functional/channel_program/lua_core/tst.integer_illegal.ksh \
	functional/channel_program/lua_core/tst.integer_overflow.ksh \
	functional/channel_program/lua_core/tst.language_functions_neg.ksh \
	functional/channel_program/lua_core/tst.language_functions_pos.ksh \
	functional/channel_program/lua_core/tst.large_prog.ksh \
	functional/channel_program/lua_core/tst.libraries.ksh \
	functional/channel_program/lua_core/tst.memory_limit.ksh \
	functional/channel_program/lua_core/tst.nested_neg.ksh \
	functional/channel_program/lua_core/tst.nested_pos.ksh \
	functional/channel_program/lua_core/tst.nvlist_to_lua.ksh \
	functional/channel_program/lua_core/tst.recursive_neg.ksh \
	functional/channel_program/lua_core/tst.recursive_pos.ksh \
	functional/channel_program/lua_core/tst.return_large.ksh \
	functional/channel_program/lua_core/tst.return_nvlist_neg.ksh \
	functional/channel_program/lua_core/tst.return_nvlist_pos.ksh \
	functional/channel_program/lua_core/tst.return_recursive_table.ksh \
	functional/channel_program/lua_core/tst.stack_gsub.ksh \
	functional/channel_program/lua_core/tst.timeout.ksh \
	functional/channel_program/synctask_core/cleanup.ksh \
	functional/channel_program/synctask_core/setup.ksh \
	functional/channel_program/synctask_core/tst.bookmark.copy.ksh \
	functional/channel_program/synctask_core/tst.bookmark.create.ksh \
	functional/channel_program/synctask_core/tst.destroy_fs.ksh \
	functional/channel_program/synctask_core/tst.destroy_snap.ksh \
	functional/channel_program/synctask_core/tst.get_count_and_limit.ksh \
	functional/channel_program/synctask_core/tst.get_index_props.ksh \
	functional/channel_program/synctask_core/tst.get_mountpoint.ksh \
	functional/channel_program/synctask_core/tst.get_neg.ksh \
	functional/channel_program/synctask_core/tst.get_number_props.ksh \
	functional/channel_program/synctask_core/tst.get_string_props.ksh \
	functional/channel_program/synctask_core/tst.get_type.ksh \
	functional/channel_program/synctask_core/tst.get_userquota.ksh \
	functional/channel_program/synctask_core/tst.get_written.ksh \
	functional/channel_program/synctask_core/tst.inherit.ksh \
	functional/channel_program/synctask_core/tst.list_bookmarks.ksh \
	functional/channel_program/synctask_core/tst.list_children.ksh \
	functional/channel_program/synctask_core/tst.list_clones.ksh \
	functional/channel_program/synctask_core/tst.list_holds.ksh \
	functional/channel_program/synctask_core/tst.list_snapshots.ksh \
	functional/channel_program/synctask_core/tst.list_system_props.ksh \
	functional/channel_program/synctask_core/tst.list_user_props.ksh \
	functional/channel_program/synctask_core/tst.parse_args_neg.ksh \
	functional/channel_program/synctask_core/tst.promote_conflict.ksh \
	functional/channel_program/synctask_core/tst.promote_multiple.ksh \
	functional/channel_program/synctask_core/tst.promote_simple.ksh \
	functional/channel_program/synctask_core/tst.rollback_mult.ksh \
	functional/channel_program/synctask_core/tst.rollback_one.ksh \
	functional/channel_program/synctask_core/tst.set_props.ksh \
	functional/channel_program/synctask_core/tst.snapshot_destroy.ksh \
	functional/channel_program/synctask_core/tst.snapshot_neg.ksh \
	functional/channel_program/synctask_core/tst.snapshot_recursive.ksh \
	functional/channel_program/synctask_core/tst.snapshot_rename.ksh \
	functional/channel_program/synctask_core/tst.snapshot_simple.ksh \
	functional/channel_program/synctask_core/tst.terminate_by_signal.ksh \
	functional/chattr/chattr_001_pos.ksh \
	functional/chattr/chattr_002_neg.ksh \
	functional/chattr/cleanup.ksh \
	functional/chattr/setup.ksh \
	functional/checksum/cleanup.ksh \
	functional/checksum/filetest_001_pos.ksh \
	functional/checksum/filetest_002_pos.ksh \
	functional/checksum/run_blake3_test.ksh \
	functional/checksum/run_edonr_test.ksh \
	functional/checksum/run_sha2_test.ksh \
	functional/checksum/run_skein_test.ksh \
	functional/checksum/setup.ksh \
	functional/clean_mirror/clean_mirror_001_pos.ksh \
	functional/clean_mirror/clean_mirror_002_pos.ksh \
	functional/clean_mirror/clean_mirror_003_pos.ksh \
	functional/clean_mirror/clean_mirror_004_pos.ksh \
	functional/clean_mirror/cleanup.ksh \
	functional/clean_mirror/setup.ksh \
	functional/cli_root/zdb/zdb_002_pos.ksh \
	functional/cli_root/zdb/zdb_003_pos.ksh \
	functional/cli_root/zdb/zdb_004_pos.ksh \
	functional/cli_root/zdb/zdb_005_pos.ksh \
	functional/cli_root/zdb/zdb_006_pos.ksh \
	functional/cli_root/zdb/zdb_args_neg.ksh \
	functional/cli_root/zdb/zdb_args_pos.ksh \
	functional/cli_root/zdb/zdb_block_size_histogram.ksh \
	functional/cli_root/zdb/zdb_checksum.ksh \
	functional/cli_root/zdb/zdb_decompress.ksh \
	functional/cli_root/zdb/zdb_decompress_zstd.ksh \
	functional/cli_root/zdb/zdb_display_block.ksh \
	functional/cli_root/zdb/zdb_label_checksum.ksh \
	functional/cli_root/zdb/zdb_object_range_neg.ksh \
	functional/cli_root/zdb/zdb_object_range_pos.ksh \
	functional/cli_root/zdb/zdb_objset_id.ksh \
	functional/cli_root/zdb/zdb_recover_2.ksh \
	functional/cli_root/zdb/zdb_recover.ksh \
	functional/cli_root/zfs_bookmark/cleanup.ksh \
	functional/cli_root/zfs_bookmark/setup.ksh \
	functional/cli_root/zfs_bookmark/zfs_bookmark_cliargs.ksh \
	functional/cli_root/zfs_change-key/cleanup.ksh \
	functional/cli_root/zfs_change-key/setup.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_child.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_clones.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_format.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_inherit.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_load.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_location.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_pbkdf2iters.ksh \
	functional/cli_root/zfs/cleanup.ksh \
	functional/cli_root/zfs_clone/cleanup.ksh \
	functional/cli_root/zfs_clone/setup.ksh \
	functional/cli_root/zfs_clone/zfs_clone_001_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_002_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_003_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_004_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_005_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_006_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_007_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_008_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_009_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_010_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_deeply_nested.ksh \
	functional/cli_root/zfs_clone/zfs_clone_encrypted.ksh \
	functional/cli_root/zfs_clone/zfs_clone_rm_nested.ksh \
	functional/cli_root/zfs_copies/cleanup.ksh \
	functional/cli_root/zfs_copies/setup.ksh \
	functional/cli_root/zfs_copies/zfs_copies_001_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_002_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_003_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_004_neg.ksh \
	functional/cli_root/zfs_copies/zfs_copies_005_neg.ksh \
	functional/cli_root/zfs_copies/zfs_copies_006_pos.ksh \
	functional/cli_root/zfs_create/cleanup.ksh \
	functional/cli_root/zfs_create/setup.ksh \
	functional/cli_root/zfs_create/zfs_create_001_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_002_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_003_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_004_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_005_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_006_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_007_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_008_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_009_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_010_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_011_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_012_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_013_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_014_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_crypt_combos.ksh \
	functional/cli_root/zfs_create/zfs_create_dryrun.ksh \
	functional/cli_root/zfs_create/zfs_create_encrypted.ksh \
	functional/cli_root/zfs_create/zfs_create_nomount.ksh \
	functional/cli_root/zfs_create/zfs_create_verbose.ksh \
	functional/cli_root/zfs_destroy/cleanup.ksh \
	functional/cli_root/zfs_destroy/setup.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_and_disable.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_races.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_dedup.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_001_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_002_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_003_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_004_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_005_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_006_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_007_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_008_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_009_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_010_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_011_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_012_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_013_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_014_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_015_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_016_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_clone_livelist.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_dev_removal_condense.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_dev_removal.ksh \
	functional/cli_root/zfs_diff/cleanup.ksh \
	functional/cli_root/zfs_diff/setup.ksh \
	functional/cli_root/zfs_diff/zfs_diff_changes.ksh \
	functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh \
	functional/cli_root/zfs_diff/zfs_diff_encrypted.ksh \
	functional/cli_root/zfs_diff/zfs_diff_mangle.ksh \
	functional/cli_root/zfs_diff/zfs_diff_timestamp.ksh \
	functional/cli_root/zfs_diff/zfs_diff_types.ksh \
	functional/cli_root/zfs_get/cleanup.ksh \
	functional/cli_root/zfs_get/setup.ksh \
	functional/cli_root/zfs_get/zfs_get_001_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_002_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_003_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_004_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_005_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_006_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_007_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_008_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_009_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_010_neg.ksh \
	functional/cli_root/zfs_ids_to_path/cleanup.ksh \
	functional/cli_root/zfs_ids_to_path/setup.ksh \
	functional/cli_root/zfs_ids_to_path/zfs_ids_to_path_001_pos.ksh \
	functional/cli_root/zfs_inherit/cleanup.ksh \
	functional/cli_root/zfs_inherit/setup.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_001_neg.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_002_neg.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_003_pos.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_mountpoint.ksh \
	functional/cli_root/zfs_jail/cleanup.ksh \
	functional/cli_root/zfs_jail/setup.ksh \
	functional/cli_root/zfs_jail/zfs_jail_001_pos.ksh \
	functional/cli_root/zfs_load-key/cleanup.ksh \
	functional/cli_root/zfs_load-key/setup.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_all.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_file.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_https.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_location.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_noop.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_recursive.ksh \
	functional/cli_root/zfs_mount/cleanup.ksh \
	functional/cli_root/zfs_mount/setup.ksh \
	functional/cli_root/zfs_mount/zfs_mount_001_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_002_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_003_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_004_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_005_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_006_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_007_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_008_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_009_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_010_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_011_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_012_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_013_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_014_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_001_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_fail.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_mountpoints.ksh \
	functional/cli_root/zfs_mount/zfs_mount_encrypted.ksh \
	functional/cli_root/zfs_mount/zfs_mount_remount.ksh \
	functional/cli_root/zfs_mount/zfs_mount_test_race.ksh \
	functional/cli_root/zfs_mount/zfs_multi_mount.ksh \
	functional/cli_root/zfs_program/cleanup.ksh \
	functional/cli_root/zfs_program/setup.ksh \
	functional/cli_root/zfs_program/zfs_program_json.ksh \
	functional/cli_root/zfs_promote/cleanup.ksh \
	functional/cli_root/zfs_promote/setup.ksh \
	functional/cli_root/zfs_promote/zfs_promote_001_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_002_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_003_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_004_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_005_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_006_neg.ksh \
	functional/cli_root/zfs_promote/zfs_promote_007_neg.ksh \
	functional/cli_root/zfs_promote/zfs_promote_008_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_encryptionroot.ksh \
	functional/cli_root/zfs_property/cleanup.ksh \
	functional/cli_root/zfs_property/setup.ksh \
	functional/cli_root/zfs_property/zfs_written_property_001_pos.ksh \
	functional/cli_root/zfs_receive/cleanup.ksh \
	functional/cli_root/zfs_receive/receive-o-x_props_aliases.ksh \
	functional/cli_root/zfs_receive/receive-o-x_props_override.ksh \
	functional/cli_root/zfs_receive/setup.ksh \
	functional/cli_root/zfs_receive/zfs_receive_001_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_002_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_003_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_004_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_005_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_006_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_007_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_008_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_009_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_010_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_011_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_012_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_013_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_014_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_015_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_016_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_-e.ksh \
	functional/cli_root/zfs_receive/zfs_receive_from_encrypted.ksh \
	functional/cli_root/zfs_receive/zfs_receive_from_zstd.ksh \
	functional/cli_root/zfs_receive/zfs_receive_new_props.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw_-d.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw_incremental.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw.ksh \
	functional/cli_root/zfs_receive/zfs_receive_to_encrypted.ksh \
	functional/cli_root/zfs_receive/zfs_receive_-wR-encrypted-mix.ksh \
	functional/cli_root/zfs_receive/zfs_receive_corrective.ksh \
	functional/cli_root/zfs_receive/zfs_receive_compressed_corrective.ksh \
	functional/cli_root/zfs_rename/cleanup.ksh \
	functional/cli_root/zfs_rename/setup.ksh \
	functional/cli_root/zfs_rename/zfs_rename_001_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_002_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_003_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_004_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_005_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_006_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_007_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_008_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_009_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_010_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_011_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_012_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_013_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_014_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_encrypted_child.ksh \
	functional/cli_root/zfs_rename/zfs_rename_mountpoint.ksh \
	functional/cli_root/zfs_rename/zfs_rename_nounmount.ksh \
	functional/cli_root/zfs_rename/zfs_rename_to_encrypted.ksh \
	functional/cli_root/zfs_reservation/cleanup.ksh \
	functional/cli_root/zfs_reservation/setup.ksh \
	functional/cli_root/zfs_reservation/zfs_reservation_001_pos.ksh \
	functional/cli_root/zfs_reservation/zfs_reservation_002_pos.ksh \
	functional/cli_root/zfs_rollback/cleanup.ksh \
	functional/cli_root/zfs_rollback/setup.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_001_pos.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_002_pos.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_003_neg.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_004_neg.ksh \
	functional/cli_root/zfs_send/cleanup.ksh \
	functional/cli_root/zfs_send/setup.ksh \
	functional/cli_root/zfs_send/zfs_send_001_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_002_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_003_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_004_neg.ksh \
	functional/cli_root/zfs_send/zfs_send_005_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_006_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_007_pos.ksh \
	functional/cli_root/zfs_send/zfs_send-b.ksh \
	functional/cli_root/zfs_send/zfs_send_encrypted.ksh \
	functional/cli_root/zfs_send/zfs_send_encrypted_unloaded.ksh \
	functional/cli_root/zfs_send/zfs_send_raw.ksh \
	functional/cli_root/zfs_send/zfs_send_skip_missing.ksh \
	functional/cli_root/zfs_send/zfs_send_sparse.ksh \
	functional/cli_root/zfs_set/cache_001_pos.ksh \
	functional/cli_root/zfs_set/cache_002_neg.ksh \
	functional/cli_root/zfs_set/canmount_001_pos.ksh \
	functional/cli_root/zfs_set/canmount_002_pos.ksh \
	functional/cli_root/zfs_set/canmount_003_pos.ksh \
	functional/cli_root/zfs_set/canmount_004_pos.ksh \
	functional/cli_root/zfs_set/checksum_001_pos.ksh \
	functional/cli_root/zfs_set/cleanup.ksh \
	functional/cli_root/zfs_set/compression_001_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_001_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_002_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_003_pos.ksh \
	functional/cli_root/zfs_set/onoffs_001_pos.ksh \
	functional/cli_root/zfs_set/property_alias_001_pos.ksh \
	functional/cli_root/zfs_set/readonly_001_pos.ksh \
	functional/cli_root/zfs_set/reservation_001_neg.ksh \
	functional/cli_root/zfs_set/ro_props_001_pos.ksh \
	functional/cli_root/zfs_set/setup.ksh \
	functional/cli_root/zfs_set/share_mount_001_neg.ksh \
	functional/cli_root/zfs_set/snapdir_001_pos.ksh \
	functional/cli_root/zfs/setup.ksh \
	functional/cli_root/zfs_set/user_property_001_pos.ksh \
	functional/cli_root/zfs_set/user_property_002_pos.ksh \
	functional/cli_root/zfs_set/user_property_003_neg.ksh \
	functional/cli_root/zfs_set/user_property_004_pos.ksh \
	functional/cli_root/zfs_set/version_001_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_001_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_002_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_003_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_feature_activation.ksh \
	functional/cli_root/zfs_set/zfs_set_keylocation.ksh \
	functional/cli_root/zfs_share/cleanup.ksh \
	functional/cli_root/zfs_share/setup.ksh \
	functional/cli_root/zfs_share/zfs_share_001_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_002_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_003_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_004_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_005_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_006_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_007_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_008_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_009_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_010_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_011_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_012_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_013_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_concurrent_shares.ksh \
	functional/cli_root/zfs_snapshot/cleanup.ksh \
	functional/cli_root/zfs_snapshot/setup.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_001_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_002_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_003_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_004_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_005_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_006_pos.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_007_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_008_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_009_pos.ksh \
	functional/cli_root/zfs_sysfs/cleanup.ksh \
	functional/cli_root/zfs_sysfs/setup.ksh \
	functional/cli_root/zfs_sysfs/zfeature_set_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_get_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_set_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_sysfs_live.ksh \
	functional/cli_root/zfs_sysfs/zpool_get_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zpool_set_unsupported.ksh \
	functional/cli_root/zfs_unload-key/cleanup.ksh \
	functional/cli_root/zfs_unload-key/setup.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key_all.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key_recursive.ksh \
	functional/cli_root/zfs_unmount/cleanup.ksh \
	functional/cli_root/zfs_unmount/setup.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_001_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_002_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_003_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_004_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_005_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_006_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_007_neg.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_008_neg.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_009_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_all_001_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_unload_keys.ksh \
	functional/cli_root/zfs_unshare/cleanup.ksh \
	functional/cli_root/zfs_unshare/setup.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_001_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_002_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_003_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_004_neg.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_005_neg.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_006_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_007_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_008_pos.ksh \
	functional/cli_root/zfs_upgrade/cleanup.ksh \
	functional/cli_root/zfs_upgrade/setup.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_001_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_002_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_003_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_004_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_005_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_006_neg.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_007_neg.ksh \
	functional/cli_root/zfs_wait/cleanup.ksh \
	functional/cli_root/zfs_wait/setup.ksh \
	functional/cli_root/zfs_wait/zfs_wait_deleteq.ksh \
	functional/cli_root/zfs_wait/zfs_wait_getsubopt.ksh \
	functional/cli_root/zfs/zfs_001_neg.ksh \
	functional/cli_root/zfs/zfs_002_pos.ksh \
	functional/cli_root/zfs/zfs_003_neg.ksh \
	functional/cli_root/zhack/zhack_label_checksum.ksh \
	functional/cli_root/zpool_add/add_nested_replacing_spare.ksh \
	functional/cli_root/zpool_add/add-o_ashift.ksh \
	functional/cli_root/zpool_add/add_prop_ashift.ksh \
	functional/cli_root/zpool_add/cleanup.ksh \
	functional/cli_root/zpool_add/setup.ksh \
	functional/cli_root/zpool_add/zpool_add_001_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_002_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_003_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_004_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_005_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_006_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_007_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_008_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_009_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_010_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_dryrun_output.ksh \
	functional/cli_root/zpool_attach/attach-o_ashift.ksh \
	functional/cli_root/zpool_attach/cleanup.ksh \
	functional/cli_root/zpool_attach/setup.ksh \
	functional/cli_root/zpool_attach/zpool_attach_001_neg.ksh \
	functional/cli_root/zpool/cleanup.ksh \
	functional/cli_root/zpool_clear/cleanup.ksh \
	functional/cli_root/zpool_clear/setup.ksh \
	functional/cli_root/zpool_clear/zpool_clear_001_pos.ksh \
	functional/cli_root/zpool_clear/zpool_clear_002_neg.ksh \
	functional/cli_root/zpool_clear/zpool_clear_003_neg.ksh \
	functional/cli_root/zpool_clear/zpool_clear_readonly.ksh \
	functional/cli_root/zpool_create/cleanup.ksh \
	functional/cli_root/zpool_create/create-o_ashift.ksh \
	functional/cli_root/zpool_create/setup.ksh \
	functional/cli_root/zpool_create/zpool_create_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_004_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_005_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_006_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_007_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_008_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_009_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_010_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_011_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_012_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_014_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_015_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_016_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_017_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_018_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_019_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_020_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_021_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_022_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_023_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_024_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_crypt_combos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_004_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_dryrun_output.ksh \
	functional/cli_root/zpool_create/zpool_create_encrypted.ksh \
	functional/cli_root/zpool_create/zpool_create_features_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_004_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_features_005_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_006_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_007_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_008_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_009_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_tempname.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_001_pos.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_002_pos.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_003_neg.ksh \
	functional/cli_root/zpool_detach/cleanup.ksh \
	functional/cli_root/zpool_detach/setup.ksh \
	functional/cli_root/zpool_detach/zpool_detach_001_neg.ksh \
	functional/cli_root/zpool_events/cleanup.ksh \
	functional/cli_root/zpool_events/setup.ksh \
	functional/cli_root/zpool_events/zpool_events_clear.ksh \
	functional/cli_root/zpool_events/zpool_events_clear_retained.ksh \
	functional/cli_root/zpool_events/zpool_events_cliargs.ksh \
	functional/cli_root/zpool_events/zpool_events_duplicates.ksh \
	functional/cli_root/zpool_events/zpool_events_errors.ksh \
	functional/cli_root/zpool_events/zpool_events_follow.ksh \
	functional/cli_root/zpool_events/zpool_events_poolname.ksh \
	functional/cli_root/zpool_expand/cleanup.ksh \
	functional/cli_root/zpool_expand/setup.ksh \
	functional/cli_root/zpool_expand/zpool_expand_001_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_002_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_003_neg.ksh \
	functional/cli_root/zpool_expand/zpool_expand_004_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_005_pos.ksh \
	functional/cli_root/zpool_export/cleanup.ksh \
	functional/cli_root/zpool_export/setup.ksh \
	functional/cli_root/zpool_export/zpool_export_001_pos.ksh \
	functional/cli_root/zpool_export/zpool_export_002_pos.ksh \
	functional/cli_root/zpool_export/zpool_export_003_neg.ksh \
	functional/cli_root/zpool_export/zpool_export_004_pos.ksh \
	functional/cli_root/zpool_get/cleanup.ksh \
	functional/cli_root/zpool_get/setup.ksh \
	functional/cli_root/zpool_get/zpool_get_001_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_002_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_003_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_004_neg.ksh \
	functional/cli_root/zpool_get/zpool_get_005_pos.ksh \
	functional/cli_root/zpool_history/cleanup.ksh \
	functional/cli_root/zpool_history/setup.ksh \
	functional/cli_root/zpool_history/zpool_history_001_neg.ksh \
	functional/cli_root/zpool_history/zpool_history_002_pos.ksh \
	functional/cli_root/zpool_import/cleanup.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_added.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_removed.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_replaced.ksh \
	functional/cli_root/zpool_import/import_cachefile_mirror_attached.ksh \
	functional/cli_root/zpool_import/import_cachefile_mirror_detached.ksh \
	functional/cli_root/zpool_import/import_cachefile_paths_changed.ksh \
	functional/cli_root/zpool_import/import_cachefile_shared_device.ksh \
	functional/cli_root/zpool_import/import_devices_missing.ksh \
	functional/cli_root/zpool_import/import_paths_changed.ksh \
	functional/cli_root/zpool_import/import_rewind_config_changed.ksh \
	functional/cli_root/zpool_import/import_rewind_device_replaced.ksh \
	functional/cli_root/zpool_import/setup.ksh \
	functional/cli_root/zpool_import/zpool_import_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_002_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_004_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_005_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_006_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_007_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_008_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_009_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_010_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_011_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_012_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_013_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_014_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_015_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_016_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_017_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_all_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_encrypted.ksh \
	functional/cli_root/zpool_import/zpool_import_encrypted_load.ksh \
	functional/cli_root/zpool_import/zpool_import_errata3.ksh \
	functional/cli_root/zpool_import/zpool_import_errata4.ksh \
	functional/cli_root/zpool_import/zpool_import_features_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_features_002_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_features_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_002_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_rename_001_pos.ksh \
	functional/cli_root/zpool_initialize/cleanup.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_attach_detach_add_remove.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_fault_export_import_online.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_import_export.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_offline_export_import_online.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_online_offline.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_split.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_neg.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_pos.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_suspend_resume.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_unsupported_vdevs.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_verify_checksums.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_verify_initialized.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_active.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_exported.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_removed.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_valid.ksh \
	functional/cli_root/zpool_offline/cleanup.ksh \
	functional/cli_root/zpool_offline/setup.ksh \
	functional/cli_root/zpool_offline/zpool_offline_001_pos.ksh \
	functional/cli_root/zpool_offline/zpool_offline_002_neg.ksh \
	functional/cli_root/zpool_offline/zpool_offline_003_pos.ksh \
	functional/cli_root/zpool_online/cleanup.ksh \
	functional/cli_root/zpool_online/setup.ksh \
	functional/cli_root/zpool_online/zpool_online_001_pos.ksh \
	functional/cli_root/zpool_online/zpool_online_002_neg.ksh \
	functional/cli_root/zpool_remove/cleanup.ksh \
	functional/cli_root/zpool_remove/setup.ksh \
	functional/cli_root/zpool_remove/zpool_remove_001_neg.ksh \
	functional/cli_root/zpool_remove/zpool_remove_002_pos.ksh \
	functional/cli_root/zpool_remove/zpool_remove_003_pos.ksh \
	functional/cli_root/zpool_reopen/cleanup.ksh \
	functional/cli_root/zpool_reopen/setup.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_001_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_002_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_003_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_004_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_005_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_006_neg.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_007_pos.ksh \
	functional/cli_root/zpool_replace/cleanup.ksh \
	functional/cli_root/zpool_replace/replace-o_ashift.ksh \
	functional/cli_root/zpool_replace/replace_prop_ashift.ksh \
	functional/cli_root/zpool_replace/setup.ksh \
	functional/cli_root/zpool_replace/zpool_replace_001_neg.ksh \
	functional/cli_root/zpool_resilver/cleanup.ksh \
	functional/cli_root/zpool_resilver/setup.ksh \
	functional/cli_root/zpool_resilver/zpool_resilver_bad_args.ksh \
	functional/cli_root/zpool_resilver/zpool_resilver_restart.ksh \
	functional/cli_root/zpool_scrub/cleanup.ksh \
	functional/cli_root/zpool_scrub/setup.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_001_neg.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_002_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_003_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_004_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_005_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_encrypted_unloaded.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_multiple_copies.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_offline_device.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_print_repairing.ksh \
	functional/cli_root/zpool_set/cleanup.ksh \
	functional/cli_root/zpool_set/setup.ksh \
	functional/cli_root/zpool/setup.ksh \
	functional/cli_root/zpool_set/zpool_set_001_pos.ksh \
	functional/cli_root/zpool_set/zpool_set_002_neg.ksh \
	functional/cli_root/zpool_set/zpool_set_003_neg.ksh \
	functional/cli_root/zpool_set/zpool_set_ashift.ksh \
	functional/cli_root/zpool_set/zpool_set_features.ksh \
	functional/cli_root/zpool_split/cleanup.ksh \
	functional/cli_root/zpool_split/setup.ksh \
	functional/cli_root/zpool_split/zpool_split_cliargs.ksh \
	functional/cli_root/zpool_split/zpool_split_devices.ksh \
	functional/cli_root/zpool_split/zpool_split_dryrun_output.ksh \
	functional/cli_root/zpool_split/zpool_split_encryption.ksh \
	functional/cli_root/zpool_split/zpool_split_indirect.ksh \
	functional/cli_root/zpool_split/zpool_split_props.ksh \
	functional/cli_root/zpool_split/zpool_split_resilver.ksh \
	functional/cli_root/zpool_split/zpool_split_vdevs.ksh \
	functional/cli_root/zpool_split/zpool_split_wholedisk.ksh \
	functional/cli_root/zpool_status/cleanup.ksh \
	functional/cli_root/zpool_status/setup.ksh \
	functional/cli_root/zpool_status/zpool_status_001_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_002_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_003_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_004_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_005_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_features_001_pos.ksh \
	functional/cli_root/zpool_sync/cleanup.ksh \
	functional/cli_root/zpool_sync/setup.ksh \
	functional/cli_root/zpool_sync/zpool_sync_001_pos.ksh \
	functional/cli_root/zpool_sync/zpool_sync_002_neg.ksh \
	functional/cli_root/zpool_trim/cleanup.ksh \
	functional/cli_root/zpool_trim/setup.ksh \
	functional/cli_root/zpool_trim/zpool_trim_attach_detach_add_remove.ksh \
	functional/cli_root/zpool_trim/zpool_trim_fault_export_import_online.ksh \
	functional/cli_root/zpool_trim/zpool_trim_import_export.ksh \
	functional/cli_root/zpool_trim/zpool_trim_multiple.ksh \
	functional/cli_root/zpool_trim/zpool_trim_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_offline_export_import_online.ksh \
	functional/cli_root/zpool_trim/zpool_trim_online_offline.ksh \
	functional/cli_root/zpool_trim/zpool_trim_partial.ksh \
	functional/cli_root/zpool_trim/zpool_trim_rate.ksh \
	functional/cli_root/zpool_trim/zpool_trim_rate_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_secure.ksh \
	functional/cli_root/zpool_trim/zpool_trim_split.ksh \
	functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_pos.ksh \
	functional/cli_root/zpool_trim/zpool_trim_suspend_resume.ksh \
	functional/cli_root/zpool_trim/zpool_trim_unsupported_vdevs.ksh \
	functional/cli_root/zpool_trim/zpool_trim_verify_checksums.ksh \
	functional/cli_root/zpool_trim/zpool_trim_verify_trimmed.ksh \
	functional/cli_root/zpool_upgrade/cleanup.ksh \
	functional/cli_root/zpool_upgrade/setup.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_001_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_002_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_003_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_004_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_005_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_006_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_007_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_008_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_009_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_features_001_pos.ksh \
	functional/cli_root/zpool_wait/cleanup.ksh \
	functional/cli_root/zpool_wait/scan/cleanup.ksh \
	functional/cli_root/zpool_wait/scan/setup.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_rebuild.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_replace_cancel.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_replace.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_resilver.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_basic.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_cancel.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_flag.ksh \
	functional/cli_root/zpool_wait/setup.ksh \
	functional/cli_root/zpool_wait/zpool_wait_discard.ksh \
	functional/cli_root/zpool_wait/zpool_wait_freeing.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_basic.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_flag.ksh \
	functional/cli_root/zpool_wait/zpool_wait_multiple.ksh \
	functional/cli_root/zpool_wait/zpool_wait_no_activity.ksh \
	functional/cli_root/zpool_wait/zpool_wait_remove_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_remove.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_basic.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_flag.ksh \
	functional/cli_root/zpool_wait/zpool_wait_usage.ksh \
	functional/cli_root/zpool/zpool_001_neg.ksh \
	functional/cli_root/zpool/zpool_002_pos.ksh \
	functional/cli_root/zpool/zpool_003_pos.ksh \
	functional/cli_root/zpool/zpool_colors.ksh \
	functional/cli_user/misc/arcstat_001_pos.ksh \
	functional/cli_user/misc/arc_summary_001_pos.ksh \
	functional/cli_user/misc/arc_summary_002_neg.ksh \
	functional/cli_user/misc/zilstat_001_pos.ksh \
	functional/cli_user/misc/cleanup.ksh \
	functional/cli_user/misc/setup.ksh \
	functional/cli_user/misc/zdb_001_neg.ksh \
	functional/cli_user/misc/zfs_001_neg.ksh \
	functional/cli_user/misc/zfs_allow_001_neg.ksh \
	functional/cli_user/misc/zfs_clone_001_neg.ksh \
	functional/cli_user/misc/zfs_create_001_neg.ksh \
	functional/cli_user/misc/zfs_destroy_001_neg.ksh \
	functional/cli_user/misc/zfs_get_001_neg.ksh \
	functional/cli_user/misc/zfs_inherit_001_neg.ksh \
	functional/cli_user/misc/zfs_mount_001_neg.ksh \
	functional/cli_user/misc/zfs_promote_001_neg.ksh \
	functional/cli_user/misc/zfs_receive_001_neg.ksh \
	functional/cli_user/misc/zfs_rename_001_neg.ksh \
	functional/cli_user/misc/zfs_rollback_001_neg.ksh \
	functional/cli_user/misc/zfs_send_001_neg.ksh \
	functional/cli_user/misc/zfs_set_001_neg.ksh \
	functional/cli_user/misc/zfs_share_001_neg.ksh \
	functional/cli_user/misc/zfs_snapshot_001_neg.ksh \
	functional/cli_user/misc/zfs_unallow_001_neg.ksh \
	functional/cli_user/misc/zfs_unmount_001_neg.ksh \
	functional/cli_user/misc/zfs_unshare_001_neg.ksh \
	functional/cli_user/misc/zfs_upgrade_001_neg.ksh \
	functional/cli_user/misc/zpool_001_neg.ksh \
	functional/cli_user/misc/zpool_add_001_neg.ksh \
	functional/cli_user/misc/zpool_attach_001_neg.ksh \
	functional/cli_user/misc/zpool_clear_001_neg.ksh \
	functional/cli_user/misc/zpool_create_001_neg.ksh \
	functional/cli_user/misc/zpool_destroy_001_neg.ksh \
	functional/cli_user/misc/zpool_detach_001_neg.ksh \
	functional/cli_user/misc/zpool_export_001_neg.ksh \
	functional/cli_user/misc/zpool_get_001_neg.ksh \
	functional/cli_user/misc/zpool_history_001_neg.ksh \
	functional/cli_user/misc/zpool_import_001_neg.ksh \
	functional/cli_user/misc/zpool_import_002_neg.ksh \
	functional/cli_user/misc/zpool_offline_001_neg.ksh \
	functional/cli_user/misc/zpool_online_001_neg.ksh \
	functional/cli_user/misc/zpool_remove_001_neg.ksh \
	functional/cli_user/misc/zpool_replace_001_neg.ksh \
	functional/cli_user/misc/zpool_scrub_001_neg.ksh \
	functional/cli_user/misc/zpool_set_001_neg.ksh \
	functional/cli_user/misc/zpool_status_001_neg.ksh \
	functional/cli_user/misc/zpool_upgrade_001_neg.ksh \
	functional/cli_user/misc/zpool_wait_privilege.ksh \
	functional/cli_user/zfs_list/cleanup.ksh \
	functional/cli_user/zfs_list/setup.ksh \
	functional/cli_user/zfs_list/zfs_list_001_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_002_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_003_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_004_neg.ksh \
	functional/cli_user/zfs_list/zfs_list_005_neg.ksh \
	functional/cli_user/zfs_list/zfs_list_007_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_008_neg.ksh \
	functional/cli_user/zpool_iostat/cleanup.ksh \
	functional/cli_user/zpool_iostat/setup.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_001_neg.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_002_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_003_neg.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_004_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_005_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_disable.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_homedir.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_searchpath.ksh \
	functional/cli_user/zpool_list/cleanup.ksh \
	functional/cli_user/zpool_list/setup.ksh \
	functional/cli_user/zpool_list/zpool_list_001_pos.ksh \
	functional/cli_user/zpool_list/zpool_list_002_neg.ksh \
	functional/cli_user/zpool_status/cleanup.ksh \
	functional/cli_user/zpool_status/setup.ksh \
	functional/cli_user/zpool_status/zpool_status_003_pos.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_disable.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_homedir.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_searchpath.ksh \
	functional/compression/cleanup.ksh \
	functional/compression/compress_001_pos.ksh \
	functional/compression/compress_002_pos.ksh \
	functional/compression/compress_003_pos.ksh \
	functional/compression/compress_004_pos.ksh \
	functional/compression/compress_zstd_bswap.ksh \
	functional/compression/l2arc_compressed_arc_disabled.ksh \
	functional/compression/l2arc_compressed_arc.ksh \
	functional/compression/l2arc_encrypted.ksh \
	functional/compression/l2arc_encrypted_no_compressed_arc.ksh \
	functional/compression/setup.ksh \
	functional/cp_files/cleanup.ksh \
	functional/cp_files/cp_files_001_pos.ksh \
	functional/cp_files/setup.ksh \
	functional/crtime/cleanup.ksh \
	functional/crtime/crtime_001_pos.ksh \
	functional/crtime/setup.ksh \
	functional/ctime/cleanup.ksh \
	functional/ctime/ctime_001_pos.ksh \
	functional/ctime/setup.ksh \
	functional/deadman/deadman_ratelimit.ksh \
	functional/deadman/deadman_sync.ksh \
	functional/deadman/deadman_zio.ksh \
	functional/delegate/cleanup.ksh \
	functional/delegate/setup.ksh \
	functional/delegate/zfs_allow_001_pos.ksh \
	functional/delegate/zfs_allow_002_pos.ksh \
	functional/delegate/zfs_allow_003_pos.ksh \
	functional/delegate/zfs_allow_004_pos.ksh \
	functional/delegate/zfs_allow_005_pos.ksh \
	functional/delegate/zfs_allow_006_pos.ksh \
	functional/delegate/zfs_allow_007_pos.ksh \
	functional/delegate/zfs_allow_008_pos.ksh \
	functional/delegate/zfs_allow_009_neg.ksh \
	functional/delegate/zfs_allow_010_pos.ksh \
	functional/delegate/zfs_allow_011_neg.ksh \
	functional/delegate/zfs_allow_012_neg.ksh \
	functional/delegate/zfs_unallow_001_pos.ksh \
	functional/delegate/zfs_unallow_002_pos.ksh \
	functional/delegate/zfs_unallow_003_pos.ksh \
	functional/delegate/zfs_unallow_004_pos.ksh \
	functional/delegate/zfs_unallow_005_pos.ksh \
	functional/delegate/zfs_unallow_006_pos.ksh \
	functional/delegate/zfs_unallow_007_neg.ksh \
	functional/delegate/zfs_unallow_008_neg.ksh \
	functional/devices/cleanup.ksh \
	functional/devices/devices_001_pos.ksh \
	functional/devices/devices_002_neg.ksh \
	functional/devices/devices_003_pos.ksh \
	functional/devices/setup.ksh \
	functional/dos_attributes/cleanup.ksh \
	functional/dos_attributes/read_dos_attrs_001.ksh \
	functional/dos_attributes/setup.ksh \
	functional/dos_attributes/write_dos_attrs_001.ksh \
	functional/events/cleanup.ksh \
	functional/events/events_001_pos.ksh \
	functional/events/events_002_pos.ksh \
	functional/events/setup.ksh \
	functional/events/zed_fd_spill.ksh \
	functional/events/zed_rc_filter.ksh \
	functional/exec/cleanup.ksh \
	functional/exec/exec_001_pos.ksh \
	functional/exec/exec_002_neg.ksh \
	functional/exec/setup.ksh \
	functional/fadvise/cleanup.ksh \
	functional/fadvise/fadvise_sequential.ksh \
	functional/fadvise/setup.ksh \
	functional/fallocate/cleanup.ksh \
	functional/fallocate/fallocate_prealloc.ksh \
	functional/fallocate/fallocate_punch-hole.ksh \
	functional/fallocate/fallocate_zero-range.ksh \
	functional/fallocate/setup.ksh \
	functional/fault/auto_offline_001_pos.ksh \
	functional/fault/auto_online_001_pos.ksh \
	functional/fault/auto_online_002_pos.ksh \
	functional/fault/auto_replace_001_pos.ksh \
	functional/fault/auto_spare_001_pos.ksh \
	functional/fault/auto_spare_002_pos.ksh \
	functional/fault/auto_spare_ashift.ksh \
	functional/fault/auto_spare_multiple.ksh \
	functional/fault/auto_spare_shared.ksh \
	functional/fault/cleanup.ksh \
	functional/fault/decompress_fault.ksh \
	functional/fault/decrypt_fault.ksh \
	functional/fault/scrub_after_resilver.ksh \
	functional/fault/setup.ksh \
	functional/fault/zpool_status_-s.ksh \
	functional/features/async_destroy/async_destroy_001_pos.ksh \
	functional/features/async_destroy/cleanup.ksh \
	functional/features/async_destroy/setup.ksh \
	functional/features/large_dnode/cleanup.ksh \
	functional/features/large_dnode/large_dnode_001_pos.ksh \
	functional/features/large_dnode/large_dnode_002_pos.ksh \
	functional/features/large_dnode/large_dnode_003_pos.ksh \
	functional/features/large_dnode/large_dnode_004_neg.ksh \
	functional/features/large_dnode/large_dnode_005_pos.ksh \
	functional/features/large_dnode/large_dnode_006_pos.ksh \
	functional/features/large_dnode/large_dnode_007_neg.ksh \
	functional/features/large_dnode/large_dnode_008_pos.ksh \
	functional/features/large_dnode/large_dnode_009_pos.ksh \
	functional/features/large_dnode/setup.ksh \
	functional/grow/grow_pool_001_pos.ksh \
	functional/grow/grow_replicas_001_pos.ksh \
	functional/history/cleanup.ksh \
	functional/history/history_001_pos.ksh \
	functional/history/history_002_pos.ksh \
	functional/history/history_003_pos.ksh \
	functional/history/history_004_pos.ksh \
	functional/history/history_005_neg.ksh \
	functional/history/history_006_neg.ksh \
	functional/history/history_007_pos.ksh \
	functional/history/history_008_pos.ksh \
	functional/history/history_009_pos.ksh \
	functional/history/history_010_pos.ksh \
	functional/history/setup.ksh \
	functional/inheritance/cleanup.ksh \
	functional/inheritance/inherit_001_pos.ksh \
	functional/inuse/inuse_001_pos.ksh \
	functional/inuse/inuse_003_pos.ksh \
	functional/inuse/inuse_004_pos.ksh \
	functional/inuse/inuse_005_pos.ksh \
	functional/inuse/inuse_006_pos.ksh \
	functional/inuse/inuse_007_pos.ksh \
	functional/inuse/inuse_008_pos.ksh \
	functional/inuse/inuse_009_pos.ksh \
	functional/inuse/setup.ksh \
	functional/io/cleanup.ksh \
	functional/io/io_uring.ksh \
	functional/io/libaio.ksh \
	functional/io/mmap.ksh \
	functional/io/posixaio.ksh \
	functional/io/psync.ksh \
	functional/io/setup.ksh \
	functional/io/sync.ksh \
	functional/l2arc/cleanup.ksh \
	functional/l2arc/l2arc_arcstats_pos.ksh \
	functional/l2arc/l2arc_l2miss_pos.ksh \
	functional/l2arc/l2arc_mfuonly_pos.ksh \
	functional/l2arc/persist_l2arc_001_pos.ksh \
	functional/l2arc/persist_l2arc_002_pos.ksh \
	functional/l2arc/persist_l2arc_003_neg.ksh \
	functional/l2arc/persist_l2arc_004_pos.ksh \
	functional/l2arc/persist_l2arc_005_pos.ksh \
	functional/l2arc/setup.ksh \
	functional/large_files/cleanup.ksh \
	functional/large_files/large_files_001_pos.ksh \
	functional/large_files/large_files_002_pos.ksh \
	functional/large_files/setup.ksh \
	functional/largest_pool/largest_pool_001_pos.ksh \
	functional/libzfs/cleanup.ksh \
	functional/libzfs/libzfs_input.ksh \
	functional/libzfs/setup.ksh \
	functional/limits/cleanup.ksh \
	functional/limits/filesystem_count.ksh \
	functional/limits/filesystem_limit.ksh \
	functional/limits/setup.ksh \
	functional/limits/snapshot_count.ksh \
	functional/limits/snapshot_limit.ksh \
	functional/link_count/cleanup.ksh \
	functional/link_count/link_count_001.ksh \
	functional/link_count/link_count_root_inode.ksh \
	functional/link_count/setup.ksh \
	functional/log_spacemap/log_spacemap_import_logs.ksh \
	functional/migration/cleanup.ksh \
	functional/migration/migration_001_pos.ksh \
	functional/migration/migration_002_pos.ksh \
	functional/migration/migration_003_pos.ksh \
	functional/migration/migration_004_pos.ksh \
	functional/migration/migration_005_pos.ksh \
	functional/migration/migration_006_pos.ksh \
	functional/migration/migration_007_pos.ksh \
	functional/migration/migration_008_pos.ksh \
	functional/migration/migration_009_pos.ksh \
	functional/migration/migration_010_pos.ksh \
	functional/migration/migration_011_pos.ksh \
	functional/migration/migration_012_pos.ksh \
	functional/migration/setup.ksh \
	functional/mmap/cleanup.ksh \
	functional/mmap/mmap_libaio_001_pos.ksh \
	functional/mmap/mmap_read_001_pos.ksh \
	functional/mmap/mmap_seek_001_pos.ksh \
	functional/mmap/mmap_sync_001_pos.ksh \
	functional/mmap/mmap_write_001_pos.ksh \
	functional/mmap/setup.ksh \
	functional/mmp/cleanup.ksh \
	functional/mmp/mmp_active_import.ksh \
	functional/mmp/mmp_exported_import.ksh \
	functional/mmp/mmp_hostid.ksh \
	functional/mmp/mmp_inactive_import.ksh \
	functional/mmp/mmp_interval.ksh \
	functional/mmp/mmp_on_off.ksh \
	functional/mmp/mmp_on_thread.ksh \
	functional/mmp/mmp_on_uberblocks.ksh \
	functional/mmp/mmp_on_zdb.ksh \
	functional/mmp/mmp_reset_interval.ksh \
	functional/mmp/mmp_write_distribution.ksh \
	functional/mmp/mmp_write_uberblocks.ksh \
	functional/mmp/multihost_history.ksh \
	functional/mmp/setup.ksh \
	functional/mount/cleanup.ksh \
	functional/mount/setup.ksh \
	functional/mount/umount_001.ksh \
	functional/mount/umountall_001.ksh \
	functional/mount/umount_unlinked_drain.ksh \
	functional/mv_files/cleanup.ksh \
	functional/mv_files/mv_files_001_pos.ksh \
	functional/mv_files/mv_files_002_pos.ksh \
	functional/mv_files/random_creation.ksh \
	functional/mv_files/setup.ksh \
	functional/nestedfs/cleanup.ksh \
	functional/nestedfs/nestedfs_001_pos.ksh \
	functional/nestedfs/setup.ksh \
	functional/nopwrite/cleanup.ksh \
	functional/nopwrite/nopwrite_copies.ksh \
	functional/nopwrite/nopwrite_mtime.ksh \
	functional/nopwrite/nopwrite_negative.ksh \
	functional/nopwrite/nopwrite_promoted_clone.ksh \
	functional/nopwrite/nopwrite_recsize.ksh \
	functional/nopwrite/nopwrite_sync.ksh \
	functional/nopwrite/nopwrite_varying_compression.ksh \
	functional/nopwrite/nopwrite_volume.ksh \
	functional/nopwrite/setup.ksh \
	functional/no_space/cleanup.ksh \
	functional/no_space/enospc_001_pos.ksh \
	functional/no_space/enospc_002_pos.ksh \
	functional/no_space/enospc_003_pos.ksh \
	functional/no_space/enospc_df.ksh \
	functional/no_space/enospc_rm.ksh \
	functional/no_space/setup.ksh \
	functional/online_offline/cleanup.ksh \
	functional/online_offline/online_offline_001_pos.ksh \
	functional/online_offline/online_offline_002_neg.ksh \
	functional/online_offline/online_offline_003_neg.ksh \
	functional/online_offline/setup.ksh \
	functional/pam/cleanup.ksh \
	functional/pam/pam_basic.ksh \
	functional/pam/pam_nounmount.ksh \
	functional/pam/pam_short_password.ksh \
	functional/pam/setup.ksh \
	functional/pool_checkpoint/checkpoint_after_rewind.ksh \
	functional/pool_checkpoint/checkpoint_big_rewind.ksh \
	functional/pool_checkpoint/checkpoint_capacity.ksh \
	functional/pool_checkpoint/checkpoint_conf_change.ksh \
	functional/pool_checkpoint/checkpoint_discard_busy.ksh \
	functional/pool_checkpoint/checkpoint_discard.ksh \
	functional/pool_checkpoint/checkpoint_discard_many.ksh \
	functional/pool_checkpoint/checkpoint_indirect.ksh \
	functional/pool_checkpoint/checkpoint_invalid.ksh \
	functional/pool_checkpoint/checkpoint_lun_expsz.ksh \
	functional/pool_checkpoint/checkpoint_open.ksh \
	functional/pool_checkpoint/checkpoint_removal.ksh \
	functional/pool_checkpoint/checkpoint_rewind.ksh \
	functional/pool_checkpoint/checkpoint_ro_rewind.ksh \
	functional/pool_checkpoint/checkpoint_sm_scale.ksh \
	functional/pool_checkpoint/checkpoint_twice.ksh \
	functional/pool_checkpoint/checkpoint_vdev_add.ksh \
	functional/pool_checkpoint/checkpoint_zdb.ksh \
	functional/pool_checkpoint/checkpoint_zhack_feat.ksh \
	functional/pool_checkpoint/cleanup.ksh \
	functional/pool_checkpoint/setup.ksh \
	functional/pool_names/pool_names_001_pos.ksh \
	functional/pool_names/pool_names_002_neg.ksh \
	functional/poolversion/cleanup.ksh \
	functional/poolversion/poolversion_001_pos.ksh \
	functional/poolversion/poolversion_002_pos.ksh \
	functional/poolversion/setup.ksh \
	functional/privilege/cleanup.ksh \
	functional/privilege/privilege_001_pos.ksh \
	functional/privilege/privilege_002_pos.ksh \
	functional/privilege/setup.ksh \
	functional/procfs/cleanup.ksh \
	functional/procfs/pool_state.ksh \
	functional/procfs/procfs_list_basic.ksh \
	functional/procfs/procfs_list_concurrent_readers.ksh \
	functional/procfs/procfs_list_stale_read.ksh \
	functional/procfs/setup.ksh \
	functional/projectquota/cleanup.ksh \
	functional/projectquota/projectid_001_pos.ksh \
	functional/projectquota/projectid_002_pos.ksh \
	functional/projectquota/projectid_003_pos.ksh \
	functional/projectquota/projectquota_001_pos.ksh \
	functional/projectquota/projectquota_002_pos.ksh \
	functional/projectquota/projectquota_003_pos.ksh \
	functional/projectquota/projectquota_004_neg.ksh \
	functional/projectquota/projectquota_005_pos.ksh \
	functional/projectquota/projectquota_006_pos.ksh \
	functional/projectquota/projectquota_007_pos.ksh \
	functional/projectquota/projectquota_008_pos.ksh \
	functional/projectquota/projectquota_009_pos.ksh \
	functional/projectquota/projectspace_001_pos.ksh \
	functional/projectquota/projectspace_002_pos.ksh \
	functional/projectquota/projectspace_003_pos.ksh \
	functional/projectquota/projectspace_004_pos.ksh \
	functional/projectquota/projecttree_001_pos.ksh \
	functional/projectquota/projecttree_002_pos.ksh \
	functional/projectquota/projecttree_003_neg.ksh \
	functional/projectquota/setup.ksh \
	functional/quota/cleanup.ksh \
	functional/quota/quota_001_pos.ksh \
	functional/quota/quota_002_pos.ksh \
	functional/quota/quota_003_pos.ksh \
	functional/quota/quota_004_pos.ksh \
	functional/quota/quota_005_pos.ksh \
	functional/quota/quota_006_neg.ksh \
	functional/quota/setup.ksh \
	functional/raidz/cleanup.ksh \
	functional/raidz/raidz_001_neg.ksh \
	functional/raidz/raidz_002_pos.ksh \
	functional/raidz/raidz_003_pos.ksh \
	functional/raidz/raidz_004_pos.ksh \
	functional/raidz/setup.ksh \
	functional/redacted_send/cleanup.ksh \
	functional/redacted_send/redacted_compressed.ksh \
	functional/redacted_send/redacted_contents.ksh \
	functional/redacted_send/redacted_deleted.ksh \
	functional/redacted_send/redacted_disabled_feature.ksh \
	functional/redacted_send/redacted_embedded.ksh \
	functional/redacted_send/redacted_holes.ksh \
	functional/redacted_send/redacted_incrementals.ksh \
	functional/redacted_send/redacted_largeblocks.ksh \
	functional/redacted_send/redacted_many_clones.ksh \
	functional/redacted_send/redacted_mixed_recsize.ksh \
	functional/redacted_send/redacted_mounts.ksh \
	functional/redacted_send/redacted_negative.ksh \
	functional/redacted_send/redacted_origin.ksh \
	functional/redacted_send/redacted_panic.ksh \
	functional/redacted_send/redacted_props.ksh \
	functional/redacted_send/redacted_resume.ksh \
	functional/redacted_send/redacted_size.ksh \
	functional/redacted_send/redacted_volume.ksh \
	functional/redacted_send/setup.ksh \
	functional/redundancy/cleanup.ksh \
	functional/redundancy/redundancy_draid1.ksh \
	functional/redundancy/redundancy_draid2.ksh \
	functional/redundancy/redundancy_draid3.ksh \
	functional/redundancy/redundancy_draid_damaged1.ksh \
	functional/redundancy/redundancy_draid_damaged2.ksh \
	functional/redundancy/redundancy_draid.ksh \
	functional/redundancy/redundancy_draid_spare1.ksh \
	functional/redundancy/redundancy_draid_spare2.ksh \
	functional/redundancy/redundancy_draid_spare3.ksh \
	functional/redundancy/redundancy_mirror.ksh \
	functional/redundancy/redundancy_raidz1.ksh \
	functional/redundancy/redundancy_raidz2.ksh \
	functional/redundancy/redundancy_raidz3.ksh \
	functional/redundancy/redundancy_raidz.ksh \
	functional/redundancy/redundancy_stripe.ksh \
	functional/redundancy/setup.ksh \
	functional/refquota/cleanup.ksh \
	functional/refquota/refquota_001_pos.ksh \
	functional/refquota/refquota_002_pos.ksh \
	functional/refquota/refquota_003_pos.ksh \
	functional/refquota/refquota_004_pos.ksh \
	functional/refquota/refquota_005_pos.ksh \
	functional/refquota/refquota_006_neg.ksh \
	functional/refquota/refquota_007_neg.ksh \
	functional/refquota/refquota_008_neg.ksh \
	functional/refquota/setup.ksh \
	functional/refreserv/cleanup.ksh \
	functional/refreserv/refreserv_001_pos.ksh \
	functional/refreserv/refreserv_002_pos.ksh \
	functional/refreserv/refreserv_003_pos.ksh \
	functional/refreserv/refreserv_004_pos.ksh \
	functional/refreserv/refreserv_005_pos.ksh \
	functional/refreserv/refreserv_multi_raidz.ksh \
	functional/refreserv/refreserv_raidz.ksh \
	functional/refreserv/setup.ksh \
	functional/removal/cleanup.ksh \
	functional/removal/removal_all_vdev.ksh \
	functional/removal/removal_cancel.ksh \
	functional/removal/removal_check_space.ksh \
	functional/removal/removal_condense_export.ksh \
	functional/removal/removal_multiple_indirection.ksh \
	functional/removal/removal_nopwrite.ksh \
	functional/removal/removal_remap_deadlists.ksh \
	functional/removal/removal_reservation.ksh \
	functional/removal/removal_resume_export.ksh \
	functional/removal/removal_sanity.ksh \
	functional/removal/removal_with_add.ksh \
	functional/removal/removal_with_create_fs.ksh \
	functional/removal/removal_with_dedup.ksh \
	functional/removal/removal_with_errors.ksh \
	functional/removal/removal_with_export.ksh \
	functional/removal/removal_with_faulted.ksh \
	functional/removal/removal_with_ganging.ksh \
	functional/removal/removal_with_remove.ksh \
	functional/removal/removal_with_scrub.ksh \
	functional/removal/removal_with_send.ksh \
	functional/removal/removal_with_send_recv.ksh \
	functional/removal/removal_with_snapshot.ksh \
	functional/removal/removal_with_write.ksh \
	functional/removal/removal_with_zdb.ksh \
	functional/removal/remove_attach_mirror.ksh \
	functional/removal/remove_expanded.ksh \
	functional/removal/remove_indirect.ksh \
	functional/removal/remove_mirror.ksh \
	functional/removal/remove_mirror_sanity.ksh \
	functional/removal/remove_raidz.ksh \
	functional/rename_dirs/cleanup.ksh \
	functional/rename_dirs/rename_dirs_001_pos.ksh \
	functional/rename_dirs/setup.ksh \
	functional/replacement/attach_import.ksh \
	functional/replacement/attach_multiple.ksh \
	functional/replacement/attach_rebuild.ksh \
	functional/replacement/attach_resilver.ksh \
	functional/replacement/cleanup.ksh \
	functional/replacement/detach.ksh \
	functional/replacement/rebuild_disabled_feature.ksh \
	functional/replacement/rebuild_multiple.ksh \
	functional/replacement/rebuild_raidz.ksh \
	functional/replacement/replace_import.ksh \
	functional/replacement/replace_rebuild.ksh \
	functional/replacement/replace_resilver.ksh \
	functional/replacement/resilver_restart_001.ksh \
	functional/replacement/resilver_restart_002.ksh \
	functional/replacement/scrub_cancel.ksh \
	functional/replacement/setup.ksh \
	functional/reservation/cleanup.ksh \
	functional/reservation/reservation_001_pos.ksh \
	functional/reservation/reservation_002_pos.ksh \
	functional/reservation/reservation_003_pos.ksh \
	functional/reservation/reservation_004_pos.ksh \
	functional/reservation/reservation_005_pos.ksh \
	functional/reservation/reservation_006_pos.ksh \
	functional/reservation/reservation_007_pos.ksh \
	functional/reservation/reservation_008_pos.ksh \
	functional/reservation/reservation_009_pos.ksh \
	functional/reservation/reservation_010_pos.ksh \
	functional/reservation/reservation_011_pos.ksh \
	functional/reservation/reservation_012_pos.ksh \
	functional/reservation/reservation_013_pos.ksh \
	functional/reservation/reservation_014_pos.ksh \
	functional/reservation/reservation_015_pos.ksh \
	functional/reservation/reservation_016_pos.ksh \
	functional/reservation/reservation_017_pos.ksh \
	functional/reservation/reservation_018_pos.ksh \
	functional/reservation/reservation_019_pos.ksh \
	functional/reservation/reservation_020_pos.ksh \
	functional/reservation/reservation_021_neg.ksh \
	functional/reservation/reservation_022_pos.ksh \
	functional/reservation/setup.ksh \
	functional/rootpool/cleanup.ksh \
	functional/rootpool/rootpool_002_neg.ksh \
	functional/rootpool/rootpool_003_neg.ksh \
	functional/rootpool/rootpool_007_pos.ksh \
	functional/rootpool/setup.ksh \
	functional/rsend/cleanup.ksh \
	functional/rsend/recv_dedup_encrypted_zvol.ksh \
	functional/rsend/recv_dedup.ksh \
	functional/rsend/rsend_001_pos.ksh \
	functional/rsend/rsend_002_pos.ksh \
	functional/rsend/rsend_003_pos.ksh \
	functional/rsend/rsend_004_pos.ksh \
	functional/rsend/rsend_005_pos.ksh \
	functional/rsend/rsend_006_pos.ksh \
	functional/rsend/rsend_007_pos.ksh \
	functional/rsend/rsend_008_pos.ksh \
	functional/rsend/rsend_009_pos.ksh \
	functional/rsend/rsend_010_pos.ksh \
	functional/rsend/rsend_011_pos.ksh \
	functional/rsend/rsend_012_pos.ksh \
	functional/rsend/rsend_013_pos.ksh \
	functional/rsend/rsend_014_pos.ksh \
	functional/rsend/rsend_016_neg.ksh \
	functional/rsend/rsend_019_pos.ksh \
	functional/rsend/rsend_020_pos.ksh \
	functional/rsend/rsend_021_pos.ksh \
	functional/rsend/rsend_022_pos.ksh \
	functional/rsend/rsend_024_pos.ksh \
	functional/rsend/rsend_025_pos.ksh \
	functional/rsend/rsend_026_neg.ksh \
	functional/rsend/rsend_027_pos.ksh \
	functional/rsend/rsend_028_neg.ksh \
	functional/rsend/rsend_029_neg.ksh \
	+ functional/rsend/rsend_030_pos.ksh \
	functional/rsend/send-c_embedded_blocks.ksh \
	functional/rsend/send-c_incremental.ksh \
	functional/rsend/send-c_lz4_disabled.ksh \
	functional/rsend/send-c_mixed_compression.ksh \
	functional/rsend/send-cpL_varied_recsize.ksh \
	functional/rsend/send-c_props.ksh \
	functional/rsend/send-c_recv_dedup.ksh \
	functional/rsend/send-c_recv_lz4_disabled.ksh \
	functional/rsend/send-c_resume.ksh \
	functional/rsend/send-c_stream_size_estimate.ksh \
	functional/rsend/send-c_verify_contents.ksh \
	functional/rsend/send-c_verify_ratio.ksh \
	functional/rsend/send-c_volume.ksh \
	functional/rsend/send-c_zstreamdump.ksh \
	functional/rsend/send_doall.ksh \
	functional/rsend/send_encrypted_files.ksh \
	functional/rsend/send_encrypted_hierarchy.ksh \
	functional/rsend/send_encrypted_props.ksh \
	functional/rsend/send_encrypted_truncated_files.ksh \
	functional/rsend/send_freeobjects.ksh \
	functional/rsend/send_holds.ksh \
	functional/rsend/send_hole_birth.ksh \
	functional/rsend/send_invalid.ksh \
	functional/rsend/send-L_toggle.ksh \
	functional/rsend/send_mixed_raw.ksh \
	functional/rsend/send_partial_dataset.ksh \
	functional/rsend/send_raw_ashift.ksh \
	functional/rsend/send_raw_spill_block.ksh \
	functional/rsend/send_realloc_dnode_size.ksh \
	functional/rsend/send_realloc_encrypted_files.ksh \
	functional/rsend/send_realloc_files.ksh \
	functional/rsend/send_spill_block.ksh \
	functional/rsend/send-wR_encrypted_zvol.ksh \
	functional/rsend/setup.ksh \
	functional/scrub_mirror/cleanup.ksh \
	functional/scrub_mirror/scrub_mirror_001_pos.ksh \
	functional/scrub_mirror/scrub_mirror_002_pos.ksh \
	functional/scrub_mirror/scrub_mirror_003_pos.ksh \
	functional/scrub_mirror/scrub_mirror_004_pos.ksh \
	functional/scrub_mirror/setup.ksh \
	functional/slog/cleanup.ksh \
	functional/slog/setup.ksh \
	functional/slog/slog_001_pos.ksh \
	functional/slog/slog_002_pos.ksh \
	functional/slog/slog_003_pos.ksh \
	functional/slog/slog_004_pos.ksh \
	functional/slog/slog_005_pos.ksh \
	functional/slog/slog_006_pos.ksh \
	functional/slog/slog_007_pos.ksh \
	functional/slog/slog_008_neg.ksh \
	functional/slog/slog_009_neg.ksh \
	functional/slog/slog_010_neg.ksh \
	functional/slog/slog_011_neg.ksh \
	functional/slog/slog_012_neg.ksh \
	functional/slog/slog_013_pos.ksh \
	functional/slog/slog_014_pos.ksh \
	functional/slog/slog_015_neg.ksh \
	functional/slog/slog_016_pos.ksh \
	functional/slog/slog_replay_fs_001.ksh \
	functional/slog/slog_replay_fs_002.ksh \
	functional/slog/slog_replay_volume.ksh \
	functional/snapshot/cleanup.ksh \
	functional/snapshot/clone_001_pos.ksh \
	functional/snapshot/rollback_001_pos.ksh \
	functional/snapshot/rollback_002_pos.ksh \
	functional/snapshot/rollback_003_pos.ksh \
	functional/snapshot/setup.ksh \
	functional/snapshot/snapshot_001_pos.ksh \
	functional/snapshot/snapshot_002_pos.ksh \
	functional/snapshot/snapshot_003_pos.ksh \
	functional/snapshot/snapshot_004_pos.ksh \
	functional/snapshot/snapshot_005_pos.ksh \
	functional/snapshot/snapshot_006_pos.ksh \
	functional/snapshot/snapshot_007_pos.ksh \
	functional/snapshot/snapshot_008_pos.ksh \
	functional/snapshot/snapshot_009_pos.ksh \
	functional/snapshot/snapshot_010_pos.ksh \
	functional/snapshot/snapshot_011_pos.ksh \
	functional/snapshot/snapshot_012_pos.ksh \
	functional/snapshot/snapshot_013_pos.ksh \
	functional/snapshot/snapshot_014_pos.ksh \
	functional/snapshot/snapshot_015_pos.ksh \
	functional/snapshot/snapshot_016_pos.ksh \
	functional/snapshot/snapshot_017_pos.ksh \
	functional/snapshot/snapshot_018_pos.ksh \
	functional/snapused/cleanup.ksh \
	functional/snapused/setup.ksh \
	functional/snapused/snapused_001_pos.ksh \
	functional/snapused/snapused_002_pos.ksh \
	functional/snapused/snapused_003_pos.ksh \
	functional/snapused/snapused_004_pos.ksh \
	functional/snapused/snapused_005_pos.ksh \
	functional/sparse/cleanup.ksh \
	functional/sparse/setup.ksh \
	functional/sparse/sparse_001_pos.ksh \
	functional/stat/cleanup.ksh \
	functional/stat/setup.ksh \
	functional/stat/stat_001_pos.ksh \
	functional/suid/cleanup.ksh \
	functional/suid/setup.ksh \
	functional/suid/suid_write_to_none.ksh \
	functional/suid/suid_write_to_sgid.ksh \
	functional/suid/suid_write_to_suid.ksh \
	functional/suid/suid_write_to_suid_sgid.ksh \
	functional/suid/suid_write_zil_replay.ksh \
	functional/trim/autotrim_config.ksh \
	functional/trim/autotrim_integrity.ksh \
	functional/trim/autotrim_trim_integrity.ksh \
	functional/trim/cleanup.ksh \
	functional/trim/setup.ksh \
	functional/trim/trim_config.ksh \
	functional/trim/trim_integrity.ksh \
	functional/trim/trim_l2arc.ksh \
	functional/truncate/cleanup.ksh \
	functional/truncate/setup.ksh \
	functional/truncate/truncate_001_pos.ksh \
	functional/truncate/truncate_002_pos.ksh \
	functional/truncate/truncate_timestamps.ksh \
	functional/upgrade/cleanup.ksh \
	functional/upgrade/setup.ksh \
	functional/upgrade/upgrade_projectquota_001_pos.ksh \
	functional/upgrade/upgrade_readonly_pool.ksh \
	functional/upgrade/upgrade_userobj_001_pos.ksh \
	functional/user_namespace/cleanup.ksh \
	functional/user_namespace/setup.ksh \
	functional/user_namespace/user_namespace_001.ksh \
	functional/user_namespace/user_namespace_002.ksh \
	functional/user_namespace/user_namespace_003.ksh \
	functional/user_namespace/user_namespace_004.ksh \
	functional/userquota/cleanup.ksh \
	functional/userquota/groupspace_001_pos.ksh \
	functional/userquota/groupspace_002_pos.ksh \
	functional/userquota/groupspace_003_pos.ksh \
	functional/userquota/setup.ksh \
	functional/userquota/userquota_001_pos.ksh \
	functional/userquota/userquota_002_pos.ksh \
	functional/userquota/userquota_003_pos.ksh \
	functional/userquota/userquota_004_pos.ksh \
	functional/userquota/userquota_005_neg.ksh \
	functional/userquota/userquota_006_pos.ksh \
	functional/userquota/userquota_007_pos.ksh \
	functional/userquota/userquota_008_pos.ksh \
	functional/userquota/userquota_009_pos.ksh \
	functional/userquota/userquota_010_pos.ksh \
	functional/userquota/userquota_011_pos.ksh \
	functional/userquota/userquota_012_neg.ksh \
	functional/userquota/userquota_013_pos.ksh \
	functional/userquota/userspace_001_pos.ksh \
	functional/userquota/userspace_002_pos.ksh \
	functional/userquota/userspace_003_pos.ksh \
	functional/userquota/userspace_encrypted.ksh \
	functional/userquota/userspace_send_encrypted.ksh \
	functional/vdev_zaps/cleanup.ksh \
	functional/vdev_zaps/setup.ksh \
	functional/vdev_zaps/vdev_zaps_001_pos.ksh \
	functional/vdev_zaps/vdev_zaps_002_pos.ksh \
	functional/vdev_zaps/vdev_zaps_003_pos.ksh \
	functional/vdev_zaps/vdev_zaps_004_pos.ksh \
	functional/vdev_zaps/vdev_zaps_005_pos.ksh \
	functional/vdev_zaps/vdev_zaps_006_pos.ksh \
	functional/vdev_zaps/vdev_zaps_007_pos.ksh \
	functional/write_dirs/cleanup.ksh \
	functional/write_dirs/setup.ksh \
	functional/write_dirs/write_dirs_001_pos.ksh \
	functional/write_dirs/write_dirs_002_pos.ksh \
	functional/xattr/cleanup.ksh \
	functional/xattr/setup.ksh \
	functional/xattr/xattr_001_pos.ksh \
	functional/xattr/xattr_002_neg.ksh \
	functional/xattr/xattr_003_neg.ksh \
	functional/xattr/xattr_004_pos.ksh \
	functional/xattr/xattr_005_pos.ksh \
	functional/xattr/xattr_006_pos.ksh \
	functional/xattr/xattr_007_neg.ksh \
	functional/xattr/xattr_008_pos.ksh \
	functional/xattr/xattr_009_neg.ksh \
	functional/xattr/xattr_010_neg.ksh \
	functional/xattr/xattr_011_pos.ksh \
	functional/xattr/xattr_012_pos.ksh \
	functional/xattr/xattr_013_pos.ksh \
	functional/xattr/xattr_compat.ksh \
	functional/zpool_influxdb/cleanup.ksh \
	functional/zpool_influxdb/setup.ksh \
	functional/zpool_influxdb/zpool_influxdb.ksh \
	functional/zvol/zvol_cli/cleanup.ksh \
	functional/zvol/zvol_cli/setup.ksh \
	functional/zvol/zvol_cli/zvol_cli_001_pos.ksh \
	functional/zvol/zvol_cli/zvol_cli_002_pos.ksh \
	functional/zvol/zvol_cli/zvol_cli_003_neg.ksh \
	functional/zvol/zvol_ENOSPC/cleanup.ksh \
	functional/zvol/zvol_ENOSPC/setup.ksh \
	functional/zvol/zvol_ENOSPC/zvol_ENOSPC_001_pos.ksh \
	functional/zvol/zvol_misc/cleanup.ksh \
	functional/zvol/zvol_misc/setup.ksh \
	functional/zvol/zvol_misc/zvol_misc_001_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_002_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_003_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_004_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_005_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_006_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_fua.ksh \
	functional/zvol/zvol_misc/zvol_misc_hierarchy.ksh \
	functional/zvol/zvol_misc/zvol_misc_rename_inuse.ksh \
	functional/zvol/zvol_misc/zvol_misc_snapdev.ksh \
	functional/zvol/zvol_misc/zvol_misc_trim.ksh \
	functional/zvol/zvol_misc/zvol_misc_volmode.ksh \
	functional/zvol/zvol_misc/zvol_misc_zil.ksh \
	functional/zvol/zvol_stress/cleanup.ksh \
	functional/zvol/zvol_stress/setup.ksh \
	functional/zvol/zvol_stress/zvol_stress.ksh \
	functional/zvol/zvol_swap/cleanup.ksh \
	functional/zvol/zvol_swap/setup.ksh \
	functional/zvol/zvol_swap/zvol_swap_001_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_002_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_003_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_004_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_005_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_006_pos.ksh
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
	index ca037fdff777..d01017f9e09f 100755
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
	@@ -1,119 +1,127 @@
	#!/bin/ksh -p
	#
	# CDDL HEADER START
	#
	# The contents of this file are subject to the terms of the
	# Common Development and Distribution License (the "License").
	# You may not use this file except in compliance with the License.
	#
	# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	# or https://opensource.org/licenses/CDDL-1.0.
	# See the License for the specific language governing permissions
	# and limitations under the License.
	#
	# When distributing Covered Code, include this CDDL HEADER in each
	# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	# If applicable, add the following below this CDDL HEADER, with the
	# fields enclosed by brackets "[]" replaced with your own identifying
	# information: Portions Copyright [yyyy] [name of copyright owner]
	#
	# CDDL HEADER END
	#

	#
	# Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
	# Copyright (c) 2021 by The FreeBSD Foundation.
	+# Copyright (c) 2022 by Delphix. All rights reserved.
	#

	. $STF_SUITE/include/libtest.shlib

	#
	# DESCRIPTION:
	# Test hole-punching functionality
	#
	# STRATEGY:
	# 1. Create a dense file
	# 2. Punch an assortment of holes in the file and verify the result.
	#
	+# Note: We can't compare exact block numbers as reported by du, because
	+# different backing stores may allocate different numbers of blocks for
	+# the same amount of data.
	+#

	verify_runnable "global"

	#
	# Prior to __FreeBSD_version 1400032 there are no mechanism to punch hole in a
	# file on FreeBSD. truncate -d support is required to call fspacectl(2) on
	# behalf of the script.
	#
	if is_freebsd; then
	if [[ $(uname -K) -lt 1400032 ]]; then
	log_unsupported "Requires fspacectl(2) support on FreeBSD"
	fi
	if truncate -d 2>&1 \| grep "illegal option" > /dev/null; then
	log_unsupported "Requires truncate(1) -d support on FreeBSD"
	fi
	fi

	FILE=$TESTDIR/$TESTFILE0
	BLKSZ=$(get_prop recordsize $TESTPOOL)

	function cleanup
	{
	[[ -e $TESTDIR ]] && log_must rm -f $FILE
	}

	-function check_reported_size
	+function get_reported_size
	{
	- typeset expected_size=$1
	-
	- if ! [ -e "${FILE}" ]; then
	+ if ! [ -e "$FILE" ]; then
	log_fail "$FILE does not exist"
	fi
	-
	- reported_size=$(du "${FILE}" \| awk '{print $1}')
	- if [ "$reported_size" != "$expected_size" ]; then
	- log_fail "Incorrect reported size: $reported_size != $expected_size"
	- fi
	+
	+ sync_pool $TESTPOOL >/dev/null 2>&1
	+ du "$FILE" \| awk '{print $1}'
	}

	function check_apparent_size
	{
	typeset expected_size=$1

	- apparent_size=$(stat_size "${FILE}")
	+ apparent_size=$(stat_size "$FILE")
	if [ "$apparent_size" != "$expected_size" ]; then
	- log_fail "Incorrect apparent size: $apparent_size != $expected_size"
	+ log_fail \
	+ "Incorrect apparent size: $apparent_size != $expected_size"
	fi
	}

	log_assert "Ensure holes can be punched in files making them sparse"

	log_onexit cleanup

	# Create a dense file and check it is the correct size.
	log_must file_write -o create -f $FILE -b $BLKSZ -c 8
	-sync_pool $TESTPOOL
	-log_must check_reported_size 1027
	+full_size=$(get_reported_size)

	-# Punch a hole for the first full block.
	+# Punch a hole for the first full block. The reported size should decrease.
	log_must punch_hole 0 $BLKSZ $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+one_hole=$(get_reported_size)
	+[[ $full_size -gt $one_hole ]] \|\| log_fail \
	+ "One hole failure: $full_size -> $one_hole"

	-# Partially punch a hole in the second block.
	+# Partially punch a hole in the second block. The reported size should
	+# remain constant.
	log_must punch_hole $BLKSZ $((BLKSZ / 2)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+partial_hole=$(get_reported_size)
	+[[ $one_hole -eq $partial_hole ]] \|\| log_fail \
	+ "Partial hole failure: $one_hole -> $partial_hole"

	-# Punch a hole which overlaps the third and fourth block.
	+# Punch a hole which overlaps the third and fourth block. The reported size
	+# should remain constant.
	log_must punch_hole $(((BLKSZ * 2) + (BLKSZ / 2))) $((BLKSZ)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+overlap_hole=$(get_reported_size)
	+[[ $one_hole -eq $overlap_hole ]] \|\| log_fail \
	+ "Overlap hole failure: $one_hole -> $overlap_hole"

	-# Punch a hole from the fifth block past the end of file. The apparent
	-# file size should not change since --keep-size is implied.
	+# Punch a hole from the fifth block past the end of file. The reported size
	+# should decrease, and the apparent file size should not change since
	+# --keep-size is implied.
	apparent_size=$(stat_size $FILE)
	log_must punch_hole $((BLKSZ * 4)) $((BLKSZ * 10)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 387
	+eof_hole=$(get_reported_size)
	+[[ $overlap_hole -gt $eof_hole ]] \|\| log_fail \
	+ "EOF hole failure: $overlap_hole -> $eof_hole"
	log_must check_apparent_size $apparent_size

	log_pass "Ensure holes can be punched in files making them sparse"
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh
	index d8f14273ca1f..47dd06cc0dc3 100755
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh
	@@ -1,119 +1,128 @@
	#!/bin/ksh -p
	#
	# CDDL HEADER START
	#
	# The contents of this file are subject to the terms of the
	# Common Development and Distribution License (the "License").
	# You may not use this file except in compliance with the License.
	#
	# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	# or https://opensource.org/licenses/CDDL-1.0.
	# See the License for the specific language governing permissions
	# and limitations under the License.
	#
	# When distributing Covered Code, include this CDDL HEADER in each
	# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	# If applicable, add the following below this CDDL HEADER, with the
	# fields enclosed by brackets "[]" replaced with your own identifying
	# information: Portions Copyright [yyyy] [name of copyright owner]
	#
	# CDDL HEADER END
	#

	#
	# Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
	# Copyright (c) 2021 by The FreeBSD Foundation.
	+# Copyright (c) 2022 by Delphix. All rights reserved.
	#

	. $STF_SUITE/include/libtest.shlib

	#
	# DESCRIPTION:
	# Test FALLOC_FL_ZERO_RANGE functionality
	#
	# STRATEGY:
	# 1. Create a dense file
	# 2. Zero various ranges in the file and verify the result.
	#
	+# Note: We can't compare exact block numbers as reported by du, because
	+# different backing stores may allocate different numbers of blocks for
	+# the same amount of data.
	+#

	verify_runnable "global"

	if is_freebsd; then
	log_unsupported "FreeBSD does not implement an analogue to ZERO_RANGE."
	fi

	FILE=$TESTDIR/$TESTFILE0
	BLKSZ=$(get_prop recordsize $TESTPOOL)

	function cleanup
	{
	[[ -e $TESTDIR ]] && log_must rm -f $FILE
	}

	-# Helpfully, this function expects kilobytes, and check_apparent_size expects bytes.
	-function check_reported_size
	+function get_reported_size
	{
	- typeset expected_size=$1
	-
	- if ! [ -e "${FILE}" ]; then
	+ if ! [ -e "$FILE" ]; then
	log_fail "$FILE does not exist"
	fi
	-
	- reported_size=$(du "${FILE}" \| awk '{print $1}')
	- if [ "$reported_size" != "$expected_size" ]; then
	- log_fail "Incorrect reported size: $reported_size != $expected_size"
	- fi
	+
	+ sync_pool $TESTPOOL >/dev/null 2>&1
	+ du "$FILE" \| awk '{print $1}'
	}

	function check_apparent_size
	{
	typeset expected_size=$1

	- apparent_size=$(stat_size "${FILE}")
	+ apparent_size=$(stat_size "$FILE")
	if [ "$apparent_size" != "$expected_size" ]; then
	- log_fail "Incorrect apparent size: $apparent_size != $expected_size"
	+ log_fail \
	+ "Incorrect apparent size: $apparent_size != $expected_size"
	fi
	}

	log_assert "Ensure ranges can be zeroed in files"

	log_onexit cleanup

	# Create a dense file and check it is the correct size.
	log_must file_write -o create -f $FILE -b $BLKSZ -c 8
	sync_pool $TESTPOOL
	-log_must check_reported_size 1027
	+full_size=$(get_reported_size)

	-# Zero a range covering the first full block.
	+# Zero a range covering the first full block. The reported size should decrease.
	log_must zero_range 0 $BLKSZ $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+one_range=$(get_reported_size)
	+[[ $full_size -gt $one_range ]] \|\| log_fail \
	+ "One range failure: $full_size -> $one_range"

	-# Partially zero a range in the second block.
	+# Partially zero a range in the second block. The reported size should
	+# remain constant.
	log_must zero_range $BLKSZ $((BLKSZ / 2)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+partial_range=$(get_reported_size)
	+[[ $one_range -eq $partial_range ]] \|\| log_fail \
	+ "Partial range failure: $one_range -> $partial_range"

	-# Zero range which overlaps the third and fourth block.
	+# Zero range which overlaps the third and fourth block. The reported size
	+# should remain constant.
	log_must zero_range $(((BLKSZ * 2) + (BLKSZ / 2))) $((BLKSZ)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 899
	+overlap_range=$(get_reported_size)
	+[[ $one_range -eq $overlap_range ]] \|\| log_fail \
	+ "Overlap range failure: $one_range -> $overlap_range"

	# Zero range from the fifth block past the end of file, with --keep-size.
	-# The apparent file size must not change, since we did specify --keep-size.
	+# The reported size should decrease, and the apparent file size must not
	+# change, since we did specify --keep-size.
	apparent_size=$(stat_size $FILE)
	log_must fallocate --keep-size --zero-range --offset $((BLKSZ * 4)) --length $((BLKSZ * 10)) "$FILE"
	-sync_pool $TESTPOOL
	-log_must check_reported_size 387
	+eof_range=$(get_reported_size)
	+[[ $overlap_range -gt $eof_range ]] \|\| log_fail \
	+ "EOF range failure: $overlap_range -> $eof_range"
	log_must check_apparent_size $apparent_size

	# Zero range from the fifth block past the end of file. The apparent
	# file size should change since --keep-size is not implied, unlike
	-# with PUNCH_HOLE.
	+# with PUNCH_HOLE. The reported size should remain constant.
	apparent_size=$(stat_size $FILE)
	log_must zero_range $((BLKSZ * 4)) $((BLKSZ * 10)) $FILE
	-sync_pool $TESTPOOL
	-log_must check_reported_size 387
	+eof_range2=$(get_reported_size)
	+[[ $eof_range -eq $eof_range2 ]] \|\| log_fail \
	+ "Second EOF range failure: $eof_range -> $eof_range2"
	log_must check_apparent_size $((BLKSZ * 14))

	log_pass "Ensure ranges can be zeroed in files"
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/auto_offline_001_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/auto_offline_001_pos.ksh
	index 17bde9a70636..0ab9317c0a06 100755
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/auto_offline_001_pos.ksh
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/auto_offline_001_pos.ksh
	@@ -1,188 +1,187 @@
	#!/bin/ksh -p
	#
	# CDDL HEADER START
	#
	# This file and its contents are supplied under the terms of the
	# Common Development and Distribution License ("CDDL"), version 1.0.
	# You may only use this file in accordance with the terms of version
	# 1.0 of the CDDL.
	#
	# A full copy of the text of the CDDL should have accompanied this
	# source. A copy of the CDDL is also available via the Internet at
	# http://www.illumos.org/license/CDDL.
	#
	# CDDL HEADER END
	#

	#
	# Copyright 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	#

	. $STF_SUITE/include/libtest.shlib
	. $STF_SUITE/tests/functional/events/events_common.kshlib
	. $STF_SUITE/tests/functional/fault/fault.cfg

	#
	# DESCRIPTION:
	-# Testing Fault Management Agent ZED Logic - Physically removed device is
	-# made unavail and onlined when reattached
	+# Testing Fault Management Agent ZED Logic - Physically detached device is
	+# made removed and onlined when reattached
	#
	# STRATEGY:
	# 1. Create a pool
	# 2. Simulate physical removal of one device
	-# 3. Verify the device is unavailable
	+# 3. Verify the device is removed when detached
	# 4. Reattach the device
	# 5. Verify the device is onlined
	# 6. Repeat the same tests with a spare device:
	# zed will use the spare to handle the removed data device
	# 7. Repeat the same tests again with a faulted spare device:
	-# the removed data device should be unavailable
	+# the removed data device should be removed
	#
	# NOTE: the use of 'block_device_wait' throughout the test helps avoid race
	# conditions caused by mixing creation/removal events from partitioning the
	# disk (zpool create) and events from physically removing it (remove_disk).
	#
	-# NOTE: the test relies on 'zpool sync' to prompt the kmods to transition a
	-# vdev to the unavailable state. The ZED does receive a removal notification
	-# but only relies on it to activate a hot spare. Additional work is planned
	-# to extend an existing ioctl interface to allow the ZED to transition the
	-# vdev in to a removed state.
	+# NOTE: the test relies on ZED to transit state to removed on device removed
	+# event. The ZED does receive a removal notification but only relies on it to
	+# activate a hot spare. Additional work is planned to extend an existing ioctl
	+# interface to allow the ZED to transition the vdev in to a removed state.
	#
	verify_runnable "both"

	if is_linux; then
	# Add one 512b scsi_debug device (4Kn would generate IO errors)
	# NOTE: must be larger than other "file" vdevs and minimum SPA devsize:
	# add 32m of fudge
	load_scsi_debug $(($MINVDEVSIZE/1024/1024+32)) 1 1 1 '512b'
	else
	log_unsupported "scsi debug module unsupported"
	fi

	function cleanup
	{
	destroy_pool $TESTPOOL
	rm -f $filedev1
	rm -f $filedev2
	rm -f $filedev3
	rm -f $sparedev
	unload_scsi_debug
	}

	log_assert "ZED detects physically removed devices"

	log_onexit cleanup

	filedev1="$TEST_BASE_DIR/file-vdev-1"
	filedev2="$TEST_BASE_DIR/file-vdev-2"
	filedev3="$TEST_BASE_DIR/file-vdev-3"
	sparedev="$TEST_BASE_DIR/file-vdev-spare"
	removedev=$(get_debug_device)

	typeset poolconfs=(
	"mirror $filedev1 $removedev"
	"raidz3 $filedev1 $filedev2 $filedev3 $removedev"
	"mirror $filedev1 $filedev2 special mirror $filedev3 $removedev"
	)

	log_must truncate -s $MINVDEVSIZE $filedev1
	log_must truncate -s $MINVDEVSIZE $filedev2
	log_must truncate -s $MINVDEVSIZE $filedev3
	log_must truncate -s $MINVDEVSIZE $sparedev

	for conf in "${poolconfs[@]}"
	do
	# 1. Create a pool
	log_must zpool create -f $TESTPOOL $conf
	block_device_wait ${DEV_DSKDIR}/${removedev}

	mntpnt=$(get_prop mountpoint /$TESTPOOL)

	# 2. Simulate physical removal of one device
	remove_disk $removedev
	log_must mkfile 1m $mntpnt/file
	sync_pool $TESTPOOL

	- # 3. Verify the device is unavailable.
	- log_must wait_vdev_state $TESTPOOL $removedev "UNAVAIL"
	+ # 3. Verify the device is removed.
	+ log_must wait_vdev_state $TESTPOOL $removedev "REMOVED"

	# 4. Reattach the device
	insert_disk $removedev

	# 5. Verify the device is onlined
	log_must wait_vdev_state $TESTPOOL $removedev "ONLINE"

	# cleanup
	destroy_pool $TESTPOOL
	log_must parted "${DEV_DSKDIR}/${removedev}" -s -- mklabel msdos
	block_device_wait ${DEV_DSKDIR}/${removedev}
	done

	# 6. Repeat the same tests with a spare device: zed will use the spare to handle
	# the removed data device
	for conf in "${poolconfs[@]}"
	do
	# 1. Create a pool with a spare
	log_must zpool create -f $TESTPOOL $conf
	block_device_wait ${DEV_DSKDIR}/${removedev}
	log_must zpool add $TESTPOOL spare $sparedev

	mntpnt=$(get_prop mountpoint /$TESTPOOL)

	# 2. Simulate physical removal of one device
	remove_disk $removedev
	log_must mkfile 1m $mntpnt/file
	sync_pool $TESTPOOL

	# 3. Verify the device is handled by the spare.
	log_must wait_hotspare_state $TESTPOOL $sparedev "INUSE"
	- log_must wait_vdev_state $TESTPOOL $removedev "UNAVAIL"
	+ log_must wait_vdev_state $TESTPOOL $removedev "REMOVED"

	# 4. Reattach the device
	insert_disk $removedev

	# 5. Verify the device is onlined
	log_must wait_vdev_state $TESTPOOL $removedev "ONLINE"

	# cleanup
	destroy_pool $TESTPOOL
	log_must parted "${DEV_DSKDIR}/${removedev}" -s -- mklabel msdos
	block_device_wait ${DEV_DSKDIR}/${removedev}
	done

	# 7. Repeat the same tests again with a faulted spare device: zed should offline
	# the removed data device if no spare is available
	for conf in "${poolconfs[@]}"
	do
	# 1. Create a pool with a spare
	log_must zpool create -f $TESTPOOL $conf
	block_device_wait ${DEV_DSKDIR}/${removedev}
	log_must zpool add $TESTPOOL spare $sparedev

	mntpnt=$(get_prop mountpoint /$TESTPOOL)

	# 2. Fault the spare device making it unavailable
	log_must zpool offline -f $TESTPOOL $sparedev
	log_must wait_hotspare_state $TESTPOOL $sparedev "FAULTED"

	# 3. Simulate physical removal of one device
	remove_disk $removedev
	log_must mkfile 1m $mntpnt/file
	sync_pool $TESTPOOL

	- # 4. Verify the device is unavailable
	- log_must wait_vdev_state $TESTPOOL $removedev "UNAVAIL"
	+ # 4. Verify the device is removed
	+ log_must wait_vdev_state $TESTPOOL $removedev "REMOVED"

	# 5. Reattach the device
	insert_disk $removedev

	# 6. Verify the device is onlined
	log_must wait_vdev_state $TESTPOOL $removedev "ONLINE"

	# cleanup
	destroy_pool $TESTPOOL
	log_must parted "${DEV_DSKDIR}/${removedev}" -s -- mklabel msdos
	block_device_wait ${DEV_DSKDIR}/${removedev}
	done

	log_pass "ZED detects physically removed devices"
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/rsend/rsend_030_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/rsend/rsend_030_pos.ksh
	new file mode 100755
	index 000000000000..a683f5befa3d
	--- /dev/null
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/rsend/rsend_030_pos.ksh
	@@ -0,0 +1,66 @@
	+#!/bin/ksh -p
	+
	+#
	+# This file and its contents are supplied under the terms of the
	+# Common Development and Distribution License ("CDDL"), version 1.0.
	+# You may only use this file in accordance with the terms of version
	+# 1.0 of the CDDL.
	+#
	+# A full copy of the text of the CDDL should have accompanied this
	+# source. A copy of the CDDL is also available via the Internet at
	+# http://www.illumos.org/license/CDDL.
	+#
	+
	+#
	+# Copyright (c) 2022 by Nutanix. All rights reserved.
	+#
	+
	+. $STF_SUITE/tests/functional/rsend/rsend.kshlib
	+
	+#
	+# Description:
	+# Verify resumability of full ZFS send/receive on existing dataset
	+#
	+# Strategy:
	+# 1. Start a full ZFS send with redirect output to a file
	+# 2. Mess up the contents of the stream state file on disk
	+# 3. Try ZFS receive, which should fail with a checksum mismatch error
	+# 4. ZFS send to the stream state file again using the receive_resume_token
	+# 5. Verify ZFS receive without "-F" option (force recvflag) fails.
	+# 6. Verify ZFS receive with "-F" option completes successfully.
	+# 7. Repeat steps on an incremental ZFS send. It should complete
	+# successfully without "-F" option.
	+#
	+
	+verify_runnable "both"
	+
	+sendfs=$POOL/sendfs
	+recvfs=$POOL2/recvfs
	+streamfs=$POOL/stream
	+
	+log_assert "Verify resumability of full ZFS send/receive on existing dataset"
	+log_onexit resume_cleanup $sendfs $streamfs
	+
	+test_fs_setup $sendfs $recvfs $streamfs
	+
	+# Full send/recv on existing dataset
	+log_must zfs create -o readonly=on $recvfs
	+log_must eval "zfs send -c -v $sendfs@a >/$streamfs/1"
	+mess_send_file /$streamfs/1
	+log_mustnot eval "zfs recv -suvF $recvfs </$streamfs/1"
	+token=$(get_prop receive_resume_token $recvfs)
	+log_must eval "zfs send -t $token >/$streamfs/2"
	+log_mustnot eval "zfs recv -suv $recvfs </$streamfs/2"
	+log_must eval "zfs recv -suvF $recvfs </$streamfs/2"
	+file_check $sendfs $recvfs
	+
	+# Incremental send/recv
	+log_must eval "zfs send -c -v -i @a $sendfs@b >/$streamfs/3"
	+mess_send_file /$streamfs/3
	+log_mustnot eval "zfs recv -suvF $recvfs </$streamfs/3"
	+token=$(get_prop receive_resume_token $recvfs)
	+log_must eval "zfs send -t $token >/$streamfs/4"
	+log_must eval "zfs recv -suv $recvfs </$streamfs/4"
	+file_check $sendfs $recvfs
	+
	+log_pass "Verify resumability of full ZFS send/receive on existing dataset"
	diff --git a/sys/modules/zfs/zfs_config.h b/sys/modules/zfs/zfs_config.h
	index 9ef55380e57d..d9bdac34eb52 100644
	--- a/sys/modules/zfs/zfs_config.h
	+++ b/sys/modules/zfs/zfs_config.h
	@@ -1,988 +1,994 @@
	/*
	* $FreeBSD$
	*/

	/* zfs_config.h. Generated from zfs_config.h.in by configure. */
	/* zfs_config.h.in. Generated from configure.ac by autoheader. */

	/* Define to 1 if translation of program messages to the user's native
	language is requested. */
	/* #undef ENABLE_NLS */

	/* bio_end_io_t wants 1 arg */
	/* #undef HAVE_1ARG_BIO_END_IO_T */

	/* lookup_bdev() wants 1 arg */
	/* #undef HAVE_1ARG_LOOKUP_BDEV */

	/* submit_bio() wants 1 arg */
	/* #undef HAVE_1ARG_SUBMIT_BIO */

	/* bdi_setup_and_register() wants 2 args */
	/* #undef HAVE_2ARGS_BDI_SETUP_AND_REGISTER */

	/* vfs_getattr wants 2 args */
	/* #undef HAVE_2ARGS_VFS_GETATTR */

	/* zlib_deflate_workspacesize() wants 2 args */
	/* #undef HAVE_2ARGS_ZLIB_DEFLATE_WORKSPACESIZE */

	/* bdi_setup_and_register() wants 3 args */
	/* #undef HAVE_3ARGS_BDI_SETUP_AND_REGISTER */

	/* vfs_getattr wants 3 args */
	/* #undef HAVE_3ARGS_VFS_GETATTR */

	/* vfs_getattr wants 4 args */
	/* #undef HAVE_4ARGS_VFS_GETATTR */

	/* kernel has access_ok with 'type' parameter */
	/* #undef HAVE_ACCESS_OK_TYPE */

	/* posix_acl has refcount_t */
	/* #undef HAVE_ACL_REFCOUNT */

	/* add_disk() returns int */
	/* #undef HAVE_ADD_DISK_RET */

	/* Define if host toolchain supports AES */
	#define HAVE_AES 1

	/* Define if you have [rt] */
	#define HAVE_AIO_H 1

	#ifdef __amd64__
	#ifndef RESCUE
	/* Define if host toolchain supports AVX */
	#define HAVE_AVX 1
	#endif

	/* Define if host toolchain supports AVX2 */
	#define HAVE_AVX2 1

	/* Define if host toolchain supports AVX512BW */
	#define HAVE_AVX512BW 1

	/* Define if host toolchain supports AVX512CD */
	#define HAVE_AVX512CD 1

	/* Define if host toolchain supports AVX512DQ */
	#define HAVE_AVX512DQ 1

	/* Define if host toolchain supports AVX512ER */
	#define HAVE_AVX512ER 1

	/* Define if host toolchain supports AVX512F */
	#define HAVE_AVX512F 1

	/* Define if host toolchain supports AVX512IFMA */
	#define HAVE_AVX512IFMA 1

	/* Define if host toolchain supports AVX512PF */
	#define HAVE_AVX512PF 1

	/* Define if host toolchain supports AVX512VBMI */
	#define HAVE_AVX512VBMI 1

	/* Define if host toolchain supports AVX512VL */
	#define HAVE_AVX512VL 1
	#endif

	/* bdevname() is available */
	/* #undef HAVE_BDEVNAME */

	/* bdev_check_media_change() exists */
	/* #undef HAVE_BDEV_CHECK_MEDIA_CHANGE */

	/* bdev__io_acct() available /
	/* #undef HAVE_BDEV_IO_ACCT */

	+/* bdev_kobj() exists */
	+/* #undef HAVE_BDEV_KOBJ */
	+
	/* bdev_max_discard_sectors() is available */
	/* #undef HAVE_BDEV_MAX_DISCARD_SECTORS */

	/* bdev_max_secure_erase_sectors() is available */
	/* #undef HAVE_BDEV_MAX_SECURE_ERASE_SECTORS */

	/* block_device_operations->submit_bio() returns void */
	/* #undef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID */

	/* bdev_whole() is available */
	/* #undef HAVE_BDEV_WHOLE */

	/* bio_alloc() takes 4 arguments */
	/* #undef HAVE_BIO_ALLOC_4ARG */

	/* bio->bi_bdev->bd_disk exists */
	/* #undef HAVE_BIO_BDEV_DISK */

	/* bio->bi_opf is defined */
	/* #undef HAVE_BIO_BI_OPF */

	/* bio->bi_status exists */
	/* #undef HAVE_BIO_BI_STATUS */

	/* bio has bi_iter */
	/* #undef HAVE_BIO_BVEC_ITER */

	/* bio__io_acct() available /
	/* #undef HAVE_BIO_IO_ACCT */

	/* bio_max_segs() is implemented */
	/* #undef HAVE_BIO_MAX_SEGS */

	/* bio_set_dev() is available */
	/* #undef HAVE_BIO_SET_DEV */

	/* bio_set_dev() GPL-only */
	/* #undef HAVE_BIO_SET_DEV_GPL_ONLY */

	/* bio_set_dev() is a macro */
	/* #undef HAVE_BIO_SET_DEV_MACRO */

	/* bio_set_op_attrs is available */
	/* #undef HAVE_BIO_SET_OP_ATTRS */

	/* blkdev_get_by_path() handles ERESTARTSYS */
	/* #undef HAVE_BLKDEV_GET_ERESTARTSYS */

	/* blkdev_issue_discard() is available */
	/* #undef HAVE_BLKDEV_ISSUE_DISCARD */

	/* blkdev_issue_secure_erase() is available */
	/* #undef HAVE_BLKDEV_ISSUE_SECURE_ERASE */

	/* blkdev_reread_part() exists */
	/* #undef HAVE_BLKDEV_REREAD_PART */

	/* blkg_tryget() is available */
	/* #undef HAVE_BLKG_TRYGET */

	/* blkg_tryget() GPL-only */
	/* #undef HAVE_BLKG_TRYGET_GPL_ONLY */

	/* blk_alloc_disk() exists */
	/* #undef HAVE_BLK_ALLOC_DISK */

	/* blk_alloc_queue() expects request function */
	/* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN */

	/* blk_alloc_queue_rh() expects request function */
	/* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN_RH */

	/* blk_cleanup_disk() exists */
	/* #undef HAVE_BLK_CLEANUP_DISK */

	/* block multiqueue is available */
	/* #undef HAVE_BLK_MQ */

	/* blk queue backing_dev_info is dynamic */
	/* #undef HAVE_BLK_QUEUE_BDI_DYNAMIC */

	/* blk_queue_discard() is available */
	/* #undef HAVE_BLK_QUEUE_DISCARD */

	/* blk_queue_flag_clear() exists */
	/* #undef HAVE_BLK_QUEUE_FLAG_CLEAR */

	/* blk_queue_flag_set() exists */
	/* #undef HAVE_BLK_QUEUE_FLAG_SET */

	/* blk_queue_flush() is available */
	/* #undef HAVE_BLK_QUEUE_FLUSH */

	/* blk_queue_flush() is GPL-only */
	/* #undef HAVE_BLK_QUEUE_FLUSH_GPL_ONLY */

	/* blk_queue_secdiscard() is available */
	/* #undef HAVE_BLK_QUEUE_SECDISCARD */

	/* blk_queue_secure_erase() is available */
	/* #undef HAVE_BLK_QUEUE_SECURE_ERASE */

	/* blk_queue_update_readahead() exists */
	/* #undef HAVE_BLK_QUEUE_UPDATE_READAHEAD */

	/* blk_queue_write_cache() exists */
	/* #undef HAVE_BLK_QUEUE_WRITE_CACHE */

	/* blk_queue_write_cache() is GPL-only */
	/* #undef HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY */

	/* Define if revalidate_disk() in block_device_operations */
	/* #undef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK */

	/* Define to 1 if you have the Mac OS X function CFLocaleCopyCurrent in the
	CoreFoundation framework. */
	/* #undef HAVE_CFLOCALECOPYCURRENT */

	/* Define to 1 if you have the Mac OS X function
	CFLocaleCopyPreferredLanguages in the CoreFoundation framework. */
	/* #undef HAVE_CFLOCALECOPYPREFERREDLANGUAGES */

	/* Define to 1 if you have the Mac OS X function CFPreferencesCopyAppValue in
	the CoreFoundation framework. */
	/* #undef HAVE_CFPREFERENCESCOPYAPPVALUE */

	/* check_disk_change() exists */
	/* #undef HAVE_CHECK_DISK_CHANGE */

	/* clear_inode() is available */
	/* #undef HAVE_CLEAR_INODE */

	/* dentry uses const struct dentry_operations */
	/* #undef HAVE_CONST_DENTRY_OPERATIONS */

	/* copy_from_iter() is available */
	/* #undef HAVE_COPY_FROM_ITER */

	/* copy_to_iter() is available */
	/* #undef HAVE_COPY_TO_ITER */

	/* yes */
	/* #undef HAVE_CPU_HOTPLUG */

	/* current_time() exists */
	/* #undef HAVE_CURRENT_TIME */

	/* Define if the GNU dcgettext() function is already present or preinstalled.
	*/
	/* #undef HAVE_DCGETTEXT */

	/* DECLARE_EVENT_CLASS() is available */
	/* #undef HAVE_DECLARE_EVENT_CLASS */

	/* dequeue_signal() takes 4 arguments */
	/* #undef HAVE_DEQUEUE_SIGNAL_4ARG */

	/* lookup_bdev() wants dev_t arg */
	/* #undef HAVE_DEVT_LOOKUP_BDEV */

	/* sops->dirty_inode() wants flags */
	/* #undef HAVE_DIRTY_INODE_WITH_FLAGS */

	/* disk__io_acct() available /
	/* #undef HAVE_DISK_IO_ACCT */

	/* disk_update_readahead() exists */
	/* #undef HAVE_DISK_UPDATE_READAHEAD */

	/* Define to 1 if you have the <dlfcn.h> header file. */
	#define HAVE_DLFCN_H 1

	/* d_make_root() is available */
	/* #undef HAVE_D_MAKE_ROOT */

	/* d_prune_aliases() is available */
	/* #undef HAVE_D_PRUNE_ALIASES */

	/* dops->d_revalidate() operation takes nameidata */
	/* #undef HAVE_D_REVALIDATE_NAMEIDATA */

	/* eops->encode_fh() wants child and parent inodes */
	/* #undef HAVE_ENCODE_FH_WITH_INODE */

	/* sops->evict_inode() exists */
	/* #undef HAVE_EVICT_INODE */

	/* FALLOC_FL_ZERO_RANGE is defined */
	/* #undef HAVE_FALLOC_FL_ZERO_RANGE */

	/* fault_in_iov_iter_readable() is available */
	/* #undef HAVE_FAULT_IN_IOV_ITER_READABLE */

	/* fops->aio_fsync() exists */
	/* #undef HAVE_FILE_AIO_FSYNC */

	/* file_dentry() is available */
	/* #undef HAVE_FILE_DENTRY */

	/* fops->fadvise() exists */
	/* #undef HAVE_FILE_FADVISE */

	/* file_inode() is available */
	/* #undef HAVE_FILE_INODE */

	/* iops->follow_link() cookie */
	/* #undef HAVE_FOLLOW_LINK_COOKIE */

	/* iops->follow_link() nameidata */
	/* #undef HAVE_FOLLOW_LINK_NAMEIDATA */

	/* fops->fsync() with range */
	/* #undef HAVE_FSYNC_RANGE */

	/* fops->fsync() without dentry */
	/* #undef HAVE_FSYNC_WITHOUT_DENTRY */

	/* yes */
	/* #undef HAVE_GENERIC_FADVISE */

	/* generic_fillattr requires struct user_namespace* */
	/* #undef HAVE_GENERIC_FILLATTR_USERNS */

	/* generic__io_acct() 3 arg available /
	/* #undef HAVE_GENERIC_IO_ACCT_3ARG */

	/* generic__io_acct() 4 arg available /
	/* #undef HAVE_GENERIC_IO_ACCT_4ARG */

	/* generic_readlink is global */
	/* #undef HAVE_GENERIC_READLINK */

	/* generic_setxattr() exists */
	/* #undef HAVE_GENERIC_SETXATTR */

	/* generic_write_checks() takes kiocb */
	/* #undef HAVE_GENERIC_WRITE_CHECKS_KIOCB */

	/* Define if the GNU gettext() function is already present or preinstalled. */
	/* #undef HAVE_GETTEXT */

	/* iops->get_acl() exists */
	/* #undef HAVE_GET_ACL */

	/* iops->get_acl() takes rcu */
	/* #undef HAVE_GET_ACL_RCU */

	/* iops->get_link() cookie */
	/* #undef HAVE_GET_LINK_COOKIE */

	/* iops->get_link() delayed */
	/* #undef HAVE_GET_LINK_DELAYED */

	/* group_info->gid exists */
	/* #undef HAVE_GROUP_INFO_GID */

	/* has_capability() is available */
	/* #undef HAVE_HAS_CAPABILITY */

	/* Define if you have the iconv() function and it works. */
	#define HAVE_ICONV 1

	/* Define if compiler supports -Winfinite-recursion */
	/* #undef HAVE_INFINITE_RECURSION */

	/* yes */
	/* #undef HAVE_INODE_LOCK_SHARED */

	/* inode_owner_or_capable() exists */
	/* #undef HAVE_INODE_OWNER_OR_CAPABLE */

	/* inode_owner_or_capable() takes user_ns */
	/* #undef HAVE_INODE_OWNER_OR_CAPABLE_IDMAPPED */

	/* inode_set_flags() exists */
	/* #undef HAVE_INODE_SET_FLAGS */

	/* inode_set_iversion() exists */
	/* #undef HAVE_INODE_SET_IVERSION */

	/* inode->i_time's are timespec64 /
	/* #undef HAVE_INODE_TIMESPEC64_TIMES */

	/* timestamp_truncate() exists */
	/* #undef HAVE_INODE_TIMESTAMP_TRUNCATE */

	/* Define to 1 if you have the <inttypes.h> header file. */
	#define HAVE_INTTYPES_H 1

	/* in_compat_syscall() is available */
	/* #undef HAVE_IN_COMPAT_SYSCALL */

	/* iops->create() takes struct user_namespace* */
	/* #undef HAVE_IOPS_CREATE_USERNS */

	/* iops->mkdir() takes struct user_namespace* */
	/* #undef HAVE_IOPS_MKDIR_USERNS */

	/* iops->mknod() takes struct user_namespace* */
	/* #undef HAVE_IOPS_MKNOD_USERNS */

	/* iops->permission() takes struct user_namespace* */
	/* #undef HAVE_IOPS_PERMISSION_USERNS */

	/* iops->rename() takes struct user_namespace* */
	/* #undef HAVE_IOPS_RENAME_USERNS */

	/* iops->symlink() takes struct user_namespace* */
	/* #undef HAVE_IOPS_SYMLINK_USERNS */

	/* iov_iter_advance() is available */
	/* #undef HAVE_IOV_ITER_ADVANCE */

	/* iov_iter_count() is available */
	/* #undef HAVE_IOV_ITER_COUNT */

	/* iov_iter_fault_in_readable() is available */
	/* #undef HAVE_IOV_ITER_FAULT_IN_READABLE */

	/* iov_iter_revert() is available */
	/* #undef HAVE_IOV_ITER_REVERT */

	/* iov_iter_type() is available */
	/* #undef HAVE_IOV_ITER_TYPE */

	/* iov_iter types are available */
	/* #undef HAVE_IOV_ITER_TYPES */

	/* yes */
	/* #undef HAVE_IO_SCHEDULE_TIMEOUT */

	/* Define to 1 if you have the `issetugid' function. */
	#define HAVE_ISSETUGID 1

	/* kernel has kernel_fpu_* functions */
	/* #undef HAVE_KERNEL_FPU */

	/* kernel has asm/fpu/api.h */
	/* #undef HAVE_KERNEL_FPU_API_HEADER */

	/* kernel fpu internal */
	/* #undef HAVE_KERNEL_FPU_INTERNAL */

	/* kernel has asm/fpu/internal.h */
	/* #undef HAVE_KERNEL_FPU_INTERNAL_HEADER */

	/* uncached_acl_sentinel() exists */
	/* #undef HAVE_KERNEL_GET_ACL_HANDLE_CACHE */

	/* kernel does stack verification */
	/* #undef HAVE_KERNEL_OBJTOOL */

	/* kernel has linux/objtool.h */
	/* #undef HAVE_KERNEL_OBJTOOL_HEADER */

	/* kernel_read() take loff_t pointer */
	/* #undef HAVE_KERNEL_READ_PPOS */

	/* timer_list.function gets a timer_list */
	/* #undef HAVE_KERNEL_TIMER_FUNCTION_TIMER_LIST */

	/* struct timer_list has a flags member */
	/* #undef HAVE_KERNEL_TIMER_LIST_FLAGS */

	/* timer_setup() is available */
	/* #undef HAVE_KERNEL_TIMER_SETUP */

	/* kernel_write() take loff_t pointer */
	/* #undef HAVE_KERNEL_WRITE_PPOS */

	/* kmem_cache_create_usercopy() exists */
	/* #undef HAVE_KMEM_CACHE_CREATE_USERCOPY */

	/* kstrtoul() exists */
	/* #undef HAVE_KSTRTOUL */

	/* ktime_get_coarse_real_ts64() exists */
	/* #undef HAVE_KTIME_GET_COARSE_REAL_TS64 */

	/* ktime_get_raw_ts64() exists */
	/* #undef HAVE_KTIME_GET_RAW_TS64 */

	/* kvmalloc exists */
	/* #undef HAVE_KVMALLOC */

	/* Define if you have [aio] */
	/* #undef HAVE_LIBAIO */

	/* Define if you have [blkid] */
	/* #undef HAVE_LIBBLKID */

	/* Define if you have [crypto] */
	#define HAVE_LIBCRYPTO 1

	/* Define if you have [tirpc] */
	/* #undef HAVE_LIBTIRPC */

	/* Define if you have [udev] */
	/* #undef HAVE_LIBUDEV */

	/* Define if you have [uuid] */
	/* #undef HAVE_LIBUUID */

	/* linux/blk-cgroup.h exists */
	/* #undef HAVE_LINUX_BLK_CGROUP_HEADER */

	/* lseek_execute() is available */
	/* #undef HAVE_LSEEK_EXECUTE */

	/* makedev() is declared in sys/mkdev.h */
	/* #undef HAVE_MAKEDEV_IN_MKDEV */

	/* makedev() is declared in sys/sysmacros.h */
	/* #undef HAVE_MAKEDEV_IN_SYSMACROS */

	/* Noting that make_request_fn() returns blk_qc_t */
	/* #undef HAVE_MAKE_REQUEST_FN_RET_QC */

	/* Noting that make_request_fn() returns void */
	/* #undef HAVE_MAKE_REQUEST_FN_RET_VOID */

	/* iops->mkdir() takes umode_t */
	/* #undef HAVE_MKDIR_UMODE_T */

	/* Define to 1 if you have the `mlockall' function. */
	#define HAVE_MLOCKALL 1

	/* lookup_bdev() wants mode arg */
	/* #undef HAVE_MODE_LOOKUP_BDEV */

	/* Define if host toolchain supports MOVBE */
	#define HAVE_MOVBE 1

	/* new_sync_read()/new_sync_write() are available */
	/* #undef HAVE_NEW_SYNC_READ */

	/* folio_wait_bit() exists */
	/* #undef HAVE_PAGEMAP_FOLIO_WAIT_BIT */

	+/* part_to_dev() exists */
	+/* #undef HAVE_PART_TO_DEV */
	+
	/* iops->getattr() takes a path */
	/* #undef HAVE_PATH_IOPS_GETATTR */

	/* Define if host toolchain supports PCLMULQDQ */
	#define HAVE_PCLMULQDQ 1

	/* percpu_counter_add_batch() is defined */
	/* #undef HAVE_PERCPU_COUNTER_ADD_BATCH */

	/* percpu_counter_init() wants gfp_t */
	/* #undef HAVE_PERCPU_COUNTER_INIT_WITH_GFP */

	/* posix_acl_chmod() exists */
	/* #undef HAVE_POSIX_ACL_CHMOD */

	/* posix_acl_from_xattr() needs user_ns */
	/* #undef HAVE_POSIX_ACL_FROM_XATTR_USERNS */

	/* posix_acl_release() is available */
	/* #undef HAVE_POSIX_ACL_RELEASE */

	/* posix_acl_release() is GPL-only */
	/* #undef HAVE_POSIX_ACL_RELEASE_GPL_ONLY */

	/* posix_acl_valid() wants user namespace */
	/* #undef HAVE_POSIX_ACL_VALID_WITH_NS */

	/* proc_ops structure exists */
	/* #undef HAVE_PROC_OPS_STRUCT */

	/* iops->put_link() cookie */
	/* #undef HAVE_PUT_LINK_COOKIE */

	/* iops->put_link() delayed */
	/* #undef HAVE_PUT_LINK_DELAYED */

	/* iops->put_link() nameidata */
	/* #undef HAVE_PUT_LINK_NAMEIDATA */

	/* If available, contains the Python version number currently in use. */
	#define HAVE_PYTHON "3.7"

	/* qat is enabled and existed */
	/* #undef HAVE_QAT */

	/* register_shrinker is vararg */
	/* #undef HAVE_REGISTER_SHRINKER_VARARG */

	/* iops->rename() wants flags */
	/* #undef HAVE_RENAME_WANTS_FLAGS */

	/* REQ_DISCARD is defined */
	/* #undef HAVE_REQ_DISCARD */

	/* REQ_FLUSH is defined */
	/* #undef HAVE_REQ_FLUSH */

	/* REQ_OP_DISCARD is defined */
	/* #undef HAVE_REQ_OP_DISCARD */

	/* REQ_OP_FLUSH is defined */
	/* #undef HAVE_REQ_OP_FLUSH */

	/* REQ_OP_SECURE_ERASE is defined */
	/* #undef HAVE_REQ_OP_SECURE_ERASE */

	/* REQ_PREFLUSH is defined */
	/* #undef HAVE_REQ_PREFLUSH */

	/* revalidate_disk() is available */
	/* #undef HAVE_REVALIDATE_DISK */

	/* revalidate_disk_size() is available */
	/* #undef HAVE_REVALIDATE_DISK_SIZE */

	/* struct rw_semaphore has member activity */
	/* #undef HAVE_RWSEM_ACTIVITY */

	/* struct rw_semaphore has atomic_long_t member count */
	/* #undef HAVE_RWSEM_ATOMIC_LONG_COUNT */

	/* linux/sched/signal.h exists */
	/* #undef HAVE_SCHED_SIGNAL_HEADER */

	/* Define to 1 if you have the <security/pam_modules.h> header file. */
	#define HAVE_SECURITY_PAM_MODULES_H 1

	/* setattr_prepare() is available, doesn't accept user_namespace */
	/* #undef HAVE_SETATTR_PREPARE_NO_USERNS */

	/* setattr_prepare() accepts user_namespace */
	/* #undef HAVE_SETATTR_PREPARE_USERNS */

	/* iops->set_acl() exists, takes 3 args */
	/* #undef HAVE_SET_ACL */

	/* iops->set_acl() takes 4 args */
	/* #undef HAVE_SET_ACL_USERNS */

	/* set_cached_acl() is usable */
	/* #undef HAVE_SET_CACHED_ACL_USABLE */

	/* set_special_state() exists */
	/* #undef HAVE_SET_SPECIAL_STATE */

	/* struct shrink_control exists */
	/* #undef HAVE_SHRINK_CONTROL_STRUCT */

	/* kernel_siginfo_t exists */
	/* #undef HAVE_SIGINFO */

	/* signal_stop() exists */
	/* #undef HAVE_SIGNAL_STOP */

	/* new shrinker callback wants 2 args */
	/* #undef HAVE_SINGLE_SHRINKER_CALLBACK */

	/* cs->count_objects exists */
	/* #undef HAVE_SPLIT_SHRINKER_CALLBACK */

	#if defined(__amd64__) \|\| defined(__i386__)
	/* Define if host toolchain supports SSE */
	#define HAVE_SSE 1

	/* Define if host toolchain supports SSE2 */
	#define HAVE_SSE2 1

	/* Define if host toolchain supports SSE3 */
	#define HAVE_SSE3 1

	/* Define if host toolchain supports SSE4.1 */
	#define HAVE_SSE4_1 1

	/* Define if host toolchain supports SSE4.2 */
	#define HAVE_SSE4_2 1

	/* Define if host toolchain supports SSSE3 */
	#define HAVE_SSSE3 1
	#endif

	/* STACK_FRAME_NON_STANDARD is defined */
	/* #undef HAVE_STACK_FRAME_NON_STANDARD */

	/* standalone <linux/stdarg.h> exists */
	/* #undef HAVE_STANDALONE_LINUX_STDARG */

	/* Define to 1 if you have the <stdint.h> header file. */
	#define HAVE_STDINT_H 1

	/* Define to 1 if you have the <stdio.h> header file. */
	#define HAVE_STDIO_H 1

	/* Define to 1 if you have the <stdlib.h> header file. */
	#define HAVE_STDLIB_H 1

	/* Define to 1 if you have the <strings.h> header file. */
	#define HAVE_STRINGS_H 1

	/* Define to 1 if you have the <string.h> header file. */
	#define HAVE_STRING_H 1

	/* Define to 1 if you have the `strlcat' function. */
	#define HAVE_STRLCAT 1

	/* Define to 1 if you have the `strlcpy' function. */
	#define HAVE_STRLCPY 1

	/* submit_bio is member of struct block_device_operations */
	/* #undef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */

	/* super_setup_bdi_name() exits */
	/* #undef HAVE_SUPER_SETUP_BDI_NAME */

	/* super_block->s_user_ns exists */
	/* #undef HAVE_SUPER_USER_NS */

	/* struct kobj_type has default_groups */
	/* #undef HAVE_SYSFS_DEFAULT_GROUPS */

	/* Define to 1 if you have the <sys/stat.h> header file. */
	#define HAVE_SYS_STAT_H 1

	/* Define to 1 if you have the <sys/types.h> header file. */
	#define HAVE_SYS_TYPES_H 1

	/* i_op->tmpfile() exists */
	/* #undef HAVE_TMPFILE */

	/* i_op->tmpfile() has userns */
	/* #undef HAVE_TMPFILE_USERNS */

	/* totalhigh_pages() exists */
	/* #undef HAVE_TOTALHIGH_PAGES */

	/* kernel has totalram_pages() */
	/* #undef HAVE_TOTALRAM_PAGES_FUNC */

	/* Define to 1 if you have the `udev_device_get_is_initialized' function. */
	/* #undef HAVE_UDEV_DEVICE_GET_IS_INITIALIZED */

	/* kernel has __kernel_fpu_* functions */
	/* #undef HAVE_UNDERSCORE_KERNEL_FPU */

	/* Define to 1 if you have the <unistd.h> header file. */
	#define HAVE_UNISTD_H 1

	/* iops->getattr() takes struct user_namespace* */
	/* #undef HAVE_USERNS_IOPS_GETATTR */

	/* user_namespace->ns.inum exists */
	/* #undef HAVE_USER_NS_COMMON_INUM */

	/* iops->getattr() takes a vfsmount */
	/* #undef HAVE_VFSMOUNT_IOPS_GETATTR */

	/* aops->direct_IO() uses iovec */
	/* #undef HAVE_VFS_DIRECT_IO_IOVEC */

	/* aops->direct_IO() uses iov_iter without rw */
	/* #undef HAVE_VFS_DIRECT_IO_ITER */

	/* aops->direct_IO() uses iov_iter with offset */
	/* #undef HAVE_VFS_DIRECT_IO_ITER_OFFSET */

	/* aops->direct_IO() uses iov_iter with rw and offset */
	/* #undef HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET */

	/* filemap_dirty_folio exists */
	/* #undef HAVE_VFS_FILEMAP_DIRTY_FOLIO */

	/* All required iov_iter interfaces are available */
	/* #undef HAVE_VFS_IOV_ITER */

	/* fops->iterate() is available */
	/* #undef HAVE_VFS_ITERATE */

	/* fops->iterate_shared() is available */
	/* #undef HAVE_VFS_ITERATE_SHARED */

	/* fops->readdir() is available */
	/* #undef HAVE_VFS_READDIR */

	/* address_space_operations->readpages exists */
	/* #undef HAVE_VFS_READPAGES */

	/* read_folio exists */
	/* #undef HAVE_VFS_READ_FOLIO */

	/* fops->read/write_iter() are available */
	/* #undef HAVE_VFS_RW_ITERATE */

	/* __set_page_dirty_nobuffers exists */
	/* #undef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS */

	/* __vmalloc page flags exists */
	/* #undef HAVE_VMALLOC_PAGE_KERNEL */

	/* yes */
	/* #undef HAVE_WAIT_ON_BIT_ACTION */

	/* wait_queue_entry_t exists */
	/* #undef HAVE_WAIT_QUEUE_ENTRY_T */

	/* wq_head->head and wq_entry->entry exist */
	/* #undef HAVE_WAIT_QUEUE_HEAD_ENTRY */

	/* xattr_handler->get() wants dentry */
	/* #undef HAVE_XATTR_GET_DENTRY */

	/* xattr_handler->get() wants both dentry and inode */
	/* #undef HAVE_XATTR_GET_DENTRY_INODE */

	/* xattr_handler->get() wants dentry and inode and flags */
	/* #undef HAVE_XATTR_GET_DENTRY_INODE_FLAGS */

	/* xattr_handler->get() wants xattr_handler */
	/* #undef HAVE_XATTR_GET_HANDLER */

	/* xattr_handler has name */
	/* #undef HAVE_XATTR_HANDLER_NAME */

	/* xattr_handler->list() wants dentry */
	/* #undef HAVE_XATTR_LIST_DENTRY */

	/* xattr_handler->list() wants xattr_handler */
	/* #undef HAVE_XATTR_LIST_HANDLER */

	/* xattr_handler->list() wants simple */
	/* #undef HAVE_XATTR_LIST_SIMPLE */

	/* xattr_handler->set() wants dentry */
	/* #undef HAVE_XATTR_SET_DENTRY */

	/* xattr_handler->set() wants both dentry and inode */
	/* #undef HAVE_XATTR_SET_DENTRY_INODE */

	/* xattr_handler->set() wants xattr_handler */
	/* #undef HAVE_XATTR_SET_HANDLER */

	/* xattr_handler->set() takes user_namespace */
	/* #undef HAVE_XATTR_SET_USERNS */

	/* Define if host toolchain supports XSAVE */
	#define HAVE_XSAVE 1

	/* Define if host toolchain supports XSAVEOPT */
	#define HAVE_XSAVEOPT 1

	/* Define if host toolchain supports XSAVES */
	#define HAVE_XSAVES 1

	/* ZERO_PAGE() is GPL-only */
	/* #undef HAVE_ZERO_PAGE_GPL_ONLY */

	/* Define if you have [z] */
	#define HAVE_ZLIB 1

	/* __posix_acl_chmod() exists */
	/* #undef HAVE___POSIX_ACL_CHMOD */

	/* kernel exports FPU functions */
	/* #undef KERNEL_EXPORTS_X86_FPU */

	/* TBD: fetch(3) support */
	#if 0
	/* whether the chosen libfetch is to be loaded at run-time */
	#define LIBFETCH_DYNAMIC 1

	/* libfetch is fetch(3) */
	#define LIBFETCH_IS_FETCH 1

	/* libfetch is libcurl */
	#define LIBFETCH_IS_LIBCURL 0

	/* soname of chosen libfetch */
	#define LIBFETCH_SONAME "libfetch.so.6"
	#endif

	/* Define to the sub-directory where libtool stores uninstalled libraries. */
	#define LT_OBJDIR ".libs/"

	/* make_request_fn() return type */
	/* #undef MAKE_REQUEST_FN_RET */

	/* hardened module_param_call */
	/* #undef MODULE_PARAM_CALL_CONST */

	/* struct shrink_control has nid */
	/* #undef SHRINK_CONTROL_HAS_NID */

	/* using complete_and_exit() instead */
	/* #undef SPL_KTHREAD_COMPLETE_AND_EXIT */

	/* Defined for legacy compatibility. */
	#define SPL_META_ALIAS ZFS_META_ALIAS

	/* Defined for legacy compatibility. */
	#define SPL_META_RELEASE ZFS_META_RELEASE

	/* Defined for legacy compatibility. */
	#define SPL_META_VERSION ZFS_META_VERSION

	/* pde_data() is PDE_DATA() */
	/* #undef SPL_PDE_DATA */

	/* Define to 1 if all of the C90 standard headers exist (not just the ones
	required in a freestanding environment). This macro is provided for
	backward compatibility; new code need not use it. */
	#define SYSTEM_FREEBSD 1

	/* True if ZFS is to be compiled for a Linux system */
	/* #undef SYSTEM_LINUX */

	/* Version number of package */
	/* #undef ZFS_DEBUG */

	/* /dev/zfs minor */
	/* #undef ZFS_DEVICE_MINOR */

	/* enum node_stat_item contains NR_FILE_PAGES */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_FILE_PAGES */

	/* enum node_stat_item contains NR_INACTIVE_ANON */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_ANON */

	/* enum node_stat_item contains NR_INACTIVE_FILE */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_FILE */

	/* enum zone_stat_item contains NR_FILE_PAGES */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_FILE_PAGES */

	/* enum zone_stat_item contains NR_INACTIVE_ANON */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_ANON */

	/* enum zone_stat_item contains NR_INACTIVE_FILE */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_FILE */

	/* GENHD_FL_EXT_DEVT flag is not available */
	/* #undef ZFS_GENHD_FL_EXT_DEVT */

	/* GENHD_FL_NO_PART_SCAN flag is available */
	/* #undef ZFS_GENHD_FL_NO_PART */

	/* global_node_page_state() exists */
	/* #undef ZFS_GLOBAL_NODE_PAGE_STATE */

	/* global_zone_page_state() exists */
	/* #undef ZFS_GLOBAL_ZONE_PAGE_STATE */

	/* Define to 1 if GPL-only symbols can be used */
	/* #undef ZFS_IS_GPL_COMPATIBLE */

	/* Define the project alias string. */
	-#define ZFS_META_ALIAS "zfs-2.1.99-FreeBSD_gc629f0bf6"
	+#define ZFS_META_ALIAS "zfs-2.1.99-FreeBSD_gd62bafee9"

	/* Define the project author. */
	#define ZFS_META_AUTHOR "OpenZFS"

	/* Define the project release date. */
	/* #undef ZFS_META_DATA */

	/* Define the maximum compatible kernel version. */
	#define ZFS_META_KVER_MAX "5.19"

	/* Define the minimum compatible kernel version. */
	#define ZFS_META_KVER_MIN "3.10"

	/* Define the project license. */
	#define ZFS_META_LICENSE "CDDL"

	/* Define the libtool library 'age' version information. */
	/* #undef ZFS_META_LT_AGE */

	/* Define the libtool library 'current' version information. */
	/* #undef ZFS_META_LT_CURRENT */

	/* Define the libtool library 'revision' version information. */
	/* #undef ZFS_META_LT_REVISION */

	/* Define the project name. */
	#define ZFS_META_NAME "zfs"

	/* Define the project release. */
	-#define ZFS_META_RELEASE "FreeBSD_gc629f0bf6"
	+#define ZFS_META_RELEASE "FreeBSD_gd62bafee9"

	/* Define the project version. */
	#define ZFS_META_VERSION "2.1.99"

	/* count is located in percpu_ref.data */
	/* #undef ZFS_PERCPU_REF_COUNT_IN_DATA */
	diff --git a/sys/modules/zfs/zfs_gitrev.h b/sys/modules/zfs/zfs_gitrev.h
	index e933621159d5..e0c6ae76e8cc 100644
	--- a/sys/modules/zfs/zfs_gitrev.h
	+++ b/sys/modules/zfs/zfs_gitrev.h
	@@ -1 +1 @@
	-#define ZFS_META_GITREV "zfs-2.1.99-1404-gc629f0bf6"
	+#define ZFS_META_GITREV "zfs-2.1.99-1439-gd62bafee9"

File Metadata

Mime Type: application/octet-stream
Expires: Mon, Apr 29, 10:26 PM (2 d)
Storage Engine: chunks
Storage Format: Chunks
Storage Handle: .xHBjyoM1iR1
Default Alt Text: (6 MB)

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions